Where behavioural economics goes wrong
We talked yesterday about the general issues that Wright and Ginsburg raise in their critique of behavioural economists. I don’t agree with their overall conclusions about the dangerous incoherency of the “behaviourists’ agenda”–if such a thing even exists–but they do make some excellent points about the limitations of the science as it stands.
The first stage of the behavioral economics research program is best described as developing a comprehensive theory of errors. The theory-building exercise thus far has focused largely upon the effort to catalog circumstances in which economic decision-makers appear systematically to depart from rational choice behavior. The second step required to make the theory of errors policy-relevant is to map the conditions under which specific errors are more or less likely to affect decisions and then to generate estimates of the social costs imposed by those errors. This step is particularly important when the incidence of a particular decisionmaking error is context specific, unevenly distributed throughout the population, and likely to interact systematically with other errors. The third step is to compare the costs of any proposed corrective intervention against the social benefits produced by reducing the rate of error. At present, however, research in behavioral economics does not appear to have moved much beyond the first step.
I think this is true, and similar to the point Eric made in his original comments. However, it is not a criticism of the science, but merely a statement about how young the discipline is. From the discussions so far we can add a couple more points related to policy development:
- Policy conclusions need to be drawn from causal models, not correlations. That is the driving force behind the need for a coherent model supporting the idea of cognitive bias.
- The behavioural approach needs to be applied to regulators, too. Public choice was a fairly late development in microeconomics so it’s not surprising that behavioural economists haven’t got there yet, but it is certainly a development that needs to occur so that we can think more deeply about policy failure.
There is a secondary problem in the first stage of the BE research programme. There are two opposing views of these errors. One view sees them as rational, optimising responses to newly discovered constraints. That is, I rationally minimise my cognitive effort because it is expensive, which leads my to make certain kinds of errors. The other view sees these behaviours as regularities that are not driven by optimisation, but by biology or sociology.
This is important for the second stage. policy mapping, because it determines whether economists see the problem as constrained optimisation (and therefore probably socially and personally optimal) or as truly ‘behavioural’.
Seemed to me they exaggerated the extent to which behavioral measures infringe on individual liberty. For example, the explicit statement that “the behaviorists’ regulatory toolkit includes … overt coercion” for which no example or evidence was offered.
If it’s overt coercion, then it is not a behavioural measure, it is regulation. The whole point of behavioural measures is that they encourage but do not enforce behaviour which is thought to be in the individual’s best interest.
Their point that regulators cannot be certain that they know the individual’s best interest any better than the individual does is far stronger.
I think they exaggerated quite a few things so I’ve queued up a few posts on those issues. It is really nice to see them put the views of BE’s opponents so strongly, though. And, as you say, some of the points are well made.
One of my worries is that embarking on policy for behavioural-justified reasons puts us quickly on the path to hard coercion when the behavioural targeting fails to achieve the desired end. Politics doesn’t handle sunk costs well. So it’s easy for “We think X failing leads to bad outcomes on Y, so we’ll try B non-coercive policy” to turn into “We have failed to achieve our common goal of Y and so stronger measures are warranted.”
Examples? Look at how quickly “Oh, imperfect information is what makes people smoke too much” turned into “Let’s force people to quit smoking by eradicating tobacco by 2025”. All kinds of slippery slopes in paternalistic policy.