3

Foundational Behavioral and Economic Ideas

In the words of Richard Thaler, people behave as humans, not as econs. Individuals’ choices are highly malleable, often shaped by what would otherwise be considered irrelevant factors, particularly the behavioral phenomena discussed in Chapter 2. Behavioral economics is a response to the fact that the traditional economic model, which assumes that rational individuals behave in predictable ways, is incomplete and fails to account for important aspects of human behavior. Evidence from the behavioral sciences has demonstrated the critical role of phenomena such as biased perception, attention bias, memory bias, and complex influences of context in decision making that are not considered in traditional economic models. Behavioral economists have integrated understanding of these phenomena with tools and methods from traditional economic analysis. In this chapter we look first at traditional economic modeling and its role in economic analysis and then discuss the behavioral phenomena that have been particularly important for behavioral economics.

We emphasize that behavioral economics is only a subset of the larger field of behavioral science, which encompasses numerous disciplines that explore human behavior, including psychology, sociology, anthropology, and cognitive science. It is a vast field that has developed many insights into human behavior. Behavioral economics is a narrower field that draws on insights from behavioral science; often builds on those insights; and, most important, incorporates those insights into economic models of human behavior.

Modeling, a key tool of economic analysis, is the development of simplified representations of reality that can be used to test hypotheses.

Behavioral economists have incorporated behavioral insights into such models. In this chapter and in our discussion of findings from the six policy areas (Part II), we show how insights from behavioral disciplines have been used by economists to develop new economic models, often yielding important new policy implications.

ECONOMIC MODELING

As noted above, modeling is a fundamental tool in economic analysis. Economists use formal mathematics to portray the world in a set of equations intended to illustrate the key mechanisms in a given scenario they wish to analyze, taking into account many kinds of behavior. Behavioral economics models are not radically different from traditional economic models and do not constitute a complete rejection of those models. Instead, they build on, extend, modify—and, indeed, enrich—those models by adding behavioral features to them. Formally modeling behavior allows economists to make quantitative forecasts of the effects a particular intervention might have that can be more precise than qualitative or intuitive predictions. Quantification using mathematical models can also serve as a guide to empirical estimates of key parameters of those models; this, in turn, allows economists to make quantifiable policy recommendations, which are the main subject of this report. But not everything can be captured in formal models, and many behavioral features are captured only in simplified versions.

There is no single traditional economic model; different designs capture different behaviors. However, most traditional models share common features, such as representation of individuals (or firms) who know their own objectives or preferences, who always act to maximize those objectives to the extent possible, who accurately perceive the world around them (at least in terms of probabilities of uncertain events), and whose preferences and objectives are stable over time. Most economists understand these assumptions to be only approximations of reality but regard them as close enough to reality to support policy predictions. The types of policies most commonly addressed in traditional models are those based on incentives, such as taxes, transfers, and subsidies, and the usual assumption of traditional economists is that individuals respond to the incentives roughly in the way the model predicts. A chief difference between traditional economic models and behavioral economic models is the nature of the policy prescriptions they support, as we discuss throughout this report (also see Madrian, 2014).

Box 3-1 describes a stylized version of one traditional economic model in mathematical terms. In this example, individuals are assumed to take actions that increase their current well-being, their future well-being, or

both. The future is uncertain, but the assumption is that people have an accurate perception of the probabilities associated with possible future events. In economists’ terms, people “discount” the future—that is, they prefer well-being in the present to well-being in the future, everything else being equal—and they take current actions that maximize the discounted sum of well-being over the future. Their actions are consistent over time. This is a form of the well-known “expected utility” model, which Kahneman and Tversky argued is not the way humans behave (see Chapter 2).

There are many other types of models that extend this version of the traditional model in various ways. For example, some models incorporate the idea that individuals do not know the probabilities of future events with certainty but learn by observing events and constantly update their estimates of those probabilities over time. These are called learning models,

and they are based on the assumption that individuals update their beliefs in a mathematically accurate fashion (Wald, 1950). Other models factor in the idea that beliefs are subjective and that individuals hold expectations about future events—expectations that can be measured (De Finetti, 1970; Manski, 2004).

In addition to these models of individual behavior, there are many models that incorporate the assumption that individual behaviors are influenced by those of others. Early work by Samuelson (1947) on consumption assumed that one person in a family represents all the other members in making consumer choices. This framework was later used by Becker (1981) and many others in the design of models of decision making within families (and in models that incorporate the idea of altruism). Other models that incorporate social interactions portray individuals as members of networks who share information and resources with each other (Schelling, 1978; Manski, 1993; Brock & Durlauf, 2001; Jackson, 2014).

Although these and other models developed by traditional economists address many important aspects of behavior, they do not capture the full spectrum of behavior that behavioral economists emphasize. Because they address other behavioral phenomena, behavioral economic models lead to policy recommendations that differ from those supported by traditional models. Decades of research have explored the possibilities for conducting economic analyses that incorporate behavioral concepts. In that work, formal modeling is used in order to improve predictions about the effects of policy interventions, explore new tools for analysis, and develop more accurate assessments of outcomes than traditional economic modeling could provide (Chetty, 2015).

