Why did long-term unemployment rise in June 2018?

Ernie Tedeschi
Jul 20, 2018 · 8 min read

Long-term unemployment jumped by 289,000 in June 2018, bucking a mostly-steady decline since 2010. I show that this rise was statistically significant and driven entirely by a decline in outflows from long-term unemployment — particularly outflows into nonparticipation — rather than a surge in inflows into long-term unemployment. However, I caution that even a statistically-significant one-month change in the CPS is not dispositive of a sustained change in trajectory.

The June 2018 jobs report was generally a solid one, with payrolls growing by 213,000 — well above trend, even at this late stage in the recovery — and the labor force participation rate rising by 0.2 percentage points.

There were blemishes however. For one, average hourly wages rose 2.7% over the preceding 12 months; but with consumer prices measured by CPI rising 2.9% over this same period, that means inflation-adjusted wages actually fell slightly over the past year. [1]

Another negative surprise was long-term unemployment (people who say they’ve been actively looking for work for 27 weeks or more), which had been falling steadily since mid-2010 but then jumped by 289,000 on a seasonally-adjusted basis in June. One would have to go all the way back to the economic dark days of May 2010 to find month-on-month growth that high.

So this raises an apparent contradiction: the US labor market is in many ways strong or improving, but long-term unemployment rose. Why?

It may have been a one-off, but it wasn’t just statistical noise.

The first thing to note is that the Current Population Survey (CPS) — the monthly survey of around 60,000 households that gives us the unemployment rate — can exhibit fluctuations over the short-term. We should never interpret a single month’s change as definitive signal in the CPS.

I repeat: everything you read in this post could end up being a one-off fluke that disappears in a month or two.

Now, that said, it’s also important to note that June’s rise in long-term unemployment was statistically significant, with a 95% confidence interval of +225,000–353,000.

Again, this doesn’t mean that June’s rise wasn’t caused by other one-time idiosyncratic factors, such as, say, the weather. But we can confidently conclude that in June, at least, the rise wasn’t simply a case of random noise.

Long-term unemployment is like a bathtub. No, seriously.

Let’s pretend for the rest of this post that the June 2018 rise in long-term unemployment wasn’t a one-off. We need a deeper autopsy of the CPS to see what drove it.

The number of people in any labor market status over time— such as long-term unemployment — can be thought of as a bathtub with multiple faucets and multiple drains. From month to month, the level of the water changes depending on the amount of outflows — how much water (how many people) drains (leaves) from the bathtub (long-term unemployment) in each of the drains (reasons for leaving long-term unemployment) — as well as inflows, or how much water is added from each of the faucets (reasons for entering long-term unemployment). Here is my crude attempt at illustrating this:

The net change in the level of long-term unemployment is just the difference between total outflows and total inflows.

In the CPS, we can think of inflows into and outflows from long-term unemployment as involving three different sources or destinations:

  • Short-term unemployment, if an unemployed person passes the 27-week threshold;
  • Employment, if an employed person becomes unemployed again and tells the government they’ve been looking for more than 27 weeks (this might happen if the worker had a temporary job); and,
  • Nonparticipation, if people who aren’t looking for a job start actively looking again, but tell the government they’ve been looking for more than 27 weeks.

As a side note: yes, every month there is a nonzero number of people who switch from being long-term unemployed to short-term unemployed. Remember, the CPS is a survey, so not all of the answers make immediate sense to those of us who can’t follow up with the respondents!

LTU inflows have fallen but flattened in recent years; outflows continue to fall.

BLS produces some broad month-to-month flow data from the CPS, but without the distinction between short- and long-term unemployment that we need. [2]

So for this analysis, I use the raw CPS microdata to link the same individuals from month-to-month, and then I look at what happened between May and June 2018. Since it’s difficult to adjust for seasonality with individual microdata records, I compare the results of May-June 2018 with what occurred in May-June in prior recent years. [3]

Below, I’ve graphed the May-June long-term unemployment (L) flows for each year since 2008, identified by short-term unemployment (S), employment (E), and nonparticipation (N). My notation is written as “start2finish”, so e.g. “L2E” means the outflow of workers from long-term unemployment to employment.

One thing that’s interesting to note from the summary data right off the bat is that both inflows and outflows increased in response to the Great Recession. Overall long-term unemployment rose during those dark economic days, but below the surface of those aggregate increases was a lot of churn.

