Experiments in behavioural economics

In this chapter. I talk about:

  • The rise of experimental economics; how it came to become part of mainstream research in economics;
  • The similarities and differences bet ween the approaches adopted by economists and psychologists;
  • The history of experimentation in economics;
  • Issues regarding experimental design;
  • Criticisms of the experimental method, including issues of external validity;
  • Experimenter demand effects;
  • The prospects of experimental economics in the coming years.


So, you want to figure out why someone did what they did or what someone might do when confronted with a particular circumstance or choice? What do you do? Traditionally, researchers have followed two different paths. The first of these is to rely on surveys. This is essentially the same as asking people. Surveys are straightforward and usually yield valuable insights. But at the same time, there are drawbacks to this approach as well. The problem is that sometimes peoples response to what they would do in a particular situation does not predict accurately what they would really do when actually placed in that situation; that is, there is a disconnect between attitudes (stated preferences) and behaviour (actions). This implies that, at times, people’s attitudes do not correlate well with their behaviour. It is also the case that we are not always good at articulating our reasons.

This essentially means the following: suppose I asked you whether you were willing to contribute $50 for a good cause and you said yes. But when eventually the envelope gets passed around and you have to actually part with the money, you may renege on that promise completely or put in less than $50. I am not saying that you will do it, but it has been known to happen. Moreover, responses in these questionnaires may differ substantially from behaviour, not because the respondent is trying to mislead the researcher, but because the respondent may possess an incorrect perception of his/her own and of others’ views or reactions. That is, the respondent might honestly think that he/she will behave in a certain way in a particular situation, but when that specific situation comes to pass, he/she actually behaves differently.

Here is an example of such dichotomy between attitudes and behaviour, taken from the literature in social psychology. In the early 1930s, Richard LaPiere wanted to discover if people who had various prejudices or negative attitudes towards members of other ethnic groups would actually demonstrate these behaviours in an overt manner. For approximately two years, LaPiere travelled around the USA with a young Chinese couple. They stopped at 184 restaurants and 66 hotels. They were refused service only once, and, on the whole, received a better than average standard of service from the establishments visited. After returning from two years of travelling around, LaPiere wrote to all the businesses where he and the Chinese couple had dined or stayed. In a letter, which gave no indication of his previous visit, he enquired whether they would offer service to Chinese customers. While virtually none of the establishments had actually refused service, in the survey, the majority expressed the opinion that they would not serve the Chinese visitors. There are many other examples of such dissonance between attitudes and behaviour.

The second avenue of exploration, as opposed to relying on survey questionnaires, has been to look at naturally occurring held data generated by a real-life economic phenomenon. That is, if you wanted to understand whether and why people contribute to charity, then you might dig up data on charitable contributions and analyse those data. This has been the more traditional and usual approach in economics. In order to understand behaviour, one needs to look at data that pertain to a particular phenomenon. For instance, suppose we want to know the impact on unemployment of an increase in the minimum wage or large-scale migration into a particular region.1 In such cases, the recourse is to compare two otherwise similar regions, one of which has recently experienced an increase in the minimum wage (via new legislation) or a sudden influx of new immigrants as opposed to a second region that has not experienced anything similar.

In fact, the famous American economist and the recipient of the Nobel Prize in 1970, Paul Samuelson, wrote in his undergraduate textbook (which, until recently, was the most popular text in universities, not only in the US but across the world):

(e)conomists cannot perform the controlled experiments of chemists or biologists because (they) cannot easily control other important factors. Like astronomers or meteorologists, (economists) generally must be content largely to observe.

Richard Lipsey, prominent economist from the London School of Economics, makes the impossibility of experimentation in Economics even more explicit:

Experimental sciences, such as chemistry and some branches of psychology, have an advantage because it is possible to produce relevant evidence through controlled laboratory experiments. Other sciences, such as astronomy and economics cannot do this.

Given this non-experimental view, economists have traditionally adopted a more theoretical approach that relied on building mathematical models of behaviour in order to explain, understand or predict behaviour in a variety of economic transactions. These models start from a series of ex ante assumptions that are typically based on the researcher’s intuition about the state of affairs. Then they go on to make predictions about changes in behaviour that would result from those underlying assumptions. The success of such models is measured by their internal coherence. In fact, the 1986 Nobel Laureate, Milton Friedman, suggests that the assumptions made by economists in building theoretical models do not propose to represent how the world works exactly. They merely proceed on the assumption that these are "as if” propositions, distilling regularities in behaviour that happen to be useful in deriving predictions. Therefore, even though a lot of theorising in economics depends crucially on the assumptions we make about individual preferences and behaviour, these assumptions should not be treated as empirical hypotheses to which the theory is committed.

This is also because any attempt to test theoretical models is subject to the Duhem— Quine problem. Since theories must be applied in a specific context in order to test them, it is virtually impossible to test a single theoretical hypothesis in isolation. Researchers need to make a series of supplementary assumptions in order to apply the theory to the context within which it is being tested. Consequently, if the theoretical postulates are not borne out by the data, it is often difficult to disentangle whether the hypothesis itself is incorrect or whether the problem lies with one or more of the supplementary assumptions.

This prompted Vernon Smith to argue:

Consequently, we come to believe that economic problems can be understood fully just by thinking about them. After the thinking has produced sufficient technical rigor, internal coherence and interpersonal agreement, economists can then apply the results to the world of data.

As a result, often economists were not overly concerned with empirical validation of the assumptions or predictions of theoretical models. Even when empirical validation was sought, it was usually via finding a natural experiment that might generate data suitable for testing a particular theory." However, one problem with field, that is, naturally occurring, data is that such data may not always be available, or may not be available in the exact form that is needed to answer a particular question. Moreover, since the data are generated by a one-time economic phenomenon, they may not necessarily be in the form that allows us to make causal inferences; that is, whether a particular phenomenon, X, caused another phenomenon, Y. A natural experiment is also impossible to replicate.

< Prev   CONTENTS   Source   Next >