Thursday, September 1, 2011

Mercatus Study Shifts My Priors on Fiscal Stimulus

...considerably toward the pro category.

I should have read this more closely yesterday - I read Tyler Cowen on the Jones and Rothschild paper up to his punchline: "…hiring people from unemployment was more the exception than the rule in our interviews" and didn't read any farther, which was a big mistake. I figured this meant that hiring from unemployment was very rare, but actually 42.1 percent of new hires at ARRA receiving firms were previously unemployed. This is surprisingly high (so high that I'm personally curious if there was some selection bias or misreporting biasing it up). The other estimates of this figure that I'm familiar with are associated with job creation tax credits, which I'm writing about for a conference this fall. Bartik and Bishop (2009), two of the most knowledgeable guys out there on these tax credits, estimate using BED data from the Bureau of Labor Statistics and some elasticities reported in Hammermesh (1993) that they expect their tax credit proposal to have 17.7 percent of their newly created jobs filled with unemployed workers. Job creation tax credits are widely considered to be an extremely effective form of stimulus, so to have ARRA funding create 42.1 percent of jobs for the unemployed is truly impressive to the point of being a little unbelievable.

Of course, as many have pointed out by now, you increase employment of previously employed people with these ARRA funds by increasing demand for labor. If these workers are moving from non-ARRA funded employment to ARRA funded employment, presumably they quit, they were not fired (otherwise they more than likely would have been moving from unemployment) and the original firm still demands labor and will try to hire someone else. So we can still expect indirect job creation in addition to these impressive direct job creation figures.

Of course these sorts of studies still can't get at the unseen - the crowding out that may or may not occur from ARRA. But I don't personally need convincing that crowding out is not a problem now, so this Mercatus study seems to suggest a strong stimulus - at least relative to benchmarks we have about what to expect.

Also, I have to complain about one point by Tyler Cowen. He writes: "The paper also sets a new standard for disaggregated data on this macro question". No, it really doesn't. It's a great approach that they take, don't get me wrong. But people have been doing field work of recipients of these sorts of funds for a long time. The Urban Institute does this. Lot's of people do it. And of course - disaggregated analysis of aggregate phenomenon has very real liabilities, as I've discussed on the state-level studies of the stimulus.

Some other reactions to the paper:

- Kevin Drum was one of the first to point out that this is one of the most ringing endorsements of the stimulus we've had.

- Yglesias concurs.

- Steve Horwitz raises a non-sequitor about how labor isn't fungible. Of course it isn't fungible, but it is substitutable. I'm not sure exactly what Steve is so worried about here. He gives no indication of how substitutable he thinks labor is, but it seems like he thinks there's very little prospect of it. I don't know why. And he seems to act like nobody is aware that different workers aren't perfectly substitutable. Not sure what to make of this post.

- Cowen responds to Drum. His first point is about labor market polarization which is what I mentioned the other day that Autor talks a lot about. I think this whole idea of labor market polarization is not as robust or meaningful as people like to pretend it is.

UPDATE (UPDATE 2 below): In the comment section of his post, Steve Horwitz is getting very concerned with my comments about what I thought was a simple point that "fungible" traditionally means perfectly substitutable and that economists think of a range of rates of "substitutability" between factors of production. I'm at home today finishing up a paper on a much less substitutable factor of production: engineering labor. That low substitutability has important implications for the labor market that I spend a lot of time talking about. Other jobs have much higher levels of substitutability. I haven't done construction work in years, but (if there were demand for construction workers) I could probably get a construction job and be more productive at it than I would be at an engineering job (which I wouldn't be hired for anyway - precisely because I wouldn't be very produtive). But even when we think there is low substitutability, we usually think there is still some substitability. My father-in-law was trained as a chemical engineer but currently works as a nuclear engineer. I guess whatever he specifically works on is similar enough that he could get started and that - with some on-the-job experience - he could be productive at. There is often some non-zero rate of substitution for any given worker with any range of jobs. Steve just thinks it's pretty low. Anyway, the idea that this is controversial boggles my mind, and I wanted to repost a comment that I have on that comment section:

"Fungibility is the interchangability of some good. We do not think labor is fungible. We do think it can be substituted at some discount. Am I wrong? Is there any difference in Chait and Steve's understanding except that Steve thinks the discount is steeper (but presumably not total - so that even under Steve's construction of the situation there is bound to be some indirect job creation in addition to the direct job creation)?

