Anda di halaman 1dari 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

The Patient Investor:


How to Build Wealth by Purchasing Shares in Exceptional Companies at Bargain Prices
By Brian Gaudet
Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form. Any person purchasing the PDF format version of this book is entitled to print a single copy for their personal use.

Disclaimer: The investment results obtained by following the approach described in this book will largely depend on the knowledge and experience of the investor that implements them. By itself, this book will not convey sufficient knowledge and experience to analyze a company and industry to the extent necessary to insure satisfactory results, but rather provides a framework to do so.

Page 1 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Table of Contents
1 Estimating an Investments Intrinsic Value (2005)................................................. 12 1.1 10-Year Investment Grade Corporate Bond ................................................... 14 1.2 Ten-Year Junk Bond...................................................................................... 17 1.3 Bonds Sold Prior to Maturity ......................................................................... 19 1.4 Common Stock .............................................................................................. 22 Historical Evidence (2005) .................................................................................... 26 2.1 Historical SP500 Returns ............................................................................... 26 2.2 Market Efficiency .......................................................................................... 31 2.3 The Low PE Strategy..................................................................................... 34 2.4 The Dow 10 Strategy ..................................................................................... 37 2.5 The Nifty Fifty............................................................................................... 38 A Companys Demonstrated Earnings Power (2009) ............................................. 40 3.1 Adjustments to Reported Earnings ................................................................. 40 3.2 Free Cash Flow.............................................................................................. 45 3.3 A Companys Business Cycle ........................................................................ 50 3.4 Earnings Adjustment Example....................................................................... 57 Estimating a Companys Intrinsic Value (2006)..................................................... 60 4.1 Qualities of a Good Valuation Model............................................................. 60 4.2 A Simple Valuation Model ............................................................................ 63 4.3 Customizing the Model for Different Asset Classes ....................................... 68 4.4 Analyzing the Model ..................................................................................... 71 4.5 Examples ....................................................................................................... 76 Company Analysis: Cashflow Predictability (2009)............................................... 89 5.1 Sources of Competitive Advantage ................................................................ 90 5.2 Quantifying Demonstrated Competitive Advantage ....................................... 92 5.3 Future Prospects ............................................................................................ 99 5.4 Financial Strength........................................................................................ 104 5.5 Company Quality & Purchase Thresholds.................................................... 109 Analysis Examples: Three Exceptional Companies (2009) .................................. 113 6.1 Automatic Data Processing (ADP)............................................................... 113 6.2 The Coca-Cola Company (KO).................................................................... 119 6.3 Johnson & Johnson (JNJ)............................................................................. 127 Analysis Examples: Some Lemons (2007)........................................................... 135 7.1 Unproven Business Models.......................................................................... 135 7.2 Poor Financial Strength................................................................................ 137 7.3 Poor Management Decisions........................................................................ 137 7.4 Substitution ................................................................................................. 139 7.5 Supplier Bargaining Power .......................................................................... 140

Page 2 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

7.6 Product Concentration ................................................................................. 140 8 Portfolio Considerations ...................................................................................... 142 8.1 Asset Class Allocation ................................................................................. 142 8.2 Level of Focus ............................................................................................. 142 8.3 Dividend Policy ........................................................................................... 144 8.4 Frequency of Research................................................................................. 147 8.5 Benchmarking ............................................................................................. 148 9 Appendix............................................................................................................. 150 9.1 A Portfolios Internal Rate of Return ........................................................... 150 9.2 Internal vs. Realized Rate of Return............................................................. 151 9.3 Geometric Vs. Average Rate of Return ........................................................ 153

Page 3 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Preface
This book describes an investment strategy suitable for building wealth over long time horizons. The book should be of interest to both investors saving for retirement and investors desiring a stable and growing income during retirement. Although the approach suggested in this book is much more time consuming than creating and managing a portfolio of index funds, the potential rewards - in the form of both superior long-term returns and greater stability of earnings and dividends - are worth it. To achieve the benefits of this strategy, you will need to first obtain a working knowledge of accounting sufficient to allow interpretation of a companys annual and quarterly SEC filings. Beyond that, you will need to become deeply familiar with the industry of any company you decide to analyze, the source and sustainability of the companys competitive advantage, and any company or industry specific risks. But if you have the time and interest to become proficient in company analysis, it is possible to create a portfolio with higher utility within the context of your financial goal than is possible through indexing. But why not just create a portfolio from actively managed mutual funds with a history of outperforming their benchmarks, and thereby gain the benefits of an actively managed portfolio without the effort of evaluating the common stock of hundreds of potential companies? The problem with this approach is that you lose visibility into the company selection process. An impressive track record could just be the result of randomness; with thousands of active managers pursuing multiple strategies it should not come as a surprise when some of these managers achieve remarkable performance, sometimes this is just a matter of being over-weighted (through luck) in industries that have performed well. And since you probably cannot get an active fund manager to sit down with you and explain in detail his valuation model and selection criteria, how can you be certain whether a funds history of exceptional performance will continue? Typically it does not; there are many studies that show that the performances of mutual funds that have outperformed their benchmarks over even ten-year periods have a tendency to revert to the mean. As a consequence, the only way to insure that a portfolio of individual stocks has been constructed with a sensible investment strategy is to find an investment strategy that you believe is rational, and construct the portfolio yourself. There are two tenets that are central to our investment strategy. The first is to never pay more for a share than our conservative estimate of its intrinsic value, which we will define soon. The second is to focus on exceptional companies. These are companies that have a strong, sustainable competitive advantage, have solid financial strength, are well managed, and are well positioned in a highly profitable and stable industry. Obviously, the market typically prices shares of such companies at a premium to a conservative estimate of their intrinsic value. The way we reconcile this seeming contradiction is by waiting for the share price of an exceptional company to get depressed by some event that is unlikely to cause a long-term impairment to the companys earnings power. This occurs more often than you might think; Altria, McDonalds, UST inc., Proctor & Gamble, and Johnson & Johnson were all available at bargain prices at some point Page 4 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

between 2000 and 2006 and by 2009 you could purchase shares in pretty much any company at a reasonable valuation. This investment strategy requires discipline and patience - once you find an exceptional company, your natural inclination is to want to own its shares immediately, and regardless of price. A shares perfect foresight intrinsic value is equal to the total discounted cash that flows from the company to us over the period we own the share, with each cashflow discounted by our required rate of return. These cashflows consist of both dividends and the price for which we sell our share at the end of our holding period. It turns out that if we purchase a share for less than its perfect foresight intrinsic value, we will realize a return higher than our required rate of return, and vice versa. Consequently, if we can on average purchase shares for less than their perfect foresight intrinsic value, we will realize an overall rate of return in excess of our required rate of return. To accomplish this, we need a valuation model that will be accurate on average, or failing that, the valuation model should have a tendency towards underestimating a stocks intrinsic value. The primary obstacle encountered when estimating a shares intrinsic value is that unlike an investment-grade bond, where the future cashflows are contractually specified at the time of purchase and are well protected by a large buffer of the companys earnings, the future cashflows arising from owning a share of a companys common stock is often completely unpredictable. Although it is sometimes possible to predict next years dividend with some accuracy, the shares market price at time of sale will dominate the sum of discounted cashflows over such a short holding period and who can really say what the market will be willing to pay for a dollar of earnings in one years time. And for longer holding periods, where a shares price will contribute less error to our estimate (due to the larger discount factor), it becomes increasingly difficult to predict future dividends, which will depend upon the companys future earnings power. About the only thing we can state with any certainty is that the companys dividends and share price will over long time periods track changes in the companys earnings power. Predicting a companys future earnings is made easier when a company has a stable operating history indicative of a strong competitive advantage, is well positioned in a stable and profitable industry, is financially sound, and we can make a case that there is a low risk of future impairment of the companys competitive position. In this case, the companys earnings are somewhat insulated from the adverse effects of both competition and changes in the supply of credit. Consequently, the earnings of such a company can actually be easier to predict than that of a stock index, in the sense that we can be reasonably confident the earnings will exceed a modest long-term growth hurdle. Although the companys share price will still be largely unpredictable - aside from a loose tendency to track the companys earnings power - we get around this source of inaccuracy by holding a companys shares until the market makes us an offer we cant refuse in which case the final cashflow resulting from the sales proceeds is much larger than predicted by our valuation model. With this in mind, we use a valuation model that is calibrated such that the ratio of a companys estimated intrinsic value to demonstrated earnings power is equal to that of an

Page 5 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

average company with a given level of financial strength. Although this will rarely result in an accurate prediction of a companys perfect foresight intrinsic value, like John Maynard Keynes, I would rather be approximately right than precisely wrong. Moreover, since we will be applying this model only to exceptional companies, our model will be biased such that it underestimates a shares intrinsic value, giving us a margin of safety. This bias occurs because a strong competitive advantage often results in above average earnings stability, a higher than average return on retained earnings, or both, which in turn implies a higher than average intrinsic value per dollar of demonstrated earnings power. Another useful characteristic of our model is that it takes only two company specific inputs, demonstrated earnings power calculated over the last ten-years and financial strength, both of which are quantifiable at the time of purchase (financial strength is quantified as an annualized expected default rate and default rate volatility). Consequently, we cannot tweak our model to provide a higher intrinsic value for shares we want to own. This helps us to avoid the common pitfall of paying too much for a good company, thereby turning the shares of a good company into a poor investment. Recall that the second tenet of our investment strategy is to only purchase shares of exceptional companies. We use a three-pronged approach to evaluate and quantify a companys quality. First, we quantify the extent of a companys demonstrated competitive advantage using three metrics - return on adjusted assets, operating profit margin, and return on retained earnings - each calculated over the same trailing ten-year period used to calculate the companys demonstrated earnings power1. Our minimum purchase threshold for each of these metrics is set above that of an average company, and we require any company we purchase to meet the minimum threshold of all three metrics. To get an idea of how selective this is, using the trailing 10 years ending in 2007, only 65 companies in the SP500 would have met our minimum requirement for all three metrics2. Next, by undertaking a detailed investigation of the source of the companys competitive advantage and an evaluation of the stability and profitability of the companys industry, we estimate the sustainability of the companys competitive advantage. And lastly, since the extent of a companys competitive position becomes moot if the company defaults on its debt, we analyze the companys financial strength, and if necessary, adjust the companys long-term issuer credit rating downward. These three factors demonstrated competitive advantage, future prospects, and financial strength are combined into a score, with our minimum purchase requirement being a score of at least 1.0 for each metric, and a maximum score capped at 4.0. Our required margin of safety measured as the ratio of estimated intrinsic value to market value depends on this score. The reason that this investment strategy works is that although both change over time, a stocks market price tends to be quite a bit more volatile than its perfect foresight intrinsic value; this led Benjamin Graham to observe that over the short-term, the market is a voting machine (the most popular stocks provide the highest returns, and vice versa), but over the long-term, it is a weighing machine (it weighs a stocks intrinsic value). What this means is that although the stock market is efficient on average - in that a
1 2

Later in this book we will show how these metrics relate to demonstrated competitive advantage Similar results were obtained in 2008

Page 6 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

shares market price and intrinsic value are roughly equal on average over long time periods - a stocks market value typically fluctuates around its intrinsic value, and is more volatile. To some extent, all investments exhibit this discrepancy between the markets current and long-term average evaluation of intrinsic value, but the discrepancy is greatest for those investments with the highest cashflow volatility, such as a companys common stock. One reason for this is that with less certainty regarding an investments future cashflows, it is easier to get irrationally exuberant about a companys future prospects when the company is doing well, and just as easy to get irrationally despondent about the prospects of the same company when it gets into trouble. Another contributor to market inefficiency is that most investors have a short time horizon over which they measure their success, and consequently pay inordinate attention to news that might cause a companys share price to significantly change in the near future. As a result, events that in reality will have only a transitory effect on a companys earnings power often have a much larger effect on the stocks market price. The prevalence of this myopic investing style provides an opportunity to investors that have the patience to realize their return over a longer time horizon; the trick is to determine if some event that has driven down a companys share price actually has a good chance of impairing the companys long-term earnings power. When we purchase shares of a companys stock, our realized rate of return depends on both the magnitude of the purchase price discount to the perfect foresight intrinsic value and the time it takes for the market to realize its mistake and drive up the stocks price until it approaches its intrinsic value. This last factor deserves some elaboration; by selling stocks when their market price significantly exceeds their intrinsic value, and reinvesting the proceeds into the stock of companies with comparable or better future prospects that are trading at a good discount to their intrinsic value, we convert part of a portfolios excess price volatility3 into excess return. When the relationship between intrinsic and market value reverts quickly (say over a few years time), this excess return can be significant. Although this rebalancing can significantly boost our return, it is not required to realize exceptional returns, which are obtainable by essentially holding forever, in which case the increasing dividend stream becomes the sole source of cashflows. This recipe of purchasing exceptional companies at a reasonable price to generate investment rates of return well in excess of that possible by investing in stock index funds is certainly nothing new. However, our investment strategy does have some differentiating factors that make it unique. First of all, we make no attempt to integrate the effect of a companys competitive advantage into our valuation model, but instead estimate a companys value assuming average performance for even the best companies. I believe this difference is crucial, in that it avoids the common investing mistake of paying too much for a competitive advantage that may not give you an adequate return on the premium you pay for the companys shares. Another differentiating factor is that we attempt to quantify a companys demonstrated competitive advantage, and use this as a sanity check against our analysis of a companys future prospects. Moreover, our
3

Price volatility in excess of intrinsic value volatility

Page 7 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

minimum requirements for demonstrated competitive advantage, future prospects, and financial strength are such that our valuation model is likely to underestimate a companys perfect foresight intrinsic value. Finally, we demand a large margin of safety as measured as the ratio of estimated intrinsic to market value for companys that fall towards the low side of our purchase requirements, and reduce this margin of safety for companies that score higher. But regardless of how exceptional we believe a company to be, we never pay more than our conservative estimate of its intrinsic value. Keep in mind that this strategy will not outperform the market every year. Any strategy that did so would have its effectiveness diminish over time as more and more investors implement the strategy, thereby driving up the price and consequently driving down the future return of the investment. It is human nature to want a sure thing. Consequently, the best investment strategies occasionally disappoint, with the result that they are unlikely to be pursued to the extent necessary to decrease their effectiveness. I believe that purchasing shares in exceptional companies when they are out of favor, and therefore trading at a discount to a conservative estimate of their intrinsic value, to be such a strategy. Although it is quite a bit of work to pore through a companys annual reports, familiarize yourself with the companys industry, determine a companys demonstrated earnings power, and come up with an opinion regarding the companys future prospects, this is not the primary impediment to successfully implementing this stock selection strategy. Instead, it is actually a lack of patience manifested as a shortsighted investment horizon that is most likely to derail someone implementing the strategy. Other investors in the stock market are not stupid, and can also calculate a reasonable estimate of a companys earnings power and evaluate a companys competitive advantage, perhaps more accurately than you can. Your real advantage is the patience to purchase shares only when you have an adequate margin of safety, and then the patience to allow your total return from owning a stock to unfold over a long time period, as opposed to guessing which company is likely to experience the largest share price appreciation during a given year. But if you cannot be content with watching your companys earnings power steadily build over the years, even when the increase is not reflected in the companys stock price, you are unlikely to succeed. ******** Although the calculations for estimating a companys intrinsic value, quantifying a companys demonstrated competitive advantage, and evaluating a companys financial strength can be a bit tedious, they are automated in The Investing Tool Box, a software package that is available at: http://web.me.com/briangaudet/InvestingToolBox/Investing_Tool_Box.html The software also contains useful financial planning tools, such as a portfolio analysis tool, Monte Carlo Simulator, Mean Variance Optimizer, and Internal Rate of Return Calculator.

Page 8 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Page 9 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Page 10 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Page 11 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

1 Estimating an Investments Intrinsic Value (2005)


The concepts of intrinsic value, internal rate of return, and uncertainty are central to the stock selection methodology presented in this book. In the introduction, I stated that a shares intrinsic value is equal to the sum of the shares discounted future cashflows, with the cashflows consisting of both dividends and the price for which you eventually sell the share, and with each cashflow discounted using your required rate of return. Because it is impossible to know with certainty an investments perfect foresight intrinsic value, we have to settle for estimating an investments intrinsic value; this slightly changes the definition of intrinsic value to the real expected value of the sum of the investments discounted future cashflows. The real expected value of a single future cashflow is the estimate for the future cashflows value, in todays dollars, that we believe will minimize our prediction error. By real we mean inflation adjusted. The reason we want to estimate the real value of a future cashflow is that we are concerned with an investments real rate of return. After all, a 10% rate of return is not all that impressive if inflation has averaged 10% over the period you held the investment. When you purchase an investment at a price equal to its perfect foresight intrinsic value, your realized rate of return will be the rate at which you discounted the investments future cashflows; if you pay more, the realized return will be less, and vice versa. This can be illustrated with the very simple example of an investment that will return $121 to you in one years time. In the perfect world of this example, there is no inflation, and you know with certainty you will receive exactly $121. If your required annualized return is $121 = $110 . Here 10%, then the investments intrinsic value is easily calculated as (1+ 0.10) we discount the expected $121 future cashflow by dividing the cashflow by one plus our required annualized rate of return, which we will refer to as the discount rate. In the future, we will refer to one plus the discount rate as the discount factor. If we were to purchase this investment for its intrinsic value of $110, our annualized return would be ($121 $110) = 10% ; as advertised, you realized a return equal to the discount rate by $110 purchasing the investment at a price equal to its intrinsic value.

Now what if instead the $121 future cashflow was to be received in two years time? In this case, we need to realize our 10% required return for two years rather than one, and $121 = $100 ; we need to square the the investments intrinsic value is calculated as (1+ 0.10) 2 discount factor because the cashflow occurs two years into the future. Again, it is easy enough to confirm that if we purchase this investment for $100, our annualized return will be 10%4. In general, the intrinsic value (IV) of a future cashflow (C F) is given
4

The ratio of the cashflows future value (FV) to purchase price (PP) is 1.21, and the annualized return after two years can be calculated as r = e ln(FV / PP )/ 2 1

Page 12 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

CF , where DR is the discount rate (your required annualized rate of (1+ DR) N return5), and N is the number of years into the future you expect to receive the cashflow. Clearly, the farther into the future a cashflow occurs, the less the cashflows intrinsic value, which is fairly intuitive. One nice property of intrinsic value is that superposition holds, and therefore we can calculate the intrinsic value of a complex investment with multiple cashflows by summing the intrinsic value of each of the investments cashflows. H CF[N] This can be expressed using summation notation as IV = , where CF[N] is (1+ DR) N N =1 the expected cashflow in year N, and H is the holding period in years.

by IV =

Your required rate of return a.k.a. your discount rate - should reflect the level of the investments real future cashflows. This certainty you have regarding your estimate of is also pretty intuitive; you would likely be willing to pay more for $100 in cash you are certain to receive in a years time than for $100 you expect to receive, on average, in a years time, but where the actual amount received would be either $80, $100, or $120, with each outcome equally likely. In practice, I use a discount rate equal to an investments real long-term historical rate of return. The rationale for this is that markets tend to be efficient on average (we will present evidence for this in a later chapter), which implies that on average, an investments market price is equal to its intrinsic value. Moreover, we learned that when an investment is purchased at its intrinsic value, the realized return from holding an investment equals its discount rate; it follows that by setting our discount rate to an investments historical rate of return, we are using the discount rate that the market has - on average - used to value the investment. Later in this book, we will show that using the expected value of the market discount rate to estimate intrinsic value makes our estimate independent of holding period, which is a useful attribute of a valuation model. When calculating an investments real long-term historical return, it is important to use data that covers multiple economic cycles. If a particular investment does not have at least fifty years of history, I instead estimate a discount rate based off of the historical rate of return of similar investments6 with more history, the rationale being that it is better to be approximately correct than precisely wrong.

At this point, we are ready to illustrate the estimation of intrinsic value using a few examples. We will start with the example of an investment grade ten-year corporate bond held to maturity, then turn to that of a ten-year junk bond held to maturity. We will then explore how selling a bond prior to maturity can have a large effect on your realized rate of return; this will lead into a discussion of how a bonds temporal distribution of cashflows affects the sensitivity of the bonds market price to changes in the market discount rate and the markets expectations of future cashflows. We will then end by looking at the differences between holding debt and equity in a company, and how we might estimate the intrinsic value of a share of a companys common stock.
5

Note that you could also express your discount rate as a monthly return, in which case N would be measured in months. 6 Specifically, investments that should have a similar rate of return volatility

Page 13 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

1.1 10-Year Investment Grade Corporate Bond


Companies issue bonds as a way to raise cash; this cash is typically used to fund capital expenditures, make an acquistion, or for other general corporate purposes. When you purchase a bond, you are making a loan to the company, which is then legally required to pay you periodic interest payments (often referred to as the bonds coupon) and to pay you the bonds face value (sometimes called par value) at a pre-determined future date. The maturity of a bond is the number of years you need to wait in order to receive the bonds face value. Some bonds, known as zero coupon bonds, do not pay periodic coupons, but instead just return the bonds face value when the bond matures. A bonds yield is calculated as the bonds annual coupon divided by current market price, which often differs from the bonds face value. Corporate bonds are also differentiated by their seniority, with senior bonds having the first claim on a companys assets in the case of default. Now imagine you have the opportunity to purchase a ten-year senior bond issued by an investment grade corporation. The bond pays a $600 coupon at the end of each of the first nine years, and $10,600 at the end of the tenth year (a $600 coupon and $10,000 for the bonds face value). In order for us to estimate the bonds intrinsic value, we need to estimate the bonds future cashflows and determine our required rate of return. Since inflation reduces the real purchasing power of a future cashflow, we will express each future cashflow in todays dollars, and use a real (inflation adjusted) rate of return for our discount rate. Lets start out by estimating the bonds future cashflows. In order to do so, we need to first make a couple of assumptions regarding the rate of inflation over the next ten years, as well as the probability that the company might default on its debt sometime during the next ten years. Together, these assumptions will be used to determine the real expected value of each future cashflow. One approach to cashflow estimation is to use historical data taken over a large number of economic cycles. For example, we could estimate a future rate of inflation by looking at consumer price index (CPI) data from 1950 to 2007, where we find that the CPI had an average annual increase of 3.9%. Alternately, we could use the markets expectation of future inflation, which can be estimated by looking at the difference in yields between a 10-year treasury and the 10-year inflation indexed treasury (TIPS). Since the default risk and maturity are both the same for the 10-year treasury and 10-year TIPS, we can attribute the difference in yields to inflation expectations. Since a major source of inflation is the inflation expectations of consumers and businesses, I prefer to use the TIPS spread as an estimate, with the caveat that the market can occasionally be spectacularly wrong, as when the inflation that occurred in the 70s caught the bond market completely by surprise, and in the mid eighties when the bond market overestimated future inflation. For this example, lets assume that the spread is 3%, and use 3% for the expected value of the CPI growth rate over the next ten years.

Page 14 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

When a company defaults on its bonds, it stops making interest payments, and investors typically recoup only a fraction of the bonds value7; depending on when the default occurs, this can considerably reduce the bonds future cashflows. For simplicity, in the following examples we will assume that bondholders receive nothing in the case of default. The probability of a company defaulting on its bonds depends on its financial strength, as well as the general economic conditions over the next ten years. Fortunately, credit rating agencies have compiled vast quantities of data that tabulate the default rates of cohorts of companies with a given credit rating over many economic cycles. I prefer to use the default statistics compiled from 1920 to the present, as this covers a depression and several major recessions. Because these statistics are averaged over many economic cycles, they will tend to be pessimistic during an economic boom and optimistic during a recession. But regardless of current economic conditions, I believe it is best to use the average; in other words, we treat economic growth as a random walk with trend, and make no attempt to predict whether the next ten years will be above or below the longterm trend. Lets assume our company has an A2 long-term issuer credit rating. Using data from Moodys, we find that companies with this credit rating have historically experienced a ten-year cumulative default rate of 3.61% (i.e., 96.39% of companies are still solvent after ten years). We can convert this into an annualized default rate of 0.356%. Combining the effects of inflation and the probability of default, the real expected value of the $600 received in years one through nine will fall at a rate of 0.356% (the combined effects of inflation and default probability8) per year. So the first years coupon will have a real expected value of 600 * (1-0.00356), the second years coupon will have a real expected value of 600 * (1 0.00356) 2 , and so on. Now that we have estimated the bonds future cashflows, we need to determine our discount rate. We earlier made the case for using the real long-term historical rate of return realized on similar investments; lets assume that over the last fifty years, A2 rated corporate bonds provided a 2.5% real annualized9 rate of return, and use 2.5% as our real discount rate. The real expected value of the coupon (and face value) in each year, as well as the discounted real expected value of each coupon (the coupons intrinsic value) is shown in Table 1-1. The first row of the table indicates the year in which the cashflow is received, the second row indicates the value of the cashflow, unadjusted for default rate or inflation, the third row indicates the inflation adjusted expected value (taking into account the default rate) of each cashflow, and the fourth row indicates the discounted value of the real expected value of each cashflow, or in other words, the intrinsic value of CF each cashflow as given by the equation IV = , where CF is the real expected (1+ DR) N
7

Although sometimes investors will also gain an equity stake in the company after it emerges from bankruptcy 8 To be precise, we should sum these rates geometrically: (1+0.00356)(1+0.03)-1=0.0337, but the error incurred by simply summing the rates is small, and summing the rates is more intuitive 9 Keep in mind that the historical annualized rate of return (also known as the geometric return) will always be less than the historical average annual rate of return; the reason for this is explained in Appendix 9.3

Page 15 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

value of the cashflow, DR the discount rate, and N the year in which the cashflow is received. Note that the nominal cashflow for year 10 includes both the coupon for year 10 and the bonds face value. Table 1-1
End of Year Nominal Cashflow Real expected Cashflow Discounted Cashflow 1 600 580 566 2 600 561 534 3 600 542 503 4 600 523 474 5 600 506 447 6 600 489 422 7 600 472 398 8 600 457 375 9 600 441 354 10 10600 7535 5892

We can determine the bonds intrinsic value by summing the discounted cashflows for year 1 through year 10 (found in the last row), doing so gives us an estimated intrinsic value of $9965, and if we purchase the bond for this amount, we can expect (on average) a 2.5% real rate of return. Now that we have estimated how much the bond is worth, lets turn our attention to what the market price is likely to be. Since we used market expectations for our estimate for future inflation, and set our discount rate to the long-term historical market discount rate, it is likely that the market price will at least be in this ballpark. Nevertheless, if the demand for bonds of this maturity is particularly high (perhaps due to foreign countries accumulating U.S. denominated currency reserves, or a temporary increase in investors risk aversion that results in a shift in demand from stocks to bonds), the market price will likely be above our estimate. Similarly, if the demand for bonds is unusually low, the market price will likely be below our estimate of intrinsic value. In either case, the difference between market and estimated intrinsic value would stem primarily from a difference in the discount rate used by the market as compared to that used in our estimate. Earlier we learned that if the market price is higher than our estimate of intrinsic value we should expect a lower rate of return, and vice versa. Unfortunately, when the market price is not equal to the bonds intrinsic value, there is no closed-form solution for calculating the bonds expected rate of return. However, we can determine this rate of return by plugging in discount rates until we find one that brings the bonds intrinsic value equal to the market price; this discount rate is sometimes called the internal rate of return. This is fairly intuitive: if the estimate of intrinsic value exceeds the market price, the only way to bring the estimate of intrinsic value in line with the market value is to increase the discount rate (which will reduce each discounted expected cashflow); once market price and intrinsic value are equal, this discount rate will be our expected return. In practice, we do not blindly plug in discount rates, but instead implement an algorithm called a binary search; this allows us to efficiently close in on the internal rate of return. Note we could alternately use the same discount rate as the market. If we do so, then we would just be taking the market price as the best estimate of intrinsic value. This is actually a reasonable approach for bonds with short maturity and high credit rating, and the only approach for money market securities, where the price never deviates from $1. But as uncertainty increases, we will see that it can be dangerous to assume that the Page 16 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

market price is always the best estimate of intrinsic value. If you invested money in certain high-tech companies in late 1999, you probably already know this.

1.2 Ten-Year Junk Bond


To illustrate the effect of increasing uncertainty on the discount rate, lets now attempt to estimate the intrinsic value of a ten-year junk bond with the same nominal coupon ($600) and face value ($10,000) as in the previous example. A junk bond is a bond with a credit rating that is less than investment grade (BA1 or lower). Lets assume that the bond we are valuing is rated B2, which gives it a cumulative ten-year default probability of 31.9%, and an annualized default rate of 2.8%, which combined with inflation, would reduce the real expected value of each coupon by 5.8% annually. In addition, the expected annual default rate volatility (as measured by the standard deviation) increases from 0.3% to 4.5%10 as the rating decreases from A2 to B2. This increased volatility implies a larger dispersion of possible outcomes around the expected value, which means we are less certain about the actual outcome, and we would therefore be willing to pay less (as compared to the investment grade bond) for the expected value of the bonds future cashflows. This gets reflected in the discount rate, which should be higher than for the A2 rated bond. Historically, junk bonds have provided (on average) a 2% higher rate of return than investment grade bonds, so here we will use a 4.5% discount rate, 2% higher than we used for the investment grade bond. Repeating the procedure from the last example yields Table 1-2, and an intrinsic value of $7133, quite a bit less than the intrinsic value of $9965 for the investment grade bond with the same coupons and face value. Table 1-2
End of Year Nominal coupon Real expected coupon Discounted Coupon 1 600 565 541 2 600 532 488 3 600 502 441 4 600 472 398 5 600 445 359 6 600 419 324 7 600 395 292 8 600 372 263 9 600 350 238 10 10600 5832 3789

Now if we were to purchase this bond for $713311, then our real rate of return (on average) should equal our real discount rate of 4.5% (recall that when we purchase an investment at a price equal to our estimate of its intrinsic value, our expected return will equal the discount rate). It should come as no surprise that the junk bond has a higher expected rate of return than the investment grade bond; we have learned that paying less
10 11

From Moodys historical default tables Note that typically bond underwriters do a better job of determining market expectations, and a bonds market price at the time of issue is usually closer to its face value; in this case that would be achieved by increasing the bonds coupon relative to face value (thereby raising the interest rate). However, I wanted to keep the coupons and face value the same as in the previous example to better illustrate how the discount rate and expectations of future defaults impacts the bonds intrinsic value. That said, bonds issued by companies that subsequently experience a decline in financial strength often do trade at a significant discount to their face value.

Page 17 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

for a given sequence of future cashflows increases our rate of return, and the higher discount rate means the market price of the expected value of the bonds future cashflows is quite a bit lower than for the investment grade bond, with the higher discount rate being a result of less certainty regarding the bonds future cashflows. This relationship higher uncertainty causing a higher discount rate and higher expected returns is common on average for all asset classes. The key word here is on average; investments like junk bonds often provide a lower rate of return than investment grade bonds, along with higher rate of return volatility; stocks with high price to earnings ratio typically produce similar outcomes, One consequence of this increased uncertainty is that the market price of junk bonds tends to be much more volatile than that of investment grade bonds. This market price volatility has two major sources, the market discount rate and the markets estimate of future default rates. During an economic boom, junk bond12 default rates plunge, and investors eventually start to believe that these low default rates will persist over the period they hold the bond. Moreover, investors risk13 aversion tends to decrease during economic booms, and as asset prices appreciate, the demand for riskier assets increases as investors search for higher expected returns. Both of these effects compress the junk spread, which is the difference in yields between investment grade and junk bonds of the same maturity. On the other hand, the reverse occurs during a recession. Investors start to believe that the sharp increase in junk bond defaults that typically occur during recessions will persist indefinitely, and the market discount rate increases along with risk aversion. It is with these type of investments, where the market discount rate and expectations of future cashflows are both volatile, that the benefits of using the historical average discount rate and default rates becomes most apparent. To illustrate, lets assume that we are five years into an uninterrupted economic expansion, and junk bond default rates have fallen close to zero. In consequence, the combination of decreased risk aversion and lowered expectations of future default rates causes the price of the bond in our example to increase to $9500. But if we were to purchase the bond at this price, we leave ourselves exposed to the risk that a slowdown in the economic expansion might push up default rates. If default rates were to quickly revert to the historical average, and inflation remained at 3%, our expected rate of return would be only 0.6%14, with the realized rate of return depending on if and when the company defaulted on its debt. And if the economy fell into recession, and junk bond default rates hit 5%, our expected rate of return would fall to 1.7%. But, if we use historical discount and default rates in our estimate of intrinsic value, we would not have bought this bond when it was priced at $9500, as the ratio of estimated intrinsic value to market value would have been only 0.75, indicating that the bond is overpriced.
12

Junk bond default rates vary much more with the economic cycle than investment grade bond default rates; this can be seen by the huge difference in historical default rate volatility between junk and investment grade bonds 13 We will sometimes use the word risk as a synonym for uncertainty, as is the case in most financial literature. 14 Here and in the remaining examples I used a computer program to calculate the expected return based off of the new set of expected cashflows and the indicated purchase price.

Page 18 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Of course, even if we were to purchase this bond at a price equal to our estimate of its intrinsic value, there is still some risk, as default rates could easily spike above their historical levels. Still, the risk would be less than if we purchased the bond at a price greater than our estimate of its intrinsic value. But why not decrease risk even further, and only purchase such a bond if the market price was significantly less than our estimate of its intrinsic value? This would leave us a nice buffer if default rates spiked. To illustrate this, lets assume that the market price of the bond is only $5000. If we purchase the bond at this price, and the annualized default rate ended up being 5%, then our expected rate of return would still be 7.5%. Of course if default rates ended up being close to the historical average, our expected rate of return would be even higher. This example illustrates the reasoning behind one of the pillars of our investment strategy, which is to purchase securities when they are trading at a good discount to our estimate of their intrinsic value: if our estimates are close, we end up with a rate of return well in excess of our discount rate, and if our estimates end up being too optimistic, our rate of return will at least be higher than if our margin of safety had been less. Unfortunately we cannot just wait for the ratio of estimated intrinsic value to market value to reach some arbitrarily high value, if we did so, we might never get a chance to invest! As an extreme example, if you estimate that the shares of a money market fund are really worth only $0.99, you would be out of luck waiting for the market price to drop that far. That said, investment classes with an inherently higher level of uncertainty regarding their future cashflows exhibit higher price volatility, and it is occasionally possible to purchase the investment at a good discount to our estimate of its intrinsic value. The catch is that you need to have the patience to wait for an investment to become priced at a discount sufficient to give you a margin of safety. Lets end this example by noting that the use of long-term historical default and discount rates implies that our estimate of intrinsic value will be quite a bit less volatile than the markets; this keeps us insulated from the markets current level of exuberance or pessimism. The estimate might not end up being accurate (it probably wont be), but we can compensate for this by only purchasing such a bond when our estimate of intrinsic value exceeds the bonds market value by a nice margin; this gives us a margin of safety.

1.3 Bonds Sold Prior to Maturity


Up till now we have assumed that we hold a bond till maturity. If we discard this assumption, will find that our expected rate of return volatility increases; this occurs because a bonds market price can change significantly over our holding period. This is due to changes in the market discount rate and the markets estimate of inflation and default rates, which occur for reasons described in the last example. The result is that the value of our final cashflow will no longer be the sum of the last years coupon and the bonds face value, but will instead be equal to the sum of the coupon received in the year we sell the bond and the bonds market price at the time of sale. The concepts presented

Page 19 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

in this example are very much applicable to shares of common stock; after all, shares of common stock have an infinite maturity, and all are eventually sold Lets illustrate this by looking at a ten-year Treasury bond with a face value of $10,000 and a $500 annual coupon, and where the market expects inflation to average 3% over the next ten years, and applies a 1.85% real discount rate. Under these conditions, the market price of the bond would be $10,000, but if at a later date investors inflation expectations increased from 3% to 4% and the real required rate of return remained at 1.85%, a bond with a face value of $10,000 would need to pay a $600 annual coupon to be fairly priced. Obviously, with newly issued bonds yielding 6%, we probably would not find anybody willing to pay $10,000 for our bond the yield at purchase would only be 5% at that price - and the price would need to fall to attract any buyers. How much the price falls depends on when the change in inflation expectations occurs; the drop would be largest if it occurred one day after we purchased the bond, and would decrease as the bond approached maturity, approaching zero on the day the bond matures (when the market price will be equal to the bonds face value15). We would see the same behavior if either the discount rate or expected default rate increased. Conversely, we would see the reverse if either inflation expectations, the expected default rate, or the discount rate decreased; here the bonds market price would increase, with the increase being largest the day after we purchase the bond, with the price eventually converging on $10,000 as the bond approaches maturity. Now lets say that one day after purchasing this bond, inflation expectations increase to 4% (just our luck!), and the bonds price falls to $9199. Assuming inflation actually does increase to 4% over the next ten years, our expected rate of return (assuming we hold to maturity) would fall to 0.80%, down from 1.85% if inflation had remained at 3%. Although definitely suboptimal, things get much worse if we decide after one year we suddenly need the $10,000, and must sell the bond at the prevailing market price. Since we have already received the first years coupon, the bonds remaining cashflows consist of the nine remaining coupons and the $10,000 face value. With the increase in inflation expectations, the market would likely price the bond at this point of time at $9259, higher than the market price of $9199 that prevailed when inflation expectations first increased because the bond is closer to maturity. If we sold at this point, our cashflows would consist of the $500 coupon received at the end of the first year and $9259, giving us a total of $9759, and an annual rate of return close to 0.241%, quite a bit lower than the 0.80% we could have obtained by holding to maturity. On the other hand, if inflation fell to 2% the day after we purchased the bond, our expected rate of return assuming we hold to maturity would be 2.9%. But if we sold the bond after receiving the first coupon, our rate of return would be 13%. The main point here is that a bonds rate of return will tend to be more volatile if sold prior to maturity, and that the change in a bonds market price due to changes in either the discount rate or the expected value of future cashflows exposes the bond holder to liquidity risk in that if
15

Since at that point a purchaser would be receiving $10,000 immediately, with no risk. This is the same as purchasing $10,000 in cash for $10,000.

