Saturday, March 22, 2014
A grand unified theory of behavioral economics?
Tim Harford has a great article about the backlash against behavioral economics. You should read it.
There are two main knocks against behavioral econ. These are:
1. There is no "grand unified theory" of behavioral econ; instead it's a bunch of specific little theories for different situations. What we want is a widely applicable, unified theory of economic behavior.
2. The behaviors produced in a psychology lab are extreme effects produced by extreme, coordinated manipulations; in the real world, lots of stuff is going on, and what we care about is the average.
I never really understood Critique #1, because maybe there just isn't a grand unified theory of economic behavior; maybe it really is just a bunch of little situational things. But #2 does indeed seem like a big problem, because if you don't have external validity, you're spinning your wheels.
Now a thought occurs to me. What if Critique #2 contains the secret to answering Critique #1?
Suppose that there are a whole ton of different behavioral biases, and that these vary across time, across people, and across situations so much that even with a billion lab experiments we couldn't find them all. Only once in a while will the forces be aligned to make one behavioral bias dominate; most of the time, the net effect of all the biases will be unpredictable by the outside observer. When you have an unpredictable mishmash like that, you have to model it as a stochastic process. In other words, if it's too complicated to explain deterministically, then you treat it as randomness.
So what if psychology usually just ends up injecting randomness into our decisions? What would that theory look like? I think it would look like a Random Utility Model.
For those who don't know, Random Utility discrete-choice models are one of economic theory's great success stories. They've been around for over forty years, they're widely used in the private sector, and they've already won a Nobel Prize for their creator, Dan McFadden (currently of UC Berkeley). In a legendary example of theoretical badassery, McFadden used the model to predict the number of people who would ride the new BART train in the San Francisco Bay Area. The government predicted that 15% of Bay Area commuters would use BART; McFadden's model predicted 6.3%. The actual number? 6.2%.
(Next time someone tells you that "economic theory can't predict anything", tell them about McFadden and the BART.)
A Random Utility model treats human decision-making as if it has two components - a predictable, deterministic component, and a random component. But if there are a huge jumble of behavioral effects going on, it seems to me that outside of the lab, that's usually just going to be observationally equivalent to randomness in the objective function. Which is exactly a Random Utility model.
So what if Random Utility models are the grand unified theory of behavioral economics? What if the upshot of all of these psychological effects is simply that the random component of utility is partially irreducible - that there must be a random component of utility in almost any theory if that theory is going to have a chance of predicting human behavior accurately? Maybe sometimes the randomness is so negligible that people act like homo economicus, and sometimes the random part dominated so much that their decisions are completely unpredictable with science?
In other words, maybe the Grand Unified Theory of Behavioral Economics was invented and validated more than 40 years ago, and we just didn't recognize it for what it was?
Anyway, I'm sure someone has thought of this idea long before me, and I'm sure it won't be long before a commenter provides me with a link. But old or new, it's still a really interesting idea, which I think has the potential to answer both of the main criticisms of behavioral economics.