Chinn and Thoma hit it out of the park

Robert Waldmann

Thinks that this post by Mark Thoma discussing this Post by Menzie Chinn is unusually excellent. I suggest you go read it. If interested my thoughts on it after the jump.

ideology. Its just not done to write that the debate is ideologues on one side non-ideologues on the other. In fact, he lists two ideologies. However they are both anti-stimulus bill ideologies.

Furthermore they go together not just in congress but in the economics profession. The debate among economists prominently included the views of people who think government spending is horrible (except they don’t complain about defense spending) and the market is perfect so it should be left alone and whose views haven’t been noticibly affected by the recent downturn.

I think the form of Chinn’s sentence suggests extreme reluctance to say this but, among economists, there are hard core ideologues on one side of the debate and not the other (which isn’t to say all stimulus critics are hard core ideologues).

Chinn might get himself in trouble by saying RBC models can’t fit moments, but I doubt anyone will notice the deadliness of the line “even H-P filtered.” He briefly mentions an extreme weakness of RBC models. The data are H-P filtered markedly reducing variances and covariances, the output of the models are not H.P. filtered.

In my personal experience it is hard to fit the moments H.P. filetered data, but it is much harder if the output of the model must be run through an H.P: filter too. Is there any excuse for not running the output of models through H.P. filters.

Let me quote Tom Sargent “Ed Prescott thinks growth theory should be used to study everything but growth.” It is very very strange that stationary growth models are written and compared to H.P.filtered data. Wouldn’t it be more natural to write down a growing growth model then H.P. filter it and compare it to H.P. filtered data ?

Third point. Chinn asserts that parameters are chosen so the model’s output fits the data. Prescott claims that parameters are not chosen so the model’s output fits the data, but rather based on independent data (micro data and trends), Chinn confidently asserts that, when Prescott claimed this, he lied.

Now on multipliers from RBC and DSGE models. An assumption of many of these models is that aggregate consumption corresponds to that of a rational representative consumer with no liquidity constraint. This is a feature, an important part of the GE ness.

It also implies that tax shifts affect output only via incentives on the supply side, that is, that a multiplier is zero. So, sure they are not suited to the evaluation of a proposed stimulus involving tax cuts and deficit spending.

But what was the advantage of this assumption ? The models certainly don’t fit the data on consumption. People still argue that rational intertemporal substitution with no liquidity constraint is consistent with the data (often even though the data analysis in the paper with that conclusion rejects that hypothesis) but only by arguing that there are extremely complex non separable utility functions which are not found in real or New Keynesian DSGE models. So why is it a feature to make an assumption which is rejected by the data ? Why is it a bug to make an assumption about consumption which is, at least, not demonstrably false ?