To challenge Chait's point you have to think that the firms whose workers were poached by ARRA are unable to hire any new workers to replace the lost workers. If you don't think that you have to agree with Chait that there is some additional indirect job creation.

How much and whether it offsets the cost of ARRA is an empirical question (and a tough empirical question at that)

UPDATE 2: As Mathieu Bedard points out to me, Chait has a really dumb title to his post. It is wrong. As I subsequently point out, Chait has a really good post, despite the dumb, wrong title. Challenging the point he makes in the post is wrong, and challenging the word choice of the title strikes me as largely pointless - although I suppose right.


  1. If fungiblity requires perfect substitutability, you've set the bar ridiculously high.

    Are US dollars fungible?

    Are a ten dollar bill and ten singles perfect substitutes when you're in the Sahara and come across a bottled-water vending machine that take singles?

    By definition, the first is true, and the second is not. Therefore, by your definition, dollars are NOT fungible.

  2. Oh, bother. "Therefore" was left over from an earlier draft. It should not have been in this version. Start, instead, with "By your . . . ."

  3. You usually hear fungible with reference to oil or other commodities - it doesn't matter where it comes from. The only thing that may matter is transport costs.

    Oil. Corn. Magnesium. Cotton.

    I'm not sure this bar is ridiculously high.

    Labor is not "fungible". It is substitutable to a certain extent. Workers can be replaced by roughly identical workers or by less identical workers at a discount.

    If we've gotten to the point of vending machines in the Sahara we've gotten to the point where the conversation isn't producing useful information anymore.

  4. So, you're denying that money is fungible? I can come up with other examples, and you know I can. What does it do to the concept of fungible that nothing is truly fungible?

    Essentially your defence of the "stimulus" is that labor is fungible. So, now you are reminded that labor is not fungible, because it's not a perfect substitute.

    And, lest you object to the first line of the last paragraph, there was no crash in the asphalt (or steel welding) industry in 2007, so unless you were saying that the ARRA would not stimulate employment, you were assuming fungibility in the labor market.

    And yes, as Tyler has pointed out, but you seem to reject, transportation matters in the labor market, so what does that do to the fungibility of labor? More, or less?

    Finally, "If we've gotten to the point of vending machines in the Sahara we've gotten to the point where the conversation isn't producing useful information anymore." Why? Are dollars fungible? Change my example, then. Make it the post office lobby (you know what a post office is, right?) after hours, and you need stamps. Repeat the above (unless the machines now take $10, if so, then bump up to $20).

    Is there such a thing as a perfect substitute?

  5. I'm not sure what you mean by talking about money being fungible. I think of fungibility as being perfect substitutability between the products of different producers. Someone drilling a certain type of oil in Saudi Arabia's project is said to be producing a "fungible" product because it is interchangable with other producers of that oil. I'm having a hard time translating that to money, but I suppose you could call it "fungible". I'm not sure.

  6. re: "Essentially your defence of the "stimulus" is that labor is fungible. So, now you are reminded that labor is not fungible, because it's not a perfect substitute."

    No, not at all. Labor is not fungible and the argument for stimulus does not depend on that.

    re: "And yes, as Tyler has pointed out, but you seem to reject, transportation matters in the labor market, so what does that do to the fungibility of labor? More, or less?"

    Wait - when did I reject that? Usually transportation costs are not thought of as changing the fungibility of a product. So you'll hear about how Japan has to pay more because it ships in all its oil, but oil is still considered a "fungible" product. It's the canonical example of a fungible product.

  7. "Usually transportation costs are not thought of as changing the fungibility of a product. "

    Huh? So, cotton grown in TX and cotton grown in Egypt are fungible to someone in, say, Texas? If that's the case, how are they "perfect substitutes"?

    "I'm not sure what you mean by talking about money being fungible."

    You're not sure if currency is fungible? It's essentially the definition of fungible! See, e.g., the second sentence of the Wikipedia entry for fungible.

    "Wait - when did I reject that?"
    When you claimed that housing prices were the most important factor in explaining the lack of labor mobility in this downturn.

    ARRA was a gigantic claim that labor is fungible. The logic went essentially like this:
    Construction workers are being laid off
    Roads are constructed
    Therefore, if we pay for more roads, construction workers will be rehired to meet demand.

    That treats construction labor as fungible. If you want to say it's about the multiplier, you'd better look again at how road crews spend their money. And, before you try to tell me how you should know I'm well acquainted with how they do from experience.

