Uncloaking Cape: A new Look At An Old Valuation Ratio

31 Mar 2016


Professor Robert Shiller’s Cyclically Adjusted Price-Earnings (CAPE) Ratio has proven treturns in the United States and some global markets. In recent years, though, it has been criticized for being overly pessimistic about the prospects for equity returns, its lack of robustness to distortions in corporate earnings, and for overstating the predictability of returns at long horizons on account of overlapping observations and endogeneity, particularly when estimated using Ordinary Least Squares (OLS).

In this paper, we explore various definitions of CAPE, present new construction techniques that make it robust to a wide range of accounting and index construction biases as well as to changing fundamentals in equity markets, and evaluate its forecasts using econometric methods that account for endogeneity and overlapping observations. We show that most of these enhancements have a minimal impact on CAPE for the US equity market, but can prove useful in smaller markets and in markets that have experienced significant dislocations. We also show that certain accounting flow variables such as cash flow and revenues can be useful supplements to earnings and cyclically adjusted earnings.


Since its publication in Campbell and Shiller (1998), CAPE (an acronym for Cyclically Adjusted P/E), which is defined to be the ratio of an equity index’s price level to the ten year average of its real (i.e. inflation adjusted) earnings, has become a popular measure of equity market valuation. CAPE is not a new idea: it has its roots in Graham and Dodd’s Security Analysis (1934), in which the authors urge investors not to focus on a company’s current earnings, but instead to average its earnings over a business cycle of seven to ten years. But their suggestion did not become standard practice, and Campbell and Shiller’s (1988) application of the averaging step to the entire equity market was novel, as was their introduction of an inflation adjustment to make earnings more comparable over a decade.

It was but a short step from using CAPE to forecast long horizon returns to using it as a market timing tool, and CAPE soon attracted criticism. If periods of overvaluation are identified by their elevated level of CAPE relative to its long term average from January 1881 to the end of the prior month, the market has appeared undervalued in only 16 out of the 336 months from January 1987 to December 2014. Furthermore, these periods of undervaluation were concentrated in two clusters: one after the crash of 1987 (9 out of 13 months following October 1987) and the other following the collapse of Lehman Brothers in September 2008 (7 consecutive months starting in November 2008).

A variety of enhancements have been proposed to rectify CAPE’s shortcomings. Siegel (2013) suggests the use of per-share NIPA earnings in place of as-reported earnings, which can be distorted by one-time write-downs, and correcting for trend changes in per-share earnings growth. In an interview with Alan Abelson (1998), John Hussman proposes using the ratio of price to ten-year peak earnings. Asness (2012) explores the use of the median in place of the mean to make CAPE robust to outliers. Unfortunately, none of these modifications has proven to substantially improve upon the ability of the basic CAPE methodology to describe and forecast returns, which begs the question of whether CAPE can be improved upon at all. We believe it can, and in this article, we constructively show how, by addressing the following three questions:

Download to read more