Everything that needs to be said has already been said.
But since no one was listening, everything must be said again.

Andre Gide

Friday, November 6, 2009

Fiscal Multipliers in War and in Peace

It truly is breathtaking how certain some people are of what they know to be true about the way a macroeconomy operates. The Krugmans and DeLongs of this world really make it sound like everything we really need to know has been settled long ago. Yes, the science is "settled" (where have we heard this before?).

I recently came across this piece by Brad DeLong: A Guide for the Perplexed. Consider the following quote:
But when fiscal boost was tried on a large enough scale, it certainly did the job. And it is reasonable to infer (with all the caveats provided by the CBO) that what is true in the very large will be true in the merely large as well. Eugene Fama says that it is theoretically impossible for fiscal stimulus to boost output: World War II proves him wrong. Robert Barro says that the multiplier is zero: World War II proves him wrong. Benn Steil says that Jacques Rueff in 1947 conclusively proved that fiscal policy could not boost employment: World War II proves him wrong.

Implication: a WWII style fiscal stimulus will "do the job" in a peacetime recession. WWII "proves" it. Egad...how does he know this? Why do I not feel as confident that this is the case? Am I truly that dense? (an invite to some rather rude comments, I'm sure!)

In any case, I decided to gather my thoughts on the subject and post them here for public review and criticism. The piece is a bit too long for a blog posting; so if you're interested, please click here. Looking forward to any comments.


  1. Hi David,

    I like the model you lay out. Quite in keeping with you book. Took me nearly twenty minutes to solve the equations though... getting rusty.

    How would you interpret theta and lambda?

  2. Don't feel bad, it took me more than 20 minutes!

    An increase in theta means that a household now attaches relatively more value to the government good (g) vis-a-vis the private goods (c,h). An increase in lambda means that a household now attaches relatively more value to the home-good (h) vis-a-vis the consumption good (c).

    Does this help; or were you looking for something else?

  3. That helps confirm what I was thinking. It seems to me that one of the benefits of thinking this way (following your model) is we can clearly define how theta would be expected to behave in times of war and peace.

    If one thinks of what the government provides, it is relatively stable during peace time, so theta shouldn't change much, because people don't change much. But in war, when security is threatened, of course we are more willing to work to produce g because it is more highly valued.

    What I would be interested in seeing is a model that captures the displacement of the production of c and the production of g picks up. My parents lived through WWII and told me all about the rationing. Yes, GDP was through the roof (so to speak), but few people had a nice life for all that expenditure.

  4. J:

    In fact, the rationing of private consumption during war is exactly what the model predicts (observe that c is a decreasing function of theta; even though GDP is an increasing function of theta).

    The point is that while people are not happy with the privation associated with war-time rationing, they understand the need for it. The wartime boost in g, along with the associated increase in employment and output is exactly what society demands during a national security threat. This does not, however, imply that the same sort of response is necessarily desirable during a peacetime recession (as DeLong and others would have us believe).

  5. I had to go back and read your piece again to remind myself what you were refuting. I got a little excited thinking about some other angles.

    Anyway, DeLong and others want to say that dy/dg > 1, but you are showing that dy may be function of theta, not because the government has super powers. Great, agreed. You finish by saying that you are agnostic about fiscal multipliers. So maybe we can continue the discussion?

    The thing that pops into my head first, which may or may not be important, is the effect of overinvestment in war-specific machinery and skill. For example, several higher-order capital goods are necessary in the production of tanks, artillery and other materiel that have little or no peace-time value. The same goes for skills acquired during a war (not all skills, obviously).

    To make a long point quickly: do we not need to worry about the costs of reversing the overinvestment after the war is concluded? Think of the GI bill: while honorable and (my opinion) just, it still was costly.

    If my thinking is unclear, please let me know and I'll get into more detail. I really didn't want to post a long comment.

  6. J:

    To think clearly about investment, we need a dynamic model. The model I presented was static in nature; but then again, so is the standard Keynesian Cross.

    The more thoughtful proponents of fiscal stimulus argue that recessions present the government with a good opportunity to deploy "underutilized" resources for the purpose of rebuilding or augmenting public infrastructure (roads, bridges, damns, etc). There would be no need to unwind any "overinvestment" if the proper amount of investment is chosen in the first place.

    Conceptually, this argument makes some sense to me. Much depends on how the government executes the expenditure (e.g., whether resources are truly devoted to positive NPV projects, or whether it is pure pork). In any case, I think that the Obama stimulus package devotes relatively little for this purpose (most of it is in the form of transfers). I could be wrong though; I am not exactly sure how the individual states are using their federal transfers.

  7. Does a coordination failure-caused recession necessarily imply that the economy does not have a "self-correcting mechanism"?

    Maybe some are prepared to make that argument. Are they really representative of believers in discrete fiscal stimulus?

  8. I think BD would say that unemployment is missing from your toy model. In your model there is no unvoluntary unemployment. If there was perhaps government expenditure would be necessary to achive higher employment.

  9. D: I'm sure that this is one thing BD would say. But he would be wrong. It would be easy to incorporate unemployment into such a model (a la my AER 1996 paper). It would not change the basic message.

  10. Sorry I am not a macro guy so I am not familiar with your AER paper. Why don't you expand your simple model to include this possibility? Would it be too complicated? I am not sure that I see why unwanted unemployment would not change the story.

