Upjohn Research

Stevenson, Betsey

Swonk, Diane

Horrigan, Michael

Shift or Shock? Long-Term Balance of Power in Labor Markets

Bolter, Kathleen

Hershbein, Brad J.

Hernandez, Alfonso

Timmeney, Bridget F.

Reports Upjohn Research home page 
7-3-2023 
The Kalamazoo Promise: 2022 By the Numbers 
Kathleen Bolter 
W.E. Upjohn Institute for Employment Research, Bolter@upjohn.org 
Brad J. Hershbein 
W.E Upjohn Institute for Employment Research, hershbein@upjohn.org 
Alfonso Hernandez 
W.E. Upjohn Institute for Employment Research 
Bridget F. Timmeney 
W.E. Upjohn Institute for Employment Research, timmeney@upjohn.org 
Follow this and additional works at: https://research.upjohn.org/reports 
 Part of the Higher Education Commons 
Citation 
Bolter, Kathleen, Brad Hershbein, Alfonso Hernandez, and Bridget F. Timmeney. 2023. "The Kalamazoo 
Promise: 2022 By the Numbers." Report prepared for The Kalamazoo Promise. 
https://research.upjohn.org/reports/290 
This title is brought to you by the Upjohn Institute. For more information, please contact repository@upjohn.org. 
The Kalamazoo Promise: 2022 by the Numbers
2022 BY THE NUMBERS
THE KALAMAZOO PROMISE
Across the board, enrollment in college remains lower than it was pre-
pandemic for Kalamazoo Promise–eligible students. However, the generous 
10-year eligibility window for the Promise scholarship shows that extra 
time is having a large impact in helping students finish their degrees. 
ENROLLMENT TRENDS
Compared to the state of Michigan overall, students 
eligible for the Kalamazoo Promise enrolled in 
college at a rate 5 percentage points higher within six 
months of graduating high school.  However, when 
compared to urban districts with similar levels of 
socioeconomic diversity, enrollment for Kalamazoo 
Promise–eligible students was 14 percentage points 
higher.
Kalamazoo Promise– 
Eligible Students
High School Graduates 
in the State of Michigan
Percent of Students Enrolling in College Six Months after Graduation
58% 53%
High School Graduates 
in Similar Urban Districts
44%
The Kalamazoo Promise: 2022 by the Numbers
Enrollment of Kalamazoo Promise scholars in community colleges remains lower than it 
was prior to the pandemic, but enrollment in universities is about the same, and enrollment 
in select private colleges (Michigan College Alliance, or MCA, colleges) is higher than before.
Universities
Community Colleges
MCA Colleges
Total Students Enrolling in College Six Months after Graduation
Kalamazoo Promise–eligible graduates continue to 
enroll primarily at local colleges and universities. 3 in 5 
graduates who 
use the Promise 
enroll in colleges 
or universities in 
Kalamazoo
ENROLLMENT TRENDS 
(continued)
The Kalamazoo Promise: 2022 by the Numbers
SCHOLARSHIP USE
 by Socioeconomic Status
Usage of the Kalamazoo Promise within six months of graduation remains lower than before the 
pandemic for both students who were and were not eligible for free/reduced lunch.
2019 2022
81%
66%
2019 2022
60%
43%
Not Eligible for 
Free/Reduced Lunch
Eligible for 
Free/Reduced Lunch
One innovative design decision by the Kalamazoo Promise is to allow students to access 
funding for up to 10 years and to extend the number of credits covered by the scholarship 
from 130 to 145. Comparing 6-year graduation rates to 10-year graduation rates shows 
that those extra years matter, especially for students eligible for free/reduced lunch 
where, with a bit more time, 31% more students finish their degrees. 
Kalamazoo Promise 
Students 
Not Eligible for Free/
Reduced Lunch
6-Year 
Bachelor’s Degree
Completion Rates
10-Year 
Bachelor’s Degree 
Completion Rates
50%
16%
60%
21%
Percent of Kalamazoo Promise–Eligible Students 
Enrolling in College Six Months after High School Graduation
Kalamazoo Promise 
Students 
Eligible for Free/     Re-
duced Lunch
Change in 
Degree 
Attainment
20%
31%
The Kalamazoo Promise: 2022 by the Numbers
SCHOLARSHIP USE
 by Race, Ethnicity, & Gender
Across all demographic groups, the number of eligible graduates who used the Promise 
within six months of high school graduation is lower than it was before the pandemic. 
Hispanic men and Black women have seen the largest drop in enrollment. 
2019 2022
All Women
53%
70%
2019 2022
All Men
54%
68%
2019 2022
Black Women
38%
67%
2019 2022
Hispanic Women
61%72%
2019 2022
White Women
66%75%
2019 2022
Black Men
40%54%
2019 2022
Hispanic Men
47%
77%
2019 2022
White Men
64%74%
Percent of Kalamazoo Promise–Eligible Students 
Enrolling in College Six Months after High School Graduation
The Kalamazoo Promise: 2022 by the Numbers
25%
31%
49%
19%
32%
46%
44%
29%
SCHOLARSHIP USE
by Race, Ethnicity, & Gender
(continued)
6-Year 
Bachelor’s Degree
Completion Rates
10-Year 
Bachelor’s Degree
Completion Rates
All Women
Black Women
Hispanic Women
White Women
33% 42%
19% 25%
19% 28%
49% 59%
All Men
Black Men
Hispanic Men
White Men
26% 34%
11% 16%
15% 22%
39% 50%
Across all racial and ethnic groups, 10-year completion rates for bachelor’s degrees are 
higher than 6-year completion rates. About 25% more women and 32% more men complete 
their bachelor’s degree with some extra time. Hispanic women and Black men particularly 
benefit from having some additional time to finish their degrees.  
Change in 
Degree 
Attainment


English

The Kalamazoo Promise: 2022 By the Numbers

Ghilarducci, Teresa

Truesdale, Beth C.

The False Promises of Working Longer

Asquith, Brian J.

Bock, Margaret C.

The Case for Dynamic Cities

Stuart, Bryan A.

