Participation and the Hot Labor Market

Summary

  • The labor force participation rate of prime-age (25–54 year old) Americans has risen almost 2 percentage points since its trough in 2015.
  • Labor force growth must be due to some combination of higher labor force entry rates and/or lower labor force exit rates; it is an open question how much each has contributed to the recent rise.
  • A recent study published by the Federal Reserve Bank of San Francisco (Barnichon (2019)) attributes virtually all of this rise to lower labor force exit and concludes that the tightening labor market of the last five years has not pulled in new workers.
  • By contrast, I show that this finding is not robust when looking at flows that are: 1.) adjusted for internal errors and inconsistencies; 2.) measured year-on-year rather than month-on-month; and, 3.) calculated from broader or narrower measures of labor supply.
  • I conclude that labor force entry likely explains between ⅓ to ½ of the rise in US prime-age participation from its recent trough, and that the level and slope of labor force flows suggest prime-age participation may have even further to rise in the absence of exogenous shocks.

Introduction

The Great Recession hit the American labor force hard. By 2015 Q3, the share of Americans age 25–54 who were classified as in the labor force — those with or actively looking for a job — had dropped to 80.7%, the lowest prime-age participation rate since 1984.

On bathtubs and participation

Labor supply can expand through two different mechanisms: by workers coming back into the labor force (a rise in labor force entry), or by a stem in the flow of workers leaving the labor force (a fall in labor force exit).

A recent FRBSF analysis suggests all of the participation rise stemmed from lower labor force exit.

In a May 2019 Economics Letter, San Francisco Federal Reserve labor economist Regis Barnichon looks at the behavior of entry and exit rates in the CPS and concludes that virtually all of the rise in prime-age participation after 2015 was due to a fall in labor force exit. Based on these estimates, the hot economy had little to no impact on labor force entry rates.

Note that unadjusted month-to-month labor force exit is falling and may still have further to fall.

However, even on Barnichon’s own terms, a flattening of one mechanism (entry) doesn’t necessarily mean that the level of prime-age labor force participation has peaked. Exit rates, the other mechanism, are falling but remain elevated from their 2007 and 2000 lows. If one concluded that labor force exit had further to fall, then the implication is that prime-age participation by extension still has further to rise.

The distinction between unemployment and nonparticipation is often inconsistent within the CPS.

Economists should be cautious about drawing strong conclusions solely from unadjusted month-to-month CPS flows, because the distinction between participation and nonparticipation can be inconsistent for some CPS respondents over time.

Once we adjust for these issues, labor force entry becomes a more important factor in month-to-month transitions.

To test the robustness of the Barnichon (2019) approach, then, I make two adjustments to account for possible errors mentioned above.

This is confirmed when we look at year-to-year transitions.

But just as economists should not draw strong conclusions solely from unadjusted month-to-month CPS flows, so too should they be reticent to rely only on adjusted flows, no matter how internally-consistent or transparent those adjustments are. After all, the adjustments here literally involve changing responses to the CPS.

The entry story is also confirmed when looking at employment flows instead…

Economists focus on labor force participation often because it is a proxy for how we think of labor supply — the population of workers who have or are pursuing a job in any given month.

…as well as with a broader labor force measure…

Another cross-check, which bounds the other side of the analysis, uses a broader measure of labor supply than the official labor force. To try to keep the analysis here similarly straightforward to the one using employment, I define an “augmented prime-age labor force” equal to the official labor force plus prime-age Americans who are nonparticipants but say they want a job, as well as prime-age Americans who say they are out of the labor force due to school enrollment (based on the logic that Americans investing in more education probably intend to reenter the labor force at some point). This alternative definition is about 5 percentage points higher in level terms than the official prime-age labor force participation rate in 2019, a gap that has grown over time.

..and also, finally, with a synthetic labor force measure.

As a final robustness test, I construct a synthetic labor supply metric. All workers with a job are included 1:1, but I weight those without a job, whether unemployed or nonparticipants, based on the real-time probability that they will find jobs the next month. I use a rolling centered three-year probit model. The model predicts the probability of one-month job finding based on sex, age, parenthood, marriage status, race, nativity, education, size of metro area, and reason for nonemployment (short- or long-term unemployment, discouragement, retirement, enrollment, disability/illness, home/family care, or other). To calculate flows, I randomly assign the nonemployed CPS records to synthetic labor force status based on these binomial probabilities.

Conclusions: Putting it All Together

Analyzing the contribution of exit and entry on labor supply changes depends crucially on three dimensions — the frequency of the flows, the adjustments made for consistency and survey error, and the very definition of “labor supply” itself.

End Notes

[1] My data here are 12-month moving averages whereas Barnichon (2019) shows calendar year averages; the two series nevertheless appear to be consistent with one another. I weight my flow metrics using a constructed weight that constrains the sum of longitudinally-linked monthly records to be equal to the total population along sex, 5-year age group, and labor force status interactions. Later on in this analysis when I look at year-on-year flows, I use a similarly-constructed weight but limited to records longitudinally-linked across 12 months.

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Employ America

Employ America

We write, crunch #’s, and tweet about the labor market and economic policy.