FIVE CORE PRINCIPLES

There is no single definition of behavioral economics because the field arose from observations of a set of behaviors that appear to deviate from what is predicted by traditional models—not from the conceptualization of a new, all-encompassing theory of behavior. The committee identified five core principles, as behavioral economists have framed them, that have been the basis for large bodies of research designed to test behavioral hypotheses and develop strategies to influence behavior: limited attention and cognition, inaccurate beliefs, present bias, reference dependence and framing, and social preferences and norms.

This is necessarily a selective review of key behavioral principles for the development of policy. The five principles we highlight here do not capture all the behaviors and findings from the research on cognitive and social psychology, and this is not the only possible way to organize these ideas. A more comprehensive review can be found in the Handbook of Behavioral Economics (Bernheim, DellaVigna, & Laibson, 2019). However, these five

have direct application to the kinds of decision making that behavioral economists study and the choices that policy makers hope to influence. We used them in analyzing the research on the six domains discussed in Part II and to assess what behavioral economics research has accomplished to date.

Behavioral economics research illustrates how each of these core principles functions in the context of decision making.

Limited Attention and Cognition

One cannot overstate how important limited attention and cognitive difficulties are for public policy. To fully understand their role it is useful to

step back and consider how economists have traditionally thought about designing public programs. Although no economist believes that people can pay perfect attention to everything and have infallible cognitive abilities, it is assumed that they are paying close enough attention and are sufficiently cognitively aware that policy recommendations can be based on the expectation that they will consistently respond in a “reasonable” way. Consequently, the key idea has been to put in place monetary incentives, such as taxes and subsidies, in order to induce individuals to change their behavior. Since it is in anyone’s interest to take this economically relevant information into account, economists simply assumed that people do so.

Such incentive-based designs are intended to work in a variety of contexts. In designing tax rates, for example, economists assume that the tax elasticity—that is, the degree to which people respond to the imposition of a tax by changing their behavior—is nonzero. This assumption is critical for estimating who really pays a tax (e.g., if the behavioral response is to shift the cost to someone else) and whether there are unanticipated costs from imposing the tax (such as decreases in tax revenue). The tax subsidies in place for individual retirement accounts, for example, are intended to encourage lower-income filers to contribute to their retirement savings. The cost-sharing provisions that are part of many health care plans, such as deductibles or copayments, are designed to help align the incentives of consumers and insurers, with the goal of reducing health care spending by encouraging more cost-efficient utilization. In the field of energy, utilities often use nonlinear pricing, in which the marginal cost of energy or water increases with additional usage, to encourage customers to conserve. In all of these cases, the implicit assumption is that people are well informed of those features—they know the tax rate, the savings subsidy, the costs and benefits of health decisions, the price of energy consumption—and they take these incentives into account in making decisions.

Counterintuitive Responses to Incentives

Behavioral economists have carried out empirical research to determine whether and how these assumptions operate and to what extent limited attention and cognitive limitations may affect people’s decisions. A first key lesson from this work is that people have a very limited understanding of tax, subsidy, and price incentives; moreover, they respond to the incentives they perceive, not necessarily to the incentives as they were designed to function. Multiple studies have shown that people often respond to the existence of incentive programs but not to their precise structure and that they miss features that are complex or not in plain sight.

For example, a study of the extent to which grocery store shoppers take into account state taxes when they make purchasing decisions showed

that when the tax was explicitly identified as part of the price, consumers purchased less (Chetty, Looney, & Kroft, 2009). This finding violates the traditional economic model, which assumes that the total price should be the deciding factor, regardless of how much of the price is the tax. Instead, it was shown that making the tax salient affected people’s behavior.

Another example concerns retirement savings (see Chapter 6). Research has shown that only a tiny fraction of low-income earners (1–2%) participated in individual retirement account savings plans despite tax incentives providing a substantial match for each dollar contributed. But when a similar subsidy was made salient—because it was newly offered through H&R Block at the time of tax filing—the response was about 10 times larger (Duflo et al., 2006). The key difference was the salience, the behavioral awareness, of the subsidy.

Other examples of the importance of salience come from research on health care and energy. For example, a health insurance company that imposed a high deductible for most care but no deductible for preventive care actually resulted in employees using less preventive care (Brot-Goldberg et al., 2017). That is, the employees responded to the change in incentive but disregarded the precise structure of the incentives, ignoring the special treatment of preventive care. Other work has shown that older individuals make many mistakes in their choice of Medicaid Plan D drug plans, though some do correct some of their mistakes over time (McFadden, 2006; Ketcham et al., 2012). Similarly, a study of energy utility consumers’ responses to nonlinear pricing (which rewarded those who used less energy) showed that they did not respond to the marginal price of electricity, but rather to the average price, which undermined the conservation goal for the pricing structure (Ito, 2014). These examples show that people often respond to the existence of incentive programs but not to their precise structure.