  11. Take any Pissarides-style matching model where job-matches are created and destroyed (with creation involving time and resources).

    The planner's solution entails a positive level of "involuntary" unemployment; and the optimal level will vary with the business cycle.

    As in the simple neoclassical model, the government may invoke policies to increase employment (lower unemployment) via fiscal policies designed to "stimulate" demand. The ability to do so, however, does not imply that such a policy is socially desirable.

    I'm afraid that you'll have to do some reading if you want to get into the details and convince yourself of this.

  12. David,

    I don't know how you feel about Austrian economics, but here is a very revealing piece about BD from Robert Murphy of the Mises Institute. http://mises.org/daily/3841

  13. I think that Robert Murphy wrote a thoughtful piece. It would be interesting to see how BD might respond to it; but it looks like he is not returning emails. BD = BS, in my view.

  14. Hi David,
    as I said I am not a macro guy and although I am quite interested in this topic unfortunately I have no time to do a careful reading of the literature. However, I have found a job market candidate from Berkeley with a paper exactly on the same topic (the link attached.) His model seems to be similar to your (he even cites yours) but he conclude that government spending coudl increase welfare. I do not claim that I have understood or even read it but it is interesting.


  15. Hi D,

    Well, I'm not sure what we're debating here. I have taken a look at the abstract: http://sites.google.com/site/pmichaillat/research (I was unable to find the paper).

    The point of my original post was that just because an increase in G may be socially desirable (and stimulative) in a time of war, this does not necessarily mean that the same holds true during a peacetime recession. My point is that DeLong assumes this is so. He has no theoretical, or empirical, justification for believing so strongly that this is true.

    Whether there is unemployment or not is a red herring (and this was my point to you).

    You have found an article where an increase in G may be desirable in a model of unemployment. Sure. The author has decentralized the search economy in a particular way such that the equilibrium level of unemployment is not socially efficient.

    But I could have done the same thing with my simple static model; for example, by introducing an externality and decentralizing the economy via competitive equilibrium.

    In all of these cases, the best government policy is to correct the distortion directly. It is not to tackle the distortion indirectly by changing G.

  16. This comment has been removed by the author.

  17. D,

    If you read my paper carefully, you'll note that I never accuse BD or PK of being "wrong." I accuse them of pretending to know more than what the science warrants. I try to inject a little humility into the exercise by developing alternative plausible interpretations of the data. I'm afraid that you missed the whole point of the exercise (but perhaps this is my fault for not communicating the message well).

  18. Hi David,
    I am not really arguing. I do not have a clue.
    I just observed that your toy model is just as simplictic as BD's. I do not know which one is better, if either. I believe that you are right and most of BD and Krugman's argument based on beliefs rather than knowledge. But I doubt that anyone else know the right answer. They may be wrong, they may be right. We do not have enough empirical evidence to decide.

    I was talking about this guy's job market paper. "DO frictions explains unemployments." I thought you would be interested in it since it seems to be somewhat similar to your AER paper (I did not read it. I was searching for it that is how I found this) In the conclusion (I have read only that) he says that in bad times governmeny direct employment is efficient. "The effectiveness of direct employment is conditioned on how much it crowds private employment out; in bad times competition for workers is weak and crowding out is marginal; thus this policy is very effective."

    This sounds just like BD.

    So yes I can see your graphs and understand that perhaps BD, Krugman, Summers et al are wrong but I am not totally sure. We will see in a couple years, at least in this case. I am sorry but I hope that they are right and you are wrong. I think it is better for all of us if at least Summers and Bernanke are right. Unless you are shorting the whole US market. :)

  19. Ok. Sorry then. I guess I juts had a strong desire to learn the truth and I misread your argument. :)

  20. Hi David,
    I understand that your point was that we cannot be sure. Fine.
    However, the FED and the Treasury have to make decisions. So what is your opinion on this recent CBO document?

  21. D: I haven't the time to read it carefully, but the following quote caught my eye:

    CBO used the evidence from models and historical data to determine estimated “multipliers” for each of several categories of tax and spending policies in ARRA (see Table 2 on page 4). Each multiplier represents the estimated direct and indirect effects on the nation’s output of
    a dollar’s worth of a given policy. Thus, a policy’s multiplier can be applied to the budgetary cost of that policy to estimate its overall impact on output.

    In other words: Let's assume that the Econ 101 fiscal multiplier theory is true; and then let us use this theory to estimate the impact of fiscal policy. Caveat emptor.

  22. I have to make a comment regarding D's suggesting that Fed and Treasury have to "do" something. Politicians have to do something to make it look they're supposed to have a job. Bureaucrats are the same. So even if we all "knew" that doing nothing is optimal, it is never actually an option. That's gub'ment.

    The Fed, on the other hand, is not political. But, if it doesn't act - then what? I don't know the consequences of the Fed not acting, but I bet they're undesirable in a political sense. Even if it might be desirable in an economic sense.

  23. Presumably, if we are to take things seriously, we accept that the magnitude of "the multiplier" is state-dependent. Then the question becomes, what's a credible empirical strategy to estimate it (as a function of the state!) without imposing too much structure? I haven't seen much work on that. The focus on WWII-like episodes is misguided, in my humble opinion.