Behavioral economics – beyond (ir)rationality and other-regarding preferences
(ECO MW Fellow 2014-2016)
What do you think of when you hear the term ‘behavioral economics’? Maybe you associate it with a specific methodology, maybe with a specific topic. If you think of a methodology, you most probably think of lab experiments, where subjects (usually BA students) are playing games or making decisions in front of computer screens. Indeed, it is most common that a behavioral economics conference will be advertised as a conference on behavioral and experimental economics, with 90% of presentations displaying experiments, mostly done in the lab. If you think of topics, you most probably think of either bounded rationality or other-regarding preferences. Indeed, behavioral economics gained its worldwide popularity with the groundbreaking experiments of Kahneman and Tversky, who neatly demonstrated how our reasoning diverges from pure rationality. While this stream of research emphasized our shortcomings when compared to homo economicus – the super rational though utterly selfish prototype – the second wave, focusing on other-regarding preferences, emphasized our advantages over this prototype. A series of experiments demonstrated that we tend to cooperate, contribute, trust and bestow others more than the selfish homo economicus would. Both the negative aspects (bounded rationality) and the positive ones (other-regarding preferences) are focused on demonstrating deviations from the assumptions of classical economic theory and gain their fame (or attract criticism) based on this feature. However, behavioral economics is broader, both in terms of methodology and in terms of topics. My own research for example, considered by most to be an integral part of behavioral economics, is not experimental in terms of its methodology and not focused on deviations from the homo economicus model in terms of its topic.
Broadly looking at my research agenda, I would say that I develop theoretical models that aim to highlight the potential effects of self-image considerations on societal outcomes. Before explaining what I mean by that, let me note that this line of research diverges from the agendas described above because a person may be completely rational and selfish yet care about his or her own self-image. However, homo economicus is traditionally interpreted in the narrow sense, as someone caring only for material payoffs. Thus, considering self-image puts the theory in a clear social context, deeming it “behavioral” in the eyes of many observers. This social context is treated in the theory either directly, as when considering the effect of adherence to social norms on a person’s self-image, or indirectly, as when considering the effect of adherence to one’s own set of values in the course of social interactions. In the former case one judges oneself by the standards of others, while in the latter case one judges oneself by one’s own standards. To make things more concrete, I will turn now to discuss one example for each of these two aspects of self-image using simplified versions of my own models, while showing how these can be used to analyze social outcomes.
The first example considers the effect of a social norm, and social pressure to conform to that norm, on the extent of conformity of different individuals in society. A standard model incorporating this aspect goes as follows. Imagine a social or political issue that is under some controversy, where there exists a social norm, i.e., a consensual (‘right’) opinion or norm of behavior. Suppose now that each individual in society has some private opinion regarding this issue, and everyone needs to declare their stance in public (this could also be an action, like choosing what to wear to work). An individual whose private opinion differs from the social norm will need to consider the trade-off between a social cost, associated with the social pressure for violating the norm, and a psychological cost, associated with stating an opinion different than his or her private one. The social cost is naturally increasing in the distance from the norm to the chosen stance and the psychological cost is naturally increasing in the distance between the private opinion and the chosen stance. Standard models restrict both costs to be linear or quadratic in the relevant distances. However, my own model (coauthored with Daniel Spiro from the University of Oslo) relaxes this restriction, thus highlighting the effect of the curvature of these costs on inter-individual differences.
To see the potential effect of the curvature, suppose that the psychological cost is linearly increasing in the distance between the private opinion and the chosen stance of the individual, but the social cost is very concave. Very concave social cost means that even small deviations from the norm incur a high social cost, but larger deviations do not increase the incurred cost significantly. Then, on the level of the whole society, those individuals with opinions quite close to the social norm will fully conform to the norm, while those with opinions sufficiently away from the norm, who cannot fully conform because this would imply a too high psychological cost, will speak their (very deviant) minds in public. They will choose to do so because any smaller deviation from the norm would not make sense given that all norm deviations are punished similarly. However, what should happen if we reverse the modeling assumptions (think of this as a different kind of society or a different norm)? That is, suppose now instead that the social cost is linearly increasing in the distance from the norm to the chosen stance, while the psychological cost is very concave. Very concave psychological cost means that the individual is very reluctant to state a stance that deviates from his/her private one, but, once deviating (due to large social pressure), the individual is quite indifferent between small and large deviations from the private opinion. This time, on the level of society, it will be those with opinions quite close to the social norm who will speak their minds in public, while those with opinions far away, who cannot speak their minds freely because this would imply a too high social cost, will choose to fully conform. In other words, they will conform even more than those who have only slight disagreements with the norm, because the concavity of the psychological cost implies that once they do deviate from their private opinions they don’t care too much to what extent they do so. Comparing now the two cases, representing two different societies or norms, one may notice that in the first one the conformists are those with opinions quite close to the norm, while in the second the conformists are the ‘extremists’. Moreover, public (i.e., visible) deviations from the norm are large in the first case, where the extremists are those speaking their minds in public, while small in the second, where these are the ‘moderates’ who are doing so. More generally, such a framework allows us (after adding some complexity) to compare norm conformity and endogenous norm locations in strict societies and in liberal ones, where social costs of norm deviations are assumed to be concave in the former and convex in the latter.
Going back to the general theme of exploring the effect of self-image on societal outcomes, but now from the angle of adhering to one’s own values, my second example stems from other work I have done. This work considers the effect of reluctance to cheat others on the formation of social groups. First, note that reluctance to cheat others may be easily interpreted as a form of other-regarding preference, so a word is needed here in order to explain how I model it to capture self-image and not other-regarding. Consider a situation where a person is engaged in many bilateral relations with different individuals, in each of which this person may choose to cheat for the sake of gaining some material benefit. From a modeling point of view, if cheating is purely related to other-regarding, one should model the personal cost for the cheater as linearly increasing in the number of ‘victims’ – cheating 10 people hurts others twice as much as cheating 5 people. But if the personal cost of cheating has to do only with self-image, then the cheater cares only about the violation of his/her own value ‘do not cheat’, and therefore cheating 10 people incurs the same (or at least almost the same) cost as cheating 5 people. The way to model such a cost is to make it very concave in the number of cheated people, thus capturing the significant reduction in self-esteem when one becomes a cheater (at the extreme this will be a step function – some fixed cost for cheating any strictly positive number of people).
So consider now such a model, where people interact with each other bilaterally and have a very concave cost of cheating others. What effect could this cost have on the structure of society? Here the intuition is rather simple. If cheating each person gives a material benefit to the cheater, then cheating a high enough number of people makes cheating worth the accompanying damage to one’s self-image (as a self-test of your own ‘cheating threshold’ think of an opportunity to secretly bypass a queue of people standing in line for some service). This in turn sabotages any attempt for large scale cooperation (in the absence of enforcing institutions). It thus implies that people can sustain cooperation only in limited-size group. Hence, on the level of society, one should expect to find people forming separate groups and restricting their cooperation only to within-group interactions. This basic framework then allows me to analyze more complex situations, where people differ in their personal cost of cheating, and the inclination to feel bad because they cheat is private information, but I will leave this to a future post.