Behavioral economists have developed models that accommodate the kinds of behaviors documented in these studies, and there is a vast and growing literature on such instances of limited attention and cognition. This work is also identifying which information is likely to be neglected and exploring settings in which information neglect can be mitigated (e.g., Gabaix, 2019). One key to the neglect of information is that it can stem from poor financial literacy (Lusardi & Mitchell, 2011). That is, people overlook, neglect, or misunderstand information that is in principle available to them, such as on instructions for taking advantage of tax incentives or maximizing mutual fund returns. Broadly, these behaviors deviate from those assumed in the most common traditional models (described in Box 3-1) because the individuals are not taking into account all the relevant incentives that matter for the utility maximization.

In this ongoing research, an open question is the extent to which limited attention can be understood as rational inattention (Sims, 2003; Maćkowiak, Matějka, & Wiederholt, 2023). That is, does the limited

attention reflect a person’s decision not to devote more attention to an economic opportunity because of the high cost of acquiring and processing information, or does it reflect a person’s naïve inattention? In exploring the idea of rational inattention, researchers use models that allow for the possibility that people know that they have limited attention and that their time and effort is limited, and they choose which things to pay attention to for their own reasons. Providing more evidence on this margin—of the degree to which attention is allocated in accord with the perceived value of acquiring information—could be important in the future if it leads to policies that encourage people to pay more attention to opportunities or risks that are important to their well-being. For policy developers, however, both the behavioral economic models of limited attention and of rational inattention imply that processing of complex information is costly and that complex incentives, such as taxes, or the details of health insurance plans, will often be missed.

Varied Effects of Limited Attention

A second key lesson from the research on limited attention is that people’s responses vary by their level of education and income. For example, in a study of a case in which a company accidentally offered health insurance plans that were demonstrably inferior compared with others they offered, the workers who stuck with the inferior plans—and incurred hundreds of dollars in losses—were likely to be low-wage employees (Bhargava, Loewenstein, & Sydnor, 2017). Similar results were found in a study of employer matches for retirement plans: lower-wage employees were much less likely to take advantage of the benefit than their higher-wage peers (Madrian & Shea, 2001). Below-median earners (of an already low-income sample) were more than twice as responsive to simplified notices (designed to overcome lapses in attention) as those with above-median incomes (Bharghava & Manoli, 2015). Another study showed that lower-income people systematically make poorer financial choices than do higher-income people: they have lower rates of participation in asset markets and consequent larger losses of the high historical investment returns in those markets (Campbell, 2016).

These results are surprising because lower-income individuals have the most to gain from such programs and incentives and can least afford to lose sources of income. One might expect that they would scrutinize financial options particularly closely. But lower-income people do not have the benefit of the same education or networks as do higher-income people, and they more often face pressing emergencies and challenges as they try to make ends meet (e.g., Mullainathan & Shafir, 2013). Thus, they spend less time on what could be important financial information for them. This research shows that, in general, misunderstanding of taxes, subsidies, and

other complex programs is more common among people with relatively lower education and lower socioeconomic status and thus tends to increase inequality.

This research strongly suggests that simplifying communication about financial choices, such as those related to health insurance policies, taxes, subsidies, and the like—making the features that will have the most impact for consumers more salient—is an important strategy for limiting the inequitable effects of biases of limited attention and comprehension. If a policy is expressed in clear and plain language and if people are reminded at the moment they make a decision about the relevant features, they are significantly more likely to respond as the designers of the policy intended. This finding implies that the communication of public policy programs, including the work done in nudge units (see Chapter 2) that is focused on such communication, is of first-order importance. These small changes are inexpensive to implement but can yield sizable effects.

A study of communication about the Earned Income Tax Credit (EITC), which provides a large tax credit to low-income families with earnings if they file taxes, illustrates the importance of clarity in communication (see Chapter 7). Despite the large size of the benefits, about one-quarter of eligible people do not claim them. A study of possible solutions showed that when the Internal Revenue Service (IRS) sent letters to eligible households who had previously filed tax returns but had recently failed to take advantage of the EITC, the take-up of the benefit increased by more than 10 percent (Bhargava & Manoli, 2015). The letters were written in simple and clear language, illustrating the value of clarity in communication.

Similar results were found for communication about Medicare Part D plans, which cover drug costs for seniors. There is a large variety of plans designed to suit individuals with different health care needs, and the standard model assumes that individuals study the plans and choose the best option for them. Yet a clear communication to individuals about potential savings had a large impact on their plan choices: 11 percent of participants switched plans after receiving such information about the features of the plans, yielding an average savings of about $100 a year (Kling et al., 2012).

Inaccurate Beliefs

Holding inaccurate beliefs related to economic decisions—beliefs or understandings that do not objectively reflect the actual facts or situation—is a factor closely related to but distinct from cognitive and attentional issues. Inaccurate beliefs in this context does not mean occasional or unsystematic errors: as noted, above, traditional economic models assume that people do not have perfect information. Behavioral economists have focused on instances in which inaccurate beliefs are systematic and could be corrected

with access to available information. (In the language of the traditional model in Box 3-1, this amounts to incorrect probabilities p(s_t).)

