The cyclically-adjusted price earnings ratio (CAPE), also known as Shiller’s P/E, continues to fascinate stock market watchers. The ratio, developed by the Nobel laureate Professor Robert Shiller of Yale University, and further popularised by market commentators such as John Authers, Andrew Smithers and Russell Napier, quantifies the relationship between the price of the stock market and its average 10 year real earnings. The CAPE has become widely used to gauge how over or undervalued a market is relative to its own history. It is often cited as a more useful metric than the conventional P/E ratio, based on a single year’s earnings, as the longer data essentially smooth out the volatilities over one business cycle.

Exhibit 1

Source: Dr Robert Shiller, Yale University

According to data published by Dr Shiller[1], current CAPE at 26.5x is 61%, or 1.5 standard deviations, above its long-term mean of 16.6x, which does appear overvalued relative to its own history. The last time it reached this height was in October 2007, where it peaked at 27.5x before the market’s subsequent crash.

Cerno Capital studied this subject in some detail back in 2012, with a particular focus on the US market. We concluded that there is an inverse relationship between CAPE levels and future stock market returns, but that this is only significant over long horizons of 10 years and beyond. Comparing CAPE against forward market returns for short (1Y) and long (20Y) time horizons in Exhibit 2, the one year returns produce much more volatility and noise, while a clear trend is observable in the twenty year returns.

Exhibit 2

Source: Cerno Capital

The scatter diagrams in Exhibit 3 illustrate this as well. As the time horizon stretches outward, the negative correlation becomes more evident with a tightening of returns dispersions.

Exhibit 3

Source: Cerno Capital

To test the level of significance, we mapped out the correlations between CAPE and future returns for different time horizons. In the short to medium term (1-5 years) correlations lays between a weak -0.15 to -0.35. The relationship strengthens as the years increase. Long term returns of over 10 years become more interesting as correlations rise to -0.5 and eventually tail off at -0.66 by the 20^th year, a far more significant reading.

Exhibit 4

Source: Cerno Capital

So what does the current level tell us? We assessed the probability of losing capital associated with different levels of CAPE using past 100 years data. Buying into the stock market at current levels of 26x makes the future look rather bleak with over 60% chance of losing money over the next 10 years.

Exhibit 5: Probability of loss at different levels of CAPE over different time horizons

Source: Cerno Capital

The next biggest question, however, is much more difficult to answer.  And that is when will it happen?

The major caveat with CAPE is the fact that it has very limited, if any, use as a market timing tool.  Previously, it correctly signalled that the US market was overpriced prior to the dotcom bubble, and levels stayed stubbornly high for the first half of the previous decade leading up to the 2008 financial crisis. However, had investors exited too early pre-2002 or stayed out of the market between 2002-2007, they would have also missed out on fine opportunities to make money. It is difficult to foresee how long an over-exuberant market will continue its upward trajectory or how long bearish sentiments will last before mean reversion finally kicks in.

There has been increasing debate in the financial world in recent years scrutinising the statistical details of CAPE, making adjustment for changes in tax regimes and/or accounting practices. Critics of the metric argue that CAPE has not signalled cheapness in a long time, and often cite earnings as the main source for error (e.g. reported earnings are too high). Jeremy Siegel, a Professor at Wharton Business School, proposed alternatives such as using National Income and Product Accounts (NIPA) for the US series. The exhibit below shows that NIPA CAPE had always been less bearish than the standard method, and in particular the past few years display the largest discrepancy between the two measures. Indeed, on this measure, the valuation does not appear stretched.

Exhibit 6

Source: GMO, Shiller, WorldBeta

Advocates of CAPE such as James Montier from GMO on the other hand, rebuffed this argument in his paper ‘A CAPE Crusader ─ A Defence Against the Dark Arts’ earlier this year. He constructed a long-term series of 7 year predicted returns by mean reverting CAPE towards the average over the next 7 years, adding a constant growth factor (6%) to reflect growth and income. Comparing the actual realised returns with the predicted returns side by side (as shown below), he demonstrated that CAPE has done a reasonable job of capturing realised returns over the long run, countering that if anything, it over- rather than underestimates returns.

Exhibit 7

Source: GMO, Global Financial Data

Once again, this brings us to the conclusion that CAPE is still a sensible valuation metric, but only for long horizons. There are reasons to suggest that the UK and Europe, whose performances have lagged of late, will trail the trajectory of the US in terms of valuations. However, for other markets such as Japan, due to the lack of long-term data, and because valuations are wildly distorted by the 1980s bubble where levels reached over 100x, it makes no sense to suggest that at current levels of 21x the CAPE will mean revert back to the 54x average.

Exhibit 8

Source: MSCI

Our view remains that CAPE is a useful indicator for investors who are truly long term where the long term is defined as being periods of over 10 years or more. Over that time frame, US equity market equity returns could well disappoint. We find the inherent probabilities sufficiently convincing to have already trimmed our broad S&P exposures. On other markets, CAPE cannot be relied on to any degree as the data series are not long enough to store any trust in the average ratings.