This chapter consists of commentary from the meeting participants.
Good theories are often based on the few things that really matter, and ignore large amounts of detail. What makes a theory successful, however, is our ability to use it to distinguish between detail that can be ignored at low cost and detail that is needed for a successful characterization of the relevant facts. A theory can run into trouble when it dismisses as irrelevant factors that are at the heart of important outcomes.
With regard to modern economic theory as a descriptive enterprise, we have increasingly good evidence for, and some insight into, a variety of behavioral factors that appear to be highly consequential for economic outcomes. Many of these factors, however, are dismissed by the classical theory, explicitly or implicitly, as unimportant detail. The list is long. It includes a variety of cognitive, perceptual, social, cultural, and other factors that appear to play a central role in people’s decisions, in ways that are profoundly different from what was envisioned by the progenitors of rational agents. A number of these factors have been discussed in this book, and some are further considered in Tirole’s commentary below.
Now, on the descriptive front, it is true that what we have so far consists largely of “a list” of factors, rather than a characterization that is more general or theoretical in nature. And this lack of generalization makes any attempt to incorporate the new behavioral insights into the models quite daunting. It is also at the heart of Tirole’s admonition that we resist the temptation to incorporate “stylized facts” into economists’ relatively unified, “portable” theory.
As a modeling strategy, that may be exactly right. But for those of us who want an account that is descriptively faithful, some fairly radical revisions to the portable theory seem unavoidable. The problem is deeper than merely choosing which of the stylized facts are in and which are out. The problem we have, if you will, is one not just of content, but also of structure.
The dialogue between economics and psychology presents a classic divergence in disciplinary sensitivities. Whereas the psychologist is happy to focus on the fascinating events happening behind the eyes and between the ears, the economist wants to know what specific behaviors are generated thereby. This tension has had a substantial impact on psychologists who are engaged in the dialogue with economists: their work has veered in the direction of a greater focus on actual outcomes, in real circumstances, with real incentives, and so on. At the same time, economists interested in a behaviorally more sophisticated theory need to take into account some fundamental facts about our mental lives. And by that I do not mean any specific brain processes underlying any particular activity but, rather, a much more general, somewhat mundane, fact about psychology that is critical to decision-making.
Whereas the notion of “construal” is second nature to psychologists, it is far less obvious to the untrained observer and even less so to economists. Economic thinking about optimization concerns itself with things in the world, with alternatives as they really are (or, as they are best understood by a rational actor). The axioms that underlie the normative theory are stated in terms of extensional outcomes, where each option, representing some state of the world, is clearly and uniquely designated, independently of how precisely it may be described or perceived. The problem with this perspective is a terribly trivial but profoundly consequential fact about human nature: people do not actually choose between items in the world—they choose between those items as they are represented in the mind. Decisions are not about objective (extensional) states of the world, but about the mental (intensional) representations of those states (see Tversky and Kahneman 1983). And the mental representation of states of the world is subject to a rich psychology involving perception, embellishment, interpretation, and distortion.
Construal processes are central to many areas of psychology, ranging from auditory and visual perception, memory, language, and self-perception, to judgments, decisions, opinion surveys, and social interaction. In the case of decision-making, as we all know well, nuanced differences in otherwise logically equivalent representations of options can generate significantly different perceptions, trigger different reactions, and yield systematically different choices.
We thus are left with multiple sources of tension. First, there are the “contents” of decision: what attributes enter into people’s calculations. These include things such as prosocial behavior, a concern with fairness, or interpersonal comparisons, which turn out to have a more substantial then normatively anticipated impact. Then there is the structure, or context, of decision. This includes unavoidable contextual nuances, ranging from how the options are described (even when providing otherwise identical information), to the process used to elicit preference, the presence or absence of other (“irrelevant”) alternatives, the point of reference, and so on.
