Tuesday, October 25, 2011

Evidence: Stimulus spending creates jobs

Delightfully, it is evidence from a Coalition program

It’s been one of the most fiercely-fought debates of financial crisis, conducted mainly in the absence of evidence: Do government stimulus programs create jobs?

The Coalition has derided Labor’s stimulus programs as “wasted money”. Labor says without the tens of billions it spent building school halls, insulating home roofs and sending cheques to taxpayers Australia’s unemployment rate would have skyrocketed.

Today the journal Economic Letters, publishes an attempt at an answer.

Economists Christine Neill and Andrew Leigh have been able to do what those before them have not: measure the effect of stimulus spending on individual regions by comparing those that received stimulus dollars with those that did not.

Christine Neill is an Australian at Wilfrid Laurier University in Canada. Andrew Leigh was until recently an economist at the Australian National University and is now a Labor MP. He completed the research while at the ANU but the long delays in the journal process mean it has only now been published.

The usual problem in such a study is finding a controls - regions as economically depressed as those that received government payments that had to make do on their own...

It can’t be done for Labor’s stimulus payments. They went to all Australians who fitted the financial criteria and to all regions throughout the country.

But Dr Leigh discovered this wasn’t the case for an earlier program run by the Coalition.

The so-called Roads to Recovery program begun in 2001 directed money for roads to some local councils and not to others.

Dr Leigh found “clear evidence” the money wasn’t evenly directed to regions that needed it. Electorates held by Liberal and National MPs got more funding than those held by Labor.

By using the poorly funded but economically-similar Labor electorates as controls he was able to work out what the big licks of money directed to National and Liberal electorates actually did.

His finding: a 10 per cent increase in stimulus spending in an electorate creases an extra 26 to 78 jobs. The cost per job amounts to $10,000 to $31,000 over a three year period.

“That’s not that expensive,” Dr Leigh told The Age. “At times when the economy is depressed such as during the global financial crisis you would expect the effect to be bigger.”

“As far as I know we are the first to use this method. It has gone through the peer-review process in an internationally recognised journal, so I think it can withstand criticism.”

Published in today's Age

Email from Christine Neill:

If you want to know whether in some particular instance government spending can reduce unemployment, you've got to worry about the fact that an increase in unemployment often causes government spending to increase (eg welfare spending, etc). So if you just look at correlations in the raw data, you will often find that higher government spending is associated with higher unemployment. In econometrics terms, there's an endogeneity problem. There are all sorts of techniques used to get around that problem in the macroeconometric literature (policy experiment case studies (wars studies), vector autoregressions, or various calibration-type techniques). All are somewhat contentious. A typical microeconometric approach is to find an instrumental variable - in this case, something that is correlated with a change in government spending, but that is not itself likely correlated with a change in unemployment. Here, we used politics. Previously, Andrew Leigh had found that electorates with (then government) National/Liberal reps got more spending in the Roads to Recovery program. We find that those electorates also had a bigger drop in unemployment than other electorates, suggesting that the higher spending in an electorate via Roads to Recovery led to lower unemployment in that electorate. While you can't extrapolate from this to overall fiscal policy effects (including because you wouldn't expect any crowding out via higher interest rates, which is a concern with national-level fiscal policies), it is interesting that in even a very small jurisdiction government funded local infrastructure spending seems to be sticky, and to increase local employment.

There are a couple of other recent papers that try to take the same basic econometric approach: Nakamura and Steinsson: "Fiscal Stimulus in a Monetary Union: Evidence from US Regions" (uses military buildups in particular regions); and perhaps more interestingly: Mafia and Public Spending: Evidence on the Fiscal Multiplier from a Quasi-Experiment (this one is somewhat closer in spirit to our paper). They both have fairly similar findings, of quite large effects of government spending at the local level.

Related Posts

. I wonder if the stimulus had any effect?

. Those stimulus measures, did they work?

. Swan delivers... stimulus of sorts