Public Spending and Crowding Out - a very cool test

Megan McArdle links to what sounds like a fascinating study of the impact of federal expenditures. What is most intriguing to me about it is the identification strategy they use.

Studies of the impact of fiscal policy are notoriously difficult because of endogeneity concerns. If we're thinking strictly of stimulus-type spending, this spending goes on precisely when the economy is bad. What you want to know is the effect of the spending relative to a counter-factual of what would have happened if the spending hadn't been there. But you don't generally have that. You often can't even compare to other recessions where fiscal stimulus wasn't used, because its very likely that it wasn't used because the recession wasn't quite as bad. People who ignore this counter-factual problem and pronounce fiscal stimulus D.O.A. confuse correlation with causation.

So you need some exogenous change in fiscal policy to test the effect. In the past, Robert Barro has used defense spending. The study McArdle links uses changes in committee appointments. When a Congressman wins a chairmanship, federal spending in his or her district goes up dramatically. This increase should be unrelated to the economic conditions of the state, so it should be useful for identifying fiscal impacts.

And they find that private corporate spending and employment goes down when government spending goes up. McArdle provides lots of thoughts. Here are some of mine:

1. First and foremost, this should not be a surprising result for anyone. In normal times, under normal conditions, crowding out is an important problem and the fiscal multiplier should be very small. So I have the same critique of running with this finding that I have always had with Robert Barro's work: it's important research as far as it goes, but be careful not to apply the results "out of sample". Don't take a coefficient from a growth period and apply it to depression conditions. Robert Barro is very sloppy about this consideration when he promotes his findings.

2. Corporate spending and employment goes down, but I'm curious what total output does. In other words, is this a net gain, but with a transfer away from the private sector? That wouldn't justify massive government spending even if it were true, but if it is true it would make us less uneasy about spending on valuable things.

3. These sorts of impacts are likely tied to the scale of the spending. In other words, as in so many things, there may be diminishing returns to federal spending. If you're judicious and pick a few very important projects federal spending may very well improve private economic performance. Of course if you dump cash on a community much of it will be wasted and distortionary. In econometrics lingo, this is a "local average treatment effect" - it's an impact estimate associated with very high spending levels, and it may not translate to lower spending levels.

So there are some caveats and questions, but it's a very interesting identification strategy. I think it's also promising because unlike Barro's defense spending, there is probably more variation in chairmanship turnover during downturns. This could help us differentiate between multipliers during downturns and multipliers during normal times. Of course, it still doesn't provide too much help comparing liquidity trap periods to non-liquidity trap periods, which is perhaps the most important thing to test when it comes to fiscal policy.

UPDATE: This sentence stood out for me: "The results show up throughout the past 40 years, in large and small states, in large and small firms, and are most pronounced in geographically concentrated firms and within the industries that are the target of the spending. "... there might still be some endogeneity. I'm not quite sure what being the "target of the spending" means, but if this is a matter of helping dying industries, we might have the same counter-factual (or for Austrians - "seen and unseen") problem that biases fiscal multipliers downward.