As a matter of fact, there are additional factors that are somewhere in between context and content. There are features that—unlike fairness, say—people do not profess to care about, and may even be unaware of, but which, once introduced into the context of a decision, nonetheless can have a substantial impact on choice. For example, people gently touched on the shoulder by a waitress at a restaurant tip higher than those who were not touched. And when asked later, all attribute their tipping to the quality of service, with no awareness of the influence of the touch (Crusco and Wetzel 1984; Schwarz 1990; Schwarz and Clore 1983).
We have recently conducted a field experiment in South Africa to assess the relative importance of price and of various subtle psychological features in the decision to take-up a loan offer from a local lender (Bertrand et al. 2005). Some 57,000 incumbent clients of the lender were sent letters offering large short-term loans at randomly chosen interest rates. Consistent with standard economics, those offered higher rates were less likely to take up a loan then those with access to lower rates. In addition, various psychological features on the offer letter, which did not affect offer terms or economic content, were also independently randomized. One of the features consisted of the number of sample loans displayed: the offer letter displayed either one example of a loan size and term, along with the respective monthly repayment, or it displayed multiple (four) examples of loan sizes and terms. In contrast with standard economic prediction, we found higher take-up under the one-example description than under the multiple-example version. The magnitude of this effect was large: the simpler description of the offer had the same positive effect on take-up as dropping the monthly interest on these loans by two percentage points. (In a similar vein, Huberman et al. (2004) found that employees’ participation in 401(k) plans drops as the number of fund options proposed by their employer increases.) We also randomized the presence or absence of a smiling woman’s picture in the bottom corner of the offer letters. For the men in the sample, the presence of that picture had the same positive effect on take-up as dropping the monthly interest on the loans by 4.5 percentage points! On average, any one psychological manipulation we tried had the same effect as a one-half percentage point change in the monthly interest rate.
Unlike a concern with fairness or even an aversion to losses, nobody would acknowledge or, for that matter, be willing to entertain the possibility that a touch on the shoulder or a picture in the corner of a letter predisposes them to make a substantially different economic decision. Choice behavior is the outcome of a variety of cognitive and affective processes. Some of these we consciously endorse and even insist on; others we have no introspective access to, nor control over (Bargh 1997). And many of both kinds of processes depart in significant ways from economists’ unified, “portable” theory.
Tirole correctly observes that (see p. 298, this volume)
When parents report that they “value” education, when Americans “believe” they should be saving more, or when people “believe” that they perform better than average in a number of dimensions, they may indeed be making mistakes, but they may also be engaging in cheap (although therefore not very effective) reinforcement of their identity or self-image.
Similar remarks may apply to economists’ “belief” that their model is “doing just fine, thank you.” Perhaps they are not actually making an error, just reaffirming their identity. And meetings such as this may help alleviate the pressure and allow for a more permissive self-image. It is, in fact, quite respectable to be in possession of a theory that is in many ways beautiful, but that nonetheless requires substantial restructuring before it is able to predict our quirky behaviors.
Bargh, J. A. 1997. The automaticity of everyday life. In Advances in Social Cognition (ed. J. R. S. Wyer), Volume 10, pp. 1–61. Mahwah, NJ: Erlbaum.
Bertrand, M., D. Karlan, S. Mullainathan, E. Shafir, and J. Zinman. 2005. What’s psychology worth? A field experiment in the consumer credit market. NBER Working Paper 11892.
Crusco, A. H., and C. G. Wetzel. 1984. The Midas touch: the effects of interpersonal touch on restaurant tipping. Personality and Social Psychology Bulletin 10:512–17.
Huberman, G., S. S. Iyengar, and W. Jiang. 2004. How much choice is too much: determinants of individual contributions in 401K retirement plans. In Pension Design and Structure: New Lessons From Behavioral Finance (ed. O. S. Mitchell and S. P. Utkus), pp. 83–96. Oxford University Press.
Schwarz, N. 1990. Feelings as information: informational and motivational functions of affective states. In Handbook of Motivation and Cognition: Foundations of Social Behavior (ed. E. T. Higgins and R. Sorrentino), Volume 2, pp. 527–61. New York: Guildford Press.
Schwarz, N., and G. L. Clore 1983. Mood, misattribution, and judgments of well-being: informative and directive functions of affective states. Journal of Personality and Social Psychology 45:513–23.