One example is overoptimism: people tend to overestimate the positive aspects of their own lives, such as their likely success in a future job search effort or commitment to investing in retirement savings. For example, people who are unemployed overestimate the likelihood of finding a job compared with the actual empirical probability (e.g., Spinnewijn, 2015). Similarly, people who consider enrolling in a gym overestimate their future attendance (DellaVigna & Malmendier, 2006).

This kind of overestimation has consequences. For example, people who overestimate the likelihood that they will start saving later may save less in the present and fail to save enough to meet their retirement needs (Ganong & Noel, 2019; Gerard & Naritomi, 2021). Similarly, sales agents who overpredict future sales in their forecasts end up selling less than predicted (Huffman, Raymond, & Shvets, 2022). People who join a health club often pick the wrong membership contract because they overestimate how often they will go in the future (DellaVigna & Malmendier, 2006).

Why do people hold such overly positive beliefs in these instances? Two key factors seem at play: people like to have a positive opinion of themselves (positive ego utility), and they (consciously or not) tend to remember past events or behaviors as more positive than they were, which is an example of memory bias (see Chapter 2). For example, in one study, individuals were given (truthful) positive or negative feedback about how they did on an IQ-type test. One month later they accurately recalled the positive feedback but had largely forgotten the negative feedback (Zimmermann, 2020).

Indeed, in several of the empirical studies that have documented systematic overestimation, the overestimation was tied to biased memories. Sales agents who overestimated their future sales are also more likely to have an excessively rosy memory of their past sales (Huffman, Raymond, & Shvets, 2022). Gym club members who overestimated future attendance similarly were likely to recall more past gym visits than they actually made (Sial, Sydnor, & Taubinsky, 2023).

Similar findings of overestimation of future earnings have been found for gig workers, who typically experience more uncertainty than employees about future earnings because of variability in both hours worked and earnings per hour (Pires, 2022). The costs of vehicle use and depreciation are also an important part of net earnings. When asked to forecast a paycheck expected in the near future, gig workers overestimate it by an average of 30 percent (and even more if car depreciation is accounted for). This overestimation is related to selective memory, because gig workers also recall their past earnings as higher than they were. This study points out that this overestimation may be the reason gig workers continue in their work, for

they claim they would be working in different jobs if their wages were as low as they actually are.

It is important to point out that the research does not suggest that individuals are overoptimistic all the time. The research on the elicitation of beliefs (e.g., Dominitz & Manski, 2004) suggests nuanced patterns, and it will be important to accumulate more evidence on settings in which overoptimism is at play versus others where individuals have on average accurate beliefs, or beliefs that diverge in unsystematic ways.

A second example of inaccurate beliefs, recency bias, or availability bias, occurs when people pay more attention to recent events than older ones, especially ones that they have directly experienced or witnessed. This phenomenon is illustrated in a study of people who own property in flood-prone areas: many who had available to them the entire history of floods in their areas nevertheless responded more directly to recent experiences (Gallagher, 2014). Homeowners who were not directly affected by a flood but witnessed other nearby properties affected by a flood were significantly more likely to purchase flood insurance the next year. This pattern is especially pronounced if the county of residence of the homeowner shares a media market with the county affected by the flood, which suggests that salience of the news in the media is largely responsible for the finding. In another example, investors who, by the traditional model, should incorporate all past investment returns of companies instead overweight their own experiences (Kaustia & Knüpfer, 2008). The study showed that Finnish investors who had previous investment experience with an initial public offering were more likely to invest again if they had earned a positive return from the earlier investment, despite the fact that it would be prudent to look at the average over all past returns.

That such personal, direct salient experiences can affect people for a long time is illustrated in a study showing that early exposure to traumatic financial episodes affected later risk taking (Malmendier & Nagel, 2011). The authors found, for example, that people who lived through the Great Depression in the 1930s were less likely to invest in stocks later in life than those who did not (holding wealth constant); they found similar results in other contexts. Such findings suggest that individuals are more responsive to recent or salient events, such as events in the media and personal experiences, than to accurate long-term information.

Another example of incorrect beliefs occurs in the context of predicting the future utility associated with a particular choice: individuals tend to assume that the future utility will be too close to the one they are currently experiencing. This phenomenon is called projection bias (Loewenstein, O’Donoghue, & Rabin, 2003). Examples of this include findings that people are more likely to purchase a convertible on an unusually hot day and to cancel solar energy contracts when the weather is cloudy

(Busse et al., 2015; Liao, 2020). Another example, from China, is that on days with (temporarily) high pollution, people are more likely to purchase health insurance, even though the health insurance does not take effect until later in the future, when the temporary pollution has faded (Chang, Huang, & Wang, 2018).

