The recency effect—that the most recent observations have the largest impact on an individual’s memory and, consequently, on perception—is a well-documented cognitive bias. This bias could impact investment behavior if individuals focus only on the most recent returns and project them into the future. Such behavior may lead investors to experience a reduction in their risk tolerance (which, in turn, can lead to selling) after a bear market, when valuations are lower and expected returns are higher. Conversely, recency may lead investors to experience an increase in their risk tolerance (which, in turn, can lead to buying) after a bull market, when valuations are now higher and expected returns are lower.

The recency effect nudges investor behavior in a direction contradictory to economic theory, which states that relative risk aversion is a function of increasing wealth (the marginal utility of wealth declines as wealth increases). Strong market returns would increase investor wealth, and thus we should see a reduction in investor risk tolerance.

Rui Yao and Angela Curl—authors of a 2011 study, Do Market Returns Influence Risk Tolerance? Evidence from Panel Data, which appeared in the Journal of Family and Economic Issues—hypothesized that the recency effect would dominate rational economic behavior. They posited that risk aversion is negatively related to recent market returns (or, in other words, that risk tolerance is positively related to recent market returns).

Their study used data from the 1992, 1998, 2000, 2002 and 2006 interview waves of the Health and Retirement Study (HRS), an ongoing biannual study conducted by the University of Michigan and funded through the National Institute of Aging. The target population for the HRS is noninstitutionalized men and women, born from 1931 to 1941, living in the contiguous United States. Based on responses to a set of income gamble questions, researchers assigned participants to a risk tolerance level for each wave: most risk tolerant, second-most risk tolerant, third-most risk tolerant and least risk tolerant. Stock market performance was measured as a continuous variable using the S&P 500 Index’s trailing 12-month returns prior to each interview. The following is a summary of the authors’ findings:

• Consistent with the recency theory and their hypothesis, there was a significant positive linear relationship between S&P 500 returns and respondent risk tolerance.

• Controlling for time and other independent variables, a one percentage point increase in market returns increased the probability of taking substantial or high risk by 1%. A one-standard-deviation increase in S&P 500 returns increased the likelihood of taking substantial or high risk by 15.7%.

• When the stock market is falling, average monthly investor risk tolerance scores are strongly correlated with changes in the S&P 500. However, when stock prices start to rise, changes in average risk tolerance seem to be largely uncorrelated with the market.

Yao and Curl also found that:

• Each additional year of age above the sample mean decreased the likelihood of taking some risks by 2%—consistent with theory and prior research showing that the likelihood of being in the high-risk or some-risk groups decreases as people age.

• Men tended to be more risk tolerant than women. This finding, which might be called the “testosterone factor,” is consistent with the findings of Brad Barber and Terrance Odean in the study Boys Will Be Boys: Gender, Overconfidence, and Common Stock Investment.

• Higher educational attainment was consistently predictive of higher levels of risk tolerance.

• Investors with greater financial assets reported lower levels of risk tolerance. This is consistent with the theory of declining marginal utility of wealth.

The key finding is in direct conflict with rational economic theory. When market return becomes negative, wealth decreases. Therefore, risk aversion should decrease (and risk tolerance should increase). But Yao and Curl’s analysis found that risk tolerance fluctuated positively with market returns. While the loss of money, combined with loss aversion, contributes to an increase in risk aversion during a bear market decline, gains during a bull market lead to the well-documented “house money” effect and a decrease in risk aversion.

The authors concluded that investors don’t behave according to rational economic model assumptions, and that “such changes in risk tolerance in response to market returns may be an indication that investors, and possibly their financial advisors, overestimate their ability to understand risk and assess individual risk tolerance.”

These findings suggest that individuals invest more after periods when market returns are high and withdraw partially or even completely from the market after periods when returns have been poor. Yao and Curl reached the conclusion that their findings support “the projection bias hypothesis and confirms the recency effect.” What’s more, their findings on investor behavior are consistent with those from the field of behavioral finance.

For example, Richard Thaler and Eric Johnson, authors of the 1990 study Gambling with House Money and Trying to Break Even: The Effect of Prior Outcomes on Risky Choice, found that individuals experience less dissatisfaction from losses after a prior gain and greater dissatisfaction after a prior loss. Thus, risk aversion is time-varying and dependent on prior outcomes.

Yao and Curl’s findings are also consistent with those of Robin Greenwood and Andrei Shleifer, authors of a 2014 study, Expectations of Returns and Expected Returns. They were able to document a strong negative correlation between investor expectations of stock returns and recent returns for the S&P 500—investors change their expectations of the reward from taking risk based on recent changes in stock market returns.

The financial crisis of 2008 provided a good example of how recency impacts investor risk tolerance. During the crisis, individual investors pulled out hundreds of billions of dollars from the equity market. The result was that, by 2010, portfolio allocations to risky assets had declined to their lowest level for people under the age of 35 in the history of the Survey of Consumer Finances.

A more recent example can be found by examining the returns from emerging markets and investor flows. From September 2014 through September 2015, the MSCI Emerging Market Index lost more than 23%. Investment Company Institute data shows that beginning in July 2015, emerging-market funds experienced net withdrawals in every single month. For the period from July 2015 through January 2016, total net withdrawals exceeded $13 billion.

Next week, we’ll examine some additional support in the research for Yao and Curl’s findings, as well as explore the relationship between the recency effect and loss aversion and investor overconfidence.