Is unemployment overdetermined?
There’s been a lot of recent discussion of long term unemployment. I would caution against putting too much weight on any single factor. For instance, the drop in the labor force participation rate is partly due to boomers retiring, young people staying in school, and the unemployed going on disability. But none of those factors plays a dominant role.
Another mistake is that people sometimes forget that unemployment can be “overdetermined,” i.e. that multiple factors are each powerful enough to explain most unemployment. I hate to pick on Kevin Erdmann, because his post is actually one of the best I’ve read on the subject:
Upon further reflection, I think the approx. .75% of age-related unemployment, anomalous to the recent recession, which I found in the demographics section, is probably closely related to the .7% excess unemployment that I found in the later section on EUI related to excess unemployment duration above 26 weeks. So, in total, of the approx. 5% in cyclical unemployment that we saw at the peak, I am attributing approx. .5% to age demographics and at least .75% to EUI. This leaves 3.75% attributable to other factors, although EUI would likely be responsible for some of the remaining 3.75% in ways that I haven’t been able to isolate here.
Those numbers sound plausible to me, however be careful when partitioning the effects of various factors. To see why let’s go back to my “musical chairs model” of recessions. Suppose 10o people play the game of musical chairs, but there are only 95 chairs to sit in. When the music stops 5 people will end up sitting on the floor.
Each time the music stops, a different 5 people will be sitting on the floor. However it is possible that some people will end up missing out 2 or even 3 times in a row, particularly if there are differences in quickness and agility among the players. By analogy, the 5% frictional unemployment we observe will represent different people each year, although less skilled workers are more likely to show up multiple times.
Now assume two things happen at once. The number of chairs is reduced from 95 to 90, and the people who end up sitting on the floor each time have heavy lead weights attached to their feet. They will be less agile, and thus far more likely to end up on the floor multiple times in a row. However by construction the lead weights have no impact on the total number of people who sit on the floor. If there are 90 chairs, then 10 people sit on the floor. End of story.
This is essentially Paul Krugman’s argument for why extended UI benefits don’t raise unemployment in a recession. The reduction in chairs from 95 to 90 is analogous to the reduction in the employment rate from 95% to 90% when NGDP falls very sharply and nominal wages are sticky. The lead weights are like the disincentive effects of extended UI, which makes it less likely that those workers will find new jobs. (A better analogy might be if those sitting on the floor have soft cushions affixed to their rear ends, to making sitting on the floor less painful. So they don’t try as hard as the other players.)
I’ve challenged Krugman’s argument in the past, but there is clearly a grain of truth in this sort of reasoning. Just because a study shows that the unemployed are suddenly consisting disproportionately of the long term unemployed, as compared to past recessions, doesn’t really prove that the factors leading to long term unemployment have any effect on the overall unemployment rate. They might, but that fact would have to be established in some other way. Unemployment might be overdetermined, i.e. without extended UI there would but just as many unemployed, albeit a different mix of people and durations, and for a different reason.
The problem with my simple example is that extended UI benefits probably do reduce the number of chairs, by making nominal wages less more sticky than if an army of unemployed workers were desperate to find new jobs. Disability benefits and boomer retirement also have that effect, BTW. And if wages are more sticky then nominal GDP shocks have bigger effects on employment than otherwise.
In the end I find the 0.50% and 0.75% figures to be plausible, but they don’t necessarily imply that the remaining 3.75% is an upper limit to the part of excess unemployment that is due to effects of deficient demand. One problem is overdetermination, as I’ve just showed. The other is endogeniety. The extend UI program was itself caused by the demand shock, and would probably be eliminated if demand returned to normal.
Still, it’s a useful exercise to partition the various factors as well as we can, if only because Congress needs accurate policy counterfactuals when making the decision as to whether or not to extend the extend UI benefits. So I applaud Erdmann’s thoughtful post.
PS. In the final paragraph I am assuming that the Dems in Congress are Matt Yglesias-types and the GOP is composed of a bunch of Tyler Cowens. Both groups want to maximize aggregate utility, but disagree slightly on the size of disincentive effects. I think that’s a reasonable description of Congress these days, isn’t it?
PPS. I may have been unfair to Erdmann, as the final sentence I quote suggests that he understands the overdetermination issue.
PPPS. Here’s my favorite part of the Erdmann post:
Here’s the kind of reminder that makes Scott Sumner slap his forehead: The EUI was instituted in June 2008 when the unemployment rate was 5.6%. At the time, the Fed Funds rate was still at 2%.
HT: Tyler Cowen
Tags:
29. September 2013 at 08:38
“The problem with my simple example is that extended UI benefits probably do reduce the number of chairs, by making nominal wages less [more?] sticky than if an army of unemployed workers were desperate to find new jobs.”
29. September 2013 at 08:45
Men under 54 have dropped out, men over 54 have participated more.
http://www.bls.gov/emp/ep_table_303.htm
29. September 2013 at 08:52
“In the final paragraph I am assuming that the Dems in Congress are Matt Yglesias-types and the GOP is composed of a bunch of Tyler Cowens.”
