Human Error Is Predictable

How expectation errors explain factor investing

Every earnings season, Bloomberg terminals come to life with reports of firms that have missed or beaten expectations, sending the diving or soaring stocks as a result. Equity analysts spend these periods furiously updating their models to correct for the new information.

Despite this seasonal rhythm being such an obvious fact of life, serious financial academics have spent remarkably little time studying expectation errors. A few years back, the economist H. Woody Brock observed that the words “mistake” and “forecasting error” don’t appear in the index of any textbooks of modern financial theory.

But three important new studies by the powerhouse behavioral finance team of Pedro Bordalo, Nicola Gennaioli, Rafael La Porta, and Andrei Shleifer look set to finally put expectation errors on the map as a major driver of asset prices.

The researchers find that analysts’ forecasting errors and revisions explain “a large chunk” of factor returns and that the volatility of aggregate expectations can quantitatively explain Shiller’s excess volatility puzzle.

Their research shows that factors like value, investment, size, profitability, and momentum work because they predict surprises relative to expectations. “Average spreads materialize because the realized earnings growth of stocks in the portfolio’s short arm systematically disappoints compared to that of stocks in its long arm,” they write. The factors, therefore, aren’t proxies for exotic risk. They are proxies for non-rational beliefs about future growth.

The papers rely on analyst forecasts of future earnings growth and then compare actual results to historical expectations. The researchers argue for the “preeminence of forecast errors” in explaining how the equity factors predict returns. “Analysts and the market appear to hold systematically bullish expectations about firms in the short portfolios, compared to firms in the long portfolios, and the former do worse on average because that relative optimism systematically decreases,” they write.

Figure 1. Expected Value Returns (Red) vs. 1Y Log Value Returns (Blue) (High-Minus-Low)

Source: Bordalo et al.

It turns out that the relative spread between the analyst forecasts for the long and short arms can predict the future returns of the factors. For example, when analysts are extremely bullish about the growth rates on high-investment firms relative to low-investment firms, the investment factor will have abnormally high returns, and vice versa. The scholars argue that 60% of the variation in factor return spreads can be accounted for by expectations.

Equity returns—and future growth rates—may be highly unpredictable. But there is “systematic predictability of forecast revisions and errors,” according to the authors. Our forecasts, in other words, are predictably wrong. And betting on factors is a way of betting on systematic patterns in how humans tend to err in forecasting the future. Contrary to rational expectations theory, humans might actually be predictably irrational, especially when it comes to making predictions about the future.