A key task in value investing is determining the “normal” earning power of a company in any given year.
People use a variety of methods to do this, the most famous being Benjamin Graham’s average earnings from the company’s previous 7-10 years of operation.
Here’s a technique I use to get a sense of a company’s historical normal earning power while weighing the highest and lowest earning years a little less.
The method is called the Program Evaluation and Review Technique (PERT) and is a project management tool invented by Booz Allen Hamilton and the U.S. Navy to build the Polaris missile system in the late 1950s. It was designed to determine likely schedule outcomes for specific project tasks based on estimates of the most likely, optimistic, and pessimistic times for completion.[1]
Here’s how it works.
Let’s say you have earnings from operations per share for a company for a 10-year period.
Instead of using a straight average, note the best year and the worst year.
Label the best year the optimistic earning power of the company and the worst year the pessimistic earning power of the company.
You could determine the most likely earning power in a variety of ways.
What I do is take the average of the remaining 8 numbers.
Then multiply the most likely earning power by 4 and add the optimistic and pessimistic numbers to that figure.
Divide the total by 6.
The result is called the expected value and is a normalized answer.[2]
For example, let’s say the reported earnings per share for a company for the years 1998-2007 are:
$1.22, $2.12, $2.62, $2.75, $2.59, $3.42, $3.26, $3.82, $4.25, and $.72.
$4.25 is your optimistic year and $.72 is your pessimistic one.
The straight average of the rest is $2.73 and is your most likely number.
4 X 2.73= 10.92.
Adding $4.25 and $.72 gives you $15.89.
Now divide by 6 to get $2.65.
Using the PERT method, $2.65 would be your mean or “normalized” earnings per share for this company based on the 10-year historical record.[3]
The standard deviation (SD) is equal to the pessimistic number minus the optimistic number divided by 6.
In this case the SD is (.72-4.25)/6= -.59
If this company’s earnings are normally distributed, that means that 68.4% of the time they will fall within plus or minus 1 SD of the $2.65 mean (between $2.06 & $3.24), 95.4% of the time they will fall within plus or minus 2 SD of the mean (between $1.47 and $3.83), and 99.7% of the time they will fall within plus or minus 3 SD of the mean ($.88 and $4.42).[4]
But don’t get too attached to this number: a business is a very dynamic organism and earnings may not necessarily follow a normal distribution.
Even if they do that .3% does happen: the company in our example earned $.72 in 2007, below the 99.7% probability low of $.88.
Plus, statistically speaking, 10 data points is not a whole lot.
Regardless, I’ve found the method a useful place to start, especially when examining companies with stable track records and business models.
If you’re interested in learning more about the PERT method and the normal and beta distributions as they relate to project management check out the following link:
http://www.interventions.org/pertcpm.html
[1] http://en.wikipedia.org/wiki/Program_Evaluation_and_Review_Technique
[2] The PERT calculations are based on a beta distribution but the result is normal in the statistical sense. My thanks to Mr. Ajit Mani of Intervention Ltd. of Bangalore, India for helping me understand why: “The weights are based on an approximation of the Beta Distribution, NOT the Normal. The originators of PERT selected the Beta distribution for the following qualities: a) It is unimodal; b) Has finite and non-negative end points; c) Is not necessarily symmetrical. The Normal distribution does not satisfy qualities b) and c). The Normal curve is used for calculating Probability of Completing a Project by a Given Date. This is possible because the Project Length Te is calculated by simply adding the te’s along the Critical Path. Since the te’s are all random variables, so is Te. ‘But the interesting (and fortunate, from a statistical standpoint) result is that Te does not have the same distribution as the te’s but follows what is called a normal distribution…’ Please refer Wiest and Levy, ‘A Management Guide to PERT/CPM’, Prentice-Hall International, Inc., Englewood Cliffs, N.J. USA, 1969, Chapter 4. The PERT Model.”
[3] Of course, whether the company can maintain and/or grow this normal earning power in the future without dilution is the key.
[4] http://en.wikipedia.org/wiki/Normal_distribution