Tversky, A., and D. Kahneman. 1983. Extensional vs. intuitive reasoning: the conjunction fallacy in probability judgment. Psychological Review 91:293–315.
Churchill once famously stated that “democracy is the worst form of government except for all those others that have been tried”; economists have long regarded their neoclassical model in a similar light: clearly unsatisfactory, yet superior to alternatives. This unhappy state is of course what makes research and volumes like this so exciting.
The economist’s self-view is currently being challenged. Building in particular on work in psychology, the standard model of maximization of present discounted expected utility has been recently amended to allow for
• a broader description of preferences with the introduction of anticipatory feelings and memories, concerns about self-image, social values and even emotions, and the use of reference points and path dependency,
• a broader approach to discounting, as with hyperbolic preferences,
• a broader take on belief formation, including errors in Bayesian updating, imperfect memory, and biases in hedonic forecasts, and
• a challenge to the optimization hypothesis—rules of thumb, mental representations, analogies, and categorization strategies, with an ongoing debate as to whether limited optimization is rational given memory, mental search and time constraints.
As has already been noted in this volume, while other social scientists often do not aim at an overarching theory of decision-making, economists often tend to prefer a relatively unified, “portable” theory that helps explain a wide range of behaviors with a small number of core assumptions. They use a relatively parsimonious framework that has predictive power and normative implications across a broad range of situations. Thaler once rightly remarked that behavioral economics will be successful once it stops being “behavioral” and it becomes just economics. To this purpose, and in my opinion, we sometimes need to resist the temptation to incorporate “stylized facts” about human behavior into reduced forms for economic modeling and should rather build more structural models, “dig deeper,” and run experiments to discriminate finely among alternatives in order to perform normative analysis. Our profession has sometimes learned this lesson the hard way, as when, in reaction to the (correct) observation that firms do not maximize profits, the literature of the 1950s and 1960s added objectives into a firm’s reduced form payoff function (this literature was abandoned when agency theory built an alternative approach based on the unifying principle of asymmetric information). Understand me well: warning against the use of reduced forms has its limits; after all, we all to some degree always use reduced forms in order to limit complexity, and even reduced forms that are by any measure very sloppy can prove stimulating and ultimately useful (witness IS–LM theory).
My first illustration may be tainted by recent work with Roland Benabou (Benabou and Tirole, forthcoming), but I will be discussing the challenge to modeling, which I feel rather confident about, more than the way we address it, for which I am a poor judge. This illustration refers to how economics can take on board prosocial behavior. People frequently engage in activities that are costly to them and mostly benefit others: they vote, volunteer time, help strangers, give to political or charitable organizations, donate blood, join rescue squads, and sometimes even risk or sacrifice their life.
One tractable approach for economists would consist in positing a reduced form in which individuals—or at least a fraction of individuals—have a taste for prosocial behavior in the same way macroeconomists use warm-glow preferences in order to build a tractable model of bequest behavior. One could perhaps also add some situational parameters to account for the wide variations in prosocial contributions across individuals, circumstances, and causes. Even so enriched, the “taste-for-prosocial-behavior-in-the-utility-function” approach may encounter some difficulties.
First, it would be difficult to explain, except in an ad hoc manner, why incentives sometimes crowd out prosocial behavior. In this reduced form, incentives would just add to the natural proclivity of individuals to engage in prosocial behavior.
Second, reduced forms would not offer clues as to why prosocial behaviors are sometimes socially determined.
Third, and quite importantly, the “prosociality” of prosocial behaviors often needs to be questioned. Consider the following.
• We tip taxi drivers in rich countries and fail to give money to starving children in Africa.
• In studies reported in Glazer and Konrad (1996), less than 1% of donations are anonymous.1
• People often react with disapproval when someone reveals how generous and disinterested he is.
• In a transportation-related survey of about 1,300 individuals, Johansson-Stenman and Svedsäter (2003) find that people who are asked which attributes are most important to them in a car systematically put environmental performance near the top and social status near the bottom; but when asked about the true preferences of their neighbors or average compatriots, they give dramatically reversed rankings.
