I generally agree but I guess why I ask this about behavioural economics in particular is that the application of broader models to to specific human behaviours is that that is where things have gone so very very wrong.  I suspect we’d be better off in many cases if we asked how will this intervention work out and tested the relevant behaviours robustly rather than drawing too long a bow from other models. 
 

Good question, and one that I’m probably not qualified to answer. Matt probably has some deep and meaningful thoughts on the matter, but I’d say yes. Forming a consistent model of choice allows us to generalise beyond the experiment. Essentially, we want to be able to abduce from the results of the experiment the reasons for the behaviour we observe. From there we can make predictions about other behaviours and use our model to draw conclusions that can’t be deduced from the experiment itself.

… and that’s why it worries me when a not insubstantial proportion of papers in behavioural economics draw policy implications without considering comparative institutional analysis….

I’m not saying that we should shut down the behavioural programme. Rather, we should be a bit more reticent to advance any policy conclusions from it until it’s matured a bit. 

I wholly agree that you can’t point to a cognitive bias and then say it justifies intervention. That’s like pointing to the failure of the conditions for Pareto optimality and then saying one should obviously intervene. However, that feels more like a critique of policy-makers, rather than an indictment on behavioural economists.

In many ways, behavioural economics feels like primary data gathering at the moment, rather than a mature theory that could be used to justify intervention. It’s natural that people will seize on some results to justify their prejudices, but we shouldn’t let that hinder the progress of the science!

I’m not saying dismiss behavioural results. I’m saying that we should set a high bar on providing policy advice based on those results. Lab experiment results ought to be confirmed by field experiments. And we ought to have a close eye on comparative institutional analysis: will the bureaucrats running the programme be less subject to behavioural failings?

Behavioural econ can give some great advice as in self-help books – things to watch for in your day to day decision making. I get nervous when it’s applied too quickly to policy-making. Especially when you can point to a purported lab behavioural failing to ex post justify just about any damned thing you like.