Place-Based Consequences of Person-Based Transfers: Evidence from Recessions

Pittelko, Brian

Lendel, Iryna V.

Mbwana, Kassim

Ehrich, Kami

The National-Level Economic Impact of the Manufacturing Extension Partnership (MEP): Estimates for Fiscal Year 2022

Pepin, Gabrielle

Employment Research Newsletter 
Volume 30 Number 1 Article 1 
2-2-2023 
State Tax Strategies to Reduce Care Costs 
Gabrielle Pepin 
W.E. Upjohn Institute for Employment Research, pepin@upjohn.org 
Upjohn Author(s) ORCID Identifier: 
https://orcid.org/0000-0002-3156-0268 
Follow this and additional works at: https://research.upjohn.org/empl_research 
 Part of the Labor Economics Commons 
Citation 
Pepin, Gabrielle. 2023. "State Tax Strategies to Reduce Care Costs." Employment Research 30(1): 1-4. 
https://doi.org/10.17848/1075-8445.30(1)-1 
This title is brought to you by the Upjohn Institute. For more information, please contact repository@upjohn.org. 
 
 
 
 
 
 
 
 
JANUARY 2023  • VOL 30, NO 1 
ARTICLE HIGHLIGHTS 
n States wishing to extend 
Child and Dependent Care 
Credit (CDCC) benefits to 
low-income families should 
offer refundable credits that are 
a function of the “allowable” 
federal CDCC. 
 
n A promising strategy to 
maintain quality standards is 
to condition CDCC benefits 
on state-administered quality 
ratings of child care providers. 
n Well-designed state 
CDCCs can increase work 
and earnings while 
transferring income toward 
parents and caregivers.
n Nearly one-quarter of one-
parent households who work 
and pay for child care have 
incomes too low to receive 
benefits. As these households 
are disproportionately Black 
and Hispanic, nonrefundability 
perpetuates both income 
and racial inequality. 
ALSO IN THIS ISSUE 
How Do Broad Non-
Disclosure Agreements 
Affect Labor Markets?
Jason Sockin, Aaron Sojourner, 
and Evan Starr
page 4
State Tax Strategies to Reduce
Care Costs
Gabrielle Pepin 
Child and elder care in the United States is 
notoriously expensive. For instance, in recent years, 
the average household with children in the bottom 
fourth of the income distribution spent nearly 
20 percent of their income on child care alone 
(Herbst 2018). Researchers have found that these 
costs keep many caregivers from working, as they 
would easily eat up a good chunk of take-home 
pay (Averett, Peters, and Waldman 1997; Guner, 
Kaygusuz, and Ventura 2020; Michalopoulos, 
Robins, and Garfnkel 1992; Miller and Mumford 
2015; Pepin 2020). Although many policymakers 
continue to recommend policies designed to 
promote employment opportunities, income 
support for child and elder care at the federal 
level remains quite limited. If states and localities 
wish to create shared prosperity and to help their 
residents get and keep good jobs, they should thus 
act independently to reduce care costs. 
Some states already provide a road map: 
supplements to the federal child and dependent 
care credit (CDCC). When states supplement 
the federal CDCC, residents get an additional 
tax credit based on their income and care 
expenses, efectively lowering their out-of-pocket 
costs. Because the credit is tied to work, it also 
encourages labor force participation. Moreover, 
my research suggests well-designed state CDCCs 
can increase work and earnings while transferring 
income toward parents and caregivers. 
How Does the CDCC Work? 
Te federal CDCC is available to all working 
families with children younger than 13 or with 
a coresident spouse or dependent “physically or 
mentally incapable of self-care.” To qualify for 
benefts, all nondisabled taxpayers in the household 
must have positive annual earnings, including 
both spouses among taxpayers married fling 
jointly. Households meeting this work requirement 
may claim up to $3,000 in care expenses for each 
of up to two qualifying individuals, receiving a 
nonrefundable tax credit worth up to 35 percent of 
those expenses, or $1,050. However, benefts begin 
to decrease once household adjusted gross income 
(AGI) reaches $15,000, plateauing at $600 per 
qualifying individual once AGI exceeds $43,000. 
Conditioning the amount of the
state CDCC on state-administered
quality ratings is appealing because
it creates an incentive for high-
quality care without requiring new
bureaucracy. 
Expenses eligible under the credit include 
fees paid to child and adult daycare facilities, 
family child care homes, and attendants assisting 
dependents with activities of daily living. Claiming 
the credit requires flling out Form 2441 on the 
household’s tax return, reporting expenditures and 
the care provider’s tax identifcation number. 
Te federal CDCC is not as generous as it may 
seem, however. Te federal credit is nonrefundable, 
which means that it can only ofset income taxes 
owed—low-income working households who 
owe no income taxes do not receive benefts. Tis 
stands in stark contrast to other refundable tax 
credits targeted at working families, such as the 
Earned Income Tax Credit, for which households 
may receive a tax refund. 
In practice, this limitation is a big deal. For 
instance, a one-parent household with two children 
and $15,000 in annual earnings has no federal 
income tax liability and therefore will not receive 
any federal CDCC benefts, regardless of how 
much they spend on child care. Meanwhile, a two-
parent household with two children and $50,000 
 
 
 
