I actually really disagree with Paul Krugman this time

Krugman argues that the Bank of Englands worst case scenario for no deal Brexit is implausibly bad. I agree with his conclusion, but strongly disagree with one argument (on a point which he stresses is quantitatively minor)

… the BoE includes some nonstandard effects of trade: they assume that reduced trade (and foreign direct investment) will reduce productivity more than the direct impacts on resource allocation would predict. They cite some statistical evidence, but it’s important to realize that this is black-box, reduced-form stuff: there’s no explicit mechanism through which it’s supposed to happen.

However, these assumed nonstandard effects aren’t what’s driving the really bad scenarios; they only, as I understand it, contribute something like 1 percentage point of GDP to the predicted costs.

[skip]

On the substance: I’m skeptical about the supposed effects of trade on productivity. I know that there’s some evidence for such effects; trade seems to favor more productive firms. But relying a lot on effects we can’t model seems dubious.

In particular, I have strong memories of the openness-growth debacle of the 1990s.

I comment.

To me sentence “But relying a lot on effects we can’t model seems dubious” seems dubious. What do you mean “we” bright man ? I am willing to bet you could whip up a model where trade causes high productivity growth within 15 minutes.I am not willing to bet on you against you as being the guy who bet he couldn’t do it would creat a bit of conflict of interests.

I will attempt to do it in 30 minutes (OK I have begun thinkin already).

Then against data you have an example (one (1)). I can think of many debacles of people who decided not to rely on an effect because they couldn’t model it.
1) we can’t explain why nominal stickiness might be optimal & in our models firms maximize profits. The claim is true. Menu costs don’t do the trick as firms synchronize. Calvo fairies are embarassingly implausible. Akerlof said near rational (not optimizing). So they decide they must assume prices are flexible and we get a RBC debacle.
2) “zero isn’t an especially important number” Paul Krugman 1988 (at 1050 Mass avenue). There is no reason why people should accept constant nominal wages with 2% expectable inflation and not accept a 2% wage decline with 0% expectable inflation. So it can’t be true. But it is.
3) There can’t be a liquidity trap because of the Pigou effect. Also there is Ricardian equivalence. No one noticed that Pigou and Ricardo contradict each other until … *you* remember when — it was in the 1990s (and that example is *not* an elephant).

Over at the New York Times I ran out of allowed space so I will continue here with examples after the jump.

But now I wan’t to start a clock. Trade causes higher productivity growth in the model which I will present in 30 minutes or less.

4) Simple models said that the minimum wage caused markedly lower employment of low skilled workers and didn’t help the people it was supposed to help. Data suggested otherwise. Less simple models were easily developed.

5)The same simple models said either the minimum wage must redduce employment a lot (because the slope of the labor demand curve is low) or migration of unskilled immigrants must cause the wages of the unskilled to fall a lot. Data disagreed again. It is hard to fit both the small effects of minimum wage increases on employment and the small effect of the Mariel boatlift on wages in Florida. But, and here I play my trump Card, both facts are facts.

More generally, some empirical work is no good (ok most is no good) but some is based on serious efforts to find natural experiments (or is based on actual experiments) and is useful especially when the results puzzle theorists.

That’s just economics. There are also much more dramatic cases of physicists who didn’t believe something happened because it was logically impossible. But it keeps on happening.

My maxim is that if you think theory is telling you something, think again. Theory in general can’t tell us anything. It is the assumptions made for convenience or habit which tell us things. The things they tell us are never reliable. Fit the facts. If there is a model which fits so many facts that you can’t remember them except as the bunch of facts fit by the model, then fit the model. This is not elephant to economics as we sure don’t have any such models.