• A clever experiment by Dana et al. (2003) reveals that when people know how their choices will affect other people, they often display substantial altruistic behavior; when they are given the opportunity to remain ignorant of how their choices affect others, or of their precise role in the outcome, many choose not to know and revert to selfish choices. They exploit the “moral wriggle room.” To illustrate this, suppose that an individual plays a “dictator game,” i.e., a game in which he chooses between (say) two actions affecting his welfare and that of another, passive individual. A “selfish action,” A, gives the highest material payoff, 5, to the dictator. The other action, B, gives him only 4. Suppose that the selfish action yields 0 and action B yields 8 to the passive player. As is well-known, a substantial fraction of dictators in those circumstances would choose action B, sacrificing a bit of material payoff for the psychic payoff associated with a good deed. Suppose, next, that there are two states of nature, and that the dictator’s payoff is as above regardless of the state of nature. By contrast, the passive player’s preferences across actions depend on the state of nature. In the first state, they are as above. In the second state, action B yields 4 to the passive individual while action A yields her 5. Finally, the dictator is asked whether he wants to know which state of nature prevails. It would seem that the dictator would want to know which state prevails and, provided that she enjoys the psychic payoff from raising the passive individual’s payoff from 0 to 8, choose action B in the former case, and action A in the latter case. A substantial fraction of dictators, though, choose not to know and select the selfish action, A, with the “excuse” (in their own eyes at least) that action A may actually benefit the other individual.2
• Identification with a group and the concomitant altruistic or even heroic acts are often associated with a hostile or even cruel attitude toward the out-group.
These six observations, and others, suggest that the economists’ rendering of prosocial behavior should combine heterogeneity in the individuals’ degree of altruism and greed with concerns for social reputation and self-respect. This is in line with (many) psychologists’ approach to these issues in terms of the “over-justification effect.” For instance, a prominent explanation of the “crowding out effects” of prosocial behavior by incentives alluded to earlier is that the presence of rewards or punishments changes the meaning that people attach to actions (by others or their own) in a detrimental way. My own work with Benabou on prosocial behavior combines heterogeneous social preferences with cognition. That is, prosocial actions are undertaken both because a fraction of individuals are genuinely other-regarding, and because individuals either are in a quest for social esteem or strive to maintain a certain self-view of “what kind of person” they are. In such an environment, incentives then give rise, in economics parlance, to a signal extraction problem, which gives a formal content to the over-justification effect. (This cognitive approach is then used to study various issues such as the form and confidentiality of prosocial behavior, the emergence of multiple social norms enforced by the interplay of honor and shame, and the competition among the recipients of altruism.)
A digression here: when discussing survey evidence on attitudes and beliefs quite generally, we should take social signaling and especially self-signaling more seriously than we currently do. This was pretty clear in the transportation survey just discussed. Similarly, when parents report that they “value” education, when Americans “believe” they should be saving more, or when people “believe” that they perform better than average in a number of dimensions they may indeed be making mistakes, but they may also be engaging in cheap (although therefore not very effective) reinforcement of their identity or self-image.3
With prosocial behavior as with any other theme of behavioral economics, social scientists may have to dig deeper in order to find the primitives and to perform a welfare analysis. Other illustrations include three concepts that many at the meeting, including myself, think economists should take on board: fairness, endowment effects, and bounded willpower.
Why do people turn down unfair offers in ultimatum games; or, as Bewley documented (see Chapter 5, this volume), have their morale reduced when their wage is cut? Is it because of intention-based preferences (an intrinsic preference for reciprocity), social preferences (a preference for certain forms of income distribution—perhaps one that refers to a moral ideal of equal pay for equal work, a rule highly sensitive to adverse selection, or equal pay for equal output, a rule highly sensitive to moral hazard), a preoccupation with self-image (for example, the management of one’s self-reputation as someone who does not turn the other cheek), a taste for procedural fairness (also perhaps stemming from a moral ideal associated with organizational efficiency), or a more standard economic explanation, suggested by Holmström at this meeting, in which the victim learns about others’ views of himself and project the implications of the signal for the future of the relationship?
