What is structural unemployment?

There seems to be a disconnect between how economists (and Fed economists in particular) use this term and how bloggers (and left bloggers in particular) use it.

Some definitions:
“The unemployment resulting from wage rigidity and job rationing is sometimes called structural unemployment.” — Mankiw’s intermediate text

“Frictional unemployment [w.a. where the only other kind of unemployment they mention is cyclical unemployment] is the unemployment that exists when the economy is at full employment.” — Dornbusch, Fischer and Startz

“The part of unemployment associated with the institutional features of an economy, including hiring and firing costs and the structure of the unemployment compensation system.” — Jones’ intermediate text

There may be confusion about this term because there seems to be such a multitude of definitions (wikipedia’s definition talks about skill mismatch). I like Jones’ definition the best, but the point I want to make is that the lefty bloggers are reading more into these definitions than necessary.

There’s no reason to think the relationships between “wage rigidity”, “full employment”, “institutions” or “skill mismatch” and unemployment are constant over the business cycle. You can make good arguments that the mechanisms producing these relationships are responsive to overall conditions in the economy (I won’t). And… and!… estimates of structural unemployment suggest that it is time-varying: higher during recessions and lower during booms (anyone have a good cite for an estimate of time-varying NAIRU?).

When the Fed folks say structural unemployment is high right now (e.g. Altig or Kocherlakota), they’re not saying its permanently high. They’re not trying to pull an inception making us feel like permanent high unemployment is ok. They’re observing momentarily high structural unemployment that fits in a pattern consistent with historic experience.

17 thoughts on “What is structural unemployment?”

  1. The CBO published a “natural rate” estimate. It seems its up from 4.8 to 5%

    http://www.cbo.gov/ftpdocs/117xx/doc11705/2010_08_19_SummaryforWeb.pdf [pdf]

    I’m not familiar with another source that estimates it.

    Regarding structural unemployment, this shift in the Beveridge curve is nothing new- it is well-known. It shows up in Pissarides’ textbook on search-and-matching theory (p. 26)- I’m surprised Kocherlakota failed to cite this or mention it, as he seems to have spent a lot of time with Shimer’s mismatch paper.

    See for example:Why Has the U.S. Beveridge Curve Shifted Back? New Evidence Using Regional Data by Valletta, 2005, SF Fed

    “For example, under the reasonable assumption that vacancies adjust to shocks
    more quickly than does unemployment, the return to an initial Beveridge curve equilibrium after a cyclical shock will follow a counterclockwise loop, with vacancies adjusting upwards more quickly than unemployment falls (Bowden 1980, Blanchard and Diamond 1989). Such
    adjustments should be relatively short lived, however, occurring at business cycle frequencies, as distinct from longer term shifts in the position of the Beveridge curve that suggest persistent, structural shifts in match efficiency.”

    The uptick in vacancies without a corresponding decrease in unemployment is a short-run phenomenon. It’s easy to test- the standard theory would predict that the Beveridge curve will start moving northeast soon, while eventually returning to its original curve. But to draw the conclusions that Kochelakota do doesn’t follow at all from search and matching theory. All this is at the end of Altig’s piece, but Altig doesn’t seem to believe it.

  2. northwest* soon

    Economist’s view had a good post on this from 2006


    We already had two “jobless recoveries” where unemployment was slow to recover. Why should this recession be any different? We returned to full employment of 4-5%. No structural unemployment problems, though the vacancy rate increased faster than unemployment in the early stages.

    That doesn’t mean that Kocherlakota can say that monetary policy can’t affect unemployment much because of structural unemployment. I never cease to be amazed the intellectual pretzels some will tie themselves in to avoid talking about liquidity traps. Maybe mismatch is causing all the excess reserves too?

  3. I don’t follow. Are you saying these cyclical movements in the beveradge curve do not represent cyclical changes in the efficiency of the matching function? And do you think the efficiency of the matching function is a policy variable?

  4. The beveridge curve is generally a curve, however, during a recovery (when a country is on the lower right edge of the Beveridge curve), vacancies rise initially before unemployment falls. So the points move up and then counterclockwise. Once you get to the upper right hand quadrant. So this isn’t anything new with this recession or some sort of “structural unemployment” that was different this time than during other recessions. Monetary policy ended the 1990 and 2000 recessions, and the current crisis could have been resolved with monetary policy if monetary policy wasn’t ineffective now.

    Basically, the rise in vacancies without a fall in unemployment is taken by Kocherlakota as evidence for structural unemployment, and he proceeds to stick the numbers in a model (sShimer’s mismatch). I’m saying that this rise in vacancies without a corresponding fall in unemployment is a normal part of the American business cycle and has nothing to do with structural unemployment.

