IPG Photonics – charts up to Q1 2015

Disclosure – I’m long IPG Photonics Corporation (IPGP).

There are a few remarks about particular issues, but I mostly let the charts speak for themselves. My sources are the SEC filings and the press release “IPG Photonics Reports First Quarter 2015 Revenue Growth of 17%” on IPG’s site.

Charts of quarterly data

In the first chart, for Q2 2015 “Net sales”, I used the figures for “revenue” in the guidance ($215 million to $225 million). “Revenue” is likely to mean the same as “net sales” because the earnings release gives similar figures for Q1 2014 “Revenue” and “Net sales” ($170.6 million and $170,575 thousand).

IPG results to Q1 2015 bars and line

In the short period charted, I’d say Q2 and Q3 look strong, but the 10-K for 2014 has –

    “Historically, our net sales have been higher in the second half of the year than in the first half of the year.”

IPG results to Q1 2015 stacked

IPG EPS to Q1 2015

IPG EPS growth to Q1 2015

I have not charted quarterly Other Comprehensive Income, but annual charts further down show a big OCI loss for 2014 due to “translation adjustments”, i.e. the effect of changing exchange rates on the U.S. dollar value of assets and liabilities. For Q1 2015, the OCI loss was $38,276 thousand, leaving comprehensive income of only $19,070 thousand.

Cash flow charts

In the cash flow charts below, cash from operations includes variation in current assets and liabilities, particularly inventories, accounts receivable and accounts payable. If I’d also shown cash from operations without those movements (still with depreciation, stock-based compensation, etc. added back to net income), the resulting chart might look smoother, but it’s a lot of work.

For some companies, it would be essential to use capex rather than cash invested, because cash put into short term investments is more like parking cash for a while than an investment in the company’s operations. Some long term investments can also have little to do with investing in operations.

In IPG’s case, there were “Purchases of short-term investments” of $25,451 thousand in 2011, and exactly the same amount was “Proceeds from short-term investments” in 2012, due to the investments maturing. (Financial activity in 2012 included a stock offering, a special dividend, and buying out a minority interest.) As a result of the short-term investment, the investment flow I’ve used overstates the cash invested in IPG’s operations in 2011, and understates by the same amount in 2012. I only use the charts to get a general impression and I’m not concerned by shifting a portion of old investment by a year.

The short-term investment accounted for 32.2% of the investment flow in 2011, and 46.1% in 2012. There were no purchases or proceeds from short-term investments in other years from 2009 to 2014. In 2013 there were “Proceeds from sale of investment”, but only worth $495 thousand.

For comparison, Net cash used in investing activities rose from $10,639 thousand in 2009 to $90,080 thousand in 2014.

The cash invested is dominated by investment in the company’s operations. “Purchases of property, plant and equipment” accounts for most of the investment flow ($88,601 thousand, or 98.4% in 2014). There’s also “Acquisition of businesses, net of cash acquired”, which hit $11.6 million in 2012, the highest amount in the past six years, and the “Purchase of intangible assets” of $2 million in 2014 also counts as an investment in the company’s operations.

IPG cash from ops against cash into investment

IPG cash from ops against cash invested per share

A more conventional chart of cash flow per share –

IPG cash from ops and cash invested per share

This chart is logarithmic, so the slope of a line corresponds to the compound growth rate –

IPG logarithmic cash flow - to 2014

The charts for growth in OCF between any two years are probably understood best if I show the spreadsheet first. Although there’s no “2014”, the table is up to date, for example the bottom row is for 2013, the only entry is “49.6%” in the column for 1 year ahead, and the increase from 2013 to 2014 was 49.6%. The figure in the top right shows growth of 30.2% (CAGR) between 2004 and 2014.

IPG CAGR CFO per share - spread

IPG CAGR CFO per share by year

Sometimes growth over five years is used as a metric. In the chart below, find the columns grouped above “5”, and you can see that growth measured over the previous five years has been over 20% for each of the past five years, i.e. for 2004 to 2009, 2005 to 2010 … up to 2009 to 2014.

IPG CAGR CFO per share by number of years

Sales, costs and profit charts

In some cases I’ve used a stacked and a non-stacked chart of the same data. If any subjective bias results from the way a chart represents data, a different type of chart might not produce the same bias.

In order to get the stacked charts, I needed to change the sign of expenses, which are negative in the 10-Ks (as indicated by brackets). For some years in some stacked charts there will be minor distortion due to the inclusion of a negative quantity, for example “Interest expense (income)” has mostly negative figures, but the absolute amounts have declined from about 5% of net sales in the early years, to 0.01% in 2014.

I changed some item-names, e.g. “OTHER INCOME (EXPENSE), Net:” in the 2014 10-K is “Other expense (income), net” in my charts, to make the name consistent with the sign change which was needed to get a stacked chart.

Interest expense is the only expense I found to be capitalized as investment, instead of charging it as an expense on the income statement. From the 10-K –

    “Expenditures for maintenance and repairs are charged to operations. Interest expense associated with significant capital projects is capitalized as a cost of the project. The Company capitalized $383 , $524 and $142 of interest expense in 2014 , 2013 and 2012 , respectively.” (thousands)

That compares to an expense of $77 and $1 for 2014 and 2013 on the income statements (i.e. not capitalized), and interest income of $319 in 2012 (all in thousands). The capitalization is small and probably conservative. The capitalization of expenses is too complicated a subject to explain in detail here, I’ll just mention that a few companies have taken it to fraudulent levels.

The items reported changed between 2006 and 2014, which presented problems. A few expenses were only significant or relevant in 2005 or in early years, for example the Series B warrants have not been relevant since 2007. 2006 is the earliest 10-K I used.

IPG sales breakdown by cost and income

IPG sales breakdown by cost and income as percent

In the next chart I’ve zoomed in and the detail is clearer, but some percentages for 2002 and 2003 are off the scale. The percentages are represented in the chart above, and they’re in the spreadsheet titled “Data as percentage of sales” near the end.

IPG income and cost as percent of sales

IPG sales costs and income

IPG operating expenses

IPG sales and profit

The following chart is unorthodox. It’s often supposed that a drop in margins warns of problems. When a company has pricing power, falling margins can warn of the erosion of pricing power through increased competition. Analysts are aware of other causes, such as variation in product mix, but it still occasionally happens that an officer at a company with high margins will remind analysts of just how high the margins are, when the analysts are probing a drop of 1% or 2%. The chart is intended to let you judge if margins predict results, as in sales, operating income, net income or cash from operations. By showing the quantities as a percentage of 2014 net sales, I got everything to fit the percentage scale on the left.

IPG margins and main results

I do not see evidence that a fall in margins predicts disappointing results, but this is a subjective exercise and you may interpret the chart differently. 2013 was a relatively poor year for margins, cash from operations, net income and operating income, but 2014 was a good year all round.

IPG earnings per share

IPG annual EPS growth

IPG Logarithmic chart of sales and profit

Comprehensive income charts

Other comprehensive income (OCI) in 2014 consisted mostly of an unusually large loss on Translation adjustments, of -$110,734 (thousand). The loss on translation adjustments was more than twice as big as the rise in net income in 2014, with a sharp fall in comprehensive income as a result.

“Translation” refers to the currency exchange rates at which assets valued in foreign currency are converted to U.S. dollar values. For the Russian operations, a low Ruble keeps costs low, and is good for net income, but a fall in the Ruble’s value against the dollar reduces the dollar value of Russian assets, which affects OCI. If the Ruble stays at a constant low level, there would be no further effect on asset values and OCI, while costs in Russia could continue to stay low, depending on inflation. I have not seen the impact of the Ruble broken out.

The impact of a low Yuan would be different. China is not a production center and does not have large long-lived assets (find “long-lived assets” below). It follows that local costs are likely to be small relative to sales in China, so there would be little benefit to costs from a lower Yuan, while the exchange rate would be bad for sales, although the size of the hit would depend on various factors including the price elasticity of demand and competitors’ costs. I don’t mean to imply that the Yuan is still falling – the Yuan’s big fall ended in January 2014.

Recently, the high U.S. dollar and low Ruble have dominated the currency effects, although the Ruble has partly recovered this year and the U.S. dollar index (DXY) has weakened since March.

25.45% of the $110,734 (thousand) translation loss is explained by the currency hit on Cash and Cash Equivalents. From the 10-K for 2014, in thousands –


The 10-Q for Q1 2015 has a “Total other comprehensive loss” of $38,276, due to Translation adjustments of -$38,319.

This is from the 10-K, about valuing foreign assets, and where they put the cash –

    “Foreign Currency — The financial information for entities outside the United States is measured using local currencies as the functional currency. Assets and liabilities are translated into U.S. dollars at the exchange rate in effect on the respective balance sheet dates. Income and expenses are translated into U.S. dollars based on the average rate of exchange for the corresponding period. Exchange rate differences resulting from translation adjustments are accounted for directly as a component of accumulated other comprehensive loss.

    Cash and Cash Equivalents — Cash and cash equivalents consist primarily of highly liquid investments, such as bank deposits, marketable securities with original maturities of three months or less with insignificant interest rate risk and marketable securities with remaining maturities of three months or less at the date of acquisition.”

Don’t worry about bars you can’t see in the next chart. Charting an insignificant quantity is my way of demonstrating the insignificance visually.

IPG comprehensive income

IPG comprehensive income per share

Geography and sales breakdown charts

Russia is included in CIS, which is in IPG’s region “Other including Eastern Europe/CIS”. CIS is the Commonwealth of Independent States, which includes Azerbaijan, Armenia, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Russia, Tajikistan, Turkmenistan, Uzbekistan and Ukraine.

IPG geography stacked

IPG geography stacked percent

IPG physical assets by location

    “Our major manufacturing facilities are located in the United States, Germany and Russia.” (10-K for 2014)

The three countries account for nearly all the long-lived assets. China accounted for 32% of net sales in 2014, but only 2.30% of the long-lived assets. (IPG have two application centers in China, in Shenzhen and Beijing.)

The evidence suggests that long-lived assets are dominated by manufacturing, and the relative size of long-lived assets by physical location is likely to be a reasonable proxy for relative production, although it is affected by foreign exchange rates and changes in them. There could also be differences in asset turnover, for example diode production is centered in the U.S., and the turnover is not necessarily the same as for other production.

IPG physical assets by location percent

IPG’s materials processing sales have grown, but other applications have not. Three small categories were consolidated into “Other applications” in 2012.

IPG materials processing and other

IPG materials processing and other - percent

Days Sales Outstanding and days of inventory charts

I fitted trend lines to DSO and days of inventory in the next chart. The DSO trend line formula shows that every year on average, DSO increases by 0.3771 of a day. However, the R-squared of 0.0707 shows the line only explains 7.07% of the variation. Taking the square root gives R, the correlation coefficient, of about 0.2659 which is low, with a 45.78% chance that there was no underlying linear trend, with random variation probably the most likely alternative. It’s conventional to reject a hypothesis with 5% or more chance of being false, as here.

The trend line formula for days of inventory shows that every year on average, days of inventory falls by 2.8046 days. The R-squared of 0.5212 shows the line explains 52.12% of the variation. Taking the square root gives R, the correlation coefficient, of 0.7219418, with only a 1.84% chance that there was no underlying linear trend. I’m wary of projecting the trend, for example a step change between the first five years and the last five years would also fit the data.

If you want to check the figures, there’s a calculator on The Chinese University of Hong Kong‘s site. Navigate – Stats toolbox Home / Correlation / Stat Sig of r. For the days of inventory, enter 10 for the sample size, and 0.7219418 for R. The results include “p=0.0184”, which translates to a 1.84% chance of the hypothesis being false.

IPG DSO and days inventory

Analysts and markets generally don’t like rising DSO or rising days of inventory. Some less sophisticated investors might even look at accounts receivable and inventory without adjusting for sales. There is some statistical evidence for the predictive power of the metrics, for example in Beneish’s work on his M-Score for earnings manipulation. If rises in DSO or days of inventory are a strong indication of trouble for IPG, the effect ought to be discernible in the next chart. I suggest looking at the chart and seeing if DSO or days of inventory predict anything for sales, income or cash from operations.

IPG DSO days inventory - sales income and CFO

So far as I can tell, rises in DSO or days of inventory have not foreshadowed problems, and neither have above average DSO or days of inventory. The only serious blip was in 2009, when IPG were affected by the recession which followed the financial crisis. I would not blame the results on the lengthening of IPG’s cash conversion cycle, especially as cash from operations was hardly affected. I would also not try to predict the next recession by looking for a rise in IPG’s DSO or days of inventory. The chart shows that the metrics fell along with sales and income in 2009.

Analysts will continue to probe any increase in DSO or days of inventory, some authors will take a gloomy view of any such rises, and so might the markets, but so long as the metrics don’t rise by or to unprecedented levels, I don’t see evidence that they predict trouble for IPG. I have not completed a full investigation, and in particular I have not looked at quarterly figures.

IPG hold high levels of inventory, keeping stocks of parts that can be assembled quickly when orders are received. This helps sales by giving customers low lead times, but the low lead times reduce the visibility of future sales. Given the high net cash and the good free cash flow, IPG might as well operate with high levels of working capital so long as it benefits sales. Growth adds to the need for inventory, and occasional preparation for higher growth and the launch of new models adds more. From the 10-Q for Q1 2015 –

    “Given our vertical integration, rigorous and time-consuming testing procedures for both internally manufactured and externally purchased components and the lead time required to manufacture components used in our finished products, the rate at which we turn inventory has historically been comparatively low when compared to our cost of sales. Also, our historic growth rates required investment in inventories to support future sales and enable us to quote short delivery times to our customers, providing what we believe is a competitive advantage. Furthermore, if there was a disruption to the manufacturing capacity of any of our key technologies, our inventories of components should enable us to continue to build finished products for a reasonable period of time. We believe that we will continue to maintain a relatively high level of inventory compared to our cost of sales. As a result, we expect to have a significant amount of working capital invested in inventory. A reduction in our level of net sales or the rate of growth of our net sales from their current levels would mean that the rate at which we are able to convert our inventory into cash would decrease.”

Balance sheet charts

The value of overseas assets and liabilities is affected by exchange rate movements, and for 2014 both will be lower than otherwise due to the higher dollar. The lower Ruble will have had an effect on asset values, particularly on Property, plant, and equipment, net. From an accounting perspective, the translation adjustments which dominated the negative Other Comprehensive Income in 2014 would have the effect of pushing asset values down (but that was more than offset by income pulling the value up). The translation adjustments imply that overseas assets are bigger than overseas liabilities. That’s to be expected as the overall assets are bigger than liabilities, and it’s common for an international company with head office costs in the U.S. to have its liabilities concentrated there.

The unrealized gain on derivatives in 2014 was only $172 thousand, and I found no indication of a realized gain on currency hedges in 2014. IPG report Cash Flow Hedges but they were for fixing interest rates, not hedging currencies. From the 10-K for 2014 –

    “We have no foreign currency derivative instrument hedges as of December 31, 2014 . We will continue to analyze our exposure to currency exchange rate fluctuations and may engage in financial hedging techniques in the future to attempt to minimize the effect of these potential fluctuations.”

The 10-Q for Q1 2015 has a similar statement with “We have no foreign currency derivative instrument hedges as of March 31, 2015 .”

IPG assets

IPG assets per share

IPG assets per cent of total

IPG liabilities and equity

Regarding the note in the graphic above, in the balance sheet spreadsheet at the end, the equity is called “Total IPG Photonics Corporation stockholders’ equity”. The name is used in 10-Ks for 2009 to 2014. My calculation agrees with Total liabilities for 2012 to 2014, but is greater than stated total liabilities for 2011 as my calculation includes minority interests. In 2010 the term “Total liabilities” was used but not on the balance sheet, and in 2006 the term was only used in connection with a covenant ratio.

There were many claims on assets in 2005, including series A, B and D, warrants, Notes receivable from stockholders, and Deferred compensation. Most of those items were gone by 2006, but I still preferred to use my simplification so I could take the charts back to 2005. The term “non-stockholder claims on assets” would have been more accurate, as my “calculated liabilities” includes liabilities, minority interests, and the variety of claims on the 2005 balance sheet.

IPG liabilities and equity per share

IPG liabilities cash and current assets

IPG liabilities as percent of cash


I don’t repeat the spreadsheet for cash flow growth which I’ve shown above. The only omission I know about is for the “Logarithmic chart, base 2, of sales and profit” (before the title “Comprehensive income charts”, above), where I don’t show the results or formulas of the logarithmic functions. The logarithmic formulas all have this format, with only the cell address varying –


IPG results to Q1 2015 spread

IPG cash from ops and into investment - spread2

IPG Photonics sales costs and profit - spread

IPG Photonics sales costs and profit - sign adj - spread

IPG Photonics sales costs and profit - sign adj - formulas

IPG Photonics sales costs and profit sign adj percent - spread

IPG Photonics sales costs and profit sign adj percent - formulas

IPG comprehensive income - spread

IPG comprehensive income - formulas

IPG sales breakdowns spread

IPG DSO days inventory - sales income and CFO - spread

IPG balance sheet spread

That’s all.

DISCLAIMER: Your investment is your responsibility. It is your responsibility to check all material facts before making an investment decision. All investments involve different degrees of risk. You should be aware of your risk tolerance level and financial situations at all times. Furthermore, you should read all transaction confirmations, monthly, and year-end statements. Read any and all prospectuses carefully before making any investment decisions. You are free at all times to accept or reject all investment recommendations made by the author of this blog. All Advice on this blog is subject to market risk and may result in the entire loss of the reader’s investment. Please understand that any losses are attributed to market forces beyond the control or prediction of the author. As you know, a recommendation, which you are free to accept or reject, is not a guarantee for the successful performance of an investment.


Gilead Sciences – a dose of information

Disclosure: I’m long Gilead Sciences, Inc. (GILD).

Share price as at April 1, 2015: $97.72.

Contentious profile

Gilead say they are a “… research-based biopharmaceutical company that discovers, develops and commercializes innovative medicines in areas of unmet medical need.” (from the 10-K). Pure “bio” companies can make things like stem cells or complex proteins that are hard for anyone else to make (see Biosimilar on Wikipedia). Producers in India, China and Bangladesh are capable of making Gilead’s big sellers or the intermediates they can be made from. Gilead have the same reliance on patents as any other pharma innovator, and similar risk, except for the unusual attack from AbbVie’s controversial combination-finding model.

About this piece

Investors who are new to Gilead might like to read from “Litigation regarding Sofosbuvir” (below) first, to see if they are comfortable with the risk and with their ability to assess it.

My simple explanations may bring more pain than joy to well informed readers, who might prefer “Gilead Sciences – the charts” (wordpress.com), where I’ve cut nearly all the text.

I cover many aspects of the business, with a lot about hepatitis C but not much about HIV.

I start with a table and chart for customer concentration, then break down revenue by geography, type and product, which leads into product descriptions, the pipeline, and more.

Customer concentration

Gilead customers over 10pc rev - spread

Gilead customers over 10pc rev

Revenue by geography

Gilead revenue by geography - spread

Gilead revenue by geography

Gilead revenue by geography percent

Revenue by type

This is overwhelmingly product sales. The other category is Royalty, contract and other revenues.

Gilead rev by type - spread

Gilead rev by type

Revenue by product

Gilead product breakdown spread

(The spreadsheet formulas are shown here.)

Gilead product pie detail

Gilead product pie type

Harvoni was only on sale for a short time in 2014, with approval from the FDA in October 2014 and from the European Commission in November 2014. Quoting from “Gilead Sciences’ (GILD) CEO John Martin on Q4 2014 Results – Earnings Call Transcript” Feb. 3, 2015 (www.seekingalpha.com) –

    “prescription data at year-end indicate that for each Sovaldi patient, three patients started therapy with Harvoni.”

It’s hard to estimate 2015 sales of Harvoni by extrapolating from the short period of sales in 2014, partly because –

    “Although after the launch of Harvoni we saw payor restrictions increasingly in place across all patient types and especially for those with lower fibrosis scores.” (the Seeking Alpha transcript, linked to above)

and there has been news in 2015 which affects price and volume. Payor restrictions have been eased, but in return for lower pricing. If you want to estimate 2015 sales of Harvoni, you may be better off looking at the weekly prescriptions and multiplying by the average price, if you can estimate it, but the prescription data I found is U.S. only and might not catch every prescription.

In the two pie charts that follow, I set Harvoni sales equal to Sovaldi sales in 2014, to give a view of the product concentration which does not grossly under-represent Harvoni. The simple kludge is not intended as a projection, or as an adjustment to 2014 results. Some Harvoni sales will replace Sovaldi sales, and the charts take no account of that.

Gilead product pie detail extra Harvoni

Gilead product pie type extra Harvoni

Gilead’s products

I’ve quoted selectively from a list on page 6 of the 10-K PDF. Lots of crucial info is omitted to keep the length down. I put in the percent of Total product sales I calculated, when the sales of the product were disclosed.


• 4.89% ~ Stribild is an oral formulation dosed once a day for the treatment of HIV-1 infection in treatment-naive adults. Stribild is our third complete single tablet regimen for the treatment of HIV and is a fixed-dose combination of our antiretroviral medications, Vitekta, Tybost, Viread® and Emtriva® (emtricitabine).
• 5.02% ~ Complera/Eviplera is an oral formulation dosed once a day for the treatment of HIV-1 infection in adults.
• 14.18% ~ Atripla is an oral formulation dosed once a day for the treatment of HIV infection in adults. Atripla is our first single tablet regimen for HIV intended as a stand-alone therapy or in combination with other antiretrovirals.
• 13.65% ~ Truvada® (emtricitabine and tenofovir disoproxil fumarate) is an oral formulation dosed once a day as part of combination therapy to treat HIV infection in adults.
• 4.32% ~ Viread is an oral formulation of a nucleotide analog reverse transcriptase inhibitor, dosed once a day as part of combination therapy to treat HIV infection in patients two years of age and older.
• Emtriva is an oral formulation of a nucleoside analog reverse transcriptase inhibitor, dosed once a day as part of combination therapy to treat HIV infection in adults.
• Tybost is a pharmacokinetic enhancer dosed once a day that boosts blood levels of certain HIV medicines.
• Vitekta is an oral formulation of an integrase inhibitor, dosed once a day as part of combination therapy to treat HIV infection in adults without known mutations associated with resistance to elvitegravir, the active ingredient of Vitekta.

Liver Diseases

I’ll start with some info about the seven major genotypes of HCV viruses, after describing “genotype”.

Every person has a gene for eye color. There’s an allele for brown eyes, an allele for green eyes, and an allele for blue eyes. Another example of an allele is the allele for lactose intolerance. A person’s genotype is their complete set of alleles. (I’m ignoring details like the existence of dominant and recessive alleles.)

Although virus reproduction is very different, they still have genes, alleles, and genotypes. Genotypes are grouped into classes, and HCV is currently divided into seven major genotypes, numbered 1 to 7 (according to Wikipedia, but most sources I’ve seen only mention 1 to 6). Genotype 4 is broken down into 4a, 4b etc. up to 4j, while genotype 5 has no sub-types (I’m not sure that’s up to date).

Approximate relative prevalence of genotypes in the U.S. –

Genotype 1 ~ 70%
Genotype 2 ~ 20%
Other genotypes ~ Wikipedia claims about 1% for each other major genotype, presumably leaving about 5% for all the minor genotypes.

Genotype 1 is also the most common genotype in South America and Europe.

Genotype 1 is subdivided into genotype 1a and genotype 1b, while genotype 2 is subdivided into genotypes 2a, 2b, 2c and 2d. I assume that when it’s stated that a product can be used for treating genotype 1, that includes the sub-variants. The 10-K for 2014 has “genotype 1” several times in connection with HCV, and no instance of “genotype 1a” or “genotype 1b”.

Sources: “Hepatitis C” (wikipedia.org), Genotype (wikipedia.org), “Hepatitis C” (hepatitiscentral.com – I’m not sure who they are), What are Genotypes? – Definition, Examples & Quiz (study.com).

For more simplified science, find “Sofosbuvir-aka-Sovaldi bio-chemistry”, below. I also write about “Compensated and decompensated cirrhosis of the liver”, below.

• 8.69% ~ Harvoni is an oral formulation of the NS5A inhibitor with a nucleotide analog polymerase inhibitor dosed once a day for the treatment of HCV genotype 1 infection in adults. … Harvoni is also indicated for certain patients … those with HCV/HIV-1 co-infection.
• 42.2% ~ Sovaldi is an oral formulation of a nucleotide analog polymerase inhibitor dosed once a day for the treatment of HCV as a component of a combination antiviral treatment regimen. … Sovaldi’s efficacy has been established in patients with HCV genotypes 1, 2, 3 or 4 infection (in United States and Europe) and genotypes 5 and 6 infection (in Europe), including … those with HCV/HIV-1 co-infection.
• (4.32%, listed under HIV) Viread is an oral formulation of a nucleotide analog reverse transcriptase inhibitor, dosed once a day for the treatment of chronic HBV in adults with compensated and decompensated liver disease. … Viread is also approved for the treatment of HIV infection.
• Hepsera® (adefovir dipivoxil) is an oral formulation of a nucleotide analog polymerase inhibitor, dosed once a day to treat chronic HBV in patients 12 years of age and older.


• Zydelig is a first-in-class PI3K delta inhibitor, in combination with rituximab, for the treatment of certain blood cancers.


• 2.43% ~ Letairis (ambrisentan) is an oral formulation of an endothelin receptor antagonist (ERA) indicated for the treatment of pulmonary arterial hypertension (PAH) (World Health Organization (WHO) Group 1) in patients with WHO Class II or III symptoms to improve exercise capacity and delay clinical worsening.
• 2.08% ~ Ranexa® (ranolazine) is an extended-release tablet for the treatment of chronic angina.
• Lexiscan®/Rapiscan® (regadenoson) injection is indicated for use as a pharmacologic stress agent in radionuclide myocardial perfusion imaging (MPI), a test that detects and characterizes coronary artery disease, in patients unable to undergo adequate exercise stress.


• Cayston® (aztreonam for inhalation solution) is an inhaled antibiotic for the treatment of respiratory systems in cystic fibrosis (CF) patients seven years of age and older with Pseudomonas aeruginosa (P. aeruginosa).
• Tamiflu® (oseltamivir phosphate) is an oral antiviral available in capsule form for the treatment and prevention of influenza A and B. Tamiflu is approved for the treatment of influenza in children and adults in more than 60 countries, including the United States, Japan and the European Union.


• 1.59% ~ AmBisome® (amphotericin B liposome for injection) is a proprietary liposomal formulation of amphotericin B, an antifungal agent to treat serious invasive fungal infections caused by various fungal species in adults.
• Macugen® (pegaptanib sodium injection) is an intravitreal injection of an anti-angiogenic oligonucleotide for the treatment of neovascular agerelated macular degeneration.


Gilead pipeline

The data is from counting blue bars on Gilead’s pipeline page. I’m not sure what a blue bar means, e.g. a blue bar completely across Phase 2 suggests the phase is complete, but all the blue bars go all the way across. I’ve emailed Investor Relations and not seen a reply. (I failed to draw a definite conclusion by looking into GS-4774, which has blue bars up to Phase 2, and is the subject of “Safety and Efficacy of GS-4774 in Combination With Tenofovir Disoproxil Fumarate (TDF) for the Treatment of Subjects With Chronic Hepatitis B and Who Are Currently Not on Treatment” (clinicaltrials.gov).)

There’s a statement at the top of the ‘pipeline’ page –

    “Safety and efficacy of the following compounds have not been established.”

In some cases, Gilead won’t get 100% of any profit, for example –

    “Complera/Eviplera is a fixed-dose combination of our antiretroviral medications, Viread and Emtriva, and Janssen’s non-nucleoside reverse transcriptase inhibitor, Edurant (rilpivirine).”

Some of the revenue from that would go to Janssen Pharmaceuticals (and when other companies use a Gilead product in a combination, it’s good news, rather than competition).

The pipeline page omits anything pre-clinical.

On Gilead’s site, the Earnings page has the PDF “Fourth Quarter 2014 Earnings slides” which includes diagrams for the pipeline which I prefer to the diagrams on the pipeline page, because it’s clear which phase a candidate is in. The pipeline page should be more up to date, though. The Earnings page also has the Q4 results and guidance for 2015, and audio for the earnings call.

Approvals timeline

I’ve reduced Gilead’s page showing year, picture and name, to number of approvals in this chart –

Gilead approvals numbers

The numbers add up to 22. The trendline I added should not be taken too seriously, but the four approvals in 2014 are probably above any reasonable trend, and a fall looks fairly likely. The R-squared of 0.416 means the trend line only explains 41.6% of the variation. The number of approvals in a year says nothing about the commercial potential of each of them.

Recent Phase 3 results for TAF (for HIV)

See “Gilead Announces Phase 3 Results for Investigational Once-Daily Single Tablet HIV Regimen Containing Tenofovir Alafenamide (TAF)” (investors.gilead.com).

The TAF regimen being tested was found to be “Non-Inferior” for treating HIV, with less side-effect damage to the liver and bones. The reason why TAF is needed is that HIV patients are living longer than previously, increasing the need for treatments which cause less damage. The trial subjects were adults with HIV-1 infection who had not previously received treatment.

Patent expiration timeline

Shareholders will know that Gilead have a good patent expiration timeline, with the big sellers Sovaldi and Harvoni not due to expire until 2029 and 2030. The expiry of Atripla and Truvada in 2021 is not so far off but is not imminent either. There’s a full list on page 17 of the 10-K for 2014 PDF. My table below shows the 2014 sales. It misses out the products in the two ‘other’ categories (in the product sales breakdown), but they only accounted for 1.04% of product sales. I compiled the table by combining the disclosures for sales per product, with the patent expiry table. I’ve shown a simplified (or kludged) version, because it only gives the date of U.S. expiry, along with global sales.

    “For our product that are single tablet regimens (e.g., Truvada, Atripla, Complera/Eviplera and Stribild), the estimated patent expiration dates provided correspond to the latest expiring compound patent for one of the active ingredients in the single tablet regimen.” (10-K for 2014)

The Harvoni combination gets it’s expiry dates based on the latest expiries of its two components. The combination only has patent applications and pending patents (there’s more on that later).

Gilead US patent expiry 2015 spread 3

The next chart tries to show the benefit of the U.S. expiry years, compared to the assumption of an equal amount of U.S. sales expiring every year (expiring with the patents), from 2015 to 2030.

The charts are only meant to give a very rough view, because various assumptions and fudges were needed due to the lack of detailed data. I assumed 2014 sales would repeat until patent expiry, followed by zero sales. I split the ‘rest of the world’ between U.S. and E.U. I did not adjust for Harvoni’s launch date, and because of that the timeline should be better than in the chart.

Gilead US patent expiry 2015

What counts in the chart is that the brown line is above the straight blue line, which means that annual sales are greater than they would be if the effect of expiry was spread evenly throughout the period.

The benefit is not so big for the E.U., but E.U. product sales were less than a third of the U.S. sales in 2014.

Gilead EU patent expiry 2015

One thing that seemed certain enough was the strength of the IP, because an expiration date tends to imply that a patent protects a product until expiry.

Litigation regarding Sofosbuvir

While the risk from the patent disputes may be hard for some people to judge (including myself), it should not be forgotten that nearly half of Gilead’s revenue in 2014 was not from hepatitis C (the disputed area), Gilead have a strong pipeline, and the valuation is relatively low.

Details of Gilead’s litigation start on page 18 of the 10-K PDF.

Sometimes, the U.S. Patent and Trademark Office (USPTO) notices that the claims of a pending patent clash or “interfere” with a patent that’s already been granted. They told Gilead about interference from pending patents for a class of compounds which includes metabolites of sofosbuvir, the drug in Gilead’s Sovaldi and one of two ingredients in Harvoni. (Metabolites are the result of the body’s metabolism turning one compound into other compounds. Technically, sofosbuvir (Wikipedia) is a “protide prodrug”, which is metabolized to an active antiviral.) The pending patents for the class of compounds belong to Idenix Pharmaceuticals, Inc.

In cases of interference, it was necessary to establish who was first to invent (under the law at that time). In March 2013, the USPTO Patent Trial and Appeal Board decided in favor of Gilead. The reason was that Idenix had not given instructions on how to make the disputed compounds. After that, the USPTO board decided Pharmasset (which Gilead acquired) were the first to invent sofosbuvir, because Idenix couldn’t prove they thought of it first, or identified the structure of the compounds or any use for them, or tested them, in the relevant time period. Idenix has appealed to the U.S. District Court for the District of Delaware. That’s only the “First Idenix Interference”. In the “Second Idenix Interference”, Gilead have won the first phase.

(The law has changed, as most of the Leahy-Smith America Invents Act (Wikipedia) became effective on September 16, 2012 and March 16, 2013. America’s principle of “first to invent” was out of step with the rest of the world, which operates on “first to file”. It’s easy to see which filing date was earlier, but the content still needs to be examined, for example to see if it described how to make a compound. The law does not affect previous patents, and it’s the pre-Leahy-Smith law that applies to Gilead’s patent disputes over sofosbuvir.)

The dispute has spread to other Idenix patents, such as ‘191 and ‘572, although patent numbers could be different because they are filed overseas. The dispute has gone on a world tour, with Idenix losing in Norway and the U.K., and withdrawing in China. A trial in Canada started in January 2015. A decision from a court in Düsseldorf, Germany, was expected in mid-March 2015. A trial will start in Sydney, Australia, in September 2015.

From the 10-K,

    “Idenix was acquired by Merck in August 2014. While the acquisition does not change our view of the lack of merit in the claims made by Idenix, Merck has greater resources than Idenix and may therefore choose to fund the litigation at higher levels than Idenix.”

Litigation with Merck

From the 10-K for 2014 –

    “In August 2013, Merck contacted us requesting that we pay royalties on the sales of sofosbuvir and obtain a license to U.S. Patent Nos. 7,105,499 and 8,481,712, which it co-owns with Isis Pharmaceuticals, Inc. We believe that Merck’s patents are invalid and are not infringed by” etc.

For more about Merck, find “Merck’s HCV pipeline” and the link “Merck’s hep C drug no longer a ‘breakthrough’”, below.

See also “Gilead Sciences: Clear The Dockets And Settle With Merck Already” by Small Pharma Analyst, Oct. 21, 2014 (seekingalpha.com). There are some good points in the comments about why Merck’s case is probably weak. It’s still true that judges don’t always decide correctly or as expected. Chemist357 pointed out that Merck’s provisional patents expired worthless. For a provisional patent to stake an invention date, the patent application had to be filed within 12 to 13 months.

Litigation with AbbVie, Inc.

USPTO (the U.S. Patent and Trademark Office) is likely to refuse a patent for a combination of two drugs which attack different vulnerabilities of the same pathogen, on the grounds that it’s an obvious thing to try. It’s not too surprising that Gilead has no patent for the combination of sofosbuvir and ledipasvir in Harvoni. More surprising is the granting of five patents for the combined use of the same two drugs to Abbvie. Instead of claiming their scientists had discovered the combination, Abbvie claimed it was the result of a virtual model. That seems to have been enough to convince USPTO that AbbVie’s combinations did not fail the test of obviousness.

You can find statements by Gilead starting at the bottom of page 20 of the 10-K PDF.

      “We own published and pending patent applications directed to the use of combinations for the treatment of HCV, and, specifically, to the combination of ledipasvir and sofosbuvir. Certain of our applications were filed before the AbbVie Patents. For this reason and others, we believe the AbbVie Patents are invalid.

Accordingly, in December 2013, we filed a lawsuit in the U.S. District Court for the District of Delaware seeking declaratory judgment that the AbbVie Patents are invalid and unenforceable, as well as other relief. We believe that Abbott Laboratories, Inc. and AbbVie conspired to eliminate competition in the HCV market by falsely representing to the USPTO that they, and not Gilead, invented methods of treating HCV using a combination of ledipasvir/sofosbuvir. In February and March 2014, AbbVie responded to our lawsuit by filing two lawsuits” etc.

Obviously, “If a court determines that the AbbVie Patents are valid and that we have infringed those claims, we may be required to obtain a license from and pay royalties to AbbVie to commercialize sofosbuvir combination products.”.

I’ve included a few links –

AbbVie (ABBV) Uses Patents To Ambush Gilead Sciences, Inc. (GILD)” by Mark Terry, Breaking News Staff, 11/13/2014 (biospace.com)

Gilead Sciences Gets Ambushed By The Patent Troll, AbbVie” by Small Pharma Analyst, Nov. 12, 2014 (seekingalpha.com). With 325 comments, most of the arguments are likely to have been made already.

The Seeking Alpha piece prompted – ‘“Seeking Alpha” labels Abbvie a patent troll.‘ by Lawrence B. Ebert, November 12, 2014 (blogspot.co.uk), although it was the author Small Pharma Analyst and not Seeking Alpha who used the ‘troll’ label.

The Seeking Alpha article is long and highly informative. The biospace and Seeking Alpha articles both use the word “ambush”. Both authors expect the case to drag out over years. Small Pharma Analyst supports the “troll” label by explaining the obstacles to Abbvie commercializing a copy of Harvoni, which only leaves the extraction of cash from Gilead as the way to exploit the relevant patent claims. Although firms that specialize in “patent trolling” have faced some headwinds in recent years, the ‘trolling’ is not illegal. (For the basics of patent claims, see “How do I read a patent? – the Claims” (bpmlegal.com) and find the “motor vehicle” example.)

Gilead’s accusation of conspiring to eliminate competition is more colorful than you normally see in a 10-K. I can understand them being outraged, but fear of massive loss would also be consistent with the tone of the statement. Gregg H. Alton, an Executive VP, is a member of “the U.S. Government’s Industry Trade Advisory Committee on Intellectual Property Rights” (under Senior leadership), and the company has been through enough litigation that inexperience is not a likely factor. See Small Pharma’s article for more fairly extreme accusations in Gilead’s lawsuits.

About a lawsuit against AbbVie – “Gilead in 2013: AbbVie sought to eliminate competition and dominate market for HCV drugs” by James Love, July 16, 2014 (keionline.org).

You might not expect gas pedals to have much to do with drug combinations, but see “Is Chunky Monkey an Obvious Combination?” by Peter Pitts, 5.23.07 (spectator.org).

About obvious combinations and prior art, in order of increasing unreadability – the U.K. (genericsweb.com), the European Patent Office (epo.org), and USPTO (uspto.gov).

Litigation with generic manufacturers

This mostly concerns Teva Pharmaceuticals, over Tenofovir Disoproxil Fumarate, Emtricitabine and Fixed-dose Combination of Emtricitabine, Tenofovir Disoproxil Fumarate and Efavirenz, and you can find the details on page 21 of the 10-K PDF.

It’s an obvious tactic for a generics manufacturer to try to break patents. It’s less expensive than developing new drugs, but when patents are successfuly broken, the generics manufacturer has no monopoly on the benefit.

Doctors of the World

Charity challenges Gilead’s European patent on hepatitis C therapy Sovaldi” (Ref: Bloomberg, The Guardian, CNBC, Doctors of the World – Médecins du Monde) by Joe Barber, February 10th, 2015 (firstwordpharma.com). Médecins du Monde claim that Sovaldi depends too much on a breakthrough at Cardiff University, and have filed a patent challenge with the European Patent Office. Médecins du Monde (Wikipedia) are a non-profit that provide medical care and campaign for equal access to healthcare.

Also in the piece, in January, India refused a patent for Sovaldi, implying that the inventive step was too small. I suspect their conclusion, because even if Gilead only tweaked a molecule as alleged, the effect of tweaking a molecule is not necessarily obvious. The tweak might not be easy to make, even if the chemical diagram does not change much. Both those points seem likely to be magnified when it’s only the metabolites that are active. Roche were an early leader in HCV, but their products were never good enough, which supports the view that success was not easy. The same applies to Bristol-Myers Squibb, who acquired key IP (see “Hepatitis C drug set to give Cardiff University financial boost” 12 January 2012 (bbc.co.uk)). If there was an obvious and easy way to build on the university research, it begs the question, why did the researchers at the university or at Bristol-Myers Squibb not make that last inventive step.

Awards for patent infringement

Awards in patent disputes are meant to compensate for lost profits and the extra costs incurred. That naturally includes backdated royalties. I’ve read that big awards for patent infringement are scaled back on appeal more often than not, but I can’t find a link which confirms that.

I googled “record patent awards” and the biggest I could see was “Jury Awards $1.67B to Drugmaker in Record-Breaking Patent Verdict” by Martha Neil, Jun 29, 2009 (abajournal.com).

Following that up – “Abbott Wins Reversal of J&J’s $1.67 Billion Patent Victory” by Susan Decker, February 23, 2011 (bloomberg.com).

The 2011 piece gives Abbott’s worldwide sales of Humira (the relevant product) at around $6.6 billion. If I assume the sales were the same in 2009, the ratio of award / sales = 0.2530303 or 25%. Although that’s for a record-breaking award (for patent infringement), the size of the annual sales would be a large factor. The ratio of award / sales is probably not record-breaking, but IMO it’s likely to be above average. 25% of sales is big when sales are big, but not likely to be a company-killer. Because pharma costs are mostly up-front, with high gross margins, an award like that does not look as bad as failing phase 3, not getting approval, or having very low sales due to a superior competing product, all normal risks for a pharma company.

I did not find another case where damages and the infringer’s sales were both reported, and the studies I found showed the academics had no interest in the relation between the two quantities. A single case is not much to go on but better than nothing.

The amount in a settlement depends on how the case develops, perceptions of the strength of a case and the likely size of an award, attitude, etc.

Pfizer Reaches $2.15 Billion Protonix Accord With Teva” by Sophia Pearson, Susan Decker and David Voreacos, June 12, 2013 (bloomberg.com)

Assuming Teva and Sun’s sales equaled Pfizer’s lost Protonix sales, their annual sales were –





      = 1094

Settlement / Annual sales
= 2150 / 1094
= 1.9653
= 197%

That’s much higher than the award / sales of 25% for the Abbott vs J&J case.

Assuming the infringing sales only occured in 2008, 2009 and 2010 (the patent expired in January 2011) –



      x 3 years


      = 3,282

Settlement / Total sales (over the three years)
= 2150 / 3,282
= 0.6550 or 65.5%

It was a clear case of jumping in before expiry, and Pfizer’s lost sales really were evident in a lower sales figure. Generics manufacturer’s have smaller gross margins, and the settlement would have hurt. Abbott vs J&J seems more relevant, but it’s just one case.

This lists various IP deals, awards and settlements – “Patent/copyright infringement lawsuits/licensing awards“.

These academics claim to have figured it out – “Explaining the “unpredictable”: An empirical analysis of U.S. patent infringement awards” (PDF) by Michael J. Mazzeo, Jonathan Hillel, Samantha Zyontz, Accepted 2 March 2013 (kellogg.northwestern.edu). I’m not convinced that forward citations are a good proxy for the economic value of patents. I’d guess that applying the methods to cases could produce an endless stream of ‘cookie-cutter’ articles.

Sofosbuvir aka Sovaldi bio-chemistry

Some knowledge of the bio-chemistry should help when assessing competitors’ pipelines, because there are no sales figures to go on. This is meant for people who don’t know much of the science, and while I’ve made it simple, it won’t be as reliable as if I’d regurgitated the technicalities.

The hepatitis C virus (HCV) hijacks a host cell’s machinery to make copies of itself. The virus has a strand of RNA that codes for making proteins, and those strands of RNA need to be reproduced. Each strand of RNA is actually a chain of nucleotides, and a new strand of virus RNA is made by adding one nucleotide at a time to the end of a growing chain.

One way to stop or reduce the virus’s reproduction is to give it some fake nucleotides. That’s a tricky proposition, because the fakes have to be close enough to the real thing to be added to the chain, but they can’t be exactly the same because the whole point is that a chain which includes a fake does not work like a normal strand of virus RNA. The fakes also have to be very effective against virus reproduction compared to the harm they do to the patient.

In the body, Sofosbuvir is turned into metabolites which are “nucleotide analogs”, which means fake nucleotides.

RNA can be understood as a kind of template for building proteins, with the template being read one nucleotide at a time. The nucleotide analogs (the fake building blocks) are designed so that when one of them is read, it’s interpreted as a kind of end-marker, with the effect of saying “stop what you’re doing, we’re finished”. Getting technical, the end marker is called a “stop codon”, and when it causes the chain-building to stop, it’s called “RNA chain termination”.

The chain-building involves a thing called a polymerase. A polymerase is an enzyme that builds a polymer, “polymer” is the word for any kind of molecular chain, and DNA and RNA are both polymers.

NS5B is the Hepatitis C virus’s RNA polymerase, which it uses to replicate its RNA.

If you see Sofosbuvir described as a “nucleotide analogue inhibitor of HCV NS5 B polymerase”, you should now have some idea of what that actually means –

nucleotide analogue – fake building block with a sneaky “stop” sign.
inhibitor of … polymerase – stops the polymerase from finishing a long molecule (by putting an early “Stop!” sign in the template).
HCV – the hepatitis C virus
NS5B – the virus’s RNA polymerase, which it uses to replicate its RNA.

When I read “nuc”, I assume it’s short for “nucleotide analogue” and that it probably inhibits something vital for virus reproduction. “nuc”s are generally good for most genotypes, because all versions of a virus have to make use of the same four or five kinds of “building-block” nucleotides. They are good to have in drug combinations for HCV because they work across many genotypes and act differenctly to other kinds of antiviral. The more efficient a “nuc” is, the less need there is for combining it with drugs that cause severe side-effects. It wasn’t easy to get a good “nuc”, and getting a “nuc” as good as sofosbuvir or better, which is also different enough to be patented, will not be easy.

It’s harder for a virus to evolve resistance to a combination of effective anti-viral agents, because a mutation that gives resistance to one of them should not allow the virus to reproduce and spread the resistance, due to the effect of the other antivirals.

One more term. You might see “Uridine” in a description of Sofosbuvir, as in

“Uridine nucleotide analogue inhibitor of HCV NS5 B polymerase”

There are five kinds of nucleotides. Three of them occur in both DNA and RNA, but the nucleotides built around thymine are only in DNA, and the nucleotides built around uracil are only in RNA. Uridine is a kind of nucleotide analog built around uracil (well, not quite, because it’s a nucleoside, but I can’t explain everything). You don’t need to make every building block a fake, you just need to have at least one fake in a sequence often enough (and not at the end), and faking the uracil nucleotides is good enough.

I’ve missed a lot out, (ribosomes, positive sense RNA, messenger RNA, etc.) which you can tell if you look up the links.

Wikipedia: Nucleoside analogue, Nucleotide, RNA polymerase, Hepatitis C virus, Translation (biology).

Other sources: Nucleoside and Nucleotide NS5B Polymerase Inhibitors (medscape.com), Nucleoside/Nucleotide Analogues (emedicinehealth.com, about hepatitis B).

Protease inhibitors

HIV and hepatitis C viruses make a number of different proteins by making them all in one long molecule, and then slicing the molecule into the parts they need. The slicing up is done by their protease enzymes. A protease inhibitor is an antiviral molecule which attaches itself to the protease enzyme and stops it from working. See “Protease inhibitors” (aidsmap.com).

Gilead have two HIV products which work in combination with protease inhibitors, and have a combination for HCV in Phase 2 which includes a protease inhibitor. (From the 10-K, slightly ediited – Fixed-dose combination of GS-9857, sofosbuvir and GS-5816 – GS-9857 is a pan-genotypic NS3 protease inhibitor being evaluated in combination with sofosbuvir and GS-5816 for the treatment of HCV.)


    “If successfully developed and approved by regulatory authorities, this treatment would represent the first protease inhibitor-based STR and thereby continue Janssen’s commitment to providing its HIV products in more simplified dosing presentations.” (10-K for 2014)

(STR is probably “single tablet regimen”.)

Merck’s HCV pipeline

Of the three Hepatitis C candidates listed on Merck’s pipeline page, one is a combination in phase 3, and the other two are combinations in phase 2.

The phase 3 combo is of grazoprevir and elbasvir, each of which is oral and once-daily, and for treating chronic hepatitis C. The combo is an Investigational New Drug (IND), which only means the company has permission from the FDA to ship the drug across state lines before a marketing application has been approved. The 10-K has –

    “MK-5172A, a once daily, fixed-dose, combination, chronic HCV treatment regimen consisting of MK-5172, grazoprevir, an investigational HCV NS3/4A protease inhibitor, and MK-8742, elbasvir, an investigational HCV NS5A replication complex inhibitor, began Phase 3 clinical trials in June 2014. MK-5172A is being investigated in a broad clinical program that includes studies in patients with multiple HCV genotypes who are treatment-naïve, treatment failures, or who fit into other important HCV subpopulations such as patients with cirrhosis and those co-infected with HIV.”

Elbasvir is described as a “HCV NS5A replication complex inhibitor”, which has some similarity to Gilead’s NS5A inhibitor ledipasvir, one of the drugs in Harvoni (the other is sofosbuvir, a NS5B polymerase inhibitor). Harvoni is on sale, but Gilead also have GS-5816, a pangenotypic NS5A inhibitor, and a combination with sofosbuvir is currently in Phase 3 clinical trials. AbbVie and Achillion each have a NS5A inhibitor In Phase 2 (AbbVie’s is ABT-530, and Achillion’s is ACH-3102).

The phase 2 entries on the pipeline page don’t have much detail. One has both grazoprevir and elbasvir, as for the phase 3 combo, and the other has grazoprevir without elbasvir. They both have MK-3682, which I describe later in this section.

From the news item “Interim Data from Proof-of-Concept Study of Merck’s Investigational Hepatitis C Treatment Grazoprevir/Elbasvir in Combination with a Nucleotide Inhibitor (C-SWIFT study) Presented at The Liver Meeting®” Sunday, November 9, 2014 (mercknewsroom.com),

    “To date, no discontinuations due to an adverse event and no drug-related serious adverse events have been reported. Most adverse events were mild or moderate in intensity, with no apparent dose effect.”

I don’t know just how serious the adverse events were. It remains to be seen just how bad they’ll be in phase 3, and the effect on demand, pricing and sales if the combo gets through the remaining stages.

(This section is about Merck’s HCV pipeline, but I’ll mention that the pipeline page has a HIV candidate in phase 3 – “MK-1439 is an investigational orally available HIV non-nucleoside reverse transcriptase inhibitor (NNRTI) being evaluated for the treatment of HIV infection.”, with a link for the clinical trials. The HIV candidate is “investigational”.)

There are two relevant tables in “Is Merck Ready To Soar Or Flop? The Challenging Question Of Predicting Pipeline Potential” by Pharma Doc, Mar. 17, 2015 (seekingalpha.com). The tables compare Gilead’s Harvoni, Abbvie’s Viekira Pak (both on sale), and Merck’s MK-5172A aka Grazprevir/Elbasvir (in Phase 3). The data used is from phase 2 and phase 3 clinical trials. Taking into account Abbvie’s higher pill count and the side effects, Merck’s candidate seems to be slightly inferior to Harvoni, but better than Abbvie’s Viekira Pak. “GT1” means “genotype 1”, and the author seems to regard MK-5172A as being only good for that genotype, probably for a good reason, although the quote above from Merck’s 10-K included “patients with multiple HCV genotypes”.

The author is also hopeful about an HCV combination made possible by Merck’s acquisition of Idenix. It looks like the combo is MK-3682/MK-8742 (elbasvir)/MK-5172 (grazoprevir), the phase 2 candidate with the mild or moderate adverse events. It seems to be aimed across genotypes, and would compete with Gilead’s Sovaldi. The 10-K says very little except “The Company expects to begin Phase 3 studies in 2015.”, and does not give the combo a name. The component MK-3682 was acquired from Idenix, and it gets –

    “MK-3682 is a nucleotide prodrug in Phase 2 clinical development being evaluated for potential inclusion in the development of all oral, pangenotypic fixed-dose combination regimens.” (Merck’s 10-K)

“prodrug” means it’s the metabolites which are active against viruses (as for Gilead’s sofosbuvir, aka Sovaldi).

“nucleotide” is also in common with Gilead’s sofosbuvir, which is described as a “nucleotide analog polymerase inhibitor” in the 10-K (nucleotides are described in “Sofosbuvir aka Sovaldi bio-chemistry”, above).

In other words, Merck’s MK-3682 looks like a “nuc”, and because of the adverse events it’s probably not as good as Gilead’s “nuc” (Sofosbuvir), even though the “adverse events were mild or moderate in intensity”.

Achillion’s HCV pipeline

The pipeline shows four compounds, two in phase 2, one in phase 1, and a compound between discovery and preclinical. Clicking the names gets more detailed information.

The first compound, in phase 2, is called ACH-3102. It’s an HCV NS5A Inhibitor, which Achillion claim is “second generation”. They also say it’s been “fast tracked” by the FDA, but I wouldn’t be surprised if technically it’s been given a Breakthrough Therapy Designation. There’s a pilot Phase 2 for the use of ACH-3102 in combination with Gilead’s sofosbuvir, and that might be all the compound’s phase 2 activity. Gilead, Merck and AbbVie have NS5A inhibitors in trials, find “Elbasvir is described as a” above. Gilead’s Harvoni includes a NS5A inhibitor and is selling well.

Achillion’s phase 2 candidate pairs their HCV NS5A Inhibitor with Gilead’s HCV NS5B Inhibitor (sofosbuvir). I’m sure they would rather use their own NS5B polymerase inhibitor if they could. That’s the second compound on the pipeline page, in phase 1, a pro-drug uridine nucleotide analog NS5B polymerase inhibitor (after rearranging the words), the same description as Gilead’s Sovaldi/Sofosbuvir (so it’s a “nuc”). It does not seem to be as good as sofosbuvir, as it needs to be taken with ribavirin or pegylated interferon or both, which have side effects. There are some other criticisms in Achillion: The Dark Side by Kanak Kanti De, Feb. 26, 2015 (seekingalpha.com).

Nasdaq give Achillion a market cap around $1.24 billion. Revenue so far has been zero. Liquid assets from cash to receivables are nearly $160 million.

UPDATE: Achillion Pharmaceuticals (ACHN) Will Be Left Behind, Barclays Starts at Underweight” March 3, 2015 (streetinsider.com) – the short piece looks a few years ahead, and puts Gilead and Merck ahead of Abbvie, but the opinions are not explained.

For more research into Achillion, you could start with Seeking Alpha. One recent news item is the resignation of the Chief Regulatory Officer.

My own research into Achillion has not gone much farther than what you see here.

Selling a combination which is mostly sofosbuvir is great news for Gilead, but it might not be such good news for Achillion. I haven’t seen anything from them that looks like it would beat or equal sofosbuvir in a combination, which seems to make them less of a threat to Gilead than almost any company with positive cash flow, ambition in the area, and some relevance to the market. Cash-burners can also find it harder to raise cash when the market they hope to address is declining, although in this case the peak has not yet been reached, and Achillion have been able to raise capital. (Find “prevalence peaked in the 1990s”, below.) Sometimes genius is surpassed by even greater genius, and that’s probably what Achillion would need to be a threat to Gilead. I’m not saying Achillion are doomed, just that they are not likely to be a serious threat, based on the little information I’ve gathered.