I do not know, but two points are worth making: first, different hypotheses have different positive and policy implications. Second, and relating to the first four hypotheses, it would be helpful to understand the origin of the preferences. Ernst Fehr’s research agenda should prove very helpful in this respect.
Again, a fascinating and relevant effect (as documented for example, by Bewley (see Chapter 5, this volume) for the quick development, following an increase in pay, of a feeling of entitlement). On the other hand, a key challenge for the theory of loss aversion and endowment effects is that of the determination of the reference point. As Honkapohja noted above (Chapter 5, this volume), empirical observations cannot be interpreted through the lens of an application-specific reference point. Rather, some broad principles are called for. My hunch is that a cognitive approach building on self-signaling and social signaling (and sometimes on informed principal theory) is likely to be relevant. But this is largely unexplored territory at this stage.4 Furthermore, other approaches, for example, building on neuroscience, are likely to be relevant as well.
Other examples come to mind. For instance, hyperbolic discounting is a tractable reduced form capturing bounded willpower, but one that for some applications will prove rather unsatisfactory. (Incidentally, the classical economists’ rendering of bounded willpower—for instance that of Smith, Bohm-Bawerk, or Jevons—was richer than that of behavioral economists.) My guess is that, as for the rest of economics, behavioral economics must build on a back-and-forth interaction between careful and parsimonious modeling combined with structural tests in the field and in the laboratory.
Benabou, R., and J. Tirole. Forthcoming. Incentives and prosocial behavior. American Economic Review.
Dana, J., R. Weber, and J. Kuang. 2003. Exploiting moral wriggle room: behavior inconsistent with a preference for fair outcomes. Carnegie Mellon Behavioral Decision Research Working Paper 349. (Available at http://ssrn.com/abstract=400900.)
Glazer, A., and K. Konrad. 1996. A signaling explanation of charity. American Economic Review 86:1019–28.
Johansson-Stenman, O., and H. Svedsäter. 2003. Self image and choice experiments: hypothetical and actual willingness to pay. Göteborg University and London School of Economics, Working Paper 94.
Koszegi, B., and M. Rabin. 2004. A model of reference-dependent preferences. EconWPA Working Paper 0407001.
It is gratifying to see how ideas in social psychology have been applied to behavioral economics and real-world problems. The two disciplines share a common goal: to understand how people process information and make decisions in ways that impact their lives and the lives of others.
It is important, though, to point out differences between the disciplines of social psychology and behavioral economics that have, perhaps, kept them apart. Behavioral economics has largely been concerned with understanding policies and real-world applications, whereas social psychology has focused more on basic research—on what makes people tick. There are also important methodological differences. Psychology is dominated by laboratory experiments, whereas behavioral economics relies more on field studies. There are, of course, many exceptions to these rules in both disciplines, but you are more likely to find a social psychologist doing basic laboratory studies with college students, and more likely to find behavioral economists doing large-scale field studies of important real-world problems.
I think each field can learn from the other and that a lot of progress will be made if we join forces. How might we facilitate such interactions and what form might they take? I suggest that we collaborate on small-scale studies that employ the experimental method in applied settings.
Social psychologists have been reticent to do field research for a number of reasons. One is that it can be difficult to know how the processes we study in laboratory studies, in which we focus on a small number of independent variables, will play out in natural settings. Armchair thinking on how intervention might work is not much help here. We need to take our best shot in small-scale interventions, with randomly assigned control groups, much as argued by Ross and Nisbett (1991).
Admittedly, it is not easy to conduct well-controlled experiments in natural settings. With a little cleverness, however, we can do it, as with the following examples.
