Why do some states have such low unemployment rates?

That is for June, Kentucky is at 4.3 percent but West Virginia at 10.4?  Here are the underlying BLS data.  Here is some description of Kentucky in June.  Note that Kentucky cases are now rising rapidly.  Here is the case pattern for Idaho.  The three states with the highest unemployment rates — New York, New Jersey, and Massachusetts — have been moving toward relative safety after some very tough times early on.  One interpretation of these numbers is that serious lockdowns were necessary to stem the virus, but those lockdowns caused high unemployment.  More plausible to me is the view that a high initial virus burden led to high unemployment — consumers were scared — but also superior case and death results later on.

Via Nick Clerkin.

13.3% unemployment rate

That one surprised me, as indeed it did most other economists.  What should I learn from this episode?  After all, labor market adjustment was relatively slow coming out of the 2008 crisis.

My tentative hypothesis is that “matching” is more important than I had thought (and I already thought it was quite important, relative to other macro commentators).  One feature of the current layoffs and rehirings is that the ties between workers and firms apparently were not so severed in the first place.  For most sectors (cruise ships aside, etc.), no “rematches” were required, and so rehirings were accomplished very quickly.  As demand (partially) returned, employers wanted at least some of the old workers back, and workers wanted their old employers back, and then it happened.  “Figuring out where I belong” did not slow down the process very much.

That is good news for the remainder of the recovery, provided the recovery happens soon, and it is at least one factor (not necessarily decisive, of course) militating in favor of a speedier reopening.  “Reopen before the worker-employer ties are lost!”

It also implies that during regular, non-pandemic downturns a lot of the slowness of labor market recovery has to do with matching rather than demand per se, noting that the two interact.  And that is a sign of a more general pessimism for the future, since demand problems are easier to fix through policy than matching problems are.

Another possible implication of the new numbers is that employers realized that “F*** it, I want to get back out there” is the prevalent consumer and also worker attitude, whereas Twitter-bound intellectuals were slower to see the same.

Unemployment is always about the supply side too

We document that unemployment is increasing with GDP per capita. Furthermore, we show that this fact is accounted for almost entirely by low-educated workers, whose unemployment rates are strongly increasing in GDP per capita, rather than by high-educated workers, whose unemployment rates are not correlated with income.

That is from Ying Feng, David Lagakos, and James E. Rauch.  In their core model, reallocating low-education workers to the formal sector makes them harder to reemploy at short notice, in contrast to the informal sector and self-employment.  An alternative view, not mutually exclusive, is that in poor societies low-education workers simply have to take jobs, due to extreme need.

You will on Twitter, and in blogs, see various attempts to mock supply-side theories by showing increasing employment, often accompanied by remarks such as “I didn’t know video games were getting so much worse.”  Such comments are a mistake and a misunderstanding.  Proper supply-side theories do not deny the relevance of the demand-side, and so nor should demand-side theories deny the relevance of the supply-side.  It is possible to believe both “supply-side factors made the labor market recovery slower than usual,” and “demand-side forces have at this moment overcome many of those problems.”  Just look at the disability rolls.  The ability to receive disability kept many people out of the labor force in earlier years, slowing down labor market recovery.  Yet it is also true that currently demand-side forces are creating jobs good enough that many of those same people finally are leaving the disability rolls.

The deeper lesson of course is that — outside of the short-run — demand-side forces are supply-side forces.  And right now we are out of the short run indeed, at least when it comes to macroeconomic shocks.

The immigration–unemployment nexus: do education and Protestantism matter?

That is the title of a new paper by Jakob B. Madsen and Stojanka Andric, here is the abstract:

Using annual data from 1850 to 2010 for Argentina, Australia, Brazil, Canada, New Zealand, and the USA, this paper examines the impact of immigration and the immigrants’ educational and cultural background on unemployment. Instruments for 27 emigrating countries are used to deal with the feedback effects from unemployment to immigration. The results show that educated immigrants, in particular, and immigrants from Protestant countries significantly reduce unemployment, while poorly educated and non-Protestant immigrants enhance unemployment.

For the pointer I thank the excellent Kevin Lewis.

Why did the U.S. unemployment rate fare better than gdp growth?

That is a new paper by Bob Hall (you must scroll down to get to the pdf), here is the abstract:

Answer: Between 2007 and 2014, GDP growth was held back by shortfalls of
 4.4 percent in productivity
 4.0 percent in capital input
 3.6 percent in labor-force participation
 2.2 percent in growth of the working-age population

Any further questions you might have?

There are other interesting macro papers at the link, and hat tip goes to Greg Mankiw.

