Oliver D. Bunny and Robert J. Shiller May 9, 2014 H/T Meb Faber
Changing Times, Changing Values: A Historical Analysis of Sectors within the US Stock Market 1872-2013
We construct a price, dividend, and earnings series for the Industrials sector, the Utilities sector, and the Railroads sector from the beginning of the 1870s until the beginning of the year 2013 from primary sources. To infer about mispricings in the sector markets over more than a century, we investigate the forecasting power of the Cyclically Adjusted Price-Earnings (CAPE) ratio for these sectors. With regard to the CAPE ratio, which has originally been devised and employed by Campbell and Shiller (1988, 1998, 2001) as well as Shiller (2005), we define a methodological improvement to this ratio to not only be robust to inflationary changes, but also to changes in corporate payout policy. We then update the original evidence from Campbell and Shiller (1998, 2001) of the return predictability of the CAPE ratio for the overall stock market and furthermore extend this evidence to the three aforementioned sectors individually. Whereas this part of our analysis focuses on each sector of the US economy in isolation, we subsequently construct an indicator from the CAPE ratio that enables us to perform valuation comparisons across sectors. In addition to establishing the prediction of subsequent return differences based on differences in the CAPE-based valuation indicator, we also suggest a hypothetical, historical, and simple value investment strategy that rotates between the three sectors based on the valuation signals derived from the CAPE-based indicator, generating slightly more than 1:09% annualized, inflation-adjusted excess total return over the market benchmark during a period of nearly 110 years.
The CAPE Ratio and Price Fluctuations
There is a substantial literature on the value factor in predicting returns, but this literature has not yet exploited the historically-longest available stock price series, sector indices. Studying the very-long-term value effect with these series allows us to test the value proposition out of its original sample, where out of sample means extending the sample back in time, as well as forward. The long time series enables us to evaluate a very effective measure of mispricing relative to fundamentals, the Cyclically Adjusted Price-Earnings (CAPE) ratio, originally shown to be highly effective in predicting returns in the aggregate US stock market by Campbell and Shiller (1988), who did not, however, look at sector data. Sector data, in contrast to firm-level data, is especially interesting because the sectors represent major, almost macroeconomic, activities, and much analysis, accordingly probably also much mispricing, is sector-related. It is also interesting since we can accurately evaluate a sector’s valuation over a longer historical period than has previously been studied in the literature.
Subsequent to Campbell and Shiller (1988), the CAPE ratio has been employed by Campbell and Shiller (1998), Campbell and Shiller (2001), and Shiller (2005). In the analyses since the late 1990s, it has been defined as the ratio of real (inflation-corrected) per-share price divided by a long average of real earnings per share.3 Our present analysis modifies the definition of the CAPE ratio from Campbell and Shiller (1998), Campbell and Shiller (2001), and Shiller (2005) and considers the ratio real total return per-share price divided by a long average of real scaled earnings per share, where the scaling of earnings takes into account the transition from price return to total return numbers.
In the original analyses, it was concluded that when the average of earnings is taken over the past ten years, the ratio is a significant forecaster of long-term, ten-year stock market returns. When the ratio has been high (i.e., stocks are overpriced) prices tend to come down, not necessarily in the next year or two, but some time over the next ten years, and conversely when the ratio is low. The CAPE ratio differs from the conventional price earnings ratio in that earnings are measured over a much longer interval, and because of the longer-term view they are necessarily inflation-adjusted. The longer interval appears justified, since the earnings in any given year tend to be noisy and influenced by the business cycle. The idea of using price relative to a long average of earnings was not new with Campbell and Shiller (though, it appears that the inflation adjustment was new and important). As long ago as 1934, Benjamin Graham and David Dodd advocated a ratio that is an analogue to CAPE in their classic book Graham and Dodd (1934):
“A conservative valuation of a stock issue must bear a reasonable relation to average earnings. In addition it must be justified by whatever indications are available as to the future. This approach shifts the original point of departure, or basis of computation, from the current earnings to the average earnings, which should cover a period of not less than five years, and preferably seven to ten years.”