Ayres (see p. 145) mentioned a fascinating correlational study he has done in New Haven on tipping taxicab drivers. Taxi passengers gave lower tips to African-American drivers than to white drivers. As with any correlational study, the causal interpretation of this finding is unclear. To make a stronger causal inference, the study could be done experimentally. The driver of the taxi could obscure his face with a hat or hood, making his race unclear. Then, the researcher could randomly assign passengers to race of driver, by changing the picture of the driver that is posted in the back seat. With such an experiment we would know for sure whether the race of the driver is causing people to tip differently (the driver, of course, should not know which picture is posted during any particular ride). This would fall in the domain of social psychology that looks at context effects on people’s prejudice attitudes, in particular their uncontrolled prejudiced attitudes.
Second, Mullainathan (see p. 86) described the problem of low school attendance in developing countries. It has been suggested that building in incentives such as feeding kids at school at lunch time might be a way to solve the problem. That could be tried experimentally with a small intervention, in which incentives are offered in some school districts but not in others. There are practical and ethical problems but the power of such designs would probably warrant the experiment.
Third, many universities and businesses are instituting diversity training programs, in which the attendees are asked to undergo some sort of educational exercise to enlighten them about diversity and presumably to change their behavior in more egalitarian direction. Little is known about how these programs affect people. If you ask the organizers how they evaluate whether the program has the intended effects, they often point to the responses of the people who participate in the program. This might have some value, but a much more powerful design would be test these programs experimentally. There is a big industry offering diversity training services but, for obvious reasons, they are not always interested in in-depth evaluation of their effectiveness. However, it would not be difficult to come up with control groups that are randomly assigned to different treatments in order to evaluate their true effects.
Fourth, think of a new drug for cancer. We all agree that it should be tested very carefully for harmful effects before marketing. We seem not to have the same attitude toward psychological interventions. In Canada, for example, the law now requires that cigarette packs have pictures of people with cancer and other scary images to dissuade people from smoking. To my knowledge, this intervention was never experimentally tested to see whether the pictures have any effect, desirable or not. When it comes to psychological interventions, we rely more on common sense about what we think will happen and, unlike testing how a drug will work, see little reason to test these interventions experimentally. I suspect that the pictures on cigarette packs in Canada do have some effect, but that is just my common sense talking, and I think it is important to find out empirically what those effects are.
One often hears from social psychologists that they would like to do more applied research but there are all kinds of practical and institutional barriers. This is one area in which social psychologists should learn more about neighboring fields such as behavioral economics. I have been very impressed with the kind of field studies many economists are conducting, often using experimental designs to study important real-world problems. There is a natural collaboration here. Social psychologists know a lot about social influence and cognitive processing, and many are eager to see how these processes play out in the real world. By partnering with economists who have much more experience studying real-world problems, substantial progress could be made.
I share the hope, expressed above by Tirole, that in time behavioral economics will cease to exist. Currently, a first-year economics graduate student learns statistics, math, micro, and macro. At some point, one of these students interested in behavioral economics studies psychology in some form, taught by an economist or, if the student is lucky, by a psychologist. Integration of the underlying intellectual basis for behavioral analysis into the first-year core is the long-term hope.
In designing this conference we wrestled with the question of planning a conference and a volume that would enhance behavioral economics. This issue was addressed explicitly by Camerer (see Chapter 7), who said that behavioral economics really started with hammering away at anomalies. This is a lot easier to do when relevant data are present. I think it is always important to keep anomalies in mind and I think everybody recognizes that merely hammering at anomalies does not move paradigms all by itself. More than that has to be done.
I hope that the conference and this book can generate an increasing respect towards the potential for psychology in economics. I think this is helped by having all the excitement generated by brain pictures in neuroeconomics. However, to go from these brain images to outcomes is both hard and a long route. But I think that through experiments we can increase awareness of the potential contributions of social sciences to better policy generally. And that will make psychology more acceptable inside economics as well. A primary target for generating excitement for behavioral economics is obviously Congress. Congress votes to provide money for building bridges, but it does not vote on a blueprint for the bridge. Yet when it comes to designing a pension system, Congress lays out the blueprint. If we can get the idea of the importance of experiments more embedded in the public consciousness, i.e., more of the sense that there is real expertise in social science, then maybe Congress will rely more on that expertise, as they do with bridges.