An important question is how researchers measure beliefs, and thus how they conclude that in some cases they are systematically incorrect. In some instances, the beliefs are inferred from the observed behavior (e.g., Chang, Huang, & Wang, 2018), but researchers also measure beliefs directly in order to compare them with the real situation, if possible. An example is a study in which people were asked what they remember about their past health club attendance, and the results were compared with administrative records of attendance for that person (Sial, Sydnor, & Taubinsky, 2023). Such studies follow an extensive literature on belief elicitation in economics, which predates behavioral economics research and provides evidence on methods for capturing beliefs (e.g., Manski, 2004; Hurd, 2009; Delavande, 2014). Some of the lessons from this important literature are captured in the recent Handbook of Economic Expectations (Bachmann, Topa, & van der Klaauw, 2022). One of the findings in this literature is the considerable heterogeneity in beliefs even within a group of seemingly similar individuals.

In the cases in which individuals have biased beliefs, a relevant question is how the biases in beliefs appear to be sustained over time, despite the opportunity for feedback. In some instances, the imperfect information that individuals have is consistent with optimal inattention: people disregard available information rationally, as the cost of information processing may exceed the value of the information (Sims, 2003; Caplin & Dean, 2015). However, in other cases it appears that behavioral forces are also at play (Schwartzstein, 2014; Gagnon-Bartsch, Rabin, & Schwartzstein, 2018). For example, people focus their attention on what they perceive as relevant and important, but if they misconceive what really matters, they may miss essential information (Hanna, Mullainathan, & Schwartzstein, 2014; Esponda, Oprea, & Yuksel, 2022). Studies have also shown that at times people make errors in evaluating information in order to maintain strongly held beliefs, such as a positive self-image or an ideologically motivated belief (Fryer, Harms, & Jackson, 2019; Santos-Pinto & de la Rosa, 2020; Thaler, 2021). Confirmation bias, which occurs when people discount information that contradicts their past choices and judgments, is a prominent example (Nickerson, 1998; Rabin & Schrag, 1999). People may also place more weight on favorable news than on unfavorable news (Eil & Rao, 2011; Möbius et al., 2014; Coutts, 2019). Furthermore, there is evidence that people have asymmetric memory recall, more easily recalling positive feedback than

negative feedback (Enke, 2020; Zimmermann, 2020; Huffman, Raymond, & Shvets, 2022).

Present Bias

Limited attention, cognitive barriers, and incorrect beliefs are factors in the way people perceive and comprehend economic information and the probabilities of future events. Other cognitive traits affect the way individuals respond to the information and incentives that they perceive. Present bias is one of the most widely documented behavioral factors, and it is particularly important for the development of public policy. Box 3-2 shows how it is modeled by economists.

There are two key features of the way people perceive time-related issues that behavioral economists attempt to take account of in their work. First, most people are tempted to put off activities that are beneficial but require an immediate cost, such as going to the gym, writing a report, or filling out a tax return. At the same time, people are tempted by pleasant activities that may have a future cost, such as lighting one more cigarette or eating an extra scoop of ice cream. In planning for the future, people display tension: they would very much like their future selves to exercise, write reports, get their taxes done, stop smoking, and eat salads rather than

ice cream, but they do not necessarily want to do those things right now. The tension is between the lure of immediate gratification in the present and longer-term hopes and goals for the future.

Behavioral economists have added a second, more subtle, component to understanding of present bias by making a distinction between the naïveté or sophistication that individuals bring to their assessments—or how aware they are of their present bias. A sophisticated individual is aware not only that they do not feel like filling out tax forms today but also that they will still dislike doing so in the future. A naïve person, in contrast, might expect their future self to not have the same present bias in the future and thus to be more willing to face those IRS forms at a later time.

Sophisticated and naïve individuals often behave quite differently. More sophisticated individuals, aware of their future selves’ intended behavior, look for ways, such as commitment devices, to tie the hands of their future selves. They tend not to delay tasks on the to-do list because they are aware that things that do not get done today may not get done tomorrow either. Naïve individuals, in contrast, tend to have unrealistic expectations that their future selves will take on the hard tasks and thus are more likely to procrastinate. The evidence suggests that sophistication and naïveté are not crisp categories but are reflected in varying degrees in everyone—although naïveté may predominate (Augenblick & Rabin, 2019).

Present bias has important implications for understanding how people respond to public policy. Two key findings mirror each other: people are especially aversive to, and tend to delay, undertaking bureaucratic tasks, such as signing up for a benefit for which they are eligible; and people are not very sensitive to future benefits, even substantial ones, but are quite responsive to benefits perceived as immediate.

Aversiveness to bureaucratic tasks is a phenomenon almost anyone would likely recognize: in the face of tasks such as calculating one’s taxes, signing up for the Supplemental Nutrition Assistance Program, or changing health insurance plans, the cost of the effort of figuring out complex forms and the administrative burden looms large. Especially when there is no clear deadline—and when naïveté is dominant—people will push the effort to the future under the (usually incorrect) assumption that it might be less unpleasant later and they will get to it at some future time.