    For supporters of structural unemployment, the definition of structural unemployment should be more clearly stated (as you said) and there need to be more testable implications. For me, temporary structural unemployment is kind of contradictory, but if that’s the way its defined, I can believe it. For me structural unemployment is relatively long-term unemployment that’s related to mismatch (no jobs for a worker’s skill), technological change (50 year old rust belt worker), or, labor market rigidities (ex: Eurosclerosis).

    In response to your question: yes, the efficiency of the matching unction varies over the business cycle. If this is what you mean by “structural unemployment”, then that’s fine. I don’t see this as how Kocherlakota sees structural unemployment. Instead he sees unemployment now as immune to monetary policy, so there is nothing that can be done for the unemployed (as he doesn’t consider fiscal policy). My point is that, yes, vacancies don’t make a dent in unemployment initially in a recovery, but eventually, with sufficient vacancies, unemployment will fall. So monetary policy is not ineffective- it’s just slow out of the gate. The unemployment is not structural, its just that, given the less efficient mathcing function, you have to create more vacancies than you would otherwise.

    I think the issue that I have with theories of structural unemployment thoeries, which might be what other “leftys” have a problem with, is that it seems to be a theory that justifies monetary and fiscal policy ineffectiveness. I see no evidence that structural unemployment is the driver of the current persistent unemployment- its a liquidity trap and weak fiscal policy (once states and localities are considered). Instead of attempting new measures to bring an end to the recession, the structural unemployment thesis implies that nothing can be done. It smacks of Mellonism.

    I don’t know if your definition of structural unemployment and Kocherlakota’s are similar.

    I think there’s two important questions to ask:

    1. Is there anything policy can do to reduce unemployment now? If you’re answer is no, then I see why lefty bloggers don’t like that definition.

    2. What will unemployment look like at the next business cycle peak? For me, any really strucutral unemployment theory should believe that unemployment will be significantly higher at the next business cycle peak than at the previous peak. Maybe yours doesn’t, but the structural unemployment theory seems to be gaining adherents rapidly, so it should be better defined than it is now (as you stated).

  5. How is a movement of the beveridge curve, whether or not it is correlated with the business cycle, not a measure of the efficiency of the matching function?

    Forget the definition of structural employment: if right now it’s harder for unemployed workers to be mapped to job openings, what can the Fed do about it?

    It’s obvious that this isn’t the only thing driving unemployment. The debate should be over effect size, but you seem to be denying this effect exists!

  6. “How is a movement of the beveridge curve, whether or not it is correlated with the business cycle, not a measure of the efficiency of the matching function?”

    This is a measure of the efficiency of the matching function.

    “Forget the definition of structural employment: if right now it’s harder for unemployed workers to be mapped to job openings, what can the Fed do about it?”

    Just because it’s harder (for the time being) doesn’t mean that policy is totally ineffective. Also, this is temporary- if vacancies keep increasing, the beveridge curve will return to normal, just shifted out. The “structurality” is only temporary, and thus not very structural. My main point is that this recession is not that different in terms of recovery phases, but for some reason many see some kind of structural problem that is intractable.

    It may be obvious that SU is not the only thing driving unemployment, yes. But this is no different than what we’ve seen in other recession- why wouldn’t policy work?

    Just look at the following graph for the earl 1980s recovery. We could clearly see a counterclockwise loop, so the early stages of the recovery looked “structural”. But monetary policy was obviously very effective in bringing an end to the recession.


  7. No need to cite the loops in the Beveridge curve. These are well known. The point is that structural unemployment increases in recessions. This is what the loops are telling us. If you think structural employment is constant through recessions, then we’re using different definitions of structural employment. And that was the point of my post. I don’t care how you decide to define structural employment, the Fed guys are using the definition for which its NOT constant.

    To the extent mismatching (which the Fed folks call SU) is driving unemployment, standard AD policy won’t get anybody more employed. At least if you follow the usual mechanisms connecting policy and unemployment.

    Let me ask, how can monetary policy make it easier to match the unemployed to jobs? How can straight fiscal policy?

  8. “To the extent mismatching (which the Fed folks call SU) is driving unemployment, standard AD policy won’t get anybody more employed.”

    I won’t cite the loops again, but the loops imply that, after a brief period where vacancies rise and unemployment doesn’t fall, unemployment will start to fall due to increased AD. The is what happened in the early 1980s recovery, as shown above.

    If increasing AD wouldn’t get more people employed, then we should see V’s instead of loops. There’s something I’m not understanding in your argument. The Fed’s actions have ended almost every recession in the post-war period. This is AD is it not? How else do recessions end?

    “Let me ask, how can monetary policy make it easier to match the unemployed to jobs? How can straight fiscal policy?”

    The Beveridge curve shifting out means that for a given level of vacancies there is higher unemployment. But the Beveridge curve still slopes down, so increasing vacancies will decrease unemployment. The Beverdige curve loop just means that to reach a full employment level of say 5% soon a country needs to increase vacancies more than in a year like 2005. The loops mean that there needs to be a larger policy response than with a stable Beveridge curve, which is quite different than a position where policy is ineffective.

