Monday, July 25, 2011

More on macroeconometrics

The other day I said that macroeconometric role models should be people like Robert Barro and Christina Romer, who understand the issues surrounding identification of macroeconomic models. Of course, there's one major problem with using a lot of this work in practice, which Jonathan Parker points out in a new NBER working paper. From the abstract:

"We do not have a good measure of the effects of fiscal policy in a recession because the methods that we use to estimate the effects of fiscal policy — both those using the observed outcomes following different policies in aggregate data and those studying counterfactuals in fitted model economies -- almost entirely ignore the state of the economy and estimate 'the' government multiplier, which is presumably a weighted average of the one we care about — the multiplier in a recession — and one we care less about — the multiplier in an expansion. Notable exceptions to this general claim suggest this difference is potentially large."

His solution is a more careful look at general equilibrium effects. This is similar to my response to the Conley-Dupor paper, where I suggest that some sense of the magnitude of interstate trade would help us understand exactly what the Conley-Dupor results are telling us. This also smacks of Old Keynesian econometrics to me: identify all the micro-reaction components (marginal propensities to consume, import, etc.) and then slap it all together to get a rough estimate of a fiscal multiplier.

Andrew Bossie advocated a mixed-methods approach to multipliers, and to a certain extent I agree. We can't just embrace all clearly biased estimators and call that "mixed methods", but we can take weighted averages of multipliers like Barro's and Romer's, and compare that to cross-national studies, and studies that build a multiplier up from its constituent parts, etc. I just don't like the idea of taking an estimator we know is going to be biased - like these cross-state studies - and throwing them into the pot, just assuming that the average that we get out is going to be useful.

Parker suggests: "Microeconomic estimates of the partial-equilibrium causal effects of a policy can discipline the causal channels inherent in any DSGE model of the general equilibrium effects of policy. Microeconomic studies can also provide measures of the dependence of the effects of a policy on the states of different agents which is a key component of the dependence of the general-equilibrium effects of fiscal policy on the state of the economy." It sounds like calibrating a DSGE model instead of the traditional calibration of an Old Keynesian model.

1 comment:

  1. To clarify a little. I prefer "nonparametric" and other techniques that estimate things other than the mean effect. I'm a big fan of quantile regressions.

    My mixed-methods approach is definitely not to look for a mean among studies but rather to approach the whole thing less formally and think about things in a "general sense" since I don't think econometrics really produces capital T truth anyway.

    Also, have you seen this paper? It's an example of trying to estimate fiscal policy effects that isn't simply a mean effect. It's in a framework I think you are more comfortable with:


All anonymous comments will be deleted. Consistent pseudonyms are fine.