How do we advance behavioral economics beyond the anomalies and generate more respect? I think we need to convince a lot of people, and that means primarily young people, that this is a good research opportunity. That is what this volume is trying to do. So what does it take to make a good research opportunity? A first element is that it has to include the type of work that people like to do. An example is the kind of extremely complex simulations that are perhaps best exemplified by the work of Tom Sargent—this really appeals to a set of people who like to get into the code and have the computer skills to do very complex calculations. If it is not fun for you to do that kind of research, you are not going to get caught up in that style of work. Some people like to do hard theory where they use their math techniques. There were some references in the conference to work by Gul and Pesendorfer, which I think has much potential to help people recognize that there is scope in behavioral economics for doing this kind of work. I say this even though I think the psychology in their work is not well-grounded. We need to recognize that different people like to do different kinds of research and the research endeavor benefits from different people doing different kinds of research.
And here, with some reluctance, I may disagree a little with Tirole. I think that sometimes the insights from reduced-form models are worth having as long as one does not overinterpret them. But this goes with the basic point that the real distinction is not between a good model and a bad model, but between a model that is good for some question and one that is bad for that question.
A second element in having a good research opportunity is that there is an audience. Most people do not write for themselves (although a few do). They want an audience so that they can think they are going to be read, think they are going to be applauded. I think this volume can make it clear that there is a range of areas here with audiences and potential audiences. The way we picked topics was to stay away from the fields that are already well-developed or well-advertised—behavioral finance does not need this kind of advertising and behavioral macro just had all the fuss from a Nobel prize to a practitioner. There was no behavioral industrial organization covered at the conference because there is not yet an audience there.
A third element for a good research opportunity, and the one that gives me the most pause, is the ability to judge the quality of a contribution. Most people do not want to write for an audience if the bulk of the audience is not going to like what is written. A researcher needs a sense of what it takes to do good work. One of the things that bothers me a lot in behavioral economics is that it is much harder to judge what will be well-received than it is in a well-established field, where a researcher can judge whether a potential paper seems more interesting than similar ones that have been published—and so it is easy to see where the subject can go. I feel very nervous when talking to graduate students who are thinking about behavioral research; I think the risk, inherent in doing any research, is larger in behavioral economics than in well-developed fields.
The chapters in this volume can help since they spell out what some researchers like. And of course, the bottom line for the definition of a good research opportunity is from the jobs that follow from research, to which economists pay a lot of attention. It is easier for hot prospects who will end up with jobs in top schools to find a supportive environment than it is for the average job candidate.
Long before coming to the conference, I had the feeling that one should never use the term “preferences” without a modifying adjective, whether it be real preferences, underlying preferences, or choice preferences. Preference, I think, needs to be modified since it is so easy to slip verbally from one interpretation of “preference” to another. One of the things I learned at this conference is that the same should be done with “incentives.” We have financial incentives and economic incentives. But we also have other ways of changing behavior that are changes in the nature of incentives. I think we need some terminology here remembering that economists have great interest in incentives as well as in outcomes and in equilibria.
Economics and psychology are different disciplines for the good reason that they try to do different things. Sometimes having different disciplines is important for getting results. Sometimes they create barriers, but barriers that one tries to break down. The participants at the meeting are working at this. I think it is useful to remind fellow economists that we tend to use psychology differently from psychologists. Sometimes the difference is a good idea and sometimes it is not. Another goal of the conference was to show that there is a wide range of psychology that can be useful in economics: that is why we wanted diversity in the type of psychologists in this meeting.
This conference was great fun for us; let us hope it will be good for others.
We should promote the possibility of carrying out experiments about social or economic issues more aggressively to policy makers, and require the testing of laws before they are implemented. An example of an arena where this would be highly desirable is in firearms legislation.
An experiment in law is very different from an experiment in medicine because of the Heidelberg Principle: in the case of law, performing the experiment changes the environment.
If you start with a small experiment, this concern about general equilibrium effects is not so serious, and we will still obtain valuable results.