Achillion’s NS5A inhibitor in phase 2 would be more valuable to Gilead than to anyone else, because owning it would mean owning the whole combination. If the combination of the inhibitor and sofosbuvir has not been spat out by AbbVie’s magic model, owning the combination would reduce Gilead’s vulnerability if they lose the patent dispute. But, the combination might not be as good as the candidates in Gilead’s liver disease pipeline, and Gilead might not want to show interest in an acquisition while Achillion’s stock is high on acquisition hopes.

If Merck acquires Achillion, it would be with the intention of using Achillion’s HCV NS5A Inhibitor with their own HCV NS5B Inhibitor (because they don’t need another NS5B “nuc” which is not as good as Gilead’s). Trials for the combination would have to start from scratch. While the combination is likely to be inferior to the NS5A Inhibitor/Sofosbuvir combination which Achillion are putting through trials, it might still improve Merck’s position in HCV relative to Gilead, but the necessary clinical trials would take years, and the market is due to peak. Merck could face complications if they apply for a “breakthrough” designation with the FDA, because the combination would probably be inferior to Achillion’s combination (with Gilead’s Sofosbuvir), and killing that combination to make their own rank higher would probably not impress the FDA. Against that last point, Achillion were hoping to replace Sofosbuvir with their own version as much as they could, and the problem seems to be with the quality of their version, not the attitude of the FDA.

When the FDA rescinded its Breakthrough Therapy designation for Merck’s HCV combination of grazoprevir and elbasvir, Achillion’s stock fell, apparently because Achillion’s breakthrough designation might be at risk, and future breakthrough designations are less likely, as the FDA’s action shows they believe the number and standard of HCV treatments warrants raising the bar. See “Why Achillion Pharmaceuticals, Inc. Stock Is Crashing Today” by George Budwell, February 4, 2015 (fool.com).

Abbvie’s HCV pipeline

Abbvie’s HCV pipeline has a combination for HCV genotype 1 which does not need Interferon, in “PHASE III/SUBMITTED”, and a combo in phase 2. There’s no information about phase 1.

From “AbbVie’s (ABBV) CEO Richard Gonzalez on Q4 2014 Results – Earnings Call Transcript” Jan. 30, 2015 (www.seekingalpha.com) –

    “Clearly another important driver of performance in 2015 will be our interferon-free HCV therapy Viteron [ph] which is now been approved in the US, EU and a number of other countries around the world. We are pleased with Viekira product label and updated AASLD treatment guidelines and we believe both reflect the strength of the product’s clinical profile across genotype-1 patient population.”

I can’t find anything else about “Viteron”, and I’m wondering if it’s a transcription error. I can’t see it on “2014 FDA Approved Treatments For Hepatitis C” or “New Hepatitis C Drugs Coming in 2015” by Nicole Cutler L.Ac., January 12, 201 (hepatitiscentral.com).

All I could find in AbbVie’s 10-K about their HCV pipeline was –

    “AbbVie is also currently conducting Phase 2 studies of its next-generation HCV program which includes ABT-493, a potent protease inhibitor, and ABT-530, AbbVie’s new NS5A inhibitor.”

Gilead, Merck and AbbVie have NS5A inhibitors in trials, find “Elbasvir is described as a” above. Gilead’s Harvoni includes a NS5A inhibitor and is selling well.

I have not found much detail about Abbvie’s HCV pipeline. In addition to their normal development pipeline, Abbvie’s virtual model and legal action can be seen as an another way to get income from intellectual property in addition to acquisitions or developing the pipeline.

Further research into competition in the pipeline

One line of research is to go to www.clinicaltrials.gov and search for HCV. Not every result will be relevant. The site was not designed for investors and I haven’t seen any info linking trials to companies.

Recent financial results

In the table below, I’ve noticed that I have not been consistent about making expenses negative and red. I needed most of them to be positive, for the appearance of the chart (the 3D chart after the table).

Gilead 3 years earnings to 2014 spread

Gilead 3 years earnings 3D

Gilead margins

Gilead cash flow 2012 - 2014 spread

Gilead cash flow 2012 - 2014

Gilead balance sheet spread

Gilead balance sheet

About the “Long-term marketable securities”, I believe they are relatively realizable. On page 93 of the 10-K PDF for 2014, they are classed as “available-for-sale securities”, of which only $18 million have a contractual maturity over 5 years (compared to $1,598 million of Long-term marketable securities). None of the assets recorded at fair value are “Level 3”, the hardest of the three levels to assess (for 2013 or 2014).

To see figures for the fantastic growth from 2005, and excellent returns metrics, see Morningstar’s Ratio tab (morningstar.com/ratios).


Sometimes companies are too enthusiastic about capitalizing costs to make them look like investment instead of expenses. Nothing in the investment cash flow suggested that. I found “We had unamortized capitalized software costs on our Consolidated Balance Sheets of $80 million as of December 31, 2014 and $84 million as of December 31, 2013.”, and “We capitalized $20 million related to the milestone incurred in connection with the FDA approval of Stribild and $12 million related to the milestone we incurred in connection with the European Commission approval. Both milestones are being amortized over the useful patent life of elvitegravir, which is approximately 10 years, expiring in 2023.”. The “accumulated depreciation and amortization” of $620 million includes “$2 for 2014 and 2013 relating to capitalized leased equipment”, i.e. $2 million, and while the value of capitalized leased equipment will be a lot bigger than that I’d guess it’s still small. Because life is short and the items are small compared to the $2,854 million R&D expense, I did not worry about whether or not the capitalizations were proper.

Full Year 2015 Guidance

The guidance is in the earnings call, the earnings release and the “2015 Guidance” PDF on Gilead’s Earnings page. In the following table, the first three columns show the 2015 guidance, and they are likely to be the most reliable columns.

Guidance only gave Product sales, and not Royalty, contract and other revenues. Instead of adjusting guidance to include royalties etc., I adjusted the 2014 figures where necessary to exclude the royalties etc., to give a reasonably fair comparison with the guidance, although I’m not asking anyone to trust the comparison a lot. A full explanation would have taken too much time and space. If you need to know what I did, you can find the formulas by finding “The spreadsheet formulas” above, for a link. I calculated the 2014 non-GAAP gross margin backwards from the non-GAAP EPS, and if you find a better source, that might not be necessary.

It was reasonable to put the highest R&D expenses under the ‘Worst case’ columns, but I don’t mean to imply that high R&D expense is bad. Similarly, SG&A expense is necessary to build the business, and a higher expense is only likely to be bad if it’s the result of inflation or inefficiency.

Gilead guidance for 2015 spread

Gilead beat Q4 analyst estimates after Q3 fell short (according to Zacks on Yahoo). Apparently the outlook disappointed. Estimates were “crushed” back in Q1 according to Gilead Beats by a Billion” by: Ian Wyatt, 05/09/2014 (moneyshow.com), which shows that analysts can be wrong about Gilead.

2014 net product sales beat revised guidance. Non-GAAP R&D expenses were bigger than guided, due to two one-offs – buying a voucher for FDA priority review, and a collaboration with Ono Pharmaceutical, possibly buying the rights to Ono’s Once daily BTK inhibitor outside of Japan, China and ASEAN countries (for the treatment of B-cell malignancies and other diseases).

I thought Gilead’s 2015 guidance was probably conservative, and I’m not alone – find “Michael Yee and RBC Capital” below.

Compensated and decompensated cirrhosis of the liver

Anyone with a liver problem should refer to reputable sources and not rely on my writing (investors can do the same, if they have the time).

According to Wikipedia, 30% of cases of cirrhosis are caused by hepatitis B, and 27% by hepatitis C. Those are global figures and they could be lower for the U.S. and other developed countries. The second most common cause is alcohol consumption. Wikipedia gives bullet points to 16 causes.

Cirrhosis of the liver is permanent scarring, which results when the liver tries to repair damage. The damage means it doesn’t work as well.

Compensated cirrhosis is the first stage, with relatively light damage, and patients might not be aware of anything wrong at first. Symptoms include feeling tired, nausea, abdominal pain, loss of appetite, weight loss, and small red spots on the skin called spider angiomas.

Decompensated cirrhosis is the second stage, with serious liver damage. Symptoms include fluid buildup in the legs, feet and abdomen, itching, bruising and bleeding, and the skin yellowing (jaundice). For the non-squeamish, there’s a picture on Wikipedia.

For more symptoms and detail, see Wikipedia or “Cirrhosis” (umm.edu/health).

Pharmasset acquisition

The Pharmasset acquisition (for hepatitis C drugs) was criticized at the time due to the price and the lack of meaningful revenue. For one example of the criticism, find “Paying Too Much” in “Gilead to Buy Pharmasset for $11 Billion to Win in Hepatitis” by Margaret Tirrell and Ryan Flinn, November 21, 2011 (bloomberg.com). Gilead were already the leaders in HIV, and now lead in hepatitis C. One risk is that another big deal might not go as well, although the company is now bigger and could survive a setback better.

With Pharmasset, Gilead acquired Sofosbuvir which is now sold as Sovaldi. Harvoni contains Ledipasvir (which Gilead already had) and Sofosbuvir.

The hepatitis C cures

Sovaldi is generally useful against HCV across genotypes as part of a combination, while Harvoni targets genotype 1, the common genotype which infects 70% of HCV patients in the U.S..

Sovaldi is the commercial name for the drug sofosbuvir. It was approved in the U.S. in December 2013, and by the European Commission in January 2014 for genotypes 1 to 6. Gilead claimed high rates of cure and a shortened course of therapy (12 weeks), patients who can’t take Interferon had their first option for a completely oral treatment, and previously there was no effective regimen to stop HCV from recurring after a liver transplant, that could be taken while waiting for a transplant. See this News release about the E.C. approval. The claim that Sovaldi was good for genotypes 1 to 6 was qualified by –

    “The clinical data supporting the use of Sovaldi in patients with genotypes 5 and 6 is limited.”

but the 10-K for 2014 has –

    “Sovaldi’s efficacy has been established in patients with HCV genotypes 1, 2, 3 or 4 infection (in United States and Europe) and genotypes 5 and 6 infection (in Europe),”

The problem may be that genotypes 4, 5 and 6 are uncommon in Europe.

Harvoni combines ledipasvir with sofosbuvir (aka Sovaldi). It was approved in the U.S. in October 2014, and in the E.U. in November 2014. Harvoni is a single pill once a day, benefiting the “type 1” patients who needed to take interferon and ribavirin with Sovaldi, or use Viekira Pak from competitor AbbVie with a higher pill count. The pill count affects how regularly patients take the pills, and the regimen is obviously less effective if the pills are missed or taken at the wrong time.

From “Gilead Sciences’ (GILD) Management Presents at Cowen & Company 35th Annual Health Care Conference (Transcript)” Mar. 3, 2015 (www.seekingalpha.com),

    “Harvoni is recommended for genotype 1 patients and requires a dramatically lower pill count compared to Viekira Pak.”.

Sovaldi sold $10 billion worldwide in 2014, with 32,000 patients treated. 31,000 patients started treatment with Harvoni between launch and the end of 2014.

Market share in HIV and Hepatitis C

Biotech Head-to-Head: Gilead vs. Celgene” by Todd Campbell and Michael Douglass, February 26, 2015 (fool.com). An industry analyst (and enthusiastic shareholder) implies that around 80% of new HIV patients have a treatment which involves a drug from Gilead.

This site has too much page junk, IMO – “Gilead Sciences, Inc. Continues To Lead The Hepatitis C Market; Its HCV Drugs See 4.5% Growth” by Hannah Ishmael, Feb 28, 2015 (bidnessetc.com). The prices in the article look higher than the prices likely to apply in practice, find the link with “US pharmacy deals could cut costs”, below.

Why Gilead Sciences Doesn’t Seem Too Worried About AbbVie’s Hep-C Treatment” by Ben Levisohn, February 17, 2015 (blogs.barrons.com).

This is probably the most relevant link – “The Hepatitis C Scorecard: Gilead is Trouncing AbbVie, but at a Price” by Ed Silverman, Feb 12, 2015 (blogs.wsj.com). The ‘price’ is the discounts Gilead had to make, find “46% overall discount”, below. Also find “Info about prescriptions” below, about getting weekly figures for US prescriptions.

Hepatitis C (HCV) numbers

2% to 3% of the world’s population are estimated to be infected by hepatitis C, with a lower rate of infection in developed countries, see “Hepatitis C” by Deborah Holtzman (Centers for Disease Control and Prevention). Applying 2% to 3% to a world population of 7.2 billion gives 144 million to 216 million.

Many cases are undiagnosed, and when diagnosed, many cases are not treated unless or until there are serious effects, such as cirrhosis of the liver. About the delay in treating diagnosed cases –

    “in many countries patients are unlikely to use prescription drugs that are not reimbursed by their governments. In addition, negotiating prices with certain governmental authorities can delay commercialization by 12 months or more.” (10-K for 2014)

Infection is through sharing or reusing hypodermic needles, or unscreened blood transfusions or unscreened organ transplants. In developed countries, screening is much more effective than in the past. It’s harder to reduce infection by intravenous drug users who share needles.

The WHO has reported the prevalence of transfusion-transmissible infections (TTI) in high, middle and low income countries. For HCV prevalence in blood donations, the figures are –

      0.02% for high-income countries


      0.37% for middle-income countries


    1.07% for low-income countries

but with quite wide ranges in each case.

The figures are from “Blood safety and availability” Fact sheet N°279, Updated June 2014 (who.int).

Myanmar is an exception I did not expect, see “Myanmar National Blood Center: Receiving a Prestigious International Award” June 1, 2014 (jica.go.jp).

The piece in the next link claims that prevalence peaked in the United States in the 1990s and will be down to a third of the peak level by 2030. Serious liver disease peaks later than prevalence, and the delay should mean that the incidence of disease can be predicted more easily. “Revolutionizing Treatment Outcomes in Hepatitis C: Managed Care Implications and Considerations—The New and Evolving Standards of Care” by Gary M. Owens, MD, March 24, 2015 (ajmc.com)

This is a bit old, but it looks quite far ahead – “U.S. Hepatitis C-Related Health Care Costs to Peak at $9.1B in 2024” January 24, 2013 (hepmag.com).

The paper – “Chronic Hepatitis C Virus (HCV) Disease Burden and Cost in the United States” by Razavi H, Elkhoury AC, Elbasha E, et al., Hepatology June 2013, online May 6, 2013 (ncbi.nlm.nih.gov) – looks like the source of at least some articles with projections. The abstract, conclusion, and much of the rest are readable by non-academics, although there’s a high density of facts and figures. It’s easy enough to get the general idea from four charts showing various peaks. Bearing in mind the title (so the costs should be HCV-related, and U.S. only), paraphrased from the paper –

      Prevalence of compensated cirrhosis was expected to peak at 626,500 in 2015.


      Prevalence of decompensated cirrhosis was expected to peak at 107,400 in 2019.


      The annual cost of compensated cirrhosis of the liver was expected to peak at $1.9 billion in 2022.


      The annual cost of decompensated cirrhosis of the liver was expected to peak at $4.2 billion in 2025.


    The annual cost of chronic hepatitis C (CHC) was expected to peak at $1.4 billion in 2025.

(Technically, prevalence should be a proportion, but I’m not trying to write a Phd thesis.)

I probably shouldn’t add peaks for different years, but I have, and it comes to $7.5 billion. Those projections may be out of date already, but it’s hard to be sure because I have not found figures which are higher, more recent, and confined to the same categories. For comparison, a chart near the top of this piece shows Sovaldi sold $10,283 million, and Harvoni sold $2,127 million, in 2014. The sales will be heavily U.S. weighted, and Harvoni was not on sale for long. While prices have come down, access has been improved.

The article “New hepatitis C drugs predicted to place a dramatic financial strain on health care system” University of Texas M. D. Anderson Cancer Center, March 16, 2015 (sciencedaily.com), gives a projection which amounts to an average cost of $27.2 billion per year over the next five years, for hepatitis C drugs. Another source projected spending on hepatitis C drugs in the U.S. to reach over $20 billion in 2018, getting there with growth over 6% p.a., find the link “At $84,000 Gilead Hepatitis C Drug Sets Off Payer Revolt”, below (the info is under the heading “Increasing Costs”).

The figures in the bullet points below are from “Gilead Sciences (GILD) CFO Robin Washington Presents at RBC Healthcare Conference (Transcript)” Feb. 27, 2015 (on www.seekingalpha.com). I’m not sure what all the figures mean, for example the prevalence in Italy is probably all cases of HCV infection, but it was not specified.

• Patients treated in 2014 in the United States with either Sovaldi or Harvoni, 140,000.

• Expected to be treated in the U.S. in 2015, under 250,000 (240,000 and 250,000 were also mentioned).

• Diagnosed patients in the U.S., 1.6 million (that would last 6.4 years if 250,000 were treated per year, without counting new cases).

• Undiagnosed cases in the U.S, 2 million, but prevalence is estimated at over four million (which is more than the sum of diagnosed and undiagnosed cases, 1.6 + 2 million = 3.6 million).

(It’s also been claimed that 3.2 million Americans are infected with HCV, and more than three quarters of them develop chronic infection, which can be without symptoms until damage to the liver causes them. Find the link “FDA Approves Olysio (simeprevir) in Combination with Sofosbuvir for Genotype 1 Chronic Hepatitis C Infection”, below.)

Advertising and education could help to get more cases diagnosed. The current attitude in the medical profession is that early-stage liver disease does not need treatment urgently, so there will be a delay between diagnosis and treatment, and not every case is likely to be treated.

The limited number of doctors and the patients they can see is also a ceiling on the rate at which diagnosed patients can be treated.

The size of the market puts a ceiling on the opportunity, and competitors will take at least some market share.

While not stated, the numbers above seem to be U.S. only.

• Expected to be treated in 2015 in Europe, 100,000.

• Prevalence in Italy, 1.5 million (unusually high, and prevalence could be above average in other southern European countries).

• Run rate in Europe, $2 billion per year. The run rate probably assumes that sales are level with the most relevant recent period, and is not a projection. It’s based on sales in only two countries, France and Germany, Sovaldi only (no Harvoni sales), with restrictions that are likely to become looser because prices have been agreed.
Agreements in Europe vary between countries, and some are confidential. In Europe, lower prices result from higher volume, rather than better access. Until the data is in, it’s not easy for Gilead to give an average European price, probably for any product with potentially high volume, and it’s harder for anyone else, due to the confidential agreements.

The suggestion that prices in Europe would be 65% to 70% of the U.S. price was not “totally out of the ballpark” according to the CFO.

• Number of HCV patients in Japan, a million. (The article “A No-Nonsense Projection Of Gilead In 2015” by Anthony Clarke, Mar. 18, 2015 (seekingalpha.com), has a lot about Japan. See also “Japan’s Ministry of Health, Labour and Welfare Approves Gilead’s Sovaldi® (sofosbuvir) for the Treatment of Genotype 2 Chronic Hepatitis C” Mar. 26, 2015 (gilead.com).)

• Proportion of genotype 2 and 3 patients in Japan, 30%. Harvoni won’t be prescribed for genotype-2, but Sovaldi could be.

    “Marketing applications for sofosbuvir and the fixed-dose combination of ledipasvir and sofosbuvir are pending in Japan.” (10-K for 2014)

(I’ll remind that sofosbuvir is known commercially as Sovaldi, it’s one of the two ingredients in Harvoni, and Harvoni targets HCV genotype 1.)

According to the analyst Michael Yee, genotype 1 represents the other 70% of the HCV market in Japan. Bristol-Myers Squibb got there first with a product that’s only good for genotype 1, with about a quarter of a billion sales (USD) in a quarter.

Revenue from Japan (for HCV) is likely to start near the end of 2015.

BTW I don’t usually have much praise for analysts, but Michael Yee asked some pertinent questions about how hepatitis C sales will trend over several years.

A ‘promising’ HCV vaccine

A “Promising new HCV vaccine” by Audrey Lin, 11 November 2014 (applevir.org), seems some way off.

The genetic diversity of HCV is a problem mentioned in the paper. Some people with HCV spontaneously clear the virus, and the vaccine is based on the T-cell response seen when that occurs. 15 healthy volunteers showed the right kind of T-cell response, and the vaccine appeared safe. The author says the vaccine “shows promise”.

Hepatitis B (HBV)

Worldwide, about 240 million people have chronic HBV, see Hepatitis B by Francisco Averhoff (Centers for Disease Control and Prevention).

Viread is used to treat HIV, but it’s also listed under “Liver diseases” and is …

    “dosed once a day for the treatment of chronic HBV in adults with compensated and decompensated liver disease.” (10-K for 2014)

It’s licensed to GlaxoSmithKline Inc. (GSK) for chronic HBV in China, Japan and Saudi Arabia.

Hepsera is listed under “Liver diseases” and is dosed once a day to treat chronic HBV, in patients 12 years old or older. It’s licensed to GSK for the treatment of chronic HBV in Asia Pacific, Latin America and some other places.

The pipeline includes –

      TAF (nucleotide reverse transcriptase inhibitor) (in phase 3)


      (Phase 3 has been completed for TAF for HIV.)


      GS-4774 (Tarmogen T cell immunity stimulator) (in phase 2)


    GS-9620 (TLR-7 agonist) (in phase 2)

Apart from the phase 3 results for TAF for HIV, the info is from Gilead’s Q4 2014 earnings slides PDF.

I don’t know what proportion of HBV patients could be treated with Gilead’s approved or candidate products.

HIV numbers

See “The HIV/AIDS Epidemic in the United States” Apr 07, 2014 (kff.org). Over 1.1 million people are infected in the U.S. New infections peaked in the 1980s, but have remained stable at around 50,000 for over a decade.

The WHO puts the number of people with HIV/AIDS worldwide at 35 million, in 2013 – “HIV/AIDS” (who.int)

From “UNAIDS reports that reaching Fast-Track Targets will avert nearly 28 million new HIV infections and end the AIDS epidemic as a global health threat by 2030” 18 November 2014 (unaids.org) –

    “We have bent the trajectory of the epidemic,” said Michel Sidibé, Executive Director of UNAIDS. “Now we have five years to break it for good or risk the epidemic rebounding out of control.”

Fast-tracked competitors

The FDA can give a drug in development either “Fast Track” status, or a “Breakthrough Therapy Designation“. In each case, the FDA will expedite the approvals process, although it depends on the company not delaying its communications with the FDA. The links give clear explanations. Investors tend to say that a treatment has been “fast-tracked” if it’s actually been given the breakthrough designation.

The criteria in the links for both designations do not include increasing competition. I can’t say “The FDA will never fast-track a product in order to increase competition.”, but I know of no evidence beyond the fact that most approvals probably hurt a competitor.

Two competing hepatitis C products have been fast-tracked, and then taken off the fast track, due to the existence of four licensed products. See “Merck’s hep C drug no longer a ‘breakthrough’” by Thomas Meek, 5th February 2015, (pmlive.com), where the list of four new treatments starts with two from Gilead, and see “BMS follows Merck as glut of HCV drugs leads to loss of fast-track status” By Dan Stanton+, 11-Feb-2015 (in-pharmatechnologist.com). The ‘glut’ is the same four products as in the previous link.

Pricing in the U.S. and other developed countries

The piece “US pharmacy deals could cut costs of AbbVie and Gilead’s HCV treatments” By Dan Stanton+, 06-Jan-2015 (in-pharmatechnologist.com), allows the quote –

    “Gilead and AbbVie’s exclusive deals with payers CVS and Express Scripts could help cut the cost of the new class of treatments for America’s three million Hepatitis C sufferers.”

Governments don’t like high prices for drugs. The tax in France (later) was an open measure to claw back cash one way or another. It’s normal for drugs to start at a higher average price and sell mostly to private buyers, and for the price to come down as more sales are made to public agencies, and as agreements are made with the large buyers. Management have stated in earnings calls that the process has been accelerated (a physicist would say no, it’s been faster). The average price is also affected as sales build in other countries.

Typically, in return for rebates (from Gilead), restrictions can be eased, such as how badly diseased a liver has to be before an insurer pays for the relevant medication. (Find “The insurance companies that Gilead” for a comment by rational2168 under the Q4 2014 earnings call transcript, and there are other well-informed comments there).

This gives a flavor of the opposition to Gilead’s early pricing – “Activists Hold Die-In to Protest High Price of Gilead’s Hepatitis C Drug” by Michel Sidibé, UNAIDS, July 24, 2014 (treatmentactiongroup.org). The article grives a price range of US$84,000 to $168,000 for the Sovaldi regimen in the United States. Less committed sources stick with $84,000 – “At $84,000 Gilead Hepatitis C Drug Sets Off Payer Revolt” by Drew Armstrong, January 27, 2014 (bloomberg.com). For comparison – “U.S. FDA approves AbbVie hepatitis C drug, costs $83,319 for 12 weeks” by Caroline Humer, Mon Dec 22, 2014 (reuters.com). See also “Payers hit back at Gilead for $94,500 price tag on brand-new hep C combo pill” by Carly Helfand, October 13, 2014 (fiercepharma.com).

About interest from the U.S. Senate –

    “In July 2014, we received a letter from the U.S. Senate Committee on Finance requesting information and supporting documentation from us related to Sovaldi and the pricing of Sovaldi in the United States. The letter raised concerns about our approach to pricing Sovaldi, its affordability and its impact on federal government spending and public health. We are cooperating with the inquiries. It is both costly and time consuming for us to comply with these inquiries. We cannot predict the outcome.” (10-K)

See “US lawmakers question price of Gilead’s hepatitis C drug Sovaldi” by: Joe Barber, July 11th, 2014 (firstwordpharma.com).

That might not be much of an issue now, as I could not find “letter”, “Senate” or “Committee” in the earnings call transcript, or the RBC or Cowen & Co conference transcripts, although there might not have been much point in asking about it before any further news.

The media frequently highlighted the “$1,000 a day” or “$1,000 per pill” price tags (google – Gilead $1,000 a pill). A cure is better than a lifetime palliative, and requires fewer pills. A short course of treatment is better and requires fewer pills. A single pill is better than mutliple pills, and requires fewer pills. Once per day is better than more than once per day, and requires fewer pills. Pills are better than most other methods of treatment (exceptions, such as acidic pills given to elderly patients, don’t seem to be relevant here). The total cost of a course of treatment matters, as well as all the benefits including the benefits of convenient administration. “$1,000 per pill” does not reflect those considerations.

While initial prices may have been too high, it’s normal for the price to come down. There’s a third factor for Harvoni – volumes were surprisingly high, or shockingly high for the the PBMs (Pharmacy Benefit Managers) who had to pay, so they were keen to get rebates, and in return, Gilead got improved access (meaning Harvoni was available to a wider set of patients). I don’t go into the ability of PBMs to pass on the cost, because the issue here is not who bears the ultimate cost, and PBMs don’t say “never mind, we’ll just pass the cost on”, they resist high prices.

Price variation in the U.S.A.

There is not usually a single price for a drug. Pharmacy Benefit Managers (PBMs) like Express Scripts (ESRX) buy in bulk and negotiate discounts. They can get a bigger discount in return for an exclusive agreement, to buy only one company’s product for a particular condition. Some patients with the condition could have a variation which means they can’t be prescribed the product with the exclusive agreement, and the buyers have to buy from a pharma company which won’t be happy about being excluded. The excluded company will not want to give any discount on such sales (and would like to charge as much as possible). The resulting price will be higher than the price charged where the company has agreed a lower price in return for volume. These links have information about exclusive deals, but not about the revenge of excluded companies – “Sorry, Gilead. AbbVie cuts exclusive hep C deal with Express Scripts” by Tracy Staton, December 22, 2014 (fiercepharma.com). That’s fairly anti-Gilead, and for balance there’s “How Drug Company Gilead Outpaces Its Competitors—And Common Diseases” by J.J. McCorvey (fastcompany.com) which also mentions Express Scripts. Gilead’s CFO Robin Washington said –

    “If you take a look at the large PBMs, nine out of 10 have ultimately chosen Harvoni. And of those nine, eight have chosen exclusive.” (seekingalpha.com)

For the source and a link, find “RBC Healthcare Conference (Transcript)” above.

It’s possible that sometimes when exclusive deals are referred to, that might be short for “preferred or exclusive formulary status”, a phrase used in the link “The Hepatitis C Scorecard: Gilead is Trouncing AbbVie, but at a Price” (above). I can’t prove that any use of “exclusive” ought to have been “preferred or exclusive”.

So I can write more about price variation, I’ll assume here that Gilead’s IP is rock-solid – If a competitor sells a combination which includes a Gilead product, the combination could cannibalize the product’s sales. Gilead’s management would have to be crazy to agree a discount big enough to make the total revenue smaller, the more of the competitor’s combination is sold. Gilead could agree a discount if the combination gained access to patients, if the cannibalization of Gilead’s sales was low enough or non-existent. That’s another kind of case where pricing can be at the high end. When two companies use each other’s products in combinations, the terms are more likely to be reasonable, to the extent that the relationship is equal. In such cases the pricing is likely to be closer to the pricing that a single company would choose based on normal commercial factors.

For an example of a competitor using a Gilead product in a combination, see “FDA Approves Olysio (simeprevir) in Combination with Sofosbuvir for Genotype 1 Chronic Hepatitis C Infection” Nov. 5, 2014 (drugs.com). Olysio / simeprevir belongs to Janssen Therapeutics. Gilead’s 10-K mentions the drug at the end of …

    “Our HCV products, Sovaldi and Harvoni, compete with a product marketed by AbbVie Inc. (Abbvie) and Janssen R&D Ireland’s Olysio (simeprevir) in the United States.”

… under “We face significant competition.”.

The relationship is reversed for Gilead’s HIV combo Complera/Eviplera, which includes a Janssen product.

    “Complera/Eviplera is an oral formulation dosed once a day for the treatment of HIV-1 infection in adults. The product, marketed in the United States as Complera and in Europe as Eviplera, is our second complete single tablet regimen for the treatment of HIV and is a fixed-dose combination of our antiretroviral medications, Viread and Emtriva, and Janssen’s non-nucleoside reverse transcriptase inhibitor, Edurant (rilpivirine).”

The 10-K lists Janssen under “Commercial Collaborations”, as well as naming them as a competitor. Janssen are owned by Johnson & Johnson.

I’ll repeat here that government agencies pay less.

About guidance, the CEO said –

    “What we have in there is the range of different prices depending on the mix of private to public and there are a range of different discounts offered across all those areas.” from the Seeking Alpha transcript “Gilead Sciences’ (GILD) CEO John Martin on Q4 2014 Results – Earnings Call Transcript”, linked to above.

From the same transcript, about “gross to net” –

    “We expect our 2015 growth to net adjustments for our HCV products in the United States to be approximately 46% or a little more than double of that where we ended 2014 which was around 22%.”

The 46% overall discount is sometimes called “Gross to net”, where “gross” is at the list price and “net” is net of rebates or discounts. “growth to net” in the quote looks like a transcription error.

The effect of exclusive deals

The exclusive deals that Gilead and Abbvie have signed with the major PBMs don’t seem to leave much room in the U.S. market for treatments in other companies’ pipelines (find “Merck’s HCV pipeline” and “Achillion’s HCV pipeline”, above). I don’t know how long the agreements last, but I”d guess that anything much more than a year or two would be anti-competitive, because it would deny patients the benefit of innovation from other companies. Gilead have stated that they prefer letting doctors choose, but they can’t be expected to refuse exclusive agreements when that’s what the PBMs want, and when the giant Express Scripts signed an exclusive agreement with a competitor.

One reason for the PBMs’ enthusiasm for exclusive deals in this area may be that the drugs work, and they are safe, leaving limited room for improvement, even though Gilead’s liver disease pipeline is aimed at doing just that.

Obviously a U.S. deal does not affect sales in the rest of the world. If the U.S. market is sewn-up for a while, it does not necessarily mean that excluded companies have poor prospects in HCV, although the U.S. is a major market and it’s expected to peak.

The TRIPS agreement

The Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS), is generally what you’d expect from the name. The HIV/AIDS epidemic in Africa prompted some loosening of IP protection to allow access to essential medicines. A modification in 2003 allowed exports from one developing country to another, with restrictions such as packaging which is clearly different from the packaging used in developed countries. Those simple facts can cause complications for pharma companies and give rise to financial journalism of the kind light on nuance.

Production for low-price sales

Gilead can’t stop unauthorized production in Bangladesh and Egypt, because the World Trade Organization allows the non-enforcement of patents in some poor countries. Authorized production in India is sold at much lower prices than in the U.S., and is authorized for sale into agreed countries (poor ones). My guess is that unauthorized production in Bangladesh would leak into India if the price in India was much higher. In any case, India refused to grant a patent for Sovaldi (Gilead will appeal). India also has a big problem with counterfeit drugs.

See “$10 version of Sovaldi available in Bangladesh” by Douglas W. House, SA News Editor, Mar 9 2015 (seekingalpha.com), and Indian drug firm cleared to sell generic sofosbuvir” by Douglas W. House, SA News Editor, Mar 12 2015, (seekingalpha.com/news).

I don’t mean to imply anything about the Indian companies Gilead have agreements with, but it’s worth knowing about the problems in India. See India becomes a hub for fake medicines on safemedicinesindia (not dated), and if you can bear the page-junk, “Is your medicine a fake? Government report warns counterfeit drugs are flooding India” by Neetu Chandra, March 5, 2014 (dailymail.co.uk), which includes the claim that India has over 10,000 drugs manufacturers.

This looks exaggerated – “How India’s Patent Office Destroyed Gilead’s Global Game Plan” by Bruce Einhorn, January 15, 2015 (bloomberg.com). The most credible claim in the article is probably that India’s ruling will help challenges to Gilead’s Sofosbuvir patent in other countries. IMO courts and agencies in developed countries will not take a lead from India. No block-buster drug is likely to be sold at developed country prices in developing countries, by any company, and yet it seems to be big news when Gilead don’t achieve that. Bloomberg say Bruce Einhorn is an “Asia regional editor” and was an “Asia technology correspondent” (you can check the frequency and variety of his article titles here.)

I suggest finding the comment by Ptatty which mentions “TRIPS” under “Gilead Sciences Inc. Faces An Uncertain Future After Patent Loss In India” by Rami Alsamaraee, Mar. 30, 2015 (seekingalpha.com).

Overseas online pharmacies

The sellers that don’t abide by the law in the country they operate from are likely to supply sub-standard or dangerous pills. I suggest reading Wikipedia’s page or finding your own source for more detail.

Wikipedia describes verification programs. This is one of them – Verified Internet Pharmacy Practice Sites (VIPPS) program. When I put “h t t p : / / w w w.drugs******.com” into the box and clicked “Verify” I got “This pharmacy is not in the VIPPS database. Please report this occurrence to NABP by using our feedback form.”. (The spaces are so the URL won’t work as a link, and the asterisks are because I don’t want to publicize the drug seller.)

This story suggests there were allegations in China that Alibaba’s employees allowed fake drugs to be sold. “Chinese square off against Alibaba in palace intrigue as online drug sales loom” by EJ Lane, February 2, 2015 (fiercepharma.com). The story is vague, because it avoids saying directly that the fakes included drugs, or even that sales of fake drugs were alleged, yet “drug” occurs 21 times and the story is on fiercepharma.com. It isn’t hard to google-up things like “High Quality Sofosbuvir API/ Purity 99% Sofosbuvir ( sovaldi )” on alibaba.com. For ledipasvir, one result on Alibaba quoted “US $ 10,000 – 50,000 / Gram”, while a “Verified Supplier” quoted “US $ 1 – 100 / Kilogram” for what seems to be 99% pure ledipasvir in powder form, from any port in China. I don’t know how fussy the supplier is about who they sell to (I could guess). Some of Gilead’s HIV drugs seem to be on sale, or at least intermediates for them. I’d appreciate comments here because I’m not sure how to interpret the offers on Alibaba, but on the face of it, the active ingredients are made and sold at low cost in China.

Opinion on the low-cost trade

Everything from the problems patenting sofosbuvir in India to unauthorized cross-border trade is not special to Gilead. Instead I see an unresolvable tension between the need for pharma companies to have an incentive to develop drugs, and patients who desparately need a cure and can’t afford it, as well as other patients and other parties who want to keep costs down. The tension creates a suitable environment for generics companies, lawyers, online pharmacies of varying legitimacy, etc. A crucial factor is that the costs are both up-front and intellectual in nature, so there’s little up-front cost for copyists.

Gilead in France

See “France uses tax to put pressure on hepatitis C drug prices – (Daily Mail via NewsPoints Desk)” September 30th, 2014 (firstwordpharma.com).

It’s a concern because there doesn’t seem to be any downside for France, which other countries are likely to notice, and unlike the rebates in the U.S., there’s no volume or increased access for Gilead in return for the tax.

Tax in the U.S.A. and overseas

See “Gilead Avoids Billions in U.S. Taxes on Its $1,000-a-Pill Drug – (Bloomberg via NewsPoints Desk)” February 26th, 2015 (firstwordpharma.com).

Info about prescriptions

On the blog site “Gild – Gilead Science Shareholders” (gileadscienceshareholder.blogspot.co.uk), find “GILD Harvoni’s scripts were”, and you’ll see a list of weekly figures for US prescriptions. As at March 13, 2015, the figures show a steady rise for the year so far. “nrx” means new prescriptions, and I’d guess “trx” means total prescriptions (new plus repeat prescriptions).

The sources are “Symphony/BBG”, or Symphony Health Solutions and Bloomberg. I don’t know who their sources are, and I would not assume that every body gives them the data they would like to have. I expect that every prescription counted was real, but I would not assume that all prescriptions were counted. The conservative approach is to not assume that Gilead’s U.S. prescriptions are much more than reported (at least, not without a good reason), while not assuming that low numbers for a competitor are accurate. So long as the exclusions stay the same, a rise in the weekly figures is very like to reflect a real rise in the U.S. total. I hope that someone who knows more about it can simplify my qualifying statements.

See also the “BMD Asset Management and Research” instablog (on seekingalpha.com).


The employee reviews on Glassdoor can give a valuable insight into a company, although employees are no more unbiased than other stakeholders, and the “disgruntled employee” is common enough for the adjective to have become fairly standard. Unfortunately, I can only get the UK version, and I suggest checking the US version if you can. The score I see is very nearly three whole stars out of a possible five.

Generally, the drug devlopment side is praised. There’s much complaint about the work/life balance (which is not unusual, but it seems to be worse than average). There’s one low opinion of each of these – the commercial side, the IT, and the bureaucracy. It’s worth finding the review which contains “Little company syndrome”. The most severe criticism includes “turf fights”. The lean model is blamed for demanding too much from staff. It’s reported that staff changed tasks fairly frequently, but it was not said to be excessive or reduce productivity.

Employee reviews on indeed.com give an average of four stars (out of five), and were mostly very positive. Reviews on Indeed seemed more effusive and less critical of management, but I haven’t compared reviews from the two sites in any systematic way.

A bear’s view, and insider deals

This is bearish – “Forget Gilead Sciences (GILD), Buy These Biotech Stocks Instead” by Zacks Research Staff, March 06, 2015 (zacks.com). The EVP and a director have recently sold over $700,000 worth of shares, and earnings estimates have gone down.

A director bought nearly $1 million worth of shares on February 20, 2015, see “Gilead insider buys 10000 Shares“. As Peter Lynch said in “One up on Wall Street”, there are many reasons why an insider might sell shares in the company, but there’s only one reason to buy.

Michael Yee and RBC Capital

I noticed Michael Yee’s questions were pertinent and unusually relevant to the long term in the RBC Healthcare Conference. Michael Yee of RBC Capital is in tipranks‘ top 5% of analysts. He has rated Gilead more times than he’s rated any other company. I’ll just say the number is in double figures but under 20, and his performance is likely to have been assessed over a few years at least. The performance of each rating is based on returns over the following year.

Before 2014 results were out, RBC were reiterating “outperform”. Soon after the results – “Gilead (GILD) Price Target Lowered to $118 at RBC Capital” February 4, 2015 (streetinsider.com), although “outperform” was maintained. Michael Yee believes Gilead’s guidance is conservative, see “Bigger drug discounts put question mark over Gilead’s stellar run” by Reuters, Fortune Editors, February 4, 2015 (smarteranalyst.com).

On Tipranks I noticed a buy recommendation by Michael Yee for Uni-Pixel (UNXL) from two years ago. They are a tiny company that were developing touchscreens. There was much bitter argument about the company on Seeking Alpha, with serious accusations levelled at management (I think the management has been mostly replaced, but I stopped following the company). Two years ago, a short term profit could have been made, but UNXL were not a safe company to hold for one or two years. I also saw a ‘hold’ for Dendreon Corp. (DNDN) two years ago, and they went bankrupt. He may be weak at assessing some major risks, or maybe his outperformance was only possible because he learned from experience. Michael Yee’s performance is good relative to other analysts, but I saw no info about performance relative to sector or market indexes.

Less factual …

My piece – Gilead Sciences – highlights, opinions and unconfirmed points (kitchensinkinvestor.wordpress.com).

Last point about Gilead

Management have delivered.

Seeking Alpha transcripts quoted from

“Gilead Sciences’ (GILD) Management Presents at Cowen & Company 35th Annual Health Care Conference” Mar. 3, 2015 (www.seekingalpha.com)
Words: 19

“Gilead Sciences (GILD) CFO Robin Washington Presents at RBC Healthcare Conference (Transcript)” Feb. 27, 2015 (www.seekingalpha.com)
Words: 26

“Gilead Sciences’ (GILD) CEO John Martin on Q4 2014 Results – Earnings Call Transcript” Feb. 3, 2015 (www.seekingalpha.com)
Words: 120

“AbbVie’s (ABBV) CEO Richard Gonzalez on Q4 2014 Results – Earnings Call Transcript” Jan. 30, 2015 (www.seekingalpha.com)
Words: 63

Total words: 228

Thanks SA

With thanks to Seeking Alpha for their policy about quoting from transcripts, which can be found at the end of the transcript. I hope they don’t mind me quoting from four of them.

That’s all

Thank you for reading this.

DISCLAIMER: Your investment is your responsibility. It is your responsibility to check all material facts before making an investment decision. All investments involve different degrees of risk. You should be aware of your risk tolerance level and financial situations at all times. Furthermore, you should read all transaction confirmations, monthly, and year-end statements. Read any and all prospectuses carefully before making any investment decisions. You are free at all times to accept or reject all investment recommendations made by the author of this blog. All Advice on this blog is subject to market risk and may result in the entire loss of the reader’s investment. Please understand that any losses are attributed to market forces beyond the control or prediction of the author. As you know, a recommendation, which you are free to accept or reject, is not a guarantee for the successful performance of an investment.

IPG Photonics – the hard-to-gauge reliance on their Russian operations

Disclosure – I’m long IPG Photonics Corporation (IPGP).

About this piece

This is about possible disruption to IPG’s vertically integrated supply chain, if political action restricts supply from the Russian operations to the rest of the company. If you are sure that sanctions against Russia won’t get worse, or that escalated sanctions would not hit IPG’s Russian operations, there isn’t much point in reading any more.

Company profile

    “As a pioneer and technology leader in fiber lasers, we have built leading positions in our various end markets with a large and diverse customer base.” (10-K for 2014)

IPG claim “significant competitive advantages” (2013 Annual Report) from having vertically integrated operations, which are likely to be a big part of why the company has industry-leading margins.

Main sources

I’ll be quoting from the 10-K for 2014, which you can find on IPG’s SEC page, where there’s also a link for the Annual Reports. For SEC filings in .htm format you could try these EDGAR Search Results (if that doesn’t work, try the EDGAR Company Search and put IPGP in the ‘Fast Search’ box).

Find “Conference Call” and “Q3 2014 Results” (below) for the transcripts on Seeking Alpha I’ve quoted from.

The risk

The bold highlighting is mine. From the 10-K for 2014 –

    “Recent events in Ukraine have resulted in the United States and the European Union imposing and escalating sanctions on Russia and certain businesses, sectors and individuals in Russia. The United States and the European Union also suspended the granting of certain types of export licenses to Russia. Russia has imposed its own sanctions on certain individuals in the U.S. and may be considering other sanctions on the U.S. and the European Union or certain businesses or individuals from them. We have a large manufacturing facility and research and development operations in Russia which supplies components to our U.S. and German manufacturing facilities and finished lasers to our subsidiary in China. In addition, we supply components from our U.S. and German manufacturing facilities to our Russian facility. To date, we have not experienced any material disruptions or impact from current sanctions. Should there be disruption of our supplies from or to our Russian operations, or should the United States, the European Union or Russia implement different sanctions, our production and/or deliveries as well as results of operations would be affected.”


An important question is, if the Russian operations were severely restricted and the company faced a shortage of components as a result, just how badly would IPG have to be affected before the statement above becomes too mild to protect management from regulatory consequences? The phrase “would be affected” does not imply an upper limit to the damage, so I expect that management are well covered by the statement, though I have no expertise in the area and I’d appreciate a comment if you know better.

IPG’s production sites

It would be nice to have some idea of how dependent the non-Russian manufacturing facilities are on components from Russia. I found nothing that settles the matter, but there’s some information in the 10-K for 2014. If you find –

    “Our main facilities at December 31, 2014 include the following:”

there’s a list of 13 locations. Five locations include “Components” under “Primary Activity”, and the other eight do not. I’ve extracted info for the five locations with component production, and included the two locations with other manufacturing. The locations I’ve omitted only have administration and service listed.

    Owned or Leased (with expiration date if leased)
    Approximate Size (sq. ft.)
    Primary Activity

    Oxford, Massachusetts
    Diodes, components, complete device manufacturing, administration

    Marlborough, Massachusetts
    Manufacturing, administration

    Mountain View, California
    Leased February 2015
    Components, complete device manufacturing, administration

    Burbach, Germany
    Optical fiber, components, final assembly, complete device manufacturing, administration

    Fryazino, Russia
    Leased July 2016
    Components, complete device
    Manufacturing, administration
    (I explain further down why I believe the activities are not split as shown between the Leased and Owned floorspace.)

    Manchester, New Hampshire
    Leased December 2016
    Components, complete device manufacturing, administration

    Cerro Maggiore, Italy
    Complete device manufacturing, administration

    Total (Owned plus leased)

    Total Russia (Owned plus leased)

    Total Russia / Total (Owned plus leased)

    (Total sq.ft. occupied, including locations with no manufacturing: 1,647,900.)

There’s no reason to suppose that the proportion of the total floorspace in locations with manufacturing is a good guide to the dependence of the company on component production from a country. It’s a massive leap from the result for “Total Russia / Total” to say that about 40% of IPG’s component-production is in Russia, but I can’t see any better way to estimate it. The problem is not just that 39.33% is a lot, but that the true figure could be much more (although it could also be much less).

There may be some ambiguity about the “Primary Activity” in Russia. Either the activities are split between the leased and the owned floorspace as I’ve indicated, or all the activities listed for Russia are spread over both kinds of floorspace.

    Leased – “Components, complete device”
    Owned – “Manufacturing, administration”

The five other instances of “complete device” are all part of “complete device manufacturing” with a lower-case “m” in “manufacturing”. My guess is that “Components, complete device manufacturing, administration” applies to both classes of floorspace (for Russia), with no allocation of an activity to only one class of floorspace. I believe that’s supported by the long quote near the top, which includes “We have a large manufacturing facility and research and development operations in Russia which supplies components to our U.S. and German manufacturing facilities and finished lasers to our subsidiary in China.”. It does not necessarily matter if I’m wrong about that, as I explain in the next paragraph.

Most of the floorspace in Russia is owned, and here I’ll assume the activity is “Manufacturing, administration” (the activities on the same line as “owned”). If Manufacturing does not include components, then only 79,000 sq ft or 13.35% of the floorspace in Russia is for “Components, complete device”. That only works out to 5.25% of the “Total” (for all locations with some kind of manufacturing). However, I regard that 5.25% as even more tenuous than the 39.33% figure (as a guide to IPG’s dependence on Russian-made components), because there is no information for the overall percentage of floorspace used for making components (relative to the manufacturing floorspace, or in fact any class of floorspace).

I did not find much hard information about IPG’s reliance on components made in Russia. Without more information, points about the reliance are likely to be speculative and arguable. Unfortunately, that’s likely to be true for claims that the reliance is low, as well as for claims that the reliance is high, although I don’t want to pre-judge arguments which I have not seen or thought of.

Diodes all made in Oxford, Massachusetts

These are laser diodes, which are more like Light Emitting Diodes (LEDs) than the kind which only function as a ‘one-way-street for current’.

If supplies from Russia are disrupted, it’s good that diodes were not in Russia’s “Primary Activity” list. (There’s some risk in only making diodes in Oxford, Massachusetts, which suggests that economies of scale apply at IPG’s diode capacity, and they are big enough to outweigh the risk of concentrating production, or managements’ perception of the risk.)

More square feet

    “We plan to continue our expansion of our operations in Russia, Germany and the United States to meet the demand for our products and our sales and support needs. We believe that we will be able to obtain additional land or commercial space as needed. The additional expansion for Russia, Germany and the United States will provide an approximately additional 212,200 square feet, 114,500 square feet, and 191,200 square feet (excluding building and land purchases in 2014 for our California and Alabama locations), respectively once these additions are completed and occupied. With the amount occupied as of December 31, 2014, once all expansions are completed in 2015, we will have approximately 2.1 million square feet of occupied space to continue to execute on our planned strategies.” (10-K for 2014)

The “additional expansion”s listed sum to 517,900 sq ft, and Russia’s 212,200 additional expansion is 40.97% of the sum. Adding Russia’s additional expansion of 212,200 to my “Total Russia” figure of 591,700 (based on data “at December 31, 2014”), gives 803,900 sq ft, or 39.75% of the total floorspace in locations with some kind of manufacturing, after the additions referred to in the quote. That’s only slightly above the 39.33% I calculated using data “at December 31, 2014”. However, my calculations based on the “additional expansion”s are fairly rough, because locations are not broken out of countries, and the U.S. includes Novi, Michigan where IPG only list “Administration, service”.

Subsidiaries and the scope of sanctions

When sanctions against Iran were extended to cover overseas subsidiaries of U.S. corporations, a wind-down period was granted (see “U.S. Tightens Sanctions on Iran: Foreign Subsidiaries Wind-Down Period Ends March 8” (foley.com).). See also “Clarity for foreign subsidiaries of U.S. companies winding down Iran ops” (worldecr.com).

It looks like U.S. companies with Iranian subsidiaries would have been forced to divest, but I have not found an explicit statement of that. It would be a big blow if IPG is forced to sell its Russian subsidiary. It would also make many of my ‘if’s and ‘but’s irrelevant.

Finished lasers from Russia

    “We are subject to risks of doing business in Russia through our subsidiary, NTO IRE-Polus, which provides components and test equipment to us and sells finished fiber devices to customers in Russia and neighboring countries as well as finished pulsed lasers to China. Further, over 30% of our sales are to customers in China. The results of our operations, business prospects and facilities in these two countries are subject to the economic and political environment in Russia and China.” (10-K for 2014)

If IPG’s supplies of finished lasers from Russia dried up, I expect that to some extent the subsidiary in China could be supplied from other locations (although any shortage caused by not being able to export components from Russia to the U.S. and the E.U. would affect IPG’s global production of finished lasers). IPG are likely to have spare capacity after substantial investment (find “near finishing a round of investment” under “Nothing about Russian components in the Conference Call”, below).

The supply of finished lasers from Russia would be affected if IPG were unable to export their diodes from Massachusetts to Russia. So long as suitable diodes and other parts are available in Russia, components produced there could (possibly) be used in finished lasers for sale in China, the Russian home market, and many other countries even if the U.S. and the European Union tighten sanctions to include lasers (unless IPG would be punished for exporting from Russia, to China, for example). If sanctions caused a shortage of Russian-made components in the U.S. and the E.U., that could be offset by increasing Russian exports of finished lasers to countries where the trade is allowed (given the conditions: suitable diodes and other parts remain available in Russia, and IPG would not be punished in the U.S. for exports such as Russia to China).

The CEO has claimed that an auto product (presumably a laser or laser-system) has been made in Russia with Russian parts except for some metal parts which could be made in Russia if needed. From “IPG Photonics’ (IPGP) CEO Valentin Gapontsev on Q3 2014 Results – Earnings Call Transcript” Oct. 28, 2014 (seekingalpha.com) –

    “… our Russian company, really sales efficient Russian entity, which produced of auto product without any impact of the devices from outside of Russia. It’s all made in Russia from Russian components. Is only a few metal parts (indiscernible) we still import outside, but it is not — this is a part we can replace if need …”

The transcript also has management stating that Russian revenue was less than a twentieth of the total revenue (“Russian revenue” must be only for the Russian home market), although seasonal factors and expectations of strong growth in Russia are mentioned (find “It is less than 5%” in the transcript).

The ‘worst-case’ and the business model

I’ll assume here that sanctions stop the export of components from Russia, and the possible offset I just described is not enough to stop a shortage of IPG-made components. The only countries that I know have imposed sanctions against Russia are in North America and Europe, which is where IPG have their manufacturing. (In addition to the U.S. and the E.U., there’s Canada, where the sanctions seem to be mostly against named individuals and some entities, and Norway.)

If IPG don’t have enough components coming out of Russia, and that continues until stocks run out, they face a difficult choice – 1) Sell fewer lasers, or 2) Use the same sources as their competitors. The consequences of selling fewer lasers are likely to include lower revenue, customers not regarding the supply as assured, and lost customers who might be reluctant to return. About the downside of using the same sources as their competitors, from the 10-K for 2014 –

    “Our vertically integrated operations allow us to reduce manufacturing costs, control quality, rapidly develop and integrate advanced products and protect our proprietary technology.”

About the second item, “control quality”, would IPG say “Some of our lasers are not as good as they were.”? If they say “All our lasers are as good as ever.” it will be hard to claim a quality advantage from vertical integration when it’s restored. If they say nothing, there could be speculation about IPG’s sourcing of components which could force a statement, and customers who believed that IPG’s vertical integration resulted in higher quality would feel cheated. Obviously I can’t accuse IPG of a crime that’s in the future, I’m just showing how none of the choices available would be attractive. A reduction in quality would be bad, and if there’s no reduction in quality, the statement about vertical integration would look rather tenuous. A claim that standards were being maintained by thorough quality control could be met by competitors claiming that’s just what they’ve always done. On top of the hard choice, some quick redesigns may be needed if components from other suppliers can’t always be simply ‘slotted in’ to replace IPG components.

IPG say –

    “we design and manufacture most of our key components used in our finished products, from semiconductor diodes to optical fiber preforms, finished fiber lasers and amplifiers.” (10-K for 2014, the bold italic highlighting is mine)

That may leave some leeway for substitution if components can’t be shipped from Russia.

Competitors are unlikely to follow Grantland Rice‘s advice –

Instead, they’re likely to take maximum advantage of any weakness at IPG that results from sanctions. Any market areas where IPG have become weak, could be examined by competitors to see if they could dominate or address the areas profitably. IPG shareholders would be hoping that competitors settle for higher margins, rather than think strategically (see “What is strategy?” June 4, 2013 (build2think.wordpress.com).)

IPG reported Cash and cash equivalents of $522 million, compared to only $164 million of Total liabilities. It’s good to have a cash cushion if a crisis hits.

Is the ‘worst-case’ too crazy to happen?