Skill mismatch unemployment is real and significant

Even during demand-driven recessions.  Part of the problem is that cyclical and structural causes of unemployment interact and magnify each other.  Here is the job market paper of Pascual Restrepo, one of the stars from MIT currently on the job market.  I turn the floor over to him:

To study the effect of structural change on labor markets, I build a model in which structural change creates a mismatch between novel jobs skill requirements and workers’ current skills. When the mismatch is severe, labor markets go through a prolonged adjustment process wherein unemployment is amplified and job creation is low. Due to matching frictions, firms find less workers with the requisite skills for novel jobs and they respond by creating fewer jobs. The paucity of novel jobs creates an external amplification effect that increases unemployment for all workers—including those who already hold the requisite skills—and discourages rapid skill acquisition by workers. Structural change is not only a secular process; it also interacts with the business cycle, causing a large and long-lasting increase in unemployment that concentrates in recessions. I demonstrate that the decline in routine-cognitive jobs outside manufacturing—a pervasive structural change that has affected U.S. labor markets since 2000—caused a severe skill mismatch that contributed to the long-lasting increase in unemployment observed during the Great Recession. My evidence suggests that this external amplification effect is important. Moreover, I find that the skill mismatch amplified and propagated demand shocks at the local labor market level.

How many times during the last five years have I read or heard critiques of structural theories which neglect their more sophisticated forms?  (“What, did everyone in 2008 simply forget…?” etc.  Be very suspicious of the structure of that argument.)

Here are two other interesting papers by Restrepo, including one on how to share income with the robots, co-authored with Acemoglu.  I agree with their conclusion: “We find that inequality increases during transitions, but the self-correcting forces of the economy limit the increase in inequality over longer periods.”

Larry Summers on technological unemployment in history

This bit is from the Q&A session:

LS: So, I guess I think there is both a, you know, Keynes as usual I think was pretty smart and you know, Keynes began his essay on economic possibilities for our grandchildren by saying that there was this really pressing cyclical problem that had to do with demand which was really important but not all that profoundly fundamental. And there was this more fundamental thing which was that technology was marching on and he thought the dis-employment effects would show up as everybody working 15 hour weeks. And it doesn’t look like that’s quite what they’re showing up as. But the basic idea that technological progress comes with reduced labor input, sometimes it’s early retirement, sometimes it’s people who aren’t able to get themselves employed, sometimes, it’s lower hours, but that is basically the story of the last 150 years.

So, I would not back off of my putting a lot of weight on technology as something important here.

The talk and dialogue (pdf) are on macro more generally, interesting throughout.  In general I believe there should be more transcribed and summarized dialogues, both the NBER and Brookings have had intellectual success with that format.

Claims about electricity adoption and technological unemployment

This is from a recent working paper (pdf) by Miguel Morin:

When the adoption of a new labor-saving technology increases labor productivity, it is an open question whether the economy adjusts in the medium-term by decreasing employment or increasing output. This paper studies the effects of cheaper electricity on the labor market during the Great Depression. The first-stage of the identification strategy uses geography as an instrument for changes in the price of electricity and the second-stage uses labor market outcomes from the concrete industry—a non-traded industry whose location decisions are independent of the instrument. The paper finds that electricity was an important labor-saving technology and caused an increase in capital intensity and labor productivity, as well as a decrease in the labor share of income. The paper also finds that firms adjusted to higher labor productivity by decreasing employment instead of increasing output, which supports the theory of technological unemployment.

You will note of course that the short-, medium- and long-run effects here are quite different, and of course electricity is a major boon to mankind.  Still, technological unemployment is not just the fantasy of people who have failed to study Ricardo.

Here is a short summary of the paper, via Romesh Vaitilingam.

What is China’s Unemployment Rate?

What is China’s Unemployment Rate? 4.1% For what month, what year? Doesn’t matter the answer is still 4.1%. That’s a slight exaggeration but for the last 3 years the unemployment rate has been 4.1% almost every month. Indeed, since 2002 the official unemployment rate has varied between 3.9% and 4.3%, an absurdly smooth series.

In contrast to the unemployment rate, China’s GDP growth rate has had massive swings. As a piece in Quartz puts it the unemployment rate exhibits an eerie stillness.


A new NBER working paper uses a newly available household survey and finds a very different series–the China-UHS series shown in black below. According to these estimates China’s unemployment rate shot up to around 11% in 2002 and has been nearly that high at least until 2009 when unfortunately the new series ends.

UE Rate China

So how high is Chinese unemployment today? No one knows but it could well be closer to 10% than to 4.1%.

Keep an eye on China and don’t be surprised by the unexpected. In China it’s not just the unemployment rate that is more volatile than it appears.

