This article describes a simple valuation model I use to estimate a share’s intrinsic value. A share’s 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 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. The required rate of return, also known as the discount rate, should increase with decreasing certainty regarding an investment’s future cashflows; i.e., the present value of a future cashflow with a given expected value falls with decreasing certainty.
The primary obstacle encountered when estimating a share’s 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 company’s earnings, the future cashflows arising from owning a share of a company’s common stock is in many cases completely unpredictable.
Although it is sometimes possible to predict next year’s dividend with some accuracy, the share’s 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 year’s time. And for longer holding periods, where a share’s 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 company’s future earnings power. About the only thing we can state with any certainty is that the company’s dividends and share price will over long time periods roughly track changes in the company’s earnings power.
Predicting a company’s 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 company’s competitive position. In this case, the company’s 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 company’s share price will still be largely unpredictable - aside from a loose tendency to track the company’s earnings power - we get around this source of inaccuracy by holding a company’s shares until the market makes us an offer we can’t 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, I use a valuation model that is calibrated such that the ratio of a company’s estimated intrinsic value to demonstrated earnings power is equal to that of an average company with a given level of financial strength. Although this will rarely result in an accurate prediction of a company’s perfect foresight intrinsic value, like John Maynard Keynes, I would rather be approximately right than precisely wrong. Moreover, since I apply this model only to companies that have demonstrated evidence of a strong competitive advantage, have a good position in stable and profitable industries, and are financially sound, our model will be biased such that it underestimates a share’s 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.
One useful characteristic of our model is that it takes only two company specific inputs, demonstrated earnings power 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.
This valuation model is an example of forecasting using an “outside view”, where instead of trying to add as much information as possible to a model in order to increase accuracy, we instead look at the base rate (in this case statistics gathered from the SP500), and assume the company we are analyzing will have similar performance. Although it seems counterintuitive, research in behavioral finance (including papers collected in “Judgment under uncertainty: Heuristics and Biases” edited by Kahneman, Slovic, and Tversky) show that restricting ourselves to known information concerning base rates is often more accurate than using all available information when we attempt to forecast the future – particularly when some of this information is at best loosely correlated with the output being predicted. There is evidence (See David Dreman’s “Contrarian Investment Strategies”) that stock analysts, who typically try to include as much company specific information as possible into their models – consistently overestimate the earnings growth rate of the companies they follow, often by 10% annualized.
My valuation model is a constant dividend growth dividend discount model, where I assume that a company paying out 50% of its earnings as dividends can grow per-share earnings at a long-term real rate of 2% (this serves as the modest long-term growth hurdle I mentioned earlier). Two percent was the real rate of earnings growth for the SP500 index from 1950 to 2005, a period over which the average payout ratio was 50%. I use a 5.5% real required rate of return for my discount rate; this was the real rate of return on the SP500 index from 1950 to 2005, with the effects of the rising price to earnings ratio factored out. An estimate for a share’s intrinsic value is then given by:
Here SEPS is the company’s sustainable earnings per share, which is calculated by creating a time series combining a conservative estimate of next year’s earnings with 9 years of trailing earnings adjusted for non-recurring items, and then translating the time series to today’s dollars. Sustainable earnings per share (an estimate of the company’s demonstrated earnings power) is then calculated as the average of the de-trended time series (See this article for de-trending algorithm).
Although most companies don’t have a payout ratio of 50%, I make the simplifying assumption that if a company’s actual payout ratio is higher or lower than 50%, the real earnings growth rate will be correspondingly lower or higher than 2%, and so the share’s estimated intrinsic value becomes a function of the company’s demonstrated earnings power. Of course this raises the question of why did I not just use an earnings capitalization model, and set the estimate of intrinsic value equal to some historic average multiple of demonstrated earnings power.
The answer is the credit rating adjustment, which needs to be applied to a growth rate and discount rate. The model equates to an earnings capitalization rate of 7% for a company with an A2 credit rating, or stated differently, would consider such a company fairly valued at a price to sustainable earnings ratio of 14.3.
CRA is the credit rating adjustment, which adjusts for the company’s financial strength. The probability of a company defaulting on its debt should be of interest to an owner of the company’s 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 company’s long-term issuer rating and the company’s subsequent default rate, and intuitively, it makes sense that, at a minimum, we would adjust a company’s earnings growth rate by the difference between the default rate implied by the company’s credit rating and the default rate implied by the SP500’s historical average credit rating, with the reasoning being that the asset class’s 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.
Our model uses the following growth rate adjustment:
Here is the sum of the annualized default rate and default rate standard deviation for the asset class (in this case large cap domestic stocks), as implied by the asset class’s historical average credit rating, and is the sum of the annualized default rate and default rate standard deviation for the company, as implied by the company’s current credit rating. In this equation, I use Moody’s default rate statistics from 1920 to the present. I also do my own analysis of a company’s financial strength, and if necessary, reduce the rating downward (but never upward).
The reason we take into account default rate volatility is that lower credit ratings are associated with higher default rate volatilities, and this increased volatility reduces the certainty of bondholders receiving the expected value of future cashflows. It is fairly intuitive that this increased default rate volatility also decreases the certainty of the expected value of shareholder’s future cashflows. To account for this, we should adjust our discount rate based off of the default rate volatility implied by the company’s credit rating.
Our model does have a potential problem, in that we do not randomly choose companies, but instead select them when they are trading at a discount to our estimate of their intrinsic value.
This introduces negative selection bias to the extent that the market is efficient, and the discount ends up being justified. For two reasons, this is not an issue. First of all, there is considerable evidence that choosing companies only on the basis of compelling valuations results in market beating returns. Second, positive selection bias is introduced by focusing on companies that have demonstrated evidence of a strong competitive advantage, have a good position in stable and profitable industries, and are financially sound.
The following table shows the sustainable earnings per share and estimated intrinsic value for Johnson & Johnson (NYSE:JNJ), Procter & Gamble (NYSE:PG), Automatic Data Processing (NASDAQ:ADP), and The Coca-Cola Company (NYSE:KO) as of 05/27/09. Also shown in the last row is the real earnings per share growth rate that would need to be sustained to validate the estimate, assuming a constant payout ratio equal to the company’s current payout ratio.
Since each of these companies has demonstrated evidence of a strong competitive advantage, has a good position in a stable and profitable industry, and has exceptional financial strength, I have high confidence that each company can meet the earnings per share growth rate hurdle given in the last row. It follows that my estimate for each company’s intrinsic value is conservative, and by purchasing shares at less than this estimate, I will likely realize a rate of return exceeding my 5.5% real required rate of return.
More details concerning this valuation model and the calculation of sustainable earnings per share can be found on my website.
Disclosure: I own shares of JNJ, PG, ADP, and KO.