   EMPLOYMENT RESEARCH  • JANUARY 2023 W.E. UPJOHN INSTITUTE 
State Tax Strategies to Reduce Care Costs 
Figure 1: Several States Provide Their Own Child and Dependent Care Credits, But Availability 
and Generosity Vary Greatly by Family Type 
W.E.UPJOHN 
INSTITUTE 
SOURCE:  Author’s calculations using state and federal tax forms. 
in annual earnings may receive up to a 
$1,200 credit each year. In my research, 
I fnd that nearly one-quarter of one-
parent households who work and pay 
for child care have incomes too low to 
receive benefts. As these households 
are disproportionately Black and 
Hispanic, nonrefundability perpetuates 
both income and racial inequality. 
How Can State CDCCs Reduce 
Care Costs? 
Twenty-two states ofer their own 
state child care credits that are directly 
tied to the federal CDCC. Additionally, 
Idaho, Maryland, Massachusetts, 
Montana, Virginia, and Wisconsin 
ofer individual income tax deductions 
for child care expenses. (Deductions 
lower taxable income while credits 
ofset taxes owed.) Colorado, Iowa, 
New Mexico, Oregon, and the District 
of Columbia ofer child care credits 
that are not a percentage of the 
federal CDCC or the expenses used to 
calculate it. In most cases, these state 
benefts are just a fraction of the federal 
credit. However, unlike the federal 
CDCC, some states ofer refundable 
credits, limit benefts to taxpayers with 
incomes below a certain threshold, or 
provide larger benefts to low-income 
households.
For example, Maine and Delaware 
each provide a state care credit equal 
to half of the federal credit, without 
refundability. Tus, the hypothetical 
one-parent household with two 
children and $15,000 in earnings 
receives no care credits in these states 
because its income was too low to 
get any federal credit. In contrast, 
California and Louisiana ofer CDCCs 
worth up to $1,050, half of the $2,100 
in federal benefts the household 
would receive if the federal CDCC 
were refundable. California’s CDCC 
is itself nonrefundable, though, so the 
hypothetical household with $15,000 
in earnings, which does not have state 
income tax liability, still receives no 
benefts. Only in Louisiana, which 
ofers a refundable CDCC based on 
2 
 
   
 
 
 
 
 
 
 
 
EMPLOYMENT RESEARCH  • JANUARY 2023 W.E. UPJOHN INSTITUTE
the allowable federal credit, may the 
household receive up to $1,050 in state 
CDCC benefts. 
Te efectiveness, or “take-home,” of 
state CDCC benefts can vary by place 
and income. For example, single-parent 
families with two children, $15,000 in 
earnings, and no additional income are 
unlikely to have access to state CDCC 
benefts. Among the fve states that 
do ofer benefts to these households, 
however, state benefts top out at over 
$1,000 per year. 
On the other hand, two-parent 
families with two children and $50,000 
in earnings are much more likely to 
have access to state CDCCs. Twenty 
states ofer these households benefts, 
with maximum benefts ranging from 
$120 to $2,100 annually. 
How Can States Help to Ensure 
Care Quality? 
One concern about the CDCC is 
that it may lead to reduced care quality 
if it encourages households to pay for 
low-quality care instead of providing 
To help alleviate concerns
about care quality, a few
states tie their CDCCs
to state-administered
ratings of provider quality. 
higher-quality care themselves. Tis 
concern is valid: the quality of child 
care afects children’s development and 
long-term outcomes (Cornelissen et 
al. 2018; Cunha and Heckman 2007; 
Havnes and Mogstad 2011). To help 
alleviate concerns about care quality, a 
few states tie their CDCCs to state-
administered ratings of provider 
quality. For example, for the single-
parent, low-income family, maximum 
benefts in Louisiana increase with 
the rating of child care provider, from 
$1,050 at any provider up to $5,250 at 
providers assigned the highest possible 
fve-star rating. 
Figure 2: Single-Parent, Low-Income Family, Maximum Benefts in Louisiana Increase 
with the Rating of Child Care Provider 
W.E.UPJOHN 
INSTITUTE 
NOTE: Maximum Louisina CDCC benefts for a one-parent household with two children younger than six 
years old and $15,000 in earnings. 
SOURCE: Author’s calculations using federal and Louisiana tax forms. 
 Conditioning the amount of the 
state CDCC on state-administered 
quality ratings is appealing because 
it creates an incentive for high-
quality care without requiring new 
bureaucracy: 40 states and the District 
of Columbia already administer 
Quality Rating and Improvement 
Systems to assess early care and 
education programs’ quality levels. 
Parents can access these ratings and 
other provider information online, 
allowing them to see how choosing 
diferent providers would afect the size 
of their CDCCs. Although these quality 
rankings tend to exist only for child 
care providers, states could extend 
the rating systems for care providers 
serving adults with disabilities. 
Conclusion 
States can supplement the federal 
CDCC with their own credits to lower 
residents’ out-of-pocket care costs 
while encouraging work. For state 
CDCCs to be effective labor market 
policies, however, they must reach 
their targeted beneficiaries. To reach 
low-income families, states should 
thus offer refundable care credits tied 
to the federal CDCC. Making the size 
of these state credits more generous at 
higher-rated care providers may help 
ensure care quality.
References 
Averett, Susan L., H. Elizabeth Peters, 
and Donald M. Waldman. 1997. “Tax 
Credits, Labor Supply, and Child Care.” 
Review of Economics and Statistics 79(1): 
125–135. 
Cornelissen, Tomas, Christian 
Dustmann, Anna Raute, and Uta 
Schönberg. 2018. “Estimating Marginal 
Returns to Early Child Care Attendance.” 
Journal of Labor Economics 126(6): 
2356–2409. 
Cunha, Flavio, and James Heckman. 
2007. “Te Technology of Skill 
Formation.” American Economic Review 
97(2): 31–47. 
Guner, Nezih, Remzi Kaygusuz, and 
Gustavo Ventura. 2020. “Child-Related 
Transfers, Household Labor Supply and 
Welfare.” Review of Economic Studies
87(5): 2290–2321. 
3 
 
 
 
 
 
 
 
 
   
 
 
 
 
 
 
 
 
 
 
 
 
