Join our community of smart investors

When good times lead to bad

Good medium-term returns tend to lead to stock markets falling. Investors should therefore worry about US equities
October 23, 2018

Despite their recent wobble, US shares have enjoyed a fantastic run: the S&P 500 has risen 60 per cent in the past five years. This fact alone should worry investors a little.

I say so simply because history warns us that good five-year periods tend more often than not to lead to bad ones in the following five years, and vice versa. If we take five-year periods ending in December 2017 (so we ignore overlapping periods), the correlation between changes in the S&P 500 adjusted for inflation in one period and those in the next has been minus 0.26 since data began in 1870. This pattern has been especially obvious in recent years. The market did well in the five years to 1997; fell in the next five; rose between 2002 and 2007; fell between 2007 and 2012; but rose strongly thereafter.

We see a similar pattern in the UK. The Bank of England has share price data going back to 1700. If we look at non-overlapping five-year periods (from 1702, 1707 and so on) the correlation between price changes in one period and those the next has been minus 0.15.

This contradicts efficient market theory. This tells us that past returns should not help predict future ones, so our correlations should be zero.

Why might this be? In a new book, A Crisis of Beliefs, Nicola Gennaioli and Andrei Shleifer suggest a reason. Investors, they say, extrapolate recent market conditions into the future: “they are excessively optimistic about the future in good times and excessively pessimistic in bad times”. This means that in good times prices rise too far and in bad ones they fall too much. This gives us a negative correlation between price changes in one five-year period and in the next.

This, they say, happens because investors are prone to a common error of judgment, which the Nobel laureate Daniel Kahneman has called the representativeness heuristic.

Imagine you are asked to guess the occupation of a middle-aged man who, you are told, is quiet, bespectacled and tidy-minded. Most people guess he’s more likely to be a librarian than a farmer because such characteristics seem more representative of librarians than of farmers. In fact, though, he’s more likely to be a farmer because there are more farmers than librarians, and these features aren’t sufficiently predictive of one’s occupation to change those odds.

In a similar fashion, we infer that good times now are representative of good ones to come, and so overestimate their likelihood. Professors Gennaioli and Shleifer say this thinking contributed to the 2008 crash. Years of stability led bankers to think that the future would be stable which emboldened them to take too much risk. The same is true for investors and share prices.

Now, there is a caveat here. The negative correlations between returns in one five-year period and the next are not statistically significant at conventional levels; the p-values are 18 per cent for the US and 24 per cent for the UK. This means we cannot rule out the possibility that they are in fact zero and our samples have just been unrepresentative; these correlations are not statistically significant from zero.

I’m not sure how important this is. Investors and academics have different thresholds for beliefs. The investor who waits for academics’ levels of statistical significance will do very little, except hold tracker funds and perhaps momentum and defensive ones, and will have no strong views about the future. For investors, 70-30 bets are often good enough: they have to be.

Also, there’s a reason why the correlations are small. It’s that sometimes, macroeconomic conditions have justified long bull markets. In the 50s and 60s, for example, share prices continued to rise because investors learned that in the golden age of capitalism high profits and low inflation could last a long time (although not forever as they discovered in the 70s). Similarly, in the 80s and 90s a rising share of profits in GDP and long fall in bond yields produced a protracted bull market in equities.

Which poses the question: will macroeconomic and political circumstances really remain favourable for shares? Will the next five years really give us surprisingly low bond yields and rising profits as the last five have? If they don’t, then the tendency for shares to fall because investors have overreacted to good times will reassert itself. In this sense, the medium-term outlook for the market is unusually risky.