In this article, we examine the employment impacts of the coronavirus disease 2019 (COVID-19) pandemic on a particular group of vulnerable workers—those with a criminal history record (CHR). Using data from the 2021 COVID-19 Supplement of the U.S. Bureau of Labor Statistics (BLS) National Longitudinal Survey of Youth 1997 (NLSY97), we find that, overall, people with a CHR faced greater employment disruptions from the pandemic than those without a CHR. By itself, this result is not surprising. People with a CHR are more likely to have unstable job histories even before they get involved in the criminal justice system.1 Moreover, involvement with the criminal justice system exacerbates employment instability, either by limiting people to the secondary job market or by leading them to exit the labor market altogether.2
However, the disruptive effect of COVID-19 on employment for people with a CHR holds even when we focus our analysis on those who had achieved a measure of stable employment in the years preceding the pandemic, despite having a CHR. The negative employment effects from the pandemic on workers with a CHR arises mostly because these workers are concentrated in industries that were more adversely affected by the pandemic. More specifically, employment declined sharply in leisure and hospitality, and there were smaller but sizeable declines in education and health services, trade, construction, and manufacturing. Employment in sectors such as information and state and federal government was much less affected by the pandemic-induced shutdowns.3 The jobs in the industry sectors that experienced the largest declines tend to have a higher proportion of workers involved in face-to-face interactions and fewer hiring restrictions for people with a CHR. By contrast, the less affected sectors tend to have more restrictions for hiring people with a CHR.4 This finding points to a largely unexplored collateral consequence faced by workers who have a CHR: they are largely limited to industry sectors that are more susceptible to serious economic shocks like the COVID-19 pandemic, even if they have demonstrated a commitment to the labor market.
For this analysis, we examined data from waves 1 through 19 of the NLSY97 Cohort and the COVID-19 Supplement fielded to this cohort between February and May 2021.5 The NLSY97 is a longitudinal survey of 8,894 individuals born between 1980 and 1984. Starting in 1997 (wave 1, the year first surveyed), respondents were surveyed annually through 2011–12, after which they were surveyed every other year through 2019–20 (wave 19, the most recently available survey data). The gender, race, and ethnicity breakdown of the respondents in the NLSY97 is as follows: 51.2 percent male, 48.8 percent female, 51.9 percent White non-Hispanic, 26.0 percent Black non-Hispanic, and 21.2 percent Hispanic or Latino.6 To ensure sufficient variation, the NLSY97 includes an oversample of the Black and Hispanic populations, which we incorporate into our analysis and properly weight so that our results are nationally representative.
The COVID-19 Supplement was fielded to a majority of the NLSY97 Cohort in early 2021 and had an approximate response rate of 66 percent. The COVID-19 Supplement asked questions about respondents’ employment conditions in the week prior to the survey and in the previous 12 months, and how COVID-19 had affected their employment (e.g., number of hours worked, earnings, change in employer), their health conditions, and their children’s schooling.7 The COVID-19 Supplement was fielded to supplement both wave 19 and the ongoing wave 20. Thus, when we analyzed the data, we used person-level weights from wave 19 of the NLSY97.
In this article, our definition of a CHR is similar to that used in earlier work in which we analyzed data from the NLSY97.8 We focus on people who have a criminal record of convictions and/or guilty pleas. We construct a cumulative measure of criminal history defined as having been convicted of, or pled guilty to, at least one nontraffic-offense charge since turning 18. We consider respondents’ criminal history as of their most recent interview in the latest wave of the NLSY97 (i.e., wave 19). Approximately 17.9 percent of our sample have at least one conviction and/or guilty plea since turning 18, with 25.7 percent of men and 10.0 percent of women having a conviction and/or a guilty plea (the difference is statistically significant).
We start by defining specific conditions for inclusion in our analysis, beginning with a definition of civilian labor force participation. We define civilian labor force participation as having a nonzero number of weeks worked full time, weeks worked part time, or weeks unemployed between the first week of January 2017 and the last week of August 2019.9 We exclude individuals with any Armed Forces involvement in the wave 19 of the NLSY97, but we count service in the Armed Forces prior to 2019 as standard employment.10 This is to ensure that we are looking at people who are currently part of the civilian labor market. Finally, we limit our sample to respondents who were age 20 years or older at the time of their interview. From this definition of labor force participation, we determine stable employment with the following two criteria: (1) we classify people as stably employed if the number of weeks they worked (either in full- or part-time status) since 2017 is greater than or equal to 75 percent of the total number of weeks in our study period; and (2) we exclude from our definition of stably employed people who worked part-time hours in excess of one-third of the time since 2017 and reported having searched for work in the previous 3 months while currently employed.
Put simply, we consider a respondent in our sample as being stably employed if he or she worked for pay for at least 75 percent of the time from January 2017 through August 2019 and did not report having searched for work recently. For example, we would classify a respondent who worked for pay for 104 of the 138 weeks of the period as stably employed. But if a respondent worked part time for at least 46 weeks (out of 138) and had recently searched for new work, then we would not include that respondent in our stably employed population.11
Our measure of employment stability allows for people to experience short periods of unemployment (either because of cyclical economic conditions, because they changed employers with only small gaps in employment, or for other reasons) without being considered unstably employed. Furthermore, our second criterion enables us to distinguish between workers who are employed part time but would prefer full-time work and those who are content with their current part-time employment. For example, we manually reclassified 30 people as working “part time for economic reasons” (see last row of table 1), which is an employment status used in the Current Population Survey but not directly captured by the NLSY97.12 Our measure also differs slightly from measures of employment stability used in previous studies analyzing NLSY97 data.13 Table 1 presents the number of civilian workers in our sample who participated in the civilian labor force (4,939 respondents), the percentage breakdown between the stably and unstably employed, and the percentage of workers in each of those groups that had a CHR.14
|Characteristic||Number of observations||Percent of labor force||Percent with CHR|
Total in labor force
Criterion 1: at least 75-percent employment
Criterion 2: worked part time and searched for work
Criterion 1: less than 75-percent employment
Criterion 2: worked part time and searched for work
 Not applicable.
Note: NLSY97 = National Longitudinal Survey of Youth 1997; CHR = criminal history record. This table presents the sample counts for individuals responding to the NLSY97 and the NLSY97 COVID-19 Supplement, according to our definitions of labor force participation and stable employment, as well as the survey-weighted percentages of stably and unstably employed labor force participants with a CHR. A labor force participant is a person who worked at least 1 week (in either full- or part-time status) or was classified as unemployed in at least 1 week between January 2017 and August 2019. Stably employed workers worked at least 75 percent of weeks between January 2017 and August 2019 in either full- or part-time status. If they worked part time, the workers are not considered stably employed if they worked at least one-third of weeks in part-time status and reported having searched for work in the previous 3 months. To get the final sample sizes for the stably and unstably employed, add each of the criteria rows (e.g., criterion 1 plus criterion 2 = stably employed). The data presented in this table are authors’ calculations from the NLSY97 and the NLSY97 COVID-19 Supplement, which was fielded between February and May 2021.
Source: U.S. Bureau of Labor Statistics, NLSY97 and NLSY97 COVID-19 Supplement.
Our definition of employment stability provides insights into employment demographics around this issue. (See table 2.) Not surprisingly, for example, we find that the proportion of workers who are stably employed increases with higher levels of education. Moreover, when we examine large industry groups (“supersectors,” as defined in the BLS Quarterly Census of Employment and Wages15), we see that leisure and hospitality had the lowest rate of employment stability, a result driven by the precarious nature of businesses in the industry and seasonal variation in demand for its services.16
|Demographic or industry||Percent with stable employment||Percent of all stably employed workers||Percent of NLSY97 COVID-19 Supplement sample|
Less than a high school diploma
High school diploma, no college
At least some college
Trade, transportation, and utilities
Business and professional services
Education and health services
Leisure and hospitality
 Includes people who have passed the General Educational Development test (commonly known as the “GED” or high school equivalency test).
Note: NLSY97 = National Longitudinal Survey of Youth 1997. This table presents the survey-weighted percentages of NLSY97 COVID-19 Supplement respondents who were stably employed during the period from January 2017 to August 2019, by selected demographic characteristics and industry groups. It also shows the percentages of all stably employed workers and the percentages of all NLSY97 COVID-19 Supplement respondents represented by each group. Stably employed workers are people who worked at least 75 percent of weeks between January 2017 and August 2019, in either full- or part-time status. If the respondents worked part time, they are not classified as stably employed if they worked part time at least one-third of weeks and reported having searched for work in the previous 3 months. The data presented in this table are authors’ calculations from the NLSY97 and the NLSY97 COVID-19 Supplement, which was fielded between February and May 2021.
Source: U.S. Bureau of Labor Statistics, NLSY97 and NLSY97 COVID-19 Supplement.
The NLSY97 COVID-19 Supplement provides data on employment disruptions, both in the week prior to the survey and at any point in the previous 12 months. In this analysis, we examine both voluntary and involuntary (employer-related) employment disruptions, as well as reductions in hours worked and declines in earnings.
We begin our analysis by examining pandemic-related employment disruptions among people with and without a CHR. We consider two definitions of employment disruption. The first definition captures employer-related employment disruptions—specifically, it includes people who worked for profit during the 12 months prior to the survey and either voluntarily or involuntarily stopped working for an employer. The second definition, which is broader in scope, includes people who experienced an employer-related employment disruption and those who experienced a reduction in the number of hours they worked or a decline in their earnings (or both). The broader definition allows us to capture employment disruptions among the self-employed as well as among those who did not stop working during the reference period. Table 3 presents the percentages of all employed and stably employed civilian workers who experienced employment disruptions according to both definitions, by CHR status. The results are consistent across both definitions of employment disruption: among all workers, those with a CHR were 12 to 14 percentage points more likely to report having experienced an employment disruption than were their non-CHR counterparts (the difference is statistically significant). Conditioning on stable employment reduces the percentage of people who experienced a pandemic-related employment disruption in the previous 12 months, with the difference in employment disruption between those with and without a CHR ranging from nearly 9 percentage points for the first definition to nearly 11 percentage points for the second definition. (The differences for both definitions are statistically significant at the 95-percent confidence level.)
|Employment status||All workers||No CHR||Has CHR|
|Percent with disruption||CI lower bound||CI upper bound||Percent with disruption||CI lower bound||CI upper bound||Percent with disruption||CI lower bound||CI upper bound|
Employment disruption, definition 1
Employment disruption, definition 2
Note: CHR = criminal history record; CI = confidence interval. This table presents the survey-weighted percentages of all workers, workers without a CHR, and workers with a CHR that experienced employment disruptions during the early stages of the coronavirus disease 2019 (COVID-19) pandemic, according to two definitions of employment disruption. The first definition includes people who worked for profit during the 12 months prior to the survey and either voluntarily or involuntarily stopped working for an employer. The second definition is broader in scope and includes people who experienced an employer-related employment disruption and people who experienced a reduction in the number of hours they worked or a decline in their earnings (or both). The data are derived from questions that were part of the National Longitudinal Survey of Youth 1997 (NLSY97) COVID-19 Supplement, which was conducted from February to May 2021; the 12-month reference period ranges from February–May 2020 to February–May 2021. A labor force participant is a person who worked at least 1 week (in either full- or part-time status) or was classified as unemployed (actively searching for work) in at least 1 week between January 2017 and August 2019. Stably employed workers worked at least 75 percent of weeks between January 2017 and August 2019, in either full- or part-time status. If they worked part time, workers were not considered stably employed if they worked at least one-third of weeks in part-time status and reported having searched for work in the previous 3 months. The data presented in this table are authors’ calculations from the NLSY97 and the NLSY97 COVID-19 Supplement.
Source: U.S. Bureau of Labor Statistics, NLSY97 and NLSY97 COVID-19 Supplement.
People with a CHR are also more likely than their non-CHR counterparts to experience multiple types of employment disruption. Specifically, 18 percent of people with a CHR experienced all three types of employment disruption, compared with 11 percent of their non-CHR counterparts (the difference is statistically significant).
The differences in employment disruption between those with and without a CHR is largely explained by the industries in which people with a CHR are most likely to work. Chart 1a and chart 1b display employment concentrations of workers with and without a criminal record across industry groups. We find that approximately 83 percent of employed (and 80 percent of stably employed) people with a criminal record are employed in 6 of the 11 industry supersectors (percentage of stably employed shown in parentheses):
(1) Trade, transportation, and utilities: 18.4 percent (16.0 percent)
(2) Construction: 13.7 percent (14.8 percent)
(3) Leisure and hospitality: 13.1 percent (11.1 percent)
(4) Education and health services: 12.6 percent (13.0 percent)
(5) Professional and business services: 12.5 percent (12.4 percent)
(6) Manufacturing: 12.5 percent (11.8 percent)
In other words, workers with a CHR are not proportionally employed across all industries. Moreover, several industries overrepresent the CHR rate in the general population. In our data, 17.9 percent of the sample respondents have a prior conviction, while 37.0 percent of the workers in construction, 27.9 percent of the workers in leisure and hospitality, 23.4 percent of the workers in manufacturing, and 23.3 percent of the workers in natural resources and mining reported having a prior conviction. (See chart 2.) Chart 1a shows that workers with a CHR are overrepresented in manufacturing, leisure and hospitality, and construction. Chart 1b shows the same results for people who are stably employed, and we see that workers with a CHR are overrepresented in the same three industries. We also see that stably employed workers with a CHR are underrepresented in education and health services and in public administration. This result may be due to workers with a CHR having lower levels of educational attainment (on average) than their non-CHR counterparts, and workers in these industries tend to have higher levels of education (on average) than workers in other industry groups.
Using our definition of employment stability, we find that there are important differences across industries, with the least stable employment during the pandemic period occurring in leisure and hospitality, which also had the smallest percentage of stably employed workers during the prepandemic period, at 76.5 percent. By comparison, public administration, at 92.7 percent, had the largest percentage of stably employed workers during the prepandemic period. (See chart 3a.) The industries with the smallest proportions of stably employed workers have nontrivial shares of their workforce reporting that they have a CHR: 20.1 percent in trade, transportation, and utilities; and 27.9 percent in leisure and hospitality. (See chart 2.) Combined, these two industries make up more than 30 percent of the total number of workers who have a CHR.
Given the lower rate of employment stability in leisure and hospitality before the pandemic, we would expect to find that workers in this industry (and other industries with relatively low rates of employment stability) were more likely to have experienced employment disruptions during the pandemic, and this is indeed what the data in our study show: more than half (53.4 percent) of workers in leisure and hospitality reported that they had experienced employer-related employment disruptions during the early stages of the COVID-19 pandemic. That proportion is more than 20 percentage points higher than the proportion in construction (30.7 percent), which had the second-highest rate among the industry groups. (See chart 3b.)
Does the industry in which people work explain all the differences in employer-related employment disruptions during the COVID-19 pandemic between those with and without a CHR among the stably employed? Chart 4 shows the differences between these two groups for the stably employed. Although people with a CHR experienced more disruptions than their non-CHR counterparts in most industries, the differences are not statistically significant.17 Interestingly, employment disruptions are experienced at similar rates between those with and without a CHR in leisure and hospitality, an industry with one of the highest proportions of workers with a CHR (27.9 percent).
Having a CHR is also correlated with other characteristics associated with employment instability, including race, gender, and education level. For example, among the race and ethnic groups, Black workers were the least likely to experience employment stability, with 77 percent in stable employment, compared with 82 percent of Hispanic workers and 83 percent of White workers. In terms of gender, 78 percent of women were in stable employment, compared with 86 percent of men. Finally, and perhaps most noteworthy among the demographic characteristics, education level had substantial effects on employment stability: 65 percent of workers who did not graduate from high school or hold a GED were stably employed, compared with 81 percent of high school graduates and 88 percent of those with some college or more.
Our analysis has some limitations. First, the NLSY97 provides data only for people born between 1980 and 1984 (the NLSY97 Cohort). Thus, our findings may not be representative of other age groups in the population. That said, using a narrow age range ensures that all survey respondents are at a similar stage of their lives and avoids conflating life-course differences with employment differences. Second, our industry comparisons have limited validity in some of the smaller industry groups. In these industries, there are both large differences between those with and without a CHR and large confidence intervals.
Despite being at a disadvantage in the labor market, some people with a CHR managed to maintain stable employment in the years prior to the pandemic (defined as having worked more than 75 percent of the time during the period from January 2017 to August 2019). In this article, we find that even after we condition on prior stable employment, the COVID-19 pandemic had disproportionately disruptive effects on those with a CHR. This is primarily due to individuals with prior convictions being disproportionately employed in the four sectors of the economy that were most adversely affected by the pandemic: leisure and hospitality; construction; trade, transportation, and utilities; and manufacturing. Future studies should seek to explain the differences that exist even after controlling for industry. Although these differences are not statistically significant in our analysis, research conducted with larger samples might provide new insights into this issue.
Daniel Schwam, Shawn Bushway, and Jeffrey B. Wenger, "The impact of the COVID-19 pandemic on workers with a criminal history," Monthly Labor Review, U.S. Bureau of Labor Statistics, January 2023, https://doi.org/10.21916/mlr.2023.1
1 For more on this issue, see Joan Petersilia, When Prisoners Come Home: Parole and Prisoner Reentry (New York: Oxford University Press, 2003).
2 See Shawn D. Bushway and Peter Reuter, “Labor markets and crime,” in James Q. Wilson and Joan Petersilia, eds., Crime: Public Policies for Crime Control (Oakland, CA: ICS Press, 2002), pp. 183–209.
3 Industry employment data are from the U.S. Bureau of Labor Statistics (BLS) Current Employment Statistics survey (also known as the establishment survey) and were retrieved from Federal Reserve Economic Data (FRED) database (Federal Reserve Bank of St. Louis, January 2022), https://fred.stlouisfed.org/categories/11. For more information on the disproportionately negative effects of the coronavirus disease 2019 (COVID-19) pandemic on certain industries and occupations, see, for example, James V. Marrone, Susan A. Resetar, and Daniel Schwam, “The pandemic is a disaster for artists,” The RAND Blog: Commentary, August 4, 2020, https://www.rand.org/blog/2020/07/the-pandemic-is-a-disaster-for-artists.htmlKathryn A. Edwards, “For leisure and hospitality, weak recovery still looks like recession,” The RAND Blog: Commentary, September 4, 2020, https://www.rand.org/blog/2020/09/for-leisure-and-hospitality-weak-recovery-still-looks-like-recession.html and Gene Falk, Paul D. Romero, Isaac A. Nicchitta, and Emma C. Nyhof, Unemployment Rates During the COVID-19 Pandemic, CRS Report R46554 (Congressional Research Service, updated August 20, 2021), https://sgp.fas.org/crs/misc/R46554.pdf.
4 For more information on the kinds of jobs that are more likely to have been negatively affected by the pandemic, see Abel Brodeur, David Gray, Anik Islam, and Suraiya Bhuiyan, “A literature review of the economics of COVID-19,” Journal of Economic Surveys, vol. 35, no. 4 (September 2021), pp. 1007–44, https://doi.org/10.1111/joes.12423. See also After the Sentence, More Consequences: A National Snapshot of Barriers to Work (New York: Council of State Governments Justice Center, January 2021), https://csgjusticecenter.org/wp-content/uploads/2021/02/collateral-consequences-national-snapshot.pdf.
5 The BLS National Longitudinal Survey of Youth 1997 (NLSY97) data used in this article were compiled and distributed by the Center for Human Resource Research at the Ohio State University, https://chrr.osu.edu/national-longitudinal-surveys. For more information on the NLSY97, see “National Longitudinal Surveys: NLSY97 data overview” (U.S. Bureau of Labor Statistics, last modified November 18, 2022), https://www.bls.gov/nls/nlsy97.htm.
7 See “Appendix 14: NLSY97 COVID-19 Supplement” (U.S. Bureau of Labor Statistics), https://www.nlsinfo.org/content/cohorts/nlsy97/other-documentation/codebook-supplement/appendix-14-nlsy97-covid-19.
8 See Shawn Bushway, Irineo Cabreros, Jessica Welburn Paige, Daniel Schwam, and Jeffrey B. Wenger, “Barred from employment: more than half of unemployed men in their 30s had a criminal history of arrest,” Science Advances, vol. 8, no. 7 (February 2022), https://www.science.org/doi/full/10.1126/sciadv.abj6992.
9 We considered data only through August 2019 to avoid any potential bias in the data from the early stages of the COVID-19 pandemic.
10 People in the Armed Forces include individuals serving in the regular forces, the reserves, or the National Guard. In addition, we include people who said they work for a military employer because they are typically people serving in the Armed Forces who incorrectly responded to the civilian-employment survey questions. For more information, see the definition for “military service” in “National Longitudinal Surveys: employment” (U.S. Bureau of Labor Statistics), https://www.nlsinfo.org/content/cohorts/nlsy97/topical-guide/employment/employers-jobs.
11 In the NLSY97, there is no way to determine if a respondent has looked for additional part-time work, replacement part-time work, or full-time work. We assume that any job-search activity is intended to achieve a better position than the current part-time job.
12 Including these 30 individuals among the stably employed (rather than among the unstably employed) probably had little effect on our analysis because this group represents less than 1 percent of the stably employed population.
13 See, for example, the definition used in Robert Apel and Gary Sweeten, “The impact of incarceration on employment during the transition to adulthood,” Social Problems, vol. 57, no. 3 (August 2010), pp. 448–79, https://doi.org/10.1525/sp.2010.57.3.448.
14 We also define employment stability as being employed for 52 weeks prior to the respondent’s last interview date. The last interviews occurred between September 2019 and July 2020, so the 52-week period prior to the last interview ranges from September 2018–July 2019 to September 2019–July 2020. This alternative definition of employment stability reduces the share of stable employment across industries and demographics, but the effects on employment disruption related to criminal history and COVID-19 are qualitatively the same. The full results are available from the authors upon request.
15 Our supersectors are slightly different from those used in the BLS Quarterly Census of Employment and Wages (QCEW). Specifically, we combine the QCEW supersectors “unclassified” and “other services, except public administration,” into a single category, “all other (including unknown),” for a total of 11 supersectors. For more information on the QCEW supersectors, see “Quarterly Census of Employment and Wages: QCEW high-level industry titles” (U.S. Bureau of Labor Statistics, last modified August 19, 2022), https://www.bls.gov/cew/classifications/industry/industry-supersectors.htm.
16 For BLS employment data, see Current Employment Statistics Highlights: December 2020 (U.S. Bureau of Labor Statistics, January 8, 2021), https://www.bls.gov/ces/publications/highlights/2020/current-employment-statistics-highlights-12-2020.pdfSee also Edwards, “For leisure and hospitality, weak recovery still looks like recession”; and Falk et al., “Unemployment rates during the COVID-19 pandemic.”
17 The differences are not statistically significant at the 95-percent confidence level. The full results are available from the authors upon request.