Thursday, August 25, 2011

Regular Economics

Barro has a piece in the WSJ which appears to be part of a series of critical opinon pieces on Keynesian economics. It's better than the last one.

I like the idea of emphasizing "regular economics" vs. Keynesian economics. The 1970s revolution that Lucas started was principally a drive to make macroeconomics regular, by using tools developed in other fields in economics to make sense out of macroeconomic problems. This was initially hard for Keynesians to swallow, but it did not take long for them to catch on, and by the 1980s they were doing the same thing.

Modern Keynesian coordination failure models (not so popular today) and New Keynesian models (very popular) are in fact regular economics - most of the time. Everything is explicit, you can see how it works, and you can quantify things, such as the effects of government policies. Then, we can go about evaluating these models. Why is this assumption there and not another one? Is this parameter measured correctly? How well do these models fit the data?

Much blogosphere Keynesian economics is not regular economics, though. What is driving much of this are the misleading notions that come from some undergraduate textbook macro - the free lunches that Barro is complaining about. Arguments are typically couched in terms of "deficient aggregate demand," by which an irregular Keynesian means "output is below trend." Not all of the blogosphere is like this though. Paul Krugman, who often engages in irregular Keynesian economics, can on occasion be regular, in which case you can actually take it apart and see how it works.

Now, since Barro is regular, you can also take his arguments apart and see what makes them tick. For example, Barro states the following:
Ironically, the administration created one informative data point by dramatically raising unemployment insurance eligibility to 99 weeks in 2009—a much bigger expansion than in previous recessions. Interestingly, the fraction of the unemployed who are long term (more than 26 weeks) has jumped since 2009—to over 44% today, whereas the previous peak had been only 26% during the 1982-83 recession. This pattern suggests that the dramatically longer unemployment-insurance eligibility period adversely affected the labor market. All we need now to get reliable estimates are a hundred more of these experiments.
We know that unemployment insurance (UI) has incentive effects. Any serious model of unemployment will deliver the result that an increase in UI benefits - working through a reduction in search effort by the unemployed, or a tendency of the unemployed to become more picky about the jobs they will accept - will imply a higher unemployment rate and longer average duration of unemployment spells. The key question is how large the effect is. We might also be interested in how the size of benefits matters relative to the duration of benefits, for example.

Barro is too careful to stick his neck out, but he certainly seems to be indicating that he thinks the effect could help in a significant way in explaining what we are observing in the data - i.e. he thinks the effect is large. He also indicates that we should have to see more data to sort this out, but I don't think that is true. We actually have a lot of cross-country data for countries with very different UI systems, for example, and some previous experience with the extension of benefits in the US. Indeed, I would be surprised if there were not many published papers that address the question directly. Anyone know?

P.S. Here is some work on this, arguing the effect is small. The source of the citation also makes the point that it is important to look at the recent data.

12 comments:

  1. "We know that unemployment insurance (UI) has incentive effects. Any serious model of unemployment will deliver the result that an increase in UI benefits - working through a reduction in search effort by the unemployed, or a tendency of the unemployed to become more picky about the jobs they will accept - will imply a higher unemployment rate and longer average duration of unemployment spells. The key question is how large the effect is."

    This is a very one-sided view of the question: increased UI benefits will have incentive effects, as you suggest, as well as demand side effects from increased income if this is financed by debt. It's not clear why the incentive effect (workers are pickier about jobs) would dominate the demand effect (more jobs available) to push UE up.

    More broadly, I think you've been a little unfair toward the Keynesian folks with this idea that they're pushing for a free lunch with stories of deficient aggregate demand. I'm sure they all recognize that increased G today (or lower taxes) will have to be financed in the future, hence this involves reallocation of resources over time (more C today, less in the future) but that it's worthwhile because MU(C) is higher today than it will be in the boom years of the future. For example, your two favorite economists, Krugman and Romer, have both been pushing the point that fiscal policymakers should be spending more now while restricting spending (or raising taxes) in the future.

    - C

    ReplyDelete
  2. I remember Assar Lindbeck talking about this at Friedman's 90th birthday party. His 1997 JEL paper talks about disincentive effects and outright cheating in Sweden's unemployment insurance system. He also says that in a country with a strong work ethic it takes time for these effects to work.

    Here is a more recent working paper that uses Estonian data: http://www.mtk.ut.ee/orb.aw/class=file/action=preview/id=825506/Febawb70.pdf

    Of course, in addition to incentive effects on recipients, you have to consider the effects on employers who pay a tax to support the system.

    ReplyDelete
  3. "This is a very one-sided view of the question: increased UI benefits will have incentive effects, as you suggest, as well as demand side effects from increased income..."