You could argue that sanctions by either side which would affect IPG, would hurt both sides, without any measurable gain. However, in conflict a simple cost-benefit analysis is not necessarily relevant, any more than it applies to a game of ‘chicken’. The relevant branch of math is game theory, not the more direct methods of optimization.

For comment on geopolitics, you could try ‘Stratfor‘ (Strategic Forecasting, Inc.) who give free weekly Intelligence Reports. Their point of view tends to be based on the long term effect of geography, such as the lack of natural barriers which made Russia hard to defend.

Before World War I, many people thought that war in Europe was unlikely due to the interdependence of the European economies (possibly due in part to not understanding a book called “The Great Illusion“). It’s not the only war which demonstrated that mutually-advantageous trade links could be broken.

In less violent conflict, trade can be disrupted when it favors one side more than the other, and also when it’s believed that something needs to be done, and the alternatives are less attractive.


Maybe IPG could quickly replace lost Russian component production by stepping up production in the U.S. or Germany, but there isn’t much information to base that on. IPG’s soon-to-be-completed round of investment suggests that there should be spare capacity, but not necessarily of the right kind to replace Russian component production.

Inventories for the latest quarter fell year-on-year (from $172,700 to $171,009, in thousands), so there’s no evidence there of increased production of components in Russia for stockpiling outside the country. While the effect on reported inventories might not be great if vital components are produced at low cost, the simplest explanation is that IPG were not stockpiling components made in Russia.

About resilience more generally, IPG were hit by the financial crisis/recession in 2009, and bounced back very quickly (see the Morningstar link at the top, and find the nine-year history on the ‘Ratio’ tab). The problem then was probably the recession causing low demand across the industry, which is different to a single company having a problem with supply.

Costs in Russia

Employment costs are probably lower in Russia. You could try finding “What is the Russian Middle Class?” on forbes.com. Doctors earn less in Russia than in other European countries, see “Russian Doctors Protest as Reforms Threaten Jobs” Agence France-Presse, November 18, 2014 (ndtv.com). That isn’t directly relevant, but google keeps thinking I want to know about sanctions when I’m trying to research staff costs. If employment costs are much lower in Russia, that could explain some of IPG’s superior margins. In any case, so long as a low Ruble translates to lower costs in Russia, news which is bad for the Ruble is good for IPG’s costs. Potentially, increased tension could increase IPG’s margins via a lower Ruble, at the same time as increasing the risk in various areas (to the Russian operations, the company’s revenue, the share price etc.).

Nothing about Russian components in the Conference Call

See “IPG Photonics’ (IPGP) CEO Dr. Valentin Gapontsev on Q4 2014 Results – Earnings Call Transcript” Feb. 20, 2015 (seekingalpha.com).

These are the Russia-related topics covered –

In Q4 2014, revenue from the “other products” category “decreased 47% year-over-year to $9.6 million”, the result of “a decline in system sales in Russia due to economic conditions and foreign currency devaluation there as well as having had a comparatively large system order” (in Russia) “in Q4 of last year”. The “other” category is volatile, and small compared to the quarter’s total revenue of $207 million.

Sales in Europe grew to $69 million, “partially offset by weakness in Russia related to the economic environment there” (and the year-ago system order mentioned previously).

There’s a paragraph about the effect of the strong dollar and weak Ruble on the value of inventory, which was about $9 million lower as a result.

The quarter’s revenue from Russia is described as “good” given the headwinds.

The company is near finishing a round of investment in Russia, the U.S., and Germany, which should provide enough capacity to double revenue (find “finishing the round” in the CC).

I saw no questions from analysts about the effect on the rest of the company if components can’t be shipped from Russia. Maybe they are wrong to ignore the risk, or maybe they have better information and regard the risk statement about Russia as legal boilerplate.

Corruption in Russia

Yukos (wikipedia.org). Oil giant Yukos was seized by the Russian government, over 2003 to 2007, under Putin’s leadership.

U.S. Investor’s Lawyer Dies in Moscow Jail by Gregory L. White, updated Nov 18, 2009 (wsj.com). Hermitage Capital were accused of not paying taxes. Hermitage claim they paid, but the taxes were pocketed by fraudsters. The company’s lawyer was jailed and denied medical treatment.

See Russia, Corruption Timeline (mapreport.com) and Corruption in Russia (wikipedia.org). I’ve read that IPG were developing their own distribution network in Russia, due to the corruption there, but I can’t find any evidence. Many of IPG’s management team were educated in Russia and some of them may be able to advise on how to avoid trouble.

All IPG say about corruption in the 10-K is that anti-corruption laws are complex and often difficult to interpret and apply, with penalties if they get it wrong. They comment on the risk associated with Russia and China –

    “The results of our operations, business prospects and facilities in these two countries are subject to the economic and political environment in Russia and China. In recent years, both countries have undergone substantial political, economic and social change. As is typical of an emerging economy, neither China nor Russia possesses a well-developed business, financial, legal and regulatory infrastructure that would generally exist in a more mature free market economy. In addition, tax, currency and customs legislation is subject to varying interpretations and changes, which can occur frequently.”

Miscellaneous links

IPG’s Management and Board of Directors (investor.ipgphotonics.com).

This article briefly mentions the proportion of IPG’s employees based in Russia – “IPG Photonics – Growth And Technological Leadership Offer Appeal” by The Value Investor, Feb. 23, 2015 (seekingalpha.com). (Some comments by myself repeat points I’ve made in this piece.)

I wrote positively (mostly) about IPG in my blog post “Three cash3 companies“, January 17, 2014 (on wordpress.com).


I’ve focused on the adverse possibilities, and it’s worth bearing in mind that maybe sanctions will not be escalated, or an escalation will not have much effect on the Russian operations, or the effect will mostly be limited to Russia, or a shortage will be dealt with by quickly by ramping up production of components outside of Russia. If the risk is material, management ought to have considered it, and plans could already be in place.

Thanks, SA

With thanks to Seeking Alpha for their policy about quoting from transcripts, which can be found at the end of the transcript.

DISCLAIMER: Your investment is your responsibility. It is your responsibility to check all material facts before making an investment decision. All investments involve different degrees of risk. You should be aware of your risk tolerance level and financial situations at all times. Furthermore, you should read all transaction confirmations, monthly, and year-end statements. Read any and all prospectuses carefully before making any investment decisions. You are free at all times to accept or reject all investment recommendations made by the author of this blog. All Advice on this blog is subject to market risk and may result in the entire loss of the reader’s investment. Please understand that any losses are attributed to market forces beyond the control or prediction of the author. As you know, a recommendation, which you are free to accept or reject, is not a guarantee for the successful performance of an investment.

Hewlett-Packard – long term charts

Disclosure – I have no financial interest in Hewlett-Packard Company (HPQ). I’m long UBNT.

Additional disclosure – I won’t make any transaction relating to HPQ within 72 hours of publication.

My piece “Hewlett-Packard – three spreadsheet tables, and the quality of cash flow” shows the data the charts are based on. The formulas are in my ‘build the spreadsheet’ pieces, “Build the big HP spreadsheet and “Build the HP Operating Cash Flow growth spreadsheet“.

Revenue and earnings, in millions:
HP revenue and earnings

HP margins

Assets, liabilities and equity, in millions:
HP assets liabilities and equity

HP liquid assets

‘Per share’ charts

My reason for using ‘Shares calculated from EPS’ is that HP disclosed EPS adjusted for stock splits, but not an adjusted number of shares. I use ‘Shares calculated from EPS’ later to get figures for Operating cash flow per share (adjusted for stock splits). The number of shares is adjusted for the stock splits in 1995 and 1996, but not for 2000. I then used an adjustment factor for the 2000 split only, when I calculated split-adjusted values for dividend, cash flow per share, etc. (Find “build” for a link with the formulas.) It’s complicated but the smoothly rising adjusted dividend means I probably got it right.

HP number of shares

HP EPS and dividend

HP EPS divi and OCF per share

IMO the most pronounced feature of the period is the steep fall in OCF from 1998 to 2001. EPS lagged, falling from 2000 to a negative value in 2002. The 2012 drop in EPS is dramatic, reflecting an $8.8 billion writedown for the Autonomy acquisition, but OCF looks like it was not affected, with a dip that does not look bigger than average. The 2012 dip in OCF was during a rise, but it’s hard to pick a start for the rise, and the rate depends on the start.

HP trends for EPS and OCF

There are reasons to not trust the trendlines in the chart above –

1) The R-squared values are fairly low. They measure the amount of variation explained by the trendline, so an R-squared of 0.60 means the trendline explains 60% of the variation, and an R-squared of 0.53 means the trendline explains 53% of the variation.

2) If both trends continued, EPS would overtake OCF per share (OCFPS), and the ratio OCFPS / EPS would approach zero. That’s because the trendline for EPS has an exponential formula, which implies a fixed annual percentage growth (on average), whereas the OCFPS trendline has a linear formula, which implies the annual percentage growth rate will fall to zero. It’s extremely unlikely that OCFPS / EPS will approach zero, so at least one of the trendline formulas must be wrong. Visually, it looks like the gap between the trendlines narrows from 2010, increasingly rapidly, while the actual quantities have diverged.

3) It’s possible to select a shorter period which gives different trendlines and higher values of R-squared, such as 2001 to 2014:

HP trends for EPS and OCF 2001-2014

In each case I’ve used the trendline with the highest R-squared (out of Linear, Log, Exponential, and Power). Whether or not it’s mathematically correct, the power curve fitted to EPS since 2001 looks suspiciously like a straight line. It also looks too steep, and too high for the latest three years.

I would say for 1991 to 2014, the trend has been up, with such big and long deviations from any simple trend you could fit, that you can’t reasonably seperate the data into trend and noise. If you can’t pick out a trend in EPS or OCFPS in 24 years of data, there is no meaningful trend other than more up than down. The reversal that started in 1998 shows the danger of trying to ride a shorter term trend. I would be wary of any investment case that relies on trends in EPS or OCFPS.

One constant feature is that OCFPS has stayed higher than EPS. While in some cases earnings can smooth out variations in cash flow, in the long term it’s the cash flow that matters (given reasonable assumptions, such as OCF not including a disguised financing flow). It’s cash that can fund investment or be returned to shareholders, not earnings.

The following chart is based on my definition:

    Free cash flow = Operating cash flow – net PPE – net cash paid for acquisitions (all ‘per share’)

Rearranging that gives:

    Operating cash flow = net PPE + net cash paid for acquisitions + Free cash flow

In the chart, I’ve stacked the three quantities on the right, and for most years the top of the blue ‘Free cash flow’ bar corresponds to Operating cash flow. For some years the quantities don’t stack well in the chart. For most of those years, the height of the stack above zero is a good approximation of OCF. Two exceptions are 2002 and 2011 –

2002 – Operating cash flow was $2.17 per share, and Free cash flow was $1.63 per share.
2011 – Operating cash flow was $5.93 per share, and Free cash flow was negative (at -$0.61 per share).

HP components of OCF per share

Before showing the ‘walk charts’ for HP’s capex and cash-from-ops, I’ll show what an excellent chart looks like with this old chart for Ubiquiti Networks (I’m long UBNT).

Ubiquiti cash walks per share to 2013

Ubiquiti’s latest chart is in my previous blog post, and it shows a dip in 2014. Still, their cash-from-ops has grown fast without much need for cash to be invested (the investment flow includes capex). Not many companies can do that (and skepticism about it could explain the shorting). HP’s chart (below) shows loops. There isn’t much point in working out if they went round an old loop clockwise or anti-clockwise, what matters is the lack of clear progress during the loop. HP stayed mostly on the positive-free-cash-flow side of the chart, and finished in a better area than where they started.

HP walk chart OCF and capex

HP walk chart OCF and PPE per share

HP walk chart cumulative OCF and capex

HP’s Trade receivable sales

Increasing sales of receivables can imply a deteriorating quality of cash flow, because the sales cannot be increased indefinitely, so they potentially represent an unsustainable source of cash flow growth. I don’t think that’s likely to be a serious problem. There were claims relating to the 2012 results that sales of receivables were propping up the cash flow, with a strong rebuttal from HP. I write more about it in “Hewlett-Packard – three spreadsheet tables, and the quality of cash flow” (already linked to), where I give reasons why the increasing sales of receivables might not be a problem.

HP Trade receivable sales and more

HP’s growth of Operating Cash Flow per share

HP OCF growth CAGR - spread



I’ll repeat that for EPS and OCFPS, there’s no clear trend in the 24 year period, other than more up than down. Overall growth in OCF per share has been an unexciting 6.42% (1991 to 2014, adjusted for stock splits), but the rate depends very much on the period you look at. OCF per share has been more stable than EPS, maybe because downturns or acquisitions require writedowns which affect net income but not OCF, although variation in the optimism applied to accounting judgments is also a possibility. It’s natural for a long-lived company to have a succession of CEOs and variable earnings, and shareholders need to consider the possibility that a new CEO might at some point have ‘kitchen-sinked‘ to some extent (for example, see “Some Analysts Question Numbers in H.P.’s Write-Down” by Peter Eavis, November 21, 2012 (nytimes.com)). Free cash flow has also been less stable than OCF, which is often the case due to capex being more lumpy than other expenditure.

At the risk of stating the obvious, HP’s changing fortunes will be due to some combination of industry cycles, the changing market for its products and services, and the varying quality of its leadership. The dotcom boom and bust may have had some effect on HP’s business around 2000, through the misallocation of capital in tech companies.

The walk charts are consistent with the fairly common view that HP are good at generating cash flow, but have allocated capital badly, through poor or overpriced acquisitions. Poor performance has also been blamed on uncertainty or a lack of clarity about HP’s purpose and what it is or should be trying to achieve. That problem seems to be over, with the advanced technology HP are developing, and the planned split into a printing and personal systems company (printers and PCs), and the ‘Enterprise’ company.

HP’s long term performance is not fantastic or abysmal, but is quite patchy. Looking at the cash flows on a per share basis, results have improved under the current CEO, but I’m not 100% certain that the rapid rise in the sales of receivables has not been a factor in that. I like the technology HP are developing, although I’m currently not sure how far I should trust management. (I write about trusting management in general, here).


I’m less sure of my facts in this part.

I can’t pin down just how well HP are recovering and catching up in the areas they operate in. I give three relevant links under “Cash rush” in my “three spreadsheets” piece. It looks like heavy cuts on the service side from the EDS acquisition have caused low morale and an exodus of expertise. Such restructuring yields cash flow, for a while, which contributes to the flow needed to develop the Moonshot servers, hybrid cloud and R&D generally. I can’t say if HP is deliberately neglecting services to fund other areas, and that would never be admitted because of the effect on customers. It may be the result of the succession of CEOs with different visions of which markets HP should be in. One risk for HP is that the CEO will leave (to sort out IBM?), and IMO whatever you think of her, HP does not need another CEO changing the direction yet again.

IBM have been criticized for cutting costs in their services side and providing no more than is agreed to in the contract, making them undifferentiated from companies in emerging markets while being no cheaper. India-based Infosys has an excellent record of return on assets and equity, and have accumulated cash. The article “Innovation And Acquisitions Will Drive Growth For Infosys” by Economics Fanatic, Jan. 9, 2015 (seekingalpha.com) suggests to me that Infosys are looking to differentiate their service, while services from IBM and HP are becoming commodity-like through cost cutting. It’s not about Infosys having access to cheap labor – IBM employ staff in India, and have been criticized for low pay which results in the best staff leaving when they have enough experience. Maybe it’s too hard for IBM and HP to compete with the Indian companies, and it makes sense for them to move out of the way. There are other competitors –

    Enterprise Services. ES competes in the IT services, consulting and integration, infrastructure technology outsourcing, business process outsourcing and application service markets. Our primary competitors include IBM Global Services, Computer Sciences Corporation, systems integration firms such as Accenture plc. and offshore companies such as Fujitsu and India-based competitors Wipro Limited, Infosys Limited and Tata Consultancy Services Ltd. We also compete with other traditional hardware providers, such as Dell, which are increasingly offering services to support their products, new players in emerging areas like cloud such as Amazon, and smaller local players.” (10-K for the year to October 31, 2014)

“HP Enterprise Services” is the name given to EDS some time after the acquisition in 2008. EDS have been described as an “IT outsourcing giant”.

I’ve listed most of those named competitors by their stock ticker, with a link to the morningstar Ratios page where you can see the financial history including some return metrics. HPQ (poor ROA), IBM (reasonable ROA), CSC (poor ROA), ACN (good ROA), WIT (good ROA), INFY (good ROA). Obviously there’s more to the financial history than ROA, and more to a company than its financial history. IMO Accenture stands out, then India-based Infosys and Wipro look good. HPQ trails. Check the leverage, which indicates roughly how ROA is levered to increase the return on equity (ROE) which in turn tends to determine growth and the return of cash.

The Autonomy dispute

After acquiring Autonomy, HP accused the former management of fraud. See my short piece – “The Autonomy dispute after the UK’s fraud-buster gives up” February 01, 2014 (On this site). The U.S. authorities are still investigating. I look into the record of the UK’s Serious Fraud Office, then describe the point of view of Autonomy’s former management, with a little comment.


I mention the technology which HP will be investing heavily in, in one of my comments (as “fnoobler”) under an article titled “Server acceleration Altera FPGAs vs nVidia GPUs” on Seeking Alpha. The article is relevant to the future of servers. There are some technically informed comments under it, and I like to think that my less-informed speculations encouraged them. Also see Hewlett-Packard’s (HPQ) Management Presents at Citi 2014 Global Technology Conference (Transcript) Sep. 2, 2014 (seekingalpha.com).

Issues with implications for cash flow

From “Hewlett-Packard’s (HPQ) Presents at Barclays Global Technology Conference Transcript” Dec. 10, 2014 (seekingalpha.com):

    “restructuring and the focus on cost, has also then allowed us to plow money back into R&D.”

I assume that “restructuring” mostly means job cuts. HP expect 55,000 jobs to have gone by the end of 2015, making 5,000 to go after about 50,000 in recent years (in Meg Whitman’s tenure, I think).

Much of the transcript is about the planned split. The printer and PC side generates cash, and with the Enterprise side not having access to that cash flow after the split, there was a need to get Enterprise to generate more cash, which may have motivated the restructuring. ‘Enterprise’ gets the one-off benefit of shifting debt to the printer and PC side (apart from the debt in the financing operations), while keeping most of the cash.

Reawakening The Sleeping Giant At Hewlett-Packard” by Gary Hirst, Dec. 3, 2014 (seekingalpha.com). Much of the article seems to be based on the Barclays and Citi conferences. One other area covered is the contrast between HP’s program for developing paradigm-shifting server technology, and IBM’s program which is big but based on pushing existing tecnnology further.

History and awards etc.

Wikipedia list the company’s history. A chart showing the stock price since 2000 above the tenure of four CEOs, currently stops at 2013 and misses the rise under Meg Whitman, to $40.08 on Jan 26, 2015. HP have won many environmental and business awards, including “Most Respected Company in China” nine times in a row (there are various categories, quite a few winners, and IBM say they won it in 11 consecutive years). Making a Forbes list in 2010 where criteria included “use of corporate assets” may seem odd or a case of ‘award curse’, given the acquisition record, and some abrupt changes of plan when CEOs Hurd and Apotheker were ditched (in 2010 and 2011).

Errors and fudges

My extraction of long time series from 10-Ks was not straightforward and may not be reliable. The issues are described under an “Errors and fudges” section in the ‘three spreadsheet tables’ piece I linked to above.

Short term charts

See “Hewlett-Packard – Trying to find a pattern in the changes in working capital“. Skip the scatter charts if you like – the results were negative (as in theories not proved).

Thank you SA

With thanks to Seeking Alpha for their policy regarding quotes from transcripts, which can be found at the end of the transcript.

DISCLAIMER: Your investment is your responsibility. It is your responsibility to check all material facts before making an investment decision. All investments involve different degrees of risk. You should be aware of your risk tolerance level and financial situations at all times. Furthermore, you should read all transaction confirmations, monthly, and year-end statements. Read any and all prospectuses carefully before making any investment decisions. You are free at all times to accept or reject all investment recommendations made by the author of this blog. All Advice on this blog is subject to market risk and may result in the entire loss of the reader’s investment. Please understand that any losses are attributed to market forces beyond the control or prediction of the author. As you know, a recommendation, which you are free to accept or reject, is not a guarantee for the successful performance of an investment.

Ubiquiti Networks – charting volatile revenue

Disclosure – I’m long Ubiquiti Networks, Inc. (UBNT).

Closing price February 17, 2015: $29.44

Short profile

From the company’s 10-K – “Ubiquiti Networks develops high performance networking technology for service providers and enterprises.” The service providers are typically Wireless Internet Service Providers (WISPs). “Our technology platforms deliver highly-advanced and easily deployable solutions that appeal to a global customer base, particularly in under-networked markets.” They are able to price disruptively due to their business model, which I describe later.

About this piece

This is mostly about illustrating the volatility of the company’s revenue, a peripheral issue which is not as important as the long term prospects, or even the prospects for launches and upgrades when they can have a significant effect on results. However you aren’t likely to find the product types and geographies charted anywhere else.

While anyone who has studied the company will know about the volatility, IMO it’s still worth charting the quarterly revenue from the geographic and product segments. The charts are more important than the writing, and the trendlines and R-squared values in the charts are secondary. In fact the R-squared values are low, indicating volatility and showing the trendlines can’t be projected reliably. There may be some irony in using quarterly charts to show you shouldn’t read too much into the latest quarterly change.

Before the quarterly revenue charts I look at mostly annual data for cash flow, revenue and earnings, for a longer and more comprehensive view.

Cash flow

The first chart is based on annual data from the cash flow statements. It tunes out most of the quarterly noise, but it has its own noise. The chart shows a recent dip in the operating cash flow (OCF) per share, which is still very high compared to the capital invested per share.

Ubiquiti cash walks per share to 2014

The OCF dip is the result of big swings under “Changes in operating assets and liabilities:” –

($ thousands)
2014 ~ 2013 ~ 2012

Accounts receivable
(18,329) ~ 38,664 ~ (36,648)
(33,764) ~ (8,996) ~ (2,266)

Those are the contribution to cash from the change over the previous year, and (33,764) is negative meaning Inventories increased by 33,764 ($ thousand). Big increases in Inventories and Accounts receivable can be warning signs of problems, but there are good explanations for the rises in Ubiquiti’s case. One factor is that with the CEO having a majority holding, and with the company having good net cash and cash flows, there’s no pressure on the CEO to manipulate earnings (some forms of earnings manipulation can lead to increased receivables).

Inventory is built up before product launches. There was also a decision to increase inventory to avoid stock-outs, which I wrote about in “Ubiquiti Networks – long term sense meets a short-sighted market” May 13, 2014.

My conclusion about the ‘changes’ part of cash flow is that receivables and inventories will not increase relentlessly, and for various reasons cash flow was effectively pulled into 2013. The drop in OCF in 2014 does not negate the underlying trend. The trend in cash flows is important and a one year blip is not, so long as the blip has a good enough explanation.

Ubiquiti’s fiscal year ends on June 30 and there have been two quarters reported since 2014 (the last point on the chart).

(thousands, except per-share quantities)
Q1 2015 ~ Q2 2015

Shares (diluted)
89,913 ~ 89,737
$46,942 ~ $32,676 (calculated for Q2)
Cash invested
$3,461 ~ $4,930 (calculated for Q2)
OCF per share
$0.5221 ~ $0.3641
Cash invested per share
$0.0385 ~ $0.0549
Annualized OCF per share (x4)
$2.0883 ~ $1.4564
Annualized Cash invested per share (x4)
$0.1540 ~ $0.2198

Cash flow has been affected by greater Days Sales Outstanding, as distributors who have built up a good credit history have been allowed to take longer to pay and bigger distributors have been signed up (who expect more time before they have to pay).


Because I mentioned distributors taking longer to pay, I need to explain why Ubiquiti do not depend on distributors creating sales. About an internet forum refered to as the Ubiquiti Community –

    “Word of mouth referrals from the Ubiquiti Community generate high quality leads for our distributors at relatively little cost.” (10Q for Q2 2015)

That makes distributors keen to sign up, and with the pull from customers there’s no need to push the products through the channels, and no pressure to offer discounts to distributors.

Using the cash

With plenty of cash and low investment needs, it makes sense to use the cash to fund inventory and receivables, within reasonable limits, when it helps to increase sales. (See the earnings call transcripts on Seeking Alpha for more information.) One alternative – to use the cash for acquisitions – is limited by the low headcount which results from the lean business model. An acquisition candidate with half of Ubiquiti’s revenue is likely to employ far more staff than Ubiquiti, demanding more management attention and diluting the benefit from the business model.

Annual revenue 2010 to 2014

I’m about to investigate five years of annual revenue data using statistical methods. Given the short data set and the variation, I’m not surprised that the results were inconclusive. If you don’t mind negative results, or if you want to repeat the online test yourself, read on, otherwise skip to the three charts below.

In this short series of revenue data, 1 corresponds to 2010, 2 to 2011, etc. –

    ($ thousands)

    1 136,952
    2 197,874
    3 353,517
    4 320,823
    5 572,464

I fed the data into the Linear Regression Analysis page on The Chinese University of Hong Kong. You can repeat the exercise if you like, but you need to navigate via “Data modelling” and “Linear Regression”. Then put “5” into the “Number of rows” box, and copy and paste the data into the “Data …” box, and then click on the “Calculate” button. You should see the results –

    Model y = 99,397.3200x + 18,134.0800
    Pearsons product moment correlation Coefficient(r) = 0.9342
    Standard Error of Slope SE(b) = 21,922.3815
    t-test for the significance of the slope = 4.5341 Two-tailed probability = 0.0201
    95% confidence interval for slope 52,044.9763 to 146,749.6643

That means the low, mid, and high cases for revenue growth are approximately $52 million, $99 million, and $147 million (increase in revenue per year). The results imply only a 5% chance that the trend is for revenue to increase by less than $52 million or by more than $147 million, with a probability of 2.5% for each of the tails (the ‘out of range’ cases). The numbers can only reflect factors that have already had an effect on revenue, for example because Ubiquiti have not been affected by key personnel risk, the risk is not factored into the range of slopes (the slope of the graph is the increase in revenue per year). On the other hand, the period includes infringement of Ubiquiti’s IP, and Ubiquiti are now better protected against the risk.

The stats site reports using “Pearsons product moment correlation Coefficient” which is the standard method. It assumes a normal distribution which in finance is a standard but not necessarily correct assumption (it’s assumed in the Black Scholes equation for pricing options, and fractals-genius Benoit Mandelbrot didn’t approve). While the calculations take the low number of data points into account, for a sample size as low as five, the results are less reliable.

Even at the high-limit of $147 million, linear growth would eventually become a miserably low percentage growth, so I put the same data into the Exponential Regression page. I’m not confident about the results because they were a bit different when I replaced 1, 2, 3 … with 2010, 2011, 2012 …, and the rate of exponential growth should not depend on the period labelling (if you decide year zero was when Elvis Presley was born, it wouldn’t affect the growth rate of trees, corporations, or anything not Presley related). The relevant lines from China were –

    Linear Regression Model of the log10 transformed value is y = 0.1452x + 5.0134
    95% confidence interval for slope 0.0916 to 0.1989

I calculated the low, mid and high annual growth rates were 23.5%, 39.7% and 58.1%. I’m not confident about those, but the mid-range growth rate of 39.7% applied to the 2010 revenue of 136,952 would turn it into 521,620 by 2014, which is less than the actual figure of 572,464, but isn’t too far off and I wouldn’t expect a close match (all in $ thousands).

Even 23.5% annual revenue growth, at the bottom of the confidence interval, is a good growth rate, and given Ubiquiti’s margins it suggests good prospects.

The mid-case of $99 million linear growth only predicts annual growth of 17.3% in 2015 (from 99 / 572.464 = 0.173, not assuming any reversion-to-trend). But, the same $99 million growth would have been 72.3% growth over the 2010 figure of $136.952 million.

The Chinese University site does not show charts and only regresses linear and exponential functions. The problem is that the exponential function is concave-up, while other functions that could fit the data such as logarithmic are concave down.

In the chart on the right, below, the ‘concave down’ trendline implies growth that deteriorates faster than the linear trend, and the R-squared value is not far off the value for the ‘concave up’ exponential trend. If I’ve made statistics look indecisive, it’s better than being too sure about a trend. Five data points is not enough for very reliable results, meaning that even though a wider range of slope will be reported, the range is still less reliable. It would help if the points were all very close to the trend line, but that isn’t the case, and the three charts below confirm that it’s hard to fit a trend line with much confidence.

Ubiquiti - revenue - three trends

The fact that two very different trends fit about as well as each other, confirms my impression that while growth has been high, the data does not reliably indicate a likely rate of deceleration (growth companies usually slow down as they get bigger).

My long term expectation is that growth will hold up better than for most companies, due to the advantage of the business model, which can be applied to new products so long as ‘high-touch’ is not required and the forum can provide a useful function. The ability to stay lean and avoid bureaucratization is also a factor in delaying the onset of aging.

Revenue, earnings and margins

Ubiquiti revenue and earnings to Q2 2015

Ubiquiti margins to Q2 2015

Quarterly revenue, total and by geography

Quarterly revenue has been variable. The sharp dip in Q1 2013 was the result of widespread IP infringement. The situation was serious but effective action was taken and UBNT recovered.

Ubiquiti quarterly revenue total and by geography

The trendline has an R-squared value of 0.742, which means it only explains 74.2% of the variation. IMO few such trendlines can be extrapolated with confidence, and this one is no exception. If you believe the trendline, then fluctuations around it account for 25.8% of the variation. The results from the Linear Regression Analysis page at The Chinese University of Hong Kong (linked to above) were –

    Model y = 6,946.1300x + 58,165.7300
    Pearsons product moment correlation Coefficient(r) = 0.8616
    Standard Error of Slope SE(b) = 1,181.4866
    t-test for the significance of the slope = 5.8791 Two-tailed probability = 0.0001
    95% confidence interval for slope 4,394.1188 to 9,498.1409

My method was to test the correlation between the sequence 1, 2, 3, … and the quarterly revenue figures 79167, 87817, 91665 … (If you want to check the result, you’ll need the data, which you can get via a link at the end under “Tables and formulas” with instructions to ‘Paste special’ into a spreadsheet.)

The last line in the results implies that you can be 95% confident that the underlying trend in revenue is between an increase per quarter of $4.4 million and $9.5 million, with a ‘model’ mid point of $6.9 million. That’s slower trend-growth than the annual data gave, which could be because the quarterly data starts at Q1 2012, while 2012 was a good year with a big rise from 2011. The “t-test” line means you can be extremely confident that the growth in the period was not just the result of random fluctuations.

By not showing the total revenue, the next chart shows the revenue from the four regions more clearly.

Ubiquiti quarterly revenue by geography

Every region has experienced volatile revenue. While the two largest regions have demonstrated obvious and signigicant growth, that’s not quite as clear for APAC, and it’s quite arguable for South America (with only incremental growth between the first and last points). Any two regions you pick are capable of moving in opposite directions for a quarter or two (I quantify the low correlation using R-squared values in charts further on). Here I’ve added trendlines and shown their R-squared values:

Ubiquiti quarterly rev by geog with trends

The R-squareds show the percentage of variation explained by the trendlines – 64%, 65%, 21% and 54%. Taking the square root of the R-squared numbers gives R, the correlation coefficient (for a trend, the correlation is between the variable with the trend, and time, for example the end date of a quarter). The values are 0.800, 0.806, 0.458, and 0.7345. Putting the R values into a page on the University of Hong Kong site (linked to above) gave “p” values of 0.0006, 0.0005, 0.0996 and 0.0028 (navigate with “correlation” / “stat sig of r”). Taking the highest p-value of 0.0996 which is for South America, it means that there’s nearly a one-in-ten chance of no linear correlation (meaning revenue is not linearly correlated with time, and may just fluctuate randomly, possibly about a mean), whereas the p of 0.0005 for North America implies only one chance in 2000 that there’s no underlying growth. The p values do not give much idea of the variability, but I believe you can get some idea just by looking at the chart, and R-squared gives the proportion of variation which is ‘on trend’. (If you want the variability of the slope and don’t mind some work, find “You can get all the data” (above), follow the instructions, and find “I fed the data into the Linear Regression Analysis page” (above).).

All the R-squareds in the chart above are lower than the 74% for the total revenue. It’s expected that a total such as total revenue will have less variation than the parts it’s the sum of, provided the parts can fluctuate independently, because sometimes the parts will move in opposite directions, and they’ll tend to move at different rates. That’s why a big conglomerate will have more stable results than most other companies, although as well as the number of subsidiares, its results are smoother the more diversified its markets are, the more stable its markets are, and it helps if the subsidiares are about equal in size. Ubiquiti will become more diversified if it becomes more established in new areas. Machine-to-machine would be such an area, but there doesn’t seem to be much focus on it at the moment, and video is more promising in the near term.

Here’s a 3D view of the same data (no special glasses required).

Ubiquiti quarterly revenue by geography 3D

Quarterly revenue by product type

The product type charts are complicated by the change in disclosure which you can see by comparing either side of the gap. (New platforms, Antenna’s/other etc. can split between the later categories “Service provider technology” and “Enterprise technology”.)

Ubiquiti quarterly revenue by product type

The next chart is for “Service provider” and “Enterprise” only, with trendlines fitted, but six points is a small number to use in a statistical test, especially when the R-squared values are low. I saw no point in putting the data into any further statistical test. There are reasons to believe that ‘Service provider technology’ will grow, but there’s a big lack of evidence for it in the chart and the R-squared value of only 0.01.

Ubiquiti quarterly revenue by type with trends


The stacked view of the same data lets you compare the revenue of a product type to the total revenue.

Ubiquiti quarterly revenue by product type stacked

And in glorious 3D:

Ubiquiti quarterly revenue by product type 3D

While Service provider technology rose and fell, Enterprise technology fell then rose. It’s highly unlikely that there’s an underlying inverse relationship which would reliably predict movement in opposite directions in the future.

It’s not worth extrapolating a few quarters

While the overall growth of revenue is evident, the conclusion I draw from the quarterly ‘geography’ and ‘product type’ charts is that it’s hard-to-impossible to make a meaningful short-term prediction based on any of them. Analysts can put too much emphasis on the latest quarter’s results (maybe it’s in their job description), and some commentary can be knee-jerk over items like an increase in inventory in a quarter.

Currency and emerging markets headwind

While a dip in total revenue could easily be the result of the two product types happening to have a random dip at the same time, or the consequence of geographies happening to move down at the same time, there is currently a headwind. Much of Ubiquiti’s revenue is from emerging markets, especially where the telecoms infrastructure is not very developed. Those markets in particular could be affected by the strong dollar. Countries with an especially weak currency will be more affected, and countries that depend on oil exports or other depressed commodities are likely to have weak currencies and weak economies.

Ubiquiti are far from unique in being affected by dollar strength (Proctor & Gamble had a fair amount of publicity when their results suffered from it). I started warning on November 3, 2014 (before most of the Ruble’s fall) of the risk that a strong dollar and massive dollar-denominated debt could put some countires into crisis, with some similarity to the 1990s when Mexico, Asia and Russia were hit. That could still happen and is under-appreciated, but at least now there is general awareness that the strong dollar is a problem for exporters, and of weakness in emerging markets and primary producers, especially where the geopolitics has been in the news. Guidance seemed to build in a lot of downside for the effect of weak foreign currencies and weak emerging markets (my impression from the CC transcript), IMO leaving upside if Ubiquiti’s overseas markets are not hit as hard as feared, or if the company outpaces those factors. While the market is aware of the adverse factors (currency risk, low oil price, depressed commodities, weak emerging markets), the share price seems to indicate it is either not aware of the advantage of Ubiquiti’s business model, doesn’t believe it, or doesn’t care due to a short-term focus.

The business model

Ubiquiti concentrates on R&D, and outsources manufacturing. A vibrant forum gives feedback about products (especially new products), provides support, and evangelizes the product. By not needing a large staff for sales and support, the company can keep prices at levels that are disruptive for the competition, while still keeping a good net margin. Ubiquiti have a prototyping facility in China which CEO Pera described as a factory in an answer during the Q1 2015 CC, but he explained it was purely so they can introduce new products faster.

The business model is intact. So long as the model stays intact, the best time to buy is when a real or imagined problem sends the share price down. The net cash, cash from operations, low investment needs and disruptive pricing mean that Ubiquiti can ride out problems and emerge stronger, as they did after their IP was widely infringed.

Keeping it brief

I’m trying to stick to the subject and not say too much about related issues. There are good authors writing about Ubiquiti on Seeking Alpha and for a wider treatment I’ll refer you to “Ubiquiti Networks: A Rare Opportunity In An Overvalued Market” by Mingran Wang, Feb. 11, 2015 (seekingalpha.com). It’s reasonable to ask “How can the business model be intact when there’s no evidence of growth in the revenue of the largest product type?”. I suggest reading Mingran’s piece and then see what you think.

Abitrary categorization

The geographies could have been organized differently, e.g. Africa could have been included with South America instead of Europe and the Middle East, or there could have been three regions or five instead of four. A different categorization would have made the geography charts look different, but the volatility can’t be magicked away by regrouping, and the regions would always be more volatile individually than their sum.

The geographies move fairly independently

I could be criticized here for applying linear regression (with a trendline) to sets of points that don’t look linear. My excuse is that the points don’t suggest any better curve, and I’m quantifyng the lack of much relationship, rather than saying “This is the formula that relates the variables.”. The lines indicate the order of the points, which start at Q1 2012 in the bottom left and end at Q2 2015 on the right. The order of the points is not relevant to the trend lines or the R-squared values.

Ubiquiti quarterly revenue scatter chart EMEA

The R-squared values in the chart above correspond to 37%, 32% and 68%. The first two of those percentages are low, and they imply that revenue in the second and third biggest regions (North and South America) fluctuates fairly independently of the revenue from the biggest region (EMEA). Putting the data for EMEA and North America into the Hong Kong site’s Linear Regression page got the result

    “95% confidence interval for slope 0.0911 to 0.9055”

In other words, on average, for every $100 million per quarter that EMEA grows by, North America would grow by at least $9.11 million and at most $90.55 million, with a 5% chance of being outside that range. (That’s about the uncertainty of the trend, not the noise around it.) The wide range reflects the 63% independent variation (from 1 minus R-squared), although the number of data points is also a factor. Common sense might suggest you can make the range narrower, but that brings in more information than the 2 x 14 array of numbers which went into the test.

Ubiquiti quarterly revenue scatter chart N America

The main point from the chart above is that the R-squared of 0.29 (or 29%) implies that revenue in the third biggest region (South America) fluctuates fairly independently of the revenue in the second biggest region (North America). With the previous result, all the regions except for the smallest (APAC) move fairly independently of each other. Although APAC has some dependence (statistically) on EMEA and North America, it’s still fairly low, and evidently not big enough to make EMEA and North America move in step.

Any time two quantities grow, there’s a statistical correlation between them, even when it’s obviously meaningless. For example, a baby walrus could happen to be growing at the same time as the economy of Japan, and statistically there would be a positive value of R-squared, very likely a higher value than some of the R-squareds above. The long term growth in Ubiquiti’s regions is clearly related, because it depends on factors like new product launches, the quality of the products, the business model that allows a low price point, and evangelizing on the forum (because the understanding of languages crosses over the regions). Without underlying growth, it’s likely that the R-squared values in the scatter charts would be even lower. In any case they are low enough to say that a short term upswing in one region can easily be accompanied by a move in the other direction for whichever other region you care to pick (though less so for APAC, the smallest region by revenue).

Regions can also move in the same direction purely through chance. As a hypothetical example, if there’s a 40% chance of any region having lower revenue in a quarter, and if the regions moved completely independently, then there’s a 16% probability of the two biggest regions both falling purely through chance, and a 6.4% probability of the three biggest regions falling purely through chance. Adjusting for the level of dependence would increase those probabilities.

If there was a 60% chance of any region having higher revenue in a quarter, and if the regions moved completely independently, then there’s a 36% probability of the two biggest regions both rising purely through chance, and a 21.6% probability of the three biggest regions rising purely through chance.

The subject raises the fairly tricky question, can the fluctuations really occur through chance when most people would agree that they have rational causes? My attitude is that, to the extent that you can’t predict the movements, they might as well be random (on top of a long term growth trend).

Ubiquiti would be more volatile if they only had one of the four regions, and the relative independence actually helps stability, just not enough to make revenue stable.

Limited evidence suggests the product types move fairly independently

Ubiquiti Enterprise tech on Service provider tech - scatter

The short data series for the current two product categories shows an inverse relationship (the slope of the trendline down to the right means that the more revenue it reads on one axis, the less revenue it reads on the other axis). The R-squared of 0.35 shows that 35% of the variation in Enterprise technology revenue is accounted for by the variation in Service provider technology. Apart from the fact that 35% is quite low and there aren’t many data points to base it on, it might not be a feasible relationship. Two explanations don’t seem likely – 1) customers with limited budgets who spend more on one product type have less to spend on the other (that probably fails because different product lines address different markets). 2) When Ubiquiti develop one product they have less resources for the other. Because Ubiquiti are not short of cash, the second explanation would have to mean a shortage of R&D staff or of management time for managing them, but it seems unlikely that either product type has been short of development effort. If there’s no underlying inverse relationship, the product type revenues are likely to be more independent of each other than the R-squared of 0.35 suggests.

While I don’t believe the product types have an underlying inverse relationship, it’s still true that they have tended to move in opposite directions, but that has not been enough to make the overall revenue very stable or stop it from falling since Q4 2014.

Why Ubiquiti’s revenue is volatile

Ubiquiti supply high-tech capital equipment, with significant sales to emerging markets. Revenue from new models is hard to predict. Upgrades are a relatively more reliable source of revenue, but there’s no recurring revenue. Given those factors, it’s remarkable how steady the growth in OCF has been, and it would have been steadier if the swings in inventory and receivables had been smaller.

Stock price volatility

The share price is also volatile, and probably over-reacts to results and headwinds. That may be understandable for a few reasons. The CEO holds most of the stock and is hanging on to it. The staff hold stock, and some customers who were impressed by the technology available for the price may also hold stock. Shareholders with experience of the company may be less likely to trade.

While the articles and comments on Seeking Alpha might not represent the attitude of small investors, there is general agreement in the articles and comments that Ubiquiti are a good long term investment. The investors there may add when the price is low, and a few might take some profit when prices are high, but the attitude seems weighted towards holding long term rather than trading.

With many of the shares effectively not for sale, the share price can move up fairly sharply. Lately the movement has been down (since September, with a small uptick after the Q2 results). The size of the slide may be greater because small shareholders are unable or unwilling to add much at lower prices, for whatever reason.

Another factor is that the company uses its cash to buy back shares, but only when the price is fairly obviously low. I don’t follow the short selling much, but the short % of float seems to be around 30% quite often. There’s info from NASDAQ, and the Morningstar link at the top gives the short % of float.

Also, while a results surprise or disappointment is likely to be temporary, there isn’t a clear trend for results to revert to, which could discourage traders who aim to ‘buy low, sell high’ rather than follow momentum. There will also be investors who add when the price is low, with the general expectation that the long term prospects are good (which is usually my approach).

Seeing patterns in randomness

It’s often said that people see patterns where no pattern exists. There’s a recent case concerning random playlists, with users believing they are not random and there’s some kind of conspiracy. The solution has been to tweak the results of the standard and well-founded algorithms to conform to common misconceptions of what a random sequence should be like. (I heard that on BBC radio but I can’t find it on the internet. The subject area is covered in the book “Fooled by Randomness” by Nassim Taleb. Also see “Patternicity” and “Apophenia“.) It seems at least possible that when analysts try to abstract something meaningful from quarterly results, sometimes they find it when it doesn’t exist. I could fall into the same trap myself, and it’s hard to ignore quarterly results.


Ubiquiti’s parts move fairly independently, but they are volatile enough that the sum of them is still volatile, though less volatile than any individual geography or product type. While volatility in the results and the share price creates opportunities for trading, it would be a waste of a good business model to only trade and avoid long term investment, especially when volatility has contributed to a good share price for entry or adding.

Tables and formulas

See “Spreadsheets for Ubiquiti Networks – charts and volatile revenue“.

Data in semicolon-seperated format with instructions to ‘Paste special’ into a spreadsheet, formulas not included – “Data for Spreadsheets for Ubiquiti Networks – charts and volatile revenue

This is fairly advanced. For each trend line, it would be nice to have a line above and a line below representing “regression channels” (aka “confidence channels”), which you might have seen in technical analysis. IMO the trend lines here have too much uncertainty about the slope for that to be more useful than potentially misleading. If you want to know more, try How do I calculate confidence channels above and below a trend line. Find “richdiesal2009-03-15 at 20:13:39” for the comment I think is most relevant.

A simple guide – “Using Linear Regression to Predict an Outcome” by Deborah J. Rumsey, PhD from Statistics For Dummies, 2nd Edition (dummies.com).

DISCLAIMER: Your investment is your responsibility. It is your responsibility to check all material facts before making an investment decision. All investments involve different degrees of risk. You should be aware of your risk tolerance level and financial situations at all times. Furthermore, you should read all transaction confirmations, monthly, and year-end statements. Read any and all prospectuses carefully before making any investment decisions. You are free at all times to accept or reject all investment recommendations made by the author of this blog. All Advice on this blog is subject to market risk and may result in the entire loss of the reader’s investment. Please understand that any losses are attributed to market forces beyond the control or prediction of the author. As you know, a recommendation, which you are free to accept or reject, is not a guarantee for the successful performance of an investment.

Hewlett-Packard – three spreadsheet tables, and the quality of cash flow

Disclosure – I have no financial interest in Hewlett-Packard Company (HPQ).
Additional disclosure – I won’t make any transaction relating to HPQ within 72 hours of publication.

The spreadsheets cover much of HP’s financial history. I intended to show them with little comment, but I spotted HP’s rising sales of receivables, which needed writing about (and I didn’t stop there). I don’t have the expertise a qualified forensic accountant could bring, but the quality of cash and cashflows is a fairly neglected subject so I feel justified in writing about it. There will be long term shareholders and other people likely to know more about HP than I do.

Make your own copies

I’ve provided column data and instructions for building the spreadsheets, see “Build the big HP spreadsheet and “Build the HP Operating Cash Flow growth spreadsheet“.

Viewing the spreadsheet images

You might need to open the images in a new tab or window, and zoom in.

A big spreadsheet shown in 4 sections

This 60-column spreadsheet is the source for “Hewlett-Packard Co – long term charts” which I intend to publish soon. The columns where the background is slightly shaded are columns where I’ve used a formula. The other columns hold data taken from 10Ks on EDGAR (.htm, early 10Ks are .txt) or HP’s site (formats include PDF).

    All dollar amounts are in millions except for per-share figures. Numbers of shares are in millions

HP big spread D-W

HP big spread X-AN

HP big spread AO-BB

HP big spread BC-BK

Errors and fudges

The most likely source of error is the inconsistency you’d expect in reporting between 1991 and 2014, which makes it complicated to get a reliable time series for some quantities. For example, “Financing income” is not in the 10K for 2000, but in the 10K for 2001 it’s broken out for 2001, 2000 and 1999. Financing receivables are listed in the breakdown of current assets in the 10-K for 1998, so it’s likely that Financing income was earned in that year even though it was not broken out. I’ve assumed zero for Financing income in years when it was not disclosed, rather than guess.

Most columns that start with a zero or a run of zeros are actually cases of ‘data not found’. I left in ten years of zero Comprehensive income, as an obvious warning that initial runs of zero can’t be trusted, for anyone who has not read and remembered this section. I thought it would be more misleading if I left long runs of empty cells which were treated as zero in calculations, but I used dashes for empty cells in the last few columns, so I’ve had a consistency failure.

In some cases only a net amount was disclosed for some years, and for other years the net amount is calculated in the spreadsheet. The net amounts should be correct, but if one of the components is zero it’s likely that only the net amount was disclosed.

HP had 2-for-1 stock splits in 1995, 1996 and 2000. I’ve used the adjusted figures HP reported in 1996, but I had to adjust 1991 myself. On top of that, I applied a ‘stock-split factor’ to adjust for the 2000 stock split. The dividend was halved at the time of the 2000 split, and my adjusted dividend rises smoothly from 1991 to 2014. That indicates my adjustments are probably correct, but it’s sufficiently complicated to be a likely source of error.

I’m not surprised that some currency symbols are wrong, with “£” instead of “$”. Much of my editing consisted of changing pound signs to dollar signs, and some escaped or reverted.

About spreadsheet errors in general, there’s a European Spreadsheet Risks Interest Group who’s horror stories include the well publicized slip by economists Reinhart and Rogoff regarding the effect of debt on the economy, and Britain’s spies snooping on 134 phones by mistake (“MI5 makes 1,061 bugging errors” Identifier: FH1215). Most of the errors seem to involve human error, and it might be convenient to blame the spreadsheet.

I hope my own errors are spotted and reported.


I include the cash cost of acquisitions in ‘capex’ or ‘capital expenditure’ (column AK), because it’s not an operating expense or a short term investment. Investopedia‘s definition is more like investment in PPE or ‘Property, Plant and equipment’ (column AB). Most definitions seem to be like Investopedia’s rather than mine.

Trade receivables sold

One feature to emerge from the spreadsheet is the rapid growth in Trade receivables sold, shown in column AH. The usual point of selling Trade receivables is to get the cash quicker (although sales ‘without recourse’ also transfer the risk of non-payment, in return for a bigger discount). The increase contributes to Cash From Operations (CFO), and (after deducting capex from CFO) adds to free cash flow. The sale of receivables is not a sustainable way to grow CFO or free cash flow. Shareholders want to be able to exclude two motivations – a weak balance sheet and wanting to make the cash flows look better than they really are. HP’s balance sheet doesn’t look too bad. In millions, 2014 current liabilities of $43,735 are covered 1.15 times by current assets of $50,145 (or, current assets are 14.66% bigger than current liabilities), and the current assets amount to 65.91% of the $76,079 total liabilities I’ve calculated (in column U).

There are reasons why the increase in Trade receivables sold might not be as bad as it looks –

1) Each dollar increase in the sale of receivables does not automatically translate into $1.00 more cash flow, due to timing – a sold receivable may have been collected anyway, inside the reporting period.

2) It’s stated in the 10K that cash from operations is the primary source of liquidity.

3) The sales could be a response to local liquidity needs.

4) The rising sales of receivables could be a normal part of HP’s financing operations (I mostly rule that out, below).

5) Guidance for lower free cash flow in 2015 has not changed with the latest quarter.

6) HP are priced at a low multiple of their operating cash flow.

About 1), the effect on cash flow being less than the increase in the sale of receivables, ‘Utilized capacity’ might give some clue. From the 10K for 2014 (rearranged by myself):

    Utilized capacity represents the receivables sold to third parties, but not collected from the customer by the third parties. Transferred trade receivables included in the utilized capacity that HP has not collected from third parties are as follows:
    As of October 31
    2014 ~ 2013
    In millions
    Non-recourse arrangements
    $ 78 ~ $ 54
    Partial-recourse arrangements
    $ 381 ~ $ 118
    Total arrangements
    $ 459 ~ $ 172

From here on, when I write “Uncollected utilized capacity” it’s short for “Transferred trade receivables included in the utilized capacity that HP has not collected from third parties”. Using that term, the figures above are for Uncollected utilized capacity. (When a receivable is collected it’s turned into cash and any net liability due to not collecting it disappears.)

The figures are much smaller than the year’s $5.4 billion increase in the sales of receivables, which is hopeful. When a receivable is sold with recourse, it means that the buyer can get the cash from the seller (HP) if the receivable is not paid (by HP’s customer). A receivable sold with recourse might be a sale in legal terms, but in economic terms it’s a loan with collateral. Searching the 10K for “Partial-recourse” did not find an explanation of exactly what recourse is conceded and how sales with Partial-recourse are accounted for (as an operating cash flow or as a financing cash flow).

Adapted from the Consolidated Statements of Cash Flows in the 10K for 2014:

    Changes in operating assets and liabilities (net of acquisitions):
    Accounts receivable

[Changed February 4, 2015 – Less than 1% of the $2,017 million contribution to OCF from the change in receivables, can be explained by the fall in total net revenue from $112,298 in 2013 to $111,454 million in 2014. The figure presumably excludes the relatively small ‘uncollected utilized capacity’ which relates to the financing operations, because there’s $420 million disclosed for “Financing receivables” (the contribution to OCF from the change in them), and the size is about right for the uncollected utilized capacity.

Apart from the minor effect of the fall in total net revenue, factors that could make up the rest of ‘OCF from the change in receivables’ are 1) timing (because items billed early have more chance of collection in the year), 2) quicker collection from the customer, and 3) sales of receivables (i.e. of normal receivables, not as part of the financing operations). Such sales ought to be non-recourse, because it would be odd if HP did not disclose the liability associated with recourse, having made the disclosures about utilized capacity (although you might like to check the 10K for every instance of ‘recourse’, in case I missed something).

The first point “1) timing” seems unlikely to have had much effect, as there wasn’t much quarterly variation, and the pattern showed little variation from the year before:

Net revenue (millions)
Q1 ~ Q2 ~ Q3 ~ Q4
$28,154 ~ $27,309 ~ $27,585 ~ $28,406
$28,359 ~ $27,582 ~ $27,226 ~ $29,131

I have no information for saying how much ‘OCF from the change in receivables’ was due to sales of receivables (outside of the financing operation). I noted earlier that the $5.4 billion increase in the sales of receivables does not feed dollar-for-dollar into OCF, and that’s because it does not feed dollar-for-dollar into the change in receivables (due to timing).

Having considered the change in receivables, it’s worth looking at the change in payables, inventory and other amounts. I’ve made a spreadsheet and charts of those in “Hewlett-Packard – Trying to find a pattern in the changes in working capital“, but I couldn’t conclude much – only that the change in receivables and payables has recently been unsustainably positive (for OCF). I could not find good evidence that sales of receivables have been consistently used to compensate for changes in other areas.

About 2), the OCF providing liquidity, the “LIQUIDITY AND CAPITAL RESOURCES” section of the 10K has –

    “We use cash generated by operations as our primary source of liquidity. We believe that internally generated cash flows are generally sufficient to support our operating businesses, capital expenditures, restructuring activities, maturing debt, income tax payments and the payment of stockholder dividends, in addition to investments and share repurchases.”

suggesting that HP do not “generally” need to raise cash. When the cash and cash flow is not enough –

    “We are able to supplement this short-term liquidity, if necessary, with broad access to capital markets and credit facilities made available by various domestic and foreign financial institutions.”

Sales of receivables have risen over a few years and IMO can’t be called exceptional, at least not without an explanation. So if cash from operations is “generally” enough for the company’s needs, the sales were not necessary to provide liquidity. It follows that they were for some other purpose, such as a normal part of the financing operations. That seems logical, but with logic, confidence in the conclusion depends on the premises, and in this case I think it’s more likely that the sales have provided liquidity where it was needed, geographically.

From further into the statement –

    “Our cash position remains strong, and we expect that our cash balances, anticipated cash flow generated from operations and access to capital markets will be sufficient to cover our expected near-term cash outlays.”

Under the “Liquidity” sub-heading, the statement …

    “Net cash provided by operating activities increased by $1.0 billion for fiscal 2013 compared to fiscal 2012, due primarily to improvements in working capital management and a reduction in payments associated with webOS contract cancellations.”

