Tuesday, March 22, 2011

Hal Cole's Take on the Krugman/DeLong Debate

Here's what Hal Cole (U Penn) has to offer on the Krugman and DeLong affair.
One thing that struck me is that Krugman and Delong are really asking and answering a different question than many of the opposing voices they cite:

1. Can increasing government spending, especially during a downturn, increase output?
2. Does increasing government spending, especially during a downturn, make us better off?

They're answering "Yes" to the first question, based upon either assertion or simple empirics, and taking it as given that this makes us better off. The opposing voices they're citing are typically answering the second question. To answer this question you need a model, typically a quantitative model, to evaluate the welfare counterfactual. These models generally have the feature that government spending can raise output by making people poorer, but in order to a get a positive welfare impact of this spending (assuming that spending was set efficiently without the downturn and hence the marginal production of government spending is at least at its cost), requires more action from the model than the simple negative wealth effect can deliver.

This confusion about what question is being asked and answered is still a bit surprising given that the conventional RBC model, like the one in your [Williamson] textbook delivers an increase in output as a result of an increase in government spending (assuming lump-sum taxes). With marginal taxation this becomes a bit more problematic, and the extent to which government spending substitutes for private more of an issue as we move away from military spending. However, the conventional wisdom - which I take to be Barro's - holds that the multiplier is about 0.8 for military spending and perhaps a bit less than that for nonmilitary.

A lot of this confusion started with Cochrane's original comments on this issue. Cochrane wanted to rule out any model that got an increase in output from a negative wealth effect since that meant that welfare was necessarily being lowered. In the restricted class that he was considering as a result of this, government spending didn't raise output. But that's essentially by assumption.

There are a variety of the saltwater and freshwater types who are trying to re-examine the role of government spending under the circumstances that we now face. In doing so, they are using essentially identical methodologies, but somewhat different assumptions with respect to the stickiness of prices, etc. Mankiw argues recently that their may be a role if we're on the zero bound and cannot commit to future monetary stimulation in a recent piece for Brookings. Christiano et al constructs a standard neo-Keynesian model and shows that the impact can be large when we hit the zero bound. Jesus Villaverde has a paper with financial frictions and also finds that government spending can have a multiplier close to one in its initial impact. Both of the later papers don't (I think) do the welfare analysis, but that is largely because one has to take a stand on the value of the government spending, which is problematic. Mankiw's main point seems to be that there are more efficient ways to undo relative price distortions than government spending.


  1. OK. I think I got it basically right then!

    1. Yes.

    2. No, if the government spending is useless. Yes, if the government spending is useful.

    But I would add:
    3. If government investment has an NPV>0, it might more than pay for the future taxes needed to finance the bond issue, so that the Ricardian effects are actually negative. Government investment causes people to expect lower future taxes (or, more generally, higher future disposable income), so the multiplier could be greater than one.

    Now, you might say the government should be doing NPV>0 investments anyway, whether or not it wants to increase AD. Fair enough, but the Keynesian argument that the opportunity cost of unemployed resources may be very low in a recession is part of what makes NPV likely to be >0.

    Think of it this way: giving the unemployed a transfer payment would be dominated by requiring them to produce something useful in exchange for that transfer payment. Not only would we get something useful, but it might crowd in current consumption, since expected future consumption would be higher.

  2. Nick:

    Isn't that crowd out "less" in current consumption?

  3. Think of it this way: giving the unemployed a transfer payment would be dominated by requiring them to produce something useful in exchange for that transfer payment.

    Not if the value of the "useful" thing was less than the welfare loss from making it. Or, to put it another way: how do you define what's useful, and why isn't the government already doing it?

    You could say that the construction workers building houses were producing something useful, but it's not clear that was the best use of their time. Maybe a transfer payment would have allowed them to learn a new trade.

    In practice, most stimulus spending seems to be of the "bridge to nowhere" variety, i.e. of negligible value. In such cases it seems to me people would be better off with a transfer payment.

  4. Bill: (Where have you been? Busy Mayoring? We have missed you!)

    Normally, government spending crowds *out* current consumption. But if government spending is on productive investment, and so raises future disposable income, it may crowd *in* current consumption (i.e. negative crowding out). We are not disagreeing, it's just terminology.

    Anonymous. Agreed. The value of the government spending must be greater than the value of the unemployed's "leisure" (or whatever).

    "...and why isn't the government already doing it" Well, because it is waiting for our advice on whether or not to do it ;-)

    But underlying your question is perhaps this one: would the microeconomic CBA perspective of doing government spending if and only if NPV>0 give different answers to a macro perspective? Once we take into account the low real interest rates and low opportunity costs in a recession, I'm not sure it would give different answers.

    I'm still getting my head clear on this. Maybe (in a demand-deficient recession) macro NPV = k times micro NPV, where k is a positive number. Same cutoff, in other words, but different values. This could easily be wrong. Still thinking about this.

    Purely anecdotally, from my own casual observations in Canada, the stuff the government was building seemed like stuff we would eventually have built anyway. They just brought it forward a bit. If I'm right on that, future consumption will be higher (because it won't be crowded out by future bridge building), so current consumption will be higher via the Euler equation.

    Generally though, all this in just expanding on the point that we can't just talk about "G". We need to look at what is in that G. It matters, both on micro grounds and on macro grounds.

  5. Not only does G matter but so does C and I.

    But C and I are private sector. So they must be efficient. HAHA

    You economists are hilarious. I hope efficient G crowds out inefficient C and I.

    In fact it is inefficient banking sector I created the mess we are in now.

    A lot of C is good and needed like food but some C is crap, see "5-Hour Energy".

    Are these things in your models?

  6. Not a single thing that the previous post contains is remotely correct.

  7. That's obviously a strong statement. Care to elaborate?