Procrastination is a natural explanation for the very large default effects (tendency to accept whatever is the default option) that are found repeatedly in behavioral economics. For example (noted in Chapter 2 and further discussed in Chapter 6), employees are far more likely to enroll in 401(k) matching plans when opting in is set as the default choice. Similar work shows that individuals do not optimally adjust their retirement and health insurance choices over time, so that they get stuck in suboptimal plans (Handel, 2013; Bhargava, Loewenstein, & Sydnor, 2017). The health

insurance example noted above as an example of limited attention also demonstrates present bias because even attentive workers might put off enrolling in better plans because they naïvely believe they will do it in the future. Similarly, people generally itemize deductions on their tax returns only when the amount of deductions they claim is hundreds of dollars higher than the standard deduction (Benzarti, 2020), even though they would benefit from itemization of smaller totals. This high avoidance of a seemingly clear benefit could be partly the result of limited attention, but present bias and naïve procrastination also play a role.

The evidence that individuals are not very sensitive to future benefits is also supported by research from Denmark that suggests that people do not respond at all in their savings decisions to variation in the rate at which their employers contribute to retirement savings, even though the traditional economic approach would predict substitution (Chetty et al., 2014). Interestingly, information about the benefit alone is not generally enough to stimulate action. Even people whose attention is directed to the possibility of receiving the employer’s matching funds are slow to respond unless they have an opportunity to act right away as well as a deadline (Choi, Laibson, & Mandrian, 2011). In another case, the offer of a gift card did yield increased savings, but it is striking that employees responded far more substantially to an immediate incentive of $10 than to information on benefits worth thousands of dollars at retirement (Bhargava & Conell-Price, 2022).

Reference Dependence and Framing

One of the most influential models in behavioral economics is reference dependence, which is based on prospect theory (see Chapter 2). Prospect theory itself was based on observations of how people respond when faced with uncertain and risky choices. The traditional economic model assumes that people will rationally calculate the potential gains and losses under different scenarios, and will pick a riskier choice only if their average well-being could be significantly improved.¹ But, as discussed in Chapter 2, Kahneman and Tversky (1979) found that decisions involving risk often deviated from the traditional model. In facing risky choices that have only small magnitudes of gains and losses, people are more averse to risk than is predicted by the traditional model. For example, people often purchase insurance to protect themselves against losses (a risk-averse behavior) but simultaneously engage in risky lotteries that could bring gains or losses. Furthermore, if behaving according to expected utility, people

___________________

¹ The idea of rationally assessing the scenario that is most likely to yield overall benefit, given uncertainties, is captured in what is called expected utility theory, “a weighted average” of the potential utilities of various possible outcomes (Briggs, 2019).

would consider all aspects of a decision, integrating all relevant elements of the context. For example, in deciding to participate in a lottery, an individual should take into account their underlying wealth, including expected returns to any investments.

Kahneman and Tversky proposed that anomalies—deviations from expected utility–based behavior—could be explained by a simple theoretical model incorporating three key psychologically based insights. First, people exhibit loss aversion, which means that people respond more strongly to losses than the positive effect of equal-sized gains—this effect is much stronger than is predicted by the standard model. Second, people’s choices are framed relative to a reference point, not in absolute terms, and are thus highly contextual (the reference point will depend on the context). A reference point could be a previous choice, a comparison group, or a default choice that is presented. (Reference dependence is an example of the influence of context discussed in Chapter 2.)

A third aspect of the model is an implicit assumption (made more explicit by recent research) of narrow framing: people tend to consider a particular decision in isolation, taking it as framed, instead of integrating it with all the other economically relevant factors. For example, consider an individual participating in a lottery that offers a 50 percent chance to gain $10 and a 50 percent chance to lose $8 (or the choice to opt out). The concept of narrow framing suggests that a person will consider this decision in isolation and not assess it in light of other aspects of their financial situation, such as wages or salary and returns from a portfolio investment. If those considerations were integrated with the lottery choice, the risk from the gamble would appear much smaller (or larger, if the amounts at stake were higher). Thus, the framing of a particular choice, and the reference point in particular, will affect how an option is perceived and thus the ensuing decision. An example of this was documented in a study of investors in 401(k) retirement accounts, which showed that they were more likely to take into account the asset allocations in their other investment accounts in making decisions about the 401(k) plan when they were prompted to do so (Choi, Laibson, & Madrian, 2009).