Page 20 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

the bond must be liquidated for cash, changes in market price can make your realized return quite different than it would have been if the bond were held to maturity. It turns out that the sensitivity to changes in a bonds market price to changes in the discount rate, inflation expectations, and the expected default rate is smaller for bonds whose cashflows (with the assumption that the bond is held to maturity) are weighted more heavily towards the present. To see why, consider that with more cashflows weighted towards the present, a greater percentage of the bonds discounted cashflows occur sooner, and therefore the bonds intrinsic value depends less on discounted cashflows weighted farther into the future. Moreover, as a cashflow is received further into the future, its intrinsic value becomes more sensitive to changes in both the discount rate and assumed coupon growth rate (which is determined by the annualized default rate and CPI growth assumptions)16. As a simple example, consider the intrinsic value of a $100 cashflow received in one years time using a 5% discount rate, which is equal to $95.24. Increase the discount rate to 10%, and the intrinsic value falls to $90.91, a decrease of 4.55%. Now compare this to the intrinsic value of a $100 cashflow received in two years time using a 5% discount rate, which is $90.70. If the discount rate increases to 10%, the intrinsic value will fall to $82.64, a drop of 8.89%. The fact that the sensitivity of a cashflows intrinsic value to changes in the discount rate increases as the cashflow is received further into the future is why bonds with shorter maturities have a market price that is less sensitive to changes in the market discount rate. Typically, we quantify the temporal distribution of a bonds cashflows by calculating the bonds duration, which is the weighted average time to receipt of a bonds cashflows, with each weight calculated as the ratio of the discounted cashflow (estimated at the time duration is calculated) to the bonds intrinsic value at the time the duration is calculated. The duration of a bond will always be equal to or less than the bonds remaining time to maturity; duration and maturity are only equal in the case of a zero-coupon bond, which has only a single cashflow equal to the bonds face value. If this seems a bit complicated, the main points to keep in mind are that duration increases with higher yield17 and shorter maturities, and that the sensitivity of the bonds market price to changes in assumptions
16

Recall that a cashflow received in year N has an intrinsic value given by IV =

CF . If we (1+ DR) N

include a constant growth rate G for the annual cashflow, then the intrinsic value is given by:
IV = CF(1+ G) N . The sensitivity of intrinsic value to changes in the discount rate is then given as (1+ DR) N IV N * CF(1+ G) N , here the factor of N in the numerator dominates, and the sensitivity = DR (1+ DR)(N +1)

increases with increasing N. Similarly, the sensitivity of intrinsic value to changes in the assumed N 1 growth rate G is given by IV = N * CF(1+ G) ; again the factor of N in the numerator dominates,
G
(1+ DR) N

and sensitivity increases with increasing N, although here intrinsic value increases with increasing growth, whereas it decreases with increasing discount rate. 17 For a given rate return implied by an investments market and intrinsic value, a higher yield of is associated with a lower growth rate, which in turn implies cashflows weighted more towards the present.

Page 21 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

relating to future cashflows and the discount rate is roughly18 equal to the duration; i.e., if a bonds duration is 5%, a one percent change in the discount rate will result in close to a 5% drop in price.

1.4 Common Stock


Owning a share of a companys common stock gives you a share of the companys future earnings. Part of these earnings are typically paid out to you as dividends, while the remainder is re-invested on your behalf back into the companys business by the companys management. The proportion of earnings retained depends primarily on the opportunity to expand the companys business. For example, when Home Depot was initially expanding its market share, it retained almost all of its earnings in order to fund expansion; but when the company started to saturate the domestic market, expansion slowed, and the company started returning more cash to shareholders by increasing the dividend payout ratio and using more of retained earnings to repurchase outstanding stock with the repurchase increasing shareholder returns by dividing the companys earnings over fewer shares of stocks. In this case, the retention of a larger percentage of earnings earlier in the companys lifecycle enhanced shareholder wealth by allowing the company to grow earnings at a faster rate, which eventually resulted in higher dividends and a higher stock price. The future cashflows arising from owning a share of a companys stock consist of two parts: dividends and the price for which you sell the stock, and both of these components depends on the companys future earnings. Lets now take a look at some of the factors that can affect a companys future earnings. As with bonds, inflation has a large effect on the real value of a companys future earnings. Although inflation might cause the price of the companys products to increase, there is often an imbalance between this benefit and the effect of inflation on the cost of inputs such as labor, leasing costs, power, and raw materials, as well as the cost of debt capital. The net effect can be a large drop in earnings, or if the imbalance occurs in the other direction, a large increase in earnings; either way, the effect of inflation on the earnings of some companies can be significant, and lead to highly volatile earnings. On the other hand, these imbalances tend to average out over time, and in the long-term, the earnings of most companies tend to increase somewhat faster than inflation. Just as with a bond, a companys probability of default also impacts the expected value of future earnings. Once a company defaults on its debt, earnings fall forever to zero19. Aside from inflation and default, many other factors have an impact on a companys future earnings. If demand for a companys products or services falls, the company will need to cut prices to maintain volume, and revenue will fall, leading to a drop in earnings.
18 19

Due to the effects of convexity, this is really only true for small changes in our assumptions. Even if the company emerges from bankruptcy and earnings increase, as a shareholder of the bankrupt company, you will not own any of the new shares.

Page 22 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Even with constant demand, competition can lead to an increase in supply, which will also lead to falling revenue and earnings, although the effect is less for companies with a strong competitive advantage. With supply and demand constantly changing during the economic cycle, and the costs of labor, power, raw materials and debt capital in constant flux as well, and with many of these factors being interdependent (changing one can have an effect on others), it is pretty easy to see that forecasting a companys earnings with any degree of accuracy is futile. This uncertainty will exist no matter how well you understand a business; keep in mind that even a companys own management (who hopefully have a good grasp of their business) do not know precisely what their earnings will be over the next year; the best they can manage is a range of possible values. And the uncertainty just gets worse as you extend the time horizon. To compound the problem, if you are not planning on holding shares in a business forever, the sales price will affect your rate of return, and therefore even if you manage to get within the ballpark when estimating the businesss future earnings, how the market values these earnings at the time of sale will throw yet more uncertainty into your realized rate of return. To give an idea of how much this can vary, the price to earnings ratio (a measure of how much value the market places on a dollar of earnings) of the SP500 has ranged from a low of 6.5 to a high of 55 from 1950 to 2007. This is largely due to the extremely long duration of most stocks, which makes their market price very sensitive to changes in the market discount rate and expectations of future cashflows. So not only do you need to predict earnings, you also need to predict market sentiment, although fortunately the effect of market price diminishes with increasing holding period (because the cashflow arising from the sale of the share occurs farther into the future, it is discounted by a higher factor). Although in the limit all of these factors will affect the cashflows from a corporate bond as well, as a bondholder you are much more insulated from these effects. Assuming that inflation expectations are accurate, and a company does not default, you should receive all of the bonds cashflows and receive a rate of return equal to your discount rate. And because investment grade companies typically have high ratios of earnings to interest and maturing debt, the companys earnings can fall quite a bit without impacting the cashflows due to bondholders, whereas stockholders could at the same time experience a large negative return. As a simple example, consider a ten-year bond issued by a company with an interest coverage ratio of ten and maturing debt covered five times over by earnings, not an uncommon situation for a company with a high credit rating20. Here the limiting factor is maturing debt, and the companys earnings could fall at an annual rate of 17% for ten years and you would still receive your interest payments and the bonds face value. Unfortunately, shareholders would not fare so well. Lets assume that you purchase a share for $100 and the company starts out with a 50% dividend payout ratio, and is earning $10 per share. Here the dividend would need to be reduced after 4.4 years (and would probably be cut earlier), and even assuming the price to earnings ratio remains at ten (a very optimistic assumption for a company in this situation), the market price would fall to $21 at the end of the ten years. In reality, the price would be much
20

One reason cashflow uncertainty is higher for junk bonds is that there is much less insulation between a companys earnings and cashflows due to bondholders.

Page 23 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

lower, as the company would default in the eleventh year unless earnings started to increase again. Even if you decided to sell the shares earlier, you are almost certainly going to realize a large loss. So why would anybody in their right mind own shares when they could own bonds? Well, if instead earnings increased at 17% a year, the situation reverses. No matter how high the companys earnings grows, the bondholder just gets his coupons and face value, and will consequently realize a modest return. But assuming a constant payout ratio and price to earnings ratio, a shareholder would realize almost a 23% real rate of return. And since the market would likely assign a higher price to earnings multiple to a company with such rapidly expanding earnings, this actually understates the likely rate of return. Since we have made a pretty good case that the cashflow uncertainty from owning a stock is much greater than that of owning a bond issued by the same company, how do we go about estimating a stocks intrinsic value? As in the bond valuation example, we will start by choosing a discount rate equal to the asset classs long-term historical real rate of return. But since a share of common stock does not have a future series of nominal cashflows, but rather the cashflows will depend on the companys future earnings, we need to take a different approach with forecasting future cashflows and assume that a company with a given payout ratio will grow its earnings at a rate equal to the long-term historical average of similar companies. To illustrate, if we are estimating the intrinsic value of a large market capitalization domestic stock with a payout ratio of 50%, we would assume an earnings and dividend growth rate equal to the long-term historical average earnings growth rate of the SP500 index, which had a similar payout ratio from 1950 to the present. And if a company has a payout ratio differing from 50%, we make the additional assumption that the companys ratio of intrinsic value to demonstrated earnings power is independent of payout ratio; if a company pays out less of its earnings as dividends, the increased growth rate will exactly offset the lower initial dividend, and vice versa. Finally, to take into account a companys financial strength, we will adjust our estimated earnings growth rate based off of the expected annualized default rate and adjust our discount rate based off of the expected annual default rate volatility, using the expected default rate and default rate volatility implied by the companys long-term issuer rating. We need to keep in mind that this estimate of intrinsic value will rarely be accurate; the actual ex-post intrinsic value will often be much lower or higher than our estimate. The reason we are proposing such a simple approach is that it will likely minimize our prediction error (although it will still be huge) as opposed to other methods that attempt to use a detailed analysis of the company, industry, and economy. Although a more complex and detailed result might provide a more precise estimate, it will rarely provide a more accurate estimate; we will present an argument for this in Chapter 4. One of the benefits of our approach is that it forces us to acknowledge the likely magnitude of our prediction error, and therefore encourages us to stack the odds in our favor by purchasing shares only at a good discount to our estimate of their intrinsic value, and by focusing on companies that are financially sound, compete in stable and profitable industries, and have a strong sustainable competitive advantage. In short, we relegate a detailed analysis

Page 24 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

of a companys competitive advantage to the stock selection process, and leave it out of the stock valuation process. Later in this book, we will present a methodology for estimating a companys current earnings power, and then break down the historical price to earnings ratio of common stocks into two components, a discount rate and expected earnings growth rate. We will then modify these rates using a companys long-term issuer credit rating. The result is a valuation model that uses only two company specific inputs, an estimate of the companys current earnings power and long-term issuer credit rating, and two global inputs (based off of the long-term track record of common stocks): a discount rate and an expected earnings growth rate assuming a 50% payout ratio.

Chapter Summary
When we purchase an investment at a market price equal to our estimate of its intrinsic value, our expected return is equal to the discount rate used to discount each cashflow. If the ratio of estimated intrinsic value to market price is greater than one, our expected return will exceed our discount rate, and vice versa. Since it is logical to prefer a higher degree of certainty regarding a future cashflow, we should increase the discount rate as the level of certainty decreases; consequently, investments with less cashflow certainty tend to have higher expected returns. Investments with less cashflow certainty tend to have more volatile market prices Using a discount rate and future cashflow growth rate based off of long-term historical averages can keep us from getting carried away with market exuberance and pessimism, and can even enhance our expected return when we employ a margin of safety. Investments with lower durations have a market price that is less sensitive to changes in either the market discount rate or the markets expectations of the real value of the investments future cashflows, and duration decreases with increasing yield and decreasing maturity. There is much more uncertainty inherent in the cashflows arising from stock ownership as compared to bond ownership.

Page 25 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

2 Historical Evidence (2005)


In this chapter we are going to start off by looking at historical stock market returns, and the factors that led to these returns. We then ask ourselves what are the necessary conditions to allow us to achieve returns in excess of that obtainable by investing in a stock index fund, and then look at some historical evidence showing that this excess return is possible - even by blindly selecting stocks based off of a price to earnings ratio lower than the market average and paying no attention to the quality of the company issuing the shares. Next we look at the historical results of selecting stocks with below average valuation ratios from a set of above average companies, and then end by looking at the results of paying a premium for a group of companies thought to have exceptional future prospects. In addition to demonstrating that value investing has worked in the past, we will also use historical stock index data to determine some performance parameters (earnings per share growth rate and market discount rate) for the shares of an average company that will be used in the valuation model we will develop in Chapter 4.

2.1 Historical SP500 Returns


A look at historical stock market returns gives us some perspective when the stock market has recently had a run of unusually high or low returns. A common mistake made by investors is to extrapolate a short sample of past history into the future, either for an asset class or an individual stock; this tends to result in investors selling when they should by buying, and vice versa. So lets take a look at the historical rate of return that could have been realized by investing in an SP500 index fund, and then look at the factors that contributed to this rate of return. We will measure historical stock market performance using Robert Shillers data that combines the SP500 index data from 1926 to the present with the Cowles index from 1871-1925; this gives us 135 years of data. However, we need to keep in mind that prior to 1926, the earnings for many companies were not available, and were estimated from their dividends; therefore the earnings data from 1926 to the present should be considered more accurate. The reason we are focusing on the SP500 index is that it gives us a lot of data covering many economic cycles, and accurate data for other stock asset classes are much more limited; accurate foreign stock data starts in 1970, and Real Estate Investment Trusts (REITs) have only been around since the mid 1960s. In the future, I would expect an index tracking large market capitalization foreign stocks to have a risk-return profile similar to the SP500, provided that the index is weighted towards stocks issued by companies in countries similar to our own, i.e., democracies with strong property rights and a market based economy without excessive government regulation. As for REITs, they have in their brief history provided a total rate of return and rate of return volatility not too different from that of the SP500. Lets briefly discuss stocks with a small market capitalization. These have historically provided a slightly higher rate of return than

Page 26 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

owning stock in larger companies, but at the expense of over twice the rate of return volatility; moreover, some research has shown that a large portion of the excess return was due to the fact that they had on average a lower price to earnings ratio, which we will soon see is associated with above average returns. For this reason I usually avoid small cap stocks, preferring to focus on larger companies with a longer track record. Figure 2-1 charts the inflation-adjusted price, earnings, dividends and total return index on the primary (left) y-axis, as well as the forward price to earnings ratio on the secondary (right) y-axis. The price is the index value at the start of each year, while the dividends and earnings are the total for the year. The total return index shows the growth of $100 invested into the index at the start of 1871 assuming that all dividends were reinvested. Figure 2-121
Historical SP500 Index Performance 1871-2005
1000000 60

100000

50

Real Dollars per Share

1000

30

100

20

10

10

1991

1895

1879

1907

1967

1975

1887

1923

1931

1939

1943

1947

1955

1963

1971

1979

1899

1915

1871

1911

1919

1927

1935

1951

1959

1983

1883

1987

1995

1891

1875

1903

Year Real Dividends Real Earnings Real Price Real Total Return Index Forward PE Ratio

The first thing that probably strikes us about Figure 2-1 is the large amount of volatility in how much the market is willing to pay for a dollar of earnings, as shown in the price to earnings ratio time series. This is the factor that dominates returns over all except the longest holding periods. Also note that the price to earnings ratio exhibits a slight upwards trend; over time the market has been willing to pay slightly more for a dollar of earnings. This might be because the economy has become a bit more stable from 1950 to
21

Source: Book/AssetClassChapter/SP500_chart.xls

Page 27 of 156

1999

2003

Forward PE Ratio

10000

40

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

the present as compared to the period from 1871-1949; the consumer price index, real gross domestic product, and real SP500 earnings have all become less volatile, and the index contains companies representing a much broader range of industry than in the years prior to 1926. Or it might just be a spurious trend, although it does appear to be statistically significant22. Whether or not this trend of an increasing price to earnings ratio continues is anybodys guess, but I would not count on it. So what has the average annual return been from holding shares representing the SP500 index? The average real total return from 1871-2005 has been 8.23%, and has been quite volatile, with a standard deviation of 17.63%, and has varied from a low of 37% (1931) to a high of 51% (1933 & 1935). Before we get too carried away with this nice fat real rate of return, we need to realize that this is the arithmetic average, which (see Appendix 9.3) is always less than the geometric rate of return for any multi-unit holding period, and it is the geometric return that determines the rate at which your wealth increases. Using the total return index, we can compute the geometric rate of return using the index values at 2005 and 1871, which gives us a geometric rate of return equal to 6.77%. One thing we need to keep in mind is that this 6.77% return is partly due to a slowly rising average price to earnings multiple over the period from 1871-2005, and it would probably not be prudent to extrapolate this into the future, doing so is akin to assuming that in the future, somebody else will put more value on a dollar of the indexs earnings than you do today. We can remove the effect of a changing price to earnings multiple by adjusting the geometric rate of return (R) using the geometric growth in the price to earnings (PE) ratio ( GPE ) over the period as follows:
Radj =

(1+ R) 1 (1+ GPE )

Doing so we find that our real rate of return is about 6.42%. In Table 2-1 we repeat this process for several other historical periods, and find that there does not seem to be any sort of trend towards either higher or lower stock market returns, but we do see that higher adjusted returns are associated with a higher average earnings yield23 over the period. This makes sense, as when a stocks (or indexs) intrinsic value is a function of its earnings power, when the earnings yield is higher, we are paying less for a given amount of intrinsic value, and we should therefore expect a higher rate of return. We also see that the real earnings per share growth rate is highest when the dividend payout ratio is lowest; this also makes sense, as a lower payout ratio means that on average, the indexs companies had more retained earnings to invest for growth. Note that during the two earliest periods, the adjusted geometric return is greater than the un-adjusted return because the geometric growth in the price to earnings ratio was negative over the period.
22

To be precise, it appears that there is a 97% probability that the average PE from 1950-2006 being greater than the average PE from 1871-1950 is not due to chance, and there is a 91% probability that the difference in PE ratios being greater than one is not due to chance. 23 A stocks earnings yield is the reciprocal of a stocks price to earnings ratio, a more widely used measure of stock valuation.

Page 28 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Table 2-124
Period Ave. Return 8.23 7.72 8.73 8.66 8.08 SD Return Geo. Return 6.77 6.54 7.57 6.93 6.22 Adj. Geo. Return 6.42 6.78 5.63 6.18 6.98 Ave. Geo. EPS Earnings Growth Yield 6.75 1.82 7.91 1.48 7.23 2.12 7.44 2.06 7.99 1.61 Ave. Payout Ratio 0.62 0.70 0.50 0.57 0.71

1871-2005 1871-1926 1950-2005 1926-2005 1871-1950

17.63 14.94 16.08 19.28 18.71

Perhaps right about how you are thinking that this 6.4% historical real rate of return sounds a little low, and you would rather have a 50% or higher annual rate of return on your stock portfolio, so why not just create a portfolio of the next Microsoft, Wal-Mart, Home Depot, (you get the idea)? If you can do this, great, but I dont think you can. That said, I do believe it is possible for you to obtain returns consistently higher then that possible by purchasing index funds such as the SP500 (that is the whole point of this book). But for now, lets take a look at the factors that led to the historical rate of return that could have been realized by purchasing shares in an SP500 index fund. Factors Contributing to the Historical Rate of Return Up to now we have looked at the historical rate of return for the SP500 index over various periods. We need to keep in mind that the future does not necessarily have to follow the past, and even if it does, the historical rate of return has diverged markedly from its long-term trend, sometimes for fairly long periods of time. So lets look at what factors actually contributed to the historical rate of return, doing so will allow us to better estimate the future rate of return for stocks in general. During a given year, the annual rate of return is given by the difference between the indexs price at the end of the year and the indexs price at the start of the year plus the dividend received during the year, all divided by the indexs price at the start of the year, or R = (PEOY PSOY + D) /PSOY , which can be rewritten as the forward dividend yield at the start of the year plus the indexs price change during the year. So we already know that just two factors determine our annual return, the forward dividend yield when we purchase the shares in the index and the change in the indexs market value during the year. It turns out that this equation is also quite accurate for estimating the indexs internal rate of return over longer holding periods as well; the only inaccuracy comes from the fact that the indexs price and dividends do not grow at a constant rate over the holding period. Table 2-225
24 25

Book/AssetClassChapter/SP500_chart.xls Source: Book/AssetClassChapter/Regress/ReturnFactors

Page 29 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Holding Period R-Squared Yield Coefficient Price Growth Coeff.

1 1.00 1.00 1.00

5 0.99 0.89 0.96

10 0.98 0.85 0.90

15 0.97 0.82 0.86

20 0.94 0.89 0.79

This is shown in Table 2-2 where we performed a multiple linear regression of the internal rate of return vs. the dividend yield at the start of the period and the price geometric growth rate over the period. In this case, we are using multiple linear regression to find the coefficients ( 1, 2 ) and intercept that minimizes the mean square error for the equation:
IRR = 1 *Yield + 2 PriceGrowth +

An R-squared of 0.95 means 95% of the variation in the internal rate of return can be explained by a linear dependence on the forward dividend yield and geometric price growth rate. Note that for one year holding periods, the prediction is perfect, and this equation becomes IRR = Yield + PriceGrowth . Because price changes are so volatile, and get the most attention in media, most investors focus here, getting exuberant when prices are rising and pessimistic when they are falling. But price changes really cannot be predicted with any accuracy, except to say that the long-term price trend will follow long-term earnings growth. This long-term earnings growth is in turn loosely tied to economic growth, as world GDP grows, consumer consumption grows, and companies sell more of their products; this is partially offset by increased competition as new companies are created. This long-term rate of GDP growth has been fairly constant despite the many technological innovations over the last century, so it probably makes sense to assume a similar earnings growth rate going forward. It might be somewhat higher if on average companies retain more of their earnings, and vice versa, but since this will reduce the dividend payout ratio, the net effect to total return should not be significant. From Table 2-1, we see that if companies continue to pay out roughly 50% of their earnings as dividends, then a real earnings per share growth rate of 2% seems reasonable for the future, and since the indexs price will fluctuate around its earnings, this is also a reasonable estimate for the indexs long-term price per share growth rate. And since the index is a composite of a large number of companies, this also seems like a reasonable estimate for an average company as well. I would not count on any future innovations boosting this average growth rate, you can see from Table 2-1 that the earnings per share growth rate hasnt changed all that much since 1871 (the increase in the earnings per share growth rate is explainable by the decrease in the payout ratio), despite the fact that the index was composed of completely different industries, and dominated by railroads prior to 1926. I would guess that this historical earnings growth rate is more a function of how companies compete in a free market than dependent on the industry composition of the SP500 index. Granted, technological innovations can improve a companys productivity, but it will improve the productivity of any company that adopts it. So instead of cost savings falling to the bottom line, the competition will force the average company to cut its prices.

Page 30 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Perhaps you are wondering why historically, the indexs earnings per share growth rate has lagged gross domestic product (GDP) growth. After all, if a company maintains a constant market share, it seems reasonable that it can grow revenue at this rate, and with a constant ratio of costs to revenue, earnings as well. However, that is exactly the problem: on average, companies do not maintain a constant share of their industry. As GDP grows, so do the potential profits in different industries, and this brings in more competition. Only the exceptional company with a sustainable competitive advantage can actually hope to maintain market share in a growing economy. The other factor that determines our total return from investing in stocks is the forward dividend yield at the time we purchase the stock index. But a major problem arises when we attempt to use these two factors to predict future return although the forward dividend yield can be predicted fairly accurately at the time of purchase, the change in market price cannot, except to say that over long holding periods, it should roughly follow the growth in earnings. Nevertheless, we can say that on average, a company that pays out less of its earnings as dividends will tend to grow earnings at a faster rate than a company that retains more of its earnings, as the company will have more cash to reinvest in the business, repurchase stock, or both. This in turn implies an inverse relationship between dividend payout ratio and price growth rate. So although a higher dividend yield does contribute to higher returns, if the higher dividend yield is just due to a higher dividend payout ratio, the impact of the increased initial dividend will be offset by a lower expected earnings, and therefore price, growth rate. On the other hand, if the reason for a high dividend yield is merely a high earnings yield rather than an increased dividend payout ratio, then we might also expect a more typical price growth rate. Consequently, if we want to identify one factor that is known at the time we purchase a companys shares, and that should have the maximum impact on our realized rate of return, it would have to be the companys earnings yield. And since a companys earnings yield increases when the market values a stock using a higher discount rate, your best bet for achieving above average returns is to buy when pessimism is highest, and most everybody else is selling. In other words, you should get optimistic when prices are falling, and pessimistic when they are rising. To quote Benjamin Graham, be greedy when others are fearful, and fearful when others are greedy.

2.2 Market Efficiency


How well does the market determine a stocks intrinsic value? This is an important question, because if a stocks market value equals its intrinsic value at all times, then attempting to find stocks that are priced at a discount to their intrinsic value is futile, and we will be better off creating a portfolio of index funds. The version of the efficient market hypothesis that is the most widely accepted today states that the market does a good job on average of estimating a stocks intrinsic value, but that this estimate does have some amount of prediction error; consequently, at any given time, the markets estimate of a stocks intrinsic value may be high or low. The hypothesis also states that as Page 31 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

a stocks intrinsic and market value vary over time, they are on average equal, and have similar volatility, but are imperfectly correlated. In order for us to, on average, gain excess return by selecting a subset of stocks from a given asset class, we need to, on average, purchase a stock when the difference between its perfect foresight intrinsic value and market value is greater than zero. Earlier we argued that a stocks intrinsic value is a function of its future earnings, so if we want to increase the probability that this difference is greater than zero, we want to pay as little as possible for a companys earnings. The problem with this strategy is that when a stock has a below average price to earnings ratio, it is due to the belief that the company will have below average future cashflows, these cashflows will have a higher than average volatility, or both. And if we assume an efficient market, then on average, the market will be right, and when it is not, the stock has an equal chance of either performing better or worse than what the market expected. So if the efficient market hypothesis is correct, gaining excess return by purchasing companies with below average price to earnings ratios should not work. But what happens if a stocks market value is more volatile than its intrinsic value? In this case, a stocks market value and intrinsic value tend to move in the same direction, but the market value tends to overshoot the intrinsic value in both directions. So when a stock is expected to outperform its asset class, the stocks price increases on average more than the perfect foresight intrinsic value, and when a stock is expected to under perform its asset class, the stocks price will decrease on average more than the perfect foresight intrinsic value. In other words, the market tends to be overly enthusiastic about companies with above average prospects, and overly pessimistic about companies with below average prospects. Under these conditions, you should be able to obtain excess return by purchasing stocks with a price to earnings ratio below the market average and selling stocks with a price to earnings ratio above the market average. Note that it is not necessary to have any clue what a stocks actual perfect foresight intrinsic value is in order to obtain these excess returns; it is enough to, on average, know when a stock is priced at a higher discount to its intrinsic value than the asset class from which the stock is selected. Before we look at the evidence, lets first postulate why a stocks market value would be more volatile than its intrinsic value. One thing that comes to mind is that sometimes investors are heavily influenced by recent performance in their estimate of future performance; this has been observed at both the asset class level and individual stock level. Research has shown that after a period of above average stock market returns, most investors believe that future stock market returns will be above the historical average, while the reverse holds after a period of below average stock market returns. In the midst of a bull market, stock valuations seem to matter less as investors manage to convince themselves that prices will go up forever. Similarly, a stock that has had a nice record of above average earnings growth will typically command a higher than average price to earnings multiple, because most investors expect that performance to continue. But the historical record shows otherwise - unusually good or bad stock market returns revert to

Page 32 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

the mean, and there does not appear to be any correlation between a companys past and future earnings growth rate. One reason that the stock market is less efficient than the bond market is that a stocks rate of return is less bounded than a bonds. What I mean by this is that there are many stories of stocks that have rewarded investors with incredible rates of returns; early investments in Micro Soft, Wal-Mart, or Home Depot would have returned an investors initial investment hundreds of times over, whereas an investment in these companys bonds would have returned only slightly more than an investment in Treasury bonds. Stocks are also less bounded on the downside, as there are many more stories of stocks that have destroyed billions in shareholder value. I think that during a bull market, investors tend to focus on the unbounded upside (think Micro-Soft and Wal-Mart), whereas in a bear market, they tend to focus on the unbounded downside (i.e., Enron, World-Com). Another cause of market inefficiency is market overreaction to events that do not really impact a companys long-term future prospects. Consider Home Depot, which at the start of 2002 had a split adjusted price of $50, but which fell to $20 by 2003 because of a decline in same store sales caused by overzealous remodeling. This price decline occurred despite the fact that the company maintained double-digit earnings per share growth during the period. Soon the market realized it overreacted, and pushed the stock price back up to over $40 within two years. One reason that the stock fell so fast so quickly is that in the years leading up to the sell-off, the market had overreacted to the companys impressive earnings per share growth rate, and bid the stock up to a price to earnings multiple of 32. The market can be pretty fickle, changing its mind frequently about the same solid company. This overreaction is probably due to investors using a short time horizon to value what should be a long-term investment. We earlier mentioned that the market assigns above average price to earnings ratios to companies that are expected to experience above average performance, and vice versa. Over a short time horizon, this performance will be dominated by the change in companys share price. This means that if investors are estimating a stocks intrinsic value and expected return assuming a short holding period (say one year) they will be focused more on events that will likely change a companys price over the next year, instead of basing their estimate of intrinsic value on the companys likely long-term earnings power. Some examples of these events are the changes in short term interest rates, a rise or fall in the dollar, a short-term change in the companys performance (such as the example given for Home Depot), a war, or basically anything that might make the market more pessimistic or optimistic. In essence, what we have is investors valuing a stock based upon how they believe other investors will value the stock at the end of the year. To some extent, this short-term investing horizon is rational, as most fund managers (and pension managers) are evaluated on how well they perform relative to a benchmark, and this performance is graded on a pretty short timescale. Significantly underperforming a benchmark for even just one year can often be enough to cause the fund to start looking

Page 33 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

for a new manager. This tends to promote a myopic investing style, where a manager might be tempted to bail out of a stock at the first sign of trouble. Ironically, this focus on maximizing short-term returns typically results in below average short-term performance. While not all funds suffer from this (Dodge & Cox and some of the Vanguard managed funds are notable exceptions), most do. There is quite a bit of evidence showing that the market in general employs a short investing horizon. Just look at the number of analysts that are paid to predict what a stocks price will be in a year. And an interesting paper26 shows that the historical equity risk premium is explainable if we assume that investors evaluate their performance on an annual basis. For more theories on the causes of market inefficiency, see the suggested reading list for several books on behavioral finance. Regardless of the causes of excess market volatility, there is some compelling evidence27 that a stocks market value is more volatile than its intrinsic value. Some of this research attempts to estimate the intrinsic value of a stock index and compare it to the indexs price. One paper shows that stocks with above average price to earnings ratios tend to revert to a lower price to earnings ratio, and vice versa; this would seem to imply that when the market votes, it does so with excess exuberance and pessimism, but eventually correctly weighs the prospects of a given company. But why not get to the root of our question, and see if blindly selecting stocks with lower than average price to earnings ratios has provided above average returns, and if so, what has been the cost in terms of increased rate of return volatility.

2.3 The Low PE Strategy


Lets examine the historical excess return that could have been realized using a simple relative valuation investing strategy known as the low price to earnings (PE) ratio strategy. We would implement the strategy by selecting stocks with a price to earnings ratio lower than the asset class average. For example, each year we might sort all of the stocks listed on the NYSE into deciles28 (lowest PE in decile 10) based off of their price to earnings ratio, and each year, sell any stocks that are no longer in the upper N deciles, and purchase the stocks in the upper N deciles. We would choose N based off of how focused we want our portfolio to be. At one extreme we would set N to 1, and we would only purchase the 10% of stocks with the lowest price to earnings ratio; at the other extreme we could set N to 5, and purchase the 50% of stocks with the lowest PE ratio.

26 27

Myopic Loss Aversion and the Equity Premium Puzzle, Benartzi and Thaler Robert Shillers Market Volatility is a good collection of papers 28 In other words, if there were 10,000 stocks listed, we would sort them by their price to earnings ratio, and put the 1000 stocks with the highest price to earnings ratio into decile 1 and the 1000 stocks with the lowest price to earnings ratio in decile 10.

Page 34 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Table 2-3 tabulates the performance of each PE decile from 1952-200529. Shown in the first row is how we would have fared if we decided to index this asset class by purchasing shares in an SP500 index fund. While the SP500 index only contains a sub-set of the stocks listed on the NYSE, it does contain those with the larger market capitalizations, and since we are market cap weighting the returns from each decile, it is an appropriate benchmark. In the last two columns, we show the relative performance between the SP500 index and each decile. Table 2-330
Decile Index 1 2 3 4 5 6 7 8 9 10 Mean PE 16.6 35.3 21.9 17.7 15.2 13.5 12.1 10.9 9.6 8.2 4.3 Mean Return 12.24 10.87 10.74 12.45 12.55 12.71 14.93 16.24 17.84 18.61 20.68 SD Return 15.69 21.45 16.60 16.47 16.05 16.93 17.35 17.80 20.31 22.54 24.62 Geometric Return 11.15 8.69 9.46 11.22 11.36 11.41 13.61 14.85 16.04 16.48 18.15 Delta Geo. Delta SD Return -2.46 -1.69 0.07 0.21 0.26 2.46 3.70 4.89 5.33 7.00 5.76 0.91 0.78 0.36 1.24 1.66 2.11 4.61 6.84 8.93

We see that as the average price to earnings ratio of each stock group decreases, the geometric rate of return increases, while the rate of return volatility appears to increase as the average price to earnings ratio diverges from the average in either direction. This kind of contradicts the efficient market hypothesis, as we would expect stocks with higher PE ratios to be priced that way because of better than average future expected performance, but these stocks under perform the average from the standpoint of both rate of return and volatility31. You are probably wondering if the increased rate of return is statistically significant. The results are mixed when applied to the difference between adjacent deciles, (due to the small difference in mean PE between deciles) but become stronger when we look at the difference between the indexed returns and the more extreme (higher and lower) deciles. When we sort the stocks by quintiles rather than deciles, we find that the increased rate of return between adjacent quintiles become consistently statistically significant, while the increasing variance is only significant between quintile 1 & 2 (the highest PE quintiles).

29

Data from professor Frenchs website (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html), using the market cap weighted returns for each decile. 30 Source: /Book/ValueInvestingChapter/PE_deciles 31 This probably means it would not be a good idea to purchase shares of a growth index, as these tend to have above average price to earnings ratios.

Page 35 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

So where does the excess return in this strategy come from? One component comes from the fact that when PE ratios are low, we are purchasing more earnings for each dollar we spend. And since a stocks intrinsic value is a function of its earnings power, purchasing a stock with a low price to earnings multiple is equivalent to paying less for a dollar of intrinsic value, which boosts our expected return. Of course this is only true to the extent that the market is not perfectly efficient; if the market is at least somewhat efficient, then a lower than average price to earnings ratio would imply that a stock would have below average future cashflows. However, the historical data indicates that if this is the case, the benefits of paying less for a companys earnings more than compensates for any reduction in future cashflows. The other excess return component comes from mean reversion. Specifically, when a stock is priced at a higher or lower price to earnings multiple than the market average, it is because the market expects the stock to either outperform or under perform the asset class average. Since sometimes the market is wrong, and the difference in performance does not materialize to the extent that the market predicts, the price to earnings ratio will subsequently adjust, moving back closer to the mean. When the original price to earnings ratio was too high, this causes the price to fall, which decreases the realized return, while if the original price to earnings ratio was too low, the price will subsequently rise, increasing the realized return. Mean reversion might also be responsible for the observed increase in rate of return volatility for both high and low PE stocks. In Contrarian Investing, David Dreman notes that part of a stocks rate of return volatility is caused by earnings surprises, and that stocks with below average PE ratios tend to have little reaction to negative surprises (when reported earnings are less than analysts expect) but tend to experience large positive price movements after positive surprises (when reported earnings are better than analysts expect). However, this behavior is reversed for high PE stocks, where positive earnings surprises typically have little effect, and negative earnings surprises tend to create large negative price movements. If this is the case, then we can certainly live with a bit of increased volatility on the upside. Note that the excess return provided by the low PE strategy does not come from providing a more accurate estimate of a stocks intrinsic value than the market does, but instead identifies stocks that, on average, are priced at a greater discount to their intrinsic value than the asset class from which the stocks are chosen. In other words, it is a relative valuation strategy, and can provide excess returns (relative to the market) regardless of whether the market as a whole is priced at a premium or discount to its intrinsic value. At this point, you might be convinced that investing in stocks with lower than average PE ratios provides excess return, but perhaps you are wondering whether the excess rate of return volatility negates any benefit of the increased return. Although an in depth treatment of this subject is outside the scope of this book, I have run Monte Carlo simulations for a wide range of financial goals and found that when the excess rate of return volatility is less than 3 times the excess return, the trade-off is typically a good

Page 36 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

one, and you are improving the probability of meeting your goal32. Since we see in Table 2-3 that the increase in return and increase in rate of return volatility are roughly equal in magnitude for the groups with a lower than average PE ratio, the low PE strategy should add value regardless of your financial goal. Historically, it appears that the sweet spot for increased return for the smallest amount of increased volatility is in deciles 6-8. One possible reason for this is that some of the stocks with extremely low PE ratios are priced that way for a reason; for example, perhaps it is obvious that the company is headed for bankruptcy. Although in an efficient market this should not matter, because the market should price the stock taking into account the likelihood of bankruptcy, this is apparently not the case. In In Search of Distress Risk, a paper by John Campbell et al, they investigate the pricing of financially distressed stocks from 1963-1998, and find that these stock have historically provided below average returns with increased rate of return volatility. With this in mind, it seems like we might be able to stack the odds in our favor if we restrict ourselves to a universe of stocks with at least average financial strength. Lets test this hypothesis by looking at a different value investing strategy that focuses on companies that are a market leader in their industry and have above average financial strength.

2.4 The Dow 10 Strategy


The Dow 10 value investing strategy is implemented by purchasing the ten stocks in the DJIA industrial average with the highest dividend yield, and rebalancing annually. Lets also define the Dow 30 as an equally weighted portfolio containing all 30 DJIA stocks. The premise behind this strategy is that earnings can be volatile, and a companys earnings in any given year might not give that good a picture of the companys actual earnings power, but dividends are paid based off of the managements estimate of the companys future earnings power. Consequently, a higher than average (as compared to the Dow 30) dividend yield is an indication that the stock is trading at a higher discount to its intrinsic value than the Dow 30, and should therefore provide a higher expected return. Since the companies that comprise the DJIA are chosen because they are market leaders in their industry, and they tend to have above average financial strength, intuitively it seems that it is less likely that the reason these stocks are priced at a higher than average dividend yield is because the issuing company is headed for a serious decline. Since 1928, the Dow 10 strategy has provided a geometric rate of return 1.81% above that obtained by an equally weighted Dow 30 portfolio, and 2.57% above that obtained by indexing the SP50033. Moreover, the rate of return volatility of the Dow 10 portfolio was slightly less than that of the Dow 30 portfolio, and only slightly higher than that of the SP500. This is pretty remarkable, since we would expect that a portfolio of ten stocks to
32

More accurately, we can show that if the ratio of the increase in rate of return to the increase in rate of return volatility exceeds the slope of the efficient frontier at the optimal portfolio, then the benefits of the excess return outweigh the penalty of the excess volatility 33 Stocks for the Long-Run Jeremy Siegel

Page 37 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

have a much higher rate of return volatility than an index containing hundreds of stocks, even if we dont account for the fact that they are purchased because they are out of favor with the market.