  
EMPLOYMENT RESEARCH  • JANUARY 2023 W.E. UPJOHN INSTITUTE 
State Tax Strategies to Reduce
Care Costs 
Havnes, Tarjei, and Magne Mogstad. 
2011. “Subsidized Child Care and 
Children’s Long-Run Outcomes.” 
American Economic Journal: Economic 
Policy 3(2): 97–129. 
Herbst, Chris M. 2018. “Te Rising 
Cost of Child Care in the United States: 
A Reassessment of the Evidence.” 
Economics of Education Review 64: 
13–30. 
Michalopoulos, Charles, Philip K. 
Robins, and Irwin Garfnkel. 1992. “A 
Structural Model of Labor Supply and 
Child Care Demand.” Journal of Human 
Resources 27(1): 166–203. 
Miller, Benjamin M., and Kevin 
J. Mumford. 2015. “Te Salience of 
Complex Tax Changes: Evidence from 
the Child and Dependent Care Credit 
Expansion.” National Tax Journal 68(3): 
477–510. 
Pepin, Gabrielle. 2020. “Te Efects 
of Child Care Subsidies on Paid Child 
Care Participation and Labor Market 
Outcomes: Evidence from the Child 
and Dependent Care Credit.” Upjohn 
Institute Working Paper No. 20-311. 
Kalamazoo, MI: W.E. Upjohn Institute 
for Employment Research. 
This article is part of a report from the Upjohn
Institute’s “Promise: Investing in Community”
initiative, which ofers strategies to create good jobs
and guidance for local and state policymakers to
implement them. The report, Bridging Research and
Practice to Achieve Community Prosperity, features
short articles demonstrating key takeaways from
research and practice on each of these themes.
Focusing on economic and workforce development,
it showcases how state governments can do more
to help their distressed places, highlighting the role
of neighborhood employment hubs that bring job
assistance directly to residents. 
Gabrielle Pepin is an economist at the Upjohn
Institute. 
4 
How Do Broad Non-
Disclosure Agreements
Afect Labor Markets? 
Jason Sockin, Aaron Sojourner, and Evan Starr 
As anybody who has ever changed 
employers knows, taking a new job is 
risky. Even aside from career concerns, 
there is a chance you may face an 
unpleasant working environment or 
even toxic behavior, such as sexual 
harassment, discrimination, or wage 
thef. Employers can exacerbate the 
uncertainty job seekers face by using 
broad non-disclosure and non-
disparagement agreements (NDAs) to 
prohibit current and former employees 
from speaking about their negative 
experiences at work. Tese policies 
can keep job seekers in the dark about 
important aspects of working at a 
potential new employer. 
Te #MeToo movement put 
a spotlight on how common it is 
for frms to use broad NDAs to 
silence workers, helping perpetuate 
misconduct at work. Indeed, 
management increasingly requires 
workers to sign broad NDAs as a 
condition of employment, with the 
express goal of limiting employees from 
sharing negative information about 
the employer, even on supposedly 
anonymous websites like Glassdoor. 
Over half of U.S. workers are now 
bound by NDAs (Balasubramanian, 
Starr, and Yamaguchi 2022). 
Despite this recent attention, the 
subsequent debates about whether 
frms should be allowed to silence 
workers with NDAs, and recent 
legislation passed by Congress to 
prohibit frms from using broad NDAs 
to conceal sexual harassment, little if 
any research has examined how broad 
NDAs infuence labor markets. Our 
new study helps fll this gap. 
Core to our study is the passage of 
laws in California, Illinois, and New 
Jersey, in the wake of the #MeToo 
movement in the late 2010s, that 
prohibited frms from using NDAs 
to conceal unlawful conduct at work. 
Using the passage of these laws as a 
natural experiment, alongside variation 
across industries where NDA use is 
low versus high—where the “bite” of 
the laws is greater—we ask how broad 
NDAs afect three outcomes: 
1) public measures of frms’ 
reputation 
2) the ability of competing 
employers to diferentiate 
themselves within the market 
3) labor market sorting of workers 
across frms 
Using data drawn from worker 
reviews of employers on Glassdoor, 
ARTICLE HIGHLIGHTS 
n Non-disclosure and non-disparagement agreements (NDAs) cover more than half 
of U.S. workers.
n Firms use broad NDAs to prevent workers from speaking out about the harms they 
experience at work.
n Silencing workers with negative information to share makes it harder for better 
employers to stand out, and harder for workers to avoid bad employers.
n State laws that narrow NDAs mitigate these negative spillovers and help improve 
the flow of information about employers.


State Tax Strategies to Reduce Care Costs

Bartik, Timothy J.

Presentations Upjohn Research home page 
7-18-2023 
Discussant Comments on Hyman et al. Paper on California 
Competes Tax Credit (CCTC) 
Timothy J. Bartik 
W.E. Upjohn Institute for Employment Research, bartik@upjohn.org 
Follow this and additional works at: https://research.upjohn.org/presentations 
 Part of the Labor Economics Commons 
Citation 
Bartik, Timothy J. 2023. "Discussant Comments on Hyman et al. Paper on California Competes Tax Credit 
(CCTC)." Presented at the 12th Annual Municipal Finance Conference, Hutchins Center for Fiscal and 
Monetary Policy, Brookings Institution on July 18, 2023. 
https://research.upjohn.org/presentations/97 
This title is brought to you by the Upjohn Institute. For more information, please contact repository@upjohn.org. 
Tim Bartik 
W.E. Upjohn Institute for Employment Research
Comments presented at 12th Annual Municipal Finance Conference, 
Hutchins Center for Fiscal and Monetary Policy, Brookings Institution
July 18, 2023
Discussant Comments on Hyman et al 
paper on California Competes Tax Credit 
(CCTC)  
Paper convincingly shows large effects of  CCTC on firm’s short-run job creation; if  
long-run California job effect, CCTC cheap, pays for itself, and has high benefits. 
• One of  best papers ever on state business tax credits, because it convincingly estimates 
causal effects using “natural experiment”: variation in scoring cutoff  for CCTC.
• CCTC provides one-time credit averaging $10K/job. Estimates imply job creation equal 
to 82% of  job creation at proposed project site. Job creation costs, after credit 
recaptures, of  $7,721 per job. 
• $7,721 net cost less than 1/20th the average cost per job created of  state/local business 
tax cuts from research literature, of  $165K per job created. (Bartik 2020).
• If  these represent effects on California LR jobs, program pays for itself  within 2 years. 
(Average pay of  $60K/job, California taxes of  at least 8%) Given other benefits (e.g., 
higher employment rates), CCTC should be adopted by other states. 
But CCTC effects seem too large to be plausible LR effects on 
California jobs
• If  estimated firm job effects are long-run job effects, and are due to effects on 
probability of  project being undertaken, then effects on jobs are extremely large relative to 
reduction in payroll costs. One-time incentive of  $10K is 17% of  1-year payroll per job 
of  $60K, but much lower % of  present value of  future payroll costs: at 12% real 
discount rate, $10K represents 2% reduction in present value of  payroll costs. Effect on 
probability of  project being undertaken is 82 percentage points, so this 2% wage subsidy 
QUINTUPLES jobs. (100% over 18% probability of  project w/o CCTC).   
• If  CCTC effects are LR California job creation effects, CCTC probably increased 
California job creation from 2014-2021 by 250K jobs. (=153K jobs incentivized times 
82% effectiveness ratio times multiplier of  2). This is 18% of  California’s job growth 
during this period. But this is not apparent in aggregate job trends (see figure). 
California aggregate job statistics show no sign of  increased growth starting in 
2014, when CCTC started
Alternative interpretations of  this paper’s CCTC effects
• Alternative interpretation 1: Effects are SR effects on timing of  firm’s job creation, with 
much smaller effects on LR job creation by firm. Firms that are flexible about when to 
undertake a project may initially ask for large incentives, and lower ask in subsequent 
incentive rounds. Depending upon when their incentive request passes CCTC score 
cutoff, firm may accelerate or delay project.   
• Alternative interpretation 2: As in other states, California’s state economic development 
agency provides extensive assistance to helping selected firms get environmental permits 
and overcome zoning and other challenges. States are selective on who receives this 
assistance. Advent of  CCTC could mean that state uses CCTC as part of  screening 
process in who receives zoning & permitting help. But if  number of  jobs in firms 
receiving such help does not change after CCTC, zero effect on aggregate job growth in 
state; effect is on who adds jobs, not on job growth. 


