Monday, March 5, 2012

Thoughts from Brad and Andrew on Nakamura and Steinsson

Brad has more biases to think through:

"There are, I think, the places to question the applicability of Nakamura and Steinsson:

1. Supply-side regional factor mobility biases their estimate up.
2. Demand-side regional factor mobility biases their estimate down.
3. In the current conjuncture, the counterfactual is not one of higher government purchases with constant monetary and financial conditions--a constant real interest rate and constant risk premia. A stronger economy means faster inflation and fewer bankruptcies, and thus lower real
interest rates charged to firms. There is not crowding-out but crowding-in, and N&S's constant-monetary-and-financial-conditions multiplier misses this effect

He also stresses that Nakamura and Steinsson are estimating a constant-monetary-conditions multiplier. He adds at the end (about the biases) "No, I do not know how these three effects add up. But I have to have a view by 1 PM or else...". And that - of course - is 100%, entirely legitimate. When the rubber meets the road, I'm going to be coming down on Brad's side in how the multiplier is assessed. I just want to point out that Cochrane made a quite valid critique of some of the evidence hinted at by Larry Summers, which you can agree with without holding to a crowding out view.
Andrew asks "If taxes are distributed evenly (I don't know if I buy that but it seems like its a key assumption of N&S) don't you have a treatment and control group that both have the negative tax effect built in?". Yes, and that's precisely the problem. Let's say you're looking at exogenous spending in one state financed by taxes divided equally between both states. You are calling one state a counterfactual for another, so you are comparing a state with all the spending and half the tax burden and asking what it looks like relative to a state with half the tax burden and none of the spending. You are differencing out the tax burden, in other words, and that's the problem. Now you are literally asking what the benefit of a free lunch is. It's not surprising that you're going to get a big multiplier.
This is definitely not a good way of answering the question "what was the fiscal multiplier from 1966 to 2006?", because you're completely differencing out what we think should be one of the biggest drags on fiscal policy.
What I think Brad may be getting at is that differencing that out may give us bad estimates for 1966-2006, but it's pretty good for providing an estimate for the impact of deficit financed stimulus now. Maybe. But that depends on the assumption that funds aren't being crowded out now. I buy that. I am onboard with the point that large deficit financed stimulus would crowd other spending in right now. But that very convincing chain of argument doesn't make estimates using federal financing at the state level unbiased estimates. They are biased. Brad's case is that they are biased in a way that we want them to be to explain our current situation.


  1. I'm going to go read the paper now, since I now have a golden opportunity to justify this argument on the internet as working on my dissertation!

    Also, I don't think their multiplier is that big, it seems pretty mid range.

    1. I should, but I might not be able to until later in the week.

      Emailing brad and talking to you... I am changing my mind a little bit on the point I make about taxes... I think I've got my head stuck too much in thinking about how this is a problem in my job training evaluations.

      If you want to know the impact of deficit spending, you want to difference out the taxes, right? That's bad for a "what is the impact of the the government on output" estimate (my original point), but that's good for the impact of deficit spending on output, which is what we're interested in.

      Spillover and factor mobility concerns still apply.

      Hmmm... turning more against Cochrane now, at least on the tax point. Let me know if that sounds right to you.

  2. Yeah, I think we have reached a convergence. I read the paper, I don't have much to say about it. At least without risking sounding dumb since I only read the paper casually. I also skimmed the theory part of it.

    They do argue reasonably well the conditions under which their results can be generalized. They do totally neglect the issue of whether you can claim that tax policy is constant across states. But most authors who do this state level stuff just take that idea as given and I've never seen anyone actually argue the case.

    It's possible that net tax burdens could be different across states. For instance, New York and Connecticut possibly may have a larger proportion of income taxed as capital gains than say in Alabama. Any changes in tax rates on ordinary income would effect the two states differently. There are other ways I think net tax burden can be different across states, but I don't know if that is a real problem or if it is significant if it is a real problem. It's just one of those things everyone who does this kind of study just declares as obvious.


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