2.5 The Nifty Fifty


Lets end by looking at the nifty fifty34, a group of companies in 1972 that were thought to have exceptional future prospects. This bright consensus on these companys future prospects translated into a very high price to earnings multiple, which averaged 41.9 in 1972. Later, the stock price of these companies got clobbered during the 1973-74 bear market, and again during the twin recessions of the early 1980s. Although these companies are often used as an example of why it is important to purchase companies at a reasonable price, the subsequent track record of these companies indicates that the high premium was justified. The sustainable competitive advantage of most of these companies allowed them to grow earnings at an annualized rate of 11% over the next twenty-five years, as compared to an 8% rate for the SP500. This translated into a 12.4% annualized total return, as compared to 12.9% for the SP500, which is pretty remarkable considering that the SP500 has a price to earnings ratio of only 18.9 in 1972. Clearly, if you had purchased shares in these companies sometime after 1972, when they were trading at a more reasonable price, the total return would have exceeded that of the SP500 by a large margin, probably something close to the 3% difference between the nifty fiftys and SP500s annualized earnings growth rates. The companies in the nifty fifty with the highest returns over the 25-year period were those in the consumer staples and pharmaceutical industries, which are both profitable and stable; the worst performers were technology companies, where the industry changes more rapidly, and therefore competitive advantages tend to be shorter lived. At any rate, these companies in aggregate demonstrate how a sustainable competitive advantage can translate into above average earnings growth.

Chapter Summary
We started out by looking at historical returns for the SP500 index, and isolated the factors responsible for those returns. Later, we argued that the stock market is likely not completely efficient, and presented evidence that purchasing stocks at a lower than average price to earnings ratio has historically provided higher than average returns with an acceptable increase in rate of return volatility. We then looked at the Dow 10 strategy and found that a value investing strategy that is focused on exceptional companies has produced excess return without any perceptible increase in rate of return volatility. This was followed by a look at the nifty fifty companies of 1972, which demonstrated how a sustainable competitive advantage translates into above-average earnings growth. We might conclude from these examples that a sensible investment strategy would be to limit our investment universe to stocks issued by exceptional companies, and then choose from
34

Data in this section is from Stocks for the Long-Run Jeremy Siegel

Page 38 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

that universe those stocks that are available at a lower than average price to earnings ratio. This type of strategy is also easier for the individual investor to implement (as compared to purchasing the lowest decile of the NYSE, which contains hundreds of stocks), as a diversified portfolio of just a handful of high quality companies purchased at a price to earnings multiple below the market average can provide a higher average rate of return than can be achieved by indexing, without incurring significant excess rate of return volatility. While this strategy would work well, I am more in favor of using an absolute valuation strategy, and only purchasing the stock of exceptional companies when they are trading at a discount to my estimate of their intrinsic value, regardless of their relative valuation as compared to their asset class. This means that in a bull market, when stocks on average are selling at a PE ratio of 30, I wouldnt necessarily consider a stock issued by an exceptional company priced at a PE of 20 to be a bargain; this gives me a margin of safety when the bull market ends. To estimate a stocks intrinsic value we first need to calculate a companys sustainable earnings power, and then plug this into a valuation model; these are the subjects of the next two chapters.

Page 39 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

3 A Companys Demonstrated Earnings Power (2009)


In this chapter we present a method to calculate a companys sustainable earnings per share. We start by removing non-recurring components to earnings, and learn how to use free cash flow as a check against both the adjusted earnings time series and common accounting tricks used to boost revenue and earnings. We end by developing a method to compute sustainable earnings per share by calculating the average of the de-trended adjusted earnings time series.

3.1 Adjustments to Reported Earnings


A companys reported earnings per share is not always a good indication of a companys actual earnings power, as it often consists of one or more temporary earnings components, and can also fluctuate with the business cycle. Perhaps you have seen a company report a large drop in earnings, and wondered why the stock price did not change all that much - the reason might have been that these reported earnings were not a good indication of the companys current earnings power. This discrepancy between a companys reported earnings and current earnings power can stem from many sources; but the three largest contributors are non-recurring charges to earnings, the write-down and amortization of intangible assets, and a companys business cycle. Here a nonrecurring charge is defined as an expense or source of income that is not likely to be consistently repeated in the future. Although this chapter will illustrate several sources of non-recurring components to earnings, the examples are by no means comprehensive. In the future, we will refer to our estimate of a companys demonstrated earnings power over a full business cycle, adjusted for non-recurring items, and with any underlying growth trend removed, as sustainable earnings. We will calculate a companys sustainable earnings per share so that it represents an estimate of a companys earnings power projected one year into the future (as opposed to the companys current earnings power); the reason for this will soon be apparent when we discuss the details of our valuation model in the next chapter. Non-Recurring Components to Earnings Asset Sales One example of a non-recurring earnings component that is when a company sells a subsidiary for a profit or loss; obviously this is not sustainable, as the company would be slowly liquidating itself over time. Using the same argument, we should also consider the earnings of a subsidiary sold in a given year to be a non-recurring component to that years earnings. Fortunately, both the income from and profit from the sale of a subsidiary are identified in the income statement as discontinued operations. However, Page 40 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

this is only the case of the entire business unit is sold - if a product line within a business is sold, the income from that product line and any profit from the sale are included in earnings from continuing operations, and you need to make an adjustment to earnings from continuing operations. To illustrate, in 2006 3M sold their pharmaceutical product line. But since this was just one product line within the companys health care business unit, both the income and profit from the sale of the business unit were included in continuing operations. On the other hand, if a company has a subsidiary that is in the business of selling assets, then the gain (or loss) on such sales should be considered recurring. An example of this is an REIT with a development subsidiary that develops new properties with the intent of selling them once they are stabilized with new tenants. Although this type of income is much less stable than rental income, it is recurring. Changes in Accounting Principles Some examples of changes in accounting principles are a change from LIFO to FIFO inventory reporting or a change in how capital expenditures are depreciated (i.e., straight line to accelerated). Although technically non-recurring, I usually ignore these, unless they have a significant impact on the stated earnings. Sometimes Generally Accepted Accounting Principles (GAAP) itself changes, as when companies were no longer required to amortize goodwill after 2002, but instead periodically test acquisitions for asset impairment. In this case, if the company did not restate their previous financial statements to account for this in their ten-year financial summary, I would make the adjustments myself, as they are a specific case of the amortization of intangible assets, which we will discuss shortly. Extraordinary Events Under GAAP, an extraordinary event must be both infrequent and not related to the companys normal course of business. By definition, these are non-recurring, and earnings should be adjusted to factor out the financial impact of these events. Sometimes a company might be overly conservative, and not classify an event as extraordinary when it should be (at least in my opinion); an example of this is when UST inc. lost an antitrust lawsuit and had to pay over $1B in damages; although not explicitly identified as an extraordinary item on the companys income statement, I considered the event extraordinary enough to add back in the litigation expense to the companys income. Restructuring Charges When a company acquires another company, merges with another company, or just reorganizes internally, the company must estimate the future expenses associated with these changes, and charge them immediately to earnings. Although these charges might be considered non-recurring (unless the company is a serial acquirer) I prefer not to adjust reported earnings by factoring back out the full amount of restructuring costs. Instead, I use the notes to the consolidated financial statements to determine the amount Page 41 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

of restructuring costs that actually resulted in a cash outlay in the current year, and add the remainder back to income from continuing operations. In subsequent years, when the restructuring reserve is reduced and cash paid out for restructuring costs, I subtract these cash outlays from income. Any portion of restructuring costs that are used to account for asset impairments are added back to net income. It looks like FASB is moving to a similar accounting method, where restructuring expenses would be charged to earnings as incurred. The reason I treat cash restructuring expenses as business as normal, even if they occur rarely, is that they will hopefully result in future cash savings, and a boost to earnings (that is usually the point of restructuring). So in this respect, it is no different than a company investing in a new factory. Changes in Loan Loss Reserves Another example of a charge to earnings that should be considered recurring is changes in a banks loan loss reserves. This reserve (a liability) gets debited whenever a bank charges off a loan, and gets credited by the provision for credit losses recognized on the income statement. If a bank expects loan defaults to increase, then it might decide to take a larger provision for credit losses in order to increase loan loss reserves. Although changes in loan loss reserves tend to be cyclic, since these charges represent managements best estimate of either an increase or decrease in future cashflows due to changes in loan charge-offs, they should be considered recurring, and not adjusted for. Write Down of Intangible Assets Whether the asset is tangible (such as the case of a plant closure) or intangible (as when an intangible asset is written down below its carrying value), if the charges are both infrequent and small with respect to operating income, I usually add these back to pre-tax earnings. My rationale is that the required charge to earnings does not require a future outlay of cash. But if these charges are both large and frequent, I would not want to own the companys shares. Eastman Kodak is a good example of a company where you would have been well advised to treat the write-down of assets (which were the main component of restructuring charges) as a normal component of earnings. Here, restructuring programs were basically business as usual, and were used by management to obscure the companys diminishing earnings power, hoping that investors would focus on the companys adjusted earnings where these charges were added back in.

Amortization and Write-down of Intangible Assets Purchase Accounting When a company acquires another company or purchases a business line from another company, it must determine the value of the purchased entity so that it can adjust its Page 42 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

balance sheet to reflect the purchase. In doing so, the value of the acquired entity must be attributed to three parts, identifiable tangible assets, identifiable intangible assets, and goodwill. Some examples of tangible assets are cash, inventory, property, and equipment; what these have in common is that they either have an obvious market value (in the case of cash and securities) or they can be readily appraised to determine their market value. Intangible assets often take the form of brands, trademarks, patents, and in-process research and development. Here the value is less obvious, and is usually estimated as the sum of the discounted cashflows that the asset is expected to generate, with the estimate of future cashflows likely based on how much cash the asset generated for the acquired company. When a company acquires another company or business line, it is often necessary for a company to pay more than the market value of the acquired entitys total identifiable assets, both tangible and intangible. When this occurs, the difference between the purchase price and these identifiable assets is attributed to goodwill. Why would a company pay more for a company than the estimated value of the companys assets? One possibility is that the acquiring company might have a distribution network that could dramatically expand the potential market for the acquired companys brands. In this case, over time the acquired brands could end up providing much more cash than they did prior to the acquisition. Note that intangible assets can also arise from the internal development of software. Amortization of Goodwill Prior to 2002, goodwill was amortized based off of the estimated lifetime for the intangible asset; this meant that a fraction35 of the intangible assets value would be deducted from the companys pre-tax earnings each year. In 2002 the FASB changed the rules of Generally Accepted Accounting Principals (GAAP) to discontinue the amortization of goodwill; instead, the companys intangible assets must be regularly reevaluated to insure that the original assumptions regarding the cashflows that the goodwill asset is expected to generate are still reasonable. If the company determines that the sum of the discounted cashflows are likely to be lower than the original estimate, then the company must reduce the carrying value of the goodwill asset on the balance sheet, and record a charge to pre-tax earnings equal to the difference in carrying value. I add back the amortization of goodwill to pre-tax earnings. My reasoning is that goodwill does not represent future cashflows that must be paid by the company, as is the case with restructuring charges. Amortization of Other Intangible Assets Intangible assets with definite lives some examples being software, patents, and some brands are amortized over this lifetime. Since the asset has a definite life, the cashflows it generates will decrease over time, and require a cash outlay to replace it at some point. In this sense, it is no different from the amortization of tangible long-lived assets such as
35

For straight-line amortization, the fraction would equal the assets basis (original amount recorded in the balance sheet at the time of purchase) divided by the assets estimated lifetime

Page 43 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

manufacturing equipment. For this reason, I do not add back the amortization of intangible assets with definite lives to pre-tax earnings.

In Process Research and Development One type of intangible asset that is not amortized is in process research and development. When a company acquires another, it might decide to allocate part of the purchase price to the present value of the future discounted cashflows that it estimates it will receive from research and development that was in progress at the acquired company. When this occurs, the present value of these future cashflows must be immediately deducted from pre-tax earnings. The reason that the in process research and development is not amortized in GAAP is that in general, the future cashflows expected from research and development are too hard to quantify. However, if a company pays a reasonable price for the cash flows that the acquired entity is currently producing, I like to add charges due to in process research and development back in to pre-tax earnings. Note that if a company has a history of over-estimating the future cashflows associated with acquired intangible assets, a decline in the companys return on assets will usually be apparent. Adjusting Reported EPS The starting point for our calculation of a companys sustainable earnings per share is the companys reported earnings per diluted share from continuing operations available to common shareholders (i.e., after any preferred dividends are deducted); it is from this number that we will make adjustments for non-recurring charges and the amortization and write-down of intangible assets. Diluted earnings per share divides the companys earnings by the potential number of outstanding shares if outstanding in the money stock option grants and convertible bonds were immediately converted to common shares, whereas basic earnings per share divides the companys earnings by the number of shares currently outstanding. I like to use diluted earnings per share, as it is a more pessimistic measure of a companys per-share earnings power. We need to keep in mind that most non-recurring charges are reported pre-tax, and since earnings is an after tax measure of profit, we need to adjust for this. We can convert a pre-tax charge to an after tax charge by multiplying it by one minus the companys effective income tax rate. If they are not specifically identified in the income statement, non-recurring earnings components can sometimes be inferred from the statement of cashflows, where cashflow from operations is reconciled to reported earnings (including discontinued operations) by listing all contributors to earnings that do not actually contribute to the companys cashflow in the year being analyzed. Not all of these items should be considered nonrecurring, for example, depreciation, which is included in this reconciliation, should be considered a real expense in that it will approximate the companys required capital

Page 44 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

expenditures. Also, since the statement of cashflows includes discontinued operations, but our starting point for sustainable earnings is income from continuing operations, be careful not to use any non-recurring items arising from discontinued operations found in the statement of cashflows to adjust income from continuing operations.

3.2 Free Cash Flow


Since deciding which components of earnings are non-recurring might seem like more an art than a science, it would be nice to have a sanity check for these adjustments. I like to use a companys free cash flow for this purpose. A companys free cash flow is calculated by subtracting recurring capital expenditures from the companys adjusted operating cash flow. These recurring capital expenditures are necessary to maintain both the companys productive capacity and competitiveness, Capital expenditures are found in the companys statement of cash flows under cash flows from investing activities, where they are usually referred to as additions to property, plant, and equipment, although sometimes they are referred to as capital expenditures. Unfortunately, a company will rarely (an exception being REITs) break down capital expenditures into those required to maintain the companys productive capacity such as the replacement and upgrading of plant property and equipment - and those used for expansion; although for certain industries we can sometimes infer what this might be. Later in this section we will present a method for estimating a companys ratio of recurring to total capital expenditures, but lets first take a look at some adjustments we need to make to a companys reported cash flow from operating activities. Adjustments to Reported Cash Flow from Operations As with reported earnings, we need to make some adjustments to reported cash flow from operations, both to remove some non-recurring components and to correct for some limitations of GAAP reporting concerning cash flow treatment of income taxes paid. To begin, we want to remove cash flows resulting from discontinued operations. Most companies make this easy by specifically breaking out operating cash flows from discontinued operations in the operating section of the cash flow statement. Since the starting point for the derivation of operating cashflow is net income, we also need to subtract any preferred dividends. We also want to remove non-recurring items found in cash flows from operating activities such as gains and losses from litigation and gain on the sale of investments held as trading securities36. Other types of non-recurring gains that resulted in adjustments to reported earnings such as gains on the sale of businesses or fixed assets are already backed out of earnings in the statement of cash flows, and therefore do not require any additional adjustments, except to account for the tax effects of gains or losses.

36

Unless the company is a financial company, in which case these gains could be considered business as usual.

Page 45 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

The reason we need to adjust for taxes on gains or losses from the sale of operating assets is that GAAP requires all cash inflows and outflows related to income tax to be captured in cash flows from operating activities, even if the actual gain is reversed out of earnings in the reconciliation of net income to cash flow from operating activities in the operating section of the statement of cash flows. So in the case of a net gain on the sale of a business, the gain would not result in an operating cash flow, but would result in a decrease in operating cash flow due to the taxes on the gain. To remove this bias, we need to multiply the gain (or loss) by the effective tax rate, and then add back the result to cash flow from operations. We need to perform a similar adjustment to account for gains (or losses) on the sale of discontinued operations. The last major adjustment to operating cash flow is to account for the true cash cost of exercised stock options. Currently, the Black-Scholes method is used to estimate the likely future cost of outstanding stock options, which is then deducted from pre-tax earnings. We measure the actual cash cost by assuming that a company will re-purchase any new stock that is issued as a result of exercised options, to avoid dilution. The cash cost of the exercised options is therefore calculated as the difference between the weighted average market price (which we assume is the repurchase price) and exercise price, multiplied by the number of shares exercised during the year. The amount of exercised shares as well as the exercise price can be found directly in the notes to the financial statements, but sometimes the weighted average repurchase price is not directly reported. However, it can be found by dividing the total cash used to repurchase shares (in the financing section of the statement of cash flows) by the total shares repurchased (found in the consolidated statement of changes in shareholders equity. Recurring Capital Expenditures If a company does not break down capital expenditures into recurring and non-recurring components, the average ratio of recurring to total capital expenditures can be estimated by analyzing the companys organic37 volume growth over a ten-year period. For Diageo, a company that sells wine, beer, and spirits, organic volume is given in the annual report as cases shipped adjusted for the effects of acquisitions. We estimate the ratio of recurring to total capital expenditures over a ten-year period because the delay between a capital expenditure made with the purpose of expanding production and the resulting increase in volume can vary. Consequently, the relationship between capex and production measured over a single year may not give much insight. In our analysis, we assume that any percentage change in organic volume requires a corresponding percentage change in manufacturing capacity. Moreover, this percentage change in manufacturing capacity must be matched with a similar percentage change to property, plant, and equipment (PP&E). It follows that we can impute a value for PP&E at the

37

We use organic volume growth, otherwise acquisitions (particularly if late in the ten-year period) can distort the link between total capital expenditures over the period, production and manufacturing capacity.

Page 46 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

beginning of our measurement period by dividing the companys current net38 PP&E by one plus the percentage increase in organic volume growth over the period. The difference between the companys current PP&E and the imputed start value is an estimate of the cumulative capital expenditures used to expand the companys productive capacity over the period. Note that the actual PP&E at the start of the measurement period will often differ from the imputed value of PP&E at the start of the measurement period. One reason for this is that PP&E sometimes increases because the minimum capex required for maintaining production volume has consistently exceeded depreciation; this would make the imputed PP&E at the start of the period greater than the actual net PP&E at the start of the period. Another is that productivity improvements can sometimes increase volume without additions or improvements to PP&E; this would make the imputed PP&E at the start of the period less than the actual net PP&E at that time. Our next step is to allocate a portion of the total capex equal to the difference between current net PP&E and the imputed starting value as accruing towards expansion, with the remainder being recurring. The procedure for estimating the amount of capex used for expansion over the time period is shown in the following equation, where PP & E is current net PP&E, VCurrent is the companys current production volume, and Vstart is the companys production volume at the start of the time period adjusted for acquisitions.

Capex Expansion

PP & E VStart = PP & E = PP & E 1 VCurrent /VStart VCurrent

Using the second form of the equation, we see that capital expenditures attributed to expansion increases with increasing volume over the period, and if volume is flat, the equation results in none of the capital expenditures being attributed to expansion. We can then estimate the companys ratio of recurring capex to total capex over the period as:

CapexTotal Capex Expansion CapexTotal


I compared this estimate to two companies that disclosed their estimate of recurring capital expenditures, General Electric (industrial operations only) and Wal-Mart. In the case of General Electric in 2008, the estimated ratio of recurring to total capital expenditures was 83% as compared to a ratio of 70% disclosed in their 2008 10-K. For Wal-Mart, using data through FY 2009 resulted in an estimated ratio of 55% as compared
38

Less accumulated depreciation. We use net PP&E for an estimate of the replacement value of PP&E because accumulated depreciation is necessary to offset the value of capital expenditures necessary to maintain competitiveness. For example, consider a factory purchased for $1B, a depreciation lifetime of 20 years (so annual depreciation is $50M), and ongoing capital expenditures equal to depreciation. After 10 years, the total cost of the factory plus accumulated recurring capital expenditures would equal $1B + (10 years)($50M capex / year), or $1.5B, and it would be carried on books at $1.5B less accumulated deprecation of $500M, or $1B. So the book value (net PP&E) does approximate replacement cost except for the effects of inflation.

Page 47 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

to a disclosed ratio of 66%. Part of the error for Wal-Mart is due to the abrupt slowdown in expansion; this made the historical relationship between recurring capex and total capex less indicative of the current situation. Still, it is pretty close. Going back to FY 2008, our equation gives a ratio of recurring to total capex of 50% as compared to WalMarts stated ratio of 52%, and going back to FY 2007, our equation gives 46% as compared to Wal-Marts stated figure of 49%. Table 3-1 illustrates our estimating of the ratio of recurring capital expenditures to total capital expenditures for Consolidated Edison (ED), Southern Company (SO, General Motors (GM), The Coca-Cola Company (KO), and 3M (MMM). For the two utilities, we estimated productive capacity by adjusting revenue by the urban electricity and gas index, a component of the consumer price index (CPI). For example, using 10 years of data, if 2008 revenue was $17,127M and the ratio of the urban electricity and gas index in 2008 to the index in 1999 was 1.68, we assumed that productive capacity in 2008 was $10,194M, i.e., the rest of the change in revenue was due to price increase rather than and increase in productive capacity. Although both utilities had a different mix of electricity and gas revenues, the price of electricity and gas has similar increases over the period. For Wal-Mart, we assumed that revenue growth is a good proxy for the growth in productive capacity, and consequently price increases were an insignificant contribution to revenue growth, which seems to be the case as the growth in the number of stores has only lagged revenue growth by 2%. This is small enough difference that it could be attributed to an increase in Super Centers as a percentage of total stores, which would increase the revenue per store. For General Motors, we could not find consistent manufacturing volume, so we used unit sales as a proxy. Note that the Southern Company has a ratio of recurring to total capital expenditures of greater than one; this can be interpreted as meaning that the capital expenditures have not been sufficient to maintain productive capacity. Table 3-1
As of end of FY08 for ED, SO, GM, KO, MMM, and TM; and FY09 for WMT Company ED SO GM TM KO MMM Industry Utility Utility AutoMfg AutoMfg Beverage Conglo. Net PPE 20,874 35,878 41,877 4874 8326 6886 Years of Data 10 10 10 10 10 10 Org. Volume Change 1.08 0.88 0.96 1.91 1.83 1.42 Rec. Capex / Capex 0.88 1.20 1.01 0.77 0.76 0.81 Rec. Capex / Depr 2.38 2.41 1.71 1.14 1.12 0.83 Rec. Capex / Capex is the sum of recurring capital expenditures to the sum of total capital expenditures over the period. WMT Retail 95,653 10 1.97 0.55 1.43

The average ratio of recurring to total capital expenditures over the ten-year period can be used to estimate recurring capital expenditures in a given year, which in turn can be used to calculate free cash flow in that year39. Free cash flow (FCF) in a given year i in a
39

An alternate approach would be to estimate capex used for expansion in each year as the percentage change in volume over that year multiplied by PP&E at the end of the year, but using

Page 48 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

period of N years is calculated as the adjusted cash flow from operations in year i ( AdjCashOpsi ) less reported capital expenditures in year i ( TotalCapex i ) multiplied by the ratio of the sum of recurring capital expenditures over the period to the sum of total capital expenditures over the period:
N

RecurringCapex
k=1 N

FCFi = AdjCashOpsi TotalCapex i

TotalCapex
k=1

Comparing Free Cash Flow To Adjusted Earnings

Because free cash flow tends to be quite a bit more volatile than earnings, I like to compare free cash flow to earnings over a ten-year cycle. One way to do this is to calculate the sum of the companys free cash flow over the period, and divide it by the sum of the earnings over the period. Before computing the sum, I adjust both the earnings and free cash flow time series to state them in todays dollars; this corrects for the effects of inflation. Over a sufficiently long time period (say ten years), we should expect to see free cash flow roughly equal to earnings. If there is a significant difference, then it should be investigated. Sometimes the reason is benign, such as when a company is investing heavily to finance rapid expansion of the business. In this case, the rapid expansion of working capital can create a drag on operating cash flow. Rapid expansion can also cause depreciation expense to significantly lag capital expenditures. Other times, the reason is systematic to the industry; utilities tend to have a large and long-lived asset base, and because deprecation expense is not indexed to inflation, it tends to cause depreciation to underestimate recurring capital expenditures. For this reason, a utilitys reported earnings are typically a bit optimistic. At the other extreme are REITs, whose reported depreciation overestimates their recurring capital expenditures (since their assets tend to appreciate when properly maintained). Sometimes the difference between free cash flow and earnings can point out problems in the quality of a companys reported earnings that might make you think twice about investing in their stock, such as when free cash flow is much less than earnings (or non-existent) because the company is slowly liquidating itself by selling assets. But often the differences between free cash flow and earnings just help to highlight non-recurring components to earnings, which helps us create a more accurate estimate of a companys demonstrated earnings power.

the ten-year average ratio of recurring to total capex to compute each years free cash flow gives more accurate results, particularly when volume is volatile, or we need to estimate volume changes from revenue. And as we mentioned previously, the relationship between capacity enhancing capex and volume increases is not always apparent over one-year measurement periods.

Page 49 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

A Quick Look at Creative Accounting Practices Comparing free cash flow to adjusted earnings over a ten-year period can sometimes reveal aggressive (or even fraudulent) accounting practices. Some common forms of aggressive accounting, and how they affect the relationship between adjusted earnings and free cash flow, are given below40: Premature & Fictitious Revenue: When revenue is reported but not collected, it will increase accounts receivable. This will in turn reduce operating cash flow, eventually causing earnings to exceed free cash flow. Comparing the ratio of accounts receivable to revenue and noting any large deviation from the trend can also detect this. Often aggressive revenue recognition is also described in the companys accounting policies in their form 10-K. Cookie Jar Reserves: Sometimes a company will intentionally over reserve for restructuring with the intent of boosting subsequent earnings by funding operating expenses through these liability accounts. This should cause a corresponding decrease in operating cash flow as the liability account decreases. Capitalizing Operating Expenses: Capitalizing operating expenses into PP&E can be detected when it causes free cash flow (which takes into account the true cash outlay for these capitalized expenses) to lag reported earnings. Aggressive Capitalization: Amortization periods in excess of an assets actual expected life, as well as optimistic assumptions regarding residual value can boost reported earnings. This is detectable by earnings in excess of free cash flow, which takes into account the true cash outlays for capital expenditures. Hiding Non-Recurring Sources of Income: Sometimes non-recurring contributions to income are not explicitly stated in the financial statements, managements discussion and analysis, and footnotes. However, they will cause reported earnings to exceed free cash flow, which must exclude gains from asset sales, mark-to-market gains, and other non-operating sources of income. One exception is trading securities, where gains and losses are treated as operating cashflow. However, these are typically explicitly identified in the cash flow statement.

3.3 A Companys Business Cycle


Even without the effects of non-recurring charges to earnings, a companys reported earnings in a given year can be a misleading indicator of the companys earnings power
40

A more in-depth analysis of this issue can be found in a couple of books by Comiskey and Mulford, The Financial Numbers Game and Creative Cash Flow Accounting

Page 50 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

when the companys earnings growth is cyclic. The earnings of many companies will tend to follow an industry specific or economic cycle, where profits are lower at the bottom of the cycle and higher at the top; this can sometimes obscure the companys underlying earnings growth rate (with the effects of the cycle removed). For example, at the peak of the cycle, a companys earnings per share might be $5.00, but it might be a bit optimistic to value the company based upon this level of earnings, as they might be considerably reduced by the time the cycle bottoms out. Lets look at an actual example. Figure 3-1 illustrates the earnings per share of Caterpillar (CAT), with non-recurring items removed, and adjusted for stock splits, as well as revenue per share, price per share, and the price to earnings ratio. The period covers several economic cycles, and although the overall earnings growth trend is positive, the companys earnings tend to rise and fall with the economy, bottoming out during recessions and peaking during booms. It should be clear that even in the absence of non-recurring charges, a companys adjusted earnings at either the peak or trough of its business cycle is probably not a good indication of the companys earnings power, whatever the companys underling earnings growth rate might be. Figure 3-141
CAT
95.00 85.00 5 75.00 65.00 55.00 3 45.00 35.00 25.00 1 15.00 5.00 0 2 4 6

Price per share, Revenue per Share & PE Ratio

1986

1989

2006

1988

1991

1985

1993

1998

1992

1994

1996

2002

1997

1999

1995

2000

2001

2003

2005

1990

1987

End of Year Annual Earnings per Share Price to Earnings Ratio Annual Revenue per Share Price per Share

Sales of Caterpillars products primarily mining and construction equipment are correlated with the economic cycle, and the nature of this industry is that it is not possible to quickly adjust costs to match changes in demand. This results in earnings volatility that is much higher than revenue volatility. This volatility often results in a situation where a companys shares are a better value at a high price to earnings ratio than at a low price to earnings ratio. For example, at the end of 2003 Caterpillars price to earnings
41

/Book/ValueInvestingChapter/CAT_sustainable_earnings.xls

Page 51 of 156

2004

2007

2008

Earnings per Share

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

ratio was 26, but only because the companys earnings were temporarily depressed. A similar situation occurred during the recession of 1992, where the companys earnings completely disappeared. Both of these times were, in retrospect, a good point to purchase shares. At the other extreme, the companys price to earnings ratio was 12 at the end of 2006, but only because earnings were at a cyclical high. In retrospect, despite the relatively low price to earnings ratio, this was not such a good time to buy. Not shown on the chart is what is likely to unfold in 2009, where the company expects revenue to fall by over 30%, and earnings per share (adjusted for redundancy costs) to fall from $5.66 to $1.25. So how can we account for the companys business cycle? First of all, we are going to need more than this years earnings to do this; ideally we would want enough years of adjusted earnings to cover a complete business cycle. One problem with this is that a business cycle is not apparent until it is over, and even then identifying it can be a bit subjective. To illustrate, how can we be sure that the year 2005 in Figure 3-1 is actually the peak of a business cycle? Without knowing this, it is not obvious how many years of history we need to capture the business cycle. A partial solution to this problem is to use either ten years of adjusted earnings data or enough years to capture at least two peaks or troughs in the companys business cycle. Referring again to Figure 3-1, ten years would suffice, ten years would suffice, as it captures a trough in the earnings time series in 1992 and a trough in 2002. Be aware that not all business cycles track economic cycles; one way to make sure that you have captured the business cycle of a company that produces a commodity such as oil or aluminum is to look at a history of the commoditys price. Once we have gathered at least ten years of a companys earnings with non-recurring charges removed, and we believe this captures a full business cycle, how do we now determine the companys earnings power? Can we just average the ten years of earnings? If the earnings growth rate underlying the cyclic component of earnings is zero, then this approach would work fine. But if the company has a positive underlying earnings growth rate, then this approach would be pessimistic, in that earnings would be reduced more and more by this underlying growth rate as we go back in time. If this is not clear, then picture a company that has no cyclic component to its earnings, but has grown earnings at exactly 10% a year over the last ten years, and the earnings per share expected during the most recent year was $1.00. Now our intention is to make our estimate of a companys earnings power independent of the companys historical earnings growth rate42; therefore this company should have the same sustainable earnings per share as another company with earnings per share of $1.00, but that has had no earnings growth (and also no cyclical component to earnings) over the last ten years, as the only difference between these two companies is that the first company has substantially expanded its business over the last ten years. Yet averaging the first companys earnings per share over ten years would give sustainable earnings per share of only $0.68, while the second companys sustainable earnings per share would be $1.00. Surely we do not want to penalize the company that has grown its business over the last ten years, yet that is exactly what happens if we just average ten years worth of data.
42

This is because there is no evidence that past earnings growth rate trends persist into the future

Page 52 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Evidently, we need to somehow remove the underlying earnings per share growth rate from the earnings per share time series before taking the average of the time series. This is known as de-trending the earnings per share time series. De-trending a time series of earnings with a constant earnings growth rate trend equal to G is accomplished by multiplying each value in the original time series by (1+ G) N , where N is the difference between the current year and the year corresponding to the earnings value in the time series. Going back to our example of a company that has grown earnings per share at 10% over the last ten years, we would multiply last years earnings by (1+10%), the year before that by (1+10%)2, and so on. If we do this for the company in our example, and then take the average of the de-trended time series, we find that the companys sustainable earnings per share would be $1.00, the same as the company with zero earnings growth. So far this sounds pretty easy, the catch is that a company rarely grows its earnings at a constant rate, and the situation is further complicated because there is almost always a cyclic component to the earnings history as well. One solution that probably comes to mind is to use the geometric earnings growth rate over the period as our de-trending rate. Unfortunately, since the calculation of the geometric earnings growth rate only uses the first and last elements of the time series, we could get a spurious growth rate in certain situations. As an extreme example, imagine ten years of earnings data with the least recent earnings per share equal to $0.25, and the other nine elements equal to $1.00 per share (1,1,1,1,1,1,1,1,1,0.25); the geometric earnings growth rate over ten years would be 15%, and using this to de-trend the earnings time series would give sustainable earnings per share of $1.75, which seems a little high for a company with earnings of $1.00 per share for every year except one. So what about using the arithmetic average growth rate, which uses all elements in the time series? This is actually worse, as it suffers from being overly sensitive to one or more years of earnings growth that are not really representative of the overall trend (i.e., outliers)43. In this example, using the arithmetic average growth rate to de-trend the time series yields sustainable earnings of $4.03 per share. We clearly need a better technique to determine the underlying earnings growth rate; the technique should take into account all data points (not just the beginning and end of the time series), and should not be overly sensitive to the growth rate during any one year. Using the slope from linear regression would meet these criteria, except that the growth trend is not linear, but geometric. As part of our solution, lets first define an optimally de-trended data series as a data series with a slope close to zero, as determined by linear regression (once the series is de-trended, linear regression works fine). We could then keep guessing (as usual, we will use our binary search algorithm to aid in our guessing) at the value of the underlying growth trend until we find one that, when used to de-trend the data series, produces a de-trended data series with a slope of zero, as indicated by applying linear regression. We can then take the average of this de-trended data series to determine sustainable earnings; alternatively, we can directly determine sustainable
43

Because of this sensitivity, approximating the geometric growth rate by subtracting the estimated variance lag from the average growth rate does not work well either in extreme cases

Page 53 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

earnings from the regression intercept, both yield the same result. Applying this technique to our example (1,1,1,1,1,1,1,1,1,0.25) yields sustainable earnings of $1.15 per share, which seems fairly reasonable. This technique works well for all but the most chaotic earnings time series, but can produce spurious results in some cases where there really is no underlying trend, but more of a single step up (or down) in earnings. In favor of conservatism, I take Benjamin Grahams advice given in Security Analysis, and limit our estimate of sustainable earnings per share to the smaller of either the de-trended average or the highest year of adjusted earnings achieved in the last ten years, in todays dollars44. For example, if the companys inflation adjusted expected earnings per share for the trailing ten years of adjusted earnings is ($3.50, $3.75, $3.41, $3.10, $2.81, $2.50, $2.15, $1.85, $1.50, $1.30), with $1.30 being the adjusted earnings calculated for the earliest year, then the average of the de-trended earnings per share time series is $4.20. But since the highest adjusted earnings per share out of the ten years is only $3.75, we would use this rather than $4.20 for our estimate of sustainable earnings per share. Note that this technique can sometimes produce spurious results when the earnings time series is extremely volatile. I have found through experience that when the lower 95% confidence internal on the regression intercept falls below 75% of the regression intercept which often occurs with companies in cyclical industries - it is better to go with an alternate approach, where we take the average of the de-trended revenue per share time series (which is more stable), and then multiply this by the average profit margin (adjusted earnings divided by revenue) over the ten-year period. It turns out that Caterpillar is such a company. Returning to our example with Caterpillar, Figure 3-2 illustrates the difference between sustainable earnings calculated by taking the average of the de-trended revenue per share time series and multiplying it by the average profit margin - and reported earnings. The Adjusted Earnings Per Share time series in the chart has non-recurring charges removed, but the only significant charge was in 1992, where reported earnings were 3.01 per share. Clearly the sustainable earnings time series gives a much clearer picture of the companys earnings power at different points in the business cycle. Moreover, the price to sustainable earnings ratio has a much better correlation to the companys relative valuation than using the ratio of price to a single years of adjusted earnings. Figure 3-2

44

The concept of adjusting for inflation was not found in Security Analysis, but I think it is a good idea; consider that $2.00 a share five years ago is really worth more in todays dollars, and we are evaluating the companys earnings power in todays dollars. With positive inflation, adjusting for inflation before finding the maximum adjusted earnings per share actually earned is a bit less pessimistic than using nominal values.

Page 54 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

CAT
30.00 6

Price to Sustainable Earnings Ratio

25.00

20.00 3 15.00 2 10.00 1 5.00

1987

1990

1998

1993

1996

1985

1999

1988

2002

1991

2005

2007

1994

1997

1986

2000

1989

2003

1992

1995

2001

0.00

2004

2006

2008

-1

End of Year Adjusted Earnings per Share Sustainable Earnings per Share (rolling average of de-trended time series) Price to Sustainable Earnings Ratio

We can also determine the sustainable earnings of a stock index. In this case I usually assume that non-recurring earnings components tend to balance out, and just calculate the average of the de-trended earnings time series. Figure 3-3 shows reported and sustainable earnings for the SP500 index from 1882-2005; we used the years from 1872-1882 for the sustainable earnings calculation for 1882. Figure 3-345

45

Source: /Book/AssetClassChapter/SP500_chart.xls

Page 55 of 156

Earnings per Share

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Sustainable Vs. Reported Earnings for SP500 Index


100

Earnings

10

1882

1914

1922

1906

1902

1950

1958

1970

1978

1894

1986

1938

1994

1998

1886

1930

1918

1910

1898

1890

1926

1934

Sustainable Earnings Per Share

1942

Year Reported Earnings Per Share

Both earnings time series have the same trend, but the sustainable earnings time series largely ignores minor economic cycles, although it does track the longer cycles. The standard deviation of the changes in the sustainable earnings time series is about half that of the reported earnings time series. We are walking a fine line here, in that we want to react to medium-term (long enough to impact our realized return over a long holding period) economic trends quickly while filtering out part of the shorter trends that will not significantly impact our realized return. At one extreme, as earnings plummeted during the great depression and stayed low for quite some time, calculating sustainable earnings per share using too many years of data would have inhibited the sustainable earnings time series from quickly adapting to the underlying medium-term trend. Here, using ten years of data did allow a reasonably accurate tracking of the medium-term trend. At the other extreme, when earnings fell in the 2001-2002 recession, they quickly recovered, and the ten-year trend would have told you that the SP500 was favorably priced with respect to its sustainable earnings. Since at any given time, we have no idea what the medium-term trend in earnings will be, we need to make a compromise, and ten years seems to be a fair compromise for both individual stocks and stock indices. Implementation Details This is a good point to mention that rather than use ten years of trailing adjusted earnings data in our calculation of a companys sustainable earnings, we instead use a conservative estimate (ideally projected by management whom you trust) of next years adjusted earnings combined with nine trailing years of adjusted earnings data. By doing

Page 56 of 156

1946

1954

1962

1966

1974

1982

1990

2002

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

so, our estimate of a companys sustainable earnings becomes an estimate of the companys earnings power projected one year into the future. The need for this will become apparent when we develop our valuation model later in this chapter. As a final note on this technique for calculating sustainable earnings, I prefer to use an inflation adjusted earnings time series as an input to the algorithm. This is because we are concerned about the companys real earnings power; furthermore, large changes in inflation can add another cyclic component to nominal earnings. In practice, I first compute the adjusted earnings per share time series, after which I adjust each element of the time series so that it is stated in todays dollars, and then apply the de-trending algorithm.