Discussant Comments on Hyman et al. Paper on California Competes Tax Credit (CCTC)

Ivandić, Ria

Lassen, Anne Sophie

Upjohn Institute Policy and Research Briefs Upjohn Research home page 
8-28-2023 
Gender Gaps in Employment and Earnings after Job Loss 
Ria Ivandić 
University of Oxford and London School of Economics, ria.ivandic@politics.ox.ac.uk 
Anne Sophie Lassen 
Copenhagen Business School 
Follow this and additional works at: https://research.upjohn.org/up_policybriefs 
 Part of the Labor Economics Commons 
Citation 
Ivandić, Ria and Anne Sophie Lassen. 2023. "Gender Gaps in Employment and Earnings after Job Loss." 
Policy and Research Brief. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research. 
https://doi.org/10.17848/pb2023-60 
This title is brought to you by the Upjohn Institute. For more information, please contact repository@upjohn.org. 
POLICY BRIEF
BRIEF HIGHLIGHTS 
 
n  Job loss leads to persistently lower 
earnings and higher unemployment, 
but little is known about gender 
differences. 
n  Focusing on workers who lost 
their job due to a plant closure, we 
examine gender gaps and disentangle 
the role of child care responsibilities, 
experience, and education. 
n  We find that women face a 40–45 
percent greater risk of unemployment 
and lower earnings in the first two 
years after job loss.
n  The majority of the gap remains 
even when we compare men and 
women with similar labor market 
experience and socioeconomic 
characteristics.
n  The gender gap for unemployment 
risk is more than twice as large when 
the workers have children.
n  If these differences by presence of 
children did not exist, the earnings 
gender gap would be half as large, 
and the employment gender gap 
would be one-third smaller.
For additional details, see the full working 
paper at https://research.upjohn.org/up_
workingpapers/387/.
Research has long established that job loss leads to lower earnings and a higher risk 
of unemployment. However, we know little about how recovery from a job loss differs 
for women and men. Many job-loss studies exclude women from the analysis or pay 
scant attention to gender differences. To fill this gap, we directly investigate gender gaps 
following job losses and disentangle the roles of child care and human capital. 
More specifically, we use administrative data from Denmark to track the employment 
and earnings of individual workers following plant closures that triggered job losses. We 
compare workers who lose their job to matched control workers of the same gender. We 
find that both men and women face an increased risk of unemployment and substantial 
earnings losses after their job had been displaced. However, in the first two years, women’s 
unemployment risk and earnings losses are 45 percent greater than men’s. Gender gaps 
are particularly large among parents and among workers with little formal education. 
Prior to losing their job, women and men have different labor market histories and 
sociodemographic characteristics. We therefore estimate conditional gender gaps in 
which we match men and women who look similar on these characteristics. We find 
that the conditional gaps are smaller, but never fully close. Digging deeper, we examine 
which characteristics can explain the gender gaps. Education and predisplacement 
earnings are important drivers. Additionally, while men and women are almost equally 
likely to be parents, the impact of children is strikingly different. If men and women were 
equally affected by small children, the gender gap in earnings would be halved, and the 
employment gap would shrink by one-third. 
Our research points to two relevant implications. First, excluding women from 
analyses of job loss (or other disruptions) can lead to an incomplete picture of its 
economic effects. Second, the gender gap after job loss is large, and a likely mechanism is 
that child care responsibility prevents women’s adjustment to labor market shocks.
Background
Many workers become unemployed at some point during their lifetime, and 
research has shown that such job loss can lead to persistently lower earnings and higher 
unemployment (Jacobson, LaLonde, and Sullivan 1993)—at least for men. With more 
women working today than in the past, their risk of job loss has increased, yet we know 
little about gender differences in labor market recovery following job loss. 
We address this shortcoming by investigating the effects of job loss on future labor 
market outcomes for both women and men. To do so, we use high-quality administrative 
data from Denmark that covers all employment relationships, including information 
matching workers to firms. We also observe background information on each individual, 
such as their labor market experience, education, and family characteristics. This 
allows us not only to estimate gender gaps following displacement, but also to attribute 
how much of the gender gap is associated with child-related factors and how much is 
connected to differences in labor market experience.
Gender Gaps in Employment 
and Earnings after Job Loss
Ria Ivandić and Anne Sophie Lassen
AUGUST 2023W.E.UPJOHN
INSTITUTE
FOR EMPLOYMENT RESEARCH
Gender Gaps in Employment and Earnings after Job Loss
POLICY BRIEF | AUGUST 2023 W.E. UPJOHN INSTITUTE
2
Women on average 
experience a 45 
percent higher risk of 
unemployment than men 
in the two years after 
losing their job.
We focus on job losses from manufacturing plant closures, which reflect circumstances 
beyond the individual worker’s control. We study men and women who were employed 
at a plant that closed between 1995 and 2006 and that had been employed at that plant 
at least one year before its closure. We compare the evolution in these displaced workers’ 
earnings, risk of unemployment, and nonparticipation to similar workers, matched on 
sociodemographic characteristics, who were employed in a plant that did not close. 
These latter workers represent the trajectories that the displaced workers would have 
experienced if their plants had not closed. We compute the gender gap as the difference 
between displaced men and women compared to nondisplaced individuals of the same 
gender. 
The Gender Gap after Job Loss
As Figure 1 shows, we find substantial gender gaps in the risk of unemployment 
and earnings loss following job loss (visualized as the difference between the blue and 
black lines). For both men and women, job loss leads to an increase in unemployment 
(panel a) and a reduction in earnings (panel b) for several years. Women on average 
experience a 14.2 percentage point increase in the probability of unemployment over 
Figure 1  Following Job Displacement, Workers See a Rise in Unemployment and a 
Fall in Earnings, and a Gender Gap Emerges 
NOTE: Job displacement occurs between −1 and 0, marked by the vertical black line. Black triangles denote 
displaced men while blue circles denote displaced women, each relative to a control group of workers of the 
same gender who are not displaced. The red diamonds show effects for displaced men that on average have 
similar observable characteristics as the sample of women. The outcome in panel (a) represents the likelihood 
of  claiming unemployment benefits for at least three months in a calendar year. The outcome in panel (b) 