A classic example of reference dependence and loss aversion is the discrepancy between willingness to pay for an object and willingness to accept something in exchange for the object: this is the endowment effect described by Kahneman, Knetsch, & Thaler (1990). In studies of this phenomenon, subjects are randomly assigned either to receive an object (e.g., a mug) or not. The subjects who are given the mug are asked their willingness to accept something for parting from the mug, while the subjects without a mug are asked their willingness to pay to get a mug. These experiments have demonstrated that people’s willingness to accept an exchange is generally larger than their willingness to pay. Models of reference dependence

are natural explanations for this phenomenon, given that loss aversion increases willingness to accept (as the people would lose the object by selling) but not willingness to pay (see Marzilli Ericson & Fuster, 2014). If losses loom larger than gains, one would predict this disparity.²

An example of this behavior and the endowment effect arises in environmental evaluations. For example, after an oil spill, the compensation amount is decided on the basis of elicitations of willingness to accept an exchange and willingness to pay for clean water and clean beaches (e.g., Bishop et al., 2017). The research on endowment effects provides one reason that such evaluations will tend to be context dependent, as the elicitation is likely to depend on whether it is framed in terms of willingness to accept an exchange (“what would you need to be compensated to lose the clean water?”) or as willingness to pay (“what would you be willing to pay to have clean water?”).³

Another example of the impact of reference dependence through framing is nominal wage rigidity. Nominal (actual) wage cuts are rare, in comparison with real wage cuts (e.g., Card & Hyslop, 1997). That is, wage increases are common, but if inflation rates are high the result might still be a reduction in the value of earnings, a real wage cut. That is, if one’s wages are increased by three percent when inflation is six percent, the value of one’s wages is three percent lower than it had been. In part because of reference dependence and loss aversion, workers evaluate wages in nominal terms, comparing this year’s pay to last year’s, so a nominal wage cut is seen as a loss, while a real wage cut is not. A study conducted in India illustrates how nominal wage rigidity distorts the adjustment of markets to shocks (Kaur, 2019). An interview-based study suggesting that employers are unlikely to put in place nominal wage cuts because they damage worker morale supports this notion (Bewley, 1999). This is an example of reference-dependent statements of fairness documented by Kahneman, Knetsch, & Thaler (1986).

Other evidence comes from the housing market. Homeowners are typically averse to selling their houses at a (nominal) loss relative to the price at which they purchased their houses (Genesove & Mayer, 2001; Andersen et al., 2022). Other work showed that homeowners who purchase home insurance and pick low-deductible, high-premium plans even though the probability of a major loss is very small do so because they dislike the possibility of having to pay a high deductible in case of an accident: this finding also demonstrates loss aversion and that low-probability outcomes loom excessively large in people’s perception (Sydnor, 2010).

___________________

² See also the work of Ellen Langer on the endowment effect (e.g., Maymin & Langer, 2021).

³ A separate issue is that elicitations involve problems with validity and appropriateness of contingent valuation, but those problems do not affect the point being made here. We also note that willingness to pay is bounded by ability to pay.

In another example, unemployed workers looking for a job reduced their hours spent in job search efforts after their unemployment insurance benefits ran out (even taking into account unemployment insurance job search requirements) instead of increasing them, as the traditional economic model presumes (DellaVigna et al., 2017, 2022). The authors suggest that this occurred because the unemployed workers adjusted the reference point after their benefits ran out. A study of taxi drivers also demonstrated this phenomenon: those who experienced an above-average number of fares late in the day tended to stop work early, rather than taking advantage of the opportunity to earn even more, because they had already achieved or slightly exceeded the normal daily earnings, which they used as a reference point (Thakral & Tô, 2021, building on the work of Camerer et al., 1997).

Social Preferences and Social Norms

The textbook traditional economic model assumes that people act to enhance their own well-being and tend to ignore, in making decisions, both the consequences for others and social norms that are not directly reflected in economic payoffs. However, social psychologists have shown that an important aspect of the context for decision making is perceptions of how one’s actions relate to those of others. People often care about the well-being of others. They make comparisons between themselves and others, and they care about their social standing, how they signal their values and preferences to others, and how they conform with social norms. People think about themselves (more or less consciously) in terms of social groups, compare themselves with others, are influenced by how others behave, and put considerable weight on how others perceive them. Behavioral economists have devoted considerable attention to exploring social interactions, peer effects, and conforming behavior, using variants of the traditional model. These models capture some elements of social preferences that come from social psychology and are consistent with some simple behavioral economics models. However, behavioral economists have explored how such socially based preferences affect choices and interactions with others in a deeper and more extensive way and have brought new conceptual models to the field.

Some of the earliest behavioral work in this area explored social preferences, the ways people care about other people’s income and consumption, perhaps in relation to or in comparison with their own, in new ways. This work demonstrated the ways people’s accrual choices diverge from those that would be driven by pure self-interest. For example, many people sacrifice their income and consumption to help others because they are altruistic or have a preference for fairness (Becker, 1974; Rabin, 1993). Nuanced studies of relative income (the impression of an income in comparison with

a reference income) have pointed to the concept of inequality aversion: the idea that most people dislike inequality in outcomes relative to others’ but are more disturbed by inequality when they earn less than others than when they earn more than others (e.g., Fehr & Schmidt, 1999; Chen & Li, 2009). Furthermore, people tend to display reciprocity—responding to being treated generously by behaving more generously themselves and vice versa if they are treated unkindly (Falk & Fischbacher, 2006). Behavioral economics has used such tools as the ultimatum game, the dictator game, and public goods games in studying people’s behaviors in response to various challenges to their concepts of fairness (see, e.g., Houser & McCabe, 2014).⁴ Researchers have also specifically studied social interactions within organizations (Ashraf & Bandiera, 2018).

