Tuesday, June 9, 2009

About P/E, Normalized Earnings and Normalized P/E Ratios


How do you determine whether something is cheap or expensive? Often, valuation is expressed in terms of price per unit. For bananas, you calculate the price/pound; for homes, price/square foot; and so forth.

When you own a share of stock, you own a portion of a company. As a result, you own a portion of its earnings. Therefore, stock market valuation is often measured by the price the market is willing to pay for a dollar of yearly earnings -- that is, by the price/earnings (P/E) ratio. To calculate the ratio, you divide the price of a company's stock by that company's earnings (per share) for one year. If the result is, say, 17, your broker might tell you Company X has a P/E of 17, or Company X is selling at 17 times earnings. The P/E is also referred to as the earnings multiple.

Similarly, since you also own a portion of the company's dividends, valuation is sometimes measured by the price/dividend ratio -- i.e., how much the market is willing to pay for each dollar of yearly dividends (per share). Other methods of valuation include price/cash flow and price/book value (the accounting, or "book," value of the company's assets).

Normalized Earnings (NE)

When evaluating the earning power of a business as of some point in time, what earnings should you use? The most recently completed calendar year? The most recent four quarters? The next calendar year? The next four quarters? If you use any of these earnings, you will find that the earnings can vary significantly in a relatively short span of time. The results are much more variable than makes sense to me since they vary more I intuitively think the real long-term earning power of the business is changing. As a result, many investors look for more stable/consistent measures of earnings. Often, they will "normalize" or smooth earnings so that the earnings appear less volatile. I think of this as an attempt to estimate long-term "sustainable" earnings.

In this blog, I normalize earnings by averaging the earnings over an 11-year period -- from 5 years before through 5 years after. (See, for example, the graph of normalized earnings since 1929 in this post.) Other methods for normalizing earnings include using the prior 10 years' earnings (used by Benjamin Graham, Robert Shiller and others), and using peak earnings (used by John Hussman). I often abbreviate normalized earnings as NE.

Normalize Price/Earnings Ratios (NPE)

An important reason for normalizing earnings is so that you can calculate a normalized price/earnings ratio. If you use yearly earnings, the resulting P/E varies more than seems intuitively reasonable. This volatility can be smoothed by using a normalized P/E. I calculate the normalized P/E by dividing price by normalized earnings. I often abbreviate this as NPE. (Note: Some others first normalize the price -- e.g., by taking the average price over a 30 day period.)

Dow Jones NPE

In this blog I treat the DJIA (Dow Jones Industrial Average) as though it were one big company. Therefore, the price I use is the closing price of the Dow Index; earnings (before I normalize them) are the published Dow Index earnings for that year. The Dow normalized P/E ratio (NPE) is, then, the closing price divided by normalized earnings, and is an indication of how cheap or expensive the Dow was at that time -- and, by implication, how cheap or expensive the total stock market was. For an example, see the graphs of normalized P/E ratios in Dow P/E Ratios since 1929.

Note: By definition, my normalized, smoothed, numbers are not available until 5 years after the fact. Thus, they are first and foremost analytical concepts -- tools useful when analyzing history. However, their usefulness is not limited to historical analysis (see, for example, the series on projecting stock market performance).

Last modified 6/6/2011

This work is licensed under a Creative Commons Attribution 3.0 unported license.


  1. Nice blog... I think your discussion on retirement is pretty good.

    You may already know this, but the problem with NE (5 yrs before and 5 yrs after) is that it does not capture what a person knows *at that time* using the earnings data alone.

    If I knew what the next five years are going to be like I would do pretty well myself.

  2. Anon,

    Thanks for the comment.

    I developed my approach to calculating normalized earnings (NE) as an analytical tool to help me understand the historical behavior of the stock market; as a result, the approach is primarily backward looking. I'm trying to smooth stock market earnings so that it's easier to discover fundamental relationships and patterns. You can make a very strong case that the market price is in fact more a function of future earnings than past earnings. So, from the point of view of analyzing historical market behavior, there is clearly no problem with this approach to normalizing.

    However, the original version of this post was easy to misinterpret. I can see how you could think that I was somehow using, for example, 1990's normalized earnings in the year 1990 even though they would not be available until 5 years later; that is not the case. I've modified the original post to, I hope, reduce the chances of this misinterpretation.

    Thanks for alerting me to this possible confusion -- and thanks for stopping by.

  3. If we are looking at your use of normalised earnings it seems to me that the following applies:

    Say the next 5 years after a point has very good earnings. At the start of those 5 years no one has any idea what earnings will be. If those 5 years are good, then the normalised earnings will be higher and the starting PE lower and the good earings may be just as if not more relevant in the performance.

    If the 5 years forward are bad this would push up the normalised PE at the start and push down the return.

  4. Sorry it took me so long to get back to you. I've made some changes to the post that I hope will make it clearer.

    I'm not completely sure that I understand your comment. However, I think the point to remember is that because of the way I normalize, I don't know what the normalized earnings are until 5 years after the fact. That may seem strange. However, if you think about it I think you'll agree that if you average the earnings from 1950 through 1960 the result is a better estimate of the long-term sustainable earnings in 1955 than it is of 1950's or 1960's sustainable earnings.


No spam, please! Comment spam will not be published. See comment guidelines here.
Sorry, but I can no longer accept anonymous comments. They're 99% spam.