represents the proportional change in earnings relative to the period 2–4 years before job displacement, while 
the outcome in panel (c) represents the change in absolute earnings (Danish krone) relative to the year before 
job displacement. Panel (d) reports a measure of the fraction of the year spent neither working nor registered as 
unemployed. 
(a) Unemployment, by Gender (b) Change in Earnings, by Gender
(c) Earnings, by Gender (d) Non-participation, by Gender
-0.05
0.00
0.05
0.10
0.15
0.20
-5 -4 -3 -2 -1 0 1 2 3 4 5 6
Ef
fe
ct
Event time
Women
Matched men
Men
-0.40
-0.30
-0.20
-0.10
0.00
0.10
0.20
-5 -4 -3 -2 -1 0 1 2 3 4 5 6
Ef
fe
ct
Event time
Women
Matched men
Men
-70,000
-60,000
-50,000
-40,000
-30,000
-20,000
-10,000
0
10,000
20,000
-5 -4 -3 -2 -1 0 1 2 3 4 5 6
Ef
fe
ct
Event time
Women
Matched men
Men
-50
-30
-10
10
30
50
70
90
110
130
150
-5 -4 -3 -2 -1 0 1 2 3 4 5 6
Ef
fe
ct
Event time
Women
Matched men
Men
POLICY BRIEF | AUGUST 2023 W.E. UPJOHN INSTITUTE
3
the first two years, while for men this increase is smaller, 9.8 percentage points. This 
difference amounts to a relative gender gap of 5.4 percentage points, or 45 percent  
(14.2 / 9.8), in the risk of unemployment. This gap shrinks over time and closes four 
years after job displacement. 
Women also experience a larger relative loss in earnings (panel b). In the first year 
after job displacement, the gender gap in earnings loss is 44 percent, as men lose on 
average 19.6 percent of their earnings while women lose 28.2 percent of theirs. In the 
fourth year following displacement, the gender gap disappears. As shown in panel (d), 
we do not find a gender gap in labor force participation after a plant closing. 
A potential source of these gender gaps could be differences in human capital, 
broadly defined to include education, experience, occupation, and other types of sorting 
in the labor market (Goldin 2014). Men and women might be different along these 
characteristics, which could in turn affect their labor market recovery beyond the role of 
gender, per se. To account for gender differences in these characteristics, we construct 
a new sample of men who are similar to the women along these lines. This allows us to 
compute the gender gap taking these factors into consideration. When we compare men 
and women with similar characteristics, shown as the red line in Figure 1, we find that 
gender gaps are smaller, but the majority of the gaps remain. 
We also document which workers’ characteristics correlate to worse labor market 
outcomes after job loss. Workers with little formal training face the most adverse labor 
market trajectories after job loss, with a large relative gender gap. Meanwhile, there is 
little or no gender gap among workers with vocational training or higher education. 
While women are worse off across all age groups, older women face the greatest absolute 
risk of unemployment and the biggest drop in earnings. The relative gender gaps, 
however, are greatest among workers between ages 35 and 50. 
The Presence of Children Helps Explain the Gender Gap
In Figure 2 we also show that the gender gap in unemployment risk is more than 
twice as large when the worker has children, from 33 percent in households without 
children to 80 percent in households with children. Because other characteristics also 
might influence the gender gap, we use a decomposition technique to understand how 
much individual and household factors matter. Specifically, we examine the importance 
of human capital, such as formal education and cumulative work experience, as well 
as the presence of different ages of children, from preschool to primary school age to 
teenagers. We find that gender differences in human capital can explain one-third of 
the gap in unemployment and two-thirds of the gap in earnings, with differences in 
predisplacement earnings and education being the most important factors. However, this 
exercise also shows that having children increases the gender gap following the job loss, 
regardless of mothers’ experience, earnings, and education. 
While men and women are equally likely to be parents, we show that the impact of 
children on unemployment and earnings after job displacement is different. The presence 
of children in the household especially decreases women’s earnings and increases their 
risk of unemployment following a job loss. Preschool children have the biggest impact 
on the gender gap, while the presence of teenagers does not matter much. If men and 
women were equally affected by the presence of young children, we find that the gender 
gap in earnings would be half as large, and the gender gap in employment would be one-
third smaller. In contrast, gender differences in occupations and industries do not affect 
the gender gap in unemployment following displacement.
Conclusion
Our findings suggest that women suffer greater risks of unemployment and earnings 
loss after job displacement than do men, creating a gender gap in their recovery. In the 
full paper, we discuss how certain aspects of the Danish labor market may cause these 
The gender gap for 
unemployment risk is 
more than twice as large 
when the workers have 
children, suggesting 
that child care is crucial 
in understanding 
women’s greater risk of 
unemployment after job 
loss.
patterns to generalize and create even larger gender gaps in other settings. Our analysis 
also provides insight into the persistent gender inequality associated with children. 
When the responsibility of child care falls disproportionately on women, it likely 
imposes a barrier to labor market recovery. 
References
Goldin, Claudia. 2014. “A Grand Gender Convergence: Its Last Chapter.” American Economic 
Review 104(4): 1091–1119.
Jacobson, Louis S., Robert J. LaLonde, and Daniel G. Sullivan. 1993. “Earnings Losses of 
Displaced Workers.” American Economic Review 83(4): 685–709.
Ria Ivandić is a lecturer in comparative politics and quantitative methods at the University of Oxford. 
Anne Sophie Lassen is a postdoctoral researcher at the Copenhagen Business School. This article is 
based on “Gender Gaps from Labor Market Shocks,” which is forthcoming at the journal Labour 
Economics.
Excluding women from 
analyses of job loss can 
lead to an incomplete 
picture of its economic 
effects. 
Gender Gaps in Employment and Earnings after Job Loss
POLICY BRIEF | AUGUST 2023 
4
WEBSITE
W.E. Upjohn Institute for 
Employment Research @UpjohnInstitute
upjohn.org
Figure 2  Gender Gaps Following Job Displacement Are Larger When Children Are in 
the Household
NOTE: See note to Figure 1.
(a) Unemployment, Child in the Household (b) Change in Earnings, Child in the Household
(c) Unemployment, No Child in the Household (d) Change in Earnings, No Child in the Household
-0.05
0.00
0.05
0.10
0.15
0.20
-5 -4 -3 -2 -1 0 1 2 3 4 5 6
Ef
fe
ct
Event time
Women
Matched men
Men
-0.35
-0.30
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
-5 -4 -3 -2 -1 0 1 2 3 4 5 6
Ef
fe
ct
Event time
Women
Matched men
Men
-0.05
0.00
0.05
0.10
0.15
0.20
-5 -4 -3 -2 -1 0 1 2 3 4 5 6
Ef
fe
ct
Event time
Women
Matched men
Men
-0.40
-0.30
-0.20
-0.10
0.00
0.10
0.20
-5 -4 -3 -2 -1 0 1 2 3 4 5 6
Ef
fe
ct
Event time
Women
Matched men
Men