3.4 Earnings Adjustment Example


The following screenshot displays a portion of a spreadsheet I used to adjust GAAP earnings and Cashflow for the Coca Cola Company using the approach outlined in sections 3.1 and 3.2. I also use the companys adjusted earnings to compute an estimate of the companys adjusted operating income, which is used in several company analysis ratios discussed in Chapter 5. The definition of operating income can vary between companies, but I calculate it as adjusted pre-tax earnings, less equity in income, and add back in interest expense. The line items in the spreadsheet are explained in the following tables. The rationale for the earnings adjustments are covered in section 3.1, and those for the cash flow adjustments are covered in section 3.2. Table 2
Line # 3 4 5 6 7 9 12 Notes to Earnings Adjustments Adjustments Consist primarily of restructuring charges and asset impairments Subtract actual cash outlays incurred in restructuring L4-L3: non-cash portion of other operating charges Companys share of non-cash asset impairments recorded by bottlers L5+L6: total non-cash restructuring and impairment charges Pre-tax profit on sale of assets 2005 and 2003: lawsuit settlements. 2002: effect of change in accounting principal

Table 3
Notes to Operating Income Adjustments Line # 29 32 34 36 Adjustments We need to add back in the bottling impairments to arrive at adjusted equity income GAAP pre-tax income as reported L32 + L14: pre-tax income adjusted for non-recurring items L34-L29: less adjusted equity income from bottlers

Page 57 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Page 58 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Chapter Summary
To get a good picture of a companys earnings power, we need to remove nonrecurring components from a companys reported earnings. Free cash flow is a useful check on the adjustments we make to a companys reported earnings, and can also be used to detect creative accounting practices. Business cycles can result in a single years adjusted earnings being unrepresentative of a companys true earnings power. We can gain a better picture of a companys earnings power by taking the average of ten-years of de-trended adjusted earnings data with the most recent element of the time series being a conservative estimate of next years earnings.

At the end of the next chapter, we will provide some examples where we adjust a companys reported earnings, calculate the companys sustainable earnings per share, and then estimate the companys intrinsic value.

Page 59 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

4 Estimating a Companys Intrinsic Value (2006)


In this chapter we will develop a valuation model that estimates a stocks intrinsic value from its demonstrated earnings power. In developing the model, we will assume a relationship between demonstrated earnings power and intrinsic value that is typical for an average company; this will minimize our models prediction error when it is used to analyze a company chosen at random. Because in practice we will not be randomly selecting companies, but instead selecting exceptional companies with a strong competitive advantage that are priced at a discount to our estimate of their intrinsic value, our prediction error will not actually be minimized. To the extent that the markets pricing of the company is accurate, this will tend to make our estimate optimistic, or in other words, add positive bias. But if we agree that these companys strong competitive advantage should provide a higher intrinsic value per dollar of demonstrated earnings power, our prediction bias will be negative, and we will on average underestimate a companys intrinsic value and add to our margin of safety. We dont account for competitive advantage in our model because the future effects of a companys competitive advantage can be hard to quantify, although it seems intuitive that a strong competitive advantage will increase the probability that a companys demonstrated earnings power will persist into the future.

4.1 Qualities of a Good Valuation Model


Lets start by looking at traditional valuation models for a companys common stock. There are probably as many valuation models as there are financial analysts; what most of these models have in common is that they attempt to value a companys shares based off of the assumption that the shares will be sold after a relatively short holding period, usually less than five years. Perhaps one reason for this is that analysts rarely attempt to project a companys earnings past five years into the future. A major problem arises from the assumption of a short holding period, in that the market value of the companys shares at the end of the holding period has a large influence on the estimate of the shares intrinsic value. Why is this a problem? Because even if the analysts estimate of the companys earnings over the next five years is completely accurate46, the price per share will be affected by how much the market values a dollar of the companys earnings, or in other words, by the prevailing price to earnings ratio at the time of sale. As an example of how much this ratio can fluctuate, the average and standard deviation of the SP500s price to earnings ratio since 1950 has been 16.6 and 7.4, respectively, with a high and low of 54 and 6.5. This can actually understate the problem; because year to year earnings are so volatile, a companys price to earnings ratio might be highly inflated due to a run of several years of above average earnings, and if the earnings growth slows down at the

46

These estimates rarely are accurate, see chapter 7 in Random Walk down Wall Street

Page 60 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

time of sale, both the companys earnings and the companys price to earnings ratio might be significantly less than the estimates fed into the model47. This is probably a good time to ask the question, why cant analysts accurately predict a companys future earnings, even over relatively short time periods of one to five years? Although we touched on this issue in Chapter 1, lets investigate further by looking at how a typical analyst attempts to estimate a companys future earnings. When an analyst attempts to predict a companys future earnings, he will typically try to gather as much information about that company as possible; this information might include an analysis of the companys brand strength, quality of management, and competition. He will then use this information to predict whether the company is likely to gain or lose market share, how much the company can expect to receive for each unit of merchandise sold, and ultimately, the companys likely revenue over the next five years. With another estimate of what the company will need to fork over for raw materials, and other estimates of labor costs, interest expense, etc., he can then generate an estimate of the companys earnings over the next five years. Of course these factors also depend on economic conditions over the next five years, which requires additional estimates. What we have here is a model where most (if not all) of the inputs are estimates, and within the model, these estimates are used as inputs for other estimates, and so on, until we arrive at a prediction for the companys earnings. The problem is further compounded because many of these model inputs are not even quantifiable (such as a companys brand strength and quality of management). It is no wonder that even consensus estimates from a large group of analysts are rarely correct, even on average. No matter how well an analyst understands a companys business model, there are just too many variables affecting a companys earnings, many of which are neither quantifiable nor predictable. The typical analysts approach to stock valuation is a good example of inside reasoning. With inside reasoning, we attempt to gather as much information as possible about the company, the companys competition, the economy, etc., and put it together to form a prediction. While you might think that a large quantity of information about a company might allow an accurate forecast of future earnings, this is not the case. The problem of forecasting future earnings is a subset of problems that require interactive, rather than linear reasoning. As it turns out, the human mind48 is not set up all that well for interactive reasoning; one effect of this is that predictive ability tends to stay flat or even decrease as additional information is presented to the problem solver. At the same time, the confidence of the prediction increases greatly with more information; obviously this is a dangerous combination. This effect has been noted in psychological experiments using many different types of specialists, including doctors, psychologists, economists, engineers, and of course, security analysts49.

47

See investments by Sharpe, page 628. Company PE ratios tend to mean revert to the asset class average after a couple of years. 48 Computer models are no better, as a computer cant reason at all 49 See chapter 4 of Contrarian Investment Strategies by David Dreman, where the concepts of inside and outside reasoning are presented in much more detail

Page 61 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

An alternative to this approach, known as outside reasoning, takes a step back and examines the problem at a higher level, ignoring the details. Lets assume we want to determine the intrinsic value of a companys common stock. Instead of trying to figure out a precise estimate of the companys earnings and dividend over the next five years, we would perhaps look at the historical average earnings growth of the companys industry over much longer time periods. Or we might even make the assumption that historical industry averages mean revert to the growth rate of the underlying asset class (they do), and instead base our prediction on the long-term average growth rate of the asset class (for example, large cap stocks). If we happened to know of any factors that both affect this growth rate and can be quantified at the time of purchase, we might use these factors to adjust the asset classs long-term growth rate. The key with outside reasoning is to focus on broad, easily identifiable factors that can be compared across all investments in the asset class50. As it turns out, outside reasoning works much better than inside reasoning for problems requiring interactive reasoning, not just with investment problems, but also with the other examples of interactive reasoning mentioned earlier. What we need is a model that is simple enough that it can be used to value either an individual stock or a stock index representing an asset class. Furthermore, we must be able to demonstrate that all model inputs are quantifiable, known at the time of purchase, and have a quantifiable effect on a companys expected return. Some examples of potential model inputs that do not fit these criteria are the companys historical earnings growth rate (on average, these tend to mean revert), analysts forecasts of a companys future earnings growth, and an estimate of how much market share the company will gain over the next five years. Finally, it would be nice to be able to minimize and quantify our prediction error even under a simplified assumption such as the company has been chosen randomly from its asset class. Although we will not be randomly choosing stocks from their asset class, but instead choosing them based off of their intrinsic value and expected return as predicted by the model, as well as our opinion of the issuing companys competitive advantage, we can later address the consequences of violating the random sampling assumption, and how we might minimize the effects of those consequences on portfolio risk. Lets summarize our desired model attributes: Company specific inputs should be quantifiable and have a quantifiable effect (even if only statistically) on the stocks expected return. Company specific inputs should be known at the time of purchase, as opposed to being an estimate The model should be general enough that it can be used to value an index representing the stocks asset class

50

This precludes criteria like this company has a really hot product

Page 62 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

The prediction error should be quantifiable under a simplified assumption such as the stock being randomly selected from its asset class.

An example of a quantifiable input is a stocks sustainable earnings per share. Although it is only an estimate of a companys earnings power projected one year into the future, with the exception of the estimate of next years earnings (which in most cases has a minor effect on the sustainable earnings calculation), a stocks sustainable earnings per share are computed using data known at the time of purchase. Therefore, a stocks sustainable earnings per share qualify as a model input under our criteria. Later we will show that a companys long-term issuer credit rating also meets our input criteria, in that it is a statistical predictor of the companys future default rate, which has an obvious impact on the average future earnings per share growth rate. Lets briefly discuss prediction error. If under a random sampling assumption our estimate of a stocks intrinsic value is equal on average to the stocks perfect foresight intrinsic value, the average prediction error is zero, and we can quantify the magnitude of our prediction error as the standard deviation of the intrinsic value of the stocks in our sample (the stocks asset class). As an analogy, if you had to guess the height of a person to be selected at random from a room of people where you know the average height to be six feet, then guessing that the randomly selected persons height is 6 feet would minimize your prediction error. Moreover, if we know the standard deviations of these heights is 7 inches, then we can quantify our likely prediction error as having a standard deviation of 7 inches.

4.2 A Simple Valuation Model


Our valuation model is a dividend discount model that assumes a constant dividend growth rate and that a stocks future price is equal to its intrinsic value. It can be shown that under the assumption of a constant cashflow growth rate and infinite holding period, the general equation for intrinsic value that we learned in Chapter 151, simplifies to: Equation 4-1

IV =

CF1 , DR G

Here G is the annualized cashflow growth rate, DR the discount rate, and CF1 is the cashflow received during the first year, which for a common stock will be the first years dividend. It can also be shown that this simplified equation is valid for shorter holding periods as long as the sales proceeds equal our estimate of the stocks intrinsic value at
51

IV =

CF[N] , Where CF[N] is the cashflow received in year N, DR the discount rate, N N =1 (1+ DR)

and H the holding period

Page 63 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

the time of sale, with the intrinsic value calculated under the same constant growth rate assumption. This is why it is important to use the same discount rate that the market has historically applied to the stocks asset class otherwise, our estimate of intrinsic value at the time of sale would on average differ markedly from that of the market, thereby creating a systematic source of error for shorter holding periods. The model makes the assumption that a stocks performance will be average; specifically, that its payout ratio, earnings per share growth rate, and market discount rate will all be equal to the asset class average. The one exception is that we make a small modification to the stocks expected earnings per share growth rate and discount rate to account for the companys financial strength. Aside from this adjustment for financial strength, the model assigns the same intrinsic value to a dollar of sustainable earnings for all companies in the same asset class. Although simplistic, it is the models simplicity that protects us from becoming too optimistic (or pessimistic) about a companys future prospects Our model takes two quantifiable inputs the stocks sustainable earnings per share and credit rating - that are known at the time of purchase, combines these inputs with four parameters that are derived from the historical performance of the securitys asset class, and calculates an estimate of the stocks intrinsic value. These four parameters are the asset classs historical real earnings per share growth rate, average payout ratio, average credit rating, and real market discount rate. The reason we derive these parameters from the asset classs historical performance is simple: we dont know the future. Sure, the future will likely be different than the past, but keep in mind that despite the enormous technological advances made since 1871, and the large changes in the industry composition of the SP500, the long-term earnings per share growth rate of the SP500 has been pretty constant, with most of the variation being explained by the changing payout ratio (see Chapter Error! Reference source not found.). Similarly, the long-term average discount rate that the market has applied to the SP500 has also remained within a fairly small range, while the payout ratio and credit rating are just used to measure the conditions under which the earnings growth occurred. With this in mind, deriving these four parameters from historical data seems pretty reasonable. But remember that over shorter holding periods, the ex-post values of these parameters will likely be very different than these historical averages; we are just choosing these values to minimize our prediction error. Payout Ratio and Dividend Growth Rate Our model will determine the first years dividend by multiplying the stocks sustainable earnings per share by the asset classs historical payout ratio, and set the constant dividend growth rate equal to the historical asset class earnings per share growth rate that was measured over the period the historical payout ratio was calculated52.
52

Because we assume a constant payout ratio, the dividend growth rate equals the earnings growth rate.

Page 64 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

EPS Growth Rate Adjustment for Financial Strength This is a good time to discuss the rationale for adjusting the asset class earnings per share growth rate based off of the companys financial strength as quantified by the companys long-term issuer credit rating. A companys long-term issuer credit rating is used by creditors, such as banks or bondholders, to determine the probability of a company defaulting on its debt. The probability of a company defaulting on its debt should be of interest to an owner of the companys shares, because once a company defaults on its debt, the value of these shares falls to zero. Credit rating agencies have quantified (statistically) the relationship between a companys long-term issuer rating and the companys subsequent default rate, and intuitively, it makes sense that, at a minimum, we would adjust a companys earnings growth rate by the difference between the default rate implied by the companys credit rating and the default rate implied by the asset classes historical average credit rating, with the reasoning being that the asset classs historical earnings per share growth rate occurred with some companies going bankrupt, and the growth rate would have been higher if the asset class had a higher average credit rating (because fewer companies would go bankrupt), and vice versa. In our model, we will adjust our estimate of the stocks earnings per share growth rate by adding a growth rate adjustment calculated as follows: Equation 4-2

GADJ = (DR + DR ) AssetClass (DR + DR )Company


Where (DR + DR ) AssetClass is the sum of the annualized default rate and default rate standard deviation for the asset class, as implied by the asset classs historical average credit rating, and (DR + DR )Company is the sum of the annualized default rate and default rate standard deviation for the company, as implied by the companys current credit rating. We will use Moodys default rate statistics from 1920 to the present.

Although the rationale for adjusting the earnings growth rate by the expected annualized default rate is fairly obvious, the reason we are including the default rate volatility is less so. In Chapter 1 we learned that as a companys credit rating falls, the companys expected default rate volatility increases, which reduces the certainty of bondholders receiving the expected value of future cashflows - this was used to explain the historical spread between investment grade and junk bond market discount rates. It is fairly intuitive that this increased default rate also decreases the certainty of the expected value of shareholders future cashflows. To account for this, we should adjust our discount rate based off of the default rate volatility implied by the companys credit rating. But instead of reducing our discount rate by the difference in volatility, we instead increase the growth rate, which has the same effect on the intrinsic value calculated using Equation 4-1.
Note that even if the company we are analyzing never defaults, the companys credit rating will still affect its earnings growth by its impact on the companys cost of capital. As an example, consider the historical yield spread between seven-year U.S. Treasury Page 65 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

bonds and seven year junk bonds from 1986 to 2001, which has averaged around 6%. Lets assume we are analyzing two companies, A and B, where A has a credit rating of AAA, and B has a credit rating of BA, and therefore B must pay a 6% higher interest rate on its debt than A. Under these conditions, we can show that if both companies can achieve a 15% return on invested capital, have a payout ratio of 50%, a debt to equity ratio of 1.0, and do not issue any new shares, then company A will grow 3%53 faster than company B; this is not too far off from our growth rate delta of 2.93% shown in Table 4-3 (found in the next section) where we use our model to compute the intrinsic value and expected return of stocks from the large cap domestic stock asset class. Discount Rate At this point, we have shown how our model converts a companys sustainable earnings per share and long-term issuer credit rating into a constant dividend growth rate and the first years dividend using three parameters derived from the stocks asset class: the historical earnings per share growth rate, the historical payout ratio, and the historical credit rating. Our model also requires a discount rate. Earlier we made the case that it makes sense to use the discount rate that the market has historically applied to the stocks asset class; using this discount rate should, on average, make our estimate of a stocks intrinsic value at the end of our holding period the same as the markets. And since the market is efficient on average, the best54 estimate for a stocks future market value is its future intrinsic value; consequently, setting our discount rate equal to the markets should minimize the contribution to our models prediction error stemming from the stocks actual price at the time of sale. So what should we use for an estimate of the historical market discount rate? Recall that an investments rate of return is equal to the discount rate required to bring an investments sum of discounted future cashflows equal to the investments market price. Since the market is efficient on average, and therefore a stocks intrinsic value should equal its market value on average, it follows that the markets historical discount rate equals the asset classs historical long-term geometric real rate of return. If the market had demanded a higher discount rate, then this historical rate of return would have been higher, and vice versa. This rate of return should be calculated with the effects of a changing price to earnings multiple factored out, as we did in Chapter Error! Reference source not found. for the SP500; after all, regardless of the past trend in price to earnings multiples, a future downward trend in the price to earnings multiple is just as likely as an upwards trend. The Model

53

dGEPS D / EQ(1 PR + S% PE) = , proof omitted di (1+ S% )


In that it will minimize our prediction error

54

Page 66 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

To review, our valuation model calculates a stocks intrinsic value using two company specific inputs, a companys long-term issuer credit rating and sustainable earnings per share (SEPS), both of which can be quantified at the time of purchase, as well as four asset class specific parameters, the asset classs historical geometric earnings per share growth rate ( GRAssetClass), the historical average payout ratio under which that growth rate occurred ( PRAssetClass ), the asset classs historical long-term issuer credit rating, and the historical market discount rate ( DRAssetClass ). This is expressed mathematically in Equation 4-3: Equation 4-3
IV = PRAssetClass * SEPS DRAssetClass GRAssetClass CRA

Here CRA is a credit rating adjustment that is a function of the asset classs historical weighted average credit rating and the stocks current credit rating, as calculated using Equation 4-2. As was described earlier, the sustainable earnings per share input will be calculated as the average of the de-trended earnings per share time series, which consists of a conservative estimate for next years earnings per share55 and nine trailing years of earnings per share, all adjusted to account for non-recurring items, and with the time series adjusted for inflation prior to de-trending. And we cap the companys sustainable earnings per share at the highest adjusted earnings per share actually achieved or expected to be achieved during the first year we own the shares. Now that we have a model for estimating a stocks intrinsic value, we might also want to estimate our expected rate of return. We know that if our purchase price equals our estimate of intrinsic value, our expected rate of return is our discount rate. For the case where purchase price and intrinsic value are not equal, but we can assume an infinite CF1 holding period, we can rewrite Equation 4-1 as ER = + G , where P is the purchase P price and CF1 is the first years expected dividends. For shorter holding periods, the expected return obtained by finding the discount rate that brings the estimated intrinsic value equal to the market price will exceed this value if the intrinsic value exceeds the market price, and vice versa, with the error decreasing with increasing holding period. The reason for the error is that if we purchase a stock for less than its intrinsic value, and the sales price is equal to intrinsic value at the time of sale, the stocks price must revert to its intrinsic value over the holding period. This is just the mean reversion of price to earnings ratio that accounts for part of the excess return noted in the low PE strategy (see Chapter Error! Reference source not found.).

55

The reason for using an estimate of next years earnings is that the sustainable earnings per share multiplied by the asset class payout ratio in Equation 4-3 represents the dividend expected at the end of the first year in Equation 4-1.

Page 67 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Since we really cannot know our holding period at the time we purchase a stock (even if we plan on holding ten years, we might be persuaded to sell early if the stock quadruples in price the next year with no change in intrinsic value), I sometimes use the following equation to estimate expected return as a function of its estimated intrinsic value (IV), actual market value (MV), and our discount rate (DR): Equation 4-4

ER = DR

IV MV

Note that for the case where the market value equals the intrinsic value, this equation CF1 gives the same results as = ER + G . We will see later that in our model, the P contribution to a stocks expected return from the reversion of a stocks market value to its intrinsic value is what we would expect if the reversion occurred over a period greater than thirty years. Our models assumptions, parameters, and inputs are tabulated in Table 4-1: Table 4-1 Assumptions
A stocks EPS growth rate and payout ratio will equal that of the asset classs historical average, unless the company has a credit rating different from the asset classs historical credit rating, in which case we will adjust the EPS growth rate using Equation 4-2 A stocks market price will equal our estimate of its intrinsic value at the end of the holding period The market discount rate will be equal to the asset classs historical long-term rate of return, with the effects of a changing price to earnings multiple removed.

Company Specific Inputs


Sustainable Earnings per Share Long-Term Issuer Credit Rating

Asset Class Specific Parameters


Asset Class Historical Market Discount Rate Asset Class Historical Real EPS Growth Rate Asset Class Historical Payout Ratio Asset Class Historical Long-Term Issuer Rating

4.3 Customizing the Model for Different Asset Classes


Large Cap Stocks (Foreign and Domestic)

Page 68 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Because there has been a small rise in the average price to earnings ratio from 1871-2005, and this rise could be due to a permanent change in the markets discount rate arising from an increase in economic stability56, we will use the period from 1950-2005 to calculate the historical earnings per share growth rate, payout ratio, and discount rate for the SP500. From 1950-2005, the real geometric earnings per share growth rate was 2%, and the market discount rate (the realized rate of return) was 5.5% with the effects of the growth in the price to earnings ratio removed. We will assume that the SP500s historical weighted average long-term issuer credit rating was A2. My reasoning is that I assume that the Moodys All Investment Grade category will have a similar composition to the SP500, and this category has similar default rate statistics to the A2 rating category. Table 4-2 Asset Class Specific Parameters
Asset Class Real Market Discount Rate Asset Class Real EPS Growth Rate Asset Class Payout Ratio Asset Class Long-Term Issuer Rating 5.5% 2.0% 50% A2

We will obtain our credit rating adjustment (CRA) to earnings growth using Equation 4-2. Table 4-3 gives this credit rating adjustment (last column) as a function of the companys credit rating (first column); the last column shows the difference between the sum of the default rate and default rate volatility for the company and a company with an A2 credit rating, which we consider to be close to the historical average credit rating for the stocks asset class. Table 4-357 Rating 20 year default rate (%) Annualized 20-year default rate (%) 0.12 0.33 Annual Default rate volatility (stdev) (%) 0.0% 0.2% Default based Growth Rate Adjustment (%) (rate + volatility) -0.12 -0.22 -0.33 -0.44 -0.55 -0.66 -0.83 Shifted (+A2=0.66)

AAA AA1 AA2 AA3 A1 A2 A3


56

2.38 6.75

7.47

0.36

0.3%

0.54 0.44 0.33 0.22 0.11 0.00 -0.17

This increase in stability stems from larger government, which can stabilize demand through transfer payments and spending - during a recession. Other causes are the increased willingness of the Federal Reserve to act as lender as last resort, increased financial transparency due to SEC regulations, and the FDIC guarantee on bank deposits. 57 Source: Moodys default rates 1920-1999

Page 69 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

BAA1 BAA2 BAA3 BA B

13.95 30.82 43.70

0.66 1.35 1.83

0.5% 1.7% 4.5%

-0.99 -1.16 -1.33 -3.05 -6.33

-0.33 -0.50 -0.67 -2.39 -5.66

Bold entries are from data, others are interpolated

Our estimate of the shares intrinsic value is given by: Equation 4-5

IV =

50% * SEPS 5.5% 2% CRA

Where SEPS is the sustainable earnings per share and CRA the credit rating adjustment given in Table 4-3, and our estimate of the shares expected return is given by: Equation 4-6

ER = 5.5%

IV MV

I would be comfortable using this model to estimate the intrinsic value and expected return of any large market capitalization stock, foreign or domestic, as long as the foreign stock comes from a country with a free market, where we would expect similar dynamics to that which led to the 2% real earnings per share growth rate in the United States.
REITs Because the reported depreciation of REITs significantly overstates the recurring capital expenditures required to maintain its competitive position, we will calculate an REITs sustainable earnings per share using adjusted funds from operations (AFFO). Funds from operations (FFO) are defined by the NAREIT, and are given by GAAP earnings attributable to common shareholders, excluding extraordinary items (as defined by GAAP) and gains/losses from the sale of previously depreciated real estate58, and adding back in depreciation and amortization expenses. To calculate adjusted funds from operations we start with funds from operations, and subtract recurring capital expenditures; these are usually found in the quarterly supplemental information published
58

The gain from development real estate built with the intent to sell is included in Funds From Operations.

Page 70 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

by REITs. Because these recurring capital expenditures can be erratic, I smooth them by reducing FFO by multiplying it by one minus the ten-year average ratio of capex to FFO. Due to the relatively short history of REITs, and the very short history (1990 to present) of the modern REIT, I am going to recommend using the same discount rate and dividend growth rate as we used for the large cap stock model. From 1972 to 2007, the performance of this asset class has been similar to that of large cap domestic stocks, with a somewhat higher real geometric rate of return (8.5% vs. 6.7%) and an almost identical rate of return volatility (15.64% vs. 15.12%). However, REIT returns have taken a substantial turn for the worse in 2008 and to date in 2009. This is largely due to REITs taking on so much debt that they could no longer pay off maturing debt using AFFO, and credit spreads have increased to the point where refinancing is not an option. On the other hand, REITs that have maintained a more conservative balance sheet (Realty Income being one) have had returns that beat that of the SP500 by a comfortable margin.

4.4 Analyzing the Model


Selection Bias We have argued that deriving our models parameters from historical data should minimize our prediction error, provided we randomly select companies from a given asset class. But since we are instead selecting companies based off of both their discount of estimated intrinsic value to market value and company quality, these choices for model parameters probably will not minimize our prediction error. This problem is known as selection bias. Selection bias occurs when instead of selecting a sample randomly from a population, we instead select based off of some criteria. If these selection criteria in our case, a high ratio of estimated intrinsic to market value and a strong sustainable competitive advantage - happens to be correlated with the quantity we are estimating - a companys intrinsic value - then assuming an average relationship between a companys sustainable earnings and intrinsic value will not minimize our prediction error. For example, what if the reason that the stock we just purchased is trading at a discount is because the market is assuming that the companys future earnings are rapidly heading to zero, and the company is headed for bankruptcy? Fortunately, our model does have an input that is a good indicator of a companys future default probability - the companys credit rating. But what if the stock is trading at discount due to more subtle reasons; perhaps the market has some insight into the companys future earnings growth that our model is ignoring. To some extent, this is probably the case, but we have seen that a value investing strategy where we buy stocks with a lower than average price to earnings ratio provides a higher return than indexing the asset class, without adding significantly to the rate of return volatility, this would not be the case if the market were perfectly efficient.

Page 71 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

However, the best way to counter any potential negative selection bias is to add positive selection bias. We can do this by only purchasing the stock of exceptional companies, which we would expect to have a better than average earnings per share growth rate due to some sustainable competitive advantage. As long as we purchase these companies at a reasonable discount to their intrinsic value, with our estimate of the companys intrinsic value based on an average relationship between a companys demonstrated earnings power and intrinsic value, we will be stacking the odds in our favor. Relationship between Fair Value and PE As a sanity check to our model, we can look at what a stocks price to earnings59 (PE) ratio would need to be for our model to indicate that the stock is fairly valued. If our model is reasonable, then the price to earnings ratio for a stock selling at its intrinsic value (as predicted by our model) should be close to the asset classs historical average. Table 4-4 shows this relationship for large market capitalization stocks, as represented by the SP500 index. Table 4-4 Large Market Capitalization Model
Credit Rating Fair Value PE ratio AAA 16.9 AA2 15.8 A2 14.3 BAA2 12.5 BA 8.5 B 5.5 Asset Class Average Price to Sustainable Earnings 1950-2006 Asset Class Average Sustainable Earnings Yield 1950-2006 Asset Class Median Price to Sustainable Earnings 1950-2006 Asset Class Median Sustainable Earnings Yield 1950-2006 Fair Value Earnings Yield 5.9% 6.3% 7.0% 8.0% 11.8% 18.2% 15.8 7.3% 15.6 6.4%

It appears that the model is somewhat pessimistic in that it only considers a stock fairly valued at a PE slightly below that of the asset class average for a company with a credit rating equal to that of the asset class average. This is because our model uses a stocks earnings yield as a starting point to computing intrinsic value, and the historical average PE ratio of the SP500 index is slightly higher than the reciprocal of the indexs average earnings yield; this is due to the fact that the price to earnings ratio distribution is not symmetrical. Note that the fair value earnings yield (7%) is pretty close to the historical earnings yield from 1950-2006 (7.3%), so our model is pretty reasonable with respect to both the historical average earnings yield and price to earnings ratio. Expected Return and Price to Earnings Ratio
59

The earnings capitalization rate, or earnings yield, is related to the discount rate through the payout ratio and earnings growth rate.

Page 72 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Now lets take a look how our models estimate of a stocks expected return compares to its price to sustainable earnings ratio. Figure 4-1 plots two estimates of a stocks expected return on the y-axis and the stocks price to earnings ratio on the x-axis. The first estimate (blue) uses Equation 4-6, where we assume that the market value will revert to intrinsic value, while the second estimate (pink) assumes an infinite holding period, and uses the SE equation ER = 0.5 * + 0.02 + CRA to estimate the stocks expected return. On the P secondary y-axis, we plot the implied period over which the stocks market value reverts to its intrinsic value; this is the period over which the stocks price would need to mean revert to create the difference in expected annualized return as shown by the vertical distance between the curves. We see that this period decreases with a decreasing price to sustainable earnings ratio, but remains above 25 years for all cases plotted. We can compare this to the historical period of PE reversion implied by NYSE data where stocks were sorted into quintiles and deciles based off of PE ratio, which is around twenty years. Figure 4-160
Expected Return Vs. Price to Earnings Ratio
18.00% 16.00% 14.00% 70

60

Real Expected Return

50 12.00% 10.00% 8.00% 6.00% 20 4.00% 2.00% 0.00% 10 40

30

Now lets take a look at the excess return predicted by our model as a function of a stocks price to earnings ratio, and compare it to that for large cap domestic stocks from 1952-2005. We will compute the excess return for the large cap domestic stock asset class as a function of price to earnings ratio by first splitting the stocks that comprise the
60

Source: /Book/ModelChapter/HistoricalVsModel/Model.xls

Page 73 of 156

0 5. 5 6. 0 6. 5 7. 0 7. 5 8. 0 8. 5 9. 0 9. 5 10 .0 10 .5 11 .0 11 .5 12 .0 12 .5 13 .0 13 .5 14 .0 14 .5 15 .0 15 .5 16 .0 16 .5 17 .0 17 .5 18 .0 18 .5 19 .0 19 .5 20 .0

5.

Price to Earnings Ratio Expected Return w/ Reversion Expected Return w/o Reversion Years to Revert

Historical Accuracy

Years to revert to mean

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

NYSE into ten deciles based off of their price to earnings ratio. We then compute both the market cap weighted geometric rate of return (from 1952-2005) and the market cap weighted price to earnings ratio for each decile. We consider the third and fourth decile to be average for the large cap domestic stocks asset class, as they have a weighted average price to earnings ratio of 16.46, fairly close to average price to earnings multiple of the SP500 over the period from 1952-2005. For the deciles with an average price to earnings ratio of 5, 8, 11, 13.5, 17, and 20, we compute the excess return as the difference between that deciles geometric rate of return and the average geometric rate of return for deciles 3 and 4. This is plotted as the data series in Figure 4-2 represented by the black diamonds. The models estimate of excess return, calculated as the difference in the models expected return for each price to earnings ratio on the x-axis and the models expected return for a price to sustainable earnings ratio of 14.5 (close to where we consider a stock to be fairly valued), is plotted on the purple curve in the same figure. Although not exact, the two curves are reasonable close, demonstrating that our model (including the modeling of the effect of the reversion of market value to intrinsic value in Equation 4-4) is fairly accurate. Figure 4-2
Excess Return vs. PE Ratio: Model & Historical
12.00% 10.00% 8.00%

Excess Return

6.00% 4.00% 2.00% 0.00% -2.00% -4.00%

Lets now test how well our models estimate of expected return matches the actual 10year internal rate of return for the SP500 index. Figure 4-3 shows the actual realized internal rate of return and the rate of return predicted by our model over ten-year rolling periods from 1882-2006. We had to start at 1882 because we need ten years of data to compute sustainable earnings, which is an input to our model. The two series of returns look similar, and they are; if we run a regression of realized vs. predicted return, we find

Page 74 of 156

5 6. 0 6. 5 7. 0 7. 5 8. 0 8. 5 9. 0 9. 5 10 .0 10 .5 11 .0 11 .5 12 .0 12 .5 13 .0 13 .5 14 .0 14 .5 15 .0 15 .5 16 .0 16 .5 17 .0 17 .5 18 .0 18 .5 19 .0 19 .5 20 .0
Price to Earnings Ratio Actual Excess Return Excess Return Predicted by Model

5. 0

5.

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

that the model explains 24% of the variation in the realized returns, with a coefficient equal to 1.23 and a P-value of 0.0000014. The reason the coefficient is greater than one is that the actual return is on average, greater than that of the model. This is due to two factors; first, the slow increase in price to earnings multiple over the period, and second, the fact that our discount rate was optimized for the 1950-2006 period, which experienced a lower rate of return (with changes in PE factored out) than that of the 1882-2006 period. Figure 4-361
Model Vs. Realized IRR 1882-2006 (10 YR rolling)
0.2

0.15
Predicted and Actual IRR

0.1

0.05

-0.05
18 82 18 86 18 90 18 94 18 98 19 02 19 06 19 10 19 14 19 18 19 22 19 26 19 30 19 34 19 38 19 42 19 46 19 50 19 54 19 58 19 62 19 66 19 70 19 74 19 78 19 82 19 86 19 90 19 94

Start of 10 YR rolling period Realized IRR Predicted IRR

Lets end by plotting our models prediction of the SP500s intrinsic value along with the indexs market value in Figure 4-4. Note that often (although not always), a high ratio of intrinsic to market value is followed by an increase in the indexs market value, and vice versa. To illustrate, the 1929 crash was preceded by a ratio of market to intrinsic value of 1.07, the 1973 crash was preceded by a ratio of 1.3, and the crash in early 2001 was preceded by a ratio of 2, whereas the bull markets beginning in 1950 and 1980 were preceded by a ratio of 0.54 and 0.56, respectively. So it appears that our model does a reasonable job of estimating the indexs intrinsic value; it also appears that the market is a pretty good weighing machine, with the indexs market value reverting back to this intrinsic value. That said, I am not advocating jumping in and out of the market based off this estimated ratio of intrinsic to market value, but instead to construct and maintain (through rebalancing) a portfolio such that it always has a high ratio of estimated intrinsic to market value. Figure 4-4
61

Source: /Book/ModelChapter/HistoricalVsModel/Model.xls

Page 75 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

SP500 Market Vs. Intrinsic Value 1882-2005


10000 2.5

2 1000 1.5

Value

1 100 0.5

10

4.5 Examples
To illustrate the details of calculating a companys sustainable earnings per share and intrinsic value lets look at a few examples62. The first example is a bit tedious, as it shows in detail the calculation of a companys sustainable earnings per share, which is glossed over in the examples that follow. Since the calculations are automated using the Investing Toolbox software package63, you can skip the details if you find yourself falling asleep. UST, inc. (UST) Imagine it is November 2006, and we are considering the purchase of shares in UST, which has a fiscal year ending on December 31st. Up till now, we have discussed computing a companys sustainable earnings per share using a companys annual earnings for each element of the earnings time series. But since UST last published annual earnings at the end of 2005, these annual earnings are pretty stale. We can fix this by adding a temporary element to our earnings array (lets call this CurrEPS), which is the estimate of the total adjusted earnings per share for the current year; this will be equal to the sum of the adjusted earnings earned over the trailing 3 quarters plus an estimate for the next quarter.

62 63

Source for following data: /archive/2006/Analysis Available at http://web.me.com/briangaudet/InvestingToolBox/Investing_Tool_Box.html

Page 76 of 156

18 82 18 86 18 90 18 94 18 98 19 02 19 06 19 10 19 14 19 18 19 22 19 26 19 30 19 34 19 38 19 42 19 46 19 50 19 54 19 58 19 62 19 66 19 70 19 74 19 78 19 82 19 86 19 90 19 94 19 98 20 02
Year Market Value Intrinsic Value Ratio of Intrinsic to Market Value

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Our next step is to compute the annual earnings per share, adjusted for non-recurring items, from 2005 to 1997. We only use nine years, because we are going to merge this annual earnings array with CurrEPS, and also add to the head of the array an estimate for earnings over the next four quarters. When we calculate adjusted earnings per share we need to remove the effects of a $1,260,000 pre-tax litigation charge in 2002 and a $280,000 pre-tax litigation charge in 2003. Table 4-5 illustrates the companys reported earnings per share (row 2), earnings per share adjusted for non-recurring items (row 3), and adjusted earnings per share in todays dollars (row 4). Table 4-5
Year Reported EPS Adjusted EPS CPI Adjusted 2005 3.24 3.24 3.24 2004 3.23 3.23 3.34 2003 1.93 2.96 3.15 2002 2001 -1.59 2.97 2.97 2.97 3.22 3.31 2000 1999 2.70 2.68 2.70 2.68 3.04 3.13 1998 1997 2.44 2.37 2.44 2.37 2.89 2.89

Our next step is to create a new array combining CurrEPS of $3.18 with the data in Table 4-5. Because CurrEPS is constructed using the last three quarters of 2006, we need to shift the annual data from 2005-1997 (in Table 4-5) forward by three quarters. In the general case, where our most recent quarter of earnings data is Q1, Q2, or Q3 2006, we create a new array combining CurrEPS and the annual earnings data as follows:
shifted _ eps[2006] = CurrEPS * eps _ ratio + (1 eps _ ratio) * eps[2005] for(y = 1998;y <= 2005;y + +) shifted _ eps[y] = eps[y + 1]* eps _ ratio + eps[y]* (1 eps _ ratio)

Here eps_ratio is equal to the last quarter in 2006 we have earnings per share data for divided by four, which in this case (Q3) equals three fourths. The way to interpret each year of the shifted earnings per share array is that it is the sum of the first three quarter of earnings in the current year and last quarter of earnings in the last year, i.e., 2006 is Q1-3 of 2006 plus Q4 of 2005, 2005 is Q1-3 of 2005 plus Q4 of 2004, and so on. The new array is shown below:
Year New Array 2006 3.20 2005 3.27 2004 3.29 2003 3.17 2002 3.24 2001 3.24 2000 3.06 1999 3.07 1998 2.89

In the new array, the earnings per share for 2006 are equal to the sum of the earnings from Q1-Q4 2006 (CurrEPS) multiplied by 0.75 plus the annual earnings from 2005 multiplied by 0.25, the earnings per share for 2005 is equal to the earnings per share from 2005 multiplied by 0.75 plus the earnings per share for 2004 multiplied by 0.25, and so on down to 1997, which is the earnings per share from 1998 multiplied by 0.75 plus the earnings per share for 1997 multiplied by 0.25.

