Why has US participation jumped over the last five months?
The labor force participation rate — the percent of the population who either have or are actively seeking a job — has been a fairly gloomy statistic for most of the US recovery. It fell by 3.6 per cent between December 2007 and September 2015. Not all of this was surprising: demographers have known for some time that America was ageing on average, and as we get older we’re much more likely to leave the labor force as we retire, get ill, or become disabled. But even after adjusting the participation rate to control for these shifts, we’ve still seen a 1.4 percentage point decline in the wake of the Great Recession. So demographics can explain much, but far from all, of the labor force dynamics since 2007.
Happily, since September the news on participation has been sharply, and quite suddenly, positive. The headline participation rate has risen by 0.5 percentage points. That’s an extraordinary rise: 1992 was the last time the US had seen a comparable five-month increase in the headline participation rate. And again adjusting for the underlying downward trend of demographics, the recent increase has been an even-larger 0.6 percentage points; one would have to go all the way back to 1981 to find a five-month jump of equal or greater magnitude.
This story is told in these two charts. I’ve decomposed the fall in participation since December 2007 and the rise since September 2015 by splitting up the rate of non-participation (the percent of the population not actively looking for work) into seven factors: demography, people who say they want a job but aren’t actively looking (“discouraged workers”), and, among those who say they don’t want a job, people who are disabled or ill, enrolled in school, retired, staying at home to take care of home or family, or any other reason. Note I’ve charted these factors in terms of their effect on participation: when these nonparticipation factors rise, participation falls, so they show up as negative on the charts. Conversely, when they fall, participation rises, and so their participation effect is positive.
If the contributions of these factors look different than similar charts you may have seen, it’s probably because I’ve controlled for demography separately. This is most obvious with retirement, which in an unadjusted chart would be the single biggest factor driving down the participation rate. But I’d argue that doesn’t tell us much about what we’re really interested in, which is why participation fell and whether or not it will rebound. After all, America was going to get older regardless of the economic catastrophe of 2007 onward, and older people eventually retire. So I lump the changes in retirement (and disability and any other factor) that we would have expected given the demographic changes since 2007 into a catch-all “demography” category, and show the other factors with these demographic effects stripped out. Retirement, for example, now becomes a net boost to the (demographically-adjusted) participation rate, indicating that for any given worker of a certain age and sex, nonparticipation due to retirement has fallen on average since 2007.
Demography and school enrollment have been the biggest negatives since the recession started.
In fact, the first chart shows that the single biggest factor driving down the participation rate since December 2007 has indeed been demography, which has shaved off a cumulative 1.8 percentage points. While Census projects that ageing will slow somewhat in 2016 (before re-accelerating through 2020), absent any major shifts or surprises (such as a large influx in immigration) this factor will only grow. So don’t expect headline participation to return to its pre-2007 highs anytime soon.
Of the non-demographic factors, the largest has been school enrollment — meaning high school, college, or any other educational program — which has lowered demographically-adjusted participation by about a full percentage point. The rise in self-reported disability or illness is responsible for 0.6 percentage points. Both of these effects have been fairly stable since at least 2013, though if you squint they may be trending positive as of late. Discouraged workers, who I define here as any nonparticipant who says they want a job regardless of the reason they give for nonparticipation, have taken off 0.4 percentage points, but this effect has shrunk markedly since August 2012 when it took off 0.9 percentage points. On the flip side, as I mentioned before, retirement has boosted participation by 0.7 percentage points on a demographically-adjusted basis. Home/family and other reasons have had negligible impact on an aggregate basis.
…but school enrollment and retirees have been the main positives since September 2015.
Now, on to the recent data, and first a word of caution. Five months of data is probably enough to be comfortable that there’s some real signal in the overall participation rise, but the underlying categories I’m using are still volatile so the story may change over time. Indeed, you can already see some shifting among the categories in the second chart.
Also, since I’m slicing the official data into unofficial categories and seasonally-adjusting it myself, the results won’t necessarily be exactly apples-to-apples which what the Bureau of Labor Statistics publishes. In the end, even though I’m using methods similar to what government statisticians use, these are not official releases.
All that said, we can tentatively say that the participation story of the last five months appears to be one of students and retirees. The falls in their nonparticipation rates have boosted the adjusted participation rate by about 0.2 and 0.15 percentage points respectively since September 2015. This is interesting because as the first chart showed, these factors were the bookends of the adjusted participation story since December 2007: students were the single biggest negative factor, while the fall in retirement rates had boosted participation once demography was controlled. This is a bit surprising. A reasonable hypothesis given the rapid rise in headline participation over the last five months would have that it was due to a decline in nonparticipation factors that were unusually elevated since the recession. Put another way, there was “slack” in these factors that would eventually revert to some trend, potential, or pre-crisis level. The school enrollment data is arguably consistent with this story. But retirement nonparticipation was already below its pre-crisis levels on a demographically-adjusted basis. So the “slack” hypothesis is at best only a half-explanation, at least with the data we have so far.
Note too that declines in nonparticipation due to the disabled or ill, and due to discouragement, boosted participation sharply in February 2016. I would love for this to be signal but one month of unofficial data that I’m adjusting on my own is not enough to give these factors credence yet (note again though that over a year-long horizon discouragement has been trending positive for participation). Their positives this month could very well be noise. If they do continue to push participation up however then clearly the possible explanations broaden.
Conclusion
The bottom line is that the data are showing an apparent fall in adjusted nonparticipation due to retirement and school enrollment over the last five months, with some hopeful-but-preliminary signs of effects from other factors as well. This behavior, combined with the slightly longer-term fall in nonparticipation among discouraged workers, has been positive for US labor force participation.
Of course, breaking down the changes in nonparticipation this way doesn’t really tell us the “why” of the boost in participation, it just tells us the “who”. But that’s an important first step in understanding what’s really going on in the US labor market. The hypotheses of the “why” will have to wait for another post.