… is ambiguous IMO, because “improvements in working capital management” could be HP-speak for increased sales of receivables, although they’ve also been paying attention to logistics.

Reduced DSOs (days sales outstanding) are nothing to brag about if it’s the result of selling receivables, which any company can do, though not necessarily on good terms. DSOs fell from 49 to 44 (as of October 31, 2013 and 2014). HP say –

    “For fiscal 2014, the decrease in DSO was due primarily to the impact of currency and the expansion of our factoring programs.”

factoring” means selling receivables. (I thought it was small cash-hungry businesses that used factoring, but it seems to be common for tech companies to sell receivables.)

According to Wikipedia, under U.S. GAAP, receivables are considered sold, only when the buyer has “no recourse”. The January 2013 piece from the ‘Grumpy Accountant’ (linked to below) implies that HP’s third-party arrangement got round that, and the author states that the deals amounted to borrowing.

About 3), “The sales could be a response to local liquidity needs.”

    “Our cash balances are held in numerous locations throughout the world, with substantially all of those amounts held outside of the U.S.” (10K for 2014)
    “Amounts held outside of the U.S. are generally utilized to support non-U.S. liquidity needs, although a portion of those amounts may from time to time be subject to short-term intercompany loans into the U.S.”

Bearing in mind that repatriating cash from overseas would incur tax, the head-office costs of the U.S., and the effect on U.S. cash of returning cash to shareholders, one possibility is that receivables have been sold to raise cash in the U.S.. Even if (and I stress ‘if’) the U.S. operations are chronically short of cash, it’s not as bad as if HP are globally short of cash (which they are not). HP’s situation is far from unique, with vast amounts of cash held overseas by US companies, and CFOs needing to find ways to finance the cash needs in the US. (You might like “The myth of Corporate America’s offshore cash” by Jeanne Sahadi, July 10, 2013 (money.cnn.com).)

About 4), the rising sales of receivables being a normal part of HP’s financing operations. It may seem unlikely, because buying other firms’ receivables is what you’d normally expect from a financing operation, but when third party arrangements are involved, expectations can be overturned. HP say …

    “We have third-party revolving short-term financing arrangements intended to facilitate the working capital requirements of certain customers. The total aggregate maximum capacity of the financing arrangements was $3.0 billion as of October 31, 2014, including an aggregate maximum capacity of $1.1 billion in non-recourse financing arrangements and an aggregate maximum capacity of $1.9 billion in partial-recourse facilities.” (10K for 2014)

(That’s the only item under “OFF-BALANCE SHEET ARRANGEMENTS”, after a paragraph saying they don’t set up shadowy things like special purpose entities which are often used to hide debt or obligations.)

HP Financial Services has not grown much since 2012, which makes it hard to attribute the growth in sales of receivables to the normal growth of HPFS.

    (Dollar amounts in millions)
    HP Financial Services, for the fiscal years ended October 31
    2014 ~ 2013 ~ 2012

    Net revenue
    $3,498 ~ $3,629 ~ $3,819

    Earnings from operations
    $389 ~ $399 ~ $388

    Total financing originations
    $6,425 ~ $5,603 ~ $6,590

[Changed February 4, 2015 – I’m still left wondering if increased sales of receivables could somehow be financing customers through the third party arrangements without pushing up those figures for HP Financial Services. However, it’s simpler to suppose that increased sales of normal ‘non-financing’ receivables has pushed up the change in non-financing receivables in the cash flow statement.

About 5), “Guidance for lower free cash flow in 2015 has not changed with the latest quarter”, the range is given in “Hewlett-Packard’s (HPQ) CEO Meg Whitman on Q4 2014 Results – Earnings Call Transcript” Nov. 25, 2014 (on seekingalpha.com), with a midpoint of $6.75 billion. That’s less than the $9.3 billion FCF for 2014. Shareholders will always want more free cash flow, but the cash conversion cycle has been unsustainably quick. From the transcript:

    “Our cash conversion cycle was four days, down four days sequentially, driven by improvements in days sales outstanding and days payable outstanding. … Normal sequential seasonality was more than offset by strength of collections and currency movements. … This cash conversion cycle is not at a sustainable level and is below our expectation of 10 to 12 days for fiscal 2015.”

Putting some facts together:

a) HP told analysts that the quick cash conversion cycle was not sustainable.
b) The nefarious reason for increasing sales of receivables boils down to getting a quicker cash conversion cycle and hoping that investors don’t notice the inherent unsustainability.

I think it follows that HP did not increase sales of receivables to fool analysts. Investors who only look at headline figures could be caught out, and the figures could help to get HP into selections thrown up by stock-screens, but valuations suggest that the market is aware of the unsustainable factors (and ‘headline figure’ investors would probably get a fair deal, but not the cash-machine bargain they expect). I believe that a non-nefarious motivation is most likely. Maybe it’s a cheap form of short term financing for the US operations, and HP expect increased US free cash flow (excluding sales of receivables) to kick in before that source of financing runs out. If so, I wish HP had explained that clearly (with good evidence, if possible) to avoid lines of investigation which are unlikely to give results with complete certainty, and where cases have to be considered which management might object to.

10 to 12 days is still a very short cash conversion cycle. A few businesses can get the cash up-front and have negative working capital, but many businesses would love to have the 10 to 12 day cycle. Even if sales of receivables figure in the cycle, if the discount is small enough, it seems worth it to get the 10 to 12 days and the liquidity that results. Rather like ‘Just In Time’ manufacturing, the cash-collection side of the cycle could be vulnerable to disruption. With less cash in the receivables pipeline, there’s less to come out if sales dry up. Of course HP won’t spend all the cash from selling receivables straight away, sales are not likely to suddenly dry up, and there’ll always be some receivables, but there could still be less of a cushion if sales fall.

The CFO has also said that 10 to 12 days is more sustainable, so I expect it to apply beyond 2015. The CFO has enthused about having inventory on boats, and I would not expect HP to achieve the 10 to 12 days without paying attention to logistics. HP are not single-minded about the cycle, and take advantage of discounts which suppliers tend to offer around their period ends (mostly for PCs).

About 6), “HP are priced at a low multiple of their operating cash flow”, Morningstar give a Price / Cash flow ratio under 6, with the share price between $38 and $39 on January 19, 2015. That’s very cheap, suggesting the market views the quality of cash flow as poor. That could be markets expecting some combination of:

a) The operating cash flow to fall.
b) High capex to leave less free cash flow, if the market also expects poor returns on the capex.
c) The cash to be trapped in countries where a depreciating currency is likely to outweigh the investment opportunities.

In summary

1) The rapidly rising sales of receivables might not be making such a large contribution to the operating cash flow.
2) The impact on OCF may be somewhere between $459 million (Uncollected utilized capacity Total arrangements) and $2,017 million (change in Accounts receivable).
3) Local liquidity needs may be a not-too-bad possible reason for the sales of receivables.
4) Guidance has been for lower free cash flow in 2015 (and was not changed in the latest quarter).
5) The market has priced in the cash flow issues (the point above, and the unsustainable speed of the cash conversion cycle).

Maybe I should say “there’s no mystery and nothing to worry about”, but I currently have enough doubt to avoid the stock. I hope that someone can settle the issue better than I can, preferably in favor of HP.

Motley Fool/nasdaq.com/whoever note that the ratio of accounts receivable to sales has fallen, which they claim is good, in “Motley Fool Guru Analysis for Hewlett-Packard Company” (nasdaq.com). Obviously, the ratio will fall if more receivables are ‘sold’ (either genuinely, or effectively used as collateral for loans). Unfortunately the analysis looks like it is updated frequently or automatically, and the gem is likely to vanish.

This … “More Financial Reporting Questions at H-P?” January 7, 2013 (grumpyoldaccountants.com) … is based on 2012 data, and my column AH shows rising sales of receivables since then. It was prompted by a piece in the WSJ which HP responded to strongly, claiming the sales of receivables had “negligible” impact on the 2012 cash flows. My opinion is that it’s natural for a cash-aware investor to be suspicious of the receivables sales, until there’s detailed enough disclosure about the deals for their impact on cash flows to be calculated.

HP’s response highlighted a big reduction in debt. Column U of my spreadsheet shows a decline in total liabilities which I calculated in column U, from $90,513 in 2011 to $76,079 in 2014 (in millions), although I have not shown debt or net debt. I’ve shown Retained earnings and Total stockholders’ equity, which reflect the difference between assets and liabilities, but in this case they are too affected by the Autonomy writedown in 2012 and related matters. You could argue that after those adjustments a recovery in Retained earnings and equity would be fairly automatic, and I think checking net debt is safer. From the 10Ks:

$ millions
2014 ~ 2013 ~ 2012 ~ 2011

Cash and cash equivalents
15,133 ~ 12,163 ~ 11,301 ~ 8,043

Notes payable and short-term borrowings (under Current liabilities)
3,486 ~ 5,979 ~ 6,647 ~ 8,083

Long-term debt
16,039 ~ 16,608 ~ 21,789 ~ 22,551

Total debt
19,525 ~ 22,587 ~ 28,436 ~ 30,634

Total debt – Cash and cash equivalents
4,392 ~ 10,424 ~ 17,135 ~ 22,591

I’ll call the last row ‘net debt’, although it’s miles away from HP’s version which is under $6 billion for 2012 (according to “Hewlett-Packard calls WSJ article ‘thoroughly misleading’” which I’ve already linked to).

The net debt reduction would not be so good if there were big off-balance sheet items, but there are some reassuring facts. First, the Uncollected utilized capacity Total arrangements (about customer finance, mentioned above) were a relatively small $459 (million), and they seem to be on the balance sheet because they are described under “Note 7: Balance Sheet Details (Continued)” in the 10K. The only item under “OFF-BALANCE SHEET ARRANGEMENTS” is the related maximum capacity ($1.1 billion non-recourse, and $1.9 billion partial-recourse). The maximum capacity is like a borrowing limit in that it’s only potentialy a liability to the extent that it’s reached. Also, the non-recourse capacity should not be a liability to HP in any case, and for most ‘with recourse’ sales the cash will be collected with no need for the third party to take advantage of the recourse. After disclosing the maximum capacity, this …

“For more information on our third-party revolving short-term financing arrangements, see Note 7 …”

… links it to the Utilized capacity.

If that seems complicated, what I’ve done is to close a loophole so you can be more confident that the net debt reduction is real. For many companies I would not be able to say the off-balance sheet items were not a problem. Obviously I can’t say if HP will raise the maximum capacity or have new off-balance sheet items in the future.

I only said ‘more confident’ because there’s also contractual and other obligations to consider. These have also come down since 2011. To find the data in the 10Ks, find “contractual and other obligations as of October 31, 2011” and “contractual and other obligations as of October 31, 2014”.

$ millions
2014 ~ 2011

Principal payments on long-term debt
18,539 ~ 25,953

27,691 ~ 37,020

Total – Principal payments on long-term debt
9,152 ~ 11,067

I subtracted the ‘Principal payments’ because they should be on the balance sheet and already in my ‘net debt’ calculations. The result comprises Interest payments on long-term debt, Operating lease obligations, Purchase obligations and (very little) Capital lease obligations. The result is a big amount that is not on the balance sheet, but at least it has come down. It’s only a rough calculation, because there’s a ‘1 Year or less’ component which should be included in the current liabilities (most of it is for the Principal payments which I already subtracted out).

The figures don’t include the non-U.S. pension plans, or:

    “… future cash payments of $1.8 billion in connection with our approved restructuring plans which includes $1.0 billion expected to be paid in fiscal 2015 with the remaining approximately $800 million cash payments to be made through fiscal 2021.”

Getting back to the January 2013 WSJ piece (about sales of receivables) which HP objected to, more detailed than HP’s riposte is “Debunking WSJ’s Hewlett-Packard Receivables Attack” by Avi Oster, Judicious Advisors LLC (Structured Finance and Securitization experts). It’s complicated. The author seems to acknowledge that proof is impossible due to limited disclosure, but claims that HP’s statement that costs were not material means that the discount was small, and somehow that’s enough for the author’s case against the WSJ article, although I didn’t really understand it. I would expect the discount to be small for sales of receivables with recourse (and smaller the better the credit rating of the seller), and also smaller for shorter periods until payment. Only the period has implications for the effect on cash flow. I’m going to fall back on dumb wariness. I like to think that if a banker had tried to convince me that CDOs were quite safe before the financial crisis, I’d have gone for dumb wariness rather than be bamboozled into agreement (without wanting to imply Avi Oster is like my hypothetical banker, and I’m grateful for being able to read and link to the piece). Being wary of arguments I don’t understand means that sometimes I’ll doubt an argument which is correct.

If HP and the ‘Judicious Advisor’ were right about the 2012 figures, there shouldn’t be too much of a problem even after the growth to 2014’s level of sales of receivables, on the principle that twice little is not likely to be big, just not as little.

Restricted cash

One description of restricted cash is cash that can’t be used until some condition is met, see “What is restricted cash and how to present it?” (simplestudies.com). The ‘grumpy accountant’ makes a case that cash held overseas should be reported as restricted cash – ‘“Trapped Cash” Begs for More Transparency‘ by Anthony Catanach, August 29, 2014 (grumpyoldaccountants.com). That’s too honest to happen, due to the effect on reported cash flows and financial ratios. If it was enforced, markets would dive and I’d wish I could invest in lobbyists.

From HP’s 10K for 2014:

    “Repatriation of some foreign earnings is restricted by local law.”

    “Where local restrictions prevent an efficient intercompany transfer of funds, our intent is that cash balances would remain outside of the U.S. and we would meet liquidity needs through ongoing cash flows, external borrowings, or both.”

There are restrictions on HP’s cash, but I found no instance of ‘restricted cash’ in the 10K. That’s probably not unusual for U.S./international companies, or tech companies in particular.

HP say –

    “We do not expect restrictions or potential taxes incurred on repatriation of amounts held outside of the U.S. to have a material effect on our overall liquidity, financial condition or results of operations.”

Capitalizing software development

I found no problem in this area, so it’s a bit dull but it’s only fair to report it. HP’s Intangible assets were $2,128 million in 2014, only about 2% of the $103,206 million Total assets. (Goodwill is much bigger at $31,139 million.)

    “HP capitalizes certain internal and external costs incurred to acquire or create internal use software, …” (10K for 2014).

By classing the expenditure as investment rather than expense, the operating cash flow is increased, but not the free cash flow because investment increases by the same amount. Capitalizing development is not as conservative as not capitalizing it, but it seems to be the norm for tech companies, especially for software development. Capitalizing development also increases earnings in the period, with amortization expense in later periods, and HP give the useful life of ‘internal use’ software as “generally from three to five years” (over which the asset is amortized down to zero).

In this … “Software Capitalization and Agile Development” by Sanjiv, May 4, 2012 (lithespeed.com) … it’s claimed that the capitalization of software development is more likely to be justifiable for the ‘internal use’ case (as in HP’s quote above), although the author’s expertise is in software rather than accountancy.

The 10K also has …

    “HP capitalizes costs incurred to acquire or develop software for resale …”

but …

    “The estimated useful life for capitalized software for resale is generally three years or less.”

Given the fast amortization and the relatively small size of intangibles, I don’t think that’s likely to be a problem in practice.

This lists “Capitalizable Costs” – “Overview: Capitalization of Software Development Costs” (stanford.edu) if you’re interested (but it’s quite dull!). For true fans, there’s a whole book about “Improper Capitalization of Costs” but software is only one area out of many.

The improper capitalization of expenses is not a common abuse, but there are cases – “How to Hide $3.8 Billion in Expenses” July 07, 2002 (bloomberg.com).


However right or wrong it is to sell receivables, capitalize development, and generally not lean towards conservative accounting, that all seems to be prevalent in the tech sector and I’ve no reason to believe that HP are on the wrong side of average.

Cathie Lesjak, the CFO, said

    “And I think we typically are considered conservative.”

from “Hewlett-Packard’s (HPQ) Presents at Barclays Global Technology Conference Transcript” Dec. 10, 2014 (seekingalpha.com). The quote was while answering a question about the cash conversion cycle.

It’s hard to inflate free cash flow over the long term, and I regard HP’s free cash flow and the excess of cash from operations over earnings since 1991 as evidence that the reporting of the earnings figures has been conservative.


If you don’t mind accounting tricks explained in an informal style, check “Financial Shenanigans Book Review“. Tech companies are over-represented in ‘Shenanigans’, though mostly for the “Dot-com bubble” (Wikipedia) era. ‘Shenanigans’ covers the sale of receivables and the improper capitalization of development costs.

Hewlett-Packard are mentioned only once, and that’s just to say that they now own EDS. Before being acquired, EDS had received early payment from a customer, which was not disclosed clearly, so investors could not tell that the cash flow was not sustainable. There’s more detail in the book, including the sentence from the 10-Q where EDS gave a small clue which I think most investors would have missed. HP can’t be blamed for that, so I’ll say they have no offences in ‘Shenanigans’, unlike many of the big tech companies (I’d say at least half).

An alleged stealth restatement at eBay

Meg Whitman was CEO of eBay from 1998 to 2008. The article – “Ebay’s Unannounced Restatement of Earnings” by J. EDWARD KETZ, May 2003 (accounting.smartpros.com) accuses eBay of restating earnings in a way that few investors would notice. The restatement was the result of an incorrect calculation about options. When I read it I thought, no investor would notice that, because they’d be looking at the phenomenal growth of income (although maybe that’s not the right attitude). For anyone who didn’t understand the table in Ketz’s piece, the error is the difference between the bottom two rows, for 2000 and 2001.

That’s all I found about the accuracy of the financial reporting during Whitman’s tenure at eBay (and eBay is only in the ‘Shenanigans’ book as a purely hypothetical example). I didn’t go looking for anything other than the accuracy of the financial reporting. If you want more info about the CEO you’ll have to find it yourself.

S&P credit rating

The rating that applied in October is given here – “Hewlett-Packard (HPQ) Placed on CreditWatch Negative by S&P” October 6, 2014 (streetinsider.com). The negative watch is due to the planned split into printer-and-PC and enterprise. The increased focus inevitably reduces diversification. S&P seemed to be playing it safe until they know more detail about the split.

[Added February 6, 2015 – “HPQ’s Growing Treasury Prompts an Investment Grade Rating” 01/27/15 (valens-credit.com)

[Added February 6, 2015
Cash rush

At least as important as conventional quality-of-cash-flow issues is the question, did HP sacrifice too much to get the cash flowing? An article suggests they did – “HP should use split to make up lost ground in IT services” by Karl Flinders Thursday 09 October 2014 (computerweekly.com). One of the points is that IT customers used to see IT as a cost, but now want it to transform their business. HP will miss out as a result of heavy cost-cutting, and customers will be wary of using HP when they’re heading for a split followed by more restructuring. There are other negative points in the article.

There’s more optimism in “Still in the garage” Jun 14th 2014 (economist.com). The author acknowledges that HP has a lot of catching up to do, and IMO may be too optimistic about their ability to do it.

HP get a good press on serverwatch.com. This is their reviews page. The trouble is, the reviews generally seem to avoid criticism. I can’t help wondering if the reviewers are careful not to offend potential future employers, although maybe that’s too cynical.

HP growth multiples for OCF per share

HP growth multiples - spread

HP growth CAGR for Operating cash flow per share

HP OCF growth CAGR - spread

Overall growth in OCF per share has been an unexciting 6.42% (1991 to 2014, adjusted for stock splits), but the rate depends very much on the period you look at. The split-adjusted dividend grew from $0.06 to $0.61 in the period, which is a little over 10.5% CAGR.

An article with a cash-flow focus

You might like to read “Hewlett-Packard: Our Analysis Says It’s A Buy And Meg Whitman Is Doing Well” by Mycroft, Jan. 5, 2015 (seekingalpha.com). I approve of the author’s focus on cash flow going back to 1990. There’s no mention of ‘receivable’ (so nothing about sales of receivables).

Too hard?

I’ve listed reasons why the sales of receivables might be reasonable or not too bad, but I currently have enough doubt to avoid the stock. Investors who follow Warren Buffett’s principles will need to ask themselves if HP belong in the ‘too hard’ pile.

Financial scores

If the links below don’t work, try this: http://www.gurufocus.com/stock/HPQ

Beneish M-Score for Earnings Manipulation (gurufocus.com)
Altman Z Score (gurufocus.com). The score improves as the market cap goes up, so it has fairly continuous variation.
Piotroski F-Score (gurufocus.com)

The Piotroski score is excellent. The only problem is with the Altman score which is a bit iffy.

The Beneish and Piotroski scores on the site both use the operating cash flow. If sales of receivables had a material effect on OCF in 2014, it would result in the Beneish score on gurufocus giving a slightly better impression of the company’s condition than is deserved. (In January 2013 HP denied the effect was material for 2012.) The Piotroski score gives one point for positive OCF, and one point for OCF greater than earnings, and for HP the score would not be sensitive to small changes in OCF. The Beneish score also has an index specifically for monitoring Days Sales in Receivables which will be sensitive to sales of receivables. The score will be lower (i.e. better) as a result. I’d say that gives a better-than-deserved score, but there’s room for opinion about it, for example the cash received is real and maybe you expect a low level of recourse to HP.

Altman’s Z includes Earnings Before Interest and Taxes but not the operating cash flow. The ‘IT’ in EBIT have to be paid in cash, while Earnings contains non-cash items, such as Stock-based compensation expense (which dilutes shareholders but does not add to bankruptcy risk). Companies go bankrupt when they run out of cash. I believe the Z score overstates the bankruptcy risk for HP based on the financial fundamentals (risks like competition, tech-leapfrog and currency movements are a different issue).

Good Beneish and Altman scores are also supposed to predict share price outperformance, beyond not being caught lying and not going bankrupt. The Piotroski was meant for picking the best investments out of companies with low price/book value ratios, and is also supposed to predict share price outperformance.

Thank you SA

With thanks to Seeking Alpha for their policy regarding quotes from transcripts, which can be found at the end of the transcript.

DISCLAIMER: Your investment is your responsibility. It is your responsibility to check all material facts before making an investment decision. All investments involve different degrees of risk. You should be aware of your risk tolerance level and financial situations at all times. Furthermore, you should read all transaction confirmations, monthly, and year-end statements. Read any and all prospectuses carefully before making any investment decisions. You are free at all times to accept or reject all investment recommendations made by the author of this blog. All Advice on this blog is subject to market risk and may result in the entire loss of the reader’s investment. Please understand that any losses are attributed to market forces beyond the control or prediction of the author. As you know, a recommendation, which you are free to accept or reject, is not a guarantee for the successful performance of an investment.

The M-Score and The Female Health Company

Disclosure – I own shares in The Female Health Company (FHCO), and also in EMN, GE, GIS, IPGP, ITW, LSCC, MKL, SNCR, SUNE, TIBX, UBNT (links are to Seeking Alpha).

FHCO Share price $7.76, Market cap $224 million, Price/Earnings per share (ttm) 18.2, Yield (ttm) 3.61%, Price/Cash from operations per share (ttm) 18.71. As at March 31, 2014.
(ttm figures based on ttm EPS $0.4257, ttm dividend $0.28, ttm Cash from operations per share $0.4148)

Seeking Alpha FHCO, Morningstar FHCO, Yahoo FHCO
Female Health SEC filings (femalehealth.investorroom.com)


This blog piece is about the Beneish M-Score for earnings manipulation, using The Female Health Company as an example. You might expect me to pick a company guilty of grossly overstating their earnings, but that’s not true of Female Health. Showing how a company can be innocent is more complicated than explaining why a ratio can point to guilt, and the explanations benefit from an example. Articles about forensic analysis always seem to be about nailing the bad guys, with little about the case for the defense.

Material that’s purely about the company is in “A quick intro to The Female Health Company”, “Female Health – ethics and incentive to manipulate earnings”, and from “The Female Health Company – 1Q 2014 results release” onwards, with charts at the end.

There’s a reason to believe that the M-Score is less effective now than it was when statistical tests were made. Readers who want to know about that first can skip to ‘What are accruals and do they matter?’ below.

The articles and papers I’ve found about the M-Score are either meant for investors but lack depth, or are highly technical and aimed at academics and regulators. I hope this piece provides depth, while avoiding phrases like “t-statistics based on the Huber/White sandwich estimator”.


I claim that the Beneish M-Score has two minus signs which are either wrongly reported or are not what was intended, indicating that a company is more honest if its leverage increases or its SG&A expense rises as a proportion of sales.

I claim there are serious problems with the ratios used in the M-Score, and improvements are possible.

I claim that The Female Health Company have booked nearly all the future benefit from their tax assets into Net income already, and a major contribution to their reported income will soon run out.

Many clever and/or well paid people must have checked the signs on the M-Score’s indexes, so I’m likely to be wrong. The M-Score has statistical support, and if I’m right, it ought to work even better when it’s fixed.

I’ve tried telling investors in Female Health about the tax situation. The few who have responded were not convinced. If I’m proved right, I’ve demonstrated the usefulness of looking into components of a company’s M-Score, because that’s how I was alerted to the unsustainable tax-related component of net income. Earnings that depend on tax benefits derived from previous operating losses are regarded as low quality. While tax assets are an easy target for manipulation, the evidence is against that, and in favor of Female Health’s tax assets having been estimated conservatively.

What’s an M-Score?

Here’s the formula:

M-Score = -4.84 + 0.92 * DSRI + 0.528 * GMI + 0.404 * AQI + 0.892 * SGI + 0.115 * DEPI – 0.172 * SGAI – 0.327 * LVGI + 4.679 * TATA

An M-Score above -2.22 is commonly supposed on various websites to indicate a serious risk that a company’s earnings have been manipulated. For anyone who can’t remember that lower is better, remember “Score high, more lie.”, and -2 is higher than -2.22.

Of the eight indexes in the M-Score, the simplest is SGI, the Sales Growth Index, which is this year’s sales divided by the previous year’s sales. Each index is multiplied by a number called a coefficient, for example 0.892 is the coefficient of the index SGI, and “+ 0.892 * SGI” is part of the M-Score formula. LVGI has the negative coefficient -0.327, and “- 0.327 * LVGI” is part of the M-Score formula. I explain later why I believe the two negative coefficients should be positive.

The constant of -4.84 at the front is an inconvenience for investors. Removing it would require adding 4.84 to any threshold the score is compared to, and would cause confusion about which version of the score was being used.

These M-Score indexes indicate distortions in the balance sheet which result from earnings manipulation:

    DSRI – Days Sales in Receivables Index
    AQI – Asset Quality Index
    TATA – Total Accruals to Total Assets

The indexes don’t prove or disprove balance sheet distortion, but provide some measure of how likely it is.

This index indicates earnings manipulation that will distort the value of Property, plant and equipment (net) in the balance sheet:

    DEPI – Depreciation Index

These indexes measure the pressure to manipulate:

    GMI – Gross Margin Index
    SGI – Sales Growth Index
    SGAI – Sales General and Administrative Expenses Index
    LVGI – Leverage Index

Even when the pressure to manipulate is high, in some cases management will stay honest.

The score was invented by Messod D. Beneish, with work published in 1997, but it’s the 1999 version which is most widely used, and used here. The coefficients in the M-Score were optimized to indicate the risk that regulators or auditors would require the accounts to be restated. That comes under the heading “earnings manipulation”, which is not usually as serious as fraud, as the restatement of accounts is not usually accompanied by prosecution. The term “earnings management” is sometimes used to mean arranging the accounts in a way which gives a truer picture than otherwise, or massaging the accounts when it isn’t serious enough to require a restatement.

The M-Score calculation needs numbers for the latest year and the previous year. The numbers will all be in a single 10-K form filed with the SEC. The score has statistical support, but I haven’t heard of the coefficients being retested or recalibrated to make sure they are up-to-date and still valid. Even when the TATA index was replaced by an index called ACCRUALS (which had been used before TATA), all the coefficients stayed the same.

Six of the indexes in the M-Score consist of one ratio divided by another, such as LVGI which is this year’s leverage divided by the previous year’s leverage, where leverage is the ratio of total debt to total assets. Ratios with that form can be misleading in the same way that a big percentage rise is often not as important as it sounds if the rise is from a small base.

Technically, a ratio should be written like this: 2 : 5, meaning two of something for every five of something else. In practice it’s easier to treat the ratios as decimals, and 2 : 5 translates to 2/5 or 0.4.

Calculating a list of financial ratios for a company leaves the problem of how to weight them to draw an overall conclusion. The M-Score and other financial scores solve that problem by using statistical procedures to weight the ratios. I do the opposite, by looking into each index in the M-Score, not just to explain the reasoning behind the index, but because the context matters. Here’s an analogy – a BMI (Body mass index, Wikipedia) of 27 generally indicates a person is overweight, with too much fat, but an athlete with a BMI of 27 is unlikely to be overweight or have too much fat. The interpretation is also different for children, and there are international variations. The additional information helps to put the number in the correct context.

I’ve written a lot about each index, and there’s a short list near the end under “M-Score definitions” for reference.

Why I show a spreadsheet that could have mistakes

My spreadsheet gives The Female Health Company a Beneish M-Score of -3.31. Gurufocus reported -2.5 on January 25, 2014, when they must have used the same source as I used (the 10-K for 2013). I expect from Gurufocus’s wide coverage of companies that they don’t copy and paste from SEC filings (a possible source of error), and they should have noticed by now if their calculation was wrong, so I could be showing a spreadsheet with serious mistakes. Different versions of the M-Score formula are possible, and Ycharts use a narrower definition of low quality assets. You have to subscribe to Ycharts before they let you see their M-Scores, and I haven’t subscribed.

    [ADDED APRIL 18, 2014 – Gurufocus have improved their M-Score calculator and now show the intermediate results. This means some of my writing about online M-Scores and Gurufocus’s in particular is now out of date. It’s a very useful feature, but there’s currently a glitch. As well as the sign-up windows that have to be clicked away, the first attempt to get an M-Score is likely to get an error message instead. If you click on “Beneish M-Score” in the alphabetical list on the left, you should then get the M-Score. Recently I’ve only used the calculator for SUNE (SunEdison) which needed the workaround, and SCTY for Solar City (I read in a comment they’re being investigated regarding their accounts). SCTY only gave a history of zeros, meaning the calculator didn’t work. I won’t know the success rate until I’ve tried more companies.

    The difference between the Gurufocus score and my own might be explained. The index they call TATA has a formula which looks more like the ACCRUALS index. I explain both indexes and why I use TATA, later. There were also two program errors warning of division by zero. While that causes some doubt about the result, the intermediate results show the working is alright. If the calculator was a ‘black box’ without the intermediate results, there would be no way of knowing if the end result was reliable.

    Find ‘M-Scores’ in my “SunEdison still set for growth and profit” piece, for a quick and rough analysis of SunEdison’s high M-Score in a comment, based on info from the gurufocus calculator.]

Gurufocus only give you the final score, and the M-Score is a measure with statistical support rather than a perfect indicator. If you like a company enough and find the M-Score indicating a serious risk of earnings manipulation, you might prefer to investigate rather than simply trust or not trust what the score is saying. It’s reasonable to target the investigation by seeing which component terms of the M-Score are mostly accounting for the signal. Until a free service turns up that shows the workings, some investors will need an M-Score spreadsheet. (For a free spreadsheet if you register (I haven’t) find “Beneish M Score to Detect Earnings Manipulation” below or under the “Links to free resources” tab.)

As an example, for Female Health Co the only two areas of possible concern about earnings manipulation are the gross margin and asset quality, highlighted by the GMI and AQI indexes. The signal from GMI is small. The AQI index measures the change in asset quality, and looking into it meant thinking about tax assets, and learning how the benefit to Net income had already been booked as tax assets were recognized. That kind of analysis is not possible if you only see the number spat out by an online M-Score calculator.

The intermediate results can be informative even when they aren’t signaling a problem. For example, Female Health’s sales and the SG&A expense have both fallen, which is obvious just by looking at the figures, but the change in the ratio between them isn’t obvious. The spreadsheet shows that as a proportion of sales, the SG&A expense has fallen from 0.2763 to 0.2452 (or from 27.6% of sales to 24.5% of sales). Further research shows that a fall in incentive payments was a large factor in this.

In other words, just as financial ratios can be looked at individually or composed into a score, the indexes in the M-Score can be looked into individually, as well as getting the overall picture from the M-Score.

The spreadsheet also gives readers something definite to refer to. Although it involves some work, you could look up the AQI calculation in the formula-view version, and trace the cell references I29, I30 and I33 back to Total assets, Total current assets and Property, plant and equipment, net, in the normal view of the spreadsheet. This provides an alternative if I don’t explain a calculation properly.

The file sizes of the graphics are only 362 KB and 162 KB, but the images are wide and you’ll need to view them separately. In Firefox, right click, select “Open link in new tab”, and use the magnifying glass. I don’t mind if you save the spreadsheet images.

Female Health M-Score 2013

Female Health M-Score 2013 formulas

A quick intro to The Female Health Company

Female Health’s only product is the female condom known as FC2, with the trademark “FC2 Female Condom” registered in the United States, and with registration or applications around the world for 14 trademarks including femidom, femy, and Reality. The company mostly sells to agencies in the public health sector, who supply the condoms to women who need them, mostly in less developed countries, and usually at no cost or low cost. Female Health’s customer concentration is high, the ordering process can be bureaucratic, and orders are lumpy. The three biggest customers are the United Nations Population Fund, USAID, and Sekunjalo, the Company’s distributor in the Republic of South Africa (RSA). The Bill and Melinda Gates Foundation supports programs with funding, some of which is likely to find it’s way to Female Health’s revenue.

Female Health have a history of growing cash from operations, with low investment costs, and have accumulated cash which exceeds their total liabilities, although they need some cash to see them through lean times. The growth has been ‘per share’ as well as in aggregate.

The M-Score spreadsheet is based on the 10-K for Female Health’s 2013, which ended on September 30, 2013. I ignore more recent results until I finish with the M-Score and only write about the company, from “The Female Health Company – 1Q 2014 results release” onwards. I’ll be referring to the spreadsheet and Female Health’s 10-K to illustrate some points, with more referencing when I get to the indexes in the M-Score.

Why the M-Score matters

1) Because when earnings manipulation results in the SEC or auditors requiring a restatement of earnings, the share price typically drops by about 40%.

2) Because an academic paper shows that the M-Score predicts future share price performance.

Both of the reasons could be undermined by recent investor awareness of the ‘accruals anomaly’, see ‘What are accruals and do they matter?’ below.

About 1), the 40% drop is under “The Model as a Classification Tool” in “The Detection of Earnings Manipulation” by Messod D. Beneish, June 1999. (PDF)

From the same PDF, it’s supposed to take about 20 to 40 similar non-manipulators in a portfolio to compensate for the discovery of earnings manipulation in a quarter, based on assuming that share prices will go up 1% to 2% on average per quarter. If the 40% drop has changed since 1999, it’s still likely to be big.

The PDF “The Predictable Cost of Earnings Manipulation” by M. D. Beneish and D. Craig Nichols, 2005 – 2007, starts by describing the ‘considerable losses’ caused by financial fraud.

The PDFs I link to are aimed at academics and regulators, but they’re worth referring to, and tend to have relatively plain English at the start and sometimes at the end.

About 2), the M-Score predicting future share price performance. I expect the prediction is with both real-world and statistical significance, to be worth publishing, which still leaves room for plenty of exceptions. The summary in this academic paper …

Earnings Manipulation and Expected Returns” by Messod D. Beneish, Charles M.C. Lee, and D. Craig Nichols, March/April 2013. Financial Analysts Journal

… suggests that the lower the M-Score, the better the share price performance, over a wide range of scores. As usual for statistical effects, it’s likely there will be plenty of exceptions.

In “The Predictable Cost of Earnings Manipulation” (above), it’s suggested that even sophisticated investors haven’t caught on to fraud detection, so mispricing occurred, or did in 2007.

Forensic accounting usually involves looking into the detailed accounts which companies don’t make public, but some of the leads are derived from information in the SEC filings (SEC – Securities and Exchange Commission), as in this case where the M-Score was used – “Using analytical tests to zero in on possible fraud” There’s enough to back up the claim, but you have to be a member to read the full story. (Some readers might have seen it already, from my “Links to free resources” tab.)

Some of the information about likely manipulators will be priced in, but I expect that for small companies the pricing-in is not a smooth process, with under-informed investors driving the price up until the overvaluation attracts short selling.

Limitations of the M-Score

The sample used to build Beneish’s model excluded financial services companies, and companies with sales under $100,000 or market capitalization under $50 million. That was in 1999 and there’ll have been some inflation since.

The M-Score is about earnings manipulation, and is less likely to pick up cash-based tricks. Sometimes cash takes a ’round-trip’, like A buying from B, B buying from C, and C buying from A. A round-trip of cash might have economic substance in a village with a butcher, a baker and a candlestick maker, but between companies it’s likely to be a ruse to create fake revenue. Any of the companies involved can class the purchase as investment, with the result that earnings, cash from operations, and cash into investment are all inflated. Detection depends on finding clues outside of the accounts, such as in the ‘Liquidity and Capital Resources’ section of a 10-K (which Female Health Co call “Liquidity and Sources of Capital”). If you spot that a company has bought from a customer, you may have spotted a short round trip of cash. Longer round trips are more complicated to arrange but are harder to detect. When a company has a history of substantial free cash flow, meaning there’s generally an excess of cash from operations over cash into investment, the excess can’t be faked by round trips and there’s less point in faking the rest of it.

The classic inflation of revenue by stuffing the sales and distribution channels can be detected through a build-up of Accounts receivable, and the M-Score covers this with DSRI, the Days Sales in Receivables Index. This tell-tale sign can be covered up by selling some of the receivables. If the receivables are sold ‘with recourse’ to a bank, it means the bank has a claim on the company if the customers don’t pay up, and the supposed sale of receivables is in fact a loan dressed as a sale. Finding that receivables are sold ‘with recourse’ is a clear indication that receivables are understated, but sales of receivables are not in the main accounts and not in the M-Score calculation. An alternative for manipulators is to sell the good part of the receivables without recourse, and receivables should usually be adjusted for any sales of them when calculating days sales outstanding.

Fraud evolves, and while big fraud gets the attention, it’s likely that lesser forms of manipulation also evolve. The practices used will reflect what is convenient for individual companies, what will influence share analysts and investors, what is less likely to cause suspicion among analysts, investors and regulators, and what is easier to defend in court and avoids the most serious personal punishment.

The phrase “the bottom line” became popular and along with revenue, earnings were an obvious target for manipulation. Savvy investors learned to distrust earnings and focused on cash. There are still investors who believe that cash can’t lie, although manipulators have long since focused much of their effort on cash. This is likely to make pure earnings manipulation and therefore the M-Score less relevant.

Another possibility is that when manipulation of earnings reaches a point where further manipulation becomes too obvious, companies switch to ‘pro forma’ (or non-GAAP) accounting in the earnings press release. That’s my understanding of this abstract: “The Relation between Earnings Management and Pro Forma Reporting” (five authors, two universities, May 1, 2013)

The paper also looks specifically into switching to non-GAAP measures where recurring items are excluded. It’s common for non-GAAP measures to exclude recurring items, and while EBITDA excludes four of them, anything starting ‘EB’ (Earnings before) is likely to exclude some recurring items. Tech companies in particular like to exclude high stock based compensation, although I can’t back that up with research. The paper implies that when companies use aggressive non-GAAP reporting, their GAAP earnings are accurate, with no further balance sheet distortion. The abstract is worth reading and I haven’t covered everything in it.

Researching a company would be easier if all the manipulation was diverted into non-GAAP measures, because it’s easy to check the ‘GAAP gap’ between GAAP and non-GAAP results. I believe that non-GAAP measures have their uses, but they are easy for management to abuse. A common method of stage magicians is to frame the area where they want the audience’s attention, which is away from where the real action is. Some believers in management’s preferred metrics and non-GAAP figures can be distracted from the GAAP figures, which are usually worse, not as smooth, and more risky to manipulate. Acacia Research are an example where anyone writing or commenting about them on Seeking Alpha would be told if they had strayed from management’s preferred metrics, and I saw no-one argue that you shouldn’t place so much trust in management. The issue died when those metrics deteriorated.

Scale-sensitive ratios

The ratio income / dividend is known as the ‘dividend cover’. The ratio is scale-insensitive, for example if a company’s Net income covers the dividend payment 3.4 times, the meaning is the same if the calculation is $3.4 million / $1 million, or $34 billion / $10 billion.

If staff turnover doubles, it sounds very bad, but if the doubling was from 0.1% a year to 0.2%, it’s not a problem at all. The doubling is a ratio of change and also a ratio of ratios, because the percentage turnover is the ratio of staff departures to staff. Six of the indexes in the M-Score have the same structure. For most of the six, this is not a big problem. For example, the ratio of SG&A expense to Sales is unlikely to be vanishingly small, although a small company is more likely to have a big increase in the ratio than a big company, quite possibly for a good reason such as employing more sales staff before launching a new product. GMI, the Gross Margin Index, can be quirky, but only when the gross margin is abnormally low or negative. The biggest problem is with leverage.

Beneish recognized the problem of division by small numbers, and used a standard fix called “winsorizing” which turns very small denominators into acceptably small denominators (so you might divide by 0.05 instead of by 0.001, which is like multiplying by 20 instead of by 1,000).

The winsorizing is at the 1% threshold. These are my simple but idiosyncratic instructions for how to winsorize a list of numbers (making the small numbers bigger), and limit outliers (making the big numbers smaller):

    1) Find the lowest 1% of numbers
    2) Call the biggest of those numbers WTD (for world’s tallest dwarf)
    3) Replace any numbers smaller than WTD with WTD
    4) Find the biggest 1% of numbers
    5) Call the smallest of those numbers WSG (for world’s shortest giant)
    6) Replace any numbers bigger than WSG with WSG

That might not be how Beneish would describe it. Winsorizing at 1% still leaves about 98% of the variation of scale, which matters, and is ignored by the M-Score calculation. I regard winsorizing as a rather brutal procedure which loses information.

The fixes were applied to the data Beneish used to build the model and get useful coefficients. That leaves two problems that could occur when calculating an M-Score –

1) Finding the cut-offs. You might have to buy the right academic paper, and understand it, or get access to a suitable financial database and know what to do with it.

2) If you know Beneish would have upped a 0.001 to 0.05, and you do the same, you don’t know how that affects the reliability of the result, without the right expertise or data.

If you need to know how gurufocus deal with the problem for their online M-Scores, you could try asking them.

Flaws in the M-Score – LVGI, a scale-sensitive ratio

LVGI, the Leverage Index

this year’s ratio Total debt / Total assets
divided by the previous year’s ratio

where Total debt = long term debt + current liabilities

Coefficient -0.327

To take an extreme example, imagine a software company with plenty of cash and no long term debt, the main expense is compensation, and the only current liability is a small Accounts payable. Near the end of a successful year, the company buys some new kit, increasing the Accounts payable. In such a situation, leverage defined as (Current liabilities + long term debt) / Total assets could have a huge percentage increase, as a result of change from a low base rather than high liabilities at the end. The definition of leverage is the ratio Beneish uses, and with that definition, LVGI can be shortened to –

    LVGI = This year’s leverage / the previous year’s leverage

To illustrate the “increase from a low base” case, suppose that leverage increases from 1% to 4%.

    LVGI = 0.04 / 0.01 = 4

That’s a big index. and after multiplying by the coefficient -0.327,

    the term for leverage = 4 * -0.327 = -1.308

When there’s no change in leverage, LVGI = 1, and after multiplying by the coefficient -0.327, the “neutral” term in the M-Score = 1 * -0.327 = -0.327.

The difference made when LVGI = 4 is therefore -1.308 – -0.327 = -0.981. That’s far enough below zero to have a substantial effect on the M-Score. (WordPress turns some minus signs into dashes, and 1 – -2 means 1 minus the number minus 2.)

If you can stretch your imagination to an increase in leverage from 0.001 to 0.004 (or 0.1% to 0.4%), LVGI gets the same result as before, because:

    LVGI = 0.004 / 0.001 = 4

with an effect on the M-Score which is completely disproportionate to the change in LVGI (in common sense terms. It’s exactly proportionate in terms of the arithmetic.)

At the other end of the scale, if leverage jumped massively from 20% to 80%:

    LVGI = 0.8 / 0.2 = 4

with exactly the same effect of adding -0.981 to the M-Score.

Find “Winsorizing” above, for Beneish’s fix and the problems with it.

Flaws in the M-Score – LVGI and the minus sign

LVGI, the Leverage Index

this year’s ratio Total debt / Total assets
divided by the previous year’s ratio

where Total debt = long term debt + current liabilities

Coefficient -0.327

Beneish defined leverage as (long term debt + current liabilities) / Total assets. By that definition, Female Health’s leverage fell from 5.83% to 2.57%, giving LVGI = 0.0257 / 0.0583 = 0.440.

Multiplying by the coefficient, the term added to the M-Score is:

    -0.327 * 0.440 = -0.144

Now suppose instead that Female Health Co’s leverage had shot up to ten times the previous level. That would make LVGI = 10, and the term added to the M-Score would be:

    -0.327 * 10 = -3.27

making the M-Score much smaller, and giving a much weaker indication of earnings manipulation.

The authors of “The Predictable Cost of Earnings Manipulation” (already linked to) change “LVGI” to “LEVI”, but it’s still the same formula. They state, as you’d expect, that LEVI is a measure of financial risk, and more of it increases the chance of earnings manipulation to meet the debt covenants set by lenders (but find “relatively low cost of breaking covenants” below). So, why does the coefficient have a negative sign, meaning that more LEVI and more financial risk indicates less chance of earnings manipulation, as I’ve shown above?

It looks to me as if the minus sign must be wrong, or is at least surprising enough to need an explanation. Either I’ve made a stupid mistake and have a complete blind-spot to it, or Beneish made a simple mistake which has not been picked up by peer review (I suppose they have that in the subject) or by at least several website publishers, or by the regulators, auditors, forensic accountants and short-sellers that use the M-Score. Or, academics spotted a mistake but it hasn’t filtered through to anywhere I’ve looked. Take your pick.

A different possibility is that while Beneish expected a high LVGI to correlate with earnings manipulation, the data said otherwise, so the software spat out a negative coefficient. Such a counter-intuitive result would need an explanation, and I haven’t seen any.

Using the positive coefficient 3.27 gives Female Health an ‘M-Score 3’ of -2.97.

No positive leverage for all practical purposes

LVGI, the Leverage Index

this year’s ratio Total debt / Total assets
divided by the previous year’s ratio

where Total debt = long term debt + current liabilities

Coefficient -0.327

The reasons given for including LVGI in the M-Score are financial risk, and not wanting to break debt covenants. (The “financial risk” reason is from “The Predictable Cost of Earnings Manipulation” (already linked to), where the term LEVI is used instead of LVGI.) This poses a problem, because a company with zero debt has no covenants, but could be short of cash to see it through shocks and lean periods. The company could reduce its liquidity risk by long term borrowing. But, should “financial risk” include liquidity risk, or did the term mean only the risk of breaking debt covenants?

Liquidity risk is a kind of financial risk, and borrowing to reduce it will increase leverage. One complication is that a company can increase its liquidity with a credit facility, which could have covenants attached, but won’t affect the leverage number until funds are drawn down from the facility.

A factor here is the relatively low cost of breaking covenants (1% or 2% of market cap, according to Beneish and Press 1993). Beneish states that the coefficients for LVGI (leverage), DEPI (depreciation) and SGAI (sales, general & admin) are not significant, and suggests they might only indicate “earnings management” which he considers to be less serious than earnings manipulation. For LVGI, that’s consistent with a relatively low cost of breaking covenants.

There’s a five index M-Score which excludes LVGI, DEPI and SGAI. That’s confirmed in this short description, where you can also download nine free spreadsheets including an M-Score spreadsheet which gets the numbers it needs from the internet, if you register (I haven’t registered).

Using Beneish M Score to Detect Earnings Manipulation” by Jae Jun, June 24th, 2010 (oldschoolvalue.com)

As well as some doubt about the reasoning behind using long term debt + current liabilities on the top of the leverage calculation, I’ve questioned the sign of the LVGI coefficient (and the SGAI coefficient, later). Also, the indexes where I’ve questioned the sign are two of the three indexes which Beneish says have insignificant coefficients, and may only indicate earnings management.

I’m going to give an opinion which is not based on Beneish’s writing. Leverage matters, but not necessarily as Beneish defined it. A company with insufficient cash and no debt might need to borrow in a hurry, and have an incentive to manipulate earnings when they need to borrow. If they borrowed before a crisis, they have more liquidity and some debt, but have the same net cash as before borrowing. The situation changes if they invest the cash they borrowed, which cancels the gain in liquidity and leaves the company with debt, and with financial risk unambiguously increased.

What I think matters are these ratios:

    All liabilities net of cash / Total assets
    Available cash / Total assets

The first ratio is instead of other definitions of leverage, and the second ratio is the liquidity of the assets.

The first ratio will be more like the usual definition of leverage when a company has high debt and little cash, which could result if a company borrows heavily to invest, and gets by with a credit facility instead of keeping much cash to cover the fluctuations that happen with normal operations. When a company has more cash than total liabilities, the ratio shows ‘negative leverage’. Some liabilities don’t ever have to be paid, such as Female Health’s Deferred grant income (explained below). Other liabilities are unlikely to be paid, such as deferred revenue (usually). In Nanometrics’ case, deferred revenue occurs when tools have been delivered and paid for, but none of the tools in an order have been tested and signed off by the customer, so revenue has not been officially recognized. In contrast, when deferred taxes are listed as a liability, the tax bill has been been deducted from taxable income but not paid yet, and it’s unlikely that the payment can be avoided. When there’s enough time to look into the liabilities, some could be excluded from the leverage calculation, some could be scaled back, but in my opinion some should remain, even if they aren’t current or classed as long term debt. I could have written “All liabilities scaled in proportion to how payable they are, net of cash” instead of “All liabilities net of cash”, but I preferred to use a number that can be found easily, with the option to refine the number. It’s more conservative to include liabilities that don’t have to be paid, than to exclude liabilities that do have to be paid.

‘Available cash’ means you can include the unused part of credit arrangements if you believe they won’t be taken away when they are needed, and you can include marketable securities if they’re included in current assets and you believe they can be sold at the value reported. That isn’t always a safe assumption. From Lattice Semiconductor (LSCC)’s 10-K for 2013:

    “Due to continued multiple failed auctions and the resultant illiquidity of these investments, we have classified our investment in auction rate securities as long-term.”

With hindsight, those securities should not have been classed as short term, showing that judgment can be involved in the accounting as well as the investing, and can be wrong.

Check a company’s securities for ‘investment grade’ (the alternative is ‘junk’), if they’re from the U.S. government or a very reliable government, and duration. Long durations are particularly risky when interest rates are low and could rise, or inflation could erode the value. Denomination in a foreign currency could be appropriate if the company has obligations in the same currency. That’s very brief. You could write a whole book about security analysis.

There are three levels of financial instruments and other assets. In my words:

    Level 1 – The instrument is reliably priced by the market. Little or no judgment is required to get the price.
    Level 2 – Either a market exists but it’s thin or infrequent, or the value of the instrument is gauged by the value of similar but not identical instruments.
    Level 3 – It’s anybody’s guess. Lattice had a long paragraph which included:

    “third party valuation results, investment broker provided market information and available information on
    the credit quality of the underlying collateral.”

and it went wrong. Currently:

    “Our Level 3 instruments are classified as Long-term marketable securities”

and that’s about the problem investments (auction rate securities) which were reclassified as long-term and they don’t sound very marketable to me.

Level 1 means the price ought to be reliable, but does not necessarily mean it’s a safe investment. Junk bonds are particularly liquid when they are generally overpriced.

A different variation is that non-recourse debt does not pose the same financial risk as ordinary debt. SunEdison (SUNE) use non-recourse debt to finance some of their solar projects. If a solar project does not generate enough income to pay the interest and repay its debt, the lender can take ownership of the project, but has no other claim on the company. It’s a bit like dividing a ship into watertight compartments – one or two big holes won’t sink the whole ship. The non-recourse debt is not included in leverage calculations included in SunEdison’s debt covenants, which means that lenders making ordinary loans regard the existing non-recourse debt as having much less risk than existing (senior) normal debt (that’s less risk to the ability to repay new debt). SunEdison’s non-recourse debt is also not amortized, meaning it stays on the books at full value until the last payment is made, which is usually about 20 years after the money was borrowed. Anyone or anything that simply takes the debt figures from SunEdison’s accounts and drops them into a leverage calculation will over-estimate the financial risk posed by the debt.

If there’s plenty of cash, you might as well just stick with cash. The ratio Available cash / Total assets says nothing about the particular liquidity needs of a company, which is a more complex matter.

Female Health Co have $8,922,430 of cash and only $3,766,996 Total liabilities. My spreadsheet uses Total liabilities instead of debt (for the main calculation), but otherwise sticks to Beneish’s definition of leverage and the LVGI index, because it doesn’t net-out the cash as I proposed. The result is 10.7% leverage (for 2013).

Going purely by Beneish’s definition, the leverage is a tiny 2.6%. The difference is because the company has long term liabilities but none of them are classed as long term debt, so by Beneish’s definition, the top of the ratio for 2013 only consists of the $904,049 current liabilities.

I don’t see Female Health as having financial leverage in any real economic sense. Even if they somehow had to pay all the liabilities immediately, and needed to repatriate the cash from overseas and pay 35% tax as a result, the total cash cost would be $5,795,378, leaving $3,127,052 cash and with zero liabilities on the balance sheet. Far from being juiced-up by debt, Female Health are ‘unjuiced’ by having more cash than Total liabilities. (For more about the tax rate, see “Links for the U.S. corporate tax rate”, below.)

Because I mentioned repatriation of cash, I’ll quote this from the 10-K,

    “Cash concentration: The Company’s cash is maintained primarily in three financial institutions, one located in Clayton, Missouri, one located in London, England and the other in Kuala Lumpur, Malaysia.”

but it doesn’t say how much of the cash is held overseas or how much tax would be payable if cash was repatriated to the U.S. It’s likely that the information isn’t disclosed because there’s little chance of needing to repatriate much of the cash.

Using my alternative to leverage, (Total liabilities – cash) / Total assets, the measure is negative for Female Health, at -$5,155,434 / $35,169,953 = -14.67%, which could be roughly interpreted as ‘negative leverage’.

The ratio Cash / Total assets is $8,922,430 / $35,169,953 = 25.4%. The ratio is high, but it’s simpler to say that the $8,922,430 provides a substantial cushion against lean times, than to draw conclusions from the liquidity of the assets, which is best seen next to my version of leverage (-14.67%). I haven’t included the revolving line of credit with Heartland Bank for up to $2 million, which was unused at the year end (and at the end of 1Q 2014).

I have not attempted a full round-up of Female Health’s obligations, commitments and contingencies (as in my “Three cash3 companies”), but the total for Contractual Obligations is $2,147,677, all for lease obligations, which stretch to 2018 apart from $102,222 beyond 2018. So:

    Cash = $8,922,430
    Total liabilities = $3,766,996
    Cash net of Total liabilities = $5,155,434
    Total Contractual Obligations = $2,147,677
    Cash net of Total liabilities and Total Contractual Obligations = $3,007,757

    “The Company has no off-balance sheet arrangements as defined in Item 303(a) (4) of Regulation S-K.”

