How do we deal with evidence? Badly.

We deal very badly with empirical falsification of our reference model:

Many of these assumptions are unrealistic but they are justified as a way to set a benchmark model around which one is then allowed to model deviations from the assumptions.
…
the problem is that the model becomes (or has become) the reference in a way that sets a high burden of proof for any deviations from it. If you think individuals are not rational, go ahead and model their behavior but you should do it in a way that is realistic and backed by data (good luck). If you want to allow for government spending to be productive, you can do it but you better have perfect econometric results that prove that returns to government investment is indeed high. Given that experiments are not possible in economics, it will always be very difficult to produce supporting evidence for some of these assumptions that is not controversial. So what do we do? We claim that we do not know enough about the real world, we claim ignorance, and that would be ok except that we do not stay quiet when we are asked for our opinion. We go back to the results of our benchmark model and if someone asks us about a relevant policy question we use them to justify our answer. Do we care about the fact that the assumptions of that model have never been proven and have no connection to reality? No, we don’t.

It reflects a thread that runs through much of our comment on TVHE: we need to be transparent about the limits of our knowledge and recognise the limits of our advice. The difficulty is that the world isn’t going to wait for the profession to figure stuff out before making a decision. Managing the tension between a demand for simple, clear advice and the supply of imperfect results isn’t easy.