Page 77 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Because the first element of the earnings time series should actually be a forecast of next years earnings64, we need to estimate earnings per share for Q4 2006 and Q1-3 2007 (we will call this NextEPS). For this I prefer to use managements own estimates of earnings over the next four quarters, but since this is not available, we will instead increase CurrEPS by 5%, which is in line with the long-term growth goal of management, and is in line with the growth rate that the company has already achieved; this gives a NextEPS of $3.34. But since this occurs one year into the future, and our earnings per share array is in todays dollars, we will adjust this downwards by 2.5% (an estimate of inflation over the next year); this gives a value of $3.28 per share for NextEPS. Combining NextEPS and the new array we created from CurrEPS and the annual earnings data (translated into todays dollars), we end up with the earnings per share data shown below:
Year Final Array 2007 3.28 2006 3.20 2005 3.27 2004 3.29 2003 3.17 2002 3.24 2001 3.24 2000 3.06 1999 3.07 1998 2.89

We now apply our algorithm to de-trend this earnings time series


Year DetrendArray 2007 3.28 2006 3.23 2005 3.34 2004 3.40 2003 3.31 2002 3.42 2001 3.46 2000 3.30 1999 3.34 1998 3.18

The sustainable earnings per share are calculated as the average of the de-trended array, or $3.32. However, recall that earlier we made the argument that to be conservative, we should cap a companys sustainable earnings per share at the highest adjusted earnings per share actually earned (or expected to be earned over the next year) by the company, in todays dollars; here this is the $3.28 expected to be earned in total for Q4 2006 and Q1-3 2007 (NextEPS). Consequently, since the de-trended average ($3.32) exceeds $3.28, we should cap sustainable earnings per share at $3.28. Although this company had few non-recurring adjustments to earnings, and was therefore fairly easy to analyze, we can still do a sanity check using free cash flow. We can do this by finding the ratio of the sum of the free cash flow per share time series from 1997-2005 and the sum of the adjusted earnings per share time series from 1997-2005; doing so we find the ratio is 1.01. We also see that in most years, free cash flow and earnings are pretty close. Since this company is in a mature industry with approximately constant market share, we assumed that all reported capital expenditures were recurring. The free cash flow for UST is shown below, and can be compared to the annual adjusted EPS data in row 3 of Table 4-5:
Year Free Cash Flow Adjusted EPS 2005 3.11 3.24 2004 3.12 3.23 2003 3.09 2.96 2002 3.10 2.97 2001 3.17 2.97 2000 3.11 2.70 1999 2.46 2.68 1998 2.38 2.44 1997 2.06 2.37

64

Because it is next years expected dividend that appears in the numerator of IV =

CF1 DR G

Page 78 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Since our computation of sustainable earnings per share uses an estimate for next years earnings, lets do a quick sensitivity analysis on how our calculation of sustainable earnings per share changes if our estimate is wrong; here, in order to focus on the smoothing properties of the de-trending algorithm, we will not cap sustainable earnings at the maximum earned or expected to be earned. Lets assume that the actual NextEPS is only $3.00 per share, a forecasting error of 10.7%. In this case, our sustainable earnings per share would actually be only $3.21, an error of 3.4%; as expected, using multiple years in our computation of sustainable earnings reduces our sensitivity to an error in our estimate of NextEPS. Note that if we had earnings data for Q4, and therefore for the entire year in 2006, we would not need to compute CurrEPS and merge it with the annual data, we would merely combine NextEPS with the annual data from 2006 to 1998, and de-trend that time series. Now lets compute the companys intrinsic value and expected return. Plugging the companys sustainable earnings per share ($3.28) and A3 credit rating (resulting in a credit rating adjustment of 0.17) into Equation 4-5, we find that UST has an estimated PR * SEPS 0.5 * $3.28 intrinsic value of65 IV = = = $44.69 per share. Since DR G CRA 0.055 0.02 + 0.0017 the stocks current market price is $56.25, the stock is currently overvalued, with a ratio of intrinsic value to market value equal to 0.79, so I would not purchase shares in the company at the current price. On the other hand, I would not sell them either, although I might consider selling my shares of UST at a high enough valuation (perhaps a ratio of intrinsic value to market value of 0.60 or less), but only if there was another company of similar quality trading at a good discount to its intrinsic value. Plugging USTs current price and our estimate of the stocks intrinsic value ($47.71) into Equation 4-6 gives us $44.69 IV = 5.5% an expected return of ER = DR = 4.37% . Since our valuation $56.25 MV model assumes an average ratio of intrinsic value to demonstrated earnings power, both the intrinsic value estimate and expected return are based on the assumption that the demonstrated earnings power persists into the future. Before purchasing the company, we would first want to analyze the company and the companys industry using the techniques of Chapter 5. Duke Realty (DRE) Adjusted funds from operations (AFFO) are the preferred measure of an REITs earnings power, and we use AFFO per share rather than earnings per share as our starting point in calculating the sustainable earnings per share of an REIT. Unfortunately, REITs do not report AFFO, but only funds from operations (FFO), and it is up to you to determine the REITs level of recurring capital expenditures and derive AFFO from FFO. The capital
65

Recall SEPS is sustainable earnings per share, DR the discount rate, G the historic earnings per share growth rate for the asset class, PR the payout ration under which that earnings per share growth rate occurred, and CRA the credit rating adjustment.

Page 79 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

expenditures we are interested in are those required to keep the REITs current properties competitive and those incurred in the normal operation of the properties, and would include both building improvements and capitalized expenses such as leasing commissions. Capital expenditures incurred by purchasing new properties, redevelopment projects, or expanding a property should not be used in the calculation of AFFO. Most REITs provide supplemental data that breaks down capital expenditures to a granularity that makes the calculation of AFFO fairly easy. Because these recurring capital expenditures can vary quite a bit over the real estate cycle, we calculate the ratio of the sum of recurring capital expenditures to the sum of FFO, which we will denote as RAVE , over ten years (just as we do in the free cash flow calculation), and then compute AFFO for a given year as FFO multiplied by one minus this average ratio:

Equation 4-7
AFFO = FFO(1 RAVE )

Now that we have a method for computing AFFO, we can calculate sustainable earnings per share using the same technique as for any other stock, we just need to replace earnings per share in a given year or quarter with AFFO, which will be the companys reported FFO per share, adjusted as per Equation 4-7. We need to use some caution when taking a companys reported FFO at face value, as sometimes companies will report FFO in different ways, which makes it difficult to compare REITs. Usually these differences are small, but I have seen at least one REIT (First Industrial) include net profits from property sales in its FFO numbers. Although FFO is not intended as a measure of cash flow, but instead is intended to be an accrual based modification to GAAP earnings, gross misreporting of FFO can usually be detected by comparing cash flow from operations to FFO over several years; this easily spotted the miscalculation of FFO by First Industrial REIT. You should familiarize yourself with the NAREIT white paper (http://www.nareit.com/policy/accounting/whitepaper.cfm) on the calculation of FFO, and check each annual report to make sure the company complies. If it does not, then you will need to adjust the reported FFO as necessary. Lets assume we have 2006 Q3 earnings reported for DRE, and we wish to compute its sustainable earnings per share. Our first step is to find the ratio of the sum of recurring capital expenditures to the sum of FFO over the ten years, which ends up being 0.19:
Year Capex PS FFOPS 2005 0.72 2.40 2004 0.76 2.47 2003 0.57 2.45 2002 0.44 2.36 2001 0.45 2.57 2000 0.44 2.46 1999 0.35 2.19 1998 0.22 2.33 1997 0.18 1.91 1996 0.19 1.61

Although the ratio of capex to FFO per share seems to be increasing at an alarming rate, this often occurs during the real estate cycle when overbuilding increases competition, requiring a higher outlay for leasing commissions and tenant improvements to keep

Page 80 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

occupancy up. We can now use this ratio to compute AFFO per share from 1997 to 2005, with AFFO in a given year calculated as 0.80 multiplied by FFO in that year:
Year AFFOPS 2005 1.97 2004 2.03 2003 2.01 2002 1.94 2001 2.11 2000 2.02 1999 1.80 1998 1.91 1997 1.57

Since we have Q3 data for 2006, CurrFFOPS will equal the sum of quarterly FFO from Q1Q3 2006 and Q4 2005, or $2.29, and CurrAFFOPS will be $1.88. We now merge CurrAFFOPS with the annual data as we did for UST: Table 4-6
Year New AFFOPS 2006 1.90 2005 1.98 2004 2.02 2003 1.99 2002 1.98 2001 2.08 2000 1.96 1999 1.82 1998 1.82

Fortunately most REITS provide an estimate of next year funds from operations, and DRE is no exception; the company predicts 2007 FFO of $2.65 per share, and Q4 2006 FFO of $0.71 per share. We can combine this into an estimate of FFO over the next four quarters (Q1-3 2007 and Q4 2006) by multiplying the 2007 estimate by 0.75 and adding the Q4 estimate, giving us FFO of $2.70 per share and AFFO of $2.21. Now we place this AFFO estimate at the head of the earnings time series shown in Table 4-6, adjust the combined time series for inflation, de-trend it, and take the average (just like in the UST example), giving us sustainable AFFO per share of $2.06. Since AFFO per share of $2.08 was obtained in 2001, we can use the de-trended average of $2.06 as our estimate of sustainable earnings per share. Plugging this sustainable AFFO per share and a credit rating adjustment (from Table 4-3) of 0.99% into Equation 4-5, we estimate the stocks intrinsic value to be PR * SEPS 1.0 * $2.06 IV = = = $31.74 . With a stock price of $44 on DR G CRA 0.055 0.00 + 0.0099 November 27th 2006 the stock is overvalued, and we can estimate the stocks expected IV $31.74 return as ER = DR = 0.055 = 4.0% . Note that the stocks price fell to around MV $44.00 $30 per share by August 2007, generating quite a large negative rate of return.

Pfizer (PFE)

Lets now take a look at Pfizer (ticker PFE), a pharmaceutical company with a significant amount of charges to earnings due to the write-down of intangible assets, and goodwill amortization as well. Table 4-7 tabulates the companys reported earnings per share from continuing operations, gains from asset sales, charges to earnings due to in process research and development acquired from the purchase of a business, and amortization of goodwill. Also shown is the weighted average number of shares outstanding during the year, the companys effective tax rate, and finally, the companys adjusted earnings per share.

Page 81 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Table 4-7
Year Reported EPS Asset Sales In Proc R&D Goodwill Amt. Wgtd. Ave Sh. Eff. Tax Rate Adj. EPS 2005 2004 2003 2002 2001 2000 1.10 1.51 0.22 1.49 1.21 0.57 188 16 85 35 0 216 -1652 -1071 -5052 0 0 0 -3409 -3364 -2187 -28 -45 -120 7411 7614 7286 6241 6361 6368 0.297 0.190 0.497 0.221 0.244 0.354 1.56 1.98 0.71 1.49 1.22 0.56 1999 1998 1997 0.81 0.73 0.48 0 0 0 0 0 0 -104 -105 -68 6317 6362 6297 0.283 0.264 0.28 0.82 0.74 0.49

Before going into more detail, lets identify where these numbers come from in the companys 10-K filing. Pfizers consolidated statement of income is found on page 35 of their 2005 financial report (page 70 of the PDF file). In this statement, although the companys diluted earnings per share from continuing operations is not stated directly, it can be found by adding back in the cumulative effect of a change in accounting principles (which was zero for 2003,2004, and 2005) to income from continuing operations before effects of a change in accounting principles; this is the number found in row 2 of Table 4-7. Also found in the income statement are expenses attributed to amortization of intangible assets (goodwill amortization), merger-related in process research and development charges, restructuring charges and merger-related costs, and the number of weighted average diluted shares. For the reasons discussed earlier, we will not adjust reported earnings per share for this last item, but we will adjust for the first two, which are noted in rows 4 and 5 of Table 4-7. Three pages later, we find the companys statement of cash flows. Under operating activities, the companys reported earnings are reconciled to net cash provided by continuing operating activities; here we find that the company had a gain on the disposal of investments, products, and product lines in 2005,2004, and 2003; these are noted in row 3 of Table 4-7. Note we do not adjust for the gain on sale of discontinued operations, as this is already subtracted from reported earnings per share from continuing operations. The data from years prior to 2003 were found from older form 10-Ks found in the companys archive of SEC filings. To arrive at adjusted earnings per share for each year, we add up all of the charges to earnings that we dont believe are part of the companys recurring earnings; in this case these are found in rows 3-6 of Table 4-7. We then convert each years sum to a per share amounts by dividing by the weighted average diluted shares (row 7), after which we convert from a pre-tax to a post-tax charge by multiplying by one minus the companys effective tax rate, which was found in note 7 to the companys consolidated financial statements, and is listed in row 7 of Table 4-7. The resulting adjusted earnings per share are shown in row 8 of Table 4-7. We can use the companys adjusted earnings per share from Q1-3 2006 and Q4 2005 to compute a CurrEPS of $2.00 per share, and use the companys stated estimate for Q4 earnings and earnings growth during 2007 to arrive at a figure of $2.10 per share for NextEPS. Applying the same procedure as with UST, and using the adjusted earnings per Page 82 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

share from row 8 of Table 4-7, we arrive at sustainable earnings per share equal to $2.13. This is pretty close to the earnings per share expected in 2007 and achieved in 2004, so lets go ahead and let this number stand. We discussed earlier that it is advisable to rationalize a companys acquisition history before allowing charges associated with intangible assets to be added back into the companys pre-tax earnings. We find important information pertaining to Pfizers 2003 acquisition of Pharmacia in note 2 to Pfizers consolidated financial statements. In this note, we can find a table illustrating how Pfizer allocated the purchase price between tangible and intangible assets, and we find that Pfizer paid $21.403 billion for goodwill. If the assumptions used to value the goodwill are too optimistic, then Pfizer could very well face a large future charge to earnings. From note 2 in the 2003 form 10-K, we find pro-forma results for the combined revenue and earnings of Pfizer and Pharmacia in 2003 and 2002. Since the transaction was actually completed in mid-2003, the 2002 results of the combined company provide us with a picture of how much revenue the acquistion was expected to add; by subtracting the pro-forma revenue of $45 billion from Pfizers actual 2002 revenue of $32.4B, we find that the acquistion was expected to add $12.6B in revenue, and therefore Pfizer paid approximately 4.37 times the expected additional revenue. If we assume that Pfizer can efficiently integrate the company into its existing operations, then it would be reasonable to use Pfizers historical profit margin of 26.5% over the last ten years and assume that Pfizer paid approximately 16.5 times for the amount of current earnings that the acquisition added; this sounds pretty reasonable, particularly since Pfizer had a price to earnings ratio close to 20 in 2003, and paid for the acquisition by issuing stock. Looking at Pfizers revenue and earnings after the transaction, it appears that the transaction has benefited the company, although the companys 5-year rolling return on equity has fallen; one reason for this decline is that the restructuring charges from several recent acquisitions has temporarily reduced earnings; recall that the actual expenses related to the acquisitions will be paid out over several years. Lets also do a sanity check on the companys recurring earnings per share from 1996 to 2005 by comparing it to the companys free cash flow over the same period. Table 4-8 shows Pfizers cash flow from operations, capital expenditures, free cash flow per share, and free cash flow per share smoothed using the same technique as in the UST example. We find that the sum of capex to sum of cash flow over the ten years is 0.24, and then multiply cash flow per share for each year by 1-0.24=0.76 to arrive at free cash flow per share, which is shown in row three. Table 4-8
Year CFPS Capex FCFPS EPS (4) 2005 1.99 0.28 1.51 1.56 2004 2.15 0.34 1.61 1.98 2003 1.61 0.36 1.22 0.71 2002 1.58 0.28 1.20 1.49 2001 1.39 0.33 1.06 1.22 2000 0.97 0.33 0.74 0.56 1999 0.87 0.39 0.66 0.82 1998 0.81 0.31 0.62 0.74 1997 0.26 0.15 0.20 0.49 1996 0.33 0.12 0.25 0.44

Page 83 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Although in a given year, the free cash flow numbers differ substantially from earnings, we can look at the free cash flow per share and sustainable earnings per share using this data from 1996-2005 to check how the relationship holds on average. Doing so, we find that the sum of free cash flow per share (row four) divided by the sum of earnings per share (row five) is equal to 0.91, which is reasonably close to one. At this point we can plug the companys sustainable earnings per share of $2.13 and AAA credit rating into Equation 4-5 to estimate the companys intrinsic value. From Table 4-3 we calculate the companys credit rating adjustment (CRA) as 0.50, and the 0.5 * SEPS 0.5 * $2.13 intrinsic value per share as IV = = = $35.50 . Since DR 0.02 CRA 0.055 0.02 0.005 the company currently has a market value of $27 a share, it appears to be priced at a good discount to its intrinsic value, assuming of course that the companys subsequent performance is in line with the past. Exxon Mobile (XOM) Occasionally ten years of earnings data is not sufficient to cover a companys complete business cycle; a case in point is the Exxon Mobile Corporation. The problem here is that the companys business cycle follows the commodity price cycle for oil, and from 19962005, the growth in the price of oil has exceeded its long-term trend by a large margin. In this case, it makes sense to gather as much earnings data as possible in the hope of better covering a full commodity cycle. Fortunately, Exxon Mobile has been around for a long time, and we can obtain electronic copies of the firms form 10-K going back to 199366, and the financial summary in that report goes back to 1989, giving us 17 full years of earnings data. Calculating sustainable earnings using the EPS data from 19892005, CurrEPS of $6.59, and an estimate for NextEPS of $6.23, we find that XOM has sustainable earnings per share equal to $4.27 per share. But if we instead calculate sustainable earnings per share using the same CurrEPS and NextEPS, and earnings per share data from 1996-2005, we get sustainable earnings per share equal to $5.59, a value 31% higher. Since oil is at a historically high price (almost six times higher than it was in 1992), I believe it is prudent to use the lower number for sustainable earnings per share, as it contains a better sample of oils price history. We can plug the companys sustainable earnings per share of $4.27 and AAA credit rating into Equation 4-6 to calculate an intrinsic value per share of 0.50 * $4.27 IV = = $71.16 . Since the current market price is $72.50, the stock is 0.055 0.02 0.005 close to fairly valued.

Home Depot (HD)

66

Note that when we extract this EPS data from older form 10-Ks, we need to be careful to account for the cumulative effect of any stock splits.

Page 84 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Now lets use our model to plot a stocks intrinsic value and market value over time. Figure 4-5 plots the market value of one share of Home Depot, along with our models estimate of the stocks intrinsic value. When reading the chart, note that the y-axis for market value and intrinsic value (expressed in dollars) is on the left of chart, while the yaxis for the ratio of estimated intrinsic value to market value, is on the right. Note how much less volatile the estimated intrinsic value is compared to the market value. Despite the steady and above average earnings growth (and intrinsic value growth), there were two times when the market value fell well below the intrinsic value, providing an opportunity to purchase the stock with a nice margin of safety. This example also illustrates that, due to our conservative valuation model, we will often miss the opportunity to purchase a stock that has above average ex-post returns; by not purchasing this stock in 1995, we would have missed out on spectacular price appreciation, which was driven by above average earnings growth and an increase in the price the market placed on a dollar of the companys earnings. This is fine, because on average, most stocks wont (by definition) have above average earnings growth, and even if we can identify the ones that will, you need to know how long it will persist in order to quantify the excess return. At any rate, you can also see that even if you were lucky enough to buy in 1995, if you did not have perfect foresight, and sell at the peak in 1999, most of your unrealized capital gains would have disappeared by December 2002. The key point is to be patient, no matter how good a company looks, wait till it is fairly valued using a conservative valuation model. The Home Depot chart also serves as an example of how a companys market value can be much more volatile than its intrinsic value; there is no way to reconcile the market price at the end of 1999 and the market price in early 2003 or mid 2006 with the stream of earnings produced by the company. While we cannot know what the future holds, it must be very bright to support the 1999 price and quite gloomy to support the 2003 and 2006 prices; they cant both be true. Figure 4-567

67

Source: Book/ModelChapter/HD_model.xls

Page 85 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Home Depot
70 60 1.6 1.4 1.2 1 40 0.8 30 0.6 20 10 0 0.4 0.2 0

50

Dec-95

Dec-98

Dec-01

Dec-96

Dec-99

Dec-02

Dec-04

Dec-97

Dec-00

Aug-95

Aug-98

Aug-00

Aug-01

Dec-03

Aug-96

Aug-99

Aug-02

Aug-04

Aug-97

Aug-03

Aug-05

Dec-05

Apr-99

Apr-02

Apr-04

Apr-97

Apr-00

Apr-03

Apr-05

Apr-98

Apr-96

Month Market Price Intrinsic Value Ratio of Intrinsic to Market Value

SP500 index We can also use our valuation model to compute the intrinsic value and expected return of a share of a stock index; lets demonstrate this for the SP500 index. Our first step is to compute the indexs sustainable earnings per share. Table 4-9 gives the reported earnings per share for SP500 index from 2005 to 1997. Table 4-968
Year EPS 2005 69.93 2004 58.94 2003 48.74 2002 27.59 2001 24.69 2000 50.00 1999 48.17 1998 37.70 1997 39.72

As of 11/20/2006, Standard and Poors has released Q1 & Q2 2006 earnings for the SP500 index, and CurrEPS (the sum of Q1-2 2006 and Q3-Q4 2005 earnings per share) is $74.49. For NextEPS, we will use the sum of Standard and Poors estimates for the next four quarters (Q3-4 2006 and Q1-2 2007), which is $84.20 per share. Following the methodology for calculating sustainable earnings per share, we obtain a value of $71.41 for the indexs sustainable earnings per share. Today the price of the index is $1402.92 per share; this gives us a sustainable earnings yield of 5.09% and a price to sustainable earnings ratio of 19.65. Plugging the sustainable earnings per share of $71.41 and an assumed weighted average credit rating of A2 into Equation 4-5, we obtain an intrinsic 0.50 * SEPS $71.41* 0.5 value of IV = = = $1021, indicating that according to our DR G CRA 0.055 0.02 0
68

Source: Book/ModelChapter/IndexValuation/SP500_Valuation

Page 86 of 156

Apr-01

Apr-06

Ratio of Intrinsic to Market Value

Intrinsic Value, Market Value

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

model, the SP500 index is currently overvalued, with a ratio of intrinsic value to market value equal to 73%. Plugging this ratio into Equation 4-6, we also find that the indexs IV $1021 = 0.055 = 4% . Does this mean that the index real expected return is ER = DR MV $1403 is poised for a crash? No, but, you would not catch me purchasing shares at this price either.
We can see the impact of replacing next years estimate of earnings with our estimate of sustainable earnings by comparing the forward price to earnings ratio of 16.66 with the price to sustainable earnings ratio of 19.65. If we were to use the estimate of next years earnings in place of our estimate of sustainable earnings, then our model would indicate an intrinsic value of $1202 and an expected return of 4.72%, a significant difference.

Morgan Stanley REIT Index It turns out that if we want to determine the intrinsic value of certain stock indices, accurate historical earnings per share data is not easily available. In this case, we can take a sampling approach, and determine the sustainable earnings per share of the ten largest companies in the index. We can then use the market cap weighted sustainable earnings per share of the sample as an estimate for the market cap weighted sustainable earnings per share of the asset class. Lets illustrate this procedure by estimating the sustainable adjusted funds per operation per share (AFFO PS) of the Morgan Stanley REIT index. Since the ten largest holdings account for 37.80% of the total market capitalization of this index, our estimate should be fairly accurate. Table 4-10 lists the ten largest (by market capitalization) REITs in this index, along with their sustainable AFFO per share, price to sustainable AFFO per share ratio, long-term issuer credit rating, weight, and intrinsic value. Note that the weight is not the weight in the index, but rather the companys market cap divided by the total market cap of the ten REITs. Also, the listed intrinsic value is actually the intrinsic value assuming a share price of $100, i.e., the sustainable earnings per share used in Equation 4-5 is calculated as $100 multiplied by the reciprocal of the price to sustainable AFFO ratio. Normalizing the stock price to $100 per share allows computing a simple weighted average of the ten REITs intrinsic value. Table 4-1069
Ticker Sustainable AFFO PS 4.62 2.18 3.00 2.51 Price / Sustainable AFFO PS 21.35 23.73 17.21 25.09 LT Issuer Rating BAA1 BAA2 BAA1 BAA1 Weight (%) 15.67 11.65 10.87 10.71 Intrinsic Value 72.10 64.07 63.16 60.58

SPG EQR GGP PLD


69

Source: Book/ModelChapter/IndexValuation/ReitIndex_Valuation

Page 87 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

VNO ASN EOP BXP AVB HST Weighted Average

4.97 1.93 1.57 3.44 3.65 0.63

24.62 30.18 30.72 33.30 34.86 38.94 34.86

BAA2 BAA1 BA1 BA2 BAA2 BAA1

10.42 9.07 8.50 8.14 8.11 6.84

59.79 50.37 47.92 44.20 43.61 27.92 55.07

Seeing that the weighted average intrinsic value for every $100 invested in this index is only $55.07, this (28 November 2006) might not be the best time to start investing in REITs by way of purchasing an index fund, as our real expected geometric rate of return would be only 3%. However, even with the high index valuation, there are a few REITs out there at a price close to their intrinsic value (but not many).

Chapter Summary
We first developed the concept of a companys sustainable earnings per share, which is a metric for a companys demonstrated earnings power that takes into account nonrecurring charges to earnings, as well as the companys business cycle. Note that simply applying the concept of sustainable earnings to the low PE strategy discussed in this chapter would yield improved results by providing a better estimate of each companys true earnings power. Next we looked at what is wrong with most valuation models used by stock analysts, and how we could create a valuation model that fixes these problems. We then developed a valuation model using a companys sustainable earnings per share and long-term issuer credit rating as company specific inputs, and historical statistics derived from the companys asset class as general model parameters, analyzed the models sources of prediction error, and demonstrated how well the model fits the historical data. We ended with several examples demonstrating the calculation of a stocks sustainable earnings per share and intrinsic value.

Page 88 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

5 Company Analysis: Cashflow Predictability (2009)


In Chapters 3 and 4 we learned how to compute a companys sustainable earnings per share and use this along with the companys long-term issuer credit rating to estimate the intrinsic value of the companys shares. Sustainable earnings are a useful quantification of a companys demonstrated earnings power, but what assurance do we have that this earnings power will persist into the future? None really, as there are many forces external to a company that can cause its earnings power to diminish over time, at no fault of the companys management (management screw ups being yet another cause of diminishing earnings power). The root cause of these forces is competition. If a company does not have a competitive advantage, then when new entrants into the companys industry increase supply and push down prices, the company must lower its prices, or risk losing market share. Similarly, the introduction of substitutes can reduce demand for a companys product. Both scenarios will result in diminishing revenue, which leads to diminishing earnings. Earnings are also impacted by the relative bargaining power of buyers and suppliers, and the intensity of industry competition. So although our estimate of a companys intrinsic value in the absence of selection bias - is probably accurate on average, a companys actual perfect foresight intrinsic value will vary considerably from our estimate, and will often be quite a bit lower (it will also often be quite a bit higher, but in that case, we probably wont complain). We can compensate for this in a couple of ways. First of all, we can gain a margin of safety by purchasing a companys shares at a significant discount to their intrinsic value. For example, if we only purchase the shares of a company that is trading at a ratio of intrinsic to market value of 1.5, then even if our estimate of the companys intrinsic value is optimistic by 50%, we still end up purchasing the shares at a price equal to their perfect foresight intrinsic value, and obtain a real 5.5% rate of return (5.5% being our discount rate). Still, with the inherent uncertainty concerning a companys future earnings power, and the fact that we introduce selection bias via non-random stock selection, how do we know that even this much margin of safety is adequate? In order to gauge the likelihood that a companys earnings power will persist into the future, we evaluate the company using three criteria. First we examine the companys operating history in order to quantify the extent of a companys demonstrated competitive advantage. Next we estimate the companys future prospects by looking deeper into the source of a companys competitive advantage, and analyze risks to the companys competitive position. In this step, we also analyze the companys industry, paying particular attention to the industrys profitability and stability - profitability being important because company profitability is often limited by industry profitability and stability being important because a competitive advantage tends to be more persistent in stable industries. Finally, since even a company with a strong competitive advantage can be forced into bankruptcy if it misses an interest payment on debt, we also take a close look at the companys financial strength. We then base our required margin of safety on how well a company scores on these three criteria. Page 89 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

5.1 Sources of Competitive Advantage


There are three major sources of competitive advantage: differentiation, cost, and focus70. For example, when a company produces a product that is perceived as superior to that of its competitors, it can charge a premium for the product without losing market share; this is differentiation. A company also possesses a competitive advantage if it can produce a product similar to that of competitors at a lower cost. Finally, a company can gain a competitive advantage by focusing on the needs of a customer that are not being met by less focused competitors; this takes the form of either cost focus or differentiation focus. To illustrate, the company might achieve a cost advantage by focusing on a narrow market, or create a differentiated product that does not exist because the total market for the new product was considered too small for less focused companies. Regardless of the type of competitive advantage, a company gains a competitive advantage by creating more value for the companys customers than competitors, through either a better product or a similar product at a lower cost. Differentiation Differentiation gives a company pricing power; even if competitors drop their price, a company with a differentiated product can often keep its price stable without losing market share. Now compare this to a company with a completely generic product that cannot be differentiated from its competitors; such a company will lose significant market share if a competitor drops its price unless it quickly follows suit. With this simple example, we can generalize what it takes to create a differentiated product. First, the companys product must be recognizable from that of its competitors; clearly a cord of wood fails this test while a can of Coke passes. Beyond that, the product must be perceived as superior to its competitors. In the case of soft drinks, beer, cigarettes, or packaged food, it might simply taste better, whereas for an automobile, passenger jet, or power-generating turbine, the product might have a reputation for quality. A product can also be differentiated by the cost savings created for the companys customers, an example being GEs composite fan jet engines that allows Boeing to create a more fuelefficient aircraft, which in turn saves airlines money through increased fuel economy. Although the competitive advantage imparted through differentiation can diminish with time, this rarely happens quickly. Lets look at a delivery service such as UPS, which has built a reputation for delivering packages on time, anywhere in the world. Now even if a competitor (lets call it Deliveries R us) can invest billions of dollars and quickly build a global package distribution network, which one would you choose if your job depended on it? Even if UPS charged more, you at least know you wont get fired for using UPS, whereas if this upstart lost your package, the obvious question your boss might ask is Why didnt you use UPS? This same argument holds for many other products and
70

From Michael Porters book, Competitive Advantage

Page 90 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

services whose performance can have a material impact on the operations of a business that uses them; it can take many years for a new entrant to earn a reputation for reliability. A similar case can be built for brand persistence with consumer products. Once you have found a product you are happy with, you might not want to take the risk that another product at a lower price will be of the same quality. With more expensive items such as automobiles, the cost of purchasing a lemon is high, so you might limit your car shopping to trusted brands that either you have a long personal experience with, or where independent research indicates a high level of customer satisfaction. On the other hand, with other items such as your favorite brand of beer, research really cant tell you if you will enjoy an alternative brand, and you probably dont have the time to purchase a sample just to see if it is better. As a matter of fact, when you are in a hurry, you probably just grab your usual brand and move on, without even checking the price. Of course, given enough time and effort, it is possible that a competitor will create a product with superior quality than that of a company you are analyzing. An example of this is Japanese automobile manufacturers. At one time, any product manufactured in Japan was considered to be of inferior quality, and most people would not dream of owning a Japanese automobile. However, over time, it became apparent that Japanese companies were proficient at manufacturing, and the quality of their automobiles were quite good; eventually, they were perceived as being of superior quality to their American counterparts. Of course this change in perception took several decades, but it did happen, and it became necessary for companies like General Motors to improve the quality of their cars to rebuild their brand, which they did over time. The key point is that over long time periods, the strength of a companys competitive advantage can change, which is why I prefer to leave a companys competitive advantage out of the intrinsic value calculation, but consider it in the purchase decision. Cost If a company can produce and distribute its product at a lower cost than rivals, it can often gain market share while at the same time maintaining profitability similar to competitors. Lower production costs can be achieved in many ways, including investment in the continuous improvement of manufacturing productivity, outsourcing the production of components to a company that can do it cheaper, setting up manufacturing in a country with lower labor costs, and the automation of factories. Similarly, an efficient supply chain can decrease a businesss inventory, thereby increasing return on assets, while at the same time insuring that a factory is never short of raw materials and a store never runs out of any item. Supply chains can also be set up to minimize transportation costs by combining rail, shipping, and trucking in an optimal manner. Often, labor costs can differ substantially within the same country; a good example is Wal-Mart vs. unionized retailers. A company can build a strong brand on the basis of cost, a couple of examples being Wal-Mart and Home Depot.

Page 91 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

5.2 Quantifying Demonstrated Competitive Advantage


A sustainable competitive advantage will allow a company to consistently: Achieve above average returns on the capital employed in running the business, Retain a large portion of the value created for customers as profits, And if the companys market is growing - provide the opportunity to re-invest earnings at high rates of return.

Although profit margin and return on assets can vary widely by industry, I prefer to focus on the absolute (rather than relative) value of these metrics. My reasoning is that I would much rather own a company with average profitability in an extremely profitable industry such as Glaxo Smith Kline in the pharmaceutical industry - than a company with above average profitability in a largely unprofitable industry such as Southwest Airlines in the airline industry. My rationale for this is that the sustainability of a companys competitive advantage is intricately linked to the stability and profitability of the companys industry, a topic we will return to soon. Time in Industry When evaluating a companys track record, we need at least ten years of data over which the company has operated in the same industry, and if the industry is cyclic, we need at least enough years to capture a full industry cycle. This usually means evaluating a companys operating history over the same period used to calculate the companys demonstrated earnings power. Most often, ten years of history is sufficient, except in the case of some companies whose profits largely depend on the price of some commodity like oil or steel, or when the last ten years have witnessed abnormally stable economic growth. Note that it is critical that the company has been competing in the same industry over the time period, and that the industry has been fairly stable over the time period as well. Operating Profit Margin A companys operating profit margin is calculated as the ten-year average of the ratio of operating profit (adjusted for non-recurring items) to revenue. We assess profitability using operating margin rather than profit margin because operating margin is a better metric of competitive advantage; this is because profit margin is distorted by differences in tax rate and interest expense that while important, do not relate directly to competitive advantage. The operating profit margin indicates the proportion of the value created for its customers (measured by revenue) that is captured by the company as opposed to being competed away - and is therefore a useful measure of the extent of a companys competitive advantage.

Page 92 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Sometimes operating margin can be distorted in industries where a company makes its profits by taking on credit risk (banks) or event risk (insurance). In this case, operating margins may be very high over long time periods (sometimes decades), until some event occurs (usually an economic downturn) where losses wipe out many years (or even decades) of earnings. Obviously in such a scenario a companys operating margin might seem to be evidence of a competitive advantage, but most of the profits are illusory, and the long-term profit margin is much lower than what might be experienced over a typical ten-year period. The banking industry71 is a good example of this; in the 1980s a banking crisis wiped out many years of cumulative earnings, and a similar situation occurred in 2008, where incidentally the bond insurance industry also imploded. These types of business models are aptly described as picking up nickels in front of steam rollers; the profits may look impressive for a while, but if you stumble, its all over. Fortunately these business models can be easily spotted, as they are characterized by a relatively low return on assets, which brings us to our next metric. Return on Adjusted Assets Our second metric is a companys return on adjusted assets, which we compute as the ten-year average ratio of operating profit (adjusted for non-recurring items) to adjusted assets. Here we adjust reported assets by subtracting net cash and cash equivalents, and subtracting goodwill and any intangible asset with an indefinite life (such as trademarks). We subtract cash net of debt from assets because we want to determine the rate of return on productive assets, and do not want to penalize a company that decides a cash buffer is prudent72. The reason we subtract goodwill and trademarks is that they are not part of the capital employed in running the business, but rather they allow the company to earn a higher return on this capital, as well as on additional tangible assets purchased for expansion. This makes our definition of a companys adjusted assets roughly equal to the capital employed in running the business. It is fairly intuitive that return on adjusted assets is a good metric for determining company quality. Imagine two different businesses, each with $100,000 in productive assets. One generates $30,000 in operating income each year, and the other generates only $10,000 in operating income each year. If both were available at the same price, which one would you rather own? If there is any opportunity at all to expand production by using retained earnings to expand the businesss asset base, it is pretty clear that the company with a 30% return on assets will grow operating income faster than the company with the 10% return on assets a dollar of retained earnings will produce $0.30 of additional operating income in the first case as opposed to $0.10 of additional
71

A banks profit margin is also highly sensitive to its net interest margin, which is typically around 3%. If a bank makes half of its revenue off of this spread, then it is easy to see that a 50% compession in this margin would wipe out 1/4 of the banks revenue, and if prior to compression the bank had a 25% pre-tax profit margin, pre-tax profits would fall to zero after the compression. 72 Although an inordinately large cash buffer accumulated by using most of retained earnings to build a cash hoard will usually result in a low return on retained earnings, which we discuss next.

Page 93 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

operating income in the second case. It follows that, all else equal, a dollar of earnings from the company with the 30% return on assets is worth more than a dollar of earnings from the company with the 10% return on assets. Moreover, unless a company possesses a competitive advantage, historically high returns on assets will attract new entrants into an industry and drive down returns. Therefore a history of consistently high returns on adjusted assets is indicative of a competitive advantage. Besides providing the opportunity to invest retained earnings at a high rate of return, a high return on assets provides additional benefits, in that it shields a company from some of the pitfalls resulting from a low return on assets. Because capital expenditures are necessary to keep a companys asset base competitive, a low return on assets implies a high ratio of recurring capital expenditures to operating income. Moreover, these capital expenditures can be considered fixed costs in that if a company reduces these expenditures during a downturn, it will not be well positioned for the recovery. However, these fixed costs also make it likely that a companys profits will have a high sensitivity to changes in revenue. Besides making a companys earnings volatile, it also reduces a companys financial strength by making it likely that an extended downturn could impair the companys ability to meet its debt obligations. Although financial companies do not have sizeable physical assets, a low return on financial assets creates the risk of small changes in asset value often caused by loan losses - wiping out earnings. Another issue is that inflation (which is not captured by depreciation expense) often results in companies with a low return on assets having recurring capital expenditures that are well in excess of depreciation cost. In this case, since a sizeable portion of earnings must be retained just to make up the difference between recurring capital expenditures and depreciation expense, the companys sustainable earnings per share will overstate the companys true earnings power. Finally, a low return on assets can also encourage risk taking. To illustrate, if a company earns a low return on assets, it will likely feel compelled to use excessive leverage to provide its shareholders with an acceptable return on equity. Return on Retained Earnings Although a high return on assets is good, it gets even better when a company can consistently invest new capital at high rates of return. We measure this by calculating a companys Return on retained earnings, which is calculated as the change in a companys per share earnings over a ten-year period divided by the sum of the companys per-share retained earnings over the same period, with both earnings and retained earnings adjusted for both inflation and non-recurring items. A companys return on retained earnings measures the return a company has achieved on earnings that are either re-invested back into the business over the period or used to repurchase the companys stock.