Declining Desire to Work and Downward Trends in Unemployment and Participation

That is the next (and for me final) NBER paper from the macro workshop, by Barnichon and Figura, the pdf is here.  Their main claim is quite startling, and very important if true.  Here is the abstract:

The US labor market has witnessed two apparently unrelated trends in the last 30 years:a decline in unemployment between the early 1980s and the early 2000s, and a decline in labor force participation since the early 2000s. We show that a substantial factor behind both trends is a decline in desire to work among individuals outside the labor force, with a particularly strong decline during the second half of the 90s. A decline in desire to work lowers both the unemployment rate and the participation rate, because a nonparticipant who wants to work has a high probability to join the unemployment pool in the future, while a nonparticipant who does not want to work has a low probability to ever enter the labor force. We use cross-sectional variation to estimate a model of nonparticipants’ propensity to want a job, and we find that changes in the provision of welfare and social insurance, possibly linked to the mid-90s welfare reforms, explain about 50 percent of the decline in desire to work.

Did you get that last bit?  Wild.  The Clinton-era welfare reforms lowered the incentive to work.  Another part of the paper explains the possible mechanisms in more detail:

We conjecture that two mechanisms could explain these results. First, the EITC expansion raised family income and reduced secondary earnersís (typically women) incentives to work. Second, the strong work requirements introduced by the AFDC/TANF reform would have, through a kind of “sink or swim” experience, left the “weaker” welfare recipients without welfare and pushed them away from the labor force and possibly into disability insurance.

The authors have strong reputations, but is it true?  Stay tuned, and look for my live-blogging in the comments section of this post…

Optimal life cycle unemployment insurance

It seems we should index unemployment benefits to a person’s age.  For the liquidity-constrained, human capital-investing young, we don’t want to rush them into unsuitable jobs.  The older workers — that’s another matter.   Michelacci and Ruffo report:

We argue that US welfare would rise if unemployment insurance were increased for younger and decreased for older workers. This is because the young tend to lack the means to smooth consumption during unemployment and want jobs to accumulate high-return human capital. So unemployment insurance is most valuable to them, while moral hazard is mild. By calibrating a life cycle model with unemployment risk and endogenous search effort, we find that allowing unemployment replacement rates to decline with age yields sizeable welfare gains to US workers.

The AER version is here.  Ungated versions are here.  Elsewhere Ben Casselman considers some ways that unemployment has changed, and what that may mean for benefits.

How much did cutting unemployment benefits help the labor market?

Quite a bit.  There is a new NBER Working Paper on this topic by Hagedorn, Manovskii, and Mitman, showing (once again) that most supply curves slope upward, here is one key part from the abstract:

In levels, 1.8 million additional jobs were created in 2014 due to the benefit cut. Almost 1 million of these jobs were filled by workers from out of the labor force who would not have participated in the labor market had benefit extensions been reauthorized.

There is an ungated copy here (pdf).  Like the sequester, this is another area where the Keynesian analysts simply have not proven a good guide to understanding recent macroeconomic events.

Good sentences about male and female technological unemployment

I think that if you look only at males in isolation, you will see this in the data. That is, men are working much less than they used to. For some men, this leisure is very welcome, but for others it is not. In that sense, I think that we should look at the [technological unemployment] fears of the early 1960s not as quaint errors but instead as fairly well borne out.

For women, the story since the 1960s is different. In the economy as a whole, the share of labor devoted to preparing food, washing clothes, and cleaning house has gone down. Also, a higher share of the remaining work in these areas is coming from the market, via restaurants and cleaning services, rather than from unpaid female labor. The upshot is that, from the 1960s to about 2000, we saw a continuation of the trend for women to increase their share of market work and reduce their non-market labor. So, while men were increasing their leisure, women were increasing their market work. Combining men and women, you would not see a decline in market work.

It seems that around 2000, the trend for more market work by women reached its peak, making the trend toward technological unemployment more visible. From now on, what was happening to men before will be what happens to the total labor force. That is, leisure will go up, and some of it will be less than voluntary.

That is from Arnold Kling.

Was there mismatch unemployment during the Great Recession?

I remember this question being debated extensively circa 2009-2011, and those who said there was a (limited) role for mismatch unemployment were mocked pretty mercilessly.  Well, Sahin, Song, Topa, and Violante have a piece in the new American Economic Review entitled “Mismatch Unemployment.”  (You can find various versions here.)  It’s pretty thorough and state of the art.  Their conclusion:? “…mismatch, across industries and three-digit occupations, explains at most one-third of the total observed increase in the unemployment rate.”  The people thrown out of work could not be matched as well as the unemployed workers of the past.

Much of the matching problem was for skilled workers, college graduates, and in the Western part of the country.  Geographical mismatch unemployment did not appear to be significant.  Now, “at most one-third” is not the main problem, but it is not small beans either.  That’s a lot of people out of work because of matching problems.

Again, the Great Recession arose from a confluence of supply and demand problems.