Tuesday, August 9, 2011

Another IV multiplier paper

Robert Vienneau shares a job market paper by David Shoag (Harvard). It's an interesting approach - Shoag uses windfall returns on state pension funds to instrument for government spending. This is similar in spirit to Romer and Romer's (2010) discussion of tax increases to reduce inherited budget deficits as exogenous to current macroeconomic conditions, except in the other direction. In this paper, it's windfall returns to pension funds that change spending
behavior.

The multiplier he gets is very large - 2.12 - and he suggests it's robust to a variety of specifications. This is quite surprising. For one thing, the multiplier itself seems too big even if we had a well identified estimate. But it's particularly big for interstate estimates, which ought to bias multiplier estimates downward for reasons I've discussed on here before. Shoag talks about certain frictions as playing a role in the result.

UPDATE: See Andrew's comment below. One of the things that puzzles me most about the paper is the instrument itself. Shoag writes "This paper’s research design can be clarified by the following thought experiment. Suppose New Hampshire’s retirement system outperforms Tennessee’s system after controlling for asset allocation. This performance is not due to the relative economic conditions in New Hampshire and Tennessee. Moreover, this excess return is presumably not expected ahead of time, or Tennessee would have altered its investment strategy. In the years after this outperformance, however, New Hampshire will have a smaller actuariallyrequired contribution to its pension plan and more resources available for government purchases. This paper estimates the effect of this randomly assigned spending on income and employment."

The bolded sentence confuses me. It would have thought strong pension fund performance would be associated with relative economic position. I don't know all that much about state pensions, but it would certainly seem to explain the large multiplier (what we usually think of as a negative endogeneity bias for fiscal multipliers is - because of the choice of instrument here - a positive bias).

So I'm skeptical of this paper, but my disclaimer is I've only read the first section and I'm always a grump when it comes to state-level analyses of fiscal multipliers.
What do you think?

Does anyone know of any other interesting multiplier instrument papers?


UPDATE 2: Also, I meant to mention a recent paper by Arnold Kling in the new issue of Critical Review on all this called "Macroeconometrics: the Science of Hubris". He does a good job clearly explaining the issues I've been talking about, but he's a lot more skeptical and nihilistic than I am about it all. He's like the Ed Leamer of the macro side of things, and while I share his understanding of the issues I don't quite share his pessimism about what we've come up with.

6 comments:

  1. Romer and Romer also get pretty big multipliers in their 2010 paper. If I remember correctly their multiplier averages around 3.


    I'm not 100% sure what your issue with the instrument is here given the ability of pension funds to invest anywhere they want. I assume there isn't much of a "home bias" to state pension funds.

    ReplyDelete
  2. Also I've been meaning to comment to tell you keep posting about this stuff. It is all directly relevant to and helpful for my dissertation where I'm doing both state level IV and VAR stuff on fiscal policy.

    ReplyDelete
  3. Let me know if you ever want to guest post about your work. I haven't done VARs for about six years, so I'm really taking my labor economics and program evaluation approaches to econometric identification and whining about macroeconometric work on that basis.

    As I mentioned earlier with how good I think Jaeger's influence was on me - I think it's a good basis on which to whine about macro. But I don't always feel like I can comment intelligently on VARs.

    Of course your work also has a lot to do with economic history, which is another favorite of this host and his readers.

    Do you have a good thorough summary of your work on your blog?

    ReplyDelete
  4. This is open to anyone. I've commented a lot on endogeneity and identification problems in macroeconometrics and what I perceive to be good and bad solutions. I haven't commented much on VARs. If anyone wants to guest post on this solution, just let me know.

    And as I said before - any other solid multiplier papers are welcome. I firmly believe this is not "ex post story telling" and while the empirical task is a tough one, it's quite wrong to call it "scientism" and pretend like we have no clue what's going on.

    ReplyDelete
  5. Naw, I don't really have anything up online.

    And, yeah, like you I assume I agree with the content but not the tone of Klings article (dont have access to it). I can guess though. Beyond the fact that it sounds like a warmed over Lucas critique I find it annoying when Austrian types invoke the radical strong version of the Hayekian idea that we "can never know anything" when they want to use it criticize people they disagree with.

    Meanwhile, they never seem overly concerned with just making crap up (like PSST or whatever it is) and acting like it has legs becuase they can construct some highly artificial anecdotes to "support" it.

    Anyway, I'm perfectly comfortable with econometrics even if it does not produce big T Truth.

    ReplyDelete

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