Politicians may not agree to evaluation through experiments because many politicians would worry that the results of experiments would undermine support for the legislative outcomes they want to push.
From audience: what is the role of evolutionary psychology?
We do not know much about evolutionary psychology, because there are no fossils. It is good for developing intuitions, but it is not enough for supporting and explaining findings.
Evolutionary psychology is a bit like psychoanalysis. It can explain anything but is very difficult to test. Its value, perhaps, is in generating new hypotheses that can be tested experimentally. I think the number of novel hypotheses it has generated about social behavior, however, is very small.
I believe in neuroscience and I want to work on this. It is not only about brain imaging, it is about trying to understand which processes are activated. How the brain makes decisions is very relevant to economics. In the case of addiction, we know how particular substances activate processes in the brain. I am surprised that psychologists are even less enthusiastic than economists. Is this because of competition about resources?
I want to come back to animal studies such as the studies on bees that Shafir talked about. There are two kinds of animal studies. Some are about processes, such as monkey studies, and I find those studies valuable. Other studies are about choices, and those are less useful for economics because humans have a much more complex brain circuitry to make choices and hence animal studies are of limited use here.
I can imagine psychologists in the 1950s saying that social psychology was not useful. I think there is good research and bad research in all fields. There is serious research going on in neuroscience, and it is a legitimate field of inquiry. It has turned out to be very useful for animal and primate research, so it has a potential for understanding human behavior as well. Neuroscience might be hysterical, but it is intrinsically interesting, it will be done, and it might generate very important findings. It is important for economists to play a role, because they ask different questions (in the same way as economists ask different questions to psychologists now).
Can you give examples in neuroscience? This might be easier than in economics. I tried to convince politicians for a long time; and I realized that one good example is worth a thousand regressions.
Politicians have very short time horizons. Therefore, a more promising route to affect policies is to work with government agencies such as the U.S. Department of Education or the World Bank which have much longer time horizons.
This is already well underway. For example, the U.S. Department of Education now requires randomized experiments and careful evaluation before implementing new policies. This is often driven by the person at the head of the institution, and those top administrators are often more open to using randomized trials than [are] elected officials.
There is a tension between psychology and economics: economists like generality, to be able to make predictions, while psychologists are often reluctant to extend their findings beyond the specific context of their study.
This difference is probably not that stark. Psychologists are also willing to make predictions about the real world; it is really about their willingness to go out and test these predictions in the field.
This is a completely atheoretical conference. One concern that comes up for a theorist is that you listen to examples, and your temptation is to just go with experiments, which will work for a while, but eventually you will have to organize facts using theories (as Tirole emphasized earlier). In this sense, this conference has been disappointing because it has not yet reached the stage of proposing theories.
There are some domains in behavioral economics where we try to sort theories out: this is the case, for example, with social preferences, where theories are proposed and then tested with experiments. I hope that the cycle between proposing theories and testing those theories with empirical or experimental evidence will be short (weeks), as it is in physics, rather than long (decades) as was the case in economics in the past.
Everybody uses reduced forms and they are indeed often very useful. For example, although the IS–LM theory in macroeconomics is a very reduced form, it has played a tremendously useful role. Later on, research based on new developments in imperfect competition, costly price readjustments, search theory or for corporate finance ended up capturing aspects of the IS–LM reduced form predictions. Yet, our aim should be to develop unifying theories which are not context dependent, so that they can be used for normative purposes.
If we want to become behaviorally more sophisticated, I think that, at least at our current level of understanding, we need many different theories and that developing a unifying theory to explain everything is an impossible dream.
1Anonymous donations are still tax-deductible, so the lack of anonymity is probably due to signaling concerns.
2This evidence on the use of moral wriggle room is the self-signaling counterpart of Anne Case’s point on social signaling; namely, that people try to escape social norms of sharing by spending immediately or by investing in illiquid assets.
3Other well-known biases associated with survey evidence include insufficient thinking and the desire to please the experimenter.
4An interesting exception is Koszegi and Rabin (2004), which endogenizes the reference point using Koszegi’s notion of “personal equilibrium.”