Let’s get fiscal

Three months ago, Christine Neill wrote on Core Economics:

If you ask a reduced-form applied microeconomist like myself whether more government spending decreases unemployment, we have two initial instincts:(1) do some experimenting with an entire economy (perhaps not completely impossible, but not quite there yet) or (2) find a situation where one part of a country gets more $ spent on it than another part for reasons unrelated to their current or future unemployment rates, and compare the two.

A nice approach would be to look at changes in spending caused by purely political factors. So if, for instance, a lot of money is poured into swing electorates just before an election, you could look at whether unemployment fell more in that area than others.

This is tough to do in the US – there’s surprisingly little evidence of the feds there directing resources on political grounds in ways that affect aggregate spending statistics. Because of that, going the extra step and figuring out whether unemployment is affected by such spending is very dodgy.

Perhaps people who’ve looked at pork barrel spending in Australia or Canada might consider extending their work to look at the effects on local unemployment rates?

Well, the offer was too good to resist. Christine and I have now written a little paper that exploits variation in pork-barrel spending under the Coalition’s “Roads to Recovery” scheme. We find that more road spending did indeed reduce unemployment – a result that is potentially useful in the current policy context.

Does Fiscal Policy Reduce Unemployment? Evidence from Pork-Barrel Spending
Andrew Leigh & Christine Neill
Studies of the effect of government spending on unemployment are potentially confounded by reverse causality. To address the endogeneity problem, we exploit variation in a pork-barrel road-building program, and find that higher government expenditure on road-building substantially reduces local unemployment.

Incidentally, I think this is now the fifth paper that I’ve written which emerges from a blog-dialogue, suggesting that blogging can have unexpected spin-offs for researchers.

(xposted @ Core)

This entry was posted in Macroeconomics. Bookmark the permalink.

1 Response to Let’s get fiscal

  1. Simon says:

    I think this is a really nice paper. For me, the most interesting result is that the IV is so much higher than the OLS. I wonder whether more could be done — for example, in the paragraph running across pp.5-6 — to frame this in terms of the distinction between an ATE and a LATE. I’m actually not convinced that the IV is valid — local unemployment is surely important in determining whether the ALP or Coalition holds a seat — but I still think it’s very interesting as an estimate of a LATE. (For example, I THINK it’s suggesting that the program was much better targeted in government-held areas…yes?)

    As a general comment, I think we should be cautious about drawing general macroeconomic conclusions from any kind of applied micro work. Your conclusion says “…we are able to estimate plausible causal impacts of local area expenditure on unemployment”. However, I don’t think it’s meaningful to speak generally of “local area expenditure” — every macroeconomist would remind us that different kinds of expenditures will have very different multiplier effects, etc. However, I do think it’s a really interesting estimate of the unemployment effects of a local road-building program!

Comments are closed.