Is 'behavioral economics' not related to behavior?

[ZomgRogged]

Walker and Wooders have a really clever paper on Mixed Strategies in Wimbledon. The paper asks do experts with high stakes (finalists at one of the biggest tennis tournaments) mix play (serving left and right) better than amateurs in a lab. It concludes that at the aggregate these players are playing minimax. Is this what you mean?

[Diplomer]

Originally Posted by ccgoal

I would like to hear you guys' opinion on behavioral economics (BE). Though under the name of behavioral, I see little difference between BE and Neo-classical style in terms of the ability to solve complicated optimization problem of agents. (...)

I guess one of the reasons it to get published in mainstram journals using such math. But other than that, do you think it's really necessary to use math like this just to describe the real behavior, which is simple and random in most cases (I think)?

A very important difference between "neoclassical" approaches, although I would prefer to call them standard approaches, is the following (at least MY interpretation of what is going on): Stuff like expected utility, standard utility functions (no reference point, no loss aversion, ...) and so on, are rather normative concepts. In some sense, neoclassical questions are questions about how rational people (rational for given a specific preference structure) should behave. What is the optimal behavior?

To the contrary, Behavioral Economics want to built up a descriptive theory of human behaviour, i.e. how do people really behave?

Given the above, mathematical modeling is not only something you do for getting into journals, no, since BE is nothing more than a generalization of standard economic theories (explaining not only what rational people should do, but what people acutally do with the former often as a special case), there is virtually no reason why sb should expect that there is no mathematical modeling.

By the way, people's deviations from neoclassical predictions can be called systematic. Taking Friedman's methodological stance, we virtually have to change the theory since its predictions are inaccurate.

Is there any research that shows the relationship between the (neo-classical) modelling and the actual behavior? By this I mean on the one hand we make our decisions in the real world almost doing no calculations (Bayes Rule, complicated probability and so on), while on the other hand, in economic models, each step is math intensed. Without such connection, how can the results of researches that build on these models with smart agents (even for many BE models as mentioned) be reliable?

Well, this would be experimental economics, right? The problem, of course, is how to refute a theory? As we are dealing with unobservable concepts like utility and an a priory non-specified utility function, its hard to refute neoclassical theory. One example, imagine one of your friends would start to bumb his head against a wall, several times. What is the neoclassical answer? Well, of course, this guy must have "bumbs on the head" as an argument in his utility function. His activity, therefore, is utility maximizing behaviour. You get the problem?

Of course, in an experiment, we can control the setting and monetary payoffs. However, we do not know what kind of game we are actually playing since the monetary payoffs might be different from the utility the agents get out of them. So, we always have to assume certain "bridge principles" (Guala, "Has Game Theory been refuted?"). The big question now is: what do we refute? The theory, or the accuracy of our bridge principle? Moreover, a stance I would like to take, we might argue whether our theories can be tested at all since they are mathematical tautologies.

However, back to your initial question. Sometimes the outcomes converge to the outcome predicted by the neoclassical theory as the number of rounds increase (e.g. public good experiments). [But interestingly, if we introduced something like the possibility to punish, the contribution acutally increases (see Fehr, Gächter, altruistic punishment). Self regarding people would not do this. ;-) ] The "discovered preference" point of view (i.e. with experience the participants learn about their preferences and thus there is a drive towards the neoclassical outcome) has one problem: aren't most of our economic decisions one shot decisions without the possibility to repeat them 40 times? If this were true, would it not be reasonable to search for theories that explain these first decisions (from a descriptive point of view)?

Here we are again back to the first part, namely: what do we want? Do we want to look for optimal behaviour (neoclassic approach) or do we want to describe what really goes on in the world (BE)?

My 2 cents
Diplomer

[smart_ana]

I can put this (maybe inanely) simpler:

BE likes to look at what might normally be considered "irrational behavior" under the Neoclassical model and try to predict this irrational behavior. Just because a behavior is an irrational one, doesn't mean that it's random and cannot be predicted. Case in point: I procrastinate on my homework until the night before no matter how busy I am or how far in advance I'm assigned the hw. Under BE I'd be a sophisticated (meaning aware of self-control problems) individual with time inconsistent preferences. You could theoretically model my behavior even though it could be called irrational.