I haven’t checked Item 303(a) (4) of Regulation S-K to see what kind of off-balance sheet arrangements they could in principle have without disclosing it.

The point of this is that Female Health have not racked up their cash through off-balance sheet arrangements or by incurring excessive contractual obligations, although I haven’t made a thorough check of commitments and contingencies. The company has risks such as customer concentration and emerging competition, but for practical purposes they have no financial risk, are not ‘juiced up’ by debt, contractual obligations or off-balance sheet arrangements, and their leverage is above zero in a formal sense but not in an economic sense. That could change if the cash flow drops and they don’t cut the dividend, but currently I see no financial risk.

When I think about a company’s cash I like to check their definition, and Female Health’s definition of cash and cash equivalents is clear. From the 10-K, under the sub-heading “Financial instruments:”

    “The Company currently does not have any assets or liabilities measured at fair value on a recurring or non-recurring basis. Substantially all of the Company’s cash and cash equivalents, as well as restricted cash, are held in demand deposits with three financial institutions. The Company has no financial instruments for which the carrying value is materially different than fair value.”

I’m happy with “Substantially all” which is much better than “include” and the general woolliness of the companies in my “Three cash3 companies” blog piece.

There’s no sign of any marketable securities or investments in Female Health’s 10-K, as you’d expect from “not have any assets or liabilities measured at fair value” in the quote above.

More on Female Health’s leverage

Forgetting my version of leverage which gives Female Health negative leverage, I prefer Total liabilities / Total assets to leverage defined as (Long term debt + current liabilities) / Total assets, because liabilities usually have to be paid or offset whether they are called debt or not (but find “Deferred grant income” below). Beneish considered three alternative definitions, but not the one I prefer. Liabilities which aren’t counted as debt aren’t subject to covenants as long term debt is, but current liabilities don’t always have covenants either.

Female Health have a revolving credit facility with Heartland Bank, limited to 70% to 80% of eligible accounts receivable plus 50% of eligible inventory. Covenants restrict corporate activity such as mergers and selling substantial portions of the assets, and restrict dividends and share buybacks, but “The Loan Agreement does not contain any financial covenants that require compliance with ratios or amounts.” Dividends are likely to be limited by income and cash flow rather than the terms of the Loan Agreement, see “FHCO – Female Health’s dividend cover and covenant restrictions” below.

Using Total liabilities instead of Total debt, my spreadsheet calculated the leverage for Female Health Co for 2013 and 2012 –

    2013 leverage = 0.1071
    2012 leverage = 0.2046

    LVGI = 2013 leverage / 2012 leverage
    = 0.1071 / 0.2046
    = 0.5235

My spreadsheet also includes Beneish’s definition, where these long term liabilities of Female Health Co are excluded:

    Deferred rent $66,799
    Deferred grant income $57,819
    Deferred income taxes $235,179

The exclusions add up to $359,797. It’s reasonable to exclude Deferred grant income, which means that someone else paid for at least $57,819 of capital expenditure. U.S. GAAP rules mean the benefit has to be recognized over the life of the capital assets bought. The mechanics of this are that the grant received is initially offset by the same amount of Deferred grant income, under Long-term liabilities. The grant is released into income by a process which cancels the effect of the depreciation charge. Therefore the Deferred grant income is not a bill which ever has to be paid in cash, and could reasonably be excluded from the top of the leverage calculation (if you prefer ‘liabilities that have to be paid’ to only ‘long term debt + current liabilities’).

Deferred rent arises when an operating lease agreement includes rising payments. The increases will have to be paid, but for more detail see “Straight-line Expense Recognition of Leases” by hhadmin, May 22 2012 (hhcpa.com/blogs). The accounting rules could change, see “Operating lease obligations – billions or trillions to land on balance sheets” below.

Deferred income taxes are 65.4% of the Long-term liabilities, and they have to be paid.

Companies sometimes report liabilities which are false, so they can conjure-up income when they like by reducing the liability (find “The way ‘cookie jars’ work” below). Cookie-jar liabilities are relevant here because they’re in the ‘never have to be paid’ category, but usually you would want to avoid a company with a cookie jar, although detecting them isn’t easy.

In the spreadsheet, the version of LVGI which sticks to the conventional ‘long term debt + current liabilities’ definition is called ‘Current LVGI’ because FHCO has no long term debt, and only current liabilities remain in the top of the leverage ratio. The Current LVGI = 0.4408, and using Current LVGI gives ‘M-Score 2’ = -3.28.

Either way, leverage has come down, by 48% (using Total liabilities / Total assets) or by 56% using Beneish’s definition. I see that as an excellent reduction in leverage, and the reduction gives no cause for concern about earnings manipulation. Taking the view as in “No positive leverage for all practical purposes” above, the situation still gives no cause for concern about earnings manipulation.

Female Health – ethics and incentive to manipulate earnings

Having made the first judgment about Female Health Co, that their leverage is nothing to worry about and does not indicate earnings manipulation, I’ll explain why I believe the company is probably less likely to manipulate earnings than most. They sell to organizations with ethical aims, including the reduction of infection and both maternal and infant mortality, partly by providing barrier-contraceptives. Any lack of ethics might not go down well with the customers. Funding sources include the Bill and Melinda Gates Foundation, who I think would rather not be associated with companies noted for dubious accounting, even though the quality of the suppliers’ financial reporting won’t be their highest priority. To a lesser extent that could be true of governments, but they’ve probably dealt with companies that have behaved improperly quite often.

In business-to-business or business-to-consumer relationships, I believe there’s less reason to care about the other party’s financial reporting. Some big retailers don’t pay their suppliers until after cash has been received from customers, so they have no financial incentive to care about the honesty of their suppliers’ financial reports. Some retail customers care about fair-trade or environmental issues, but they won’t blame a retailer if a supplier was guilty of inflating earnings, or care much, if they even know about it. The issue doesn’t matter if the business relationship isn’t close, and no-one cares about the reporting standards of the company they buy paperclips from.

Some manipulation depends on cooperation from the manipulator’s customers. The transformation of receivables to financial instruments is possible because the mechanics of payment don’t make a financial difference to business customers, and the customers are prepared to cooperate. The Penn Traffic Company are a retailer that recognized promotional allowances (like payments for advertising) before they were earned, in the early 2000s. Vendors were willing to send emails that backed up the false invoices, when they were reassured that they would not be billed twice.

The honesty of the funding sources and international agencies isn’t relevant, what matters is they could have more to lose and nothing to gain if they are associated with a company which manipulates earnings. I don’t want to overstate the case. Counterexamples are possible, and programs wouldn’t be canceled due to a company restating its accounts, but there’s still some incentive to avoid the possibility of embarrassing partners.

That’s all in my own opinion.

I’ve seen claims that Female Health Co are highly ethical, and my guess is they’re probably above average. From the Motley Fool: “The 25 Best Companies in America” (fool.com)

Fool’s list seems to have been published early in 2013, with the ranking based on how companies served their stakeholders, not just investors. Female Health Co scored highly in the ‘Investor’ category, with a low score for ‘Employee’. IMO using recruitment-related websites is a particularly unreliable method for judging a company in relation to its employees, when the company is small. Female Health made number 21, and were much smaller than any of the other companies in the Fool’s top 25.

There’s more on Fool’s page by clicking ‘Expand’ on the right, then a link to see more. Then there’s ‘click here’ to keep reading, but they ask for an email address.

If you think ethics extends to management not being over-rewarded if the company is acquired, find “FHCO – Anti-takeover provisions” below, and find “FHCO – Indemnification of directors and officers” (below) if you think they ought to at least leave some clues so shareholders don’t have to open many 8-K forms to find the wide indemnification.

Flaws in the M-Score – SGAI and the minus sign

SGAI, the Sales, General and Administrative Expenses Index

this year’s ratio SGA expense / Sales
divided by the previous year’s ratio

Coefficient -0.172

Other sources usually describe SGAI as the ratio of SGA expense to sales, for year t relative to year t-1.

The simplest way to illustrate the problem is with an unfeasibly extreme example. Imagine:

    2013 SGA expense $100 million
    2013 Sales $1,000 million
    2012 SGA expense $1 million
    2012 Sales $1,000 million

    so SGAI = (100 / 1,000) / (1 / 1,000)
    = 100 / 1
    = 100

The SGAI of 100 represents a massive increase in the SGA expense as a proportion of sales.

The term added to the M-Score equals the index times its coefficient:

    term added to M-Score = coefficient * SGAI
    = -0.172 * 100
    = -17.2

Because that’s massively negative, it pulls the M-Score down, indicating an honest company with virtually no risk of manipulating earnings. That’s the opposite of what Beneish intended, because he explained that increasing SG&A expense (as a proportion of sales) indicates more risk of earnings manipulation.

I’ve seen a claim that SG&A is usually a fairly constant proportion of sales, and since fake sales have little SG&A cost, less SG&A / sales indicates that sales have been faked. That’s an obscure belief and I can’t find it again. There is a case where a falling ratio of SG&A to sales was the result of misleading accounting, but it was for fiddled SG&A rather than sales.

In 1999, IBM sold their Global Network business, and somehow the proceeds ended up as a reduction in their SG&A cost. (Find “IBM sold their Global Network business in 1999” in my “Three cash3 companies” January 17, 2014.)

As a result, SG&A fell nearly 12%, while revenue was up 7% and Cost of goods and services was up over 9%. Investors who didn’t find the right footnote in the 10-K could have applauded IBM’s efficiency, at least if they didn’t stop to think that the figures were implausible for the bureaucratic behemoth. The M-Score doesn’t check the footnotes, and the calculation is (in $ billions):

    SGAI = (14.7 / 87.5) / (16.7 / 81.7)
    = 0.168 / 0.2044
    = 0.8219

In my opinion the result means “looks good, but investigate”. A low SGAI is more feasible when sales and the SG&A expense both grow but the sales growth is faster. That’s because the SG&A expense usually includes some fixed costs. That’s especially true for automated processes, for instance the cost of a paperless billing system doesn’t change much with volume. Publishing is an industry where revenue can increase without much change in SG&A. If a book sells a hundred times as many copies as expected, the SG&A expense doesn’t have to increase by much, although it might be worth spending more on promotion and investing to cash-in with merchandise. Similar economics apply to software and anything virtual or ‘in the cloud’, if little support from staff is required.

It’s also possible for some of the fixed SG&A costs to be cut if a reduction in sales looks like a permanent condition the company has to adapt to. Inefficient companies can streamline, but there should be evidence such as a reduction in headcount, a claim to have cut out layers of management, or the disposal of inefficient parts of the company. It’s also possible to cut too much and provide a poor service in order to hit a short term target. (See also “More with less”, below.)

When a company’s management want to hide a rising SG&A-to-sales ratio, they could be tempted to find some one-time proceeds to dump into SG&A, IBM fashion. Management can tune this by splitting the proceeds into different areas, like between SG&A and investment, and keep the SG&A-to-sales ratio constant if they want to. This can only be found if there are clues in the 10-K or 10-Q, which is reasonably likely as a lack of clues makes rather odd financial reporting more like fraud. The M-Score obviously can’t go looking for such clues, but if one index in the M-Score fails because management have covered their tracks, earnings manipulation might show up sufficiently in another index.

In “The Predictable Cost of Earnings Manipulation” (already linked to), Beneish and the other authors said that a high SGAI predisposes a company towards earnings manipulation. Administration costs that increased without creating sales signaled declining prospects, and they wrote about the GMI and the SGAI indexes, “both signals of declining prospects”. There is a strong implication that declining prospects predispose management to manipulate earnings. Gurufocus echo this interpretation, and relate a high SGAI to inefficiency in generating sales, which I believe is in keeping with what Beneish says. They also use basically the same formula for SGAI, and the same coefficient, as Beneish, and myself. Again, I’m wondering what simple mistake I’ve made that I can’t see. (Find “picked up by peer review” above for why a simple sign error is likely to have been spotted by now.)

SGAI for Female Health Co

SGAI, the Sales, General and Administrative Expenses Index

this year’s ratio SGA expense / Sales
divided by the previous year’s ratio

Coefficient -0.172

By ignoring the coefficient and whether it should be positive or negative, the situation regarding Female Health Co’s SG&A expense and Sales becomes clear enough, but first there’s a little complication about including the Advertising expense.

Female Health report Advertising separately from Selling, general and administrative. The SG&A figures I use are the sum of both expenses (Advertising + Selling, general and administrative). That’s done in a section of the spreadsheet headed “SPECIFIC TO FHCO”, which should make it easier to adapt the spreadsheet. If you made a spreadsheet for Ubiquity Networks, you’d use the section for calculations that belong under “SPECIFIC TO UBNT” rather than under “STANDARD”.

Including advertising, Female Health’s SG&A fell from $9,681,083 to $7,714,761 (in 2013).

The Sales figures are in the spreadsheet under “NUMBERS FROM THE ACCOUNTS” / “STANDARD”, which is where most of the numbers copied from the 10-K are. There you can see that Sales fell from $35.0 million to $31.5 million (in 2013).

If you find the “M-SCORE CALCULATION” area (to the right of the “NUMBERS FROM THE ACCOUNTS” area), and find the “SGAI” row in the list under the heading, you’ll see these two numbers:

    0.2452 (under “top”) and 0.2763 (under “bottom”)

Those cell values show how SGAI was calculated:

    SGAI = 0.2452 / 0.2763

The proper terms would be “numerator” for “top”, and “denominator” for “bottom”.

The text on the right of the row explains the calculation:

    “SGAI – Sales General and Administrative Expenses Index (-0.172) = this year’s SGA expense / Sales, divided by the previous year’s ratio”

That text explains how the numbers were derived:

0.2452 (under “top”) is from

    this year’s SGA expense / Sales

0.2763 (under “bottom”) is from

    the previous year’s SGA expense / Sales

That can be expressed as:

    The ratio SG&A / Sales fell from 27.6% in 2012 to 24.5% in 2013.

Summing up the change in SG&A, Sales, and the ratio between them:

    SG&A fell from $9,681,083 to $7,714,761
    Sales fell from $35.0 million to $31.5 million
    SG&A / Sales fell from 27.6% in 2012 to 24.5% in 2013
    SGAI = 0.2452 / 0.2763 = 0.8875

Very often, the ratio SG&A / Sales rises when Sales falls, as some of the SG&A expense is relatively fixed and can’t be scaled back as quickly as sales fall. Because a rise is expected, the fall needs an explanation. It’s in the 10-K for 2013:

    “The decrease was primarily due to a reduction in incentive payments partially offset by increased spending in education and training and consulting expenses.”

The reduction in incentive payments could be seen as partly reverting to 2011 levels:

    “Selling, general and administrative expenses increased $3,091,144 to $9,628,134 in fiscal 2012 from $6,536,990 in fiscal 2011. The increase was primarily due to increased spending in education and training and incentive payments based on the achievement of performance goals relating to the Company’s unit sales and operating income.”

In case there’s any doubt that the payments were cash, it’s confirmed by:

    “The increase of $1.4 million in cash generated from operating activities … was primarily due to the decrease in … accounts receivable … and the reduction in incentive payments.” (10-K for 2013, about 2013)

but the incentive payments aren’t broken out of the Selling, general and administrative expenses (or I didn’t find them).

The “share-based compensation expense” didn’t vary much, with the expense recognized for 2013, 2012 and 2011 being a fairly steady $728,000, $782,000 and $706,000, and the fair value of stock awards granted was $471,000, $218,000 and $1,677,000.

Overall, the Selling, general and administrative expense has been volatile over the past three years, with the latest year’s expense in between the two previous years. The SG&A expense has fallen as a proportion of sales (which is good), and the fall is probably accounted for by lower incentive payments. The size of the incentive payments wasn’t found, but something must have lowered SG&A, and I saw no sign of any one-off benefits that could have somehow been used to reduce the SG&A cost, IBM-style. There’s no sign here of earnings manipulation.

The value of SGAI in the spreadsheet is 0.8875, under “ratio” (I could have used the heading “index” instead). The number is the result of dividing the “top” by the “bottom” in the SGAI row. The value 0.8875 is less than one, indicating that the ratio SG&A / Sales fell (because ‘this year’ is on top, for SGAI).

In the SGAI row, in the column headed “top”, the blue mark on the left indicates that the quantity on the top of the ratio relates to “this year” (2013, the latest with a 10-K), and the pink mark on the left under “bottom” indicates that the quantity on the bottom of the ratio relates to the previous year, 2012. The marks were added to the graphic, but you could use background or border colors to show which year is on top in a spreadsheet.

DSRI, the Days Sales in Receivables Index

this year’s Days Sales in Receivables / previous year’s Days Sales in Receivables

Coefficient +0.92

A bill is said to have been outstanding for 31 days if 31 days have elapsed since the date of the bill, without the bill being paid. It’s useful to know how long the average bill has been outstanding. Using a year’s sales, the common formula for DSO or Days Sales Outstanding is:

    DSO = 365 * Accounts receivable / Sales (annual)

although this is easier to understand:

    DSO = Accounts receivable / Sales per day

You can check the definitions in “Days’ Sales Outstanding (DSO) Ratio” (accountingexplained.com)

DSO as calculated above gives an accurate result for the number of Days Sales In Receivables, which is the terminology Beneish uses. It does not always give an accurate picture of how long bills have been outstanding for, when sales or collection are lumpy, but it might be the best possible estimate given the information available.

It’s logically possible for all the bills in Accounts receivable to date from the first month of the year, or from the last week of the year, if business is sufficiently lumpy to allow the cases. Even if Accounts receivable are 90% of Sales, the sales could have been made in the last week of the year. Those are extreme cases, but they illustrate the need to check the DSO figure if you want to be sure about it. The checks include looking at the quarterly Sales and Accounts receivable figures, reading the ‘Liquidity and Capital Resources’ section of the 10-K, and knowing how lumpy sales and collection can be for the business. Searching the 10-K for ‘receivable’ could find relevant information.

Replacing ‘Days Sales in Receivables’ in the definition, with the formula for DSO …

    DSRI = (365 * this year’s Accounts receivable / this year’s Sales) / (365 * previous year’s Accounts receivable / previous year’s Sales)

… the ‘365’s cancel, to give

    DSRI = (this year’s Accounts receivable / this year’s Sales) / (previous year’s Accounts receivable / previous year’s Sales)

which is the formula I used in the spreadsheet, but with cell references, and split into ‘top’ and ‘bottom’ before dividing.

Female Health Co’s DSRI is 0.3250, which represents a dramatic fall in DSO from 76 days to 25 days. (I got the DSOs by multiplying the ‘top’ and ‘bottom’ numbers in the spreadsheet by 365.) Big changes in DSO can be expected due to the lumpiness of Female Health’s sales and the high customer concentration. It’s likely that in future DSRI could be high, possibly even pushing the M-Score over the commonly used threshold of -2.22. The reaction to a high DSRI should be “don’t panic”, and the big customers will probably pay their bills. However, it would be worth checking the 10-K (or 10-Q) for anything that allows customers to delay payment or return goods, or any change in the policy for revenue recognition. The current policy for product sales starts:

    “Revenue recognition: The Company recognizes revenue from product sales when each of the following conditions has been met: an arrangement exists, delivery has occurred, there is a fixed price, and collectability is reasonably assured.”

There’s also a small amount of revenue from royalties. If this grew, it might be necessary to adjust the DSRI calculation to exclude the effect of royalties.

The full policy is four times as long as the quote, but that still leaves the policy clear and short. Ubiquiti Networks sell through resellers and distributors, and their “Recognition of Revenues” policy covers a page, with details like “Revenues allocated to the PCS are deferred and recognized on a straight-line basis over the estimated life of each of these devices …”. I expect Ubiquiti’s revenue recognition is honest and conservative, but it’s hard to tell by reading the policy.

It’s not impossible for a big customer or funding source to become short of cash. For the funding sources and international agencies, any reduction in funding that would hit Female Health is likely to show up in the news rather than the DSO number, and if it’s in both, the news will be first. See “FHCO – Some thoughts about program-funding” below.

While Sekunjalo and possibly other customers are more like regular businesses, receivables are not broken down by customer or type of customer in the 10-K.

Female Health Co’s DSRI of 0.3250 is well below 1 and points in the opposite direction to earnings manipulation.

To illustrate the customer concentration (a hyphen indicates less than 10%):

    Customer, 2013, 2012, 2011
    UNFPA, 62%, 40%, 25%
    USAID, -, 25%, 26%
    Sekunjalo, -, 20%, –

UNFPA -United Nations Population Fund
USAID – United States Agency for International Development
Sekunjalo – Sekunjalo Investments Corporation (PTY) Ltd, the Company’s distributor in the Republic of South Africa

GMI, the Gross Margin Index

previous year’s Gross Margin / this year’s Gross Margin

Coefficient +0.528

A declining gross margin is one of the indications of poor prospects that predispose management towards earnings manipulation.

GMI has a particular quirk, because it’s possible for a company to have a negative gross margin. That’s unlikely for an established company, rather than a startup, but SunEdison (SUNE) managed it in their 4Q 2013 quarter, so a whole year of negative gross margin shouldn’t be impossible. (SunEdison are more ‘fresh start’ than startup, and are probably doing quite well, see my recent comment under my piece “SunEdison still set for growth and profit”.) Suppose in a bad year gross margin fell from 20% to -20%. Then, GMI = 0.2 / -0.2 = -1. Because the coefficient is positive, a GMI of minus one pulls the M-Score down, saying “everything’s fine here”. If the gross margin recovers to 20% in the next year, then GMI = -20% / 20% = -1, and the recovery produces the same index value as the drop to negative gross margin. More generally, if the sign on gross margin flips, GMI is negative, whether the sign flipped from plus to minus or from minus to plus. The minus sign is more accurate for the recovery. In positive territory, GMI doesn’t shout if this year’s gross margin is low, so long as it’s not worse than in the previous year. In other words, like the other indexes, it’s only sensitive to change.

GMI is also scale-sensitive, as a company which essentially breaks even at the gross profit level will not have exactly zero gross profit, but a fairly random amount of gross profit or loss, so meaningless changes like 0.3% to 0.1% are possible.

Female Health Co’s gross margin fell from 58.86% to 55.65%, giving GMI = 0.5886 / 0.5565 = 1.0578. Here’s a simpler way to calculate the effect on the M-Score compared to the ‘no change’ state. It works for seven of the eight indexes, for which the ‘no change’ state = 1, but not for TATA which is closer to zero when there’s no change:

    Effect on M-Score = (index – 1) * coefficient = (1.0578 – 1) * 0.528 = 0.0578 * 0.528 = 0.0305

So, a fall in gross margin resulted in a GMI over 1, increasing the M-Score by 0.0305 compared to the neutral “no change” situation where GMI = 1.

The term in the M-Score for an index is the value of the index multiplied by its coefficient, so except for TATA the term for the neutral ‘no change’ case always equals 1 * the coefficient, which is just the coefficient. In the case of GMI, the coefficient = 0.528, and so the neutral term = 0.528.

For Female Health’s GMI, the effect on the M-Score is small (only adding 0.0305). At 55.7% the gross margin is still high, and volatility is expected, with lumpy sales translating to lower sales in 2013 and less sales to spread fixed costs over, which reduced the margin. I don’t regard the fall in gross margin represented by GMI as indicating poorer prospects or a significant risk of earnings manipulation.

I haven’t seen fixed costs mentioned for 2013, but for 2012: “The increase in gross profit was the result of higher unit sales which improved the absorption of fixed overhead costs.”.

To illustrate the lumpiness of the business, regarding 2011 and 2012:

    “During fiscal 2011, the Company’s unit shipments, revenues and net income were adversely affected by bureaucratic delays and other timing issues involving the receipt and shipment of large orders from Brazil and the RSA. Significant orders for both countries were received in the first quarter of fiscal 2012.” (RSA – Republic of South Africa)

    “Gross profit increased $10,755,823, or 109 percent, to $20,621,013 in fiscal 2012 from $9,865,190 in fiscal 2011. Gross profit as a percentage of net revenues increased to 59 percent in fiscal 2012 from 53 percent in fiscal 2011.”

Those delayed orders made 2012 a tougher year to compare 2013 against. The quotes are from the 10-K for 2013.

AQI, the Asset Quality Index

this year’s ratio (Non-current assets – Property plant and equipment) / Total assets
divided by the previous year’s ratio

Coefficient +0.404

An AQI greater than 1 indicates a deterioration in the quality of the company’s assets, and the positive coefficient means that a deterioration in asset quality pushes the M-Score up, indicating more chance of earnings manipulation.

If I define:

    Suspect assets = Non-current assets – Property plant and equipment


    Suspect assets ratio = Suspect assets / Total assets

then AQI = this year’s Suspect assets ratio / the previous year’s Suspect assets ratio

and AQI goes up as the Suspect assets ratio goes up, which translates to:

AQI goes up as the quality of assets goes down.

I can re-write Suspect assets as:

    Suspect assets = Total assets – Current assets – Property plant and equipment

which makes it more clear that Current assets and PPE (Property plant and equipment) are excluded from the top of both the Suspect assets ratios in AQI.

The exclusions are because the distortion of some assets is picked up by other indexes, and the M-Score doesn’t cover cash.

While cash flows can be inflated, and the cash held can be increased as a result, the M-Score was not intended to specifically identify the manipulation of cash flows, although it can pick up deferred costs (explained soon) which affect the cash flow figures as well as earnings. Cash is excluded from AQI as a result of excluding Current assets. The rest of Current assets are taken care of by the TATA index, which checks the change in non-cash working capital. The value of Property plant and equipment could be distorted upwards by under-depreciation, which is tested by the DEPI index. The M-Score doesn’t cover everything, see “Flaws in the M-Score – no check on non-current liabilities” and “Beyond the scope of the M-Score” below. The “Beyond the scope” section covers ways to manipulate earnings which affect the balance sheet, but it’s hard to say they distort the balance sheet.

The equation above can be rearranged to:

    Total assets = Suspect assets + Current assets + Property plant and equipment

which is true by definition (the definition of Suspect assets).

Any distortion of assets ought to show up as a change in “Total assets” on the left, which must be balanced by a change in at least one of the three categories on the right. Asset distortion can now be broken down into three areas:

    Potential asset distortion in a year =
    Increase in Suspect assets + Increase in non-cash working capital + Probable under-depreciation

I’ve morphed “Increase in current assets” to “Increase in non-cash working capital” because that’s the main focus of the TATA index. “working capital” means current assets minus current liabilities, and “non-cash” means that cash, marketable securities and debt are excluded. I explained above that the M-Score doesn’t look specifically at cash, and it’s reasonable to extend that to exclude marketable securities and debt from the top of the TATA ratios, though in my opinion it could depend on the company having a reasonable and safe definition of marketable securities.

Total assets is used as a scaling factor on the bottom of ratios in AQI and TATA, and there’s no point in excluding either ‘good’ or ‘bad’ assets from Total assets when it’s on the bottom.

There’s a case for scaling by revenue instead of by total assets, which I don’t look into, but I think it’s not a good idea as revenue usually varies more than total assets. When total assets change more than revenue, I think it’s worth scaling by revenue to see what’s happening with the assets.

The Probable under-depreciation is different to the two increases. That’s because it’s calculated using depreciation, which is a charge that reduces net asset values. The two increases would have to be calculated by subtracting the size at the start of the period, from the size at the end of the period. That’s actually done for non-cash working capital in the TATA index, although depreciation also makes an appearance there. AQI does not use subtraction to calculate the increase in Suspect assets, instead it calculates the change in the proportion of assets which are Suspect assets, as a ratio.

The Suspect assets (my term) are what’s left from Total assets after the exclusions, and the theory is that the leftovers could represent improperly deferred costs. An example would be improper capitalization of R&D (research and development) which should be an operating expense, unless there’s a good enough reason to value the intellectual property produced by R&D as an asset. By capitalizing R&D, the expense can be transformed into investment, and without deduction of the expense, the operating income, net income, cash from operations and cash into investment are all inflated. See “IMPROPER CAPITALIZATION AND THE MANAGEMENT OF EARNINGS” by Ryerson, Frank E. III, Macon State College, 2009 (8 page PDF) (asbbs.org). Find “Fine Host” to get to the actual cases, which are easier to follow.

It’s been argued that securities or long term investments should be excluded, to give:

    Total assets – Current assets – PPE – Securities – Long term investments

on the top of the ratios. I’m not completely convinced. Any securities in current assets will be excluded from AQI anyway, as current assets are excluded. When securities are not in current assets, they should only be excluded from the AQI calculation when it’s clear that they have been given a genuine salable value, just in case they could be overvalued, because no other index would pick that up. Find “multiple failed auctions” above, about Lattice Semiconductor’s problem investments. They are honest about it and I’m not suggesting earnings manipulation, but their problem investments used to be classed as short term. That shows you can’t even rely on the value of marketable securities in current assets, without digging into and assessing the details. I might be pitching anecdote against statistics, but this isn’t science and the statistics can change.

The PDF “Popular Earnings Management Techniques” (linked to below) has a section titled “Flushing the investment portfolio” which explains how investments can be sold selectively to smooth earnings, choosing when to realize a gain or a loss. While that’s earnings manipulation, the sale of investments does not leave inflated values on the books, so keeping investments included in the top of the AQI ratios won’t help in such situations. Unless management have been very inventive, investments aren’t like the deferred costs which AQI was aimed at.

Ycharts deduct securities and long term investments in their AQI calculation, and cite some academic support, see “Beneish M Score” (ycharts.com). (They also use Depreciation and Amortization where only Depreciation is wanted.)

It would not make sense to calculate a company’s M-Score and at the same time trust management’s valuations. The famously fraudulent Enron Corp. provide an extreme case of mis-valuation. They valued contracts as if they were a securities trader, not a utility company, but the markets for the contracts were largely hypothetical. With no true market price, the valuation was a management decision, and the result of that was wildly inflated values. That supports not excluding any quantity from AQI (in the top of the ratios) where management judgment has been exercised, except for the quantities tested by other indexes (which are Current assets and PPE). The test ‘where management judgment has been exercised’ is wider than only looking for deferred costs.

Suppose a company arranges a sale-and-leaseback transaction for a large amount of PPE. First, there will probably be a gain or a loss on the transaction, but that’s dealt with under “Beyond the scope of the M-Score” below. If the company uses the cash to retire debt, the leverage ratio is affected, but there’s also a reduction in Total assets. That affects ratios which have Total assets on the bottom to scale the quantities on top. AQI and TATA will both be increased, even though the low quality assets and accruals they are testing are not affected by the sale-and-leaseback transaction. (If TATA was negative, it would get more negative.) It would take a big transaction to make much difference, but when that’s the case it’s an effect to bear in mind. The effect will depend on the nature of the lease and on proposals to put all lease obligations on the balance sheet, see “Operating lease obligations – billions or trillions to land on balance sheets” below.

One possible result of a sale-and-leaseback transaction followed by debt reduction is:

    1. A reduction in leverage (and the LVGI index)
    2. New off-balance sheet lease obligations, and
    3. False increases in AQI and TATA, due to less Total assets.

I say ‘false increases’ because in my opinion, AQI and TATA are ratios with an ambiguous meaning. While it’s reasonable to scale the low quality assets and the accruals by dividing by Total assets, changing the size of Total assets doesn’t really change the size of the effect of the low quality assets and accruals, and they stay in the same ratio to income if income is adjusted for the gain or loss on the sale-and-leaseback transaction. It’s a fairly subtle point, and if a growing company kept the same AQI and TATA ratios as it grew, I wouldn’t ignore growth in Total assets and focus only on the absolute increase in low quality assets and accruals.

It’s easier to illustrate an ‘ambiguous ratio’ with a different example. Suppose management learned that after checking the stock, 1% of the stock of finished goods was damaged. Obviously, it would be much worse if 10% of the stock was damaged. Now suppose a large order was filled using 91% of the undamaged stock. That would leave 10% of the stock damaged, but the company is better off, not worse off. Management would have to be idiots to say, “We can’t accept the order because that would leave us with 10% of stock damaged.”. So, is 10% of the stock damaged a worse ratio than 1% or not? There is no unambiguous answer. If damage occurs, you want as small a percentage as possible, but when the percentage rises through concentration rather than new damage, a high percentage becomes good because it means more sales.

This is an ambiguous ratio without any concentration effect: it’s usually a bad sign if a store has a high proportion of empty shelves, but if the shelves had just been cleared out by enthusiastic customers, it’s great.

A sale-and-leaseback transaction followed by debt reduction concentrates the more doubtful assets, but with no generally harmful effect that I can see other than increasing the AQI and TATA indexes.

Similar concentration of low quality assets and accruals occurs if cash is used to pay debt, but sale-and-leaseback puts big debt repayment within the reach of more companies.

I’ll rename my Suspect assets ratio to “proportion of uncertain assets” to avoid being too provocative, and using that still-prejudicial label, Female Health Co’s proportion of uncertain assets rose from 0.3702 in 2012 to 0.4602 in 2013, giving an AQI of 1.2433, and an M-Score term of 0.5023, which is 0.0983 above the contribution from the neutral case when asset quality doesn’t change. (The term for the neutral case equals the coefficient which is 0.404, and 0.5023 – 0.404 = 0.0983.)

The extra 0.0983 on the M-Score from AQI is not very big, but it’s a higher addition to the M-Score (above the ‘no-change’ case) than for any other index. It’s accounted for by the $4,900,000 rise in Deferred income taxes from $11,148,000 to $16,048,000, and a fall of $255,046 in PPE (net) from $2,349,876 to $2,094,830.

It’s what’s left after deducting PPE (net) that matters, rather than any change in PPE (net), so I don’t have to look into the small change in PPE (net), as I do next. The fall in PPE (net) is due to a $514,580 increase in accumulated depreciation and amortization (the two aren’t separated) with more Equipment, furniture and fixtures (before depreciation) offset partly by less Construction in progress that account for the $259,534 difference. None of that looks at all suspicious. I checked the ‘Useful life’s that assets are depreciated over and they’re normal and fairly conservative (Manufacturing equipment 5-10 years, Office equipment 3 years, Furniture and fixtures 7-10 years).

That leaves the $4,900,000 rise in the Deferred income taxes. The rise opens a Pandora’s box of complication, so I list my conclusions for anyone who can’t spare the time for the full treatment.

Conclusions about Female Health’s tax assets

1) The benefit from offsetting previous losses against taxable income has been real enough in the past. This is supported by tax payments which were lower than would be the case otherwise. The signal from AQI that dubious assets have increased, is probably false. There is no evidence of earnings manipulation here, even though the index is the biggest indicator of manipulation, numerically. Tax losses provide an opportunity for earnings management, but the evidence is against any inflation of earnings.

2) There’ll be little further benefit to Net income from previous losses, as most of that cake has already been eaten. When management aren’t confident that all the potential benefit from NOLs can be realized in the foreseeable future, the Deferred income tax valuation allowance holds the amount of potential tax benefit that’s not likely to be used, and the Net deferred tax assets hold the amount of potential tax benefit that is likely to be used. Net income has been increased as Gross deferred tax assets have been transferred from the valuation allowance to the Net deferred tax assets. There’s only $2,703,000 left in the valuation allowance. In future, when Net Operating Loss (NOL) carryforwards are offset against taxable income, the tax asset will be reduced, canceling the effect on Net income. The valuation allowance has generally fallen as growth meant more NOLs were likely to be utilized before expiry. The “little further benefit to Net income” is what really matters to investors, rather than trying to guess the intention behind the transfers to the deferred tax assets, although that hasn’t stopped me from noting the temptation to manipulate, and arguing that the deferred tax assets have been valued conservatively, with an effect opposite to inflating earnings.

3) The 2013 Net income of $14,342,598 includes $2,070,947 Utilization of NOL carryforwards, and $5,845,000 Decrease in valuation allowance, which sum to $7,915,947, or 55.2% of the Net income. If the valuation allowance had run out at the end of 2012, Net income would have been only $6,426,651. The $2,070,947 Utilization of NOL carryforwards would have been offset by a $2,070,947 reduction in the value of Net deferred tax assets, with zero net contribution to the tax expense. The $7,915,947 lifted the company’s tax expense to an Income tax benefit of $4,408,744. Doubters ought to ask how the company has been able to report substantial Income tax benefits for the past five years, while the direction of actual payments has been from the company to the IRS (Internal Revenue Service).

4) The deferred tax assets will still be good for cash flow, as they reduce the tax payments, which are in cash. The benefit will continue until the Gross deferred tax assets of $21,326,018 have either been used or the associated NOLs have expired. The $2,703,000 in the valuation allowance is more likely to expire before use, than the deferred tax assets. Growth is not strictly required by the aggregate figures to avoid NOLs expiring, but that leaves the possibility that growth of taxable income in the U.S. is required in the next few years to completely avoid NOL expiration.

5) There’s uncertainty, because the deferred tax assets depend on difficult projections which involve judgment. I believe the estimates have been conservative, but there’s room for error as there are factors which do not involve growth, such as tax planning strategies.

6) If future earnings are not high enough to avoid transfers from Net deferred tax assets back into the valuation allowance, then with hindsight the asset quality won’t look good, but that’s unlikely to be through earnings manipulation. While the tax assets are probably high quality, the Net income is low quality because it contains a source which will run out soon. Management don’t seem to have alerted investors to the issue, but they could argue that the information is there and it was up to investors and analysts to process it. The new CEO might prefer clarity.

7) In general, the GAAP requirement for management to base their valuation of tax assets on long term projections offers an opportunity for earnings management. Departing management also have an incentive to make the figures look good. However, no-one can be blamed for having unavoidable incentives, and my argument that management’s projections have been conservative is evidence against earnings manipulation (see “FHCO – Female Health’s deferred tax assets imply management’s forecasts have been conservative”, below). In addition, the M-Score of -3.31 is not indicating a risk of earnings manipulation, AQI is the only index pushing the M-Score up by much, and investigation around the other indexes mostly found reasons why the index could give a false indication of earnings manipulation in the future, due to the lumpiness of orders and cash collection.

8) The figures for ‘Decrease in valuation allowance’ for 2013 and 2012, are not exceptional given the history. While the valuation allowance has dropped dramatically in the past two years, the chart below shows that the “Decrease in valuation allowance” which contributes to Net income (through the tax benefit) has been roughly in line with the income before tax. Because the big decreases were not abnormal given the income before tax, they do not indicate earnings manipulation. All the dollar figures are in thousands.
Female Health - decrease in allowance vs pre-tax income - spread n chart

All the numbers in the charts above are from either the 10-K for 2013 or the 10-K for 2011. The original 2011 figure of $2,279 became $2,145 in the 10-K for 2013. I haven’t checked the other figures for revisions.

If that’s all you want to know about the company’s tax, skip to “How management use ‘cookie jars’ to smooth earnings”. You might want to check the other charts on the way.

More about Female Health’s tax assets

The background regarding Deferred income tax is complicated and for a full explanation you’ll need to read ‘Note 6. Income Taxes’ in the 10-K. The company made losses until 2005, and has made taxable profits since, but still has losses that can be offset against income to reduce tax. These “net operating loss carryforwards” are big, but the extent to which they can be utilized depends on the taxable income the company will make in the future, and where they make it (or where they are able to shift it to).

    Sizes of net operating loss carryforwards –

    U.S. federal – $19,165,000
    U.S. state – $17,220,000
    (both expiring 2018 to 2027)

    U.K. – $63,264,000 (Can be carried forward indefinitely and offset against future U.K. taxable income.)

NOLs (Net Operating Losses) are described in “The Net Operating Loss Carryback and Carryforward” (accountingtools.com). The NOLs are offset against taxable income, not tax, and their value depends on the tax rates. For Female Health Co, that’s the U.S. federal, U.S. state, and U.K. tax rates.

The taxable income which NOL carryforwards offset is income before tax but after other adjustments, such as deducting an allowance for R&D (when it’s available). The Utilization of NOL carryforwards is the amount of tax that’s canceled when NOL carryforwards are offset against taxable income. For example the U.K. NOL carryforwards could offset $63,264,000 of taxable income, but the corresponding total of Utilization of NOL carryforwards would be much less, and is smaller the lower the statutory tax rate.

There’s an amount called the “Gross deferred tax assets”, which is an estimate of the total tax benefit if all Female Health’s NOLs were utilized. This is split into two amounts –

1) Net deferred tax assets – the amount of Gross deferred tax assets which the company expects to be utilized in the foreseeable future. “In management’s analysis to determine the amount of the deferred tax asset to recognize, management projected future taxable income for the subsequent ten years for each tax jurisdiction.” (10-K for 2013)

2) Deferred income tax valuation allowance – the Gross deferred tax assets that haven’t gone into 1) because there’s a 50% or more chance of them not being utilized in the foreseeable future (including expiry before use).

For 2013:

    Gross deferred tax assets $21,326,018
    – Deferred income tax valuation allowance $2,703,000
    = Net deferred tax assets $18,623,018

The Net deferred tax assets are split into two items on the balance sheet:

    Deferred income taxes (under Current assets) $2,552,000
    Deferred income taxes (under Other assets) $16,048,000

except for a small difference, because the balance sheet amounts add to $18,600,000, which is $23,018 less than the Net deferred tax assets. That might be accounted for by the tax asset “Other, net – Malaysia $23,018” if it didn’t get onto the balance sheet for some reason. Another complication is that the company re-used the term “Net deferred tax assets” for the amount left after subtracting $258,197 of Deferred Tax Liabilities, which also relate to Malaysia.

When the company offsets NOLs against income, they ought to use the NOLs closest to expiry in the relevant jurisdictions, rather than caring if they were classified in 1) or 2) as above.

Management acknowledge that the income projections and the estimates of usable tax losses based on the projections are not straightforward or particularly reliable, and I’d expect the projections to be difficult, complicated and require judgment. It’s the kind of situation where even capable and honest management can be wrong. Generally, when a company is in a situation where judgment needs to be exercised to arrive at figures that affect the accounts, that’s when investors need to be wary.

There’s also some uncertainty because tax authorities can audit recent tax returns and possibly want more tax.

    Tax audits –

    U.K. (the Inland Revenue) – audit within 1 year – includes 2012 (expires in 2014)
    U.S. – audit within 3 years – includes 2010, 2011, 2012. (which expire in 2015 through to 2019)
    Malaysia – audit within 6 years – includes 2007 through 2012 (which expire in 2015 through to 2019)

It’s only Malaysia that gives the company more risk of a tax audit due to the six year period before the books are closed, and this is balanced to some extent by the U.K.’s period of a single year compared to three in the U.S.. It’s hard to aggregate the risk for the three jurisdictions, as the breakdown by country (in Note 10.) lists both United Kingdom and Malaysia as having less than 5% of Net Revenues to External Customers. The United States accounted for 8.3% of the revenue. There are ways to transfer income from where the revenue occurred to where a company wants it, often involving one subsidiary charging another for a service, but the only clue that Female Health might be doing anything along those lines is that the low percentages of revenue in the three jurisdictions have not prevented the significant utilization of NOLs and the large transfers from the valuation allowance into Net deferred tax assets.

An Income tax benefit of $4,408,744 is added to the 2013 pre-tax income (instead of having to deduct tax). The benefit is calculated by starting with a tax expense of $3,378,000, which is adjusted for several factors, listed under “A reconciliation of income tax benefit”. The adjustments include items such as “Effect of lower foreign income tax rates” and “Effect of stock option exercise”.

The Income tax benefit is increased by two adjustments related to NOLs, at the bottom of the reconciliation. The first is $2,070,947 from Utilization of NOL carryforwards, and the second is $5,845,000 from the reduction in the “deferred income tax valuation allowance”. In recent years, when the company has reassessed it’s Gross deferred tax assets, the result has been to transfer an amount from the valuation allowance to the Net deferred tax assets by decreasing the valuation allowance and increasing the Net deferred tax assets by the same amount. The same amount is reported as “Decrease in valuation allowance” and added to the tax benefit (or subtracted from the tax expense, resulting in a positive benefit). The valuation allowance can also change due to expiry or changes in the statutory tax rates.

When $5,845,000 was added to the Income tax benefit, due to the decrease of the tax valuation allowance, it’s like saying “$5,845,000 was added to the Income tax benefit, due to the increase in the Net deferred tax assets.”. Then it’s apparent that the Net income is increased on the same basis as from the increase in value other assets (except that some other changes in value will go into “Income before income taxes”, and some go into “Other Comprehensive Income”). Depreciation provides a familiar analog, but in the other direction: if a machine is depreciated by $10,000, there’s a $10,000 depreciation expense which reduces income before tax, and a $10,000 reduction in the value of the machine net of accumulated depreciation. A $10,000 reduction of the Net deferred tax assets should have a fairly similar effect on net income as a $10,000 depreciation of an asset, although the accounting is different (the depreciation expense affects income before income taxes). A reduction in liabilities increases income in a similar way (this is sometimes abused by companies that set up liabilities to use as a ‘cookie jar’. See “How management use ‘cookie jars’ to smooth earnings” below). Also see “The Importance Of Other Comprehensive Income” by Ryan C. Fuhrmann, October 10, 2012 (investopedia.com).

Technically, in U.S. GAAP, the expense or benefit is recorded as if it’s the cause of the change in the value of an asset, but here it’s easier to think of it the other way round, and either way a change in the value of the deferred tax assets will need a corresponding expense or benefit.

While I believe that management’s projections have been conservative, if growth in taxable income is consistently below the growth projected, disappointing results will have misery heaped on top from lower NOLs utilization and a valuation allowance that eats the tax asset instead of the other way around, creating an expense instead of a benefit. It should take more than the poor results often produced by lumpy orders to require a reversal of the deferred tax assets, as the company has experienced considerable volatility and lumpiness without having to make such reversals. Find the table under “Although the boost to the tax benefit is fizzling out” for the deferred tax assets from 2001 to 2013, showing the figure has never fallen (although it was zero up to 2006).

$5,259,065 of Deferred income taxes is subtracted in the cash flow statement, as one of the adjustments to Net income to arrive at cash from operations. This implies that one way or another, $5,259,065 of Gross deferred tax assets contributed to Net income for 2013 without any cash flow.

There’s a detail which I can’t explain. The Valuation allowance for deferred tax assets fell from $12,600,000 (2012) to $2,703,000 (2013), a reduction of $9,897,000, which is more than the $5,845,000 Decrease in valuation allowance given in the tax reconciliation. The difference is $4,052,000.

The immediate cause of the difference is the fall in Gross deferred tax assets, from $25,900,000 to $21,326,018, a fall of $4,573,982.

The Gross deferred tax assets are reduced when NOL carryforwards are utilized or expire, or the benefit from them is revalued down because lower statutory tax rates are expected. Subtracting the Utilization of NOL carryforwards, the remaining difference = $4,573,982 – $2,279,000
= $2,294,982

The $2,294,982 can’t be accounted for by NOL carryforwards expiring, because the 10-Ks for 2013 and 2012 both describe the carryforwards as “expiring in years 2018 to 2027”. That leaves a downwards revaluation due to expectations of paying less tax anyway per $1,000 of taxable income (before utilizing NOLs), but I’m not confident that “whatever remains must be true” due to the complexity.

The fall in the valuation allowance is split into $5,300,000 Charged to Costs and Expenses, and $4,597,000 Deductions/Other, but the split isn’t explained, and neither figure is repeated anywhere else in the 10-K.

With only $2,703,000 in the rapidly decreasing valuation allowance, and no expiry until 2018, if the company keeps growing it’s fairly likely that no more NOLs will expire unused. After concluding that management’s estimates have been conservative, it looks like growth is not strictly necessary to ensure that all the NOLs can be used. It could still require growth up to 2018 in the U.S. federal jurisdiction and the right U.S. state jurisdictions, depending on exactly how much expires when and where.

What’s more important for the Net income figure in future, is that with the valuation allowance down to $2,703,000, there isn’t much left to transfer to the Net deferred tax assets, or to utilize without decreasing the Net deferred tax assets. This means there can’t be a repeat of the $5,845,000 Decrease in valuation allowance, which was added to the Income tax benefit, which was added to Income before incomes taxes to get the Net income for 2013. There will still be contributions from “Utilization of NOL carryforwards”, but these should be offset by a reduction in the Net deferred tax assets, until they’re all gone. Otherwise, the same amount of tax offset would add to Net income twice, first when it’s transferred from the valuation allowance to the deferred tax asset, and again as Utilization of NOL carryforwards.

The cash benefit will continue until the $18,623,018 of Net deferred tax assets plus the $2,703,000 valuation allowance are all gone.

What I conclude from the details is that the Deferred income tax asset is not like the improperly capitalized expenses which AQI is meant to signal, and I go on to show that the NOLs have had a real effect on cash. While the decreases in the valuation allowance have accelerated the effect of the company’s NOLs on Net income, the method is a GAAP requirement, so management can’t be blamed for it (find “Tax and the liability method” below). The value of the Net deferred tax assets is the result of projections by management, but the tax assets have genuine value provided the company earns enough, and the assets couldn’t be compared to improperly capitalized R&D, or the assets Enron could inflate due to the false markets in them.

It ought to be good news when most of the Gross deferred tax assets are judged to be in the “Net deferred tax assets” category, because the tax benefit is more assured and because management’s projections were favorable enough to allow it, but investors might not be aware of the unsustainable nature of the contribution to Net income from the Decrease in valuation allowance.

Although the boost to the tax benefit is fizzling out, it shouldn’t matter to anyone who looks at cash rather than Net income, or who has been unusually diligent. Such investors could still consider how other investors are likely to react when they become aware of the tax situation, or its effect on future Net income.

Female Health income cash and tax assets to 2013 spread

Female Health tax assets income and cash to 2013

Female Health tax assets and cumulative income and cash to 2013

Normally I’d say management would feel pressure to avoid reversing the tax assets back into the valuation allowance, due to the effect on Net income, and it would look like an admission that long term prospects are not so good. However, for the new CEO, there’s an incentive to reverse the tax asset sooner rather than later if a reversal starts to look unavoidable. That’s a huge ‘if’ which is only likely if earnings do not meet the expectations that the decreases in the valuation allowance were based on. Reversing the tax asset would create a reserve for later use, as parts of the valuation allowance could once more be transferred to the Net deferred tax assets.

Low tax payments show the benefit of NOLs

It’s obvious that tax assets mean lower tax payments for a profitable company, but in a piece about earnings manipulation I don’t want to take much for granted. The reliability of the value of the tax asset in general is supported by the relatively low tax payments:

    (in thousands)
    Total, 2013, 2012, 2011, 2010, 2009

    Income before income taxes
    $33,232, $9,934, $10,792, $3,232, $4,224, $5,050

    Utilization of NOL carryforwards
    $7,099, $2,071, $1,637, $973, $1,087, $1,331

    Tax paid (from the cash flow statements)
    $1,575, $346, $926, $57, $112, $134

    Utilization of NOL carryforwards / Income before income taxes
    21.36%, 20.85%, 15.17%, 30.11%, 25.73%, 26.36%

    Tax paid / Income before income taxes
    4.74%, 3.48%, 8.58%, 1.76%, 2.65%, 2.65%

    Tax paid / Utilization of NOL carryforwards
    22.19%, 16.70%, 56.57%, 5.86%, 10.30%, 10.07%

Figures for 2013, 2012 and 2011 are from the 10-K for 2013 where the term “Cash payments for income taxes” is used in the cash flow statement, but for 2010 and 2009 the term “Income tax paid” is used instead, under the heading “Schedule of noncash financing and investing activities:” at the bottom of the cash flow statement in the 2010 10-K. I rounded the figures to get them in thousands.

The tax paid has been only a small proportion of Income before income taxes, at 4.74%.

Because the tax paid has only been 22.19% of the Utilization of NOL carryforwards, the NOL Utilization has canceled 81.84% of the tax that would have been payable otherwise. That’s only a rough estimate, due to variation in the tax rates between jurisdictions. There’ll also be some error due to the timing differences between the income statement and the cash flows, but not much, as deferred tax liabilities only increased by $235,179 – $152,227 = $82,952, or 5.27% of the $1,575,000 total tax paid. Even with the unquantifiable total error, it’s still clear that the Utilization of NOL carryforwards has drastically reduced the tax payments.

I also show that the tax paid is low compared to ‘Normal tax’, by which I mean tax before utilizing NOLs and before transfers from the valuation allowance to the tax asset. The variation over the three years used shows that this is not an exact calculation.

    Normal tax = Tax before any adjustments + Normal adjustments

where Normal adjustments don’t include anything to do with NOLs or the valuation allowance.


    Income tax benefit = NOLs + Decrease in valuation allowance – Normal tax

That should be correct, because everything in the tax reconciliation is included somewhere in the formula. The reconciliation starts with “Income tax expense at statutory rates” and has various adjustments down to “Income tax benefit” at the bottom.

It’s possible to calculate Normal tax because the other quantities in the equation are in the reconciliation:

    Normal tax = – (Income tax benefit – Utilization of NOL carryforwards – Decrease in valuation allowance)

I could simplify the equation by changing the signs and getting rid of the brackets, but the way I’ve written it, it’s more obvious that if you start with a positive tax benefit, and deduct the positive sources of the benefit, you are left with the tax before adding the good stuff (with a bit of sign magic). Then,

    % of normal tax paid = Cash payments for income taxes / Normal tax

and aggregating the three years, only 24% of Normal tax was actually paid. This should be reasonably accurate because although there can be a delay between tax hitting the income statement and all the cash being handed over, the deferred tax liabilities only increased by $235,179 – $188,177 = $47,002 from 2011 to 2013. This lends further support to the proposition that use of NOLs has reduced the tax paid in cash, without a build up of deferred tax liabilities, although it’s obvious anyway from the figures above showing the low tax payments compared to Income before income taxes.
Female Health - Normal tax etc - spread

The NOLs used produced a cash benefit, and the Deferred tax assets are likely to be as useful. I argue later that the Deferred tax assets have been conservatively estimated, under “FHCO – Female Health’s deferred tax assets imply management’s forecasts have been conservative”, and conclude that the Deferred tax assets are fairly high quality. I can’t say they are top quality, because by their nature they are not the kind of asset which is useful in tough times, see “When tax accounting blinds” below.

I don’t believe the indication from AQI that the proportion of low quality assets has increased, and there’s no evidence here of earnings manipulation.

The quality of the company’s earnings is another matter. The 2013 Net income of $14,342,598 included the Utilization of NOL carryforwards and the Decrease in valuation allowance, which add to $2,070,947 + $5,845,000 = $7,915,947, or 55.2% of Net income that would not have been there if the valuation allowance had run out at the end of 2012. There’s only $2,703,000 in the valuation allowance as at September 30, 2013 (and now only $2,410,068 or less, due to the $292,932 Decrease in valuation allowance in 1Q 2014).

If the NOLs are utilized at the 2013 rate of about $2 million per year, the $18,623,018 of Net Deferred tax assets will last about nine years, with the $2 million per year benefiting Cash from operations (but not Net income).

There’s some evidence under “Do deferred tax assets benefit income when they are used?” below.

This is an extract from an email Female Health’s CFO sent me.

    Mar 19 (2 days ago)

    to me

    We do not anticipate any further reduction in the valuation allowance, therefore there will not be any more tax benefit as a result of the reduction.

    Utilizing the NOL will have a favorable impact on our cash flow.

    Hope this answers your question.


