Jim Tankersley at the Washington Post and Ben Casselman at the Wall Street Journal are in a “wonkfeud” about labor force participation. It started when Casselman estimated that there are currently around 3 million “missing workers” in the US (workers who are not in the labor force but who would be if job opportunities were strong). Tankersley did not dispute that figure, but pointed out that because it only counts workers who are missing due to the weak labor market in the Great Recession and its aftermath, it is a substantial undercount of the total number of missing workers because labor force participation was also weak in the decade leading up to the Great Recession.

To Tankersley’s point I will just add that while I don’t have an estimate of the number of missing workers from the 2000-2007 business cycle, it was indeed the weakest full business cycle in terms of job growth in at least three generations, and the labor market had not yet come close to regaining the health of the late 1990s before the Great Recession began at the end of 2007. It is an important reminder that estimates like the one below of how many jobs we need to get back to the labor market health of 2007 are conservative indeed, since while 2007 was the last year before the Great Recession hit, it was no labor market paradise by any stretch.

The main point I want to make, though, is that even if you only count the number of missing workers due to the weak labor market of the Great Recession and its aftermath, it’s very likely that Casselman’s 3 million figure is a substantial underestimate. A recent paper (pdf) on labor force participation in the aftermath of the Great Recession by Christopher Erceg and Andrew Levin of the Boston Federal Reserve Board unearths an excellent resource for getting to the bottom of how much of the decline in labor force participation since the start of the Great Recession is due to cyclical factors (i.e. the result of a lack of job opportunities) and how much is due to structural factors (e.g. things like baby boomers hitting retirement age). I should note that this is a critical distinction. We estimate that the U.S. economy needs 8.8 million jobs to restore the degree of labor market health that prevailed in December 2007. Embedded in this number is an estimate that much of the decline in labor force participation—from 66.0% in December 2007 to 63.3% last month—has been due to the weak economy, and not to more benign, non-cyclical factors. If that’s not right, then the number of jobs needed to restore the labor market to full health in coming years is very different. However, the work by Erceg and Levin finds that indeed cyclical factors account for the bulk of the post-2007 decline in the labor force participation rate.

Erceg and Levin found a BLS paper (pdf) published in November 2007—before the start of the Great Recession—that rigorously projects labor force participation rates to 2016. The projection assumed a healthy labor market over that period, so all of the labor force changes it forecasts reflect purely non-cyclical factors. The difference between these projections and the actual labor force participation rate is thus a very good measure of the cyclical change in the labor force participation rate, i.e. the change in the labor force participation rate that is a direct result of the weak labor market in the Great Recession and its aftermath.

The table below shows the labor force participation rate by gender and age category for the last 12 months (ending in March 2013) based on the projections in the BLS paper, along with the actual labor force participation rate for this period. As mentioned, the difference between the projected labor force participation rate and the actual labor force participation rate measures the cyclical decline in labor force participation caused by the Great Recession. The next column uses that measure of cyclical decline in the LFPR and actual population counts to calculate the number of “missing workers,” i.e. workers who would be in the labor force if job opportunities were strong. Using this approach I find that there are roughly 4.4 million missing workers, which means that around 75 percent of the decline in the labor force participation rate since the start of the Great Recession is due to cyclical factors. (This is slightly larger than the two-thirds estimate I calculated here. It is also somewhat larger than what is implied by the Congressional Budget Office’s estimates of the potential labor force, which suggests that around 60 percent of the decline is due to cyclical factors.) The last column of the table, along with the figure, shows the distribution of missing workers.

Who are the missing workers? More than half of all missing workers—53.7 percent—are “prime age” workers, age 25-54. A third are prime age men and a fifth are prime age women. Another roughly 30 percent are under age 25; 18 percent of missing workers are men under age 25 and 13 percent are women under age 25. (It is worth noting that these missing young workers are not “sheltering in school” from the labor market effects of the Great Recession. There has been no recession-induced increase in college or university enrollment for either men or women; see Figures E and AA in this paper, which show that enrollment rates have not meaningfully departed from their long-run trend in the Great Recession and its aftermath.) The remaining roughly 15 percent of missing workers are age 55 and over, mostly women; 13.1 percent of missing workers are women age 55 and over, while 2.6 percent of missing workers are men age 55 and over.

The only gender/age group that has not seen a cyclical decline in labor force participation is men nearing retirement age, those age 55-64. This is likely due to workers opting not to retire early because of the loss of retirement security caused by the effects of the Great Recession. There are many missing women in this age group, however; the labor force participation rate of women age 55-64 was expected to increase over this period as women from cohorts with stronger labor force participation throughout their 20’s, 30’s and 40’s than the cohorts that preceded them aged into this age bracket. The weak job market has meant that that increase was far smaller than expected.