Study Suggests Irrational Investing Behavior Across Demographics

In a study by the TIAA Institute asking subjects to make decisions resembling portfolio allocations, researchers found those with more financial education made more rational decisions.

A newly published white paper from the TIAA Institute dives into the topic of “subjective expected utility theory,” or “SEU,” exploring how this classical model of behavior displayed in the face of uncertainty applies to the investing landscape.  

The paper was penned by Federico Echenique and Kota Saito, of the California Institute of Technology, and Taisuke Imai, of the Ludwig-Maximilians-University Munich. According to the trio, subjective expected utility theory “is the workhorse model of decision-making under uncertainty, and economists assume routinely that agents behave according to its precepts.”

The study “evaluates SEU’s empirical validity in experimental settings in which subjects were asked to make decisions resembling portfolio allocations.” As the team explains, the study design allowed for quantification of how far each individual subject’s choices were from what the rational SEU theory would suggest, as well as a comparison of results across subject populations.

In what is perhaps not a surprising finding, the researchers found the vast majority of subjects analyzed do not conform with SEU theory—meaning they do not make their decisions about what do to in uncertain circumstances according to a rational analysis that ranks the relative utility of the various potential choices and outcomes. The researchers also consider a related theory known as “maximum utility theory,” finding similarly that study subjects do not seem to behave according to its rational tenants.

More interesting for the retirement plan industry is the fact that subjects exhibited similar responses to uncertainty generated by simulated low- and high-volatility stock prices. Further, the researchers found the degree of compliance with the SEU theory is very similar for undergraduate students, younger adults (20 to 39), middle-age adults (40 to 59) and older subjects (60 or older). This finding suggests that individuals do not naturally become more financially savvy as they age or as they experience cyclical economic conditions. Rather, they require educational support to behave rationally.

According to the research team, generally speaking, subjects with more education behave significantly closer to the rational theory than less-educated subjects. In particular, subjects with a high degree of financial education and literacy behave significantly closer to the rational theory, based on one of two financial literacy measures tested. However, there was no significant relation found with the second measure.

Bringing their results into the practical realm, the researchers give the example of a retirement plan advising its participants on how to choose a portfolio allocation.

“Such recommendations are meant to capture agents’ best interests, and seek to assess how they trade off volatility and returns,” the paper states. “As a matter of fact, the recommendations should be based on a model of agents’ preferences—otherwise it is not clear that they are in the agents’ best interest. Economic models (SEU and MEU in our case) which better capture individuals’ preferences would lead to better recommendations to the individuals in terms of their welfare. To know what is best for a client, the plan would need a model that captures the client’s welfare. Economists have long understood welfare (or preference) to be revealed through choices—giving rise to the term ‘revealed preference.’”

According to the researchers, it is possible that new models of ambiguity aversion could do a better job of accounting for the experimental data.

“We are restricted to MEU because it is the only model for which there exists nonparametric tests of the kind that we use in our paper; it is also arguably the best known, and most widely applied, model in the ambiguity literature,” the paper explains. “The testable implications of other models of ambiguity-averse choice is an interesting direction for future research.”

The full research report can be downloaded here.