    Michele Greco

    Vice President and C.F.O.

    The Female Health Company

    515 North State Street, Suite 2225

The sentence “Utilizing the NOL will have a favorable impact on our cash flow.” suggests that Net income will not benefit from utilizing NOLs.

How management use ‘cookie jars’ to smooth earnings

The way ‘cookie jars’ work is that in a good year, instead of reporting a high net income, a false expense is created that inflates a liability. The expense offsets some of the net income. In a bad year, it’s easy to reduce the liability because it was overstated or false, and the reduction in the liability creates an increase in net income.

Krispy Kreme Doughnuts used to save any earnings above their target Earnings Per Share (EPS) into the liability for their incentive plan, with an incentive compensation accrual. In years when the target would have been missed, enough of the liability was reversed to reach the target EPS. There’s more about Krispy later.

Cookie jar reserves aren’t always disclosed individually, and could be lumped under a heading like these:

    Accrued Liabilities
    Accounts payable and accrued liabilities
    Accrued Expenses And Other Liabilities
    Accrued Liabilities And Other Operating Liabilities

The categories are legitimate, but they are open to abuse. Accrued liabilities are usually under Current liabilities where the TATA index will pick them up, but can be in Other liabilities where the M-Score doesn’t look. That’s where General Mills put their accrued liabilities, according to this, which is only for confirmation as it’s not a good read: “Accrued Liabilities” (mcgraw-hill.com).

I’ve no reason to suspect General Mills of using cookie jar reserves, I was just showing that accrued liabilities don’t have to be in Current assets. The M-Score is blind to non-current liabilities, and if that’s where management put the cookie jar, the M-Score wouldn’t see the cookies going in or coming out the jar to prop up income. If investors expect cookie jars to be in current liabilities, that’s a reason for management to put the jar somewhere else, if they can.

This is one of the few articles that actually say where the cookie jar is, but the situation is unclear and contentious due to the company being complicated: “Gulfport energy’s accounts payable and accrued liabilities” Tuesday, February 26, 2013

This is about cookie jars with restructuring charges and the allowance for doubtful accounts as examples: “Aggressive Accounting: Reserves, Allowances, Contingent Liabilities” May 3rd, 2011 (oldschoolvalue.com)

The PDF “Popular Earnings Management Techniques” (linked to below) lists seven jar areas from sales returns to pension plans.

I wrote about Deferred grant income above. It’s a liability which is created and then periodically reduced to release a grant into income over the lifetime of an asset instead of all at once. That’s quite similar to how a cookie jar works, except it’s a GAAP requirement, and it would be obvious if management adjusted the timing to suit themselves.

Flaws in the M-Score – no check on non-current liabilities

Non-current liabilities follow Current liabilities on the balance sheet where they’re usually called “Other liabilities”, but they’re sometimes referred to as Long term liabilities. They’re probably less likely to be affected by earnings manipulation, or the M-Score would have an index to check them, but possibilities exist, such as ‘cookie jar’ accounting with the cookie jar in Other liabilities (see the section immediately above).

Non-current liabilities were covered by Beneish’s index ‘ACCRUALS’ which he used in 1997, but the index was replaced with TATA in 1999. The theory behind ACCRUALS is that income not backed by cash is low quality and less likely to be repeated. I write ACCRUALS as:

    ACCRUALS = (Income – Non-recurring items – Cash from operations) / Total assets

Every dollar on the top of ACCRUALS would accrue somewhere on the balance sheet, and while non-current liabilities aren’t specifically targeted, an abnormal reduction of non-current liabilities would have an equivalent contribution on the top of ACCRUALS. See also “The M-Score’s coverage of the balance sheet” below.

SGI, the Sales Growth Index

this year’s Sales / previous year’s Sales

Coefficient +0.892

Most investors want to see sales growing quickly, but fast sales growth isn’t all good news, and the SGI index is the M-Score’s speed cop. Beneish gave reasons why the growth creates incentives for earnings manipulation, which you can find in his 1999 PDF. I’ve paraphrased Beneish’s points and added to them. I also say why I believe they aren’t very relevant to Female Health, although that’s for consideration in the future, as sales fell and SGI pulls the M-Score down, relative to the neutral constant-sales case.

Sales growth promotes earnings manipulation through two incentives and a problem, in short: 1) The need for capital to fund the growth. 2) A later slowdown in growth can signal poorer prospects. 3) High growth companies often lack proper financial controls.

1) Growth companies need capital to fund the growth, and are often in debt due to the funding of previous growth. If sales or earnings fall short of targets, then raising capital is harder and more expensive.

This won’t affect Female Health Co, as they generate cash, and their investment requires little capital relative to the growth in sales and cash from operations that results, assuming that reasonable utilization of capacity can be maintained. While the timing of sales, income and cash from operations can verge on erratic, the company has plenty of cash to see them through thin years, such as 2011 when bureaucratic delays pushed orders into 2012.

2) A later slowdown in growth can signal poorer prospects. Beneish actually says more about the pressure to manipulate earnings when a company’s share price is over-sensitive to early signs of a slowdown, but the two are related. Slower growth would not make a share price dive so much if investors expected growth to get back on track.

When a growth company has a slowdown, it could signal a permanent loss of momentum. Warren Buffett has written about the problems of retail, including the “shooting-star phenomenon” where high growth ends with a sudden drop. Find “shooting” in the 1995 Berkshire Hathaway letter to shareholders (berkshirehathaway.com/letters/1995).

Krispy Kreme Doughnuts, Inc (Wikipedia) went from high-flier to a disappointment for investors, although they’re still around. In August 2003, the shares were up 235% from the April 2000 IPO price, and with a price/earnings ratio usually above 50. This was followed by missing quarterly estimates in May 2004, and then a loss.

The SEC became interested on two counts, using an incentive plan for ‘cookie-jar’ accounting (see “How management use ‘cookie jars’ to smooth earnings”, above), and round-trip transactions: “In the Matter of KRISPY KREME DOUGHNUTS, INC., Respondent.” March 4th 2009 (sec.gov) (PDF). In 2003, $528,323 was reversed from the ‘cookie jar’ incentive plan, allowing EPS to beat the target by $0.01. Round-trip transactions sent cash to franchisees and back again, inflating sales and income. The M-Score is not good at detecting cash-based fraud.

Sales growth over the previous year, for the three years to 2003 was 36.5%, 31.1% and 24.6%,
and the growth in earnings per share (EPS) was 80%, 67%, and 47%.

The highest signal from SGI would be from the 36.5% growth reported in 2001, which translates to SGI = 1.365 because:

    SGI = this year’s Sales / previous year’s Sales
    = 1.365 * previous year’s Sales / previous year’s Sales
    = 1.365

and multiplying by SGI’s coefficient of 0.892 gives 1.2179 added to the M-Score, or 1.2179 – 0.892 = 0.3256 more than if sales had stayed constant.

Then for 2002 and 2003, using
extra on M-Score = coefficient * growth as a proportion

    extra on M-Score = 0.892 * 0.311 = 0.277
    extra on M-Score = 0.892 * 0.246 = 0.219

The author of “Stop Using the Altman-Z Score.” (linked to below) suggests short-selling distressed companies with extreme growth.

Beneish cites Fridson (1993) who is emphatic about the effort by management to avoid the appearance of decelerating growth. I don’t agree with the claim if it’s meant too literally, because deceleration is normal, and Krispy’s reported sales deceleration from 36.5% to 31.1% to 24.6% growth in 2003 was not a sign of trouble by itself, and did not stop the share price rise. Taken literally, growth of 500% followed by 490% is deceleration. I can understand that there’s a degree of deceleration that would concern investors and a similar degree of deceleration that management are tempted to hide by manipulation, which seems to have happened for Krispy’s figures as they would have been without manipulation.

Beneish’s writing and the high coefficient of 0.892 make SGI one of the more important indexes in the score, but I’m not completely comfortable with suspecting earnings manipulation when management achieve sales growth which is good news provided management are honest. I would need to find other evidence of manipulation, but I also prefer to invest in companies where cash and high free cash flow removes the problem of funding growth.

Manipulation is more likely when management commit to high targets for more than a year. Anything like “We’re going to grow 35% CAGR for the next three years.” is ‘above and beyond’. Management will be tempted to manipulate earnings if that’s the only way they can keep the promise, they might even have manipulation in mind as a backstop when they make the promise, and they might have filled up the cookie jar before promising earnings growth. They could also achieve unsustainable growth in other ways, by cutting R&D or providing the minimum service that fulfills contracts.

Below I calculate that a company with no change (as far as possible) has an M-Score between -2.845 and -2.48. No change means that many indexes equal 1, for instance GMI would be the previous year’s gross margin divided by the same margin this year. It might be impossible or unlikely for a company to have no change in its results except for sales growth, but if you can imagine it, 30% to 70% sales growth is enough to lift the M-Score from the ‘no change’ values of -2.845 to -2.48 up to -2.21, which is above the threshold of -2.22 commonly supposed to indicate earnings manipulation. (Half the calculation is: M-Score = -2.48 + 0.30 * 0.892 = -2.2124.) With no contribution needed from other indexes, it’s possible for a company with 30% to 70% sales growth and honest management to be flagged as having a serious risk of earnings manipulation. That seems like a high sensitivity to sales growth. (See “M-Score thresholds, probabilities, and sensitivity to sales growth”, below.)

It’s claimed that the 1999 version of the M-Score (the version I’m using) would have flagged Enron Corp.’s fraud in 1998, three years before Enron collapsed in 2001. That’s from “The Predictable Cost of Earnings Manipulation” (already linked to), but they don’t give the M-Score. The threshold was very probably -1.89. Enron’s revenue growth in 1998 was 54.2%, giving an SGI = 1.542.

I’ve calculated that sales growth somewhere between 66.2% and 107.1% is enough to put an M-Score over the -1.89 threshold, provided the indexes other than SGI don’t collectively pull the M-Score down, relative to their ‘no change’ contributions. The revenue growth of 54.2% does at least half the work of giving Enron its hypothetical early warning. Enron’s revenue growth was unprecedented and suspicious for a company with the massive revenue they reported, but SGI only measures the growth rate, and no part of the M-Score depends on absolute size. It’s likely that Enron would have failed the M-Score test due to low quality assets (AQI) and accruals (TATA) as well as SGI.

For more about the sales growth range of 66.2% to 107.1%, see “M-Score thresholds, probabilities, and sensitivity to sales growth” below.

The reliability of growth depends on the economic moat. While that’s true by definition, the difference between growth and moat isn’t always obvious, until the growth ends.

There’s a difference between areas where demand can decelerate with a big chance of never recovering, and cyclical or lumpy demand. When lumpy demand is a well known characteristic of the market or a company’s sales, attempts at smoothing won’t usually be worth the risk.

See also “Growth, bloat, decline and recovery” below.

Female Health Co’s growth looks reasonably safe until emerging competition makes serious inroads, although there are various risks such as reliance on a single site for manufacturing (Selangor D.E., Malaysia). The company’s share price is volatile and in my opinion, the market over-reacts to results. The lumpiness of orders is well known, but the reaction isn’t surprising given the small capitalization, the customer concentration, single product and now the emerging competition. My charts near the end show how the company has achieved long term growth with high volatility, but here’s some recent data showing a slide to 2011 and a sharp recovery:

    2013, 2012, 2011, 2010, 2009
    Net revenues $31,457, $35,034, $18,565, $22,222, $27,543
    Net income $14,343, $15,299, $5,399, $6,737, $6,535

The evident volatility means there’s little or no incentive to manage the share price or reassure investors and analysts by smoothing results. In contrast, the state-sponsored mortgage giant popularly known as “Freddie Mac” had a “steady Freddie” image to keep up, they manipulated earnings to smooth out a windfall (from a change in accounting rules), and ended up hiding losses until it became untenable (the details are in my ‘cash3’ post).

In Female Health’s case, there’s no incentive to keep the share price high to enable capital to be raised (see point 1), above).

Female Health only give general guidance and don’t guide with a range of revenue and income. This won’t completely stop share analysts from having expectations, but they shouldn’t expect with much confidence, given the lack of guidance from the company and the lumpiness of orders. The pressure to meet ‘street’ expectations is probably low.

Early in this piece, the only motivation mentioned for fraud is meeting ‘street’ expectations: “The SEC’s New Financial Fraud Task Force: Part III, Cases Following the Speech – Revenue Enhancement” by Thomas O. Gorman, 08-19-2013 (lexisnexis.com)

3) Financial controls are not as developed in very small companies, and companies that grow fast from a small start often don’t implement effective controls as soon as they ought to. This can be because the limited resources available are devoted to expansion of production, sales to new customers, and development and launching of new products, with more change to manage generally, from leases to hiring. Some companies are exempt from external audits of financial controls under the JOBS Act (Jump Start Our Business Act), and they tend to be the newer growth companies.

Female Health’s internal financial controls were OK’d by their auditors: “The Female Health Company and Subsidiaries maintained, in all material respects, effective internal control over financial reporting …”. The accounts were also OK’d. The auditors didn’t actually use the term “unqualified opinion”, but the opinion wasn’t qualified. The auditors McGladrey LLP are not one of the big four, but that’s usually more of a problem when a big company uses a small auditor, especially if it’s for many years.

The Audit Fees were $374,051, and Tax Fees of $45,600 were paid to the same company. I believe it’s better to avoid using the firm that does the audit for any other work, otherwise the firm might consider the business it could lose if their opinion about the accounts isn’t well received, although the fee for tax work is only 12% the size of the audit fee.

My impression is that Female Health Co don’t have the kind of frenetic growth activity that leaves little time for arranging proper financial controls. They should be generally used to doing things properly, having complied with FDA “Good Manufacturing Practices” and been “monitored through periodic inspections by the FDA and foreign regulatory agencies.”. Their cash and cash flow means there’s little pressure to cut corners, regarding financial control or anything critical (not counting failure to develop another product).

Female Health Co’s sales fell from $35,033,897 in 2012 to $31,456,778 in 2013. Dividing the 2013 sales by the 2012 sales, SGI was 0.8979, with an M-Score term of 0.8009, which is 0.0911 below the 0.892 contribution from the neutral case when sales stay constant (0.892 is the coefficient).

I’ve explained why a high SGI is unlikely to indicate earnings manipulation in Female Health’s case, and as the SGI is low due to reduced sales, there is absolutely no indication from SGI of earnings manipulation.

The M-Score terminology uses ‘Sales’, while Female Health report Net revenue. I don’t see any point in using gross revenue, even if it was reported, and I usually write ‘Sales’ instead of ‘Net revenue’ unless I’m quoting.

DEPI, the Depreciation Index

previous year’s ratio Depreciation / (Depreciation + net Property plant and equipment)
divided by this year’s ratio

Coefficient +0.115

The reason for adding Depreciation to net PPE (Property plant and equipment) is to rewind the net PPE back by a year, to the value before the latest round of depreciation. The ratio:

    Depreciation / (Depreciation + net Property plant and equipment)

is called the rate of depreciation, and a fall in the rate can signal that management are depreciating assets by less in order to understate the depreciation expense and inflate earnings. The depreciation expense is the result of an accounting construct, and a deliberate reduction is more likely to be about earnings manipulation than about reducing real wear and tear.

Because the previous year is on top, DEPI is the inverse of the change in the depreciation rate, so the more the rate falls, the higher DEPI is. Because the coefficient (0.115) is positive, more DEPI means more M-Score, pushing it towards signaling earnings manipulation. Cutting out most of the steps:

    A lower depreciation rate


    DEPI higher, M-Score higher, earnings manipulation more likely

DEPI is more accurate as an indicator when a company’s investment is smooth every year, or more accurately, when the investment in assets which are given the same number of years of ‘Useful life’ is nearly the same each year, otherwise DEPI will be particularly high or low in some years, in the normal course of business without any manipulation.

Suppose a company only has a single asset, bought for $4 million in year zero. The asset is given a Useful life of four years, starts with zero accumulated depreciation, and using the common straight-line method of depreciation, each year $1 million is added to the accumulated depreciation until the $4 million historic cost is accumulated. This table shows how the rate of depreciation increases automatically as the fixed annual depreciation is divided by a net asset value that gets smaller.

hypothetical depreciation

The highly simplified table has blank cells in the Rate of depreciation and the DEPI columns. If the company also had some smaller assets which were depreciated every year, that’s what the blanks would show, while the other cells would still be dominated by the large asset. In year 5, when the large asset was fully depreciated, the rate of depreciation would drop sharply from nearly 100% to a more normal level, and DEPI would go through the roof. For example, if the rate of depreciation fell from 95% to 10% in year 5, DEPI would equal 0.95 / 0.1 = 9.5. That’s an extreme example, but it shows how a high DEPI can have an innocent explanation when investment is very lumpy.

Some assets are depreciated down to a salvage value rather than zero. See “What is salvage value?” (accountingtools.com). There’s the potential for the salvage value to be fraudulently high, to reduce the depreciation expense. Fraudulently high salvage values can be picked up by DEPI when large investments in PPE are made with percentage salvage values higher than previously used. DEPI will also identify any big upwards revisions to the salvage value. A company which always has high salvage values will have less volatile depreciation rates as described previously for large assets, because the rate never reaches 100% for an asset with a salvage value. Permanently high salvage values won’t be picked up by DEPI, but the auditors should notice it.

It’s also worth searching a 10-K for ‘Useful life’, to find the assumed lifetimes that categories of assets are depreciated over. I interpret high numbers as indicating aggressive accounting, while low numbers indicate conservative accounting. Wide bands such as 5 to 10 years give management leeway to pick a shorter or longer useful life for any particular asset, although bands as wide as 5 to 10 years are common. Any change for assets that are already being depreciated ought to be disclosed (but it can’t be guaranteed), even if a new useful life still fits in the band. For the Useful life assumptions of four companies, which could be used for comparison, see my previous post, “Three cash3 companies”. The cash-rich cash-generating companies might not be representative, and if you find the Useful life assumptions of Masonite International (who were at the other end of the cash scale) you’ll see they have the highest assumed lifetimes.

Female Health’s rate of depreciation rose from 16.4% to 21.0%, making DEPI = 0.7822. Because DEPI is below 1 and has a positive coefficient (of 0.115), the index is pointing in the other direction to earnings manipulation, and pulls the M-Score down by 0.025 compared to the ‘no change’ situation, because:

    Effect on M-Score = coefficient * (index – 1)
    = 0.115 * (0.7822 – 1)
    = 0.115 * -0.2178 = -0.025

The annual depreciation rose from $461,447 in 2012 to $556,304 in 2013, while net PPE after the annual depreciation fell from $2,349,876 in 2012 to $2,094,830 in 2013.

The ‘Useful life’ assumptions are –

    Manufacturing equipment 5-10 years, Office equipment 3 years, Furniture and fixtures 7-10 years

which is reasonable IMO, and the wide 5-10 year band for Manufacturing equipment is not unusual.

TATA – Total Accruals to Total Assets

(this year’s working capital other than cash – the previous year’s working capital other than cash – Depreciation)
divided by Total Assets

and everything is ‘this year’ unless labeled ‘previous year’.

Coefficient +4.679

Accruals are what you are left with after subtracting cash-from-operations from net income. There’s broad agreement that such earnings are lower quality than earnings which are supported by cash, but not everyone agrees with the definition of accruals, see “What are accruals and do they matter?” below. The section includes a link where it’s claimed that investors are now wise to accruals, and low quality earnings get a lower share price. I expect that’s true or mostly true, but not everyone agrees. In the same section I explain why I stick to TATA, and not “ACCRUALS”, an earlier measure which Beneish later reverted to.

The TATA ratio or the amount on the top of it don’t seem to be a very popular measure of accruals, but after writing about accruals, Beneish describes TATA as a proxy for how well cash underlies earnings. While TATA will catch accruals that relate to current assets and current liabilities, and catch withdrawals from the ‘cookie jar’ when the jar is in current liabilities, I prefer to ignore Beneish’s use of the word “accruals” and say instead that TATA catches distortions in Non-cash working capital, just as AQI catches distortions in Non-current assets excluding Property plant and equipment, although TATA also looks at depreciation as a proportion of assets. The difference between TATA’s check on depreciation and DEPI’s, is that DEPI looks at the change in the rate of depreciation, whereas TATA looks at the rate of depreciation (scaled by Total assets instead of PPE).

It’s the difference between ‘this year’ and ‘the previous year’ that is calculated (for non-cash working capital) on the top of TATA, whereas the other indexes are ratios between ‘this year’ and ‘the previous year’ for a quantity (with variation about which year goes on top). This makes TATA’s neutral value more like zero (when there’s no-change), although that’s complicated by subtracting depreciation from the rise in non-cash working capital.

Non-cash working capital isn’t usually disclosed. This is how I derive a formula:

Starting with these:

    1) Non-cash working capital = Non-cash current assets – Non-debt current liabilities

    2) Non-cash current assets = Current assets – Cash & cash equivalents – Marketable securities under current assets

    3) Non-debt current liabilities = Current liabilities – Short term debt

Plugging 2 and 3 into 1:

    4) Non-cash working capital =
    Current assets – Cash & cash equivalents – Marketable securities under current assets – Current liabilities + Short term debt

The principle behind that is to remove cash, and anything that can be very easily turned into cash without affecting business operations, from the Current assets, to leave the Non-cash current assets. Similarly, Short term debt is removed from Current liabilities to leave the Non-debt current liabilities. It isn’t completely obvious that marketable securities should be excluded, because while they aren’t as relevant to operations as inventories and receivables, the reliability of their reported value is variable (find “multiple failed auctions” above) and their value could be inflated. Another problem is that companies can vary the split between cash and non-cash working capital. A toy retailer will have more cash and less inventory after Christmas, and then build up inventories ready for the next Christmas. At least that case has regular variation, whereas some companies such as Female Health have lumpy orders and cash collection.

My version of the theory behind TATA is:

1) Some inflated earnings aren’t backed up by manipulating cash.

2) Some earnings that aren’t backed up by cash end up in Non-cash working capital.

3) The two ‘Some’s intersect, so some inflated earnings end up in Non-cash working capital.

If part of the Accounts receivable are sold, then inflated revenue might have no effect on the Non-cash working capital that can be derived from the balance sheet. That shows how inflated earnings might not be detected by TATA or the M-Score if cash is manipulated to support them.

If revenue is inflated and Accounts receivable are not sold, the inflated earnings that result are not backed up by manipulating cash, and end up in Non-cash working capital (as in point 2) because Non-cash working capital includes Accounts receivable.

Point 2) has ‘Some’ rather than ‘All’ because, for example, an increase in Net deferred tax assets will increase earnings without any cash flow, and does not add to Non-cash working capital. It’s picked up by AQI instead. The under-depreciation of assets will also inflate earnings, not be backed up by cash, be picked up by DEPI, and not affect Non-cash working capital.

If revenue is inflated and Accounts receivable are turned into financial instruments, so long as they are payable within a year and not classed as marketable, they will be in Non-cash working capital and TATA should pick it up. There’s a bit of a loophole here, as the term ‘marketable’ is fairly elastic, find “Level 3 – It’s anybody’s guess” above.

Beneish uses the TATA ratio as a proxy for the extent to which earnings are not supported by cash, and it’s reasonable to ask: Why not just look at cash?

Beneish’s former measure “ACCRUALS”, does look at cash.

I’m in favor of looking at cash but I prefer to look at several years history of cash flows, with less chance of noise hiding the trend (if there is one). The picture can be partial and skewed if you only look at one or two years of cash flow. Female Health Co show that you can’t just look at Net income, and you can’t just look at the latest figure for Cash from operations. In 2013, most of Net income was from tax effects which won’t repeat, and from the 10-K,

    “The increase of $1.4 million in cash generated from operating activities in fiscal 2013 as compared to fiscal 2012 was primarily due to the decrease in the amount of outstanding accounts receivable at year end and the reduction in incentive payments.”

The 68% decrease in accounts receivable can’t be repeated, and the reduction in incentive payments was due to less good performance. Therefore, the increase in Cash from operations from $10.4 million to $11.8 million isn’t anything to get excited about. Beneish’s ACCRUALS would have signaled an increase in the quality of earnings, when with hindsight, the cash collection shows the company’s receivables were always high quality. I’m not denying cases where looking at cash over shorter periods can help to spot earnings manipulation, or show that a company will probably run out of cash even if the income looks healthy. If you read about the “deferred income tax valuation allowance” above, you might agree that Female Health’s historic cash flows are more realistic overall than their historic Net income, and are a better guide to future performance.

I don’t trust the ACCRUALS index based on two cash flow figures and a few other variables to give a reliable indication, although it might flag that something needs investigating. That might be too much opinion, as the index has been in, out, and back in the M-Score, and it must have tested alright the first time.

In my previous blog piece, “Three cash3 companies”, if you find “Sun Microsystems”, you’ll see how a windfall from a legal settlement was disclosed in the income statement, but if you only looked at the cash flow statement you wouldn’t know about it.

Here’s a hypothetical case.

    Net income = $20 million
    Cash from operations = $16 million
    Windfall = $18 million

If you don’t know about the windfall, you might calculate 16 / 20 = 80%, and say, “80% of income has turned into cash flow, I don’t expect them to match exactly, and 80% isn’t too bad.”

If you know about the windfall, you subtract it, to get adjusted Net income = $2 million, and adjusted Cash from operations = -$2 million. Then it’s obvious that net of the windfall, the cash flow is in the opposite direction to the income.

Windfalls are often from legal settlements and sometimes from having to do the accounts differently (mandated changes in accounting practices). Neither of those causes sound like they belong in cash from operations, but that’s where the cash ends up so long as the windfall is included in Net income, because there’s no non-cash adjustment for it when Net income is adjusted to arrive at cash from operations (by the usual ‘Indirect’ method), when the cash is received in the same reporting period.

This shows that if you want to check that earnings are backed up by the cash flow, you need to understand the cash flow in order to make adjustments if necessary, and the cash flow statement might not have the necessary information. An M-Score index looking at cash would know nothing about windfalls.

The paragraphs about Sun Microsystems in my ‘cash3’ blog are in a long section under the heading “The trouble with cash”. The section includes much about the manipulation of cash flows. It might not be a good idea to rely too much on a quantity that could be a target for manipulation, and ignore the change in non-cash working capital, although it’s likely that a company which fiddled its cash flow would avoid a big rise in non-cash working capital, either intentionally or as a side effect.

One problem I have with TATA is that a company could be punished for asset growth, even if non-cash working capital is a constant or decreasing proportion of assets, so long as the absolute increase outweighs the depreciation. This definition would remove the problem …

    this year’s (Non-cash working capital – Depreciation) / Total Assets


    the previous year’s Non-cash working capital / the previous year’s Total Assets

… and changing the ‘minus’ to ‘divided by’ would make TATA more like the other indexes, but there’s room for argument about whether either of those would be a better definition. It could be argued that the growth of non-cash working capital matters independently of growth in other assets, but I would argue that you wouldn’t expect a big company to have only as much Non-cash working capital as it had when it was much smaller.

I’d say it’s safer to look into the company’s situation, rather than assume there’s no problem when a high TATA is the result of Non-cash working capital growing at the same rate as other assets.

Female Health Co’s TATA is -0.0824. Because the TATA is negative and TATA’s coefficient (4.679) is positive, the company’s TATA is pulling the M-Score down, and not signaling earnings manipulation.

The low TATA is the result of high collection of cash from customers, reducing the Accounts receivable, which is a non-cash current asset and part of the Non-cash working capital.

Female Health could easily and innocently have a high TATA in the future due to the lumpiness of orders, as inventories could build up, and Accounts receivable can be high. The current liabilities are also quite variable. The company manufactures their condoms “… in a single leased facility located in Malaysia.” Any disruptions could affect the inventory level, and therefore accruals, but I can’t say how big or likely any disruption would be. If inventories fall to abnormal levels, that would pull TATA down, and if inventories recover from abnormally low levels, that would push TATA up.

If Female Health gets a high TATA in future, investors will need to look into the details and decide if it’s explained well enough by the usual lumpiness. It could be hard to tell if an inventory build up without firm orders is the result of the usual lumpiness, or major customers becoming less keen on the product. Clues could be from the company, or from customers and providers of funding if they state their priorities in public. It’s not likely that major customers and fund providers will reduce the priority they give the product for a while yet:

    “On July 11, 2012, World Population Day, the U.K. Government and the Bill and Melinda Gates Foundation held the London Summit. It was an invitation only meeting attended by public health officials, government officials, and private sector companies that supply contraceptives and related products. The primary goal of the London Summit was to increase access to contraceptives to an additional 120 million poor women in 69 developing countries by 2020. Achievement of this goal will reduce maternal and infant immortality, HIV/AIDS babies and orphans and health care costs.”

I have to say, I didn’t realize “maternal and infant immortality” was a problem.

    “At the close of the London Summit it was announced that commitments of $4.6 billion had been made to fund the 2012-2020 program.”

    “On September 24, 2013, the U.K. Government announced it was donating $1.5 billion over a 3 year period to the Global Fund for HIV/AIDS, TB and Malaria treatment and prevention.”

    “FHC was one of only fourteen companies invited to attend the London Summit.”

    Also at the summit, “… the Company has pledged to significantly increase its global education and training investment over the period from 2013 through 2018.”

    “U.S. Secretary of State Hillary Clinton, as part of the President’s Emergency Plan For AIDS Relief (“PEPFAR”), issued a blueprint for an AIDS Free Generation. In the blueprint it states that female condoms are unique in providing a female controlled HIV prevention option and that PEPFAR will work with partner governments and other donors to promote female condoms wherever effective programs can build a sustained demand.”

    (all from the 10-K for 2013)

If inventories are higher than planned or expected, it’s less of a problem when future demand is relatively assured and the product doesn’t deteriorate quickly. If existing competitors develop and grow, or a big company enters the market, inventory build-ups will be more of a concern, although the usual lumpiness won’t be any less.

The M-Score’s coverage of the balance sheet

I’m using the 1999 version of the M-Score, which has small areas of overlap. Accounts receivable are in DSRI, the Days Sales in Receivables Index, and are also in the broader TATA or Total Accruals to Total Assets index. TATA also includes depreciation, which is the focus of DEPI, the Depreciation Index. The overlap does not imply inefficiency, as Beneish measured the effectiveness of the indexes, and would have thrown out any that were redundant.

Beneish had previously used the ACCRUALS index, which I write as:

    ACCRUALS = (Income – Non-recurring items – Cash from operations) / Total assets

The quantities in the top of the ratio are annual flows, and the Total assets on the bottom are only for scaling. However, the accruals have to accrue somewhere, and have an effect somewhere on the balance sheet. The ACCRUALS index covers a wide range which overlaps with DSRI, DEPI and AQI, and also with TATA if both were in the M-Score at the same time.

Beneish uses “income before extraordinary items” where I prefer to subtract Non-recurring items, but there’s a problem whichever description is used. Some Non-recurring (or extraordinary) items also affect Cash from operations, such as a jury award. If a company receives the cash from a $10 million jury award in the same period as the award is included in the income statement, the top of ACCRUALS will be $10 million lower for no good reason. If the cash is received in a later period, the same problem occurs in that period. If Beneish wanted non-recurring or extraordinary items to be excluded from both Net income and Cash from operations, IMO he didn’t make it clear, but please feel free to check rather than trust my interpretation (see “The Predictable Cost of Earnings Manipulation“, already linked to.)

I wrote about a gap in “Flaws in the M-Score – no check on non-current liabilities”, above. It’s highly unlikely that a new index for non-current liabilities will be added, but if it happened, it might be worth seeing if the old ACCRUALS index should be included, without removing TATA (previously, the two indexes have been swapped around). It’s possible that ACCRUALS would not increase the score’s predictive power, as it would cover areas already covered by DSRI, DEPI, AQI, TATA, and the new index for non-current liabilities, but it could be worth testing, and if indexes overlap, it seems more balanced to have ACCRUALS overlapping all the relevant indexes rather than only some of them. One problem is that higher Accounts receivable would show up in three indexes: DSRI, TATA and ACCRUALS.

The advantage of having more indexes covering the balance sheet is greater when a spreadsheet is used, as the problem areas show up in the index values. Once the data is in, it’s easy to check if TATA (or ACCRUALS) is much above zero, and which of the other indexes are significantly above one.

Beyond the scope of the M-Score

This link has a window where you can view “Popular Earnings Management Techniques”, which is chapter 2 of “Earnings Management” by Thomas E. McKee. There’s also a button for downloading a free PDF of the chapter. I can’t vouch for the site, but I expect it’s safer than the average free download site. The chapter covers most of the topics in this section.

“Flushing the investment portfolio” is a section which explains how investments can be sold selectively to smooth earnings. Investments which can only be sold at a loss are sold in a good year, and investments where a gain can be booked are sold in a bad year.

The value of investments is included in the AQI calculation:

this year’s ratio (Non-current assets – Property plant and equipment) / Total assets
divided by the previous year’s ratio

but selling investments will usually reduce AQI, although it depends on how the cash is used. Very little of the effect on AQI depends on whether a gain or a loss was booked, and AQI can’t tell if the sale was timed to smooth earnings. I don’t mean to be too critical by listing such details, because the M-Score is statistical in nature. It’s a little like an insurance calculation, where no-one can predict if a particular building will be damaged by a hurricane, but the risk can be estimated. An important difference is that with earnings manipulation, management will adapt by finding the tricks that are the most difficult to detect.

Subsidiaries and assets can also be sold to book a gain or a loss when it suits management.

A sale-and-leaseback transaction allows the company to keep using an asset. If a loss is realized, it’s booked immediately, but a gain is dripped into income. The company can choose either a capital lease where the gain is spread evenly over the lifetime of the lease, or an operating lease which is similar to rental, with the gain each year in proportion to the rent. The GAAP rules effectively give management control over if and how gains are realized, and management can’t be forced to restate earnings for following GAAP. The rules for lease accounting should be changing, see “Operating lease obligations – billions or trillions to land on balance sheets” below. Management can distort PPE through the depreciation expense, and sale/leaseback gives them an alternative way to exploit PPE to adjust the income figure.

Even bond debt can be manipulated, because the company’s bonds are not valued at the market price, so early retirement of the bonds will see a gain or a loss booked, which management can time to smooth earnings. There’s no gain or loss when new bonds are issued apart from fees and admin costs, so new bonds can be issued at the same time as old bonds are retired without affecting the gain or loss booked on the bond retirement. Swapping one debt for another can have economic benefits, and SunEdison recently issued $1.2 billion worth of convertible notes which they used to pay off debt carrying a higher interest rate. Eastman Chemical arranged a term loan for an acquisition, and reduced the loan quickly using cash from operations and new debt because the term loan was expensive. Even when the gain or loss booked is the same as the economic benefit, management control the timing, although I’m not accusing the companies I’ve mentioned, and SunEdison have been particularly adept with their financing.

Deals as described above are hard to prove as earnings manipulation unless they obviously make no economic sense. When the manipulation is in the timing of a transaction, the reporting is likely to be accurate, with no need to restate the accounts. While that isn’t ideal, at least there’s not much risk of the SEC or auditors requiring a restatement of the accounts and causing the share price to fall. The gains or losses on the transactions described here should be reported outside of operating income.

If Female Health Co had any such deals, the gain or loss would be under the heading “Non-operating income (expense):”. Ignoring the “Foreign currency transaction loss”, the other category under the heading is “Interest and other income (expense)” which for 2013, 2012 and 2011 was $245,545, $362, and $2,109. Because the company has $8,922,430 cash and no debt, it will earn rather than pay interest, so net other income for 2013 can’t be over $245,545. If net other income for 2013 was zero, the interest rate on the cash would have to be $245,545 / $8,922,430 = 2.752%. That doesn’t leave much scope for net gains or losses. Even with zero interest earned, the 2013 net other income would be $245,545 or only 1.712% of Net income (of $14.3 million), or 2.511% of operating income. The same figures work as a cap on the maximum possible other expense, if the interest earned is at twice 2.752% or 5.504%. I’ve used an unfeasibly low interest rate of 0%, and an unfeasibly high rate of 5.504%, to contain the maximum possible net other income and net other expense, the only category that non-operating net gains and losses could be in (not counting currency gains). It’s safe to say that Female Health have not done the kind of deals listed above which can be timed to manipulate earnings, to any material degree. There’s more scope to time a non-operating loss in 2012 and 2011, but not enough to worry about.

As well as checking the income statement, sales of investments, subsidiaries and assets for cash should be described in the ‘Liquidity and Capital Resources’ section of a 10-K. Female Health call it “Liquidity and Sources of Capital”, and there were no asset sales described in the section in their 10-K for 2013.

It’s worth checking “Other comprehensive income”, because that’s where ‘fair value fraud’ is likely to hit the accounts. Female Health’s 10-K has “In fiscal 2013, 2012 and 2011, comprehensive income is equivalent to the reported net income.”. Because Comprehensive income is the sum of Net income and Other comprehensive income, it follows that Female Health had no Other comprehensive income.

Beneish tried incorporating several ratios into the M-Score, and rejected them because they did not improve performance. One of them was the ratio of net non-operating income to income before extraordinary items. I prefer the ratio Operating income / Net income, because it mirrors the ‘quality of earnings’ ratio which is Cash from operations / Net income, but I suppose it doesn’t matter which version of the ratio he rejected. Operating income / Net income would treat non-operating gains and tax benefits as low quality, in the same way that Cash from operations / Net income treats accruals as low quality. One complication is that the ratio needs a ceiling close to 1, because a company which regularly has supposedly one-off charges probably has or had low quality management rather than high quality earnings. The ‘one-off’s could be from big accidents, losing legal cases, or asset impairments due to a deteriorating market, poor investment, poor maintenance or over-optimism meeting reality. A relatively simple score can’t be expected to check the regularity of non-operating losses, and the conservative position is to assume they are regular.

Other rejects include Unusual items / Pretax income, Cash from operations / Debt service, an index like DSRI (Days Sales in Receivables) but based on inventory instead of receivables, ratios using the amortization of intangibles and pension funds status, and changes in capex and tax expense / income before tax. The statistical tests stop the M-Score from filling up with junk, but IMO there’s a risk that the evidence-based approach leaves holes in the net.

There doesn’t seem to be much on the balance sheet that’s immune from manipulation. The actual cash balance is fairly safe, but it can be the result of manipulating cash flows, and the definition of Cash and cash equivalents is often woolly. Marketable securities need researching to check the value is reliable (find “multiple failed auctions” above). Debt could be off the balance sheet, and a bank loan could be disguised as the sale of receivables. Liabilities including debt can be manipulated in the ways I’ve described, but most of them and all of the debt still have to be paid one way or another.

Each quantity on the balance sheet could change for a genuine reason, and it’s harder to identify the result of earnings manipulation when the areas affected can’t be narrowed down much. Imagine how easy it would be if earnings manipulation always showed up in Accounts receivable. It’s still useful to have indexes which highlight the areas with the most change.

See also “Flaws in the M-Score – no check on non-current liabilities” above.

What are accruals and do they matter?

Accruals are income not backed by cash, but there’s disagreement about how accruals should be defined and measured. Accruals matter theoretically because they underlie the accounting system that incorporated companies are required to use, and they matter in practice because the cash which accruals say is on its way, doesn’t always arrive. High accruals indicate relatively lower earnings in the near future. These days the lower earnings are anticipated and are in the share price, but there’s disagreement about that as well. For small companies, the pricing-in could be sporadic, causing price volatility. The pricing-in is recent and is likely to make the M-Score less useful, particularly for predicting share price performance.

The definition of accruals which academics use is that accruals are the result of subtracting Cash from operations from Net income. That’s the simplest definition, in my opinion it’s the most useful and it could also be the most common. I’ll call it the NIMCFO definition (Net Income Minus Cash From Operations), but no-one else names their definition because they think their’s is the only one in town. The NIMCFO definition uses the same variables as the quality of earnings ratio, which is Cash from operations / Net income.

The whole point of accruals-based accounting is that cash by itself does not tell the whole story. For instance, many liabilities have to be paid, such as Accounts payable, and if accounts payable build up in a period, Cash from operations will be higher as a result, but the bills will have to be paid in cash, and probably soon. (Against that, an increase in Accounts payable could signal that growth can be financed without needing to raise as much working capital as otherwise, but I’d still like to see the figures.) Accruals-based accounting includes many estimates made by management. The receivables figure depends on the policies for revenue recognition and recognizing when the debt is bad. The depreciation figure depends mostly on the useful life assigned to items of property, plant and equipment. Management decide on how inventories are valued, and when stock is obsolete or damaged etc. In particular, Net deferred tax assets depend on long term projections.

The ‘accruals anomaly’ is the supposedly surprising claim that accruals are of lower quality than other earnings. ‘Lower quality’ is about the same as ‘less sustainable’, so it’s no surprise that lower quality earnings indicate lower earnings in the future, but I don’t mean to minimize the work on accruals which was greeted with skepticism but has been proved in the market:

Accounting anomaly gets traded to death, proving its existence” December 16, 2011 (foster.washington.edu)

Academic Mark Soliman worked on showing that companies where accruals are a high proportion of earnings had lower earnings in the future. He published a paper outlining a trading strategy based on the observation. Few believed him, and other academics disputed the finding, so he traded, profitably. Belief grew, believers made profits, but as a result markets became so efficient that the high-accruals companies became fairly priced on average, and the trading strategy doesn’t work any more (but find “small companies with high accruals” below).

That’s about NIMCFO-based accruals, but accruals by other definitions will also have less predictive power for share prices.

There’s now general agreement that the accruals anomaly exists, but disagreement over whether it has been arbitraged away. Arbitrage trading is supposed to be when traders buy in one market and sell in another where the price is higher, with the effect of equalizing the prices in the markets, but the term now includes many of the trading strategies used by hedge funds. See “Arbitrage” (Wikipedia).

The accrual anomaly: Why investors should care about accruals & earnings quality” Dec 31 2011 (stockopedia.com)

According to Stockopedia, the anomaly has not been arbitraged away, supposedly because the shares affected were too volatile to make arbitrage trading safe. I have my own theory about the volatility, find “high enough to attract short sellers” below.

Earnings Quality Revisited” by Jennifer Bender and Frank Nielsen, The Journal of Portfolio Management Summer 2013, Vol. 39, No. 4 (iijournals.com/)

The measure of earnings quality they look into combines the size of accruals and how fast they are growing or shrinking. This stopped being predictive but made a comeback in 2008. While it’s an unusually pure signal of future out-performance, it doesn’t work as a risk factor. I’ve only seen the short abstract, but it might be out of date or conflict with other writing unless they’ve found a superior metric.

If you google:
“accruals anomaly” arbitraged away
and limit the search to the past year, there are plenty of results for further research.

Overall it’s not clear if the accruals anomaly is currently arbitraged away or not, but it won’t be ignored to the extent it used to be, and the M-Score is likely to be less effective as a result. I’m more convinced by the first story, about “Academic Mark Soliman …”, where the anomaly was pronounced dead in 2011.

If management were rarely dishonest or over-optimistic, the accruals anomaly would never have been as big, but could still exist because cash now is better than cash in the future.

Here’s a possibility which I haven’t seen mentioned. Suppose liars fiddle the cash figures, and accruals are honestly reported. That would reduce the accruals anomaly, and it’s something to bear in mind if you read that the accruals anomaly has shrunk, which is not the same as whether it’s priced-in or not. Manipulating cash flows has benefits for management because some investors don’t realize it’s possible.

YCharts give the NIMCFO definition of accruals. They mention the Beneish M-Score, and imply that a consistently large excess of Net income over Cash from operations indicates a low quality of earnings.

Beneish originally used an index called “ACCRUALS”, which I write as:

    (Income – Non-recurring items – Cash from operations) / Total assets

which is the same as

    (Accruals – Non-recurring items) / Total assets

using the NIMCFO definition of accruals.

In 1999 Beneish replaced ACCRUALS with TATA, the Total Accruals to Total Assets index, but he and and other authors reverted to ACCRUALS in “The Predictable Cost of Earnings Manipulation” (already linked to). Here’s a reminder of how I write TATA:

    (this year’s Non-cash working capital – the previous year’s Non-cash working capital – Depreciation) / Total Assets

    where Non-cash working capital = Current assets – Cash & cash equivalents – Marketable securities under current assets – Current liabilities + Short term debt

    and everything is ‘this year’ unless labeled ‘previous year’.

I’ve kept the 1999 definition of the M-Score which uses TATA, even though it’s less in line with the term “accruals”, because websites with material about the M-Score use TATA, and because investors who have wised-up to the low quality of earnings which accruals represent are most likely to look at ‘NIMCFO’ accruals.

For more about how ACCRUALS covers areas already covered by other indexes, but fills a gap, find “Flaws in the M-Score – no check on non-current liabilities” and “The M-Score’s coverage of the balance sheet”, above.

The wising-up of investors to accruals affects the predictive power of the M-Score regarding share prices. The predictive power of the M-Score regarding earnings restatement or fraud will also be affected, as management are less likely to risk earnings manipulation that distorts the balance sheet without the benefit of keeping the share price up. Less earnings manipulation will mean fewer correct indications of earnings manipulation, and therefore a higher proportion of false positives (companies with high M-Scores that have not manipulated).

There are two issues here for the usefulness of the M-Score: 1) How much the score depended on accruals, and 2) the investment areas where unsophisticated investors buy a high proportion of the shares.

1) In “The Predictable Cost of Earnings Manipulation”, it’s claimed that accruals did not account for all of the M-Score’s usefulness. That’s not surprising if the indexes for gross margin, sales growth, SG&A expense and leverage are any use at all, since GMI, SGI, SGAI and LVGI deal with the pressure to manipulate, which has nothing to do with accruals. It’s also claimed that the accruals anomaly existed largely because earnings manipulation inflated accruals, and that the M-Score “subsumed” accruals, meaning the accruals are efficiently incorporated into a larger system. This leaves some hope that the M-Score is still useful. If I’m right about two signs being wrong, a corrected M-Score would be improved, in areas not connected to accruals.

2) It’s possible that small companies with high accruals are bid up by unsophisticated investors, until the overvaluation is high enough to attract short sellers. If so, the process creates volatility, but without the short selling the share price would go higher and have further to fall when investor-expectations are not met. Management might not mind attracting unsophisticated investors, who blame short sellers rather than management when the share price dives, although I’m not sure how common that is in the U.S. (it’s rife in the U.K., which has the low-governance AIM market, and investment forums where short sellers can be blamed). After the vesting period, volatility gives management an opportunity to time how they realize their stock or option compensation.

It’s worth knowing about accruals even when they are in the share price. Learning about accruals would particularly benefit any investors who think they have a portfolio of bargains, when they have a portfolio of fairly-priced companies with low quality earnings.

Companies with high accruals have more cash tied up in the business. That could mean the business soaks up too much capital, making profitable growth more difficult, but it could mean instead that cash has been put into the business to drive growth. Other factors need to be considered, because the cash tied up could be compensated for by low needs for capital expenditure (for plant, equipment etc.), or aggravated by the need to hire and train ahead of expansion, for example.

To show how that works, suppose a company increases inventories to prepare for higher sales. If the inventory is valued at cost, the build-up does not affect Net income. If the company has to pay cash to build up the inventory, then Cash from operations is reduced. More cash is tied up in the business, and accruals are increased, because there’s less Cash from operations to subtract from Net income. I’ve simplified, and there’s more to valuing inventory than just saying “at cost”, but the principle applies to real companies. However, not all increases in inventories are to prepare for higher sales. If sales were lower than expected, an increase in inventories is a likely result. Obsolete, surplus or out-of-fashion stock could accumulate in inventory and create accruals, until the value is written down and the stock is sold at a discount or disposed of.

Laser company IPG Photonics are vertically integrated and have a short lead time of only two to four weeks, during which they assemble the lasers. The company say they test their lasers thoroughly, but it looks as if much of the time is spent testing the parts before orders arrive. The vertical integration is good for cost, quality and ensuring an adequate supply of parts, and the short lead time and thorough testing are good for sales, but those factors mean the company hold a lot of inventory.

IPG’s inventories are built up ahead of anticipated sales, particularly when new models are launched and sales ramp up. That contributes to accruals, but it hasn’t stopped IPG from having a good history of growing free cash flow per share, which I’ve charted in my two previous blog posts. Some investors will be concerned by the accruals, and analysts seem to give too much weight to the gross margin which falls slightly with growth. I list the most relevant figures because I want to show the difference between accruals as a kind of short term investment and accruals as low quality income. The distinction is rarely made so I’m probably in a minority. For IPG Photonics:

    (in thousands)
    2013, 2012, 2011
    $648,034, $562,528, $474,482 – Net sales
    $155,780, $147,744, $121,009 – Net income
    $119,367, $175,276, $87,351 – Net cash provided by operating activities
    $75,886, $55,257, $79,099 – Net cash used in investing activities
    $172,700, $139,618, $116,978 – Inventories
    52.5%, 54.2%, 54.2% – Gross margin

The figures suggest that accruals are low when investment is low, but I haven’t checked for other years. When accruals (and some expenses) are similar to short term investment, they are based on an expectation of demand, and for IPG Photonics there’s currently some risk if sanctions restrict trade with Russia, a market where there’s likely to be growing demand for IPG’s laser systems.

The M-Score was not meant for financial companies. I’ve read that financial companies have high accruals, but I’d say the state of the balance sheet usually matters more than NIMCFO accruals, and banks are known for keeping liabilities off the balance sheet.

The alternatives to the NIMCFO definition of accruals seem to apply to smaller or private businesses, and there could be some international variation in the definition. Some accountants might think of accruals as the things on the balance sheet with “accrued” in the label, like “Accounts payable and accrued liabilities”, which is different to NIMCFO accruals.

Some sites describe accruals as revenue earned and expenses incurred that have not been paid. The meaning is narrower than the NIMCFO definition, as a change in the value of an asset has no cash effect, so when the change is included in Net income (instead of Other comprehensive income), it’s an accrual according to NIMCFO, but there’s no revenue involved, so it’s not an accrual according to definitions based on revenue and expense. The same revenue-and-expense based definitions tend to imply that the accruals persist until cash is paid, so the Accounts receivable at the start of the fiscal year would be accruals. Because NIMCFO accruals = Net income – Cash from operations, the accruals must start the year at zero, and so they can’t include the starting value of any assets, such as Accounts receivable. Instead, only the change in assets or liabilities can count as accruals. (Reminder – NIMCFO is my abbreviation of Net Income Minus Cash From Operations.)

The ‘Dummies’ publishers describe an accrual as a paper transaction, in “Understanding Accounting Methods” (dummies.com).

Wikipedia list some accruals – accounts payable, accounts receivable, goodwill, deferred tax liability and future interest expense. It looks like Wikipedia wouldn’t reset accruals to zero at the start of the accounting period, whereas accruals are reset to zero in NIMCFO and Beneish’s ACCRUALS. The same applies to TATA because it’s the change in Non-cash working capital that’s used. Wikipedia distinguish between accruals, where there is more certainty about payment and timing, and provisions, where payment or timing are more doubtful. NIMCFO makes a different distinction, for instance if Female Health increased their ‘Provision for obsolete inventory’ by $1 million, their NIMCFO accruals would be decreased by $1 million, as the provision only reduces Net income, and not Cash from operations (but with some complications).

This is short, simple, not NIMCFO, and agrees mostly with ‘Dummies’: “What are accruals?” (accountingcoach.com)

I prefer to ignore Beneish’s use of the word “accruals” and say instead that TATA catches distortions in Non-cash working capital adjusted for depreciation, just as DSRI, AQI and DEPI catch distortion in their own areas.

It doesn’t matter which definition of accruals an author uses, so long as you know which definition the author is using. Taking that approach saves working out which websites can be trusted and who their information is intended for. Investors hoping to learn about accruals might have only read a single source, and the ones I’ve listed can’t all be exactly right.

With the accruals anomaly generally priced-in, management could react by fiddling cash or tailoring non-GAAP measures to their aims. High accruals need to be noted and understood, but now they could be less likely to be the result of earnings manipulation. Investors who want to be thorough can use the M-Score, but also need to look at the cash situation to see if it’s either poor or manipulated. There’s a section about fiddling cash in the excellent free PDF you can find with “Shenanigans”. My previous piece (about cash3 companies) includes “The trouble with cash” with much about cash-based deception.

Accruals on Old School Value

Jae Jun, the founder of oldschoolvalue.com, uses the NIMCFO definition of accruals (NIMCFO is my abbreviation of Net Income Minus Cash From Operations). There’s plenty of material about accruals, although like myself he generally prefers to look at cash.

You Need to Determine Earnings Quality Through Accruals” December 12th, 2011 (oldschoolvalue.com)

Checking Financial Accruals of a Company in 5 Minutes” December 27th, 2011 (oldschoolvalue.com)

Jae Jun had problems finding evidence that high accruals predicted a poor share price performance, and he suggested that instead of just checking for high accruals, a company’s accruals should be compared with competitors’ accruals. I’m not against comparing a company to competitors, but the problem might be the awareness of investors, or at least the big investors. Scroll up to “What are accruals and do they matter?” for the initial skepticism and later acceptance and pricing-in of the “accruals anomaly”. There’s also my point that in some cases high accruals could be the result of building up inventories in preparation for growth.

My opinion is that if cash flows look good, accruals are less important but are worth looking into. That’s the case with Female Health Co, where the growth of free cash flow has been good (though volatile), the cash alone is bigger than the sum of Total liabilities plus the contractual obligations, and there are no off-balance sheet arrangements. The non-cash tax benefits have generally produced more income than cash from operations, in other words, accruals. Small investors don’t seem to realize that much of the company’s net income has been unsustainable (although I’ve tried telling them). I can’t predict how they’ll react, but at least I know the possibilities.

Could You Have Predicted Diamond Foods Accounting Fraud?” January 30th, 2012 (oldschoolvalue.com)

The article above is about applying three methods, including measures based on accruals, and the Beneish M-Score, but not with very much detail about the M-Score.

The Old School pieces are best read in order. I’m not convinced that the complicated accruals-based measures are worth understanding now that sophisticated investors price-in the information about accruals.

Operating lease obligations – billions or trillions to land on balance sheets

Accounting rule changes could add $1.3 trillion of operating lease obligations to corporate balance sheets, according to Wikipedia in “Accounting for leases in the United States“.

The next link says “trillions” but the rules could be softened for two reasons:

    1) Fear of the consequences of making the information easier to find.

    2) The cost of accounting for the thousands of leases a company might have for small items like office equipment.

Companies set to win lease accounting concessions” by Huw Jones, Jan 27, 2014 (uk.reuters.com)

I’ve already said that for Female Health, “the total for Contractual Obligations is $2,147,677, all for lease obligations”, while only $66,799 of Deferred rent appears on the balance sheet. I dug out the obligations for three companies in my previous blog post. That eliminates some surprise, but how badly markets will react if or when the big weight lands on the world’s balance sheets is another matter. In a rational world, share prices would move according to how the amounts disclosed compare to rational expectations, plus a positive factor for the removal of uncertainty. Investment is supposed to be inhibited by uncertainty, but some interests seem to think or argue that markets can’t bear the truth.

There are many different ways to structure a lease. A retailer might have a ‘kicker’ in the lease agreement where more rent is paid if revenue exceeds a threshold. That might be good in the sense that less rent is paid in poor years, but kickers are likely to surprise many investors when they kick in.

