[Originally published on 07/10/2014]
Almost all professional investors look at the P/E ratio of the stock market, but rather than smoothing the earnings as we do, they look at either the past 12 months of earnings or the estimates of earnings over the next year. We can’t believe how many sophisticated investors use the P/E ratio based on the past year or following year. The reason we attempt to smooth the earnings is to dampen the volatility that one year of earnings has had over the past decades. This will make it easier to make longer term decisions.
An example of the volatility of one year of earnings over the past 14 years would be that “reported earnings” (or GAAP—Generally Accepted Accounting Principles) were $50.00 in 2000, $24.69 in 2001 and came back to $48.74 in 2004. The earnings “estimates” over that period of time also had tremendous volatility and it boiled down to almost a guess. So how could you possibly base an intelligent investment decision during that time? Also earnings were $66.18 in 2007, $14.88 in 2008, and $50.97 in 2009. Earnings are among the most mean-reverting statistics in investing, and therefore there has to be a better way of determining the “true” earnings. Earnings always return to the trend and the long term trend of earnings growth over time has consistently averaged about 6% despite the year-to-year volatility.
The idea of smoothing earnings over a period of more than a single year was first proposed by Graham and Dodd in their classic book Security Analysis published in the early 1930’s. Cyclically—smoothed earnings have proven to be a much more accurate predictor of long-term market returns than any method that uses earnings over the past year or following year. In fact, we used this method in late 1999 when we published a report forecasting that the normalized S&P 500 would be about 1260 ten years from then. That was significantly below the then prevailing price. It was a projection we made near the peak of the dot-com bubble. Almost everybody thought the forecast was crazy although it eventually proved to be highly accurate.
What we do to smooth reported S&P 500 earnings over a longer period of time is to calculate an average of the past 9 years (which usually includes at least one and often two business cycles) . We then take that average of the 9 years and place it in the 5th year, the exact mid-point of the 9 years. We then increase that number by 6% per year to the last of the nine-year period and use that number as our normalized earnings for the most current period. Using that method our normalized earnings for the 12 months ended March 2014 is $90. Then we divide the S&P 500 level of 1970 by the normalized earnings of $89 and come up with a PE of about 22.. This represents a very expensive level for the stock market, and was exceeded only in 1929, 2000 and 2007.
There are others in the investment community that also normalizes earnings such as Robert Shiller, John Hussman, and Ned Davis. They each actually use a longer time frame than we do, but each of them comes to the same conclusion that the stock market with normalized earnings is expensive and overvalued.