Micro-efficient, macro-inefficient

One of the eternal challenges for behavioural finance fans is that it seems to explain the individual behaviour of investors quite well but is unhelpful in predicting the overall market. A recent study Enoch Cheng and Clemens Struck looked at 50 different indicators that have been identified as linked to behavioural biases. For example, momentum signals are linked to investor herding, while the volatility of stocks is linked to the share of noise traders in a market. Similarly, valuation metrics like the PE-ratio are linked to a preference of investors for glamour stocks, etc. Yet, if we use all of these indicators and try to predict stock market returns over the next couple of weeks, we can typically explain only about 1% of the variation in stock market returns. For some indicators like valuation signals, the predictability of stock market returns gets better if we extend the investment horizon to years and decades, but even then the explanatory power remains relatively low.

However, if one measures the correlation between different indicators linked to behavioural biases, something interesting can be observed. The chart below shows the S&P 500 together with the average correlation between the 50 indicators in the study of Cheng and Struck.

Average correlation between behavioural indicators

Source: Cheng and Struck (2020)

Most of the time the correlation between these indicators is pretty low, meaning that some indicators point to a rising stock market inducing some investors to buy shares, while other indicators point to a falling stock market, inducing other investors to sell. In essence, the different behavioural biases of individual investors cancel each other out.

However, from time to time, the correlation between these different behavioural indicators increases. In these cases, behavioural biases of different investors no longer cancel each other out but start to reinforce each other. Markets that have been efficient at the micro-level for a long time, suddenly become macro-inefficient in the sense that a large number of stocks lose touch with reality at the same time. 

The most extreme case of this collective behavioural delusion setting in was the tech bubble of the late 1990s when the correlation between all these behavioural indicators went through the roof. Whether you looked at price momentum or were attracted to glamour stocks, everything pointed in the same direction, namely to buy the high-flying .com-stocks of the day. It took until 2003 to reduce the correlation back to levels that were so low that one could speak of a market that wasn’t biased anymore.

Since the burst of the tech bubble in 2000, we have seen the occasional spikes in correlation, all of which happened during significant market corrections or the 2008 financial crisis. In these instances, behavioural biases again reinforced each other as everybody was rushing to safety. At the height of each of these corrections and crises, markets were again macro-inefficient and too cheap. This is why it pays to buy at the height of a crisis when everybody is running for the hills. This is quite simply when markets are typically divorced from reality.

This brings me to the event of March 2020 which can be seen at the very right-hand side of the chart. Yes, in the Covid-19 crisis, behavioural indicators become the most correlated since 2000 indicating that at the March bottom markets were completely divorced from reality. Since then. Markets have recovered somewhat, but it seems unlikely that behavioural biases have again become so uncorrelated that markets are efficient again. Instead, there is likely still some net bias in the markets given that we all are just on our way out of lockdown and still glued to the news flow about Covid treatments and vaccines. Hence, there are still some good investment opportunities out there.