Page 94 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

The need for this metric is obvious when we look at extreme cases. If a company has retained half of its earnings over a ten-year period, and the earnings per share (adjusted for inflation) are the same as they were ten years ago, then the company has not created any shareholder value from the reinvestment of these earnings. This is not the same thing as saying that the company would have been better off paying out these earnings as dividends; without the earnings retention, it is quite possible that todays earnings per share would be much less than what they are. Still, if you have the choice between purchasing this company today or another company that has doubled its earnings per share over the last ten years while retaining half of its earnings, and both companies are selling at the same ratio of price to sustainable earnings, which would you prefer? While there is no guarantee that each companys return on retained earnings will persist going forward, it is also unlikely that the competitive advantage responsible for a high return on retained earnings will immediately disappear. We need to keep in mind that as with our profit margin metric, a high return on retained earnings is not always indicative of a competitive advantage. A company can earn a high return on retained earnings by using high levels of leverage to boost the return on assets funded through retained earnings; this was the case for banks and some insurers until 2008 when the leverage blew up in their face. But when a high return on retained earnings is coupled with a high return on assets, it is likely that the company has a competitive advantage giving it the ability to organically grow its asset base through retained earnings while continuing to earn a high return on those new assets. The simplest way of computing a companys return on retained earnings is to take the difference between the current earnings per share and the earnings per share ten years ago (in todays dollars), and divide this difference by the sum of the retained earnings over the period, where the retained earnings in any given year are found by subtracting the annual dividend from the earnings per share, with the difference then adjusted for inflation. This approach has many of the same drawbacks of using reported earnings rather than sustainable earnings. One problem is the dependence on only two years of earnings as a measure of shareholder value created. To see why, imagine that the company had a really bad year that coincides with the start of the period, and an exceptionally good year at the end of the ten-year period; this would tend to exaggerate the value created over the period. Instead, I prefer to estimate the difference in real earnings by first calculating sustainable earnings per share, and then using the earnings per share growth rate that optimally detrends the inflation adjusted earnings per share time series (lets call this GDT ) to derive the earnings per share at the start of the time period from the companys sustainable earnings per share (SEPS):

EPSStart =

SEPS (1+ GDT )10

From here we can calculate the return on retained earnings as:

Page 95 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

RORE =

SEPS EPSStart
10

CpiAdj(EPS DPS )
i i i=1

Here EPSi is the companys earnings per share (adjusted for non-recurring items) in year i, DPSi is the companys dividends per share in year i, and CpiAdj() is a function converting the difference to todays dollars. The return on retained earnings metric does discriminate against capital-intensive industries that require recurring capital expenditures that are considerably higher than depreciation expense; but since these companies do not have as much discretionary profit left to enhance shareholder value, that is kind of the point. To give you an idea of a typical companys return on retained earnings, for the SP500, it was 7.4% in the ten-years ending in 2007 with a ten-year rolling average of 4.6% from 1950 to 2007. Increasing leverage over the measurement period can boost return on retained earnings. As a check on this, we can also compute a companys return on retained earnings plus change in net debt. The calculation is identical to that of return on retained earnings, except that we add the change in net debt over the ten-year period to the accumulated retained earnings over the period. The change in net debt is calculated by subtracting the companys net debt in todays dollars at the beginning of the measurement period from the companys net debt at the start of the end of the measurement period, which would be the most recent year. I prefer return on retained earnings to alternate measures of a companys return on reinvested earnings, such as return on equity. One reason is that return on equity can become very large when a company uses debt to repurchase stock, as Budweiser has done in the past; not only do the earnings per share increase due to the stock repurchase, but shareholders equity decreases as well, due to the decrease in paid in capital. Even without the use of debt, share repurchases can dramatically reduce paid in capital, resulting in triple digit return on equity73 for UST, inc., whereas the companys return on retained earnings is a more modest 2.73%. Historical Performance Rating A companys operating history is quantified on a scale from 1.0-4.0, with 4 being best, and 1.0 being our minimum where we would consider purchasing stock in the company. Table 5-1 roughly illustrates the thresholds used to quantify the three metrics, and we use linear interpolation between the thresholds shown in the table (capping the maximum ranking at 4.0) to compute the actual rating for each metric. We then calculate the companys combined historical performance rating by averaging the ratings for each of the three metrics. Note that to be eligible for purchase, we require a company to meet each minimum threshold for return on adjusted assets, operating margin, and return on retained earnings.
73

This is even after we add back in the companys litigation losses to shareholders equity

Page 96 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Table 5-1
Metric Return on Adjusted Assets Operating Margin Return on Retained Earnings 1 > 16% > 12% > 6% 2 > 20% > 16% > 10% 3 > 24% > 20% > 14%

The minimum thresholds were chosen to be slightly above the median value computed from 1950 to the present. Of the 415 non-financial74 companies in the SP500 at the end of 2008, only 65 met all the minimum thresholds for these metrics, and only 15 were ranked three or higher for all three metrics. Of these same 415 companies, if we remove the requirement that a company score at least a 1 in each category, average all three metrics for each company, and look at the distribution of historical performance rankings, we generate the data presented in Table 5-2. The distribution is almost identical when repeated at year-end 2007. Table 5-2
50% > 0.88 45% > 1.11 415 Non-Financial Companies 40% 30% 20% 10% >1.29 >1.62 >2.09 >2.52 5% > 2.98 1% > 3.39

***** Note that these three metrics are complimentary. The return on adjusted assets metric can act as a check against (temporarily) high profit margins obtained through the picking up nickels in front of steam rollers strategy. It does this by ruling out companies that convert a small return on assets to an acceptable return on equity using large amounts of leverage, thereby creating the risk that an economic downturn can destroy many years of profits. But sometimes a companys adjusted assets used in the return on adjusted assets calculation gives a distorted picture of the actual capital required to run the business. This is the case when the company employs long-lived assets where the depreciation expense substantially understates the recurring capital expenditures required to keep the assets competitive. When recurring capital expenditures are consistently in excess of depreciation, the difference comes out of retained earnings. When this difference is large, the companys return on retained earnings suffers from the

74

As explained later, we do not invest in financial companies due to the high leverage employed. They were left out of this study because with their exceptionally low return on assets and high profit margin (due to revenue being computed using net rather than total interest income), they skew these statistics.

Page 97 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

effects of a large portion of retained earnings going to the maintenance of the companys competitive position rather than expansion (or stock buy-backs). The reason that we are not satisfied with the performance of a company with a record of high returns on assets but a low profit margin is that we believe evidence of pricing power is indicative of a more sustainable competitive advantage. As an example, consider Wal-Mart, an excellent company with a strong cost advantage. Return on assets has historically been close to 15%, but the companys operating margin is on the low side at around 6%. The companys high inventory turnover compensates for this, by allowing the company to quickly scale inventory to meet demand, so the company has been consistently profitable. And as long as the industry as a whole faces the same changes in input costs (inventory, fuel costs, and labor being some examples), Wal-Mart should be able to maintain its cost advantage, and maintain (or grow) market share while at the same time maintain its profit margin. After all, the company does sell basic consumer necessities that must be purchased somewhere; this allows the industry to pass these costs on to consumers. Still, what would happen if their cost advantage were eroded, perhaps due to unions making inroads into their labor force? Or perhaps another rival will enter the industry that can make similar profits at even lower prices? This has already happened in the United Kingdom, where Tesco, a company similar to Wal-Mart, has begun losing market share to Aldi, a German retailer that focuses on a single low cost brand in every category and provides even lower prices to customers. Although any company faces similar risks, a company with pricing power, as demonstrated by historically high operating margins, has a larger margin of safety. For example, whereas Wal-Mart, with earnings per employee of around $7000 annually, would face a sharp hit to earnings if collective bargaining increased the average salary per employee by $5000 a year, with no increase in productivity75, Coca-Cola, with earnings per employee of $80,000 a year would face a much smaller impact to profitability. And if a rival were to sell a cheaper soft drink, it would likely have a limited effect on Coca-Colas market share. This is because CocaColas product is differentiated, unlike Wal-Mart, which sells the same product as other retailers. This differentiation gives Coca-Cola pricing power, which is evident in its high operating margins (28% over the last ten years). Differentiation can even allow a company to partially compensate for falling demand by raising prices, in which case the incremental fall in demand is less than the revenue gained by increasing prices; the cigarette industry in the United States and Western Europe is an extreme example of this kind of pricing power revenue has continued to grow despite a fall in volume. In general, we can make the argument that a high profit margin stemming from pricing power which in turn is a result of a differentiated product - reduces the impact on profitability from adverse changes in any of the competitive forces that determine industry profitability (discussed in more detail in the next section).

75

This assumes that the increased labor bargaining power brings labor costs closer to that of unionized rivals. Wage pressure applied evenly to all industry participants would result in the costs being passed on to customers.

Page 98 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Other Performance Metrics There are many other performance metrics that while not included in our rating, nevertheless give us additional confidence in the quality of the companys documented performance. Three of these that I find particularly useful are organic revenue growth, organic volume growth, and changes in the local currency price of the companys goods and services. Organic revenue growth is computed as the annual change in the companys revenue, adjusted for the effects of acquisitions and dispositions, as well as the effects of changes in exchange rates, and consequently provides a clearer picture of the growth rate of the companys business. We can break down this growth into two components. Organic volume growth is the change in the amount of product a company sells, adjusted for acquisitions; this might be measured as cases of bottled beverages, packages of cigarettes, or boxes of diapers. A long-term positive growth in organic volume means that the companys total available market is increasing, the company is gaining market share, or both. Changes in the local currency price (meaning the price at the point of sale, as opposed to converted back into the currency in which the company reports its financial statements) gives an indication of the companys pricing power. If the company can consistently increase prices for its goods and services without a corresponding decrease in volume, the company likely has a significant competitive advantage. On the other hand, provided that organic revenue growth is positive, a lack of consistent price increases is not necessarily a red flag, as the company might be growing market share through low pricing.

5.3 Future Prospects


Having found a company with a track record indicative of a strong competitive advantage, we need to convince ourselves that the competitive advantage is sustainable well into the future. To do so, we must look deeper into the source of a companys competitive advantage using the five forces that determine industry profitability: supplier bargaining power, buyer bargaining power, the threat of new entrants, the threat of substitutes, and the intensity of rivalry between existing firms. This framework for analyzing industry profitability comes from Michael Porters book Competitive Advantage, and this section draws heavily from what I learned from this book, which I recommend reading. What we are looking for is an industry where the incumbents are likely to enjoy a very long period of stable and highly profitable operations. If our company is one of the leaders in such an industry, we gain confidence that the companys demonstrated competitive advantage will persist far into the future. As an example of why industry stability is important, lets look at the retail industry. At one time Sears was considered to be a company with bright future prospects (it was a component of the 1972 nifty-fifty), but the companys competitive position was eroded with the emergence of K-mart, which had a cost advantage over Sears. In time, K-marts cost advantage was reduced by Wal-Mart, which had an even larger cost advantage. If you had invested in either Sears or K-mart on the basis of their demonstrated competitive Page 99 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

advantage, you would have been disappointed, as both eventually lost their competitive position within their industry. As for Wal-Mart, it does have an impressive track record but then so did Sears and K-mart. In general, a competitive advantage gained by becoming the low cost provider in an industry is often less sustainable than a competitive advantage gained through differentiation. Whereas multiple companies with differentiated products can often remain highly profitable, having two or more competitors with similar scale that compete on the basis of cost will often lead to a price war, which will result in falling profitability for the industry leaders. A competitive advantage in the technology industry is often even shorter lived than in the retail industry. It was not that long ago when Alta Vista was the dominant company in the on-line advertising industry, but it has recently been supplanted by Google. For some other examples of technology companies that were once though to have excellent prospects but eventually went the way of the dodo bird, there was Digital Equipment Corporation, Burroughs Corporation, and Polaroid. If you are searching for companies that at one time were expected to have extraordinary future prospects, but where the companys competitive advantage was not sustainable, the technology sector is a good place to look76. Now contrast the retail and technology industries to the disposable diaper industry, where since the 1960s, the same two companies Procter & Gamble and Kimberly Clark have dominated the market. The Coca-Cola Company and Pepsico have dominated the soft drink industry even longer. Another example of a stable industry is the jet engine industry, where General Electric, United Technologies (through their Pratt & Whitney subsidiary), and Rolls Royce have dominated for many decades. With this in mind, given a choice between Abercrombie & Fitch a company with an excellent ten-year operating history and Coca-Cola, I would choose purchasing shares in Coca-Cola. The nature of the soft drink industry, and Coca-Colas competitive position within that industry, gives me much more confidence that the companys earnings power will persist far into the future as compared to an excellent company in the retail industry. We analyze a companys industry using the five forces of competition. An industry will have favorable dynamics if the following conditions hold: Threat of new entrants is low Intensity of rivalry is low Threat of substitutes is low Supplier bargaining power is low Buyer bargaining power is low

76

Of course there are some exceptions like Microsoft and IBM.

Page 100 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

In the following chapters we will show some examples of analyzing a companys industry within the framework of the five forces of competition. In the following section, we will examine some common attributes of stable and profitable industries.

Some Attributes of Stable and Profitable Industries High Switching Costs Switching costs may deter a buyer from switching to a new supplier; this often occurs when switching to a new company would entail a significant loss of revenue. Switching costs can be incurred when switching to a new supplier requires re-tooling, learning new processes, or system downtime. Switching costs can be high when qualifying a product from a new vendor is both time consuming and costly, as can be the case with aircraft engines, power turbines, and other equipment where the cost of failure to the purchasing company can be extreme. Sometimes switching costs are stipulated in long-term contracts in the form of contract termination feeds. Switching costs increase buyer bargaining power, reduce the threat of substitutes, and can act as a barrier to entry. Proprietary Technology Some industries require extensive manufacturing know-how that can only be gained through years of experience. A case in point is the jet aircraft engine industry. Although the science of jet propulsion is public knowledge, producing jet engines that are competitive with respect to fuel economy, noise, and reliability is another matter. The difficulty of entering this industry is apparent from the fact that there are only three major players worldwide: General Electric, United Technologies, and Rolls Royce. Since it is unlikely that new entrants will be competitive until they gain the expertise of the industry incumbents, the long learning curve will tend to discourage new entrants. A companys product can also be protected from direct competition by patents. A good example of this is a pharmaceutical company, where new drugs are covered for a limited time by patents. In this case, while a competitor can attempt to come up with their own drug to meet the same medical need, they cannot copy the companys exact molecular compound. When it is difficult for a competitor to develop a substitute, a patent gives the company many years of a virtual monopoly on that particular drug, with the lack of competition enabling the company to charge a high price per unit without losing market share. Restricted Access to Distribution Channels Distribution channels can act as a barrier to entry. One example is the soft drink industrys sales to supermarkets and convenience stores, where there is typically intense Page 101 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

competition for shelf space. Consequently, a new entrant might find it hard to persuade a retail chain to allocate space for its product, as if it does poorly, the retail chain will suffer. The automobile industry is another example where distribution channels create a barrier to entry; not only must a new entrant be able to manufacture a quality product in large volume, but it must also have a dealer network in place in order to reach potential customers. Strong Brands New companies are less likely to gain market share in an industry when buyers associate the brands of industry incumbents with exceptional quality, taste, or other sources of value. With consumer goods companies, this brand loyalty is the result of the development of innovative products supported by high advertising expenditures, which enhance a consumers preference for the product. Manufacturers on the other hand might have a brand supported by a large installed base with a low failure rate. When each of the industry leaders has a strong brand, the intensity of rivalry is lessened.. For example, Coke has little incentive to cut costs to gain market share, as some people just happen to prefer Pepsi, and the cost of the product is low enough where they are unlikely to be swayed by price differences. Here the market share gains (if any) would unlikely be sufficient to offset the reduced cost per unit. Undifferentiated Inputs When a supplier has a product that is differentiated from other suppliers this enhances the bargaining power of the supplier, and thereby reduces a companys profitability. In this case, during price negotiations the company cannot make a credible threat to replace the suppliers product. Low Supplier Concentration A companys bargaining power with a companys suppliers increases when each supplier provides a small fraction of the companys purchases, and the company accounts for a large percentage of each suppliers sales. Here a company can play off the suppliers against each other by offering to increase the volume purchased in return for a reduction in price; of course the company would be careful to insure that any one supplier does not gain a large share of the companys business. This is easy to do; once the price reduction occurs, other suppliers will follow suit, and next year the company can make a deal with another supplier, and the process repeats. Low Price Sensitivity

Page 102 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

It seems intuitive that a buyer will spend more time investigating the extent of a products value if the total amount spent on a product becomes a significant fraction of total expenditures. One example of this can be found in the automobile industry. Even if you have been happy with the performance and reliability of a certain make of automobile, you might not necessarily be willing to pay a large premium for the same make when it comes time to purchase a new car. Instead, you would likely spend a considerable amount of time researching other makes in Consumer Reports, and then carefully analyze whether any price premium is worth the likely benefits. We can contrast this with the purchase of razor blades, which in any given year represent a much smaller fraction of a consumers expenditures than automobiles. Here, if you have been happy with a particular brand, you are unlikely to invest a lot of time and effort in researching which brands actually offer the best shaving performance for your money. The likely cost savings would just not be worth the effort. This creates a certain amount of market share stickiness, where an incumbent can charge a premium for their product and still not experience significant loss in demand. High Product Differentiation A companys bargaining power with buyers increases when the companys product is differentiated from that of competitors. In the absence of differentiation, a buyer can make a more credible threat to switch to buying its product from a competitor. If the product has a large effect on the buyers profits, then the companys bargaining power is increased even more. Buyer Concentration A buyers bargaining power increases when it accounts for a significant fraction of a companys sales, particularly when the companys sales are a small fraction of the buyers total purchases. This is just the flip side of the scenario described under Supplier Bargaining Power. Low Threat of Substitutes Companies can lose their competitive advantage if a new product or service is created that better meets the needs of the customers served by the industrys incumbents. An example of this is when cellular networks and Internet telephony created a substitute for fixed line telephone service. This resulted in rapidly declining revenues from what used to be the telecom industrys cash cow. Although the incumbents in the telecom industry survived (although most were acquired during a wave of acquisitions), they only did so by rapidly embracing these substitutes and turning them into new sources of revenue. The threat of substitutes cannot always be avoided by embracing the substitute, as the companys competitive advantages may not be applicable to the new product or service,

Page 103 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

and sometimes (as with the case of digital photography replacing film-based photography) the new industry may not be as profitable as the old. *************** Quantifying Future Prospects Once we have analyzed a companys industry, and the companys position within that industry, we quantify a companys future prospects by rating the company on a scale from 1.0 to 4.0. After analyzing the source of a companys competitive advantage and the dynamics of the companys industry you should be able to identify at least one major risk to the companys competitive position. If you cant find a major risk, then you have not delved deep enough into the source of the companys competitive advantage. Because we cannot quantify future prospects through ratio analysis as we do for our evaluation of a companys operating history, we will not allow the future prospects rating to exceed that of the historical performance rating. Some companies that rate over 3.0 are Johnson & Johnson and Coca-Cola. At the other end of the scale, a 1 indicates a company meeting our minimum requirement for competitive advantage. Here the companys competitive advantage must still be sustainable, but there is a higher risk of erosion over time. Some examples of companies that we would rate close to 1.5 are IBM and McDonalds. Note that a rating of 1 still puts a company among the best of its peers; it just puts it towards the bottom of the range we would consider for purchase.

5.4 Financial Strength


One obvious prerequisite for cashflow predictability is for the company to stay in business. Companies with a high degree of financial strength are better able to withstand the downturns that occur in any industry, and can often pick up assets on the cheap during the downturn. Beyond avoiding bankruptcy, a high level of financial strength gives the additional benefit of a lower and much more stable cost of debt capital. For those industries with a relatively low average interest coverage ratio (such as REITs), this stability of interest expense can give a large reduction in earnings volatility as compared to their less fortunate competitors. With this in mind, we only invest in companies with investment grade credit ratings on all outstanding debt and preferred stock. Limiting ourselves to investment grade companies substantially decreases our risk. To see why, consider the fact that in a study carried out using NYSE data from 1951 to 199877 over half of the companies in existence at the start of a ten year period were out of business by the end of the ten year period. Now compare this to investing in a portfolio of stocks at the lower end of investment grade (BAA3 credit rating), where over 90% of the companies would still be in existence
77

The level and persistence of growth Rates, Chan, et al

Page 104 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

after ten years78. The difference is even more extreme for those companies with the highest credit ratings (AA or higher), where on average 98% survive after ten years. For a given amount of leverage, a companys credit rating will be higher when its future cashflows are more predictable. It follows that beyond giving us confidence that a company is unlikely to default on its debt, a high credit rating is also indicative of a sustainable competitive advantage. As a matter of fact, the influence of competitive position dominates that of liquidity and leverage for the purposes of a companys credit rating. As an example, in the heavy manufacturing industry, Moodys only gives a 15% weight to liquidity and capital structure. Consequently, we can use a companys credit rating as an independent analysis of a companys competitive position. Our evaluation of financial strength is supplemented by the market-implied ratings (MIR) of the companys senior debt. Over long time periods, the spread of the companys bond yields as compared to that of treasuries can be correlated to a default rate, just like the companys long-term issuer rating. When the bond market believes that the expected value of a companys default rate increases, or the level of uncertainty regarding future default increases, the spread will widen, and vice versa. Since the bond market is fairly efficient, there is useful information in these yield spreads. A companys MIR can be found on the Moodys Investor Services website, where in addition to bond MIRs, there are also MIRs based off of the spreads on credit default swaps and the market price history of the companys common stock. However, I ignore the MIR derived from stock price history because the stock market is much less efficient, and therefore the price changes have less information content after all, stock market inefficiency is one of the necessary conditions for our investment strategy to work. If there is a large gap between the long-term issuer rating and the bond-market implied rating of a company we own, or the bond MIR falls below investment grade, then we will immediately try to identify the reason behind the bond markets pessimism, and evaluate any new risks that the company faces. Rather than put full faith in a companys credit rating, we also undertake our own analysis of a companys financial strength. When doing so, we look at how insulated a company is from different potential shocks, such as: Interest Rate Shocks It is best when a company can pay down maturing debt using retained earnings. When this is the case, the company is insulated from the effects of a large spike in interest rates (or even a shut-down of capital markets), as maturing debt can be paid off instead of rolled over, even while maintaining the dividend. A high ratio of operating income to fixed charges is also desirable, although the minimum requirement for this ratio depends on the expected stability of the companys earnings; four to one might suffice for consumer staple companies, whereas a ratio of 8:1 or higher would be preferred for a

78

Using Moodys default rate data from 1920-2005

Page 105 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

heavy equipment manufacturing company. A company gains additional flexibility when it can pay off a significant fraction of total debt out of a years earnings. Liquidity During normal times investment grade companies have access to the commercial paper markets, but occasionally this funding source dries up. In such a scenario, a company would need to fund current liabilities (liabilities due within the year) solely through current assets (assets that can be readily converted to cash). A companys current ratio is the ratio of current assets to current liabilities, and if greater than one, indicates that the company can fund its current liabilities (like maturing debt and accounts payable) out of current assets such as accounts receivable, cash and marketable securities, and the sale of finished inventory. The quick ratio is a more pessimistic test that assumes that inventory is not readily convertible into cash, and therefore subtracts inventory from current assets before computing the ratio. Foreign Exchange Shocks We prefer a company that largely manufactures products in the same countries they are sold; this naturally hedges currency fluctuations. When inputs are denominated in a currency different from that used to purchase the products, a sharp change in the exchange rate can have a large impact on profit. This can sometimes be moderated using currency derivatives, but this is not always cost effective in developing countries. A company should not attempt to use derivatives to prevent a weak currency in a foreign subsidiary from reducing a companys overall profit; that would be currency speculation. In other words, a company should use derivatives for transactional hedges (to compensate for a mismatch between the currency used to purchase inputs and manufacture the product, and the currency in which the product is old), but should not use derivatives for translational hedges (used to compensate for a mismatch between the currency where the product is sold and the currency used to report earnings). Finally, a company should be able to service interest and repay principal of debt issued in a given currency with earnings that stem from sales in that same currency. Customer Defaults For financial companies or companies with a financial subsidiary; we evaluate the quality of the loan book, and look at the impact to earnings and capital adequacy in the event of a spike in default rates. For REITs, we evaluate the credit quality of tenants, and the value of the vacant property. For companies that sell a product or service, the credit quality of any major customer is important. A Spike in the Cost of Sales

Page 106 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

In this analysis we will assume that depreciation expense is stable, and look at the effect of an increase in the cost of sales less depreciation expense that is included in the cost of sales. This adjusted cost of sales includes such as manufacturing labor, freight costs for raw materials and finished product, raw materials, and leasing cost; an impact in any one of these cannot be passed on to customers will result in a drop in gross profit, and ultimately, operating profit. The sensitivity of a companys operating profit to changes in the cost of sales that cannot be passed on to buyers is given in terms of the cost of sales (COS), depreciation included in cost of sales (DEPR), and the companys operating profit (OP):

OP COS DEPR = COS OP


A Spike in Employment Costs

Inflation caused by wage-price spirals can create significant increases in a companys employment costs. This can occur when employees expect prices to increase, and have the bargaining power to demand corresponding increases in wages. This is more of a problem with a unionized workforce, particularly if there are cost of living increases built into long-term contracts. Whatever the cause of an increase in employment costs, a company will be less affected when total employment costs are a smaller percentage of the companys operating profit. Some companies do not disclose total employment costs, so instead we can look at operating profit per employee, with a higher number indicating a lower sensitivity to an increase in employment costs.
A Drop in the Companys Pension Value The funded status of a companys pension plan can present a risk to a companys financial strength, as a decline in the value of plan assets (always possible during a bear market) can in some cases require a company to make contributions that are a significant fraction of the companys earnings. A pensions funding status is given by the difference between the plans assets at market value and the projected benefit obligation (the sum of future benefits discounted to the present), with a positive value indicating over-funding and a negative value indicating under-funding. Under the new ERISA requirements79, a company has seven years to make up a pension shortfall, over which the minimum company contribution is equal to the target normal cost, which is equal to the benefits expected to be paid out in the current year. We quantify pension risk as the ratio of the expected benefits to be paid in the current year to the companys pre-tax earnings. When computing this ratio, I also add in postretirement health benefits. I prefer this ratio to be less 10%, and avoid companies where it is larger than 20%. To get an idea of the level of this ratio that indicates big trouble, in 2004 General Motors paid $11,215 million in benefits, and earned $1,894 million pre-tax, giving a ratio of 590%. Being charitable, we
79

CRS Report for Congress, April 10, 2008

Page 107 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

could use General Motors cash flow from operations in place of pre-tax earnings, bringing the ratio down to 116%, still in considerable excess of our upper limit. A Decline in Revenue A company is at lower risk if it can adjust its costs to match changes in revenue relatively quickly. The initial sensitivity of a companys operating profit to a change in revenue depends on the companys operating margin and the ratio of fixed costs to revenue, where variable costs are those that the company can quickly (within one year) adjust in response to a decline in demand for the companys products. Here we will assume that those variable costs consist of manufacturing labor and raw materials, where labor is ideally adjusted by shortening shifts as opposed to layoffs, which can create severance costs. Unfortunately labor costs and the cost of raw materials are only broken out under IFRS, so the best we can do is make an approximation. We do this by assuming that the bulk of a companys cost of sales aside from depreciation expense consists of labor and raw materials. We can then compute the sensitivity of the companys operating profit (OP) to a change in revenue (R) using the companys operating margin (OPM), cost of sales ( COS ), revenue (R), and depreciation expense (DEPR):
OP 1 COS DEPR = 1 C OPM R

This equation is not completely accurate, as cost of sales also includes costs such as freight costs (which are also variable, and can be quickly cut to match shipments), leasing costs, and changes in inventory, but it is the best we can do unless the company reports under IFRS. However, if management believes that the change in revenue might persist for more than a year or two, it can make the decision to make further cost cuts to bring expense in line with the reduced revenue. Consequently, for longer-term drops in revenue, the companys profit will first fall, and then eventually rise as plants are closed, SG&A reduced, and other cost cutting actions take effect. Ideally, the operating margin would eventually stabilize close to what it was before the drop in revenue; the profit margin will be lower if there is significant interest expense. Some risks to this include long-term contracts for raw materials, long-term leases, and labor contracts that make it difficult to reduce shifts to meet a change in demand. Business Specific Risks We also take a close look at any risks specific to the business the company operates in. Some examples of these types of risk include litigation risk for tobacco companies, or the risk of huge claims payouts for bond insurers that decided to insure CDOs. When analyzing a bank, we would need to look at other factors, including the stability of liabilities used to fund assets, net interest margin, the ratio of total assets to tangible

Page 108 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

equity, liquidity profile (maturing liabilities compared to maturing assets), and much more that I am not including here because banks cannot meet our minimum return on adjusted assets metric. ***** Again, we quantify financial strength using a scale from 1.0 to 4.0, with 4.0 being equivalent to an AAA credit rating, 1.0 being equivalent to a BAA3 credit rating, and ratings in between being determined by linear interpolation. Note that if our own analysis indicates a different credit rating than that of the rating agencies, we use the lower of the two ratings. Return on adjusted assets and operating margin are also linked to financial strength, but are considered separately in our analysis of competitive advantage. Some minimum thresholds on quantifiable financial strength metrics are given in the following table, where earnings refers to the companys sustainable earnings.
Ratio Earnings to Debt Retained earnings to Ave. maturing debt 1:1 Fixed Charge Coverage Ratio 4:1 Pre-tax income to pension benefits paid 5:1

Min. Allowed

1:4

One thing to keep in mind is that these ratios compare a companys demonstrated earnings power to cash outflows interest payments, debt maturities, and pension costs) that will partly occur far into the future. That is why a companys future prospects is tightly linked to its financial strength.

5.5 Company Quality & Purchase Thresholds


Margin of Safety Our required margin of safety (the ratio of intrinsic to market value) depends on the level of confidence we have in the companys future prospects. At one extreme, we might be willing to purchase a company when its market value is equal to its estimated intrinsic value if we are extremely confident that the future cashflows predicted by our valuation model will be realized far into the future. For other companies that are still exceptional, but for which we have less confidence in the companys future earnings, we require larger ratios of intrinsic to market value. Our confidence in a companys future earnings also affects our maximum position as a percentage of the portfolios total value; the limit is 20%80 for a company for which we have the highest confidence in future earnings and moves down from there. However, we only approach these limits when the companys
80

Limiting a portfolios concentration in a single company is discussed in Chapter 8

Page 109 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

shares are trading at a ratio of intrinsic to market value well above our minimum purchase threshold. We quantify cashflow predictability by averaging the rankings (1.04.0) of the three purchase criteria (sustainability of competitive advantage, historical performance, and financial strength). An average of 4.0 indicates the highest confidence; at the other extreme, an average of one indicates a company that just meets our requirements. If any one of the three criteria historical performance, future prospects, and financial strength - do not meet our minimum requirements we do not purchase the shares, regardless of the ratio of estimated intrinsic to market value or how well the companys outlook might seem from the standpoint of the other three criteria. Examples of the company analysis process can be found in Chapter 6. We may decide to sell part of a stock position if the companys ratio of intrinsic value to market value is two-thirds of our purchase requirement. For example, if we own shares in a company where we would consider a ratio of intrinsic to market value of 1.25 sufficient margin of safety and they are trading at a 0.83 (or lower) ratio of intrinsic to market value, we might consider selling some shares, but only if there are more attractive investment opportunities available. We would then use the proceeds to invest in the shares of other companies of equivalent or better quality (as quantified by the combined rating). Note that the large difference between purchase and sale valuation thresholds should give us a low turnover; if the turnover is higher, then the excess turnover should be due to the market price of our securities significantly outpacing their intrinsic value, and therefore the turnover should result in large gains in the portfolios intrinsic value. Also note that the sell threshold depends on our current evaluation of the sustainability of a companys future earnings power, and if our perception of this changes, the purchase threshold we use to derive the sell threshold may differ from the purchase threshold used when we originally purchased the company. Finally, note that just because shares are trading at a ratio of intrinsic to market value that is 2/3 of our purchase threshold does not mean we must sell them, it just means we can. To illustrate, if we owned shares in a company that we considered to have extremely high cashflow predictability (say a combined rating of over 3.5), we might decide to continue to hold the shares even at lower ratios of estimated intrinsic to market value. The following table illustrates how our required margin of safety and maximum allocation changes as a function of the average of the three rankings: operating history, future prospects, and financial strength. As usual, we use linear interpolation to obtain thresholds for ratings not shown in the table.
Average of 3 Rankings Purchase Threshold (IV/MV) Sell Threshold (IV/MV) Maximum Allocation to Company 1 > 1.50 < 1.00 5% 2 > 1.25 < 0.83 10% 3 > 1.00 < 0.67 15%

Cyclical Industries

Page 110 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

We typically demand a higher margin of safety for companies in cyclic industries. This is partly because it can be very hard to tell where you are in an industry cycle; this can throw off our valuation model for the case where our historical data is heavily weighted towards a period of economic growth. The other reason is that we prefer stable earnings to volatile earnings; if a recession is deep enough earnings volatility can cause a temporary drop in the dividend. Financial Companies Financial companies deserve special mention here. Although a well-managed financial company can be a good investment, the high leverage employed makes them very sensitive to management mistakes and economic shocks. To illustrate the potential effects of bad management decisions, Wachovias near failure and eventual takeover by Wells Fargo (with a very large loss of shareholder value) was a direct consequence of its decision to enter the pay-option mortgage market. Now contrast this with poor decisions at a company with less leverage although in the 70s both Coca-Cola and Gillette lost their focus, this did not destroy the companies. And even if we own shares of a wellmanaged bank, new management could come along and destroy our investment. Although this is a risk for any company, the high leverage of financial companies acts as an amplifier for management incompetence. For this reason, we avoid financial companies, although we do invest in companies with a financial subsidiary as long as those financial operations are unlikely to put the entire company at risk. In these companies, the financial subsidiary can be useful in providing credit for the companys products, boosting demand when the overall credit markets are tight. Management One major difference between owning your own business and purchasing shares in a business is that you run a risk that the interests of the companys management might not be perfectly aligned with yours. This is usually referred to as agency risk. Obviously, your interest as a shareholder is the maximization of your long-term rate of return, so why would the companys management have a different agenda? One contributor to agency risk is how a companys management is compensated. For example, if the CEO just gets a salary, and does not own any shares in the company, then he might just decide it would be nice to grow the company as fast as possible - perhaps this boosts his ego. Unfortunately, this typically results in the destruction of shareholder value, as in the case of the Coca-Cola Company in the 1970s when the company attempted to diversify into areas unrelated to the companys core competence. One approach used to align managements interest with that of shareholders is the granting of stock options that only create value for management when the shares exceed a certain price. Although this improves the situation, it still leaves plenty of room for abuse, with many CEOs preferring to defer investment if it might cause earnings per share to come in under Wall Streets expectations. Moreover, in a bull market, even a poorly run company can sometimes see an increase in its stock price. The practice of issuing stock options has also contributed to the manipulation of reported earnings by a companys management. Page 111 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Since a companys management has a substantial influence on the degree to which an investment in an excellent business translates into an excellent rate of return on that investment, it makes sense to evaluate managements track record as part of our company evaluation process. We already do this quantitatively by our rating of a companys operating history. If management is not focused on the core business and misallocates capital by entering unrelated industries or chooses a poor allocation between reinvestment in the business and returning cash to shareholders through share buy-backs and dividends, it is likely that the companys historical operating margin, return on adjusted assets, and return on retained earnings will suffer. Qualitatively, we can also get an idea of how well management is focused on the companys core business through letters to shareholders in the annual reports, quarterly conference calls, and management presentations. Of course we do run the risk that fools will one day replace competent management, but we can mitigate this by looking for a company where the companys culture has a strong tradition of focusing on the companys core business and the maximizing shareholders long-term returns. When this is the case, it is more likely that a replacement CEO will have a similar focus to the outgoing CEO. Finally, a company with a strong sustainable competitive advantage and a high level of financial strength can better withstand the effects of incompetent management, so by focusing on companies with these attributes, we stack the odds in our favor.

Chapter Summary
When a company possesses a strong, sustainable competitive advantage, the probability increases that a companys demonstrated earnings power will persist into the future. We evaluate this competitive advantage by first quantifying the companys demonstrated competitive advantage and then analyzing the companys future prospects through the framework of industry analysis. We also made the case that a high level of financial strength is a necessary condition for cashflow predictability. Finally, we combined our quantification of a companys demonstrated competitive advantage, future prospects, and financial strength into a single company quality ranking number, which we use to determine our required margin of safety. At this point, we have both of our foundations for our investment strategy, the ability to determine what a company is worth assuming average company performance and the ability to recognize exceptional companies with sustainable competitive advantages.

Page 112 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

6 Analysis Examples: Three Exceptional Companies (2009)


In this chapter we will illustrate the company analysis process with three examples where we use the methods of Chapter 5 to quantify the companys quality, and determine an acceptable margin of safety for the companys shares. These examples come from my own research undertaken prior to purchasing the companys shares, but are condensed.

6.1 Automatic Data Processing (ADP)


Business Summary ADP provides payroll, tax, benefits, time & labor management, and HR services to businesses. The company has three business units, Employer Services, Professional Employer Organization services (PEO), and Dealer Services. Customers can choose between individual services or bundles, except for PEO, which is a full package. ADPs largest business unit is employer services, which provides a full range of services for small to large sized companies. These services include payroll management (including the payment of payroll taxes and workers compensation), time and attendance management, retirement plan administration, and human resource outsourcing. Human resource outsourcing includes management of salaries and pay grades, benefit plan tracking, government reporting and compliance, screening, background checks, incentives, and new hire integration. ADPs PEO services unit provides small to medium sized businesses many of the same services offered by the Employer Services Unit, but actually becomes the employer of record for the customers employees. The benefit for the customer is substantially improved bargaining power with respect to health insurance, workers compensation, liability insurance, and other areas; the arrangement also allows the customer to focus on their core business. The Dealer Services unit offers a dealer management services software platform that handles receivables/payables management, asset management, cash management, payroll compensation, financial statements, operations control, parts invoicing, inventory and ordering, and appointments / scheduling. Another Dealer Services offering is their Digital Marketing System platform that includes on-line advertising solutions, the ability for customers to view inventory on-line, on-line vehicle and parts sales, and a web-based appointment scheduler. Dealer Services also offers network solutions including office

Page 113 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

phone systems, networked video, and solutions for the management and security of data. They also provide outsourcing of employee training.
Business Unit PEO 12% 6% Geography International 20% N/A

Revenue Profit

Employer Services 72% 82%

Dealer Services 16% 12%

Domestic 80% N/A

ADPs average customer contract length is ten years, so revenue is recurring in nature.; this has historically resulted in stable earnings throughout the business cycle. Revenue per client is variable over the contract length, as it depends on the number of people employed by the client. Total revenues depend on contract renewals, new contract signings, new services sold to existing clients, and changes in the number of people employed by each client. ADPs client base is very diversified, with no single client accounting for more than 2% of annual revenues. ADP has 560,000 clients that in aggregate employ 50,000,000 employees81. One out of six U.S. workers are paid through ADP, and companies using ADPs services employs 25% of U.S. workers. According to their March 2009 presentation, ADPs Employer Services unit serves companies with the following size (by employee) distribution. Table 6-1
Number of Employees Amount of ADPs FY 2009 Revenue Number of ADP Clients Employees Paid ADP Revenue / Client ADP Revenue / Employee 1-49 $1000 M 414,000 4M $2415 $250 50-999 $2200 M 63,000 10 M $34,920 $220 1000+ $2100 M 5000 9M $420,000 $233 Intl: 1-1000+ $1400 M 75,000 9M $18,666 $155

The bulk of ADPs assets consist of client funds, some examples being funds taken from clients to be remitted to employees in the form of a paycheck, and employee taxes taken from clients to be forwarded to the appropriate government tax agency. ADP uses a laddering strategy to smooth out variations in the interest rate earned on client funds, and uses short-term commercial paper to smooth out the volatility in fund inflows / outflows. Client funds provide a significant source of operating income, averaging almost 1/3 of operating income over the last six years.