I would strongly disagree, based on my experience with lobbying our most recent Congress (a caveat being that I specialized in education politics, rather than fiscal politics). There are a LOT of congressmen and senators in the GOP who would be way better described as primarily subscribing to a form of deontological ethics with utilitarianism only in the margins, and they control a very significant and coordinated voting bloc, especially in the House. In my experience this insight is pretty key to understanding the nitty gritty of negotiation and voting patterns in our current Congress. Dems like this also exist but are a far less cohesive and influential voting bloc.
Not to mention the importance of class and commercial interests. I challenge you to come up with a non-bullshit utilitarian defense of the GOP obsession with repealing the medical device tax.
29. September 2013 at 09:08
Thanks Pietro, I corrected it.
Waffles, I should have added a smiley face, I meant that as a joke.
29. September 2013 at 09:38
ssumner, ah, my mistake. Poe’s law for false equivalence, I suppose. 😛
29. September 2013 at 09:48
“I should have added a smiley face, I meant that as a joke”.
☺May I suggest that in the future everyone insert *two* smiley faces—one at the beginning of the joke and one at the end? This should avoid further unnecessary confusion.☺
See example above.
29. September 2013 at 09:56
☺I’m glad you were here to contribute, Vivian.☺
29. September 2013 at 10:03
Vivian, 🙂 Two smiley faces would have improved your comment at the previous post. 🙂
29. September 2013 at 10:08
Scott, I’ve got to hand it to you—you’re a real sport. And, that ain’t no joke!
29. September 2013 at 10:33
:):)
(That means, “No joke”.)
29. September 2013 at 10:33
Ah, I meant this!
🙂 🙂
29. September 2013 at 10:40
Scott, don’t you neglect overdetermination yourself when you make arguments like, “The drop in NGDP in 2008 is *fully capable* of accounting for the observed unemployment. We got about the unemployment that we would have predicted, given the drop in NGDP” ?
29. September 2013 at 12:17
[…] The problem with the Cato logic, as adopted by the ME GOP, is that there is no evidence supporting the argument that removing welfare would increase demand for labor. As Scott Sumner notes on his blog today: […]
29. September 2013 at 12:36
Saturos, My point was that the NGDP drop by itself could explain unemployment in a sort of “order of magnitude” sense. I suppose I erred in leaving the impression that no other factor played any role at all. I was pushing back against the notion that our high unemployment is some sort of mystery that needs to be explained. It’s what you’d expect after the biggest adverse demand shock since 1938. But I’ll try to be clearer next time.
29. September 2013 at 21:03
Scott, I agree with everything you say here. Very fair points.
In some ways these factors are complimentary issues with the demand factors, rather than alternatives. The initial unemployment shock could have been entirely demand related, and these factors would have simply reflected how we regard the recovery. As you mention, demand based solutions might be battling headwinds from this. Also, I would argue that especially the demographic factor causes the UE rate to be less reliable.
Also, a lot of the excess reported unemployment from age and from EUI is related to flows out of the labor force, not back into employment. So, to use your analogy, there might be 5 people who would have left the party a little earlier than they had wanted to, maybe even planning to come back later for the next round of musical chairs, and now they are hanging around for a while since the host offered them another drink.
29. September 2013 at 21:13
Unemployment is never evenly distributed — the heterogeneity of labour is one of its key characteristics. A lot of employer behaviour revolves around difficulties in gauging aptitude and commitment. Hence, for example, the importance of “friends, relatives, etc” as labour market intermediaries–they provide both more information and act as inform guarantors. Hence also the problem with “job protection laws” as per much of the Eurozone, which really drive up youth unemployment in particular. (And minority youth unemployment even more so.)
The question of who will end up “sitting on the floor” is much more a consequence of how many will “sit on the floor” when demand is falling. It has rather more effect on how many will sit on the floor as demand is recovering, as you get supply mismatches between who is available and what jobs are available as being unemployed affects your desirability as an employee and does more so the longer you are unemployed. In other words, unemployed people provide less competitive pressure the longer they are unemployed.
It is in fact better (both macro-economically and micro-economically) to have more people experience shorter periods of unemployment than fewer people experience longer periods of unemployment, for any given level of unemployment.
Of course, it is better not to screw up your demand management in the first place.
29. September 2013 at 21:14
That should be “informal guarantors”.
29. September 2013 at 21:22
Oh, and your musical chairs analogy reminded me of a blog post I’ve been meaning to write. The short version is that income inequality would be much higher in a pure meritocracy compared to actual outcomes because of information asymmetries in the labor market. EUI has provided some new information to employers which leads to this bifurcation of low and high duration unemployment. This has created a more meritocratic labor market with deeper inequality.
Maybe I shouldn’t summarize the idea here without airing out all the details of my logic, since it probably seems impolite to refer to any unemployment as meritocratic. But I think this problem poses some difficult challenges for social support policies.
30. September 2013 at 05:10
kebko and Lorenzo, Both very interesting comments. Good points.