Stephen Gordon applies the Lucas critique to behavioural economics and nudges:
A key insight of behavioural economics is that people don’t always and everywhere re-optimise whenever their environments change. Instead, they will often – or even usually – make use of various rules of thumb and/or passively accept the default option. …This the idea behind ‘nudges’: you can alter people’s behaviour by making minor changes to the frames in which people operate…
But this only works if the change is subtle enough to not attact the full, direct attention of the decision-maker. If the change is big enough, people will haul out the full artillery of their rational selves in order to try figure out what optimal decision is. This means that behavioural economics is unlikely to be of much use in policy-making.
What he’s saying is a policy that takes advantage of people’s rules of thumb relies on them using the rules of thumb. If the policy is a big enough change in the circumstances they face then they might change their rule of thumb, which changes the effect of the policy. Unless you know how people form their heuristic, you can’t target it with policy.
For example, setting organ donation schemes to be opt-in by default may increase the level of donation because people don’t have a strong preference over whether they donate after death. If you applied the same reasoning to a decision that people care deeply about then it might have no effect because people will make the effort to choose their preferred option.
It’s an interesting point but I think Gordon goes too far by suggesting that it invalidates the targeting of behavioural heuristics. In the case of defaults the problem is that the policy-maker doesn’t know if changing the default will change the heuristic. If it doesn’t change the heuristic then people are hopefully better off because of the improved default. If it does change the heuristic then people are optimally choosing their preferred option so the policy is socially costless.
Since ‘nudging’ policies don’t restrict choice sets there is unlikely to be a significant down-side beyond any additional administrative burden. The goal of the policies is to help people make the decisions they would have made if they were “haul[ing] out the full artillery of their rational selves”, so the fact that they might do that in response is not a policy failure.