Labor Force Exiters around Recessions: Who Are They?

Abstract

This article identifies workers who experienced a job separation during the Great Recession or the pandemic recession and tracks their labor force status in the following year, using the Current Population Survey. Workers are classified as exiters if they leave the labor force shortly after their job loss and non-exiters if they do not. The pool of exiters is disproportionately female, less educated, and older. During the pandemic recession, there were even more older workers in the exiters pool, although they were less likely to report being retired compared with in the Great Recession. In addition, statuses were more persistent during the Great Recession: For both exiters and non-exiters, the majority had the same labor force status a year later. I then use the patterns of these samples of job separators to estimate the propensity of being reemployed in a year and apply the estimates to the general out-of-work pools during the two recessions. I find that changes in the likelihood of being reemployed as well as the composition of individuals out of work are important for understanding the differences between the labor market in the two recessions.

Introduction

Job separations are a prominent feature of the labor market in and around recessions. Upon being separated, there are many different paths that workers can take afterward. Some quickly go back to their previous jobs or find new ones, some spend considerable time searching for work, and others leave the labor force entirely. Although some of these workers' exits are permanent, many eventually reenter the labor force. Nevertheless, both the COVID-19 pandemic recession and the Great Recession spurred declines in labor force participation. Whether and how long it takes for these individuals to return are of great interest to policymakers.

Figure 1 depicts the monthly number of labor force exits (Panel A) and entries (Panel B) following the start of both the Great Recession and the pandemic recession. The two recessions show different patterns in both types of flows. For example, the pandemic recession had large spikes of both exit and entry in certain months, tied closely with broad changes in pandemic-related shutdowns. There were also differences in churn: In terms of both entry and exit, more people were moving across the nonparticipation margin in every month during the pandemic recession compared to the Great Recession. 

Figure 1: Labor Force Flows: Pandemic vs. Great Recession

NOTE: Panel A refers to the monthly number of exits from the labor force, and Panel B refers to the monthly number of entries into the labor force. The starting points are December 2007 for the Great Recession and February 2020 for the pandemic recession.

SOURCE: Current Population Survey and author's calculations.

This article aims to understand who underlies the patterns in Figure 1 and the paths they took to get there. Specifically, I examine the labor force status patterns of people who lose their jobs in recessions, with an emphasis on comparing the Great Recession with the recent pandemic-related recession. I then use these patterns to estimate the likelihood of being reemployed again within a year as a function of labor force status, individual characteristics, and reasons for being out of work and apply the estimates to a broader sample. I begin by pinpointing a sample of workers in the Current Population Survey (CPS) who can be observed losing their jobs during either one of the recessions. I record their labor force status each month starting from employment and classify them as either a labor force "exiter" or "non-exiter" based on their employment status at the end of their four-month panel. 

I then compare the observable characteristics of exiters, non-exiters, and those who became employed again by the end of the four months. I find that in both recessions, exiters were more likely to be female, older, less educated, and similar to non-exiters in terms of racial composition. In the Great Recession, non-exiters skewed male, in comparison to a 50/50 split in the pandemic recession. The age distribution of the exiter pools is also different: It skewed older in the pandemic recession, but less of the exiters in the 55-64 age group reported being retired compared to those in the Great Recession. 

The structure of the CPS allows me to follow up on these workers one year after their initial job loss. I look at their status in their last month of the survey and compare it with where they were four months after the job loss. I find that the statuses were more persistent during the Great Recession: Most of the exiters stayed out of the labor force, and most of the non-exiters were unemployed. The pandemic recession looked much different, with 30 percent of the exiters and 60 percent of the non-exiters having gone back to work. 

Last, I estimate logit models for returning to employment in the last month of the survey for exiters and non-exiters, both in the Great Recession and in the pandemic recession. This group of workers is interesting because it is representative of the matches that were in tact before the recession hit, and thus it is a group that policymakers may be interested in getting back to work. Using the CPS for this exercise is also useful because it provides the most up-to-date micro data on employment with a panel dimension. 

The results of this estimation reflect differences in the behavior of the workers who lost their jobs during the two recessions: Exiters who become reemployed tend to be younger, not have young children at home, and be female. Non-exiters who become reemployed tend to be more educated and have left their old jobs voluntarily. I apply these estimates from the exiters and non-exiters sample to the broader out-of-work pools over the two recessions, which combines the measure of behavior with the composition of people who recently separated from their jobs. Despite the fact that exiters returned more quickly after the pandemic recession, the composition of the out-of-labor-force pool in the aftermath suggests that the rest may return more slowly. 

This article contributes to the large literature on labor market flows, specifically in and out of nonparticipation. For example, Elsby, Hobijn, and Åžahin (2015) show how these flows account for a substantial part of unemployment fluctuations. This article dives into who makes these flows and the roles they play as workers eventually find (or do not find) new jobs. In studying the paths workers take in the data, Hall and Kudlyak (2019) use the CPS in a similar way. The authors string together each worker's monthly observations of labor force status to understand what "types" of workers comprise the population in terms of their labor market flows.

This article also complements the existing research on the effects of COVID-19 on the labor market. A few related examples that have major empirical components and  use the CPS are Hall and Kudlyak (2021), Cortes and Forsythe (2020), and Forsythe et al. (2020). Chodorow-Reich and Coglianese (2021) also have a related method that projects the distribution of unemployment durations.