Monday, 5 October 2009

Can feelings be modelled?

Karl at Modeled Behavior responds to Richard Posner (I am surprised to hear Posner emphasising psychology - I thought he was talking about Richard Thaler at first):
I am not sure exactly what Posner means by psychology. I have never been optimistic about feelings as an economics model. Perhaps people spend more when they feel better but how do we get a consistent measure of feelings and even more to the point how does policy consistently effect them.

What I do think makes a difference is expectations.
Expectations do fit nicely into macroeconomic models because they bear a simple relationship to real events. Macro models mostly deal with visible events such as monetary transactions. And an expectation is just an event, translated in time and perhaps given a probability weighting.

But we do model things in economics that are not that simple. The most obvious example: preferences.

Preferences are very complex - just think of all the factors that impacted on your choice of what dinner to have last night. How much money you have in your pocket, whether you're going to a movie, what was on TV, what you have left in the fridge, which foods go together, the expiry date on that piece of chicken, the leftovers from the previous day versus your desire for variety, calorie content and other nutritional values, whether you were still hung over from Saturday, the special offer at the local restaurant, whether you saw that Domino's sponsorship of the X Factor and simply what you felt like eating at the time. And that's just one decision out of dozens you make every day, multiplied by billions of people worldwide.

Clearly it would be very hard to model all that. But microeconomists still have some useful ways to model preferences. The most common is rational utility functions.

Of course rational utility functions do have some drawbacks: there are many aspects of behaviour that they fail to describe accurately. But behavioural economists are busy incorporating a more sophisticated understanding of decision making into this utility model. The basic concept is not a bad one.

The Arrow-Debreu theorem is an extrapolation from the idea of rational utility functions to describe how free markets bring about a position of maximum welfare for a society. These simple models of preferences do provide powerful theoretical outcomes.

So if we can model something as complex as preferences, couldn't we also build a model of feelings?

A cognitive model incorporating feelings is definitely possible to build. The question is how to choose the foundations of the model so that it still accurately describes behaviour, and is also mathematically tractable when extended to the level of a goods market, and then to the macro level of a whole economy.

While I don't have an immediate answer for that, the (relative) success of models based on utility functions shows that it is often possible to find a suitable model that can work at both levels.

One challenge is the divide between micro and macroeconomists - the micro people are better at developing these types of models but the macro crowd are the ones who need to use them in this case. We need people willing to work across both fields.

Another key is to have not less, but more, understanding of mathematics in economics. The application of mathematics to psychology is what led to many of the historical successes of the economics discipline. A deep understanding of underlying mathematical principles, not just a knowledge of how to apply them, is needed. This will let us develop unified economic models to transcend the splits between Keynesian, new Keynesian, monetary, new growth theory and Austrian macro.

No comments: