A recent article in the Wall Street Journal offers an example of a common criticism encountered by behavioral economists. The authors suggest that theories in behavioral economics are too abstract and malleable to offer clear predictions about market behavior and that decisions observed in laboratory environments do not necessarily generalize to behavior in markets. This “straw man” line of attack, however, reflects a misunderstanding of the research behavioral economists actually do.
But what does that look like? A good behavioral economist will typically start by identifying real market outcomes that appear inconsistent with traditional economic theory. Laboratory experiments then serve as a useful tool for reducing the number of theories that could potentially explain observed violations of economic predictions (field experiments are also becoming increasingly common). And, just as in any other subspecialty in economics, the ultimate test of theory is its predictive accuracy in the field. The belief that behavioral economists devote all their efforts to devising contrived theories that rely on esoteric laboratory experiments is simply false, and is often the source of unproductive and misguided arguments targeted at the field.
Consider research on the market for flood insurance in the United States. Many experts and policymakers are wrestling with how to increase take-up rates for flood insurance while still maintaining premiums that reflect the costs of underlying disaster risks. Indeed, even when premiums are subsidized at discounted rates, as is the case for many properties located in the 100-year floodplain, homeowners still often refuse coverage. Taken at face value, these decisions seem to violate economic theory, which predicts that rational consumers will always fully insure against risk if they are offered a fair premium.
So what prevents these homeowners from purchasing insurance for floods or other natural disasters? The possible answers to this question are numerous. For example, homeowners could be constrained financially, by income or liquidity; they might procrastinate in making the choice to purchase coverage; or they could be overly optimistic about the chances that they will experience a flood. (Here are some other theories.) Narrowing down the number of explanations using only the existing market data on insurance purchases is difficult—if not impossible.
An alternative approach is to survey homeowners about their insurance purchase decisions. Surveys, however, have well-documented limitations (see the November 7, 2016 election polls, for instance). Suppose we were to distribute a survey to homeowners to determine whether they are overly optimistic about flood risks. We could ask questions such as, “Do you have insurance coverage?” and “What do you believe is the annual likelihood that your house will be flooded?” There are at least two serious limitations inherent in this strategy. First, even if we found that homeowners underestimate their risk, a simple survey cannot prove that this underestimation causes homeowners to refuse insurance coverage. At best, a survey might prove that homeowners’ risk perceptions are correlated with their insurance coverage—but this could equally be interpreted as evidence that buying insurance causes people to report higher risk estimates to justify their purchases. Second, there is little reason to have faith in peoples’ ability to report genuine risk perceptions given widespread difficulty in understanding probabilities. In a recent survey of homeowners in New York City, many respondents reported that they had over a 10 percent probability of experiencing a serious flood in a given year, an order of magnitude higher than any homeowner’s actual risk.
This is where laboratory experiments, in which subjects make decisions with real financial incentives, can come in handy and why they have become so common in behavioral economics. Experiments designed to test decisionmaking under risk, such as those reported by Kahneman and Tversky, place subjects in a very simple setting in which they make binary choices, indicating which lottery they would prefer to play (e.g., “choose between A: $6,000 with probability 0.45; or B: $3,000 with probability 0.90”). Insurance choices represent actual market decisions involving risk, but these choices typically come with a much greater number of considerations, such as statutory requirements, household budgets, and factors affecting insurance supply. Laboratory experiments are therefore a more viable way to test fundamental theories of decisionmaking under risk. Moreover, if people display irrationality in a simple, transparent laboratory environment, why should we expect them to suddenly become more rational in real economic environments that involve higher degrees of complexity?
Although laboratories serve as a useful testbed for understanding behavior, they are not the endpoint. It may be the case that behaviors that occur in laboratory environments do not predict decisions made by actual market participants. To return to the flood insurance example noted above, the theory developed by Kahneman and Tversky, based on their observations from laboratory experiments, fails to predict low take-up of flood insurance. Prospect Theory, as it is called, instead maintains that people overweight low-probability events, thus making them more likely to insure against low-probability flooding events. Yet the failure of a specific theory to predict market outcomes does not imply that we should abandon behavioral economics entirely or conclude that laboratory experiments are inherently misleading. Instead, we should try to better understand what “omitted variables” exist that cause behavior to change across decision environments. More recent research has discovered that whether people underweight or overweight low-probability risks depends on how they learn about the risk: If people learn about risks from descriptive accounts, as in Kahneman and Tversky’s experiments, then they overweight low probabilities—but if they learn about them from personal experience, as is probably true for most homeowners, then they underweight.
What’s the upshot of all this? Broadly speaking, behavioral economics is a process of discovering how decision environments interact with human psychology to produce economic choices—choices that can have significant impacts on people’s lives, as in the case of whether or not to purchase flood insurance. Such understanding is crucial if we wish to design policies that improve market outcomes—and human well-being.