Gender Gaps in Employment and Earnings after Job Loss

O\u27Leary, Christopher J.

Kline, Kenneth J.

Stengle, Thomas A.

Wandner, Stephen A.

Employment Research Newsletter 
Volume 30 Number 3 Article 1 
8-8-2023 
Why Are Unemployment Insurance Claims So Low? 
Christopher J. O'Leary 
W.E. Upjohn Institute for Employment Research, oleary@upjohn.org 
Kenneth J. Kline 
W.E. Upjohn Institute for Employment Research, Kline@upjohn.org 
Thomas A. Stengle 
U.S. Dept. of Labor, Retired 
Stephen A. Wandner 
National Academy of Social Insurance, Stephen.Wandner@gmail.com 
Upjohn Author(s) ORCID Identifier: 
https://orcid.org/0000-0002-3372-7527 
https://orcid.org/0000-0002-0366-3390 
Follow this and additional works at: https://research.upjohn.org/empl_research 
 Part of the Labor Economics Commons 
Citation 
O'Leary, Christopher J., Kenneth J. Kline, Thomas A. Stengle, and Stephen A. Wandner. 2023. "Why Are 
Unemployment Insurance Claims So Low?" Employment Research 30(3): 1-4. https://doi.org/10.17848/
1075-8445.30(3)-1 
This title is brought to you by the Upjohn Institute. For more information, please contact repository@upjohn.org. 
 
 
 
 
 
 
 
 
 
JULY 2023  • VOL 30, NO 3 
ARTICLE HIGHLIGHTS 
n Initial claims for 
unemployment insurance have 
fallen sharply over the past 
three decades, especially given 
the growth in the labor force, 
and even after adjusting for 
economic conditions.
 