There’s a strong suspicion that companies choose operating leases rather than capital leases so they can keep the lease obligations off the balance sheet, rather than for economic reasons. I haven’t found reliable and up-to-date sources for the current rules about lease accounting, because search engines find the proposals for change which businesses object to strongly.

Do deferred tax assets benefit income when they are used?

I believe the answer is “no” when you don’t include the “deferred income tax valuation allowance” as an asset (which it isn’t). I’ve found supporting evidence, but not definitive evidence.

According to “Corporate Tax Reform, Deferred Taxes, and the Immediate Effect on Book Profits” by Jana S. Raedy, Jeri Seidman, Douglas A. Shackelford, July 2011 (editorialexpress.com), if the value of deferred tax assets decreases, the company needs to make “an immediate charge to current period accounting earnings”. The problem with the quote is that the subject was the effect of a reduction in the maximum U.S. corporate income tax rate from 35% to 30%, not the normal use of tax assets. The reason for reducing deferred tax assets shouldn’t affect the accounting treatment, so when Utilization of NOL carryforwards drains the deferred tax assets, the “immediate charge” as above should still apply, canceling the benefit to Net income from the utilization. The benefit and immediate cancellation isn’t recorded in financial reports, at least it wasn’t for Orbital Sciences (next).

From Orbital Sciences Corporation’s 10-K for 2012:

    “Income Tax Provision – Our income tax provision was $30.8 million, $20.6 million and $17.6 million in 2012, 2011 and 2010, respectively. The effective tax rate for 2012, 2011 and 2010 was 34%, 23% and 27%, respectively.”

    “We utilized net operating loss carryforwards that substantially offset taxable income in 2010 through 2012. As a result, our cash payments for income taxes, which primarily related to alternative minimum taxes, were equal to approximately 3%, 3% and 4% of pretax income in 2012, 2011 and 2010, respectively.”

The point here is that the company paid tax at the effective rates 34%, 23% and 27%, so far as the income statement is concerned, but the cash payments for income taxes were only 3%, 3% and 4% of pretax income.

From “Orbital Sciences Management Discusses Q4 2013 Results – Earnings Call Transcript” Feb. 13, 2014 (seekingalpha.com)

    “Our full year 2013 GAAP effective tax rate was 34% while our full year 2013 cash tax rate was 10%, reflecting the utilization of NOL and tax credit carryforwards.”

The NOL carryforwards didn’t do much for the GAAP effective tax rate, even though there was a considerable benefit to the cash tax rate.

That’s the situation I expect Female Health are approaching now the valuation allowance is under $3 million and the utilization of NOL carryforwards will soon have to come out of the Net deferred tax assets.

When tax accounting blinds

When large reductions in corporation tax are proposed, there’s concern about weakening the balance sheets of companies holding tax assets. In my opinion, that’s based on a misunderstanding, or on the consequences of other people’s misunderstanding. Tax assets can’t be sold separately (they can come with an acquisition), and can’t be used when there are no taxable earnings. In addition, the assets have to be reversed into the valuation allowance when earnings are below the expectations the tax assets were based on. Therefore they are unlikely to be much use in hard times, which is a key function of a strong balance sheet. They aren’t unique in that respect, for instance a coal mine could have a large value on the balance sheet, and be worth very little if the price of coal dropped permanently below the production cost. In both cases the value of the asset depends on future earnings.

I’m not denying the importance of perceptions, and maybe companies make acquisitions or return cash to investors based on what they think is a strong balance sheet, and maybe the fiction is maintained by others sharing it, but I don’t intend to invest in such perceptions. A company which suffered a big loss due to a lower tax rate affecting the tax assets, will pay less tax in future, not more, and the benefit to income and free cash flow could last indefinitely if the tax rate isn’t raised, whereas the tax assets only benefit free cash flow, and will run out. If you want to see genuinely strong balance sheets, check my previous ‘cash3’ blog post.

Here’s a hypothetical case to make my point. Imagine a small company with:

    1) $30 million of NOLs (Net Operating Losses) which never expire
    2) $10 million income before income tax every year
    3) The company utilizes as much of the NOLs as it can, every year.

Suppose the tax rate starts at 40%. Each year, the tax would be $10 million * 0.4 = $4 million, but the tax payable is canceled by offsetting the $10 million income before tax with $10 million of the NOLs, which equates to $4 million utilization of NOL carryforwards. The NOLs will all be used in three years, with a benefit of $4 million each year, so the total benefit = 3 * $4 million = $12 million. Assuming the company has not made a sudden turnaround, year 1 starts with $12 million worth of Deferred tax assets, and nothing in the valuation allowance.

The tax actually paid is:

    Year 1 – 0
    Year 2 – 0
    Year 3 – 0
    Year 4 – $4 million
    Year 5 – $4 million
    Year 6 – $4 million
    Year 7 – $4 million
    Year 8 – $4 million

    8-year total of tax paid = $20 million

The first three years of zero tax will benefit Cash from operations, but not Net income (as I’ve explained previously, find “Conclusions about Female Health’s tax assets” above). For each year:

    Net income = Income before income taxes * (1 – the tax rate)
    = $10 million * (1 – 0.4)
    = $10 million * 0.6
    = $6 million

    8-year total of Net income = $48 million

Now suppose at the start of year 1, the tax rate is halved to 20% and stays there. Each year, the tax would be $10 million * 0.2 = $2 million, but this is canceled by offsetting the $10 million income before tax with $10 million of the NOLs, which equates to $2 million utilization of NOL carryforwards. The NOLs will all be used in three years, with a benefit of $2 million each year, so the total benefit = 3 * $2 million = $6 million. The value of Deferred tax assets immediately drops from $12 million to $6 million, a $6 million reduction in value, as a result of the tax cut. For year 1:

    Taxable income $10 million
    Statutory tax -$2 million
    Utilization of NOL carryforwards $2 million
    Deferred tax assets utilized -$2 million
    Decrease in value of Deferred tax assets due to lower tax rate -$6 million
    Net income $2 million

So Net income is only $2 million as a result of the tax cut, 66.7% less than the $6 million without the tax cut. The Deferred tax assets end the year at only $4 million, when they would have been worth $8 million without the tax cut.

The tax actually paid is:

    Year 1 – 0
    Year 2 – 0
    Year 3 – 0
    Year 4 – $2 million
    Year 5 – $2 million
    Year 6 – $2 million
    Year 7 – $2 million
    Year 8 – $2 million
    8-year total = $10 million

Each tax payment is halved, compared to the previous schedule.

With the new tax rate staying at 20%, Net income has one bad year, as explained, followed by better years, indefinitely, like this:

Net income following a tax cut from 40% to 20%:

    Year 1 – $2 million
    Year 2 – $8 million
    Year 3 – $8 million
    Year 4 – $8 million
    Year 5 – $8 million
    Year 6 – $8 million
    Year 7 – $8 million
    Year 8 – $8 million
    8-year total = $58 million

Without the tax cut, Net income stays at $6 million a year, and the 8-year total is only $48 million, but $6 million accrues earlier. I say “accrues” because it isn’t cash, and the $6 million can’t be invested to earn a return, so there’s no real advantage to accruing the $6 million earlier. There could be a relative advantage regarding perceptions of the company.

In this simple model, the fall in the tax assets was only $6 million, which is less than the taxable income. A real company could have a much bigger fall if it had much bigger tax assets, but I assumed a massive cut from 40% to 20% tax, and a smaller cut would have less effect on the value of tax assets. A company with large tax assets and with total assets only slightly above total liabilities before a tax cut, could have total assets less than total liabilities after the tax cut. It’s possible that a real effect is generated as a result of the illusion, if debt covenants are triggered or perceptions of credit risk are worsened.

I assumed that the tax cut is permanent. If the cut is temporary, the company can benefit in some circumstances, and is never worse off than if the cut never happened (not counting any macro effects). It takes too much space to give examples showing that a lower tax rate for longer can be better and is never worse, but it could be demonstrated by testing scenarios on a spreadsheet.

The value of an asset is either what it can be traded for, or the benefit it provides over it’s lifetime, and tax assets cannot be traded. A tax asset really does provide less benefit over its lifetime if the tax rate is cut, but the effect is more than made up for by the benefit arising from the lower statutory tax rate each year. There could be some relative effects, because companies without tax assets benefit more from a tax cut than companies with tax assets.

The benefit arising from lower statutory tax each year cannot be capitalized as an asset. That explains how a real improvement causes a year with lower Net income and a lower value for the tax asset. The real improvement is the lower tax actually paid in the future. Investors should be aware that financial reporting can blind as well as illuminate.

The valuation allowance, management’s predictions, and earnings management

Regulators and accounting standards don’t encourage management to share their forecasts with investors. The one exception is when allocating the Gross deferred tax assets between the Deferred income tax valuation allowance and Net deferred tax assets, and projections based on a 50% probability threshold are a GAAP requirement.

Predictive Ability of the Valuation Allowance for Deferred Tax Assets” by Jung, Do-Jin; Pulliam, Darlene, Academy of Accounting and Financial Studies (questia.com)

The article above doesn’t show the full text, but from what’s visible, the case is based on argument rather than data. IMO one snag is that when a lower allowance shows optimism, the tax benefit is pre-booked into Net income. If the projections turn out to be too optimistic, the tax benefit could be reversed as some of the tax asset is reversed back into the valuation allowance. Because of that, the best situation is probably the one I explain in “FHCO – Female Health’s deferred tax assets imply management’s forecasts have been conservative” below. While improving prospects combined with tax assets based on conservative projections are attractive, the quality of earnings is likely to be low, with Net income losing the tax benefit earlier and Cash from operations losing the tax benefit later.

The article linked to above points out that the required forecasts offer an opportunity for earnings management.

The paper “Looking at Accounting for Income Taxes: Do Managers Play “Truth or Dare” with Tax Accruals?” (PDF) by Carlos E. Jiménez-Angueira, University of Texas San Antonio

cites …

“Is Deferred Tax Expense Useful in Detecting Earnings Management in Earnings Restatements” by Badertscher, B., J. Phillips, M. P. K. Pincus, and S. O. Rego. 2006.

… in a sentence which implies that the answer to the title is “Yes.”, but it isn’t all that surprising if an earnings restatement throws up evidence of earnings management, especially if the restatement was required by the SEC or by auditors.

The google search “valuation allowance” “earnings management” finds plenty of academic papers, most of which are not free.

What a Company’s Tax Disclosures Reveal About Its Quality of Earnings” 7 Dec 2010 (taxand.com)

The last two paragraphs are about the valuation allowance. Investors are warned that the release of the valuation allowance causes an increase in earnings which is low quality.

Tax and the liability method

The liability method is a GAAP requirement when there’s a tax asset or liability. More technically, the Asset-Liability Method is a GAAP requirement when there’s a temporary difference between the period of financial reporting and the period in which the tax is payable.

Under the method, the balance sheet is seen as primary and the income statement as secondary, with the implication that the cost of keeping the balance sheet useful and accurate is that the income statement could be misleading to unwary investors.

The Gross deferred tax assets equal the value that would be derived from the Net Operating Loss carryforwards if they were all used. This is split between the Deferred income tax valuation allowance and the Net deferred tax assets. I’ve read that potential tax assets go into the valuation allowance if the chance of expiry before utilization is greater than 50%, but as Female Health Co look ten years ahead, I expect their deferred tax assets only include the benefit of NOLs likely to be used in the next ten years. The requirement to apply a 50% threshold rules out being deliberately conservative, but I don’t expect the SEC to take much interest in conservative estimates, see “FHCO – Female Health’s deferred tax assets imply management’s forecasts have been conservative” below.

The Asset-Liability Method of Interperiod Tax Allocation” By Maire Loughran from Intermediate Accounting For Dummies

Wikipedia’s “Tax expense” is less clear than Dummies.

See also “The valuation allowance, management’s predictions, and earnings management” above, for the thoughts of some academics on the subject generally.

Links for the U.S. corporate tax rate

The U.S. Corporate tax rate is high at about 35% but with deductions and complications. Companies that supply software or cloud-based services etc. are particularly well placed to pay little tax if they operate internationally. Companies have very different effective tax rates, and somewhere at the upper end, metrology company Nanometrics estimate a tax rate band up to 38% for 2014. It’s possible that revisions of open years (the last three years in the U.S.) could produce higher rates. Many companies minimize the repatriation of cash held overseas, to avoid paying withholding taxes.

Corporate tax in the United States” (Wikipedia)

A list of 2013 U.S. corporate federal tax rates. (en.planiguide.ca)

U.S. Corporate Tax Rate Fails to Move with Competition” By Ellen Kant, March 20, 2013 (taxfoundation.org)

This is 100% politics: “Update: U.S. Corporate Tax Reform Likely On Hold Until November 2014” by Joe Harpaz, 1/10/2014 (forbes.com)

Share price performance and fundamental analysis

Every tool of fundamental analysis might also predict share price performance, but their usefulness for prediction depends on enough investors ignoring them, both explicitly and regarding the quantities the tools are based on. Taking the Altman Z score, which indicates the risk of bankruptcy, as an example, suppose it is under-used, and so are any equivalent or superior alternatives. Then a mechanical share-picking system could outperform the market by using the score to avoid companies in the Altman Z ‘distress zone’. If the Altman Z score became fully appreciated, the companies flagged as being risky would become fairly priced, and the score would not predict risk-adjusted share price performance, although it could still be useful to risk-averse investors. At least, that’s one theory. Warren Buffett’s advice would be to avoid the risk of a permanent loss of capital, and risk-averse investors can still exercise their stock-picking skills within the stocks in the Altman Z safe zone.

There’s a study which went beyond the usual link between fundamental analysis and share price performance, by comparing markets in two countries, the United States and India. This links to the 12 page research abstract (a PDF):

An examination of future firm performance and fundamental analysis” by Chris Luchs, Suneel Maheshwari and Mark Myring

The study showed that predicting stock price performance from fundamentals in India needed different indicators to prediction in the United States. The conclusion is that investors priced-in different information. If the information priced-in can vary between countries, IMO it’s possible that the information priced-in can vary over time in the same country, and old studies relating fundamentals to stock price performance could be out of date.

Investors valued each dollar spent on R&D as being worth the same as a dollar of tangible capital, at the start of the 1980s, but the valuation of R&D capital fell to only 0.2 to 0.3 times the value of tangible capital by 1990. (See “The Value of Intangible Corporate Assets: An Empirical Study of the Components of Tobin’s Q.” by B. Hall, University of California, 1992.) It’s possible that companies’ R&D became much less effective, but it’s more likely that perceptions changed, and if investors were right at one end of the decade, they were wrong at the other end. A well informed investor could either have avoided companies with high R&D in 1980, or invested in companies with high R&D in 1990. According to “The market valuation of knowledge assets in US and European firms” by Czarnitzki, Hall & Orian, 2005 (berkeley.edu), one dollar of additional R&D spending adds slightly less than a dollar to market value, the equivalent of capitalizing the R&D spend and depreciating it by a bit more than 15% a year. I haven’t found a recent estimate. If investors’ valuation of R&D can drop 70% in a decade, other perceptions of fundamentals can drift. See also “The Altman Z Score improved with age but has critics”, below.

Old studies are not systematically re-tested or re-optimized. Academics might think it wouldn’t help their careers, but research into the durability of investment-related studies would be useful, original in its own way, and shouldn’t be too hard for anyone who understands the original work.

The incentives academics face might explain why they are keen to expand the theory in a piecemeal fashion, and have no scheme for regular updates, in stark contrast to the way in which antivirus software is updated (although there are a lot more new viruses than the restatements of accounts relevant to the M-Score, or bankruptcies relevant to the Z Score). Life could be different at the sharp end, and firms that charge for forensic accounting services might update their analysis tools as soon as new information allows it.

Academics write papers where theories or proposals are supported by evidence, often in the form of statistical tests. That’s good, but for fundamental analysis, investors need to know if the results still apply, and if the information has become priced-in. For forensic analysis, there’s a risk of over-weighting old abuses and leaving gaps for future abuse. The first rule of making a good fishing net is to make sure the fish can’t pass through. Academics could forget that objective by demanding evidence that every scrap of material used is necessary.

The M-Score was specifically designed to indicate the risk of a restatement of accounts required by auditors or the SEC. That under-represents any form of earnings manipulation with a low risk of being caught. There’s reason to suspect gains included in non-operating income when they make up for a drop in operating income (see “Beyond the scope of the M-Score” above), but the reporting is likely to comply with GAAP, so there’s little chance of a restatement of accounts being required. It might be possible to relate future share price performance to swings from operating income to non-operating income, and to design an index to warn of the low quality income, but forcing the index into the M-Score would create problems as there’s no clear way to optimize the coefficient of the new index. One possibility, now that the M-Score is touted as predicting share price performance, is to create a variant optimized for share price performance and earnings persistence, but still only using indexes likely to catch earnings manipulation.

Beneish’s 1999 paper improved on his work dated 1997, but before Beneish devised the M-Score, Lev and Thiagarajan wrote about twelve signals to measure earnings quality, in “Fundamental Information Analysis” by Baruch Lev and S. Ramu Thiagarajan, 1993. Much of the paper is fairly readable, but the data covered 1974 to 1988 and IMO the signals are due a re-test.

Some of Lev and Thiagarajan’s measures were similar to indexes in the M-Score, for instance signal 2 is:

    Growth in accounts receivable – Growth in sales

Compare that to the M-Score index DSRI (Days Sales in Receivables), which boils down to the growth of the ratio:

    Accounts receivable / Sales

although L&T based their growth numbers on growth above the average for two previous years.

Beneish rejected some of Lev and Thiagarajan’s measures of earnings quality when they didn’t improve the performance of the M-Score. The rejects involved changes in these – the receivable provision (for customers not expected to pay up), capital expenditure, tax expense / taxable income, and sales per employee. Another reject depended on whether or not inventory was valued using LIFO (Last In First Out). The signal based on a qualified audit report (meaning the auditors found something dubious) has nothing like it in the M-Score, but it wouldn’t have been useful to auditors, and might have displaced signals that anticipated a qualified audit report. L&T were interested in stock returns and the quality or persistence of earnings over one, two and three years, and their signals could not all be expected to improve the performance of a score aimed at earnings manipulation.

Lev and Thiagarajan recognized the ambiguity of popular indicators, including the fact that while unusually high inventory generally indicates a problem, the inventory could be held because higher sales are expected. A complicated model could look for signs like increased investment and no slump in sales, to see how to interpret the information about inventory, but L&T kept the model simple, or ‘parsimonious’, and the high inventory is just seen as bad in their formal scheme. The M-Score is also parsimonious, which is one reason why it’s worth checking that a warning is the result of a genuine problem. A non-parsimonious model (‘profligate’?) is likely to be more complicated and have more coefficients which need to be optimized for the data set. That in turn needs more data to be sure that the model gives statistically significant results. There probably aren’t enough restatements of accounts to enable a much more complicated version of the M-Score, limiting any benefit from a non-parsimonious version. It’s more feasible to have a non-parsimonious model when the model is optimized for share price performance, as there’s much more data.

The Altman Z Score improved with age but has critics

The Altman Z-Score: Is it possible to predict corporate bankruptcy using a formula?” Apr. 13, 2011 (businessinsider.com)

The title is answered with ‘Yes’, but be aware that the original version applied to manufacturing companies, and a modified version applies to non-manufacturers. There’s also a version for private companies which omits the term with market capitalization. The different scores need different interpretations. The accuracy of the Z Score for predicting bankruptcy was initially measured at 72% in 1968, but this increased to between 80% and 90% in the years to 1999. That could be due to the data accumulating, or a change in the bankruptcy risk.

This PDF is the source of the 80% to 90% accuracy figure: “PREDICTING FINANCIAL DISTRESS OF COMPANIES: REVISITING THE Z-SCORE AND ZETA® MODELS” by Edward I. Altman, July 2000 (pages.stern.nyu.edu). To be more exact, 80% to 90% of the companies which went bankrupt were flagged by the Z score. However, the ‘false positives’ flagged as a bankruptcy risk but which did not go bankrupt, grew over the period to 15-20% of the companies flagged. A company was flagged if its Z score was under 1.81.

In 2009, the Z-score was used with European companies to show that companies with weak balance sheets had poor share price performance from 1990 to 2008, with the worst performance in bear markets and recessions. While I’m all in favor of strong balance sheets, there will be times when markets turn from ‘risk off’ to ‘risk on’ and companies with weak balance sheets are likely to outperform, which may have been the case in the “dash for trash” rally after the credit-crunch. The companies assumed to have weak balance sheets were companies with a Z score less than one.

Suppose an academic tested an old score using data from 1995 to 2005. The academic could pick the threshold that gives the best result. Anyone investing during that period would not know that threshold. If there was a single best threshold that stayed constant, academics would not use different thresholds for studies of the same score. Academics could be choosing thresholds for good reasons, but the variation isn’t helpful for investors.

The Z score is affected by bull markets and bear markets, as the company’s stock price is in the score and is affected by the overall market. Market sentiment is also likely to affect the ability of distressed companies to raise capital.

Stop Using the Altman-Z Score.” by wes, July 23, 2011 (blog.empiricalfinancellc.com)

The blog-piece above is about a research paper called “Predicting Financial Distress and the Performance of Distressed Stocks” by John Y. Campbell, Jens Hilscher, and Jan Szilagyi. While the abstract, formulas and statistical tables are quoted, the style is informal. A new score with eight components is proposed. Six of the components depend on valuation by the market, through the market capitalization, the share price, and the share price volatility, so there isn’t much in the score that’s purely about fundamentals.

Forecasting Bankruptcy More Accurately: A Simple Hazard Model” by Tyler Shumway, 1999 (personal.umich.edu)

Shumway finds that about half of the accounting ratios used to predict bankruptcy are not statistically significant. My view is that having net cash and good cash flow is a reliable indication that bankruptcy is far from imminent, even if a weak balance sheet and poor cash flow does not reliably indicate bankruptcy. In the second case, signals from the market are important, and can be monitored more frequently than financial reports. So far as I know, the critics of the Z-Score have not proved that investing in companies in the Z Score’s safety zone is a bad strategy. The original formula for the Z-Score is:

    Altman Z Score =
    1.2 * Working capital / Total Assets
    + 1.4 * Retained earnings / Total Assets
    + 3.3 * Earnings Before Interest and Taxes / Total Assets
    + 0.6 * Market Value of Equity / Total Liabilities
    + 0.998 * Sales / Total Assets

    Distress zone < 1.81 < Gray zone < 2.99 < Safe zone

Shumway finds that only EBIT / Assets, and Market equity / Total liabilities are statistically significant. Altman naturally thought all his ratios were statistically significant, and Shumway's explanation is that the Altman Z Score suffers from a statistical bias because samples of healthy and bankrupt firms were not chosen randomly.

Working capital is Current assets – Current liabilities. A few websites write "market capitalization" instead of "Market Value of Equity".

For non-manufacturing companies:

    Altman Z Score =
    6.56 * Working capital / Total Assets
    + 3.26 * Retained earnings / Total Assets
    + 6.72 * Earnings Before Interest and Taxes / Total Assets
    + 1.05T4 * Market Value of Equity / Total Liabilities

    Distress zone < 1.22 < Gray zone < 2.9 < Safe zone

The coefficients are very different and asset turnover (Sales / Assets) is dropped. Altman might have been thinking of consultancies where compensation is high and capital costs are low, rather than capital intensive businesses like server farms and airlines. Physical assets can be leased, which brings up whether it’s how assets are financed that matters or how tangible the product is, when choosing the Z Score formula. Asset turnover can be very high for a service industry, or in any industry if the assets used are on the books at a low value.

Synchronoss Technologies, Inc (SNCR) provide a service to telecoms carriers, but have a low asset turnover, mostly due to the intangibles they carry. For 2013 (in thousands):

    Total assets $527,019
    Cash $63,512
    Non-cash assets $463,507
    Revenue $349,047

    asset turnover = $349,047 / $527,019
    = 0.66 times, or 66%
    turnover on Non-cash assets = $349,047 / $463,507
    = 0.75 times, or 75%

    Property and equipment, net $106,106
    Goodwill $137,743
    Intangible assets, net $101,963

Ubiquiti Networks, Inc. (UBNT), 2013 (in thousands):

    Total assets $292,340
    Cash $227,826
    Non-cash assets $64,514
    Revenue $320,823

    asset turnover = $320,823 / $292,340
    = 1.10 times, or 110%
    turnover on Non-cash assets = $320,823 / 64,514
    = 4.97 times, or 497%

Ubiquiti put their own logos on their own designs, and are probably classed as manufacturers of communications equipment, but their outsourcing allows high asset turnover, several times higher than service-provider Synchronoss. An online Z score calculator might not tell you much about the version of the Z Score it uses for a company.

I prefer to strip out the cash from Total assets for businesses with no need to turn the cash over. An online Z score calculator probably just throws in Total assets, and if it doesn’t, it might not tell you.

Some companies can operate with negative working capital, which is usually considered to be an advantage, but they are likely to get low Z Scores due to the high coefficient of 6.56 for working capital. Some companies could have high working capital as a result of high inventory, which could be due to low sales, while another company could have the same amount of working capital but made up mostly of cash.

"Bankruptcy Prediction with Industry Effects" by Sudheer Chava and Robert A. Jarrow, revised 2004 (wu-wien.ac.at)

The paper starts by using an expanded bankruptcy database to show that Shumway's 2001 model (above) beats Altman's (1968) and Zmijewski's (1984).

In "The Power of Cash Flow Ratios” by John R. Mills and Jeanne H. Yamamura, October 1998 (journalofaccountancy.com), the authors argued that auditors should catch up with Wall Street and look at cash flow ratios when deciding if a company is a going concern.

Altman considered using the cash flow to debt ratio in the Z score, but the cash flow data wasn’t always available in 1968. While the Altman Z Score is convenient and seems to have tested well, it still doesn’t look at cash. The cash ratios in the link above might be an improvement, but they look at aggregates and averages, and could miss the effect of a big debt repayment just beyond the one year range of current liabilities. A more thorough check is much less convenient: in a spreadsheet, make a schedule of all cash obligations (including debt repayment), and to start with, see how much headroom there is if cash from operations stays constant. You can leave out Operating lease obligations, and any other regular payments that are an expense deducted from Net income and therefore from Cash from operations. Check “Contractual obligations”, and look under “Financing activities” for anything that looks like an unavoidable regular payment, like “Payments on capital lease obligations”. The schedule of debt repayment is known as the ‘debt profile’. I prefer to include any payments that aren’t already included in Cash from operations.

Growth, bloat, decline and recovery

When a company has high sales growth, falling gross margin, or increasing SG&A expense as a proportion of sales, it might be worth reading this section to see if any of it applies to the company. I hope the cases I’ve cited are interesting, although an editor would delete the entire section for being too miscellaneous. Skip to “The Dodd-Frank Act – clawing back performance awards” if you like.


There are many reasons why a company’s growth could slow, including – a change in fashion, product obsolescence, market saturation, competition, scarce resources within the company, scarce resources in the market, getting too big to grow fast, recession, government policy, labor disputes, poor management, and increasing bureaucracy if management don’t address the organizational issues that size brings. The ‘scarce resources’ items include everything from skilled personnel to finance. To varying degrees, these are areas which investors can research and make a judgment about before investing, and consider when thinking about a warning from SGI.

If market saturation is reached, it’s worse for durable goods where demand could fall to replacement levels. In a recession, consumers will delay buying cars but not cut down much on bread, and car-makers will still pay for electricity but will buy much less welding equipment. There are details and exceptions, for instance some capital spending could be driven by regulation aimed at reducing pollution, rather than by the state of the economy.

There are some general differences between consumer and business markets, with businesses more likely to be conservative about buying new products, and to look at the specification. Professional buyers or purchasing managers can feel they’re punished for mistakes more than they’re rewarded for buying a better product. A business customer might insist on trials and then regard them as their least urgent activity. Safety-critical products from drugs to airplanes can take a long time to prove, and sometimes the technology looks dated by the time it’s approved. Small companies trying to grow can depend on relatively large contracts with a few customers, and a single delayed or canceled contract can have a big effect on results.

Consumers can also be conservative, for example many won’t change their preferred cola or sports team, but they also buy new products with a cool image, which isn’t usually relevant when choosing the office photocopier. Sales are likely to be at least maintained, for services and ‘consumables’ where the customers are conservative, meaning less risk from a high SGI.

Early in this interview, a value-oriented professional investor describes how he liked the business model of used car dealers CarMax, but waited until they tried to expand too fast and hit problems, before buying shares at much less than their IPO price: “Markel Manager Tom Gayner Interview with GuruFocus in Omaha” May 08, 2013 (gurufocus.com).

Krispy Kreme Doughnuts, Inc had EPS growth of 80%, 67%, and 47% following their IPO in April 2000. The growth figures were fairly synthetic due to earnings manipulation, but it’s not surprising that high growth is followed by such a decline. Investors might regard the figures as declining compound growth, but I’ve calculated from the percentages that after the 80% growth, a model of growth based on simple interest of 80% rather than fixed 80% compound interest is 13.5% more accurate, and investors who thought about increasing simple-interest based growth would need to factor-in less adjustment than investors who thought about decreasing compound-interest based growth. (Simple interest just takes the absolute ‘dollars and cents’ rise in EPS that occurred in the first year, and grows EPS by that amount every later year.)

I also looked at extrapolations for the two growth schemes, for Female Health Co, based on the growth from 2006 to 2007, the first two years with positive income. The simple growth or linear extrapolation was surprisingly accurate, but this was a fluke as using the growth between other pairs of years was inaccurate. The percentage growth between 2006 and 2007 was a very high 502.81%, which extrapolated to a ridiculous EPS of $3,078 in 2013 using compound growth, compared to the actual EPS of $0.4993. The percentage growth in EPS over the previous year, from 2007 to 2013, was: 502.81%, 176.61%, 32.42%, 0.42%, -21.04%, 183.74%, -5.58%.

Krispy fiddled the figures, and Female Health have lumpy orders, but I’d still suggest not being over-impressed by high reported growth (or by projections beyond the next year, usually). Instead, ask if there’s an economic moat (next bit) and if the price is right, and consider other factors including the most relevant ones in this section. When companies are floated, investors pay for growth. The amount of cash raised is obviously important when the owners time an IPO, and it’s worth asking why they haven’t waited a year or two.

Economic moat

An economic moat is a sustainable competitive advantage, which could be the result of a durable brand, excellent management, being the biggest when there are economies of scale, the business model, innovation, or a network effect. The moat gives a company pricing power. Growth might not be fast, but it can be maintained with a good margin.

I have alternative definitions –

1) An illusion held by investors who lack the imagination to see how success can be undermined. A town’s local newspaper used to benefit from the network effect for local advertising, because the paper was where people looked to find items for sale locally, so that’s where they were advertised, so that’s where you had to look, etc. That and other services have been undermined by the internet. The internet might have looked very unthreatening in the early days and after the Dotcom-crash. Moats or apparent moats can be destroyed by innovation, a radical change in consumer preferences, incompetent management, and government policy including breaking up monopolies. Before deciding a moat is safe, learn or think about disruptive innovation. Even geography can change, and the Suez Canal is likely to lose some business if the Arctic thaws enough.

2) An advantage to be squandered by management. GEICO is a classic case from the value-investing literature. They were insurers who only insured government workers, who generally had low accident rates. Targeting government workers meant GEICO didn’t need the expensive advertising required by other insurers. Value investors saw the value of the model, bought in, and exited with a big profit. Then GEICO’s management expanded the business until it became more like other insurers, and got into trouble. Warren Buffett decided the new CEO was competent and working to put GEICO back on the right track, and GEICO became part of Buffett’s Berkshire Hathaway.

5 Ways to Identify Wide Economic Moats” by Daniel Sparks, August 20th, 2012 (oldschoolvalue.com).

The first of the five ways gives Apple Inc as an example, citing fanatical loyalty among other good points. Blackberry also had fans. Right now Google’s Android seems to have a moat, but I’m not committing myself. IMO it’s usually hard to be sure that an advantage is permanent enough to be a moat, especially in high-tech markets.

WD-40 is a spray for lubricating metal parts and preventing rust, and is a product with an enduring economic moat. The brand is known and trusted. Because use is infrequent, it isn’t worth customers trying alternatives, often on items that would be expensive to replace or repair if the alternative caused damage. Because most customers want WD-40 and turnover is low, it often isn’t worth small outlets stocking alternatives. The company even developed a superior product but couldn’t sell much of it, because customers preferred the WD-40 they knew and trusted. Because growth isn’t fast and the market isn’t enormous, it isn’t worth competitors spending a large amount on promotion in an attempt to grab market share, and there are easier markets to target.

Warren Buffett wants long term performance, so this is relevant: “This is How Buffett Interprets Financial Statements” by Jae Jun, May 21st, 2012

In “1995 Berkshire Hathaway letter to shareholders” (already linked to), Warren Buffet said:

    ‘In business, I look for economic castles protected by unbreachable “moats.” Thanks to Tony and his management team, GEICO’s moat widened in 1995.’

GEICO’s moat derived from it’s business model. Strong brands are frequently cited as supporting economic moats, and some of the writing about economic moats could leave you thinking that brand recognition is enough.

Daimler sees profit rise as new models catch on” by Edward Taylor, Feb 6, 2014 (uk.reuters.com)

From the article above, luxury car makers Mercedes-Benz, BMW and Audi all had operating margins under 10%. That’s not a high margin, given the investment required, although volumes are high. In 2012, margin targets were scrapped at Mercedes-Benz after a series of disappointing results. Mercedes-Benz is a well known and respected name in the luxury car market (and the article “Stop Using the Altman-Z Score.” above, has a picture of a gold-plated Mercedes). Whether you say the brand is strong due to recognition and appeal, or the brand is weak because margins can be low, indicating little pricing power, it’s a mistake to assume that brand recognition and appeal amount to an economic moat. Customers who want a luxury car from a recognized luxury brand still have a choice of brand, and can demand value, even if the value is superficially different to the value of keenly-priced own-brand groceries.

Many publications have struggled to find a way to monetize their once-mighty brand online.

This article is about a book from 1967 where the concept of the economic moat was understood well before Warren Buffet mentioned “economic castles protected by unbreachable moats”: “Economic Moat: A Look Back in Time” by David Foulke, Oct 21 2013 (turnkeyanalyst.com). One old moat example is Polaroid. Anyone could copy the expired patents, and the book-author stressed qualities such as talent rather than technology. Polaroid were floored by digital technology a few decades later. Another example is Avon Cosmetics, who’s moat is the unique and hard-to-copy “army of women” who sold the cosmetics door-to-door. That may be a good moat, but door-to-door selling is not invulnerable. In the U.K., the long-forgotten ‘man from the Pru‘ (‘Pru’ was short for ‘Prudential’) collected regular payments for insurance policies, savings plans etc., and used the visits to make new sales. That might have gone out even before the internet became popular. The book-author (who uses the name “Adam Smith”) emphasized the importance of price. Buying at a high valuation requires high growth or above average growth for many years to be sure of a good return, and some moats won’t last long enough. (BTW if “army of women” sounds odd these days, there was worse in the 1967 book.)

Since I mentioned the U.K., grocery retail stocks plunged recently as one of the smaller grocers decided they had to compete on price with the discount retailers, who have 7.5% of the market. Insurance companies that sell annuities have been hit by deregulation. Previously, there weren’t many options for a personal pension except to buy an annuity, and the providers have been accused of selling ‘rip-off annuities’ to naive customers. Consumers were exploited in both cases, for decades, but not forever.

You can find your own profound quote about change, to remember every time you hear about a company with a moat.


Big companies can suffer from accumulated bloat, which strangles profitability, innovation, the ability to react, and the kind of growth you want. (I probably shouldn’t write about bloat in a piece this long 🙂 )

Streamlining spans and layers” March 16, 2010 (bain.com). Bain imply that AT&T had 14 layers of management before drastically cutting them. Former Hewlett Packard CEO Mark Hurd found that the cause of many problems was 11 layers of sales management, which he cut to 8.

Can A Big Company Move Fast?” April 25, 2013 (linkedin.com). IBM’s management realize there’s a problem but fall short of saying the company needs to streamline and cut time-consuming procedures. The CEO blamed sales staff for poor sales, and supports a drive to have customer questions answered in 24 hours.

How IBM bypasses bureaucratic purgatory” By Anne Fisher, December 4, 2013 (cnn.com). IBM employees can use an internal crowdfunding system called iFundIT to bypass the company’s bureaucracy. A good initiative, but it would be better to trim the bureaucracy than to find ways to bypass it.

After reading the next link, the CEO’s response to losing sales seems superficial. As well as bureaucracy, the determination to hit an EPS target seems to have undermined IBM’s service to clients, giving clients exactly what they specify, which usually isn’t what they need. There are already overseas companies that can supply to specification competently, and IBM won’t be able to charge a premium in that area indefinitely if they provide nothing extra.

The massive recent project failures at IBM strike at the heart of why outsourcing fails” by Andrew Pollack, 08/08/2013 (thenorth.com)

I expect management will be well rewarded when the EPS target is hit, but you don’t clear a roadblock by putting a pot of money further down the road.

IBM can still make a good margin on their own software, especially if they can successfully commercialize Watson (the knowledge system that won a game of Jeopardy on TV).

Bloat is not only about layers of management. When CEO Bruno Cercley returned to rescue ski manufacturers Skis Rossignol, they had about 400 different products. No-one understood the rational for the product range, which confused customers, retailers and the company’s own employees. The range was quickly reduced to 200. Rossignol’s crisis resulted from warm winters, skiers preferring to rent skis instead of buying their own, and high sales in Japan slumping when customers switched to virtual skiing on screens. The crisis was painful but the CEO has said it was good for the company. The recovery depended on suppliers extending credit, which can’t always be guaranteed, and it’s obviously better to cut the rot before growth stops or markets turn.

Rossignol had enjoyed decades of growth through innovation, Olympic winners using their equipment, product diversification, and the penetration of new markets. The company produced the world’s first all-metal skis, which were used by the winner of the Olympic downhill event in 1960. Fiberglass skis and plastic skis followed. Rossignol began selling in the U.S. in the 1970s, and in the 1988 Winter Olympics, most of the wins were with Rossignol equipment. Product diversification got going in the 1990s with ski boots, bindings, snowboards and clothes. When about two million Japanese took up skiing, the business would have looked like a successful company dominating an attractive niche with valuable brands, and therefore able to grow within a moat. They were bought for $560 million in 2005, and sold for $147 million in 2008 to the former CEO. While Rossignol had decades of success, they seem to have lacked a coherent management system or a business model that could have helped to keep them on track as markets, management and ownership changed.

It’s partly because of such cases that I prefer companies with plenty of cash and good free cash flow, to companies which only have the cash flow, many of which have net debt. While it’s worth learning what you can about a company’s prospects, the efficiency of the organization, and the rationality of the product range, checking the net cash position and the history of free cash flow is easier and more objective. (I wrote about the benefits and risks of cash piles in my previous blog piece.)

Jack Welch was CEO of General Electric from 1981 to 2001. While highly effective, he was noted for ‘brutal candor’, and fired the worst performing 10% of managers each year. IMO that would not be good for all companies, but he solved GE’s bloat problem. GE later faced a serious problem through its financial arm, when bad debt emerged during the financial crisis. In this old interview, Welch describes how he cut out layers of management and time-wasting procedures, and generally made GE leaner and fitter: “Speed, Simplicity, Self-Confidence: An Interview with Jack Welch” by Noel Tichy and Ram Charan, September 1989 (hbr.org).

Jack Welch saw business as essentially simple but burdened with the complications that people impose on it, such as complicated hierarchies and waiting for long and complex reports to be completed. The interview above has cases where deals were completed quickly, but this was only possible because managers were trusted, well informed, and knew what was expected of them. I see some similarity with Warren Buffett’s approach, find “He called me after the match and in five minutes I basically had a deal.” in this 1998 lecture by Warren Buffett (PDF). Buffett already knew the business (shoes), I think the key word in the quote is ‘basically’, and the accounts would have been examined for a lot more than five minutes. He also wanted a quick decision when he offered to bail-out Long-Term Capital Management: “Warren Buffett gave Meriwether less than one hour to accept the deal;” (Wikipedia).

Whether or not business ought to be simple, some businesses aren’t, and Warren Buffett famously puts them on the ‘too hard’ pile if he doesn’t understand them.

This shows simplicity in action, for real estate. “Buffett’s annual letter: What you can learn from my real estate investments” February 24, 2014 (fortune.cnn.com).

Speed and decisiveness are not at the expense of prudence. From the 2013 Berkshire Hathaway letter to shareholders:

    “Berkshire Hathaway Reinsurance Group, managed by Ajit Jain. Ajit insures risks that no one else has the desire or the capital to take on. His operation combines capacity, speed, decisiveness and, most important, brains in a manner unique in the insurance business. Yet he never exposes Berkshire to risks that are inappropriate in relation to our resources. Indeed, we are far more conservative in avoiding risk than most large insurers. For example, if the insurance industry should experience a $250 billion loss from some megacatastrophe – a loss about triple anything it has ever experienced – Berkshire as a whole would likely record a significant profit for the year because of its many streams of earnings. And we would remain awash in cash, looking for large opportunities if the catastrophe caused markets to go into shock. All other major insurers and reinsurers would meanwhile be far in the red, with some facing insolvency.”

While it should be possible to grasp a business or investment case quickly, I don’t think anyone is suggesting that due diligence can be skipped, that 90% of a 10-K filing can be safely ignored, or forensic analysis is of no use.

Buffett also knows quickly when a deal doesn’t interest him, although with a large amount of capital to invest, he’s likely to have received too many investment proposals to be able to consider each one for very long.

It’s worth asking if a successful company functions through teamwork, or if it’s driven by a “genius with a thousand helpers”. The phrase is probably from the book “Good to Great: Why Some Companies Make the Leap…and Others Don’t” by Jim Collins, HarperCollins Publishers, 2001. This links to a short PDF by LtCol B.B. McBreen, with notes on the ‘Good to Great’ book.

The single genius is good (until the genius leaves), but a great company needs teamwork. There’s a lot more in the book than the popular phrase. Jim Collins studied companies that had outperformed their competitors by enough to be judged as changing from ‘good’ to ‘great’. Among the common features was a diligent, low-profile CEO, who got on with the transformation without even giving it a name.

Going from great to abysmal, you’ll have heard about this: “57-cent part would have fixed General Motors ignition switch linked to 13 deaths” by Associated Press, April 01, 2014 (lehighvalleylive.com). It looks like mindless cost-cutting. I hope the investigation throws some light on how it happened.

Companies can become inefficient through size, time, relatively easy profit, and complacency when they are the market leader. A company that survives a crisis can emerge fitter and better oriented to the market. A similar principle can extend to individuals: ‘1930: The year when Benjamin Graham re-discovered “Margin of safety”‘ Monday, July 19, 2010 (catalign.in), is about how Benjamin Graham, known as the father of value investing, had early success, but had to rethink his ideas about his lifestyle and investment after the crash of 1929. After the crash, Graham gave up his ten room apartment. I was sure he had also employed a butler, but trying to confirm that only shows that many people called Butler have written about Graham, too many to exclude in a search. The linked piece includes a good story about an old man who told Graham to sell all his stocks.

I believe that companies that stick to a pervasive discipline are less susceptible to bloat. Illinois Tool Works (ITW) are a conglomerate that have used ’80-20’ for many years. This is the “Pareto principle” (Wikipedia), which is often stated as “80% of the effects come from 20% of the causes”. The percentages don’t always add to 100%, but in management it’s good if 80% of profit comes from only 10% of the customers, because a powerful effect can be achieved by addressing the needs of only 10% of the customers.

If the customers not in the ‘80% of profit’ category aren’t happy, ITW are prepared to lose them. The process can’t be taken too literally or too far, because if they actually lost half the secondary customers, that could be a large percentage of sales disappearing every year, and companies in the ‘80% of profit’ category would shift out of it, until the principle no longer applied, which might leave very few customers. A similar argument applies to losing the worst-performing subsidiaries. I’ve read claims that Microsoft operated like that, losing the lowest-performing employees, leading to employees seeing each other as rivals rather than colleagues to cooperate with, and (allegedly) rivalry and a silo mentality within the company.

ITW are also nearly half way through a restructuring aimed at having fewer, bigger subsidiaries, with high and sustainable margins. While they are centralizing some sourcing, ITW are preserving their decentralized decision making, which I believe is possible because using ’80-20′ is a discipline applied at every level of the business. Crucially, ITW’s restructuring was not forced by a crisis. Other companies might benefit from other systems, such as lean manufacturing.

An integrated IT system could help to keep the parts of a company working together, but I’m not sure if that would give the same focus as ITW’s ’80-20′. You could try “What is SAP (System Applications Products)” (saponlinetutorials.com), about connecting modules for Sales and distribution, Human resources, etc. Big Data company Tibco would say their real-time systems are what’s needed. They recently announced an EPS target aimed at improving sales, which sounds horribly IBM-like.

IMO an excellent management system allows decentralization by referring to the system instead of going up the hierarchy. There’s no perfect system with all the benefits of decentralization without the risks.

Bloat can also be avoided through a lean business model. Ubiquiti Networks have a low headcount relative to their sales and income, because they outsource manufacturing, distribution and sales, and concentrate on research and development. Their internet community plays a key role in support, guiding the development of new products, and ‘word of mouth’ recommendation. One benefit of concentrating a high level of talent in the R&D team rather than increasing the size, is that it’s easier for everyone to keep in touch. Ubiquiti’s headcount is so low that a normal acquisition would be a problem, due to the massive increase in staff for any reasonably sized acquisition, though companies that haven’t gone beyond the R&D phase would be less of a problem. Successful outsourcing depends on managing it well, and in Ubiquiti’s case the critical areas include counterfeiting by contract manufacturers (where they had a problem previously) and maintaining quality. The model isn’t appropriate for every company, and IPG Photonics benefit from their vertical integration.

I heard about Skis Rossignol on BBC radio, and they don’t keep programs online for long so I haven’t given a link. The U.K.’s BBC has “layer on layer” of management, but no-one seems to know how many layers there are. They spent £98 million ($163 million) on their Digital Media Initiative, which they axed when they decided the money was wasted, and archived material has to be physically transported between locations as a result. See “BBC bosses ‘did not have a grip’ on doomed IT project which wasted £98million” by sao paulo, January 29th, 2014 (tvlicenceresistance.info). The BBC sent 437 employees to China for the Beijing Olympics, outnumbering British competitors. See also “Does the BBC really need a team of 175 to cover the U.S. presidential election?” November 04, 2008 (tvlicenceresistance.info). The BBC are also widely accused of a cover-up, search for “Jimmy Savile” if you’re interested, but it isn’t pleasant. It seems to be unclear who is responsible for what, when things go wrong, and previous ‘Director General’s (top management) who had to appear before parliament were keen to avoid blame.

This is on forbes.com and fairly advert-heavy: “A Bug’s Life: What Managers Can Learn From Ants” by Rick Wartzman, 5/07/2013 (forbes.com). The article relates ant behavior to decentralization, the division of labor, and the views of Peter Drucker (Wikipedia), who studied and wrote about organizations. Drucker popularized the idea of the “knowledge economy” in 1969, crediting the phrase to economist Fritz Machlup.

I haven’t read much of Drucker’s work as there’s too much vision and anecdote for my liking. The author of this book review is more enthusiastic: “The Practical Drucker” John Pearson’s Buckets Blog, January 18, 2014 (urgentink.typepad.com)

British civil servant and naval historian Cyril Northcote Parkinson (Wikipedia) wrote books and essays about bureaucracy. He successfully predicted in 1958 that the Royal Navy would eventually have more admirals than ships. According to the Daily Mail online newspaper, the Royal Navy had 40 admirals, 260 captains, and 19 ships in October 2013. There could be some blurring due to captains of smaller vessels that support 19 warships.

In World War II, the U.S. navy had one admiral for every 30 ships. See “Enlisted sailors forced out while Navy has more admirals than ships” by Mike Mather, May 2, 2013 (wtkr.com), for the number of generals and admirals, with entourages.

Parkinson’s observations include “Parkinson’s law” (Wikipedia), “work expands so as to fill the time available for its completion”. Jack Welch wanted managers to be short of time so they had to become business leaders rather than managers, and he didn’t want them to have time to defend their fiefdoms or engage in office politics.

In my opinion, some middle-managers can have valuable experience, but using them as a link in a long chain is not the best use.

The problem with government bureaucracies might be that they don’t have the “streamline or die” pressures that businesses occasionally face, and they lack the focus which making a profit brings. Bureaucratic units are more likely to be closed down for political reasons than because they are wasteful, and they illustrate just how inefficient organizations can become. Here’s some good news: “New York Department of State Awards Local Government Performance and Efficiency Program Awards” June 19, 2013 (dos.ny.gov). The profit motive isn’t guaranteed to produce perfect results, as General Motors’ lethal penny-pinching shows.

The Dodd-Frank Act – clawing back performance awards

Earnings manipulation could become less popular due to legislation. The Dodd-Frank Act of 2010 requires companies to have a “clawback” clause, so if a company has to restate its financial results, performance awards are recalculated, based on the restated results, and excess awards are clawed back. This applies to performance awards made in the previous three years, and with no need to prove intention, which is one way in which Dodd-Frank is tighter than the Sarbanes-Oxley Act of 2002.

Companies are still waiting for guidance from the SEC, and while Female Health have a “clawback” clause, it requires intention: “.. in the event of fraud or intentional misconduct that materially contributes to a restatement of financial results …” and the company acknowledge the clause might have to be changed to comply with Dodd-Frank. In my opinion the company would be demonstrating a higher standard of governance if they didn’t wait for the SEC to push them. Also IMO, management have little risk of financial loss by attempting earnings manipulation, and clawing back 150% of the difference instead of 100% would help to fix that, even though dishonest CEOs would still benefit from asymmetric risk. It won’t happen.

Clawback” (Wikipedia)

Executives: beware of Dodd-Frank compensation clawbacks” by Steven Salky, February 28 2013 (lexology.com)

Revenue Enhancement

The SEC’s New Financial Fraud Task Force: Part III, Cases Following the Speech – Revenue Enhancement” by Thomas O. Gorman, 08-19-2013 (lexisnexis.com) (already linked to above)

The piece lists some cases from 1994 to 2012. Cases are grouped into these categories:

    Falsification of revenue
    Sham transactions
    Channel stuffing
    Premature recognition (of revenue)

Well known companies featured:

    Time Warner Inc
    American International Group (AIG)
    General Re Corporation
    Bristol-Myers Squibb Company
    Lucent Technologies Inc.
    Raytheon Company

The site is a source for other forms of mis-statement.

More with less

Detecting Financial Fraud” By Colleen Kearney Rich, March 29, 2013

When a company reports financial growth, such as in revenue, earnings per share etc., but physical contraction, as in unit sales, the number of employees or how much floorspace is used for production, the explanation could be increased efficiency, or fraudulent financial numbers. An experiment at North Carolina State University showed that when investors were shown the physical information in a normal text format, they were more likely to assume the reason is increased efficiency, but if the information was tabulated to be consistent with the accounts (like the income statement or the balance sheet), investors were more likely to believe the reason was fraud.

When a company reports financial growth with physical contraction, investors need to weigh the likelihood of inefficiency that could be reduced, the will and ability to reduce it, the incentive to manipulate earnings and how well management might resist the incentives. The usual checks such as looking at cash and reading the ‘Liquidity and Capital Resources’ section of a 10-K, could help.

Who commits fraud?

Most Likely to Commit Fraud: Finance Executives” by Marie Leone August 19, 2011 (CFO.com)

Don’t expect surprises.

Cheat sheet

Forensic Accounting For Dummies” by Frimette Kass-Shraibman and Vijay S. Sampath (dummies.com)

13 signs are listed under “Spotting Business Financial Statement Fraud”.

Free and fantastic

Financial Shenanigans” by Howard Schilit and Jeremy Perler
Fully revised and updated third edition, with copyright dated 2010.

The free PDF book covers the manipulation of earnings, cash flows, and key metrics. The authors Howard M. Schilit, Ph.D., CPA, and Jeremy Perler, CFA, CPA, are evidently qualified, and their careers include forensic accounting. What you might not expect is that they are excellent communicators, and write in a style a complete world away from anything in the PDFs I’ve linked to which were written for academics. As usual for forensic analysis, they focus on nailing the bad guys. The authors operate at the sharp end of forensic accounting, and I’m not holding my breath waiting for academics to publish anything as intelligible and make it free.

The terms limit readers to keeping a single copy, but I can’t quote the terms of use without breaking them.

Neutral scores for no change or steady growth

This part and the next two are purely technical. Skip them if you like because they aren’t going to make you rich. You might want to check “that doesn’t read like a shocking headline, but”.

I’ve calculated what I call the neutral M-Score which is the result when there is no change from one year to the next. It’s a simple benchmark and academics are unlikely to approve. There’s some complication from the TATA index, and as a result I’ve calculated two neutral M-Scores which should bracket most likely cases, although ‘likely’ is relative as neutral M-Scores are only intended as a theoretical benchmark. All the indexes except TATA equal 1 when there is no change, for example:

    SGI = this year’s sales / the previous year’s sales = 1

… and when calculating the term for SGI in the M-Score:

    term = coefficient * index
    = 0.892 * 1
    = 0.892

and the term will always equal the coefficient when the index is for one year divided by the other for a quantity or ratio that hasn’t changed. This means that one approximation of the neutral or ‘no change’ M-Score is just the sum of the constant and the coefficient of every index except TATA.

    Approximate neutral ‘no change’ M-Score = -4.84 + 0.92 + 0.528 + 0.404 + 0.892 + 0.115 – 0.172 – 0.327
    = -2.480

When there is no change, there is no increase in non-cash working capital, so the increase drops out of TATA leaving:

    ‘no change’ TATA = -Depreciation / Total assets

I’ve only calculated that ratio for four different companies, so my results are fairly rough-and-ready, and likely to be biased, for example they all hold a decent amount of cash. I haven’t been consistent about the units (in thousands or not) between companies as I only needed a sample of the contributions to the M-Score. 4.679 is the coefficient for TATA.

    Company, Depreciation / Total assets = ratio, * coefficient = contribution to the M-Score
    Female Health Co, -556,304 / 35,169,953 = -0.0158, * 4.679 = -0.074
    Lattice Semiconductor, -20,807 / 447,876 = -0.0465, * 4.679 = -0.217
    Repligen Corporation, -3,508,592 / 97,010,163 = -0.0361, * 4.679 = -0.169
    Synchronoss Technologies, -41,126 / 527,019 = -0.0780, * 4.679 = -0.365

There’s quite a range, with the most negative being -0.365 from Synchronoss Technologies.

I get a low version of the ‘no change’ M-Score by adding the -0.365 to the previous approximation:

    Low ‘no change’ M-Score = -2.480 + -0.365 = -2.845

That score is based on the 7.8% Depreciation / Total assets calculated for Synchronoss Technologies, the highest ratio out of the four I calculated. The result and the previous approximation should bracket most of the likely ‘no change’ M-Scores, in other words the ‘no change’ scores should be between -2.845 and -2.480.