81

2007 AR.

Page 114 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Operating History (3.3)


Six Year Operating History (FY2003 to FY 2008) ROAA OPM RORE RORE+D GM InvTurn E/FCF E/E+D-C 22.6% 21.1% 16.8% 10.2% 49.0% N/A 0.75 0.97 2.8 3.3 3.7 <- Rating ROAA: Return on Adjusted Assets; OPM: Operating Margin; RORE: Return on Retained Earnings; RORE+D: Return on Retained Earnings & Change in Debt; GM: Gross Margin; InvTurn: Inventory Turnover; E/FCF: Earnings to Free Cash Flow; E/E+D-C: Earnings to Earnings+Depreciation-Capex

These statistics were generated from six years of operating history that were re-stated to reflect the divestiture of the companys brokerage business. Although I usually use tenyears of data, using the re-stated data gives a more accurate picture of the companys current business. However, the critical ratios of OPM and ROAA were sustained since 1990, the earliest data we can obtain SEC data. ADP consistently generates both free cash flow and earnings less depreciation plus capital expenditures in excess of adjusted earnings. In the calculation of return on adjusted assets, assets were adjusted by subtracting goodwill and cash temporarily held for clients, and operating income was reduced by the net interest income on these assets. The rationale for subtracting funds held for clients is that this does not represent capital required to run the business, but rather is capital that results from running the business. Future Prospects (2.5) ADP competes through differentiation, with major factors being service responsiveness, product quality, and the range of services that can be provided, as well as the integration of these services on a single platform. ADP provides value to customers by providing business processes at a lower cost than the customers would incur by implementing the process themselves. Customer Bargaining Power Employer Services customers pay on average around $225 per employee paid through ADP (see Table 6-1). This is a fairly small fraction of a typical employees salary, which means that ADPs services are a small fraction of buyer expenditures; this reduces buyer bargaining power. Moreover, each of ADPs customers represents a very small fraction of ADPs overall sales, which also aids in bargaining power. The low price sensitivity of ADPs customers and ADPs bargaining power are evident in the companys high operating margins. Supplier Bargaining Power

Page 115 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Being a pure service company, the most significant input cost is employment costs. To date, the bargaining power of labor has not limited ADPs profitability. There is no mention of organized labor in ADPs form 10-K, and retirement benefits extended to employees are unlikely to create future problems the total pension obligation is only 50% of earnings, and benefits paid average less than 2% of earnings. ADP does have some equipment and facilities leases and also licenses some software from third parties; both the lessors and license owners can be considered suppliers. When these contracts come up for renewal, ADPs profitability could be impacted. In aggregate, the leases and licensing amount to $903M, or close to 10% of revenue. Barriers to Entry The infrastructure and sales force required to gain entry into the industry serving small companies is relatively small. On the other hand, entering this industry with a focus on larger customers is more difficult; since it can take several months for a new provider to get up and running, switching costs are high, and a companys reputation becomes more important (nobody wants to go through the cost and inconvenience of switching to a new provider just to have the resulting service be bad enough that you need to switch again). ADP does have several advantages against local competitors in the small business market. First, as a PEO, ADP has much more bargaining power with insurers than a small competitor. Second, ADP offers services to small companies beyond traditional payroll processing, such as workers compensation, group health, retirement services, and time & labor management. Finally, as these companies grow, they can obtain additional services on the same platform. For all company sizes, the fact that ADPs different products work on the same platform, can share information using a common company database, and have a single point of service contact gives the product an advantage over a competitors products that focuses on a single business process. Economies of scale present a barrier to entry. With ADP providing payroll, unemployment tax, HR services, and more at an average cost of $225 per employee, a new entrant with limited scale would be hard pressed to match ADPs price and still make a profit. Most of ADPs future revenues are protected by long-term contracts, with a typical annual rollover of 10%. Although these contracts could be broken, this would incur high switching costs, not only the above-mentioned downtime, but also fees. Because ADPs services are a very small fraction of buyer expenditures, switching costs should discourage buyers from pursuing alternatives as long as ADP continues to provide good service.

Page 116 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Intensity of Competition The industry is fragmented, with the exception of ADP and Paychex (PAYX), ADPs largest competitor, which has about 1/4 the revenue of ADP. Paychex has approximately the same amount of clients as ADP, with negligible international revenue. This implies that revenue per client is roughly four times higher at ADP, partly because ADP targets larger customers, and partly because it sells more products to each customer. In 2004, ADP had 70% market share of businesses with 5000 or more employees. Since ADP focuses on larger customers than Paychex, there is not much overlap in their respective markets; this reduces the intensity of their competition. The industry should grow in line with employment, which should roughly track GDP growth.

Threat of Substitutes The primary alternative to ADPs customers is to integrate the services that ADP provides. Unless some shift in technology or cost of labor makes it more cost effective for customers to do this than has been the case in the past, this is unlikely. Due to the low cost of ADPs services compared to total customer revenue there is not a lot of incentive for customers to find substitutes. Demand should be resilient during a recession; although revenue per client will decrease roughly with the increase in unemployment (which is typically just a few percent), this is partially offset by an increase in unemployment tax processed by ADP, and can be further offset by price increases.

Major Risks

As ADPs contracts expire, the largest risk is probably that other business process providers such as IBM and SAP that have both a strong brand and good relationships with many of the larger companies that ADP serves will offer services that compete with ADP. Given the fact that each of ADPs customers constitutes a very small portion of total revenue, and the long-term nature of ADPs contracts, the effect of this threat will likely be a gradual, rather than abrupt, erosion in revenues. And ADPs focus on the business process outsourcing industry as opposed to IBM, where this provides only a fraction of total revenue might enable ADP to maintain its competitive position. Finally, ADP is also more focused on the small to medium enterprise market, so currently the overlap with IBM and SAP is minimal. As a matter of fact, SAP and ADP have an alliance (2003 annual report) to provide comprehensive solutions to customers.

Page 117 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

ADP must constantly innovate to provide the services required by its customers. This risk, while real, is common to pretty much any company, but is somewhat more of an issue in technology intensive companies such as ADP, as compared to a company like Philip Morris International. Technological change might make it possible for customers to dispense with payroll and payroll tax services. For example, substantially all employees being set up for direct deposit might make it possible for a company to automate payroll at low cost, although the payment of payroll tax would still be an issue. On the other hand, growth in other parts of ADPs business could partially compensate for the lost revenue. To date ADP has remained ahead of the curve; one example being a web-based service that allows customers to print paychecks on site, here ADP still gains additional revenue through services like payroll tax filing. Another risk is not related to competitive position, but would significantly decrease ADPs earnings. Currently 1/3 of ADPs pre-tax operating income is attributable to interest on funds held for clients. If some new development made it more cost effective for clients to directly deposit their cash with the relevant tax authorities, ADPs earnings would fall significantly. A mitigating factor is that while the interest gained on total cash in transit is high for ADP, it would be insignificant for any one of ADPs clients, so they have no incentive to seek out a solution where they pay the funds directly to employees and government agencies. Future Growth

There is quite a bit of upside for ADP to grow both domestically and internationally. The growth will come from both market share gains and providing new services to existing clients. Currently the PEO market is fragmented, giving additional opportunity. As an example in the small business market of the new revenue potential for expanding services to existing clients, ADPs administrative services offering typically provides 45X the revenue per client as traditional payroll, with PEO providing 10-12X the revenue per client. The mid-size company market is similar, with the addition of HR services providing twice the revenue as compared to the typical mid-size business services package, and time and labor management add-ons increasing this to 3X. Additional services for large customers have the potential to increase per-client revenue up to 7X82. The following table shows ADPs market share of employees paid:
Small Companies Mid-Size Companies Large Companies
82

Other Payroll Methods 77% 56% 75%

Competitors 16% 13% 4%

ADP 7% 31% 21%

The information source for this paragraph is ADPs March 2009 presentation

Page 118 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

International

96%

N/A

4%

Overall, I rate ADP a 2.5 out of 4.0 for its future prospects, with the significant upside balancing the identified risks. Financial Strength (4.0)
Credit Rating Ave. Maturing Debt to RE Debt to Earnings Fixed Charge Coverage Pension Benefits Paid / PBT

AAA

0.01

23

16.5

0.015

The debt used in the average maturing debt to retained earnings and debt to earnings ratios is long-term debt only. ADPs also has on average $1.5B in commercial paper outstanding during the year; this is used to smooth out fluctuations in client funds. Because this commercial paper is covered by a bank credit facility, and is also covered two times over by liquid investments in ADPs corporate investment portfolio (this portfolio is separate from client funds and is available for general corporate purposes), the commercial paper does not contribute to liquidity risk. ADPs investment portfolio used to make interest on client funds less sensitive to changes in short-term interest rates is conservatively invested, as reflected in the portfolios small mark to market gain as of March 2009. Summary
Operating History 3.3 Future Prospects 2.5 Financial Strength 4.0 Combined Rating

3.3

6.2 The Coca-Cola Company (KO)


Business Summary Coca-Cola is a global manufacturer, distributor, and marketer of non-alcoholic beverages. Coca-Cola sells syrups and concentrates to authorized bottling and canning operators. Coca-Cola has equity stakes in 20% (by volume) of these operators. In addition, CocaCola sells fountain syrups directly to restaurants. The company also sells bottled and canned products directly to retailers. The companys products include sparkling beverages, juices, and bottled water. Roughly 75% of the companys revenue is international, and 53% comes from concentrate and syrup sales of Coke. Coca-Cola has a 10% market share of the non-alcoholic beverage market. Coca-Cola has 14 brands that each have sales over $1B, and owns four out of five of the worlds top sparkling non-

Page 119 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

alcoholic beverage brands. Globally, Coca-Cola is number one in sales of sparkling beverages, juice, and ready to drink coffee and tea, number two in sports drinks, and number three in bottled water.
Revenue by Geography
Africa 4% Bottling Investments 25% Eurasia 4% European Union 16% Latin America 11% North America 26%

Southeast Asia & Pacific 14%

Revenue by Product

Juice 4% Fountain Sales 4%

Domestic Bottlers 14%

Domestic Fountain Sales, includes restaraunts & fountain wholesalers 8%

International Concentrate Sales (nonjuice) 67%

Finished Products, primarily juice and water 3%

Page 120 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Sales of Case Volume To:


Fountain Operations and other finished beverages 11% Bottlers where CocaCola has no ownership interest 25%

Bottlers where CocaCola has a controlling interest 10%

Bottlers where CocaCola has an equity interest 54%

Operating History (3.1)


Ten Year Operating History ROAA OPM RORE RORE+D GM Capex/E+D E/FCF E/E+D-C 27.6% 28.6% 11.3% 8.6% 64.5% 18% 1.03 1.02 3.5 3.5 2.3 <- Rating ROAA: Return on Adjusted Assets; OPM: Operating Margin; RORE: Return on Retained Earnings; RORE+D: Return on Retained Earnings & Change in Debt; GM: Gross Margin; InvTurn: Inventory Turnover; E/FCF: Earnings to Free Cash Flow; E/E+D-C: Earnings to Earnings+Depreciation-Capex

Monotonic revenue growth and earnings have covered dividends since 1982. Coca-Cola has had uninterrupted dividend growth since 1972. Over the last twelve years, average case volume growth has been 5%. Earnings have consistently exceeded free cash flow, largely due to capital expenditures exceeding depreciation. This is reasonable considering the companys expansion over the period. Earnings grew from 1972 to 1997 at an annualized rate of 13.5%; this period covered multiple recessions, two of them severe83. Over the same period, the SP500 had an annualized EPS growth rate of 8%. Future Prospects (3) Coca-Cola competes through differentiation. The company has many strong brands that consumers prefer and are willing to pay a premium for. This preference is enhanced through advertising; Coca-Cola spends $2.8B a year (10% of revenue) on advertising. Moreover, the companys products tend to be a very small portion of consumers
83

From Jeremy Siegels Stocks for the Long Run, chapter on nifty fifty.

Page 121 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

budgets; this makes it more likely that consumers that enjoy the companys products will absorb price increases without looking for a substitute. The most sustainable source of Coca-Colas competitive advantage is the scale of its advertising and distribution system, which can be used to market new non-alcoholic beverages. So although consumer tastes can change (albeit slowly), Coca-Colas competitive advantage will endure. Customer Bargaining Power (2) The bargaining power of Coca-Colas customers varies widely. Since they typically only carry a single type of Cola Beverage (usually either Pepsi or Coke), very large restaurant chains have the highest amount of bargaining power. Large grocery chains that typically have multiple brands of the same product type, but still buy in large volume, have considerable bargaining power but less than the large restaurant chains. Although authorized bottlers purchase concentrate in large volumes, this is where Coca-Cola makes its largest profit margins. There are several reasons for this; one being that for many of the bottlers, Coca-Cola is their largest (or sometimes only) customer. There is also the risk of forward integration if the bottlers try to capture too much of the profit. Historically, overall bargaining power has been strong, allowing Coca-Cola to realize a 28% ten-year average operating margin. Supplier Bargaining Power (3) The raw materials for the companys beverages are generally readily available from many sources, giving Coca-Cola good bargaining power with suppliers. The exceptions are several artificial sweeteners that are sourced from only a few companies. The fact that Coca-Colas gross margins are high (65%) and very stable indicates a low risk of supplier bargaining power impacting profitability. Although some employees are unionized, the companys high ($81,000) earnings per employee and gross margin (64%) indicate a relatively low sensitivity to changing labor costs. To date, collective bargaining has not led to significant pension liability. Barriers to Entry (3) Coca-Cola in combination with its bottling partners has an extraordinary global distribution network that through economies of scale creates a formidable barrier to competition. Coca-Cola also has a large economy of scale in advertising expense, and the companys strong brands provide a barrier to competition. The industry is very concentrated, with Coca-Cola and PepsiCo being the largest companies. The barriers to entry are lower for generic products sold in grocery stores under private labels. Here the store will be willing to allocate some shelf space to its private label products, and it is possible that if these are sold at a low profit margin and are well-received by consumers, then Coca-Colas bottlers may need to price its products cheaper, with the eventual result that Coca-Cola will need to drop its concentrate prices so that the bottlers can stay in business. On the other hand, even if a consumer develops a taste for the private label Page 122 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

product, it will likely only be available at that particular grocery chain, and the consumer will stick with Coke elsewhere. Since these private label brands are not available at restaurants, and the brands are not reinforced through advertising, they will most likely gain share slowly due to the low level of exposure of their product to consumers as compared to Coke and Pepsi. Store-brand awareness rose from 86% in 1991 to 91% in 1995, with the percentage of consumers who regularly purchase private label brands rising from 77% to 83%. 90% of consumers who purchased private label brands thought them equal to or better than name brands, and planned to purchase private label in the future. Retailers have an incentive to stock private label products, as the gross margin is sometimes over 35%, as compared to 25% on name brands. Retailers have started stocking more private label brands, often at the expense of 2nd tier or lower name brands, with 3rd tier or lower hit particularly hard. So to some extent, the increase in private label competition is reducing competition from other name brands. During this time, and more recently, Coca-Cola still managed to increase volume while at the same time increasing price per unit, indicating that as yet, private label product is not significantly impacting Coca-Colas sales (and since case volume has kept pace with concentrate sales, demand is still strong for the bottled and canned product). Looking in my local grocery stores, products by Coca-Cola and Pepsi dominate the shelf space, with minimal space allocated to private label drinks. Even if private label products do gain market share, because each store has an effective monopoly on its private label product, pricing should be stable, which would not be the case if several private label manufacturers offering the same product were competing in the same store.. Intensity of Competition (3) The economic growth of developing countries will allow the companys penetration of many emerging markets (as measured by beverages consumed per capita per year) to approach that of developed countries. Volume growth will likely exceed global GDP growth over the next twenty years. This should keep rivalry controlled, as companies will be able to grow volume without gaining market share. The strong brands of the industry incumbents should also reduce rivalry. Threat of Substitutes (2) The largest risk of substitution stems from Coca-Colas most profitable product, Coke, and its other sweetened carbonated beverages. The risk is that as people become more health conscience, they will drink less of a product that contains empty calories, or calories without much health benefit, and replace this consumption with juice, bottled water, and sports drinks, which currently account for a small percentage of Coca-Colas revenues. However, Coca-Cola is addressing this risk (which today is primarily in developed countries) by rolling out Coca-Cola Zero, which has (according to reviews) the taste of Coke without any calories. Coca-Colas expansion into juices, sports drinks, and bottled water should also dampen the effect of an increasingly health conscience Page 123 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

customer base. To date, sales of Coke are still holding up well in North America and Europe, and over the next twenty years most of Coca-Colas growth will be in emerging markets, where Coca-Cola classic should continue to sell well. The economics of the most likely substitutes (juice, water, and energy drinks) are similar to that of sweetened carbonated beverages. Moreover, the customers and distribution channels are the same, as are the margins. This means that Coca-Cola has a good chance of following any substitution to these types of beverages and maintain its high return on capital. Demand has historically remained fairly stable through the economic cycle, even during severe recessions.

Risks to Competitive Position The largest risk is from changing consumer preferences, as discussed under Threat of Substitutes. This risk is most prevalent in developed countries such as North America and the European Union, which collectively account for 39% of revenues. To date in 2008 (Q3), North American case volume and revenue have fallen 2%, whereas in Europe case volume has risen 3%, and revenue 10% (due to the fall in the dollar over the period). Total company case volume in 2008 has increased 4%, and if international growth continues to make up for the declines in North America, this substitution to healthier beverages should not impact profitability. Since the increased attention consumers are paying to health has been going on for several years, and case volume has still been growing, this risk is probably manageable. The risk of competition from private label products is minor, as they do not have the scale to even come close to Coca-Colas level of marketing. Private label products have been around for years, and yet Coca-Colas volume growth is still strong. Overall, I believe that Coca-Colas competitive advantage will be sustainable well into the future, and that the industry dynamics will remain stable. Opportunities for Growth (3) Globally, per capital consumption of Coca-Colas beverages has grown from 32 drinks in 1985 to 55 drinks in 1995, and on to 77 drinks per person in 2005. As the following chart shows, much of this increase in the last ten years has been driven by developing countries.

Page 124 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

5 & 10 year Volume CAGR by region (2005)


9% 8% 7% 6% CAGR 5% 4% 3% 2% 1% 0%

Africa

10 Year CAGR

Eurasia

European Union

Latin America

5 Year CAGR

North America

region

As shown in the following chart, there is considerable room for growth in per capital consumption. Moreover, this consumption is more a function of marketing and consumer taste than wealth; as Mexico has the highest per capital consumption in the world, well in excess of Europes. Since the vast majority of the worlds population resides in developing countries, where per capital consumption is low, in the long-term, there is obviously considerable room for global growth in the beverage industry. Considering the wide range of beverages Coca-Cola can bring to market in a country, and the companys success with marketing, it seems reasonable that Coca-Cola will capture a large part of this growth. Another source for growth is purchasing competitors with local brands that have the potential for a wider market. When these new brands are integrated into CocaColas distribution network and marketed globally, the return on these acquisitions can be impressive, even when Coca-Cola purchases the company for a substantial premium. One recent example is Coca-Colas acquistion of Glaceau.

Page 125 of 156

Southeast Asia & Pacific

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Per Capita Consumption by Region


450 400

Per Capita Consumption

350 300 250 200 150 100 50 0

Africa

Eurasia

European Union

Latin America

North America

Southeast Asia & Pacific

Region

Financial Strength (3)


Credit Rating Ave. Maturing Debt to RE Debt to Earnings Fixed Charge Cov. Pension Benefits Paid / PBT

AA3

0.67

1.23

17.7

3%

Coca-Colas debt should not be a problem, as maturing debt is easily payable out of retained earnings, and the company has a high fixed charge coverage ratio. Pension exposure is small in relation to earnings. Earnings per employee are high at $80,000, indicating a low sensitivity to labor costs. Gross margin (65%) is also high, indicating a low sensitivity to changes in the cost of inputs.

Summary
Future Prospects 3 Financial Strength 3 Operating History 3.1 Combined Rating 3.0

Coca-Colas historical performance indicates that is possesses a strong competitive advantage, and the review of the companys future prospects indicates a strong likelihood that this advantage will be sustainable. And Coca-Colas strong financial strength means that the company is well positioned to withstand economic shocks.

Page 126 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

The company estimates that there is a $9B gap between the fair value of its investments in bottlers and the amount carried on the balance sheet. With a current market cap of around $105B, this is a significant source of excess value.

6.3 Johnson & Johnson (JNJ)

Business Summary Johnson & Johnson has three business units, all of them focused on healthcare. The pharmaceutical business develops new chemical and biological compounds used in the prevention and treatment of various diseases. The medical devices and diagnostics business develops products used to diagnose and treat various health problems. The consumer health care segment owns many valuable brands such as the J&J baby care product line, Neutrogena, Listerine, and Tylenol.
Revenue by Region

Asia-Pacific & Africa 14% Other Western Hemisphere 8%

United States 52%

Europe 26%

Page 127 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Revenue by Segment

Medical Devices 36%

Consumer 24%

Pharmaceutic al 40%

Operating History (3.5)


Ten Year Operating History ROAA OPM RORE RORE+D InvTurn GM E/FCF E/E+D-C 32.0% 25.5% 15.7 10.3% 3.1 70.2% 0.96 1.04 3.5 3.5 3.5 <- Rating ROAA: Return on Adjusted Assets; OPM: Operating Margin; RORE: Return on Retained Earnings; RORE+D: Return on Retained Earnings & Change in Debt; GM: Gross Margin; InvTurn: Inventory Turnover; E/FCF: Earnings to Free Cash Flow; E/E+D-C: Earnings to Earnings+Depreciation-Capex

Johnson & Johnsons high return on retained earnings plus depreciation shows that it can profitably grow its business through the re-investment of capital. Excluding restructuring charges, earnings and revenue have basically risen since 1985 (earliest data), and dividends have been increased since 1970 (earliest data). Earnings grew from 1972 to 1997 at an annualized rate of 14.2%; this period covered multiple recessions, two of them severe84. Over the same period, the SP500 had an annualized EPS growth rate of 8%. Future Prospects (3.5) The companys competitive advantages include strong brands protected by patents in the pharmaceutical and medical devices business units, and strong consumer brands in the
84

From Jeremy Siegels Stocks for the Long Run, chapter on nifty fifty.

Page 128 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

consumer business unit. The company invests a good portion of its revenue (12.5%) back into research and development, which should provide a continued supply of health care innovations to replace those losing patent protection. The companys diversification is also an advantage, in that when say the pharmaceutical pipeline starts to run dry, the company can continue to invest heavily in research and development using the revenue from the other business lines. Yet another advantage is the companys management structure. The company is a collection of over 250 operating companies that are each extremely focused on a particular market segment, with the parent company providing financial strength, strategic direction, allocation of capital, access to distribution channels, and economies of scale with marketing. Financial strength is also high, and the company possesses a rare AAA long-term issuer credit rating. This gives the company a low cost of debt capital to help fund acquisitions. J&J has a well-diversified product line, with the largest revenue from a single brand (the drug Risperdal) being only 7% of the companys total revenue. Customer Bargaining Power Currently, no single customer exceeds 10% of revenue; the largest customers are probably government agencies, such as Medicare and Medicaid in the United States. In Europe the customers for pharmaceuticals and medical devices are primarily governments running national health programs. This concentration gives the governments good bargaining power, but on the other hand, the products they are purchasing are unique (and often protected by patents), which gives the company strong bargaining power as well. In the United States the customers are more fragmented. In 2006 prescription drugs accounted for 10% of U.S. health care costs; Europe is probably similar. Although this does make pharmaceuticals a significant expense, which tends to make customers bargain harder, it is by no means the largest expense, and the fact that new drugs are initially protected by patents should keep pricing power intact. The largest risk to J&Js competitive position is a potential increase in buyer bargaining power in the United States if the U.S. health care system became nationalized in the same sense as Europes. Perhaps a more likely scenario would be that Medicare and Medicaid take more advantage of their large volume of pharmaceutical purchases to increase their bargaining power. We can get an idea of how this will affect profitability by looking at sales in Europe and other countries with nationalized health care systems. A report complied in 200385 compiles a price index of pharmaceutical prices in eight different countries as compared to the price in the United States. It appears that in many countries where the price of single source patent protected drugs is lower due to price regulation, government bargaining power, or both, that generic drug use is proportionally lower; this allows a company to attain higher margin on branded off-patent drugs as compared to the United States. Combining on and off patent drugs, 7 of the 8 countries analyzed has prices roughly 75% of that in the United States, whereas Japan had prices 20% higher than in the United States. Since this it is possible that these countries might increase their
85

Prices and Availability of Pharmaceuticals: Evidence from Nine Countries, Danzon

Page 129 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

use of generics, we should also look at price differentials of on-patent drugs only. Here we find that the price differential in seven out of the eight countries averaged around 65% of the price in the United States, with Japan being around 40% more expensive. This analysis may be pessimistic, in that lower European prices may be partially attributable to lower incomes in Europe as compared to the United States. Pharmaceutical companies came to the conclusion some time ago that it makes sense to use tiered pricing based off of what a particular market can afford. As a matter of fact, when we adjust for per-capita income differences, the price discrepancy between Europe and the United States disappears. To date I have been unable to obtain comparable data on the medical device industry, but it seems likely that it is susceptible to the same risks. For a potential scenario, lets assume that prices in the United States falls to a level similar to that of Europe, while at the same time governments globally take more advantage of generic drugs. Consequently, realized prices for on-patent drugs and medical devices falls to 65% of current levels. Currently, 44% of J&Js pharmaceutical and medical device revenue is generated internationally, so total pharmaceutical and medical device revenue would fall to 80% of current levels86. Since this revenue accounts for 84% of J&Js operating profit, we could expect earnings to fall to 83%87 of current levels. Although not good, this scenario would not be catastrophic, and we could still expect long-term growth in excess of global GDP from this new equilibrium. Supplier Bargaining Power Raw materials are available from multiple sources. The bulk of the value of the companys products is created in the manufacturing process; consequently, profits are well insulated from price swings of raw materials. Total employment costs are 23% of revenue, and employees not unionized. Barriers to Entry A key to success in the pharmaceutical industry is having the resources to fund many different potential drug candidates in parallel. To see why, lets say it costs a drug company $1M to push a potential drug through development, but this only results in a marketable drug 1 out of 100 tries; but when the trial is successful, the product will generate $200M in annual revenue. So a small company that can only afford to research one possible drug at a time has only have a 1/100 chance in each year of creating anything, and will likely go bankrupt soon. But with scale, the law of averages works in your favor. Since each drug test can be considered statistically independent, there is a 99/100 chance that each attempt will end in failure. But if a company can fund 100 research projects in parallel, the odds that all of these will fail are only 0.99100 , or 36%. If a company can fund 1000 projects in parallel, the odds of all failing is nearly zero, and
86 87

Current: 44 + 56, reduce 56 to 65% of current levels -> 44 + 36.4 = 80.4 Current: 16 + 84, reduce 84 to 80% of current levels -> 16 + 67.2 = 83.2

Page 130 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

the odds that 75% of the trials will succeed are 1 0.99 250 88, or 81%. The large scale required to maintain profitability in this industry can be considered a barrier to entry. Medical devices also need to undergo FDA trials, which makes their development somewhat risky, although less so than pharmaceuticals. Many of these devices have proprietary product differences, strong brand recognition, or both. The primary barrier to entry into the consumer health industry are economies of scale in advertising and research combined with access to distribution channels, although private label products wont have a problem getting shelf space. Overall, capital requirements are extremely high for successful entry into any one of J&Js businesses. Any potential new entrant into the pharmaceutical or medical device industry would need a large sales force to communicate the advantages of new products with doctors and hospitals. All three of J&Js segments benefit from cost advantages stemming from economies of scale in both R&D and marketing, and differentiation stemming from a reputation for quality products.

Threat of Substitutes Since pharmaceuticals and medical devices are necessary for the maintenance of good health (not to mention life in extreme cases), a significant decrease in use is not likely. The largest threat of substitutes comes from generic pharmaceuticals. Once a drug loses patent protection, associated revenue usually rapidly falls to a fraction of what it was. As discussed earlier, this effect is more pronounced in the United States. One way to mitigate the effect of generic drugs would be to develop more compounds with lower sales, perhaps by targeting more precise conditions. Research has shown that for drugs with sales less than $50M, generic competition falls off rapidly. A similar threat of substitutes is possible in the medical device industry as well, although currently, the generic medical device industry is in its infancy. One reason for this is that while most drug compounds are easily copied, there is considerable engineering and technical know how required to manufacture medical devices and diagnostic equipment. For this reason, the risk of reduced profitability due to a lull in innovation is greatest in J&Js pharmaceutical subsidiary. Intensity of Rivalry Industry growth is expected to remain strong, product differentiation exists (with patents providing protection from copying), and sunk costs are relatively small compared to profits (high ROA). All these factors serve to reduce the intensity of rivalry.

88

This is just one minus the odds that 25% will fail, 250 is 25% of 1000

Page 131 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Pipeline Risk One of the reasons the barrier to entry is so high in the pharmaceutical industry is that new drug trials are very expensive, and do not often result in a profitable product. This raises the question of whether a company has a sufficiently robust development pipeline to replace existing drugs that go off patent. The average effective patent lifetime for a new drug is 11.5 years. However, most drugs do continue to generate revenue after they go off patent; a study conducted in the mid-1990s showed that revenue fell by 43% the first year after patent expiry, and 42% the next year, and roughly 10% annually thereafter. On the other hand, sometimes a generic manufacturer can successfully challenge a patent in court, so lets leave the effective life at 11.5 years. Consequently, with an evenly distributed patent expiration schedule (often an optimistic assumption), we can expect 8.7% of the companys revenue to be lost each year. Looking at a paper89 that investigated the revenue of new products introduced from 19881992 in the U.S., it appears that the mean revenue generated by new products was around $150M in 1992 dollars. Lets assume that global sales were around twice that, or $300M in 1992 dollars. In 2008 dollars, this would be around $450M per new product. Therefore a company losing 8.7% of its pharmaceutical revenue each year would need to introduce new drugs at a rate equal to total pharmaceutical sales (in millions of dollars) multiplied by 8.7% and divided by $450M. Johnson & Johnson had 2008 pharmaceutical sales of $24,500M, so they would need to introduce roughly 4.7 drugs every year. Looking at their 2008 development pipeline, we find that they have either filed or had approved 18 compounds, which gives them a comfortable margin. The margin is probably necessary, as the actual distribution of new drug revenue is highly volatile, with only 3 out of 10 recouping development costs. Their phase III pipeline has 25 compounds, and we might expect 2/390 of these to become new products over the next couple of years. Since this is a rough analysis, lets do a differential analysis with several other pharmaceutical companies:

Company

Pharmaceutical Revenue $24,500 M $48,300 M $23,000 M $44,900 M $26,300 M

J&J Pfizer Merck Glaxo-Smith-Kline Novartis

New products Needed 4.5 9.3 4.4 8.7 5.1

Filed or Approved (1) 18 (400%) 1 (11%) 2 (45%) 24 (275%) 13 (254%)

Phase III (2)

Largest Product 15% 25% 18% 18% 22%

25 (370%) 25 (180%) 9 (137%) 29 (223%) 28 (367%)

89 90

The Distribution of Sales Revenues from Pharmaceutical Innovation Research and Development in the Pharmaceutical Industry, CBO 2006 page 23, figure 3-2.

Page 132 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Abbot Labs $16,700 M 3.2 5 (156%) N/A (3) 27% (1) number is parenthesis is the ratio of filed or approved to new products needed (2) Number in parenthesis is the ratio of 2/3 of phase III compounds to new products needed (3) Abbott Labs does not publish a development pipeline, but 2008 AR indicated 5 new drugs

It is pretty clear from the differential analysis that J&J has one of the best drug pipelines in the industry. Aside from its pipeline, J&J has the highest financial strength of companies that compete in the pharmaceutical industry. Not only is J&J the only company in the industry with an AAA credit rating, but J&J also has a much broader revenue diversification, with pharmaceutical revenue accounting for roughly 50% of total revenue. Risks to Competitive Position Risks to J&Js competitive position include: Computing power increases to the point where drugs can be completely modeled using simulators, negating the need for costly trials. The FDA would also need to sign off on this. This would significantly reduce the scale required to successfully compete. This is actually fairly likely at some point in the future; the question is when. Most likely, the computing power is at least a decade away, with perhaps another decade of regulatory buy-in. Still, existing patents will lock in cash flows for a few more years after this occurs. Changes to patent law that allows earlier generic competition. The company fails to innovate at the same rate as in the past, and suffers decreased profitability due to competition from generics. This risk is greatest in the companys pharmaceutical subsidiary. To date, the company continues to innovate, with 30% of 2008 sales coming from products developed over the last five years, and 12% of 2008 sales coming from products developed in 2007. Increased buyer bargaining power. This was discussed in detail, and the modeled scenario looked manageable.

Growth Potential With the aging population in the United States and Europe, the health care market should be pretty robust for the foreseeable future, and the total available market for health care should easily grow in excess of global GDP. Older people tend to require more medical care, which should positively affect the demand for both pharmaceuticals and medical devices due to the aging population in the companys core markets, while the overall global increase in population should increase the demand for the companys consumer Page 133 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

products. J&Js three business segments are probably among the most recession-proof industries. Contributing to positive short to medium term prospects, JNJs pharmaceutical pipeline is very strong, with 32 compounds currently in Phase III trials, and 7 filings and 5 approvals during 2006. This is for a $24B pharmaceutical division. Compare this to Pfizers 8 compounds in phase III, with three filings and four approvals, with Pfizer being a $51B pharmaceutical company. JNJ has eight times the potential new drugs adjusted for revenue as compared to Pfizer. And Johnson & Johnson has a similar advantage over Merck and Novartis. Financial Strength (3.5)
Credit Rating Ave. Maturing Debt to RE Debt to Earnings Fixed Charge Cov. Pension Benefits Paid / PBT

AAA

0.07

0.62

25.5

3%

With a debt to earnings ratio of 18%, the company could easily pay off its entire debt in a single year. Maturing debt is covered over 20X by earnings in any given year. The companys total pension obligation is only 1.3 times earnings, and consequently should not become an issue, even though it is currently only 71% funded. Management To date, management has done a good job balancing organic growth with acquisitions. Basically acquisitions are not required for growth, but are pursued opportunistically, with management not being afraid to walk away from the table if the numbers dont look good (as they did with Guidant). The decentralized structure also works well; we have seen that adjusted for revenue, the pharmaceutical business is much more productive than Johnson & Johnsons rivals. Summary
Operating History 3.5 Financial Strength 3.5 Future Prospects 3.5 Combined Rating 3.5

Page 134 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

7 Analysis Examples: Some Lemons (2007)


In this chapter we will take a look at some examples where an analysis of a company indicates that the future cashflows are much too uncertain to consider the companys shares a good investment. The analysis will not be as extensive as in the last chapter, but will instead focus on the issues that make the company a poor investment. After all, to purchase a companys share you need to look at everything, but it only takes one serious problem to reject a company.

7.1 Unproven Business Models


Lets say you are investigating a company that over the last few years has shown very impressive earnings growth, but the bulk of this growth comes from a new business activity that the company has engaged in; furthermore, the new business did not exist prior to a few years ago. To be more specific, lets say a utility company started an energy trading business, and while it is apparently extremely profitable, it also has a relatively short history, both for this or any other company. Even though you might think you understand how energy trading might work, you really cant point to any examples of companies that have successfully engaged in this business - in other words, the business model is relatively unproven. Because the company has not proven that it can generate consistent profits from this business for a period of at least ten years, valuing this company based off of its demonstrated earnings power might be risk. If this example sounds familiar, it should; the company is Enron, which ended up going bankrupt a few years after entering the energy trading business. For a more recent illustration of the dangers of an unproven business model, consider Ambac Financial. Initially this company was in the rather boring business of insuring municipal bonds, but in an attempt to boost earnings growth, in 1998 the company started issuing credit default swaps on collateralized debt obligations (CDOs), with these CDOs containing a toxic mix of mortgage backed securities, other asset backed securities, and junk bonds. I started to investigate Ambac in late 2007 when the companys shares fell from around $60 per share to under $30, with the fall being due to concern over their exposure to CDOs. The first thing I looked at was the companys credit rating implied by the spread on their senior debt and credit default swaps written on that debt; both indicated a financial strength well below investment grade. I then took a look at their $71B in exposure to CDOs, and saw that even a 5% default rate not uncommon for these types of assets - would eat up half the companys equity, and their entire unearned premium reserve. As it turned out, even Ambacs core bond insurance business was shaky, as they did not appear to be taking in enough premiums to even cover the long-term average expected default rate on the bonds they insured. To illustrate, at the end of 2006, Ambac insured

Page 135 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

$519B in bonds net of reinsurance. The weighted average credit rating (using Ambacs internal ratings) was midway between A and AA91. Being optimistic and using the AA rating would imply an annualized default rate of 0.14% using Moodys statistics from 1970-1999 and a 0.3% default rate using Moodys statistics from 1920-1999. Now in 2006, Ambac received $811M in premiums net of reinsurance, which equates to 0.16% of the net amount insured. So using the more optimistic 1970-1999 statistics, the premiums barely covered the expected default risk. Although I am by no means an expert in the insurance industry, it seems common sense to leave some room for default rate volatility. Moreover, the premiums do not even cover the expected default rate using the more conservative 1920-1999 statistics. The fact that the companys realized default rate had been much less than that implied by the credit quality of the insured portfolio was immaterial; the bond market has historically had long periods of either well above or below average default rates, but a companys long-term viability cannot be determined based off of these temporary swings. After this fairly quick and simple analysis indicated that the company faced significant risk in its relatively new CDS business, and was not bringing in sufficient premiums to make even its core bond insurance business viable, it was an easy decision to pass on purchasing the companys shares. Although only time will tell, to date, it seems like a wise decision. Lets end our discussion of the perils of an unproven business model by looking at the case of Countrywide Financial Corporation. Although the loan portfolio was not well diversified, and the company was overly reliant on short-term notes and repurchase agreements to fund assets, the real problem that led to the companys downfall was a new type of mortgage known as a pay-option adjustable rate mortgage (ARM). This allowed the borrower to make payments that actually resulted in negative loan amortization, which reduced the borrowers equity over time. Since these mortgages had a loan to value ratio that was already on the high side, and the probability of default is positively correlated with the loan to value ratio (which was bound to further increase with the negative amortization), it seemed like there was considerable risk associated with this new type of mortgage, particularly if home prices stopped rising. At year-end 2006, these types of mortgages accounted for roughly half of Countrywides mortgage portfolio, with the remainder primarily being home equity lines of credit and hybrid ARMs (loans with a fixed interest rate for several years that then switches to a floating rate), with only a small fraction of fixed rate first mortgages (the type with the lowest expected default rates). As it turned out, this new type of mortgage (and the lack of stable fixed rate mortgages) caused the companys non-performing assets ratio to surge from 0.60% at the end of 2006 to almost 3% at the end of 2007; this, coupled with the flight of non-insured depositors, large charge-offs due to securitization related losses, and a credit rating downgrade to junk status, almost pushed Countrywide into bankruptcy before Bank of America agreed to acquire the company.