    Sure, there are potentially other things to worry about as well. UI is a transfer. How is that going to affect how people behave in the aggregate? How will the transfer be financed, and how does that matter? And so on. I didn't mean to say those things were irrelevant, I was just focusing on the one particular effect.

    "More broadly, I think you've been a little unfair toward the Keynesian folks..."

    I've gone into more detail on this elsewhere. I don't think the irregular Keynesian economics is good science. You need to isolate the market efficiency the government can actually correct, and think about how best to correct it. What the government spends on matters, and there are always opportunity costs to government spending that you have to take into account. A key element is uncertainty about the future. If we are not in a temporary cyclical downturn but are on a permanently lower growth path (which countercyclical Keynesian policy is not going to change) then we have a problem.

    ReplyDelete
  4. I thought New Keynesian economics was an economics of coordination failure. If everyone adjusted prices/wages together equilibrium would be reached. Or at least some versions of NK worked like that. If that isn't the case, what are some examples of coordination failure Keynesianism? Clower-Leijonhufvud?

    ReplyDelete
  5. Here's a post on coordination failure models:

    http://newmonetarism.blogspot.com/2010/08/what-happened-to-coordination-failure.html

    I can give you references if you want. Those are actually models where all the prices (if there are any) are flexible. There can be multiple equilibiria, so for example you could be in good equilibrium with high aggregate output and welfare or a bad equilibrium with low output and welfare. Which equilibrium you are in is determined solely by self-fulfilling expectations, if you like. That's very different from a Woodford-style New Keynesian (NK) model, where the prices (and sometimes the wages) are sticky. In the coordination failure model, the problem is that you can be in a bad equilibrium, and everyone would agree that there is a better equilibrium, but you are nevertheless stuck in the bad one. In the NK models, there is an inefficiency in that the prices are wrong - relative to what would happen in equilibrium with flexible prices. You can fix that problem, or at least mitigate it, with monetary policy, subject to the zero-lower-bound constraint on the nominal interest rate.

    Different models, both called Keynesian, but they work in different ways, with some different implications. You can debate whether either is what Keynes had in mind.

    ReplyDelete
  6. Here's the Ljungqvist/Sargent 2002 paper, don't know if it was the sort of thing you're looking for. You very well might already be familiar with it.

    Link is a download:
    http://goo.gl/9iGQx

    I don't know of any relevant regression-monkey stuff off the top of my head.

    ReplyDelete
  7. Yes, I know that one. That's got another interesting mechanism, which is human capital depreciation while you are unemployed, and they seem to find it useful in accounting for differences between Europe and the US.

    Here's another useful point. You would think that optimal UI would be more generous in a recession. That is what these people find:

    http://elsa.berkeley.edu/~saez/landais-michaillat-saezFeb11UI.pdf

    ReplyDelete
  8. Thanks for the pointer on coordination failure. It a bit sounds like the distinction between failure of a market to clear and clearing of a market at an undesirable price/quantity (possible interpretations of Keynes discussed by Garrison/Meltzer here).

    ReplyDelete
  9. This is pseudo-science. Experiments can be run when linear law-governed physical processes involving natural kinds are involved.

    It is brain dead nonsense to suggest that there are physical laws of nature involving natural kinds involving unemployment insurance and unemployment rates in different historical situations involving different specific individuals.

    Barro writes,

    "All we need now to get reliable estimates are a hundred more of these experiments."

    If "normal economics" = pseudo-science it's time to smash "normal economics" and try real science.

    ReplyDelete
  10. Try Vedder & Galloway _Out of Work_ for a survey on this topic from some years ago.

    ReplyDelete
  11. About Barro being 'too careful to stick his neck out' with respect to the question of the effects of an increase in the UI duration spell on unemployment:

    In this piece (http://www.economics.harvard.edu/faculty/barro/files/WSJ%2BUI%2BAugust%2B2010.pdf), Barro states that 'My calculations suggest the jobless rate could be as low as 6.8%, instead of 9.5%, if jobless benefits hadn't been extended to 99 weeks.'

    And of course, as Barro engages in 'regular economics' as opposed to all those Keynesians, who apparantly engage in irregular economics, he must be right.

    ReplyDelete
  12. Thanks. Obviously he stuck his neck out. The paper I link to in the PS is regular economics, and it comes up with a much smaller number (0.3% rather than something approaching 3%). In the link you mention, Barro has one paragraph where he does a back-of-the-envelope calculation. That's actually not regular economics, it's actually some b.s., so Barro obviously is not practicing what he preaches. Barring something better, I would go with the estimate in the PS above.

    ReplyDelete