    Unless you think the Beveridge curve is vertical, then increased AD will increase employment as the recovery continues.

  9. Ok, but I realize we’ve been talking past each other. You think the Fed should do more to shift the economy up the Beveridge curve and I’m asking why should we think the Fed could shift the curve to the left.

    I get your point and I agree with it. (But we have to be careful because the Beveridge curve is as likely to be structural, in the econometric sense, as the Phillips curve is.)

    The interesting question (from my point of view) is how much of current unemployment is due to us being low on the curve and how much is due to the fact that the curve shifted to the right. The answer to this question is intellectually interesting but it also informs policy wonks about which policies would be most effective at reducing unemployment. If, like me, you think that much of unemployment can be explained by the curve having shifted, then aggregate demand policy won’t help because AD policy only moves us up and down the curve. This suggests, to me, that other policies might be more helpful: small things like replacing UI with a lump-sum transfer or big things like worker relocation subsidies.

  10. I shouldn’t have said AD policy won’t help if unemployment is due to a shift in the Beveridge curve. I should have said that AD policy would be less effective *relative* to policies that shift the curve.

  11. Yes, I think we have been talking past each other, and we largely agree. I think that policies to shift the Beveridge curve to the left would be a good complement to AD policy. But I see your position and Kocherlakota’s as being fairly far apart, despite both being “structural unemployment” explanations. It seems to get down to *how* structural unemployment is, not not a simple dichotomous structural unemployment or not.

  12. Yes. But let me make Kockerlakota’s point for him.

    AD policy creates job openings, it can’t create job matches. Therefore, until the matching function gets more efficient, AD policy just pushes the Beveridge curve to the right.

    In other words, the Beveridge curve isn’t invariant to AD policy.

  13. I don’t equate inefficiency in the matching function to “mismatch”, maybe this is what is missing. I would have to look into the beveridge loops more, but this is completely normal part of the business cycle and doesn’t support mismatch any more than in any other recession. Do you think other postwar recession were categorized by mismatch? I don’t really see evidence of mismatch.

    Maybe some workers leave the labor force as the economy goes toward the trough, that then return to the labor force when vacancies are opened? That would explain why the unemployment rate doesn’t fall as much as it should. I’m not wedded to any theory, but conveniently rushing to stick the numbers in a mismatch model is inappropriate.

    Basically, a couple of quarters of a vertical beveridge curves doesn’t show mismatch at all, it’s a normal part of the business cycle.

    The mismatch and structural unemployment theories are hardly supported by any evidence, they’re just convenient for those that don’t want any more policy responses to the recession.

    How does AD policy push the Beveridge curve to the right? The UI extensions? Supporters need to show that the demand effect of UI (positive) is outweighed by the labor supply response (negative). Most just cite the latter, but this is incomplete. Plus, vacancies are still far too low for this too be a serious theory. We are not even close to the number of vacancies need to get to full employment. If we get there, then worrying about structural unemployment would be reasonable.

    You like Sumner- he’s exactly right on structural unemployment. Once we are out of this recession, the structural unemployment supporters will regret their support for this theory when unemployment falls below 6%.

  14. Let’s put some numbers on the unemployment insurance numbers:

    Total unemployed in July was 15.137 million. The number of uninsured unemployed as of July 31st was 4,333,520. That means that less than 30% of the unemployed are receiving unemployment insurance! The insured unemployment rate was 3.4% To reduce unemployment by 1% by removing unemployment insurance, you would need for 30% of those insured to find a job immediately! There’s no way that would happen. Unemployment insurance already doesn’t cover that many people in the USA. You’re not going to reduce unemployment very much by altering it, and the welfare effects of removing all income supports to the unemployed vastly outweighs the welfare effects of reducing unemployment.

    And I didn’t even consider the demand impact of giving money to the unemployed who are going to spend much more of that money than other groups.


  15. I don’t disagree with you that the loops are a normal empirical regularity. For policy analysis, though, its imperative to have a story that explains them. I gave you a story, mismatch, that explains them. The policy implications of that story is that AD policy (things that increase vacancies) isn’t as important as supply-side stuff (changes to UI or relocation subsidies where two of my suggestions).

    You need to come up with a story that explains the loops and then we need to see the policy implications of that story.

    BTW, why do you think I don’t want a policy response? I want unemployment to go down just as much as you and I don’t care about inflation except instrumentally (I think controlling inflation helps dampen the business cycle). There’s nothing magical about AD policy. It won’t solve all problems and we know from history that such policy can be hurtful. In this particular circumstance I don’t know if it would hurt, but I’m pretty sure it wouldn’t help. In other words, I think you and your friends’ obsession with it is a distraction from actually helping the unemployed.

  16. Your second comment seems to be an argument against a straw man. I’ve never proposed eliminating income supports for the unemployed. This would have been a terrible idea.

Comments are closed.