91

Although you might consider the usefulness of these ratings when we see that Ambac considered 16% of its insured portfolio to be AAA quality, which begs the question of what benefit was Ambac adding by wrapping those bonds?

Page 136 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

7.2 Poor Financial Strength


At one time, K-mart, a discount retailer, owned a strong retail brand and had consistent earnings power. But by the early 1990s, it was pretty clear that the company was in trouble, despite the fact that it still had an investment grade credit rating. Going back as far as 1989, the companys free cash flow was negative up until the time of its bankruptcy in 2003 (1989 is as far back as I could get form 10-Ks from the SEC website). Although the company had positive earnings per share in most of these years, and positive cash flows from operating activities in all of these years, the companys capital expenditures were consistently higher than depreciation, despite the fact that the companys total number of stores in operation was declining (so the excess capital expenditures were not being used for expansion). Partly this deterioration of financial strength was due to some idiotic decisions on the part of management, such as acquiring companies (Borders, OfficeMax) that were not part of its core business. Still, the company would have likely recovered with better management if it were not for the burden of servicing debt and rental expense. Because the company did not own their own properties, rent expense was of the same magnitude of the interest payments made on debt, and the company had a fixed charge coverage ratio of less than 2 from 1991 on. With this low of a fixed charge coverage ratio, combined with the companys erratic earnings power, it was just a matter of time before the company went bankrupt. If you had owned shares in this company, there was plenty of time to get out before the credit rating agencies even realized there was anything seriously wrong. If on the other hand the company owned (rather than leased) most of its properties, and had little debt, it would probably still be generating cash for investors today. In reality, it actually is, as most of the companys stores kept operating through the bankruptcy and are now part of Sears Holdings; but this is little consolation for the owners of the companys common stock prior to bankruptcy, as they ended up with nothing. This is a good point to note that although credit rating agencies are sometimes slow to realize that a companys financial health is permanently impaired, once they do, the resulting debt rating downgrade provides positive feedback to the companys downward spiral. Often debt has covenants that require the company to pay a higher interest rate when the debts rating falls below a specified threshold. At any rate, once existing debt is rolled over, the companys interest expense will significantly increase due to a higher market interest rate on a new bond issue, this being caused by bond investors requiring a higher rate of return to compensate them for the increased risk of bankruptcy.

7.3 Poor Management Decisions


Besides re-evaluating a companys competitive position, it is also important to evaluate the rationality of the companys management. Poor decisions by management can cause a future impairment of the companys competitive position, often resulting in a dramatic reduction of a companys future earnings power, and sometimes even lead to bankruptcy. Poor management is somewhat qualitative, but I define it with respect to how I would like to see my companies run: in a conservative manner. This means not enhancing Page 137 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

short-term shareholder returns at the expense of the long-term financial strength of the company. As an example, consider Home Depot, which under new management decided to issue debt to repurchase stock. This decision reduced the companys long-term financial strength in return for a small one-time pop to earnings per share from the buyback. Moreover, management defended the decision by looking at the resulting debt to cashflow ratio, but did not consider what might happen if interest rates were to sharply increase. I was not the only one who thought this move ill advised, as Moodys downgraded the companys debt by several notches (AA3 to BAA1). In this case, early recognition of the companys mistake proved useful, as it enabled me to sell at close to $40 per share; soon after that the companys stock plummeted to below $25. Sure the price could recover, but by selling at $40 I could lock in my gains and re-invest the proceeds into companies that had both higher financial strength and better future prospects, and were trading at a higher discount to my estimate of their intrinsic value. To be clear, I did not sell because of the companys diminished short-term prospects caused by the deflation of the housing bubble, but because I did not have confidence that the new management would run the company in a conservative manner, and at the time there were in my opinion better companies trading at a more compelling valuation. If Bob Nardelli had not been ousted from his position as CEO, I would probably still be holding my shares. Another example of management embracing an inordinate amount of risk for a small current return are the banks that securitized loans that were poorly underwritten, and in many cases backed up the conduits which held the securitized loans with a guaranteed line of credit. Once I learned that Citigroup, a company I had held for several years, had accumulated over $400B of these off balance sheet entities that were stuffed with questionable assets (and had a large exposure to similarly dubious assets held on balance sheet as well), it was an easy decision to dump my shares for a small profit, as even a small impairment to these assets could have devastating consequences for the bank. A year later, Citigroup recorded massive charges to earnings, required several dilutive injections of capital from sovereign wealth funds, and cut their dividend. Acquisitions should also be scrutinized. Although time will only tell whether I am right, the purchase of Countrywide Financial by Bank of America seemed a bit foolish. Left alone, Countrywide would have likely failed, and the major banks could have increased their own mortgage origination businesses to take up the slack. But now that Bank of America owns the company, they are saddled with a very dubious loan portfolio (5% NPL last time I checked in 2006) as well as Countrywides liabilities. Sure they get the companys mortgage origination business, but the next few years are unlikely to see a strong demand for new mortgages; consequently, Bank of America would have been better off growing its mortgage origination business organically. Another aspect of Bank of America that I am worried about is their exposure to credit default swaps. As of EOY 2006, Bank of America has $1.5T in credit derivative notional value. It is not clear how much of this is written rather than used as insurance (bought). This is twice the value at EOY 2005. It is worrisome that the company claims that the amount of credit protection outstanding fell from 2005 to 2006, while the notional amount of contracts increased by

Page 138 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

$750B. If this means that the additional $750B was created by written contracts, then BAC is very exposed to a spike in default rates. Lets end with the observation that quite a few REITs have decided that low interest rates are here to stay, and have loaded up on a worrisome amount of debt. The obvious question is what happens when interest rates climb back up to historical norms, or spike even higher? In the near term, management looks good, as the increased leverage provides enhanced earnings growth. But in the long-term, in many cases this will necessitate selling off property at fire sale prices in order to remain solvent.

7.4 Substitution
In the ten-years leading up to 2000, Eastman Kodak had ROAA, OPM, and RORE that were close to meeting our minimum threshold, and could have exceeded them if it was better run. The company had a strong brand, and the industry dynamics were favorable. What really brought the company down was the widespread adoption of digital photography. Even if the company could have become a profitable leader in this new field, the economics were completely changed, and it would have been impossible for the company to maintain its historical earnings power. To see why, we need to look closer at the companys business in 1999. The 1999 annual report states that Eastman Kodak Company (the Company or Kodak) is engaged primarily in developing, manufacturing and marketing consumer, professional, health and other imaging products and services. Turning to note 17 in the notes to the financial statements, we find that each of the companys businesses (Consumer Imaging, Kodak Professional, Health Imaging, and Other Imaging) were heavily dependent on the sale and processing of photographic film, and the sale of photographic paper, cameras, and photofinishing equipment. Although the company did manufacture digital products, which included scanners and digital cameras, they only represented 17% of totals sales, and were not profitable. Shortly after 2000, the substitution of digital photography and health imaging for traditional imaging methods progressed rapidly. Now even if Eastman Kodak had managed to make a profitable transition to the new digital technology, there was no way it could match the profitability of its traditional imaging business. The reason for this is that each traditional camera (or x-ray machine) led to a long stream of sales of film and developing chemicals, which contributed to the bulk of the companys profits. Now compare this to digital photography, where you typically only print out a fraction of the pictures you take, and when you do you probably use a printer. So to capture recurring digital photography revenue, Eastman Kodak would have needed to successfully penetrate the printer industry, and even if it did so, each camera would still lead to a smaller revenue stream. A similar dynamic holds in medical imaging, although here most x-rays are not printed at all, but viewed digitally. In this case study, it was the same forces that drove the substitution to digital photography (a much higher value to price ratio than for traditional photography) that made the new business much less profitable. Page 139 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

7.5 Supplier Bargaining Power


The big 3 U.S. Automakers are a good example of supplier bargaining power reducing a companys profitability. Here I am not referring to the companys suppliers of parts or raw materials, but rather the companys labor force. One of the reasons these companies focused on light trucks and SUVs is that their cost of labor (wages, pensions, and medical care for retirees) made it impossible to make a profit on smaller (and less expensive) models. The cost of these retirement benefits also reduced these companies discretionary cash flow, and forced them to fund more of their capital expenditures using debt. In my opinion, if these companies had had labor costs more in line with Toyotas, the recent recession would have merely resulted in a period of losses, rather than likely bankruptcy.

7.6 Product Concentration


First of all, a company with a single product that provides a large percentage of revenue is not necessarily a risky investment, provided that the product is associated with a strong brand that is protected by a trademark, Coke being one example. However, when the only significant barrier to competition is a patent, the situation can be very different. An example of this can be found in the pharmaceutical industry, where profit margins on a drug with patent protection are high, but when the patent expires, generic competition can severely reduce the drugs profitability. This is because a generic drug manufacturer does not need to conduct any costly and risky medical trials with a low average success rate this risk is borne by the pharmaceutical company - and can consequently afford to sell a copy of the drug at a low profit margin. The customer has no incentive to stick with the branded drug, as the generic drug is identical in all respects. Clearly, a pharmaceutical company with an on-patent drug that accounts for a significant fraction of its earnings faces a sharp decline in earnings power when the drug goes off patent. The risk can be mitigated if the company has a strong development pipeline with a high probability of producing enough new drugs to replace the lost revenue. Pfizer is a company that for the ten-years ending in 2007 had an operating history indicative of a competitive advantage, but a careful analysis of its future prospects told a different story. First of all, the patent for Lipitor, which accounted for 25% of the companys 2007 revenue ($48.3B), will expire in 2010. This means that the companys development pipeline will need to provide enough new compounds to generate $12B in revenue. Unfortunately, Pfizers development pipeline is not nearly as robust as some rivals (notably Glaxo-Smith-Kline, Novartis, and Johnson & Johnson). Pfizer currently has 16 unique phase III compounds in its development pipeline. Using the pipeline analysis technique we used in our analysis of Johnson & Johnson in Chapter 6, we find that Pfizer must introduce 9.3 new products each year in order to replace lost revenue from expiring patents. Unfortunately, only one product was approved in 2008, although there are 16 in phase III development of which 2/3 can be expected to lead to new Page 140 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

products over the next couple of years. In comparison, the pipelines of GSK, Novartis, and J&J are roughly twice as productive as Pfizers when compared to each companys total pharmaceutical revenue. Although it is too early to gauge how well a long-term investment in Pfizer at the end of 2007 would have fared, the early signs are not encouraging. In a desperate attempt to invigorate the scale of its development pipeline, in 2008 Pfizer announced a takeover of Wyeth, another pharmaceutical company with a mediocre development pipeline 8 phase III compounds to support 2008 revenue of 22.8B is similar to Pfizers pipeline, although with 8 compounds submitted to the FDA, it is in better shape. In order to finance this takeover, Pfizer had to cut their dividend by 50%. Of all the companies we examined in this chapter, Pfizer was the only one that met our minimum operating history criteria (with an average score of close to 2.0) measured at the end of 2007. Hopefully this example illustrates the importance of a careful analysis of a companys future prospects, which probably would have ruled out the company due to its weak pipeline; instead shares in a company Glaxo-Smith-Kline, Novartis, or better yet Johnson & Johnson could have been purchased instead.

Page 141 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

8 Portfolio Considerations
This chapter briefly discussed some issues that arise in the management of an investment portfolio.

8.1 Asset Class Allocation


The basic idea of asset allocation is to design a portfolio with an expected rate of return and rate of return volatility that is optimal for your financial goal. Although the math is fun, in my opinion you really dont need to worry about efficient frontiers or anything like that. If you are saving for retirement, your investment portfolio should contain 100% equities92. Beyond that, I believe it is more prudent to focus on the companies you invest in as opposed to following some pre-set allocation between different equity asset classes; purchasing a focused portfolio of exceptional companies at a reasonable price will likely be much more beneficial than the (theoretical) volatility reduction predicted by a given asset allocation. If you are retired, and living off of your portfolios income, then you might want to allocate a portion of your investments to short-term bonds to act as a liquidity buffer against either unanticipated expenses or a severe recession that might result in a drop in dividend income. Depending on your age and assets, you might want to consider an inflation-indexed guaranteed payment annuity for part of your retirement income needs.

8.2 Level of Focus


Imagine you have just found an incredible bargain an exceptional company trading at a ratio of estimated intrinsic value to market price of 2. Should you invest everything in this one company? It depends. If you are just starting out on saving for your retirement, and your annual contributions are a significant fraction of your current portfolio value, I would say go ahead, as the slight risk that you have missed something will have a negligible impact on your savings goal. This is actually a pretty good way to start building a portfolio, by focusing on just a few companies, and adding one or two each year. On the other hand, if you happen to be further along towards retirement, or at the point where you are making withdrawals from your portfolio to provide income for living expenses, I would not recommend that degree of focus. Why? Because no matter how good a company looks, there is always that small chance that management is cooking the books, or a future development renders the companys product obsolete. This latter could
92

I have shown using Monte Carlo simulation that gradually adding a fixed income component to a portfolio reduces the amount saved at the 75% confidence level.

Page 142 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

happen even to a business model as simple as manufacturing razor blades. Imagine what would have happened to Gillettes business model and future earnings power - if somebody invented a razor blade that never went dull. Another way of stating this problem is that there is always the chance that a black swan - a highly improbable event that is completely unexpected but also has a massive impact93 will come along. Obviously it is the black swans that contribute to the fattening of the left-hand tail that we are worried about, i.e., the massive negative impacts. Some possible events that fall into this category might be: The combination of improvements in battery powered cars and sustainable power generation have a large negative impact on demand for oil, and the profits of oil companies. Prohibition is re-enacted, with an obvious negative impact on the profits of domestic breweries. A large bank falls, and financial contagion results in a collapse of the financial sector, with widespread bank failures94. The federal budget deficit increases, and Congress decides to repeal the REIT act to increase revenue from corporate taxes. This would create a large drop in the funds from operations of REITs.

Although none of these are even remotely likely in the near to medium term future, they are all possible, along with many other scenarios. For this reason, if a portfolio is going to be used for current income in the near future (say within ten years), I believe it makes sense to create a set of portfolio diversification guidelines, such as the maximum percentage of the portfolios market value that can be invested in the stock of a single company, in the stock of companies in a single industry, or in the stock of companies in a single foreign country. On the other hand, if retirement is still a long way away, there is nothing wrong with building your portfolio one excellent company at a time. As it turns out, diversification gives us decreasing returns towards risk reduction, in that the largest reduction in risk comes from increasing the number of stocks held in a portfolio from one to two, the next largest from two to three, and so on, with only minor risk reduction after twenty stocks. This is illustrated in Figure 8-1, which plots the volatility reduction of a portfolio of stocks where we assume each stock has a normal distribution of returns and equal volatility. Each curve illustrates the effects of a different correlation of returns between companies, with the r=0.25 curve being similar to that between companies in the SP500 index. Although identical volatility and correlation is unlikely to be the case with an actual portfolio, it does nicely illustrate the diminishing returns of diversification.
93 94

For a real enjoyable book on this subject, see Nassim Talebs The Black Swan When I wrote this back in 2007, this seemed unlikely

Page 143 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

The problem with excessive diversification is that usually there are just not that many exceptional companies whose stock is trading at a reasonable price. Moreover, I believe that owning shares of exceptional companies more than compensates for the risk of focusing on just a few of these companies, as opposed to wider diversification. In our own portfolio, which is used to provide retirement income, we stop accumulating shares in a company when the total market value of the position exceeds 20% of the entire portfolio. The actual maximum allocation will depend on the companys rating, which averages three metrics: operating history, future prospects, and financial strength (as described in Chapter 5). The rating can range from 1.0 to 4.0, and the maximum company allocation is calculated as the product of the rating and 5%. However, we do not sell shares just because our position in a company exceeds this threshold. Figure 8-1
Volatility Reduction as Function of Portfolio Size
120.00%
Portfolio Volatiltiy as % of Stock Volatility

100.00%

80.00%

60.00%

40.00%

20.00%

0.00% 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 Number Stocks R=0.00 R=0.25 R=0.50 R=0.75

8.3 Dividend Policy


What is an optimal payout ratio for a company? If we are talking about a company where depreciation is roughly equal to recurring capital expenditures and consequently the companys earnings are actually available to be paid out as dividends, then the answer to this question depends on the return a company can expect to earn on its retained earnings. The higher this return, the lower the payout should be to maximize shareholder returns. Still, since a companys future return on retained earnings is unknown, I prefer to receive Page 144 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

at least some dividends from a company regardless of how bright its future prospects. In practice, I usually wont purchase a companys shares unless the ten-year trailing payout ratio has been at least 20%, and the dividend growth has been monotonic over the same time period. Some arguments for at least some payout are: The payment of a dividend can reduce agency risk. Dividends are the only way of converting intrinsic value to cash without selling a companys shares. Receiving cash from an investment sooner rather than later intuitively decreases risk.

Agency Risk Agency risk was discussed at the end of Chapter 5. One way of mitigating agency risk is to leave less money at the disposal of a companys management, i.e., to pay out more earnings as dividends. Typically, any CEO can find a way to spend 100% of a companys earnings, but these potential investments do not all have the same expected risk-adjusted return on investment. But when a company pays out a good portion of its dividends to shareholders, it must be pickier about the projects financed with retained earnings, with the result being a higher average return on investment. To me this makes much more sense than the argument put forward by private equity firms for using huge levels of debt (and associated interest payments) to keep management focused.

Retirement Portfolios Assuming that you start out with a reasonable withdrawal rate, the single most common cause of premature portfolio depletion is being forced to sell shares during a bear market. A bear market can create serious problems, as the lower a stocks price falls during a bear market, the more shares you need to sell to meet your withdrawal requirements. Consequently, even when the shares price starts to recover, you may have depleted the number of shares to the point where the portfolios withdrawal requirement creates an unsustainable withdrawal rate, with the result that the portfolios purchasing power rapidly declines, even as the price of the portfolios individual securities are increasing. Since a portfolios investment income tends to be much less volatile than its market price (compare the dividend and price time series in Figure 2-1), a portfolio with a level of investment income sufficient to meet your income needs will reduce the expected number of shares you end up selling during a bear market. This is easiest to see in the limit where we assume that investment income has no volatility; in this case, you never need to sell shares.

Page 145 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Typically it is recommended that a retirement portfolio have a large bond allocation to provide a buffer during bear markets. Unfortunately, investments with inherently low price and income volatility such as bonds - also tend to generate low rates of return. For this reason, it is often necessary to create a portfolio that generates sufficient investment income from common stocks if we desire the portfolio to both generate a sufficiently high rate of return and provide a high current level of investment income. In practice, this means creating a portfolio of exceptional companies that also pay out a significant portion of their earnings as dividends. Note that a portfolios sustainable investment income can differ dramatically from the portfolios current level of dividends and interest. To see why, consider the case where the portfolio contains bonds (which usually have a negative real income growth rate) and half of the stocks are companies that pay out so much of their earnings as dividends that their expected dividend growth rate is negative, with the remaining stocks having a very low dividend payout ratio (and insignificant current income), but a higher dividend growth rate. Here it is easy to see that the portfolios current income level is not really sustainable in the short to medium term, as it will fall until the dividends provided by the stocks with a low payout ratio (but higher dividend growth rates) have sufficiently increased. In practice, I measure the sustainability of a portfolios current level of investment income by adjusting the companys per share income as follows. If the companys payout ratio of dividends from sustainable earnings is 50% or less, the adjusted income per share is equal to the dividends per share, otherwise, the adjusted income per share is equal to 50% of the companys sustainable earnings per share (SEPS). The calculation is shown below, where SIPS is sustainable income per share, DPS is the dividend per share, and PR is the ratio of the dividend per share to sustainable earnings per share.
if (PR > 50%) SIPS = SEPS * 50% else SIPS = DPS

To illustrate, if the dividend payout ratio of United Technologies Corporation is 25% (less than 50%), we would count the entire dividend as sustainable income. On the other hand, if the payout ratio of Pengrowth Energy were 175%, we would only count 28.6% of the dividend as sustainable income.

Risk For a given expected return, if the cashflows are weighted more towards the present, then intuitively, risk decreases. After all, industry dynamics, a companys competitive

Page 146 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

advantage, and the demand for a companys products and services will all change over time, so the sooner we get our cash back, the less we are exposed to the risk that a companys earnings power will be impaired by some unexpected change. Probably the best way to visualize this is by imagining a company whose shares you own suffers some catastrophic shock at some point in the future, and the company goes out of business. Here it is obvious that you would have been better off having received as much cash as possible (through dividends) prior to the bankruptcy. As an example, back in early 2003 I started accumulating shares in a UK bank, Lloyds TSB (this was before I had developed the investment approach documented in this book). I kept accumulating shares until mid 2008, and over time, my average cost basis was around $31.78 a share. As banks go, Lloyds was fairly conservative, and I felt it was a good long-term investment. Unfortunately, management decided (with some prodding from the government) to acquire HBOS, a bank with a riskier loan portfolio and too much dependence on wholesale funding. I thought this move stupid, and sold my entire position at $17.32 a share. If the company had never paid a dividend, this would have created a loss of 46%, with the annualized loss being lower because of the long average holding period. However, back in 2003 Lloyds had a dividend yield of over 8%, and maintained its payout up to the time I sold my shares. Although this yield fell to 6% during my later purchases, an average 6% yield over six years decreased my actual loss considerably by returning a good chunk of my initial investment in cash. As a rough guess, I would say my actual realized total loss was maybe 15%, considerably less than the 46% capital loss.

8.4 Frequency of Research


I update my calculation of a companys sustainable earnings per share and estimate of its intrinsic value whenever the company releases quarterly earnings information, and update the data used to quantify company quality when the companys 10-K is published. I also reanalyze a companys future prospects once a year. Sometimes this reanalysis will change my opinion regarding the predictability of the companys future cashflows changes. For example, I might have purchased the shares of a company at a time when I believed that the companys cashflow predictability was such that a ratio of intrinsic to market value of 1.25 gave me an adequate margin of safety, and planned on holding the shares until the ratio of intrinsic to market value hit 0.7 or less. But lets say that when I reanalyze the company three years after my purchase, my analysis indicates that the companys competitive advantage has been eroded, and a more appropriate purchase and sell threshold would be a ratio of intrinsic to market value of 1.5 and 1.0 respectively. At this point, the decision to sell the companys shares depends on two factors: the current ratio of estimated intrinsic to market value and alternate investment opportunities. To illustrate, if the current ratio of intrinsic to market value is 0.8, and I can find better companies trading at a ratio of intrinsic to market value of say 1.3, then I probably want to sell the shares and re-invest the proceeds. On the other hand, if the current ratio of intrinsic to market value is 1.9, I might decide to purchase even more shares.

Page 147 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

8.5 Benchmarking
Personally, since we do not have a target asset class allocation, but our portfolio is usually weighted towards large cap domestic stocks, I use the SP500 as a long-term benchmark. Over long time periods, both foreign stocks and REITs have had similar rates of return and volatility. On the other hand, if you do have a target asset class allocation, then a better benchmark would be a group of index funds tracking those asset classes and weighted as per your target allocation. The correct way to measure a portfolios performance is by calculating its internal rate of return, which is the rate of return that brings the present value of all the portfolios contributions and withdrawals equal to the current market price (this is described in detail in Appendix 9.1. We can also evaluate a portfolios performance using estimated intrinsic value. The first step in estimating a portfolios intrinsic value is to use our valuation model to estimate the per-share intrinsic value of each company in the portfolio. We can then estimate the portfolio's total intrinsic value by multiplying the estimated intrinsic value per share for each company by the number of shares held in the portfolio, and then summing this product for each company held in the portfolio. For cash and short-term bonds, we assume that each $1 of market value equates to $1 of intrinsic value. A portfolios internal rate of return can then be calculated using intrinsic value, adjusting for portfolio withdrawals and contributions in the same way we do when we calculate a portfolios internal rate of return using market value. A portfolios internal rate of return calculated using intrinsic value will usually be less volatile than its internal rate of return calculated using market value, and gives a good picture of changes in the portfolios sustainable earnings. We can benchmark this rate of return against changes in the SP500s intrinsic value, which is easily calculated by plugging the operating earnings of the SP500 into our valuation model. These operating earnings are basically equal to reported earnings with non-recurring items removed, and can be found on the Standard and Poors website. As we mentioned earlier, the greatest impediment to successfully implementing our investment strategy is a lack of patience; specifically, the patience to wait, sometimes for years, for the market value of a companys stock to catch up with our appraisal of its intrinsic value. For this reason, I prefer to evaluate a portfolios relative market performance over five-year rolling periods, as this is generally enough time for the market to weigh a companys intrinsic value. Still, it is human nature to want to get frequent feedback. But rather than using my portfolios market value for short-term performance feedback, I prefer to track a each companys organic revenue and operating income growth on a quarterly basis. This measure of performance will tend to have a much higher correlation to a portfolios long-term performance. What you want to avoid is becoming fixated on short-term price movements of the companies in your portfolio. I guarantee that more often than not, when you purchase the shares of an exceptional company at a reasonable price, you will see the price fall

Page 148 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

significantly shortly thereafter. This is unavoidable. Ignore it, and focus on the growth of your portfolios earnings power.

Page 149 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

9 Appendix
9.1 A Portfolios Internal Rate of Return
The correct way to measure a portfolios performance is by calculating its internal rate of return; this accounts for contributions to and withdrawals from the portfolio. To illustrate the need for using the internal, rather than absolute95, rate of return, consider a portfolio that had a value of $1000 on the first day of the year, and a value of $2000 at the end of the year. At first, this looks like an incredible return of 100%. But lets say that during each month of the year, you had contributed another $75 to the portfolio; in this case, your rate of return is actually only 7.11%. To accurately determine our portfolios annual rate of return, we need to do more than compare the value at the end of the year with the value at the beginning of the year; we need to see what rate of return we are getting on each individual portfolio contribution. In the previous example, one contribution is the $1000 that the portfolio contained at the start of the year, which has had twelve months to compound at the internal rate of return; another contribution is the $75 that we contributed at the end of the first month, and has had eleven months to compound, and so on. A portfolios internal rate of return is defined as the rate of return that, if applied individually to each portfolio contribution96, will result in a sum of compounded portfolio contributions equal to the portfolios present market value. A portfolio contribution is defined by both its cash value and the elapsed time between the date of the contribution and the present (i.e., how long it has had to compound at the internal rate of return), and a compounded portfolio contribution is the contribution multiplied by one plus the monthly internal rate of return raised to the Nth power, where N is the number of months the contribution has had to compound. A contribution can be positive, as in the case of adding cash to the portfolio, or negative, as in the case of a cash withdrawal. Lets illustrate this with an example. Imagine at the start of the year your portfolio has $1000. During the year, you make a withdrawal (hence the minus sign) of $100 on 3/1, a contribution of $75 on 4/1, and a contribution of $125 on 9/1, and at the end of the year, your portfolios value is $1500. Table 9-1 shows each contribution, the date it is added to the portfolio, and the number of months the contribution has to compound until the end of the year. Note that we treat the initial portfolio value of $1000 as a contribution.

95

By absolute rate of return, we mean the rate of return calculated as value at the end of the year divided by value at the start of the year, minus one; this does not compensate for withdrawals or contributions. 96 The portfolios value at the beginning of the measurement period is also considered a portfolio contribution. Also, we will consider a portfolio withdrawal to be a negative contribution.

Page 150 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Table 9-1
Contribution Date Months to compound $1000 1/1 12 3/1 10 -$100 4/1 9 $75 9/1 4 $125

The portfolios monthly internal rate of return can be determined by finding the monthly rate of return R required to bring the sum of the compounded contributions equal to the portfolios end of year value. This is shown below, where the compounded contributions are on the left hand side of the equation, and the portfolios final value is on the right. Each compounded contribution is determined by the corresponding contribution multiplied by one plus the internal rate of return to the Nth power, where N is the number of months the contribution has to compound.
1000 * (1+ R)12 100 * (1+ R) 9 + 75 * (1+ R) 8 + 125 * (1+ R) 3 = 1500

We can solve for R by taking the logarithm of both sides and solving for R, alternately, we can find R by performing a binary search. Either way, we find that our monthly internal rate of return, R, is equal to 2.57%; the annualized return is found by taking one plus the monthly return to the 12th power, and subtracting one, given us an annualized rate of return of 35.6%. Note that this is less than what would be obtained by taking the quotient of the portfolios ending value to starting value, and subtracting one. Although the portfolio did actually increase 50% in value during the year, part of this was due to the positive portfolio contributions.

9.2 Internal vs. Realized Rate of Return


Earlier we discussed the concept of the rate of return implied by an investments current market price and a given forecast of future cashflows. We derived this implied rate of return by finding the discount rate which, when applied to each of the forecast cashflows, resulted in the investments market price being equal to the investments intrinsic value. This procedure implicitly assumes that we are consuming the dividends as they become available, rather than re-investing them; to see this, note that for the case of a holding period sufficiently long to make the effect of the market price at time of sale negligible, the implied rate of return depends solely on the current dividend yield and the forecast of future dividends, and therefore the market price after purchase (the price at which dividends would be re-invested) has no effect on this rate of return. But what happens if we re-invest a portion of our dividends back into the investment, and then sell the investment at some future time? As it turns out, if we assume a constant dividend yield over the holding period, then the rate of return is exactly the same as if the Page 151 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

dividends are consumed at the time they are received. This concept is best understood using the example of stock shares, where we can re-invest dividends by using the dividends to purchase more shares. First lets imagine a company where the dividends per share grow annually at a constant rate G, and the dividend yield remains constant at Y. In this case it is easy to show97 that if we re-invest all dividends, the return will be equal to Y + G, which is exactly the same as the rate of return implied by the market price and assumed dividend growth rate G. The difference is that instead of consuming the dividends as we receive them, we re-invest them into the company, with the result that the value of our investment will grow over time at a faster rate than if we had consumed the dividends; this excess growth is from the compounding of the re-invested dividends. At some future time, the investment would then be sold at a price determined by the dividend at the time of sale divided by the dividend yield at the time of sale, which we are assuming remains constant over the holding period. Despite the larger investment growth when the dividends are re-invested, the actual rate of return to the investor is the same as if the dividends were consumed; this is due to the fact that the large single dividend for the case where we re-invest all dividends has a large discount factor. Lets look at this with an example. Consider a company whose shares are trading at $100, the current dividend is $8, and the forecast dividend growth is 2%. Now lets say we purchase 1000 shares, and hold them for ten years, re-investing all dividends; lets further assume that the dividend yield remains constant at 8% over the holding period. Now over the 10 years, the income will grow at a rate equal to 10% (Y+G), resulting in the income growing from $8000 at the time of purchase to $20,780 at the time of sale, and with constant yield, the investment of $100,000 will grow to $259,374; this gain of 259% over ten years is equal to an annualized rate of return of 10%. Lets compare this to the case where the dividends are consumed each year rather than reinvested, and the investment sold after ten years. Since we do not re-invest dividends, the investment only grows by 2% a year rather than 10% a year, yielding sales proceeds of $121,899 after ten years, the dividends from this investment is shown below: Table 9-2
Year D MV DF DD SDD
97

1 8.00K 102K 1.1 7.27K 7.27K

2 8.16K 104K (1.1)2 6.74K 14.02K

3 8.32K 106.1K (1.1)3 6.25K 20.27K

4 8.49K 108.2K (1.1)4 5.80K 26.07K

5 8.66K 110.4K (1.1)5 5.38K 31.45K

6 8.83K 112.6K (1.1)6 4.99K 36.43K

7 9.01K 114.9K (1.1)7 4.62K 41.05K

8 9.19K 117.2K (1.1)8 4.29K 45.34K

9 9.37K 119.5K (1.1)9 3.98K 49.32K

10 9.56K 121.9K (1.1)10 3.69K 53.00K

10* 121.9K 121.9K (1.1)10 47.00K 100.00K

The value at time of sale is given by the income at the time of sale divided by the dividend yield at time of sale, while the purchase price is given by the income at time of purchase divided by the yield at time of purchase; with our assumption of constant yield, the yields cancel, and the return is determined by the growth in income, which is given by the sum of the dividend growth accruing to one share (G) and the increase in income resulting from the increase in shares funded by the re-investment of dividends. The number of shares you can purchase in a given year using re-invested dividends is given by Div / Price, which equals the dividend yield (Y); therefore by re-investing all dividends you grow income at a rate equal to G + Y.

Page 152 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

D=Dividend, MV=Market Value, DF=Discount Factor, DD = Discounted Dividend, SDD=Sum Discounted Dividends * This is from sale of shares at end of tenth year

Table 9-2 clearly shows that the discount rate (and rate of return to the investor) required to set the intrinsic value of the shares equal to the market value of $100K is ten percent, the same rate of return that results when dividends are re-invested over the holding period. So far we have assumed constant yield over the holding period, but in reality, the dividend yield of most investments will change over the holding period, causing the implied and actual rate of return to differ. Depending on the pattern of future dividend yields, the return with dividends re-invested can be either higher or lower than the implied rate of return. Over a long enough holding period, rising yields will provide a higher actual rate of return than falling yields, but over shorter holding periods, the market price dominates more of the return, and the reverse holds true. Although the internal rate of return and realized rate of return assuming dividends are reinvested will differ, the difference is usually not significant. For example, looking at tenyear rolling returns from investing in SP500 index shares from 1872-2006, we find that the average internal rate of return was 6.98% with a standard deviation of 5.04%, whereas the average realized rate of return with dividends re-invested was 7.00% with a standard deviation of 4.99%, and 85% of ten-year periods, the difference between the returns was less than one percent. The effect on price changes during your holding period on your realized rate of return is highest when you are re-investing all of the dividends, and will matter less when you are consuming most of your dividends. For example, imagine you are saving for retirement by contributing some amount of cash each year to an investment portfolio. In this case, you are going to accumulate more wealth if the market values of the securities you are accumulating are depressed with respect to their intrinsic value, as you will be paying less for a given amount of future cashflows. Now imagine you are retired, and withdrawing some annual amount from your retirement portfolio. In this case, you are better off if each year, the securities in your portfolio are trading at a premium to their intrinsic value, because you will need to sell fewer shares to provide the same annual withdrawal rate98.

9.3 Geometric Vs. Average Rate of Return


In this book we will often refer to a stocks geometric rate of return, so lets take a quick look at how this differs from a stocks average annual return. Since a precise illustration
98

This is true even if generally, your portfolio provides enough dividend income so that you do not need to sell shares. There is always the chance that your dividend income will fall enough where you need to sell some shares.

Page 153 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

of the differences is a bit technical, lets start out with a simple discussion demonstrating the difference between these two measures of rate of return. Over single year holding periods, a stocks rate of return is determined by the stocks price at the start of the year ( PSOY ), the stocks price at the end of the year ( PEOY ), and the dividends received during the year (D); this is shown in Equation 9-1:

Equation 9-1 P PSOY + D R = EOY PSOY A stocks average rate of return over some number of years is just the arithmetic average of the annual returns (as given by Equation 9-1), calculated by summing these rates of return and dividing by the number of years you are calculating the average over. For example, if the series of annual returns was 10%, -20%, and 30%, then the average 10% 20% + 30% annual rate of return is = 6.67% . 3 Before we show the calculation of the geometric rate of return, lets first define a position in a stock as the number of shares that we currently own. If we re-invest all of our dividends, then our position will change over the course of a year by the starting value of our position (the number of shares multiplied by the price per share) multiplied by one plus the annual rate of return. In our simple example where we illustrated the calculation of a stocks average annual rate of return, if we started with a stock position worth $100, it would be worth $110 ( (1+ 10%) * $100 = $110 ) after the first year, $88 after the second year, and $114.40 after the third year. Now we can define the stocks geometric rate of return as the constant rate of return R required to bring the value of our starting position ($100) equal to the value of our ending position ($114.40), which in this case is 4.6%, less than the average annual rate of return of 6.67%. An investments geometric rate of return is always less than its average rate of return. Now although an investments average rate of return does have its uses (Monte Carlo simulation being one), it is pretty obvious that an investments geometric rate of return is the preferred method of evaluating an investments potential to increase your wealth. If you are interested, and dont mind a little math, we will continue with a more rigorous treatment of the difference between an investments average and geometric rate of return. If we decide to invest some amount of cash Vo into the shares of a company, we can compute the value of our position at the end of the period as
N

VEND = Vo (1+ R1 )(1+ R2 )L (1+ RN ) = (1+ Ri ) , where Ri is the rate of return realized i=1 during year i, as given by Equation 9-1, and N is the number of years in the holding period. From here we can calculate the annualized rate of return as:

Page 154 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

ln(VEND / Vo )

R= e

For the case where the rate of return remains the same from year to year, the annualized rate of return simplifies to the average rate of return over the holding period. But when the rate of return varies from year to year, the average return calculated over a given multi-unit period will differ from, and be less than, the annualized return over that period, as we demonstrated earlier. Since it is the annualized rate of return that determines the difference between a stock positions (or a portfolios) ending and starting value, an investor will generally find the annualized rate of return to be more important than the average return. In the future, we will usually refer to this annualized rate of return as the geometric rate of return. If the period is measured in years, then the geometric return will be the annualized rate of return, and if it is measured in months, then the geometric return will be the monthly rate of return required to bring the positions ending value equal to its initial value. One way to look at the difference between a positions average and geometric return is that the average rate of return is the expected rate of return in any given year of your holding period, but the geometric rate of return is the expected rate of return over the holding period. Since we really dont know the precise sequence of a stocks future annual rates of return, we will sometimes characterize a stocks future rate of return as an average annual rate of return and the rate of return standard deviation. When we characterize future returns as a series of random variables, we can estimate the difference between the annual returns and the geometric rate of return as quantity known as variance drag. 2 99 This difference is given by : VarDrag = Raverage Rgeo = , where Rgeo is the 2(1+ Raverage ) components geometric rate of return, Raverage is the stocks average rate of return, and 2 is the stocks rate of return variance. For a single unit holding period rate of return, the variance drag will be zero, for longer holding periods, the variance drag is always greater than zero. A consequence of variance drag is that it is possible for a stock with a given average rate of return to have a higher geometric rate of return than another stock with a higher average rate of return, but a higher rate of return standard deviation as well. Lets end this section by noting that we can convert a geometric rate of return back into an 2RGeo + 2 arithmetic return as Raverage = . 1 RGeo + (1+ RGeo ) 2 + 2 2

99

When the average annual returns are unknown and expressed as a series of random variables, the difference between the average and geometric rate of return cannot be precisely quantified, but the variance drag gives an estimate of this difference that produces very little error, except for the case of variances that are larger than that typically used to characterize a securitys rate of return volatility.

Page 155 of 156

Copyright 2009 by Brian Gaudet, All rights reserved, including the right of reproduction in whole or in part in any form

Page 156 of 156

Anda mungkin juga menyukai