So far, I've referenced normalized P/E ratios in four of my posts: "A Look At 15-Yr. Normalized Dow P/E Ratios", "A Look At Normalized P/E Ratios and the Election Cycle", "A Second Look At Normalized P/E Ratios and the Election Cycle", and "On Normalized P/E Effects Over Time".

However, I've yet to explain how I calculated these normalized P/E ratios. Obviously, I took the Dow's average price for the year and divided by a normalized earnings number. But, how did I come up with a normalized earnings number – in other words, what exactly is the normalization process?

** The Normalization Process**

The normalization process is actually quite simple and straightforward. First, you need to decide upon a reasonable long-term growth rate; otherwise, you won't have a "trend" to use for comparisons between actual and "expected" earnings. Essentially, "normalized earnings" are just "expected earnings" based on a long-term trend rather than short-term considerations.

For the Dow, a reasonable long-term growth rate would be about 6%. Many different approaches (logical and empirical) will bring you to a similar conclusion. Of course, we could argue forever about what the "right" long-term growth rate assumption is.

That's because there is no right long-term growth rate. To the extent that future circumstances differ from past circumstances, there may be deviations from this trend. But, for the most part, it is not unreasonable to use an earnings growth rate of 6% per annum when normalizing the Dow's earnings.

Once you've decided upon an appropriate earnings growth rate, you simply take one plus your assumed growth rate and raise it to a power equal to the distance between the current year and the year you are adjusting. This number is the adjustment factor.

If you were calculating a 15-year normalized P/E ratio, you would use the following fifteen "adjustment factors": 1.06, 1.12, 1.19, 1.26, 1.34, 1.42, 1.50, 1.59, 1.69, 1.79, 1.90, 2.01, 2.13, 2.26, and 2.40.

You start by multiplying the first adjustment factor [1.06] by the most recent year's earnings. Then, you multiply the second adjustment factor by the second most recent year's earnings and so on.

Finally, you add up your adjusted earnings (i.e., the products of the operations you just performed) and you divide by the number of years used in your normalization process. When calculating a 15-year normalized P/E ratio, you would divide the sum of your adjusted earnings by 15. It's really that simple.

For instance, if you were calculating normalized earnings for 1995, you would multiply 1994's EPS by 1.06, 1993's by 1.12, 1992's by 1.19, 1991's by 1.26 and so on.

Please note that I am not suggesting you ever use this normalization process on an individual stock. In fact, I think that would be a rather ridiculous approach that would generally prove inferior to a careful consideration of the known facts regarding that particular enterprise and its future prospects.

I am, however, suggesting that when applied to a diversified group of very large American businesses (like the Dow) this normalization process will provide insights into whether earnings (and earnings growth rates) are sustainable.

** The Process in Pictures**

Since normalized earnings use actual [past] earnings and a growth rate of 6%, a long-term graph of the Dow's normalized earnings looks a lot like a graph of anything that compounds:

That's a boring and rather meaningless graph. I included it simply to show you that it mirrors what you'd expect to see in terms of actual earnings, if you were looking at the very, very long-term. Since investors normally have a close-up view of earnings, the graph doesn't quite look like this. But, over time, it tends to approximate this graph – which is simply the image of a perpetual compounding machine.

Using the Dow's normalized earnings, we can draw a much more interesting graph. I already told you that "normalized" earnings are really equivalent to "expected" earnings from a long-term perspective. Normally, when we talk about earnings expectations, we are talking about short-term expectations. But, that doesn't have to be the case.

In fact, these normalized earnings are well suited to making future projections, because the current year's earnings are not included in the calculation. For instance, if we were calculating expected earnings for 1995, we would start by using the earnings per share number for 1994.

When looking at historical normalized earnings data, you need to remember that we can always draw the "expected earnings" line ahead of time.

**Actual Earnings vs. Normalized Earnings**

The difference between actual earnings and normalized (or "expected") earnings is one of the most fascinating statistics in this little study.

When the difference between actual earnings and normalized earnings is positive (i.e., the green line is above zero), logic suggests future earnings will need to revert to the mean in some way. As a result, it is reasonable to expect future annual earnings growth will fall below 6% at some point to "give back" the unsustainable earnings growth of past years.

Conversely, when the difference is negative, we would expect future earnings growth will be greater than 6%, because current earnings need to make up lost ground to return to our long-term 6% earnings growth assumption.

Here is a graph of the percentage difference between the Dow's actual earnings and the Dow's 15-year normalized earnings for each year from 1935-2005:

* click to enlarge*