Recent work in behavioral economics also goes beyond traditional models by suggesting that social preferences and other preferences that drive behavior derive from people’s understanding of themselves within a social context and their identities and attachments to particular social groups. People care about following social norms more than can be accounted for by simple peer effects would suggest because those norms often indicate appropriate and inappropriate behavior for different settings. People also care about whether others view them as following the norms and as having values that align with the social group. These behaviors do not appear in traditional economic models.⁵

We emphasize two points from the research on social preferences and norms. The first is that people make choices to conform to social norms consistent with their identities, and thus conformity is especially pronounced in public spheres. A theoretical model of economic behavior that includes identity and social norms illustrates this point (Akerlof & Kranton, 2000). The model assigns each person an identity (or possibly multiple identities) that is associated with norms that prescribe behaviors in different settings. If a person with a particular identity does not follow the prescribed behavior, they pay a “disutility cost.” Some traditional models might portray behavior related to social preferences in simple descriptive

___________________

⁴ In an ultimatum game, one player is endowed with a sum of money and tasked with splitting it with another player. The first person decides how much to give the second person, who can accept or reject the offer. If the second person accepts the offer, the money is split as proposed; if the person rejects the offer, both players receive nothing. Both players know in advance the consequences of the second person accepting or rejecting the offer. The dictator game is a variation of the ultimatum game. In a public goods game, subjects secretly choose how many of their private tokens to put into a public pot. The tokens in this pot are multiplied by a set factor and this “public good” payoff is evenly divided among players. Players also keeps the tokens they do not contribute.

⁵ There has been extensive work on these issues in the larger behavioral science literature mentioned earlier in this chapter. See, for example, the important work of Bicchieri (2005).

terms, but, in contrast, the Akerlof and Kranton models portray prescriptive behavior—they portray individuals’ views of what they should do. Many studies of the role of social identity in academic performance demonstrate the importance of this factor. For example, there is evidence that students’ decisions about their academic performance are different when they are visible to their peer groups (e.g., Hoff & Pandey, 2006; Bursztyn & Jensen, 2015). In a study of the influence of stereotypes on academic performance, the students performed better when stereotypes about their racial identity associated with strong academic performance were subtly highlighted in the testing context, but performed worse when other stereotypes they might identify with that are associated with weaker performance were invoked (Shih, Pittinsky, & Ambady, 1999). We note that the stereotypes do not have to be true, but if people have internalized them, they can have strong effects through identity and social norms.

People also care about whether they are perceived as altruistic as in social signaling models (Bénabou & Tirole, 2006). For example, people may behave more altruistically when they are being observed by others because they want to signal to others that they have this virtue. Other work shows that preferences for fairness and inequality aversion are group-based—that inequality aversion is much weaker toward participants who are outside one’s own group and that people are more likely to make decisions that could harm others after seeing others in their group do so (Chen & Li, 2009; Shayo & Zussman, 2011; Bauer et al., 2018; Kranton et al., 2020).

A second key point from this body of work is that comparisons to others often trigger responses that are consistent with the importance of social norms and social comparisons of the type emphasized in behavioral economics. That is, people’s behavior is often governed by their desire to have the good opinion of others and to maintain positive relationships (e.g., Cialdini & Trost, 1998). Efforts to harness this desire to promote socially desirable choices in numerous contexts illustrate this idea (see Chapter 8). For example, people respond to social pressure by reducing their energy use when informed about the average energy use by others (e.g., Allcott, 2011). Presenting taxpayers with the message that the vast majority of their peers pay on time has been effective at increasing on-time filing (Luttmer & Singhal, 2014).⁶

___________________

⁶ The existing literature on social interactions and social networks emphasizes what is called the “general equilibrium” nature of the problem, referring to the fact that when individuals respond to other individuals’ actions and beliefs, those other individuals respond in kind, multiplying the effect. All the behavioral forces identified in this section would also have those effects.

CONCLUSION

Insights about human behavior from work in cognitive psychology, social psychology, and behavioral economics (and other fields) have been integrated in the development of a sophisticated portrait of how people make decisions. Drawing on this work, behavioral economists have identified similar and overlapping concepts and have developed distinct ways of framing questions and findings. The integration of work from behavioral domains with economic analysis has demonstrated that decision processes are dynamic, malleable, and context dependent, often shaped by what would ordinarily be considered irrelevant factors and by people’s mental representations of the world.

Conclusion 3-1: Foundational work in behavioral economics, drawing on the fields of economics, psychology, social psychology, and others, suggests the importance of five principles to consider in designing policy interventions to modify human behavior: limited attention and cognition, inaccurate beliefs, present bias, reference dependence and framing, and social preferences and social norms.

REFERENCES