    What's tremendously interesting is that now you wish to argue labor is not fungible after you have had multiple postings essentially claiming it is (see: e.g. your posts on the stimulus and your reaction to Tyler's posts on labor mobility).

    "Usually transportation costs are not thought of as changing the fungibility of a product. So you'll hear about how Japan has to pay more because it ships in all its oil, but oil is still considered a "fungible" product."

    Geez, where to start with this? Is oil from Indonesia and oil from Saudi Arabia fungible? Is oil from Australia (low sulfur) and from Venezuela (high sulfur) fungible? What about oil from Norway (on the international market) vs oil from Alaska (essentially off the international market, but does affect US demand on the international market, but probably not "perfectly")? Are they fungible for the Japanese market?

  8. Nonymous -
    I've never claimed labor was fungible. Ever. My thoughts are in multiple places now on this and other issues.

  9. "I've never claimed labor was fungible."

    So, I assume that means you have never supported ARRA? I mean, the "stimulus" responded to a loss of jobs in the home construction industry with money for road work. That's a pretty big f'in theory of labor funbibility.

    You should see the ARRA projects I've seen: road signs in Ann Arbor, MI; curbs in Cincinnati, OH; exit-ramp flowers in OK ad TX; a new set of park benches in KY . . . .

  10. Fwiw, I don't get Horwitz' post either. Workers can be trained. That's why structural adjustments are possible. Workers can be trained for stimulus funded jobs, as well.

  11. That's good to hear Jonathan.

    Ya, I'm guessing he actually agrees with Chait. It's hard NOT to agree with Chait. Of course we need to think about indirect job creation too, and of course there's a degree to which workers are substitutable, and as you say of course there are ways of making them MORE substitutable.

    I can't imagine Steve disagrees with that.

    I think he took a cheap shot at a title and ignored the content of the post, and now just isn't willing to give Chait the time of day because he's committed himself to acting like Chait is an idiot.

  12. I believe this discussion boils down to whether labour is homogenuous or heterogenuous; hence, it is substitutable if it is the former.

    Yes, I would say that's a better way of framing it as to avoid the semantics rif raff

  13. Daniel,

    Thanks for reading our survey-based paper. Our separate interview paper has more information on how organizations say ARRA worked. Link to that in my name.

    I'd claim, though, that comparing our "hired from unemployment" percentage to Bartik & Bishop's net job creation percentage is a misreading. Our "hired from unemployment" percentage is more comparable to Bartik/Bishop's *gross* job creation figure.

    It's hard to go from gross to net--and B&B are only able to because they (fairly plausibly) assume the Hammermesh labor demand elasticity.

    Another way to put this: Just as a government-subsidized gross hire isn't a government-subsidized *net* hire (B&B's point), so a government-paid gross hire isn't the same as a net increase in jobs. Some of those unemployed workers hired by government would have gotten jobs after a few more painful months of job search, as evidenced by the 4 million gross hires every month in the JOLTS survey.

    A Mortensen-Pissarides steady state is a useful comparison: In that world, if the government starts hiring people when the economy is at the natural rate of unemployment, the government can of course hire lots of people at (or slightly above) the going wage. But it's creating zero net jobs: It's crowding out private employment one for one. High gross government hiring does not equal net government job creation.

    It's important for stimulus proponents to decide how to tell whether the facts are more consistent with ARRA causing net job growth or causing mere gross job growth. Rothschild and I have offered some evidence at the organization level, where we focus on the primary effects of stimulus. We looked for signs of labor market tightness in ARRA-funded sectors.

    Our survey paper gives a few forms of evidence for the tightness of the labor market in some ARRA-favored sectors:

    1. Most organizations said it was as hard or harder to find good workers than before the crisis.

    2. New hires reported wage gains on average (driven by the job-shifters) and better benefits (across the board).

    3. Our job-switching rate was similar to the (likely overstated) medium-term U.S. average. So good help was somewhat hard to find among the unemployed, even with twice the number of unemployed compared to the pre-crisis world.

    This is broadly consistent with the hypothesis that ARRA was, on average, targeted at sectors of the labor market that were close to their steady state unemployment rate. At the very least, all three pieces of evidence point toward weaker net job creation, more supply-side crowding out.

    I hope we can agree that *if* future ARRA-style programs occur, they should focus on weak sectors of the labor market, with high, persistent rates of unemployment for well-qualified workers. I hope we can agree that targeting sectors with slack is important.


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