P/E Ratios vs. Earnings Yields

Paul Haahr
22 November 2000

I've got a question which I don't have the training to answer, but I found interesting to think about anyway: Why do people use P/E ratios rather than earnings yields?

Background: The usual measure of how ``expensive'' a stock is is its price-to-earnings ratio -- the ratio of the price of a share to the earnings-per-share or (equivalently) the ratio of market cap to earnings; often, it's phrased as ``how much you have to spend on the stock to buy a dollar of earnings.''

The earnings yield is the inverse of the P/E: the ratio of earnings to price, and is usually -- by analogy to dividend yield -- described as a perctentage; it can be understood as ``the amount of earnings you buy with one dollar of stock.'' ( When expressed as a percentage, the earnings yield is easy to interpret as ``pennies of earnings per dollar of investment.'')

The complication is that if a company is not profitable, either you have a negative P/E or (more commonly) the P/E is undefined. In the olden days (i.e., before biotech and then the internet), few public investors -- and certainly few investors who looked at fundamentals -- were interested in companies without profits, so it may have been a moot issue. But, today, with plenty of public companies in the red, understanding the fundamentals of money-losing operations is a good idea. And, I'll argue, that's why investors should use earnings yield instead of P/E.

Notably, companies with very large positive P/Es are much more similar in fundamentals to companies with very small negative P/Es than companies with positive and negative P/E ratios near zero are too each other.

This seemingly obscure detail became important to me a couple of months ago, as part of my job, because Xigo allows users to set up alerts which trigger when a company's P/E goes above or below a target value. But what does that mean for a curve, such as P/E, with a discontinuity?

To see where the P/E causes trouble, consider the article A Dozen Stocks Still Defy Gravity by Mark Martinez at TheStreet.com. (21 November 2000)

The article talks about the absurd valuations of a number of companies, citing, for example, Research In Motion's P/E of 1125. I'm not going to argue whether RIMM is worth it's present value or not, I'll just observe that (according to Yahoo), they earned $.08/share over the past year, but lost $.02/share in the past quarter. If that loss continues, their P/E will rise (assuming constant stock price), until they have a loss for a year, at which point it will suddenly drop and become a negative number with a very large magnitude. And, at that point, the author of this column at TheStreet.com will move them to a very different category. And if, perchance, RIMM exactly breaks even for a year, they will have an infinite P/E. (Don't say it couldn't happen -- everyone thought that Presidential elections wouldn't be tied, either.)

So, a user who sets up a Xigo alert for RIMM when their P/E goes below 30 -- presumably because they are waiting for them to earning more money relative to their share price -- will get an alert if RIMM starts losing enough money such that they're no longer profitable. (Oops.)

Further, I suspect that the use of P/E makes it hard to compare quarter-to-quarter numbers for a given company, because variations in EPS lead to inverse linear effects on P/E, and people seem to be much better at understand linear behavior than non-linear. In particular, people often mistakenly approximate all changes as if they have linear effects, even if they know that they don't.

Here's my rule of thumb: when formulating a ratio between a number which can be negative (in this case, earnings) and one which can only be positive (share price), use the always positive value as the denominator.

In this case, the earnings yield is exactly the right quantity. It can serve the same role as P/E, except that it doesn't have the discontinuity around zero. It has an obvious mapping to an intuitive meaning, and it's much more useful if you're further doing computations based on the ratio.

Here's a table of a bunch of stocks, with their latest prices, earnings per share and P/E ratios, with all data based on the November 22 values from Yahoo Finance, sorted by P/E ratio:

Here's the same table, but using earnings yield instead of P/E:

Which table presents a more useful ordering?

I chatted about this with a Xigo colleague who used to work at BARRA, which does portfolio management and risk management. For their risk models, they use E/P -- the earnings yield -- internally.

So, why do we use P/E?

I suspect I know the answer: that's the way it's always been done. But is there something I'm missing?