Another thing that’s interesting (and this will turn out to be key in a moment) is that while inflows into long-term unemployment have leveled off a bit in recent years, outflows from long-term unemployment (especially into employment and nonparticipation) are still falling.

Weak outflows drove the June 2018 jump in long-term unemployment

Using my approach outlined in footnote 3, and bearing in mind that this is a non-seasonally-adjusted number, I find an increase in long-term unemployment in June 2018 of 165,000. By contrast, I find that the average June change in long-term unemployment over the 2015–17 was a fall of 147,000. So the overall swing from the 2015–17 June average to June 2018 was about +313,000. It’s this swing that I decompose below into its component parts.

Remember that each of these metrics is expressed as June 2018 change minus average 2015–17 June change. So a negative inflow number indicates that that factor is otherwise lowering the growth in long-term unemployment. A negative outflow number by contrast shows an influence that is accelerating long-term unemployment growth.

Notice first that the swing in inflows was slightly negative. That means that, relative to the prior three Junes, June 2018 actually saw somewhat lower inflows into long-term unemployment.

What’s particularly surprising is that this fall in inflows was driven by a fall in nonparticipants becoming long-term unemployed. When long-term unemployment rises suddenly in a strengthening labor market such as ours in 2018, one logical explanation is that people out of the labor force are now actively looking for jobs, and simply telling the government that they were actively looking this whole time.

But the analysis here shows that this isn’t the case. In fact, there were fewer people entering long-term unemployment from nonparticipation in June 2018 than in the prior three years.

What did drive the swing was a substantial fall in outflows. And the news here is not entirely bad. Almost 233,000 fewer workers gave up looking for work in June 2018 — that is, went from being long-term unemployed to nonparticipants — relative to the 2015–17 June average. One could plausibly interpret that as a sign of labor market strength rather than weakness.

But on the other hand, 115,000 fewer long-term unemployed workers found jobs in June 2018 relative to the past three Junes. That could be a warning sign to the economy that the labor market is hitting capacity constraints, though I’m skeptical of this hypothesis given the continued falls in other margins of slack such as .

It could alternatively be a sign that at this point the remaining long-term unemployed are just better equipped to hold out for better jobs or higher wages; the increasing frequency of positive economic news may be convincing the unemployed to be patient for a bit longer.

Or, again, June 2018 could simply be a fluke. It certainly wouldn’t be the first time the CPS gave us a head fake with one month’s worth of date.

So the most important conclusion at this point is that common economist refrain: We need more data to be sure.

[1] Consumer prices adjusted for product substitution effects — “chained CPI” — rose by 2.7% over the June 2017–June 2018 period, so even under that alternative inflation measure real average wages were flat over the last 12 months. The Federal Reserve’s preferred inflation measure — growth in the Personal Consumption Expenditure Price Index (PCEPI) — hasn’t yet been released for June; PCE inflation came in at 2.3% year-on-year in May 2018.

[2] BLS produces CPS labor market status flow table that distinguish between employment, unemployment (of any duration), and nonparticipation, as well as marginal inflows and outflows. They also produce more detailed unemployment flow tables that do distinguish the final labor market state by unemployment duration (27–52 weeks as well as 53+ weeks), but that analysis only covers workers who start out as unemployed, and the analysis does not distinguish the unemployment duration of the starting position.

[3] I use a matching approach similar to . I narrow my sample to just those households who were present in both May and June of each year. I then use BLS’s included month-to-month longitudinal weights, but reweighted by labor force status to ensure that my narrower sample matches the labor force totals for May of each year. In June, each record simply takes their May weight. Keeping the same weight in June has the upside of allowing all inflows and outflows to be perfectly accounted for in the data. The downside is that June’s labor force totals each year end up being slightly off from the official totals. This reflects the absence of accounting for marginal inflows (e.g. 15 year olds turning 16) and marginal outflows (e.g. deaths), as well as other population changes. In June 2018, for example, the civilian noninstitutional population grew by 188,000 people. This is a first approximation of how far off the totals of my narrower sample will be in June. My judgment is that very little if any of this one month of population growth would be classified as long-term unemployed.


Ernie Tedeschi on labor economics, public finance, data visualization, miscellanea

Ernie Tedeschi

Written by

Economist & budget wonk. Fmr @USTreasury economist. Data viz enthusiast. Everything here is my own opinion, and RTs/favorites are not endorsements.


Ernie Tedeschi on labor economics, public finance, data visualization, miscellanea

More From Medium

More from Ernie Tedeschi

Also tagged Labor Market

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade