by Alessia Leibert
September 2023
What is the short-term influence of education in the reemployment market? Did formal education shelter people from long periods of unemployment after being laid off in the first months of the COVID-19 recession? And did it protect workers from the risk of new job losses after the recession? What other worker characteristics besides education facilitated a more successful reintegration into the workforce? This information can help identify the characteristics of workers with the highest need for targeted reemployment services.
The study population consists of 245,091 Twin Cities Metro area residents who filed for Unemployment Insurance (UI) benefits in the first months of the COVID-19 pandemic. Since we are interested in people moving from unemployment to employment, we only examine individuals of prime working age (20 to 54) who were employed in Minnesota in 2021.
An unemployed person typically seeks to avoid three kinds of risk: taking a long time to return to work; suffering future permanent job losses; and experiencing earning losses relative to pre-layoff earnings. These three goals are measured in this study as follows:
This study aims to identify the factors that helped workers get back to work quickly and permanently. A separate article focuses on wage outcomes.
The Pandemic Recession was the shortest in U.S. history, lasting only two months. Most Minnesotans were able to rejoin the workforce a few weeks after being laid off or furloughed. The average claimant in our dataset filed an average of 13.7 weeks of continuing claims from March to September 2020, and only a small minority, 10.8%, reported being on permanent layoff from Sept 2021 to June 20222 (see Table 1). The first outcome is linked to the second because workers who spent a longer time out of work initially (first outcome) were also at higher risk of falling back to the most severe form of unemployment (second outcome) a year and a half after the initial layoff.
Table 1 - Risk of experiencing unfavorable outcomes, by demographic characteristics
Workers' characteristics |
Incidence of unfavorable outcomes |
||
---|---|---|---|
1st OUTCOME: Avg weeks collecting UI benefits from March to Sept 2020* | 2nd OUTCOME: Pct who permanently lost job from Sept 2021 to June 2022 | ||
Total | 13.7 | 10.8% | |
Gender | |||
Male | 13.2 | 10.2% | |
Female | 14.2 | 11.3% | |
Race | |||
White | 12.9 | 8.0% | |
Black | 17.1 | 22.4% | |
Hispanic/Latino | 14.3 | 11.9% | |
Asian | 12.9 | 8.4% | |
American Indian | 15.4 | 14.8% | |
More than one race | 16.0 | 16.6% | |
Disability status | |||
Without disability | 13.6 | 10.5% | |
With a disability | 16.2 | 16.2% | |
Did not respond | 15.1 | 14.3% | |
Place of residence | |||
Metro, excluding City of Minneapolis and St. Paul | 13.0 | 9.4% | |
City of St. Paul | 14.8 | 13.7% | |
City of Minneapolis | 15.8 | 13.6% | |
Education (highest attained) | |||
High school or less | 14.9 | 14.7% | |
1 or 2 years of college/vocational school | 14.1 | 12.0% | |
Associate degree or three years of college | 13.5 | 9.8% | |
Bachelor's degree | 12.8 | 6.6% | |
Master's degree or above | 10.6 | 4.7% | |
*Defined as the number of continuing claims filed from March 15, 2020 to Oct 1, 2020 |
Table 1 identifies the following risk factors:
Besides being determined by demographic characteristics, some vulnerabilities stem from the types of jobs held. Table 2 lays out the most significant ones.
Table 2- Risk of experiencing unfavorable outcomes, by Job characteristics pre-layoff
Job Characteristics Pre-layoff |
Incidence of unfavorable outcomes |
|
---|---|---|
1st OUTCOME: Avg weeks collecting UI benefits from March to Sept 2020 | 2nd OUTCOME: Pct who permanently lost job from Sept 2021 to June 2022 | |
Wages pre-layoff | ||
Less than $15.00 an hour | 16.1 | 13.3% |
From $15 to $21.49 an hour | 14.6 | 12.7% |
From $21.49 to $33.00 an hour | 13.2 | 9.2% |
Above $33.00 an hour | 10.7 | 5.3% |
Full-time status pre-layoff | ||
Part-time or seasonal | 14.7 | 12.7% |
Part-time year-round | 16.3 | 13.1% |
Fulltime year-round | 11.8 | 8.7% |
Tenure pre-layoff | ||
Laid off from a job with <= 1 year of tenure | 15.0 | 14.4% |
Laid off from a job with > 1 year of tenure | 13.0 | 8.7% |
Industry sector (selected) | ||
Construction | 10.5 | 6.5% |
Manufacturing | 10.2 | 7.7% |
Retail Trade | 13.3 | 11.8% |
Temp Help | 16.8 | 18.4% |
Administrative Svc Except Temp hHelp | 14.3 | 14.7% |
Clinics, Hospitals & Home Health Care Services | 9.5 | 6.8% |
Nursing & Residential Care Facilities | 14.8 | 20.0% |
Social Assistance | 15.0 | 16.6% |
Accommodation & Food Services | 18.8 | 14.0% |
Occupational category (selected) | ||
Management | 12.7 | 8.4% |
Business & Financial Operations | 12.1 | 7.9% |
Computer & Mathematical | 11.5 | 7.7% |
Architecture & Engineering | 9.6 | 3.4% |
Community & Social Service | 12.9 | 10.0% |
Education Training & Library | 14.0 | 10.4% |
Healthcare Practitioners & Technical | 8.6 | 4.0% |
Healthcare Support | 12.3 | 14.0% |
Food Preparation & Serving Related | 18.8 | 14.3% |
Building & Grounds Cleaning & Maintenance | 14.5 | 13.1% |
Personal Care & Services | 12.3 | 12.4% |
Sales & Related | 15.7 | 11.0% |
Office & Administrative Support | 18.8 | 11.9% |
Construction & Extraction | 10.3 | 6.2% |
Installation Maintenance & Repair | 10.5 | 9.3% |
Production | 11.4 | 11.7% |
Transportation & Material Moving | 15.4 | 11.3% |
* Measured from Sept 2021 to June 2022, for a total of 10 months. |
Many of the characteristics displayed in Tables 1 and 2 are correlated with one another and can compound a worker's risk of experiencing negative outcomes. Part of the reason claimants' race is such an important driver of outcomes is that race is strongly associated with other risk factors of a socio-economic nature, such as wages pre-pandemic, educational attainment, and being a resident of St. Paul or Minneapolis, where claimants suffered from the most disruptions due to pandemic shutdowns4. Table 3 breaks down these factors by race to show how racial disparities in outcomes are often the result of multiple sources of disadvantage.
Table 3 – Socio-economic factors and their correlation with race
Race | With no education beyond high school | Resident in City of Minneapolis or St Paul | Pre-layoff wages <$21.49 (1) | Returned to pre-layoff employer upon reemployment |
---|---|---|---|---|
American Indian | 43% | 41% | 63% | 53% |
Asian | 45% | 35% | 62% | 62% |
Black | 53% | 48% | 80% | 42% |
Latino | 54% | 40% | 66% | 54% |
More than one race | 42% | 43% | 67% | 47% |
White | 23% | 26% | 34% | 61% |
TOTAL with valid reported race | 33% | 32% | 54% | 57% |
|
The key takeaways from this evidence are:
Table 4 displays factors related to job quality and job segregation by industry and occupation. Although our data cannot accurately measure a multi-dimensional concept such as job quality, we can capture two of its most important dimensions in addition to wages (already reported in table 3): part-time status, which reflects lower access to healthcare/retirement benefits and opportunities for career advancement; and tenure, which reflects job stability.
Table 4 – Job quality and job segregation factors and their correlation with race
Race/ethnicity | Working part-time pre-layoff | Employed in very short-tenured jobs (1) | Employed in the Temp Help industry | Employed in Healthcare Support occupations (2) |
---|---|---|---|---|
American Indian | 54% | 41% | 3% | 6% |
Asian | 39% | 31% | 4% | 7% |
Black | 60% | 51% | 8% | 14% |
Latino | 51% | 42% | 4% | 5% |
More than one race | 61% | 48% | 4% | 6% |
White | 46% | 32% | 2% | 5% |
TOTAL with valid reported race | 48% | 36% | 3% | 7% |
|
The results show the following:
This evidence suggests that the racial disparities in outcomes displayed in Table 1 are not caused by race per se but by a combination of socio-economic differences, job quality differences, and job segregation by race that pushes workers of color towards low-quality jobs. Workers who enter a low-quality job track (low wage, short tenure, high turnover, high shares of part-time work, in industries/occupations with few career ladders) face a higher risk of work terminations.
Racial inequalities in the returns to education and the role of job quality and job segregation by industry in driving such differences is well documented in this dashboard and corresponding report, but more research is needed to examine the effects of these inequalities on the risk of unemployment. The present study hypothesizes that the underlying factors making workers of color more vulnerable to unemployment than their white counterparts are unequal access to skills training and faster skills depreciation over time due to job segregation and/or racial bias in hiring, access to training, and promotions. A companion report lends some support to these hypotheses by documenting large disparities in lifelong earnings by race among the same population of workers analyzed in this article, suggesting that workers of color are more subject to skills depreciation over the course of their working lives. Greater risk of skill depreciation over time may contribute to higher reliance on the UI system among workers of color as a bridge between bouts of precarious, low-paid employment.
The interconnections between some of the factors listed in Tables 1 and 2 makes it hard to isolate the effect of education, job characteristics, and race on the two outcomes of interest. In order to disentangle the effects of all these factors we conducted a linear regression analysis using our first outcome, length of claim measured in weeks8, as the dependent variable. Independent variables are those displayed in Table 1 and 2, plus some additional ones9.
The factors most strongly associated with an increase in the duration of unemployment are:
The factors most strongly associated with a decrease in the duration of unemployment were:
The model, reported in Appendix 1, explains 27% of the variation10 in number of weeks claimed.
To further examine the effect of education, we ran a second model11 that interacted education level with race, allowing education level to have a different effect on the duration of unemployment depending on race. This turned out to be significant, demonstrating that whites are more sheltered from prolonged unemployment than other racial groups even within the same education level.
What about our second outcome variable, the risk of permanent lay off? We ran a logistic regression predicting the probability of a permanent layoff from September 2021 to June 2022 (see Appendix 2). The factors most strongly associated12 with an increase in the probability of a permanent layoff were the following:
The factors most strongly associated with a decrease in the probability of a permanent layoff were related to job quality (see Table 4), with the only exception of education:
This evidence shows that workers with less stable and lower quality employment histories were more likely to fall back into unemployment after being laid off during the COVID-19 recession, either by choice or because they could not secure stable employment. Also important, had we not controlled for job quality characteristics the effect of race/ethnicity would have been even more pronounced. If we remove these variables from the model, Black claimants are shown to have 88% greater probability of a permanent layoff than white claimants, instead of the 75% revealed in the full model. Therefore, job quality characteristics help explain, at least in part, why Black claimants faced higher risks of permanent job losses.
It is also worth noting that, in both models, the occupations associated with unfavorable outcomes are predominantly low-skilled. Having a work history in high-skilled, licensed occupations such as Engineers and Registered Nurses & Other Healthcare Practitioners offered strong protection against the risk of permanent layoff, while a work history in low-skilled occupations such as Food Processing, Other Personal Care & Service (including childcare), and Home Health & Personal Care Aides appears to raise the risk of a permanent layoff in the post-pandemic recovery period.
Despite some data limitations13, both models confirm that job quality, race, education level, and job sorting by occupation/industry are factors driving length of unemployment and risk of permanent job losses.
This study uses a unique dataset that encompasses UI claims data merged with Minnesota payroll records, allowing us to examine how length and severity of unemployment vary by race, education, and other characteristics. This analysis concludes that the key underlying factors that shape a worker's ability to withstand disruptive events such as a recession are education level, race, and a set of job characteristics such as choice of occupation/industry and job quality in general.
Here is a summary of findings:
In conclusion, this study highlights the importance of job quality in protecting workers from the adverse effects of job displacement. We find that the quality of jobs – including wages, job tenure, unionization, and fulltime status – matters more than simply having jobs. Hopping from one short-tenured, low-pay, dead-end job to another offers little protection from future displacements.
This body of evidence also suggests that workers who experienced unfavorable outcomes might simply lack the skills or the networks needed to secure sustainable employment and may be trapped in jobs where it's harder to develop such skills or networks. The fact that many claimants of color, who were the most likely to experience unfavorable outcomes, were overrepresented in frontline jobs such as Home Health Aides and Nursing Assistants should raise concern. Their struggles have repercussions not only on employers but also on Minnesotans who rely on these services.
Workforce boards and other workforce development program providers can use this information to deliver targeted training and reemployment services aimed at promoting more stable employment.
APPENDIX 1 - Weeks of Benefits Claimed Regression Results (dependent variable: number of weeks)
Observations: 209,610 - R-square: 0.272 | ||
---|---|---|
Variables Grey shade= characteristics related to job quality |
Coefficient | Statistical significance
***coefficient is statistically different from zero at the <0.001 level ** <0.01 level * <0.05 level |
Gender-Male (reference: Female) | -0.344 | *** |
Race-Black (reference: white) | 2.615 | *** |
Race-Asian | 0.303 | *** |
Race-Latino | 0.444 | *** |
Race-American Indian | 0.508 | *** |
Race-More than one race | 1.475 | *** |
Disability-with a disability (reference: without a disability) | 1.097 | *** |
Disability-did not answer the disability question | 0.740 | *** |
Age 25-30 (reference: 20-24) | 1.267 | *** |
Age 31-36 | 1.778 | *** |
Age 37-45 | 1.937 | *** |
Age 46-54 | 1.998 | *** |
Education-1 or 2 years of college (reference: high school or less) | -0.124 | ** |
Education- Associate degree or 3 years of college | -0.270 | *** |
Education-Bachelor's | -0.564 | *** |
Education-Master's | -0.677 | *** |
Residence- St. Paul (reference: Metro except St. Paul and Minneapolis) | 0.933 | *** |
Residence- Minneapolis | 0.615 | *** |
Laid off from a firm that closed | 2.713 | *** |
Job tenure pre-layoff-from 1 to 3 years (reference: <=1 year) | -0.096 | * |
Job tenure-from 4 to 12 years | -0.671 | *** |
Job tenure-more than 12 years | -1.841 | *** |
Hourly wage pre-layoff from $15 to $21.49 (reference: <$15) | -0.564 | *** |
Hourly wage pre-layoff $21.49-$33 | -1.206 | *** |
Hourly wage pre-layoff >$33 | -1.000 | *** |
Work Status-Employed full-time year-round pre-layoff (reference: part-time or seasonal) | -1.142 | *** |
Permanently separated from employer anytime from Mar to Jun 2020 | 6.001 | *** |
Occupation-Engineers (reference: Sales workers) | -0.723 | *** |
Occupation-Social Scientists & Related Workers | -3.285 | *** |
Occupation-Other Teachers & Instructors (Tutors, Substitute Teachers) | 1.885 | *** |
Occupation-Media & Communication Equipment Workers | 2.632 | *** |
Occupation-Home Health & Personal Care Assistants | 0.933 | *** |
Occupation-Food & Beverage Serving Workers | 3.195 | *** |
Controls for firm size | Yes | - |
Controls for 65 industry categories | Yes | - |
The number of observations, 209,610, is lower than the starting dataset due to the need to exclude claimants who did not report an occupation. Source: Author's calculations based on Minnesota UI claims data merged with MN UI wage records. |
APPENDIX 2 -Probability of Permanent Job Loss (separation) Regression Results (dependent variable: 1=permanently separated at any time Sept 2021 -June 2022, 0=not permanently separated)
Observations: 168,924 - Nagelkerke R-square: 0.215 |
||
---|---|---|
Variables Grey shade= characteristics related to job quality |
Coefficient (in percentages) | Statistical significance
***coefficient is statistically different from zero at the <0.001 level ** <0.01 level * <0.05 level |
Gender- Male (reference: female) | -9% | *** |
Race-Black (reference: white) | 76% | *** |
Race-Asian | -6% | * |
Race-Latino | 17% | *** |
Race-American Indian | 42% | *** |
Race-More than one race | 36% | *** |
Veteran | 24% | *** |
Disability-with a disability (reference: without a disability) | 46% | *** |
Disability-did not answer the disability question | 41% | *** |
Age 25-30 (reference: 20-24) | 9% | *** |
Age 31-36 | 11% | ** |
Age 37-45 | 7% | ** |
Age 46-54 | -0.1% | - |
Education-1 or 2 years of college (reference: high school or less) | -5% | * |
Education- Associate degree or 3 years of college | -17% | *** |
Education-Bachelor's | -38% | *** |
Education-Master's | -46% | *** |
Residence- St. Paul (reference: Metro except St. Paul and Minneapolis) | 21% | *** |
Residence- Minneapolis | 12% | *** |
Laid off from a firm that closed | 20% | *** |
Job tenure pre-layoff-from 1 to 3 years (reference: <=1 year) | -20% | *** |
Job tenure-from 4 to 12 years | -37% | *** |
Job tenure-more than 12 years | -49% | *** |
Hourly wage pre-layoff from $15 to $21.49 (reference: <$15) | -5% | * |
Hourly wage pre-layoff $21.49-$33 | -13% | *** |
Hourly wage pre-layoff >$33 | -30% | *** |
Work Status-Employed full-time year-round pre-layoff (reference: part-time or seasonal) | -4% | * |
Belonged to a Union before Sept 2020 | -54% | *** |
Permanently separated from employer from Mar to Jun 2020 | 25% | *** |
Number of weeks claimed Mar 2020-Mar 2021 | 5% | *** |
Occupation-Engineers (reference: Sales workers) | -38% | *** |
Occupation-Registered Nurses & Other Healthcare Practitioners | -21% | *** |
Occupation-Food processing workers | 49% | ** |
Occupation-Other personal care workers (childcare, etc.) | 44% | *** |
Occupation-Home Health, PCAs, and Nursing Assistants | 49% | *** |
Controls for firm size | Yes | - |
Controls for 65 industry categories | Yes | - |
The number of observations, 168,924, is lower for this regression compared to the Weeks of Benefits regression due to the need to exclude claimants without an employment record in two out four quarters from q2 2020 to q1 2021. Under normal circumstances, UI eligibility requires working in Minnesota during the past 12 to 15 months and having earned at least a minimum amount of wages. Therefore, these exclusions are needed to discard individuals who temporarily qualified for UI during the initial months of the pandemic but likely lost eligibility in late 2021. |
1On September 6, 2021 the additional weeks of pandemic federal benefits ended in all states.
2This time span was chosen because it begins exactly 20 months after the initial wave of COVID-related layoffs and covers a 10 months period of economic recovery.
3Full-time year-round employment is an indicator of job quality because it is typically associated with access to health care and/or retirement benefits.
4Higher frequency of business closures in these two Cities during the last three years likely stems from longer pandemic-related shutdowns as well as from downtown office buildings being emptied out as employees switched to remote work.
5From 2017 to 2021, the estimated share of foreign-born persons is 19.1% in St Paul and 14.8% in Minneapolis. In 2021, median household income was $ 63,483 in St. Paul and $70,099 in Minneapolis. In contrast, the Twin Cities Metro area's share of foreign-born persons was 10.8% (13.8% in Hennepin county) and the median household income in 2021 was $87,433. Therefore, the socio-economic characteristics in the two cities are different and less favorable than those in the rest of the Metro. Source: Minnesota Compass Saint Paul data and Minnesota Compass Twin Cities Region, household by income and Data USA Minneapolis, MN.
6Although some individuals might have chosen not to return to the same employer due to workforce safety concerns or other reasons, the most popular decision was to return whenever possible to avoid having to search for a new job at a time when hiring activity was frozen.
7The dependent variable ranges from 1 to 28 weeks.
8Including the following: age, veteran status, SOC code at the 3 digits, an indicator measuring whether a claimant reported being permanently separated, and industry interacted with permanent separation indicator.
9The R square of this model is .272. All factors listed in the article are statistically significant.
10The model with interactions is available upon request.
11All factors listed in the article are statistically significant.
12This variable is included in UI claims records. Claimants from the Mining, Utilities, and Construction industries are the most likely to be union members and get referred to a hiring hall for reemployment after being laid off. In particular, employers in the construction industry often choose to hire exclusively through referrals from union hiring halls.
13The second model fits the data less well than the first model, with an explanatory power of 21.5% versus 27.2%. A limitation of the second model is that it relies on information collected in Spring of 2020, which might have become outdated in 2021. For example, a claimant who reported having a high school education and being employed as a Cashier in March 2020 might have subsequently earned an associate degree and switched to a higher skilled job over the course of 2021-2022, which is unmeasurable with our data. Another limitation of both models is that important characteristics could not be measured, such as family income, quality of education (at the k-12 and post-secondary level), access to occupational training, criminal record status and citizenship status.