It’s reasonable to ask: is there a neutral M-Score for a company growing at a steady rate, with everything growing in proportion? There is, but it isn’t simple. Only two indexes are systematically affected by such steady growth: SGI, the Sales Growth Index, and TATA, Total Accruals to Total Assets. Steady growth is not a reason for days sales in receivables to change, for the gross margin to change, etc..

If a company grew steadily at 8% per year:

    SGI = this year’s revenue / the previous year’s revenue
    = the previous year’s revenue * 1.08 / the previous year’s revenue
    = 1.08 / 1
    = 1.08

and the effect on the M-Score = coefficient * change in index

    = 0.892 * (1.08 – 1)
    = 0.892 * 0.08
    = 0.071

More generally, the effect through SGI = 0.892 * the growth rate, where a growth rate of 5% (for example) translates to the number 0.05.

A company growing at a steady rate, with everything growing in proportion, gets a complicated term for TATA. I’ll leave out the worst of the arithmetic, and write:

    TATA = growth factor * Proportion of non-cash working capital – Proportion of depreciation

    where growth factor = r / (1 + r)
    r is the growth rate (r = 0.05 means growth of 5% pa)
    and both ‘Proportions’ are relative to Total assets

The ‘growth factor’ scales back high growth, so 10% (which is 0.1) is scaled back slightly to 9.0909%, and 100% is scaled back to 50%.

The formula makes a kind of sense, because a fast growing company with a high proportion of non-cash working capital needs more cash to fund its growth. However, that’s concerned with the pressure to manipulate. TATA is supposed to measure an area of balance sheet distortion, and by the definition I’m using, ‘A company growing at a steady rate, with everything growing in proportion’, growth in proportion is not evidence of distortion (although it could mask distortion, for example if 50% of receivables were always fake). While the formula I’ve given is accurate regarding the value calculated for TATA, the result which TATA ought to give is the same as for no change. Therefore, in some circumstances, it’s reasonable to ignore the effect of steady growth on TATA and say:

    The neutral steady-growth M-Score is in between:
    0.892 * growth rate – 2.845
    0.892 * growth rate – 2.480

where I’ve used the previous results for ‘no change’, of -2.845 and -2.480, which should bracket most of the likely range. A growth rate of 5% is represented by 0.05, for example.

I’ve been using my own terminology, and I haven’t seen Beneish or anyone else writing about neutral M-Scores.

If you think I stretched the meaning of “simple”, bear in mind that academics write about alphas from Fama regressions when they aren’t hanging subscripts off Greek letters.

M-Score thresholds, probabilities, and sensitivity to sales growth

I said that an M-Score above -2.22 is commonly supposed on various websites to indicate a serious risk that a company’s earnings have been manipulated. I haven’t seen Beneish or any academics even mention the threshold, but I could have missed it.

30% sales growth gives SGI = 1.3, which is 0.3 above the ‘no change’ value of 1. Multiplying by the coefficient of SGI:

    extra on M-Score = 0.892 * 0.3 = 0.2676

Adding that onto the high ‘no change’ M-Score of -2.480 (which I calculated in the previous section) gives:

    30% sales growth M-Score = -2.480 + 0.2676
    = -2.2124

which is above the threshold of -2.22. That means, 30% sales growth is enough to trigger a warning if the score isn’t pulled down by the other indexes, except that -2.480 is too high to be a reasonable ‘no change’ M-Score. I also calculated a ‘too low’ value, of -2.845, so cases could be bracketed. Using that value, 70% sales growth is enough to trigger a warning if the score isn’t pulled down by the other indexes. Putting that into something like English:

    The sales growth required to trigger an M-Score warning based on the -2.22 threshold is probably between 30% and 70%, depending on the company. The growth rate is enough to trigger the warning if the other indexes don’t collectively pull the M-Score down, relative to the effect if they all represented no change, with every index except TATA equal to one.

Some readers might be thinking that growth of 30% to 70% is very high, but my reaction is that the -2.22 threshold is quite sensitive to sales growth, since high sales growth is usually welcomed and nothing else is required to trigger the warning. For some mature companies, 30% to 70% sales growth is unfeasible and could arouse suspicion. If that’s a flaw in my analysis, it’s common to other measures, and the M-Score wouldn’t ask “Is that a feasible growth rate for Exxon?”.

While the Sales Growth Index (SGI) has a high coefficient of 0.892, the coefficient is the result of the data, the model and statistical procedures, and it looks like the threshold of -2.22 is responsible for the sensitivity. The sensitivity is reasonable because the cost of owning shares in a company that has to restate its earnings is high, but the sensitive threshold means excluding many companies that are innocent, if you don’t follow up the score by noting the indexes with the warning signs and investigating to see how genuine the warnings are.

From Table 3 in “The Predictable Cost of Earnings Manipulation” (already linked to), it looks like:

    An M-Score of -1.89 means a 1 in 40 chance of earnings manipulation.
    An M-Score of -1.78 means a 1 in 20 chance of earnings manipulation.

although I’ve translated relative mis-classification costs into probabilities to get the results.

If a company has sales growth of 78.5%, the amount added to the M-Score above the ‘no change’ value is:

    extra on M-Score from 78.5% sales growth = 0.892 * 0.785 = 0.7.

Adding that to the high version of the ‘no change’ M-Score:

    78.5% sales growth M-Score = -2.48 + 0.7
    = -1.78

which hits the 1-in-20 threshold, implying a 1 in 20 chance of earnings manipulation.

A similar calculation shows that sales growth of 66.2% is enough to give a 1 in 40 chance of earnings manipulation.

In the previous section I calculated the contribution that the ratio Depreciation / Total assets makes to the M-Score for four companies, and came up with a low version of the ‘no change’ M-Score of -2.845, based on the Depreciation / Total assets of 7.8% for Synchronoss Technologies, the company with the highest ratio. I now have two ‘no change’ M-Scores which can be used to bracket most of the likely range of possibilities:

    -2.845 – Low ‘no change’ M-Score, assuming a high Depreciation / Total assets of 7.8%
    -2.480 – High ‘no change’ M-Score, assuming negligible Depreciation / Total assets (so TATA = 0)

Now if a company grew it’s sales by 107.1%, had depreciation equal to 7.8% of its total assets, and no change in anything else:

    107.1% sales growth extra on M-Score = 0.892 * 1.071 = 0.955.

Adding that to the ‘low’ or ‘depreciation at 7.8% of assets’ no-change M-Score:

    M-Score for 107.1% sales growth, depreciation at 7.8% of assets, and no change elsewhere = -2.845 + 0.955
    = -1.89

which hits the 1-in-40 threshold, implying a 1 in 40 chance of earnings manipulation.

The equivalent result where the depreciation term in TATA was considered to be negligible, was:

    M-Score for 66.2% sales growth, negligible depreciation term in TATA, and no change elsewhere = -2.48 + 0.59
    = -1.89

implying a 1 in 40 chance of earnings manipulation (or a 2.5% chance).

Between the two extremes of depreciation, from negligible to 7.8% of total assets, sales growth between 66.2% and 107.1% is required to signal a 1 in 40 chance of earnings manipulation, without other indexes pulling the M-Score down more than they would (in aggregate) for no change.

That doesn’t read like a shocking headline, but it suggests that 2.5% of companies would manipulate their earnings after sales growth between 66% and 107%. Since high growth does not always create trouble and the pressure to manipulate, the proportion of companies with management willing to manipulate seems to be more than 2.5%. If half of the companies had the growth without the trouble and pressure, then 5% of management cave in to the pressure. You shouldn’t believe that fast growing companies rarely hit trouble, and nearly all management are honest, at the same time. While relatively few companies manipulate earnings, it looks like a lot more of them would if they faced some more pressure to manipulate. However, I can’t be too confident about the conclusions as I might have oversimplified a complex matter. The analysis could be improved by finding a more representative range for the ratio Depreciation / Total assets.

Summing up the results for no change except sales growth and the effect of constant depreciation:

    Probability, M-Score, Sales growth for negligible depreciation, Sales growth for 7.8% depreciation
    unknown, -2.22, 30%, 70%
    1 in 40 or 2.5%, -1.89, 66.2%, 107.1%
    1 in 20 or 5%, -1.78, 78.5%, 119.4%

Or in more-like-English:

    The threshold of -2.22 which is on many websites but not used by academics, will signal a warning for companies with sales growth above 30% to 70%.

    The threshold of -1.89, where the risk of earnings manipulation is 1 in 40, or 2.5%, will signal a warning for companies with sales growth above 66.2% to 107.1%.

    The threshold of -1.78, where the risk of earnings manipulation is 1 in 20, or 5%, will signal a warning for companies with sales growth above 78.5% to 119.4%.

In each case, a range is given for the minimum sales growth required to trigger the warning signal. The precise sales growth that triggers the warning depends on the ratio Depreciation / Total assets. The top end of each range of sales growth is only a rough ceiling, and if a company’s Depreciation / Total assets ratio is over 7.8%, the ceiling is busted, meaning that an above-ceiling growth rate is required to trigger a warning.

This is all about the sales growth required to trigger a warning, without any help from the other indexes in the M-Score. In other words, even if every index other than SGI had its ‘no change’ value, SGI would tip the M-Score above the threshold being considered. The complication arises because TATA does not have a simple ‘no change’ value. For every other index, the ‘no change’ value is one. The results quantify how a company can be flagged as an earnings-manipulation risk solely because of high sales growth.

It’s possible that if the M-Score could be improved, some of the weight would be taken off SGI, the Sales Growth Index, and the M-Score would be less sensitive to it.

I can’t translate M-Scores into the probability of earnings manipulation for a range of scores. While Beneish wrote about thresholds, they are based on assumptions about the costs of mis-classification (of manipulators as non-manipulators, and the other way around), and it’s hard to derive many probabilities. YCharts have some limited information about the probabilities associated with some M-Scores.

About fixing scale-sensitive ratios

I’ve said how A / B = 4 can be fairly meaningless for practical purposes, for instance if A = 0.004 and B = 0.001 and they represent leverage for this year and the previous year. There’s a first step to a solution, which I call an “impact function”.

impact = (A – B) / square root of (|A| + |B|)

The function is half-way between subtraction, where only absolute amounts matter, and division, which loses information about the absolute sizes involved.

|A| just means if it’s negative, you make it positive, so |-4| = 4, and |4| =4. Now for A and B where A = 0.004 and B = 0.001:

impact = (0.004 – 0.001) / square root of (|0.004| + |0.001|)
= 0.003 / square root of 0.005
= 0.003 / 0.07071
= 0.0424

which is small, as it ought to be.

For the same percentage increase, from a base big enough to matter:

A = 0.8
B = 0.2

impact = (A – B) / square root of (|A| + |B|)
impact = (0.8 – 0.2) / square root of (|0.8| + |0.2|)
= 0.6 / square root of 1
= 0.6 / 1
= 0.6

which is 14.15 times as big as the 0.0424 impact calculated for a rise from 0.001 to 0.004, or 0.1% to 0.4%.

That seems more appropriate, for a scale sensitive ratio such as the change in leverage.

One problem with using impact functions for the M-Score indexes is that the whole model would need recalibrating. There’s also a problem because impact functions as defined here are too symmetric. The result of that is like a system which ignores a warning about an airplane’s engine because another engine is in excellent condition. The problem is easily fixed, but I expect academics could come up with better solutions if they recognized the problem with scale-sensitive ratios and the possibility of better fixes than Winsorizing.

Rolling quarterly results to get trailing twelve month data

The M-Score was meant to be used with annual data, and while quarterly data could be stuffed into a spreadsheet, it would be more noisy, the resulting score would not be calibrated (so hard to interpret), and the exercise has no statistical foundation. So use quarterly data if you want to, but don’t trust the result much.

One general advantage of looking at the quarterly 10-Q forms is that companies can arrange their fiscal year so it ends when they naturally hold the most cash. In such cases, the Accounts receivable and inventories could be lower than their average value, and much lower than their peak value. The Accounts payable reported could be above their average value and further above their lowest point. That won’t systematically affect the M-Score, as the score looks at how ratios have changed over a year, rather than what they are.

Any readers with access to trailing twelve month financial data that’s good enough for an M-Score, won’t need to read the rest of this section.

Trailing twelve month (ttm) data is actually data for the most recent four quarters. Depreciation is needed for the DEPI index, and it’s unlikely to be disclosed in earnings releases, or under headings such as “Quarterly Financial Data (Unaudited)” in a 10-K. It should be in the quarterly ’10-Q’ SEC filings, but you might want to check the first 10-Q has the necessary figures before downloading all the 10-Qs you need.

This spreadsheet shows ttm results up to 1Q 2014 for The Female Health Company:
Female Health - results ttm to 1Q 2014

To get ttm results up to a second quarter, using the period 2Q 2014 and Revenue as an example, if the 10-Q had results for the first six months, the ttm revenue would be

    Revenue 1H 2014
    + Revenue 2013
    – Revenue 1H 2013

where ‘1H’ means the first half (the same as the first two quarters). Using individual quarters, the ttm revenue would be …

    Revenue 2Q 2014
    + Revenue 1Q 2014
    + Revenue 2013
    – Revenue 2Q 2013
    – Revenue 1Q 2013

… which is obviously more work.

For a third-quarter 10-Q, you can save work by doing the same kind of thing with nine-month figures if they’re given, as in Revenue ttm = Revenue first-nine-months-of-2014 + Revenue 2013 – Revenue first-nine-months-of-2013.

You can probably save more work when the latest results are for the third quarter, by using figures from the fourth quarter earnings release, or fourth quarter figures under a heading like “Quarterly Financial Data (Unaudited)” in a 10-K. These probably won’t show Depreciation, but will have most of the figures needed, although you need to avoid using non-GAAP figures. For those figures, the calculation is like this:

    Revenue first-nine-months-of-2014 (from the 3Q 2014 10-Q)
    + Revenue 4Q 2013 (from the 4Q 2013 earnings release, or 4Q 2013 figures under “Quarterly Financial Data” in the 10-K for 2013)

For depreciation and any other quantity not disclosed for the fourth quarter, revert to the previous method.

The ‘hybrid’ method involves less copy-and-paste, but might need more thought and some more searching within documents.

M-Score definitions

M-Score =
+ 0.92 * DSRI (Days Sales in Receivables Index)
+ 0.528 * GMI (Gross Margin Index)
+ 0.404 * AQI (Asset Quality Index)
+ 0.892 * SGI (Sales Growth Index)
+ 0.115 * DEPI (Depreciation Index)
– 0.172 * SGAI (Sales General and Administrative Expenses Index)
– 0.327 * LVGI (Leverage Index)
+ 4.679 * TATA (Total Accruals to Total Assets)

Index definitions –

DSRI – Days Sales in Receivables Index (+0.92)
this year’s Days Sales in Receivables / previous year’s Days Sales in Receivables

GMI – Gross Margin Index (+0.528)
previous year’s Gross Margin / this year’s Gross Margin

AQI – Asset Quality Index (+0.404)
this year’s ratio (Non-current assets – Property plant and equipment) / Total assets
divided by the previous year’s ratio

SGI – Sales Growth Index (+0.892)
this year’s Sales / previous year’s Sales

DEPI – Depreciation Index (+0.115)
previous year’s ratio Depreciation / (Depreciation + net Property plant and equipment)
divided by this year’s ratio

SGAI – Sales General and Administrative Expenses Index (-0.172)
this year’s ratio SGA expense / Sales
divided by the previous year’s ratio

LVGI – Leverage Index (-0.327)
this year’s ratio Total debt / Total assets
divided by the previous year’s ratio

TATA – Total Accruals to Total Assets (+4.679)
(this year’s working capital other than cash – the previous year’s working capital other than cash – Depreciation)
divided by Total Assets

and everything is ‘this year’ unless labeled ‘previous year’.

It’s useful to break some of the indexes down so it’s easier to plug numbers into them –

this year’s ratio Accounts receivable / Sales
divided by the previous year’s ratio

this year’s ratio Gross profit / Sales
divided by the previous year’s ratio

this year’s ratio (Total assets – Current assets – Property plant and equipment) / Total assets
divided by the previous year’s ratio

this year’s ratio (Long term debt + current liabilities) / Total assets
divided by the previous year’s ratio

I prefer to use Total liabilities / Total assets, for leverage, but that’s unconventional, and it’s reasonable to exclude the liabilities which probably won’t incur a cash payment, from the top. Find “It would be reasonable to exclude Deferred grant income” above. I also prefer to net-out the cash, which can result in negative leverage.

(this year’s Non-cash working capital – the previous year’s Non-cash working capital – Depreciation) / Total Assets

where Non-cash working capital = Current assets – Cash & cash equivalents – Marketable securities under current assets – Current liabilities + Short term debt

and everything is ‘this year’ unless labeled ‘previous year’.

Build an M-Score spreadsheet

The M-Score spreadsheet is not available for downloading, but there are instructions for making it under the “Re-alert” tab. Find “Build an M-Score spreadsheet” there. The instructions are for four copy-and-paste operations, from the text, into a blank spreadsheet. The first two pastes involve the Text Import dialogue. After that the spreadsheet needs formatting.

The Female Health Company – 1Q 2014 results release

The Female Health Company Reports First Quarter Operating Results” (femalehealth.investorroom.com)

In my opinion the results are poor but probably explained by the lumpiness which management attribute to “public sector purchasing patterns”. The bright spot is revenue up nearly 40% quarter on quarter. The comparisons are affected by the previous quarter being bad and the quarter a year ago being good with record revenue. 1Q 2014 ended December 31, 2013.

The worst sequential declines are due to having an Income tax expense of $98,875, instead of an Income tax benefit of $4,879,728 in 4Q 2013 (from the 4Q and full year 2013 results release). I haven’t calculated the NOL carryforwards and Decrease in valuation allowance for the previous quarter, but for year-on-year comparison:

    Utilization of NOL carryforwards was $125,982, down from $565,278.
    Decrease in valuation allowance was $292,932, down from $321,247.

If you missed why those are important, find “Conclusions about Female Health’s tax assets” above (it’s complicated).

This shows the Income tax expense for the latest quarter, in context. The quarter couldn’t be expected to change the tax situation significantly, and hasn’t, even though the comparisons show a big percentage gain, and a big percentage switch from a benefit in the previous quarter.
Female Health - results 1Q 2014

The fourth quarter was disappointing, generally at about half-speed compared to 4Q 2012. See:

The Female Health Company Reports Fourth Quarter and FY2013 Operating Results“(femalehealth.investorroom.com)

    “… unit sales decreased 52%”

Gross profit fell 60% to $2.3 million, from $5.8 million in 4Q 2012.
The gross profit margin fell to 48% of net revenues, from 58% in 4Q 2012.
Operating expenses fell 79% to $0.5 million, from $2.6 million in 4Q 2012.
Operating income fell 45% to $1.7 million, from $3.2 million in 4Q 2012, due mostly to lower unit sales.
Net income fell 19% to $6.6 million, or $0.23 per diluted share, from $8.2 million, or $0.29 per diluted share, in 4Q 2012.

There was a net tax benefit of $4.9 million, versus a $5.1 million benefit in 4Q 2012.

FHCO – the backlog

I haven’t checked the data to see if the backlog is useful for predicting revenue.


    Unfilled product orders totaled $2,940,710 at November 29, 2013 and $11,382,572 at November 30, 2012. Unfilled orders materially fluctuate from quarter-to-quarter, and the amount at November 29, 2013 includes orders with requested delivery dates later in fiscal 2014. The Company expects current unfilled orders to be filled during fiscal 2014.” (10-K for 2013)

That’s a 74.2% fall in the backlog, or a fall from 32.5% of 2012 revenue, to 9.3% of 2013 revenue. Comparing both backlogs to 2013 revenue, the fall is from 36.2% to 9.3%.

I couldn’t find “unfilled”, “backlog” or “pipeline” in the 10-Qs or quarterly earnings call transcripts I looked at, so there might be no more information about the backlog until the 10-K for 2014.

FHCO – risk

Risks include:

1) High customer concentration (find “To illustrate the customer concentration” above).

2) The reaction when shareholders learn that Net income won’t benefit from the tax assets, which I’ve written about.

3) Competition. Female Health are the clear leaders. From the 1Q 2014 10-Q:

“None of these female condoms marketed or under development by other parties have secured FDA approval. FDA approval is required to sell female condoms in the U.S. or through USAID. The Cupid female condom became the second female condom design to successfully complete the WHO prequalification process in July 2012 and be cleared for purchase by U.N. agencies.”

Even so, there are issues.

a) A big maker of male condoms might invade the market. If they acquire Female Health, some of the gain will go to management with “up to three years of compensation” (see “Anti-takeover provisions” below), and management might not be incentivized to hold out for a good price. They could, instead, talk about what a wonderful fit the companies are, and the benefits to stakeholders. Customers would probably prefer to see a big company develop its own products, bringing some serious competition into the market. I don’t expect a big company to acquire one of Female Health’s competitors, at least not while Female Health is a viable acquisition target and the competition is small. Very small acquisitions (by headcount) tend to be for proven intellectual property, and research teams or other talented personnel, rather than nascent low-tech businesses (in my experience, anyway).

b) Do they and will they make the best female condoms?

‘Best to the users’ might not be the same as what the agencies think is best, and it’s likely there won’t be a single ‘best’ condom. There’s a little information later about the female condoms women prefer.

In 2013, 2012 and 2011, only $4,745, $5,277, and $10,929 was spent on research and development. The expense is included in selling, general and administrative expenses.

c) How strong is product loyalty?

I can’t gauge how reluctant the women using the condoms might be to switch to an alternative product, how years of use would affect that, or how much notice the agencies involved would take of any reluctance to switch. However, Female Health’s patents mean that designs which avoid the patented features will need their own instructions. The company mention education and training as if it’s a big deal for competitors, and maybe it is. I don’t understand why a substantial market can’t be addressed with well designed printed material and online services. If Ubiquiti Networks (see my previous blog piece) can provide support for communications equipment through an online community, maybe one of Female Health’s competitors could develop an online community to help women who have relevant questions or problems. Even if that’s an inferior service, sometimes innovation survives an inferior service (it’s practically in the definition of ‘disruptive innovation’). Whatever percentage of the market could be backed by online support, the percentage is likely to rise. However, that’s all just my opinion. I’m thinking of all the times people get by with inadequate instructions, from assembling flatpack furniture to inserting SIM cards, and maybe that’s the wrong way to look at it.

At least some of the ‘education’ part of education and training will not be product-specific, and will educate about the advantages of female condoms generally.

I expect the big customers won’t have much loyalty to products or suppliers, but would need to be convinced that other female condoms match Female Health’s on quality and reliability, and have sufficient stock or production capacity, before ordering meaningful quantities. It’s less risky for the big customers to deal with a company they know and trust, but they might want to nurture competitors. It’s possible that if one agency brags about allowing women to have a choice, other agencies won’t want to look as if they refuse to allow women a choice.

d) How much will price matter to customers if the competitors develop?

I have no picture of likely demand at various price points, but I expect big customers to use competition to lever prices down when they can, or to extract better terms.

Overall, my problem regarding future competition is that Female Health’s FC2 is probably not the best possible female condom, if there even is a ‘best’, but it’s still their only product with no plan to develop a range. Women won’t be deprived of a choice forever, no matter how strong the company’s current position. Female Health benefit from FDA and WHO approval, their relationship with the big customers, and the company’s reputation, reliability, and production capacity. Shareholders like to see their company in a dominant position, but one possibility is that Female Health’s dominance could be seen as an obstacle to choice, with action taken to help competition. I’m only suggesting that because I can’t rule it out, due to little understanding of how opinion forms within gender politics and the agencies concerned.

There are plenty more risks in the 10-K under “Item 1A. Risk Factors”.

FHCO – Competing products are not inferior

At least, competing products aren’t inferior by objective criteria according to this research summary by nine authors:

Performance and safety of the second-generation female condom (FC2) versus the Woman’s, the VA worn-of-women, and the Cupid female condoms: a randomized controlled non-inferiority crossover trial” (thelancet.com)

The summary says nothing about what the women in the crossover trial thought of the products. I’m thinking, why didn’t they ask for opinions when women had completed the trial? (if they did, I didn’t see it in the summary).

FHCO – Conventional and unconventional female condoms

This BBC web page …

The return of the female condom?” By William Kremer BBC World Service, 15 December 2013 (bbc.co.uk)

… is very informative, with information about the various female condoms including radical designs available now or soon.

I’ve also linked to a highly unconventional design under “And finally …”.

FHCO – Some information about the female condoms women prefer

This and the BBC piece are all I could find (after much searching) about which female condoms women prefer:

Female Condoms: the spaghetti sauce principle” by Saskia Hüsken and Sille Jansen from UAFC, September 16, 2013 (condoms4all.org)

Women don’t all like the same condoms (just as people don’t all like the same spaghetti sauce, apparently). Most women liked Female Health’s FC2. The research is from South Africa, and it’s possible it won’t be representative of other areas. By the article’s “spaghetti sauce principle”, Female Health should be making a range of ‘sauces’, and could lose their first-mover advantage if they stick with a single product. The principle isn’t proven,
but even if it’s wrong it doesn’t follow that the FC2 is the best design or will be the most enduring design.

The BBC piece linked to above mentions a study from 2010 involving 170 South African women. The Women’s condom (from an NGO called Path) was the most popular, with Female Health’s FC2 placed second, and most of the women chose to carry on trying new condoms rather than stay with their preferred condom. The ‘spaghetti sauce’ piece cites a paper dated 2011, which implies there were two studies in the same country, unless there was a single study with some confusion about the dates and two different recent interpretations. Two studies are not enough to be sure unless one of them is very big and well conducted. I haven’t seen any research papers or authoritative critiques of the studies, and I don’t know if the studies were independent and well conducted.

It might support the ‘spaghetti sauce principle’ that male condoms are differentiated in a variety of ways, though with no variation in the fundamental design, that I know of.

It’s likely that the women who gave their opinions did not pay the full price of the condoms. In markets where price is a factor, personal preference alone won’t account for market share, although the male condom is likely to stay the cheapest option.

One possibility among many is if about twenty different products survive in the market, and Female Health’s only product (or only viable product) is still the FC2. The FC2 might not be number 1 by women’s preference, but the company could cling to market leadership through their connections with the public health agencies. I wouldn’t regard a position like that as being secure in the long term, although the company might have many years of growth before the situation materialized. Unless a big company moves in, it’s likely to be a long time before a competitor makes serious inroads into Female Health’s sales to their major customers. It would presumably have been much more difficult for Female Health if they had a superior condom but faced an established competitor when they were starting out, although there would have been less need to engage in education. I still wish Female Health wasn’t a one-product company, with the risk of being innovated into a small niche by competitors.

South Africa does not rely on external funding. This article implies that the South African government pays for the condoms, when the government is criticized for distributing less female condoms than male condoms due to their higher cost:

SA’s National Strategic Plan: female condom needs to play a bigger role” (genderjustice.org)

The article isn’t dated, but couldn’t have been written before July 2010. The non-reliance on international agencies is confirmed in the earnings call transcript (find “CEO Discusses F1Q 2014 Results”, below), if you check through the instances of “South Africa”, and find the paragraph where deliveries of the competing Cupid condom are downplayed.

Maybe it’s not a coincidence that women’s preferences were studied in a country that doesn’t rely on big international aid agencies for their condom supply. In any case, it’s as if the women in the studies were being treated as representing consumers instead of as aid targets or lab subjects. I expect the aid to include choice eventually, even if becoming consumer-oriented would be too far a stretch for big agencies. I don’t mean to diminish the practical benefits of improving choice from no female condoms to one kind of female condom. Offering a choice of female condom to aid-recipients would increase the complexity of operations and affect costs, but is likely to increase the take-up, partly because some women won’t like the FC2 but like an alternative, and also because people feel good about being offered a choice (or should I say, they would be ’empowered’).

FHCO – some thoughts

I like the history of free cash flow per share in one of the charts at the end (find “illustrate good growth”). Apart from the tax situation (find “Conclusions about Female Health’s tax assets” above), if future per-share growth is similar to the history, then Female Health should be an excellent investment, but the ‘if’ depends on the issues I’ve mentioned. To give some idea of the potential:

“The global male condom market (public and private sector) is estimated to be $3 billion annually. The global public health sector market for male condoms is estimated to be greater than 10 billion units annually. The private sector market for male condoms is estimated at 3 billion units annually. UNAIDS estimates that the annual public health sector demand for condoms, both male and female, will reach 19 billion units by 2015.” (the 10-K)

For comparison, Female Health’s Net revenues were $31.5 million in 2013.

Sales in the U.S. could help to reduce dependence on a few big customers. This is encouraging:

    “FC2 is being distributed as part of New York City’s Female Condom Education and Distribution Project being conducted by the Bureau of HIV/AIDS Prevention and Control. As of September 30, 2013, FC2 was available in 1,436 locations in New York City, as compared to 1,001 at September 30, 2012, including both community based organizations and the N.Y.C. Department of Health and Mental Hygiene units.” (the 10-K)

Los Angeles/LA County have started a similar program.

If the funding is ever cut in the U.S., there’s probably more chance of sales continuing than would be the case in less developed countries.

I believe there’s a big opportunity, but Female Health could be throwing it away by sticking to a single product and having a very low R&D spend. R&D needs to be spent effectively, and if they don’t have ideas, they could find out why some women don’t prefer the FC2 and find people who can design for their needs (although I’m making it sound too easy).

Female Health’s incentive payments are linked to unit sales and operating income, and really are lower in the ‘down’ years. That’s probably good, but the incentive payments aren’t broken out and are high enough to increase SG&A as a proportion of sales in the good years. Apart from reducing the upside for investors, there could be too much motivation for short term goals at the expense of long term objectives. There’s also the ‘2008 Stock Incentive Plan’, but no options were granted under it in 2013, 2012 or 2011.

The shares offer diversification for investors, as the business depends on the funds available to the big customers rather than directly on the world economy or a part of it.

FHCO – Female Health’s dividend cover and covenant restrictions

This is about the terms of the Loan Agreement:

    “Dividends and share repurchases are permitted as long as after giving effect to the dividend or share repurchase the Company has a ratio of total liabilities to total stockholders’ equity of no more than 1:1.” (10-K for 2013)

At the end of Female Health’s 2013, the ratio of total liabilities to total stockholders’ equity was 3,766,996 : 31,402,957 = 0.12 (as a proportion) or 12%, and $31,402,957 – $3,766,996 = $27,635,961 could be returned to shareholders without breaking the covenant. The company had $8,922,430 cash, but for dividends the limiting factor is likely to be the dividend cover. Free cash flow covered the dividend 1.46 times. That’s with free cash flow defined as Cash from operations – Cash invested, which equals $11,793,081 – $302,198 = $11,490,883, which is 1.46 times the dividend payment of $7,863,915.

Removing the unsustainable components of the Income tax benefit, which are Utilization of NOL carryforwards and Decrease in valuation allowance, 2013 Net income would only have been $6,426,651. The dividend cover from “Net income before tax asset related benefits” (my term) is 0.817.

My guess is that one or other measure of dividend cover will limit the dividend. If not, the limiting factor will be the company’s $8,922,430 cash, which is well below the $27,635,961 set by the covenant described above. If the stockholders’ equity keeps growing, the covenants might never be the limiting factor for the dividend.

The free cash flow cover of 1.46 times is overstated IMO, as the $1.4 million increase in Cash from operations (to $11,793,081) came from lower incentive payments, and cash collection that can’t be repeated (find “The increase of $1.4 million”, above). Net income depended on tax benefits to cover the dividend. Some growth would make the dividend safer, and with lumpy orders and lumpy cash collection, I can’t predict growth in the near term.

FHCO – Anti-takeover provisions

I’m not going into much detail, but the company list six Anti-takeover provisions. There’s the usual authority to issue preferred stock to dilute the would-be acquirer’s stake, making a hostile takeover unlikely, and:

    “change of control agreements we have entered into with four of our employees which provide for up to three years of compensation following a change of control as defined in the agreements” (10-K for 2013)

The generous three years of compensation obviously reduces the price that an acquiring company would be willing to pay, and management’s gain would be the shareholders’ loss.

FHCO – Indemnification of directors and officers

The info is in the 8-K “Report of unscheduled material events or corporate event” dated May 22, 2013.

There could be more up-to-date information in a later 8-K, but if there isn’t, you’d have to open all the later 8-Ks to be sure. There is no mention of indemnification in the 10-K or the DEF 14A (about the annual meeting). Publicly quoted companies don’t shout about the fact that they offer wider indemnification to officers and directors than is generally the case for private equity, but Female Health couldn’t have hidden it better. Other companies at least leave clues in the 10-K.


    “(a) The rights of directors and officers of the Corporation provided in this ARTICLE VIII shall extend to the fullest extent permitted by the Wisconsin Business Corporation Law and other applicable laws as in effect from time to time.”

The coverage is for all kinds of proceedings, including action started by the company:

    “”proceeding” means any threatened, pending or completed civil, criminal, administrative or investigative action, suit, arbitration or other proceeding, whether formal or informal, which involves foreign, federal, state or local law (including federal or state securities laws), and which is brought by or in the right of the Corporation or by any other person.”

“expenses” doesn’t seem to miss much out, and includes proceedings started by a director or officer.

Liabilities covered include paying a judgment, settlement, forfeiture, or fine, including excise tax related to employee benefit plans, plus costs, fees, and surcharges.

There are exceptions:

    (1) A willful failure to deal fairly with the Corporation or its shareholders in connection with a matter in which the director or officer has a material conflict of interest.
    (2) A violation of criminal law, unless the director or officer had no good reason to think they were breaking the law.
    (3) A transaction from which the director or officer derived an improper personal profit.
    (4) Willful misconduct.

Admitting guilt or “no contest” or losing the case doesn’t mean a director or officer can’t be indemnified.

A clause avoids paying if someone else has already indemnified the director or officer.

The company has the option to insure the indemnity, but I’ve seen no indication that any of the liability is insured, so the liability could be uninsurable or very expensive.

That’s all my own summary. Anyone who needs to be sure about the company’s indemnification needs to do their own research.

Wide indemnification of officers and directors is common for publicly quoted companies. The usual excuse is that it’s needed to attract the necessary quality of applicants.

FHCO – Related party transactions

These are found under “Transactions with Related Persons” in the SEC filing ‘DEF 14A Official notification to shareholders of matters to be brought to a vote (“Proxy”)’ dated Jan 28, 2014.

Two fairly recent retirements were with consultancy agreements.

    Mary Ann Leeper, Ph.D
    Senior Strategic Adviser of the Company
    Retired on December 31, 2013
    consulting fees $6,000 a month

    Donna Felch
    Vice President and Chief Financial Officer
    Retired on December 31, 2012
    consulting fees $100,000 per year
    plus health and other similar benefits, plus 22,500 shares of unvested restricted stock continued to vest on December 16, 2013
    Donna Felch was appointed Director in November 2012, and is nominated for election to the Board of Directors at the AGM on March 20, 2014.

The company acknowledge that Donna Felch and two other directors are not independent. Five directors are listed as being “independent under the listing standards of the NASDAQ Stock Market”.

None of the directors were paid just for being on the board, but two were paid $10,000 and $11,000 for committee work, and six (of the nine) had combinations of stock awards and other compensation, making the total director compensation $513,855.

Karen King became President and CEO in January 14. There’s a “clawback” provision in the King Employment Letter, which might have to be amended to comply with the Dodd-Frank Act, as mentioned above.

What I like about the section is the clarity, although I’d have preferred a clear statement that there were no other transactions with related persons or parties. (For a complete contrast, find Gazprom Neft’s related party transactions.)

There are no loopholes as in Ubiquiti Networks’ “We did not have any transactions… in which the amount involved exceeded or exceeds $120,000 and in which any of our directors, executive officers …”. It wouldn’t take many of those to make a substantial sum.

Clarity doesn’t prove accuracy, but I prefer clarity to fog that leaves me wondering what’s hidden in it. I also don’t like to see contracts going to relatives, officer’s other companies, etc. Clarity and nothing too bad is good enough.

FHCO – on Seeking Alpha

There have been good articles about The Female Health Company on Seeking Alpha. I’ve tried to bring a different focus.

Unfortunately this ‘Pro’ article has disappeared behind Seeking Alpha’s paywall: “The Female Health Company: Fundamentals & Commitment” by Joseph P. Porter, Jan. 17, 2014 (seekingalpha.com). I commented (as ‘fnoobler’) that the valuation allowance was low, and described the effect on Net income, but I might not have described the complicated issues clearly enough.

The latest two transcripts are:

The Female Health Company’s CEO Discusses F1Q 2014 Results – Earnings Call Transcript” Feb. 3, 2014 (seekingalpha.com)

The Female Health’s CEO Discusses F4Q 2013 Results – Earnings Call Transcript” Dec. 3, 2013 (seekingalpha.com)

FHCO – A website featuring Female Health’s condoms

This page is titled “Support” then “FC2 information & education”. Among other items, there’s an animation.

Training for whom?” lists five categories, starting with Health care providers.

There are two downloads of training materials from “Training programs“. The FC2 training manual is 62 pages and 2,317 KB.

FHCO – Investor Relations links

Press Releases, SEC filings

FHCO – Some thoughts about program-funding

It’s unlikely that the Bill and Melinda Gates Foundation will run out of money for a while yet, and Bill Gates is currently the world’s richest person. I don’t pretend to understand politics much, but I don’t expect governments to backtrack on their commitments. The U.K. government backtracked on subsidizing solar panels on rooftops when take-up was more than expected, but their $1.5 billion commitment (for HIV/AIDS, TB and Malaria) is fixed, and probably small relative to the outcry and reputational damage if it’s cut. It’s more likely that a cut would be found in aid with an older commitment, less commitment, less need, or vulnerable to corruption (the reason that would get the most popular support). One possible risk is funding cuts in reaction to some African governments killing or jailing gay people. My guess is that economic aid is more likely to be cut than programs aimed at improving health and reducing mortality. I hope the programs aren’t hit by violence, but vaccination workers have been killed, and family planning advocates in Pakistan have had fatwas issued against them, see “Nigeria: Extremists gun down nine women giving polio vaccines to children” By Mike Pflanz, West Africa Correspondent, Feb 08, 2013 (telegraph.co.uk) and “Family planning advocates face fatwas from fundamentalists in Pakistan” by Sumitra Deb Roy, Dec 5, 2013 (indiatimes.com).

FHCO – Female Health’s deferred tax assets imply management’s forecasts have been conservative

There’s a chart above titled Tax assets, Net income and Cash from operations, to 2013 (find “when they become aware of the tax situation”). The light blue line showing Net deferred tax assets is not just rising, but has curled upwards since 2007, the first year where it’s visible and reasonably meaningful. That indicates managements’ continued expectation of growth, which has allowed them to increase their estimate of the amount of tax losses they will be able to offset against taxable income.

The red line shows the growth in Cash from operations, which has been more noisy than the smooth growth in the tax assets. The smooth blue line shows that management have looked beyond the noise created by lumpy orders when making the estimates that the tax assets are based on. Other charts show that Net income has grown even faster, but Net income is not reliable here as it includes the effects of managements’ forecasts. The insight into managements’ expectations can’t have more than one future update, at least not in a good way, as there isn’t much left in the valuation allowance to transfer into deferred tax assets. (Find “Conclusions about Female Health’s tax assets” above, for the issues around the tax assets. It isn’t simple.)

In principle, the size of the Net deferred tax assets should allow a calculation of total income before tax over the next ten years. The calculation gives a disappointing result, of only $9,530,715 a year Income before income taxes on average, compared to $9,933,854 in 2013 and $10,792,023 in 2012. This begs the question, how can disappointing estimates of future income be consistent with the continuous expectation of growth, when both are derived here from the tax assets, with future income derived from the current size of the assets, and the continuous expectation of growth inferred from the smooth and fast growth of the tax assets?

I believe the answer is that management estimates have always been conservative. Conservative estimates are against the GAAP requirement to use a 50% probability threshold, but proving that management have deliberately applied some conservatism to the rapidly rising curve would be very difficult, and the SEC probably have better things to do. This argues against the possible first impression that the rising Net deferred tax assets could be the result of earnings manipulation. This is a tricky area, I don’t like having to consider over-estimation one minute and under-estimation the next, and I expect management would like my considerations even less.

When management made ten-year projections to base the size of the tax assets on, they must have projected taxable income over the ten year period. What I do next is reverse-engineer their projections of taxable income, from the size of the Net deferred tax assets. I can’t do that for individual years, but I can estimate the ten year total (of management’s projected taxable income). My estimate is based on:

    The ratio of Utilization of NOL carryforwards to Income before income taxes for the past three years. If management expected the ratio to drift over the next ten years, the reverse-engineering becomes less accurate.

I assume:

    NOLs are only represented in Net deferred tax assets if management expect they can be realized within ten years. I expect that to be true, because management use a ten year horizon, and even though the U.K. NOLs never expire, my charts show they were not immediately booked into the Net deferred tax assets, because that would not be consistent with the rising blue line mentioned above.

First, I use data from the last three years to get the ratio “Utilization of NOLs / Income before tax”.

    Total, 2013, 2012, 2011

    Income before income taxes
    $23,957,852, $9,933,854, $10,792,023, $3,231,975

    Utilization of NOL carryforwards
    $4,681,905, $2,070,947, $1,637,205, $973,753

    Utilization of NOLs / Income before tax
    0.1954, 0.2084, 0.1517, 0.3012

It’s the ratio 0.1954 which is important, the other ratios show the wide range of variation, from 0.1517 to 0.3012.

Now using:

    Utilization of NOLs = 0.1954 * Income before tax


    Utilization of NOLs over 10 years = Deferred tax assets

which follows from management’s ten year horizon, and the definition of the tax assets. It follows that:

    Income before tax over 10 years = Deferred tax assets / 0.1954
    = Deferred tax assets * 5.1177
    = $18,623,018 * 5.1177
    = $95,307,154

That’s only $9,530,715 a year on average, which is not an exciting prospect as Income before income taxes was $9,933,854 in 2013, and $10,792,023 in 2012. There’s room for error, and I probably ought to write “$9.5 million a year” instead of implying accuracy to the nearest dollar. The error could be up or down.

The effect of no more NOLs from Malaysia is insignificant, as the NOL carryforwards only amounted to $125,000 as of September 30, 2010, and that’s only the amount of income they could offset.

I explained above why I believe this unexciting result reflects conservatism by management rather than poor prospects. In fact, lower is better, because if my estimate of management’s estimates is correct, income would need to be below $9.5 million a year before it’s necessary to reverse the deferred tax assets back into the valuation allowance, creating a tax expense which would hit Net income. However, that’s based on aggregates and could be spoiled by variation in the timing and the tax jurisdictions of income.

Applying the same method to the previous three years,

    Total, 2010, 2009, 2008

    Income before income taxes
    $13,478,149, $4,224,132, $5,050,111, $4,203,906

    Utilization of NOL carryforwards
    $3,505,750, $1,087,410, $1,331,340, $1,087,000

    Utilization of NOLs / Income before tax
    0.2601, 0.2574, 0.2636, 0.2586

It’s the ratio 0.2601 which is important, the other ratios show the range of variation, from 0.2574 to 0.2636.

Now using figures as at 2010:

    Utilization of NOLs = 0.2601 * Income before tax
    Utilization of NOLs over 10 years = Deferred tax assets,
    2010 Deferred tax assets = $5,931,000

    Income before tax over 10 years = 2010 Deferred tax assets / 0.2601
    = $5,931,000 * 3.8447
    = $22,802,768

That’s only $22,802,768 total Income before income taxes projected from 2010 over the following ten years. I calculated above that the sum over the following three years (2011, 2012, 2013) was $23,957,852.

By my estimates, management’s projections in 2010 for the next ten years were more than achieved in the next three years. That’s extremely conservative, but I expect management were in a difficult position. GAAP required using a 50% probability threshold, but orders are very lumpy, and I’d also say that taxable income over the next three or four years was hard to predict, even if good long term growth seemed likely. Looking too conservative is not as bad as the appearance of deliberately inflating earnings, so it’s understandable if management erred on the side of being too conservative when making difficult long term projections. If the tax assets had to be reversed into the valuation allowance, many shareholders would not appreciate that applying the 50% probability threshold to ten year projections makes such reversals quite likely even in a steady business, as I’ve never heard of a business with ten year earnings visibility.

My reverse-engineered estimate of $22,802,768 total Income before income taxes projected from 2010 over ten years, is only $2,280,276 a year on average, compared to the actual 2013, 2012 and 2011 figures of $9,933,854, $10,792,023 and $3,231,975.

As a quick-and-dirty robustness check, I take the ratio Utilization of NOLs / Income before tax, and plug the value calculated over 2011, 2012 and 2013 into the 2010-based calculation, so for 2010 instead of calculating:

    Income before tax over 10 years = 2010 Deferred tax assets / 0.2601

as above, I calculate:

    Income before tax over 10 years = 2010 Deferred tax assets / 0.1954

to get

    Income before tax over 10 years = $5,931,000 * 5.1177
    = $30,353,078

which is higher than the more natural estimate of $22,802,768, but the result is still a very conservative ten-year projection from 2010, given the $23,957,852 total for 2011, 2012 and 2013.

I don’t expect the under-estimation to be at all consistent, but if it was, using ‘U’ for the underestimation ratio:

    U = Average yearly projected Income before tax / Average Income before tax over the next three years

basing the calculation of U at 2010:

    U = $2,280,276 / ($23,957,852 / 3)
    U = $2,280,276 / $7,985,951
    U = 0.2855

If that value of U holds for 2013:

    Average Income before tax over the next three years = Average yearly projected Income before tax / U
    = $9,530,715 / 0.2855
    = $33,382,539

I’ll repeat that I don’t expect the under-estimation to be at all consistent, so don’t expect that much pre-tax income, but given previous under-estimation, my reverse-engineered management projection of only $9,530,715 before tax per year for ten years, should be seen as conservative rather than as anything to worry about, and the conservatism indicates that earnings have not been manipulated by inflating the Deferred tax assets. Instead, the Deferred tax assets are likely to be under-estimated, although there wasn’t much left to transfer to the category. However, any risk to income before tax is also a risk to the Deferred tax assets, although the tax assets have survived lumpy orders and customer concentration since the company achieved positive income in 2006.

I’ll repeat that management don’t disclose the projections they base the Deferred tax assets on, which is why I’ve estimated the projections with some ‘reverse engineering’ arithmetic.

See also “The valuation allowance, management’s predictions, and earnings management” above for a general treatment.

FHCO – the benefit from two programs

The first spreadsheet shown below is about estimating how funding from two big projects will drop down into Female Health’s Net income, cash from operations, and dividends. The spreadsheet is complicated due to the tax-related calculations. Apart from those calculations the model is very basic, and can only give a rough indication, at best.

The projections are in the big table below row 28, which has the wide beige cell starting “Derived from $4.6 billion”. The projections are aggregate numbers, not per year.

The project funding is $4.6 billion announced at the London Summit in July 2012, which is mostly for increasing access to contraceptives, and $1.5 billion from the U.K. government targeting HIV/AIDS, TB and Malaria.

I don’t know how much of the funding will find it’s way into the company’s revenue, so I’ve given a wide range of assumptions. It’s extremely unlikely that Female Health will see the higher percentages of the funding, which I’ve included for completeness.

I’ve lumped the two funding pools together, but there’ll probably be less of the U.K. government’s $1.5 billion heading to Female Health as the program targets TB and Malaria as well as HIV/AIDS. If you can estimate the revenue from both programs, you can look up the nearest figure in the Revenue column, to the right of the column with percentages.

A better model would separate the fixed costs from the variable costs. Then, there would be a revenue ‘pivot’ where profit-related numbers would be the same as in the spreadsheet, but numbers would get smaller below the pivot, and bigger above the pivot. Operating leverage is implied by:

“Since the product’s primary market is currently the public health sector, the Company incurs minimal sales and marketing expense. Thus, as the demand for FC2 continues to grow in the public health sector, the Company’s operating expenses may grow at a much lower rate than that of volume.” (the 10-K)

Much of the complication in the spreadsheet is because of the need to remove the tax benefit in order to get representative ratios, then subtract normal tax from income before tax, and then add back the tax benefit from the Deferred income taxes asset, to Cash from operations. Only the tax asset under ‘Other assets’ is added to Cash from operations, which excludes tax assets under Current assets and the valuation allowance. The exclusions could be too conservative.

The model leaves out revenue that is not from the two funding sources, and adds as much of the tax assets as the income before tax allows, to the cash flow that comes from those two sources. For the lower-revenue scenarios (under 4%) there won’t be enough income derived solely from the two funding sources to justify adding the full amount of the tax asset. That’s included in the calculation. If 4% or more of the funding goes into Female Health’s revenue, the income produced is enough to use all the Deferred income taxes to reduce the tax payable, but it’s possible that the income would not be produced early enough and some of the tax loss would expire. There’s an underlying assumption that the income is in the right tax jurisdictions to qualify for using the tax assets, and it’s likely that more than 4% of the funding has to hit revenue before all the Deferred income taxes can be used as a result.

As well as the limitations of keeping the model simple, the accuracy depends on three ratios holding up. They’re the numbers in red borders on the sheet, for Income before tax / Revenue, (Cash from operations – Utilization of NOLs) / Revenue, and Dividend per share / Earnings per share. Each of the ratios is based on three-year sums, so the first is the sum of all the Income before tax for 2013, 2012 and 2011, divided by the sum of all the revenue for 2013, 2012 and 2011. In favor of the ratios, the three years include a bad year (2011), an excellent year (2012) and 2013 was a bit of a drop-off from 2012. While income and cash from operations can go separate ways in a single year, there are limits on the size and duration of the divergence (for an honest company), although there could be a relatively constant lag for a growing company. I’m hoping that three years are enough to smooth out the noise. The systematic difference between income and cash flow historically, due to net income benefiting from releasing the valuation allowance, has been taken into account. Charts further down show how Net income and cash from operations have diverged and caught up with each other (more or less) over many years. The ‘cumulative’ chart shows how negative net income sagged under the small negative cash from operations until 2005, but has grown faster than the growth in cash-from-ops since (due to recognizing the tax benefits in advance).

Although I calculated the payout ratio in cell S26 using ‘per share’ figures, the projections are not per share, and not per year. Keeping the historic payout ratio based on net income will produce a lower ratio based on Cash from operations in future, as Net income will drop relative to Cash from operations, due to the tax issues described in this piece (find “Conclusions about Female Health’s tax assets”, above).

The model also takes no account of time. From the UK government’s website:

    “Making lifesaving contraceptives available to an additional 120 million women and girls by 2020.”

The quote is from “Family planning: London summit, 11 July 2012” (gov.uk)

It seems fairly likely that the funding covers the ramp up until 2020, after which new funding will probably maintain the supply. I might be taking the word ‘ramp’ too literally, but these hypothetical funding flows from 2013 to 2020 ramp up linearly and add to $4.6 billion:

    $128, $255, $383, $511, $639, $767, $894, $1022 (in millions)

That’s only meant as a very rough approximation, at best. Staff costs will come out of the funding but won’t get into Female Health’s revenue, and they are likely to be fairly constant rather than ramp up.

The education and training Female Health pledged at the Summit was for the period 2013 to 2018.

There’s a little more info on “Family planning: Historic breakthrough for 120m women

To put the figure of 120 million women into some kind of perspective, Female Health’s unit sales going back from 2013 were: 54,759,925, 61,649,997, 32,872,570, 38,919,262, 40,192,600.

In principle, you could estimate the impact of reaching the Summit’s target for contraceptives if you knew these ratios:

    1) Proportion of the 120 million women whose contraceptives included female condoms
    2) Proportion of female condoms supplied by Female Health
    3) Rate of use, per woman
    4) Revenue per condom

In my opinion, the big funding available to international agencies poses a question about how management are likely to respond to incentives. Management are rewarded for performance, and incentive payments are lower in the less profitable years. If they are motivated to achieve short term goals, this could translate to pulling in as much from the well funded agencies as they can, using the existing FC2 product. So long as this provides a good reward to management, there could be little financial incentive to secure the company’s long term future by developing a range of products. Only a shareholding big enough to outweigh formal incentives can align management’s financial interests with that of ordinary shareholders taking a long term view. Present management might not be present if and when competitors gain serious market share. See “FHCO – Competing products are not inferior” and “FHCO – Some information about the female condoms women prefer”, above.

FHCO – Charts

For info about the spreadsheet in the first image below, see “FHCO – the benefit from two programs” immediately above. There are many simplifications, and the spreadsheet could be improved, but would be more complicated. If the numbers are too small to read, in Firefox, right click, select “Open link in new tab”, and use the magnifying glass.

Female Health - Dropdown from revenue - spread

Female Health - Dropdown from revenue - formulas

In the table below, the heading ‘Shares’ means Diluted weighted average common shares outstanding.
Female Health cash walk to 2013 spread

These walk-charts illustrate good growth in cash from operations with relatively little cash needed for investment, but with plenty of noise.
Female Health cash walk to 2013

The heading ‘Shares’ means Diluted weighted average common shares outstanding.
Female Health non-cash walk to 2013 spread

Female Health income and cash walk to 2013

Female Health income and cash to 2013

Female Health cumulative income and cash to 2013

Below, the Income tax benefit is what Female Health have booked instead of a tax expense, although the actual cash paid is from the company to the IRS. Deducting the two components related to the tax assets leaves the tax expense that would have been booked without the tax assets, allowing an estimation of the ‘normal’ tax rate.
Female Health - Normal tax etc - spread

The spreadsheet below shows results for the latest quarter, compared to the previous quarter sequentially and the quarter from a year ago. It was used above, find “shows the Income tax expense for the latest quarter”
Female Health - results 1Q 2014

A chart showing Net income and decreases in the valuation allowance can be found with “the big decreases were not abnormal”.
Charts showing how the falling valuation allowance has driven the increase in tax assets and contributed to income and cash flow, can be found with “Although the boost to the tax benefit is fizzling out”.
The M-Score spreadsheet is shown above “Female Health’s only product is the female condom”, near the top.

And finally …

Designers Turn Female Condoms Into ‘Fashionable’ Dresses” By John Yong, 15 Jul 2013 (designtaxi.com). When I saw the page, the dresses looked sensible compared to some of the fashion items on the right, including a fishbowl handbag shown with goldfish in it.

The highly unconventional condoms in the next link have a serious point: “Anti-Rape Underwear in India, condoms with teeth in South Africa – protecting women and girls, but at what cost?” by Emma Saloranta, April 24, 2013 (girlsglobe.org). It’s a balanced piece about the issues rather than the product design, but it doesn’t mention possible misuse.

This link is only here because I can’t stay serious for too long. You need to magnify the diagrams: “Apparatus for facilitating the birth of a child by centrifugal force US 3216423 A” (google.com/patents).

Copyright ©2014 sinksmith

DISCLAIMER: Your investment is your responsibility. It is your responsibility to check all material facts before making an investment decision. All investments involve different degrees of risk. You should be aware of your risk tolerance level and financial situations at all times. Furthermore, you should read all transaction confirmations, monthly, and year-end statements. Read any and all prospectuses carefully before making any investment decisions. You are free at all times to accept or reject all investment recommendations made by the author of this blog. All Advice on this blog is subject to market risk and may result in the entire loss of the reader’s investment. Please understand that any losses are attributed to market forces beyond the control or prediction of the author. As you know, a recommendation, which you are free to accept or reject, is not a guarantee for the successful performance of an investment.