n We investigate this 
phenomenon using a 
decomposition analysis, 
through which we try to 
explain the decline through 
changes in workforce 
characteristics and changes in 
how these characteristics 
affect claims.
n We find that changes in the 
industry and occupation mix 
of workers play an important 
role, as more people work in 
service jobs, which have lower 
historical claims rates.
n One of the most important 
factors is the cut in UI 
generosity, both in the 
duration and the size of weekly 
benefits, as implemented 
by certain states after the 
financial crisis of 2007–2009.
ALSO IN THIS ISSUE 
How the Manufacturing 
Extension Partnership Can Anchor 
U.S.  Workforce Development
Matthew D. Wilson, Nichola Lowe, 
Greg Schrock, Rumana Rabbani, 
and Allison Forbes
page 4
Why Are Unemployment
Insurance Claims So Low? 
Christopher J. O’Leary, Kenneth J. Kline, Thomas A. Stengle, and Stephen A. Wandner 
Why are weekly unemployment insurance (UI) 
claims so much lower now than 30 years ago? 
Te dramatic decline in average weekly UI claims 
means the program now provides weaker wage 
insurance for unemployed workers and is less 
efective as an automatic macroeconomic stabilizer. 
Although the number of weekly initial claims 
remains a leading indicator of aggregage economic 
activity, its importance has diminished. 
We use a statistical decomposition methodology 
to identify the main factors for the decline in 
weekly UI claims. Tis decomposition answers 
the following two questions: 1) What would the 
level of claims have been in recent years had the 
characteristics of workers and state UI programs 
remained as they were 30 years ago? 
2) Alternatively, what would the level of claims 
have been recently if worker characteristics and UI 
programs changed, but their relationship with 
UI claims had stayed the same? 
Our analysis of state-level data over the past 
three decades, with additional detail found in our 
Upjohn Institute working paper, suggests that the 
decline in UI claims stems from three factors: 
1) changes in both the industrial and the 
occupational mix of employment, 2) UI program 
changes made by individual states, and 3) the 
interactions of these factors. Specifcally, declines 
in manufacturing employment—which historically 
has had high claims rates—and increases in the 
health care and education workforce—which 
have had relatively low claims—play a substantive 
role. However, state policy changes that have 
led to lower potential durations of UI benefts 
and reduced wage replacement rates also have 
contributed to the decline in claims. Tis decline 
could be ofset by federal requirements for states 
to improve beneft access, wage replacement rates, 
and beneft durations. Such changes could restore 
the role of UI as meaningful social insurance 
against job loss and as a potent automatic stabilizer 
of the macroeconomy. 
Background 
In the year 2000, with the unemployment rate 
at 3.9 percent and a labor force of 142.6 million 
workers, there were about 300,000 UI claims 
The dramatic decline in average
weekly UI claims means the
program now provides weaker wage
insurance for unemployed workers
and is less efective as an automatic
macroeconomic stabilizer. 
per week. In 2019, right before the pandemic, 
the unemployment rate held at 3.7 percent, the 
labor force had grown to 163.5 million workers, 
and UI claims averaged just over 216,000 per 
week. Although the labor force participation rate 
had declined from 67.1 percent to 63.1 percent 
over those 19 years, that decline cannot explain 
why weekly claims dropped 28 percent while 
the labor force increased by 15 percent and the 
unemployment rate was only a hair lower. 
To illustrate the decline in UI claims, we graph 
the annual average of weekly initial UI claims 
between 1990 and 2019, along with the annual 
average of the count of unemployed workers 
(Figure 1). Te two curves moved together tightly 
(on diferent scales) until the fnancial crisis of 
2007–2009, afer which average weekly UI claims 
fell faster than unemployment. We confne our 
analysis of UI claims to the prepandemic years 
because of the major disruption in UI trends 
caused by the dramatic rise in layofs and the 
   EMPLOYMENT RESEARCH  • JULY 2023 W.E. UPJOHN INSTITUTE 
Why Are Unemployment Insurance Claims So Low? 
Figure 1  Regular UI Weekly Initial Claims and the Level of Unemployment for the 50 States workers and state programs had been 
plus the District of Columbia, 1990–2019 as they were in the 1990s. Te most 
important factors behind these changes 
are shifs in the relationships relating 
625,000 16,000 to industry and occupation, labor force 
575,000 14,500 demographics, and beneft generosity. 
Initial claims 
Unemployed 
W.E.UPJOHN 
INSTITUTE 
With regard to changes in industry 
525,000 13,000 shares of employment, the most 
notable factors are a substantial 
decrease in manufacturing (of 4.6 
percent), which historically has a high 
claims rate, and increases in both 
health care and professional services 
(a combined 3.6 percent), which 
historically have lower claims rates. Un
em
pl
oy
m
en
t (
in
 th
ou
sa
nd
s)
 
475,000 11,500 
W
ee
kl
y 
in
iti
al
 c
la
im
s 
425,000 10,000 
375,000 8,500 
325,000 7,000 
275,000 5,500 
Te relationships between claims and 
225,000 
175,000 
4,000 
2,500 
employment shares in manufacturing 
and health care both became more 
important over time, so that the 
decline in manufacturing employment 
reinforced the decline in UI claims, 
while growth in health care somewhat 
ofset the decline. 
For occupations, the shares of 
employment in transportation and 
sales declined (by a combined 6.0 
percent), while growth occurred 
in occupations with relatively low 
historical claims rates, including 
services, education, and health care 
(by a combined 5.3 percent). Changes 
in the relationship between these 
occupations and UI claims were 
modest, although the association 
between health-care jobs and claims 
turned negative. Tis latter change, 
and health care’s growing share of 
employment, put downward pressure 
on UI claims. 
Among demographic groups, the 
shares of employment by Asians, 
Blacks, Hispanics, and workers aged 55 
and above all increased, also putting 
downward pressure on UI claims, 
as these groups have lower-than-
average claims behavior. Moreover, 
the relationship between most of 
these groups and claims became more 
negative over time, intensifying the 
decline. On the other hand, women’s 
share of employment increased slightly, 
as did the relationship between female 
employment and claims. Tis put 
SOURCE: Initial claims data: U.S. Department of Labor, Employment and Training Administration; downloaded from 
https://oui.doleta.gov/unemploy/claimssum.asp. Unemployed data: U.S. Department of Labor, Bureau of Labor 
Statistics, Local Area Unemployment Statistics; downloaded from https://www.bls.gov/lau/data.htm. 
introduction of UI pandemic programs 
in 2020. However, we note that by 
the fall of 2022, weekly UI claims had 
declined below their level in 2019. In 
February 2023, afer nearly a year of 
the Federal Reserve having steadily 
increased the target interbank lending 
rate, the seasonally adjusted four-week 
moving average of initial UI claims 
began to rise, reaching about 240,000 
by the beginning of April. As of this 
writing, it is not clear whether the 
increase in claims—although still at 
historically low levels—is a trend or 
just a temporary blip. 
Reasons for the Decline in Average 
Weekly UI Claims 
Our decomposition allows us to 
isolate how changes in characteristics 
of workers and the UI program afect 
UI claims with respect to the role 
of any changes in the relationship 
of those characteristics to claims. 
Te lef three columns of Table 1 
summarize how average initial weekly 
claims in 2019 would be predicted to 
change if the relationship were held 
fxed at the 1990–2001 period but 
characteristics of workers and the UI 
program stayed at their 2019 levels. Te 
right three columns of Table 1 show 
nearly the reverse: holding fxed the 
average relationship over 2002–2019 
but simulating the characteristics of 
workers and state UI programs from 
1990–2001. Each row shows the efect 
on predicted claims from changing one 
characteristic at a time. 
Te last row of Table 1 indicates 
that had the relationship between 
characteristics and claims stayed as it 
was in 1990–2001 but characteristics 
were held at 2019 levels, the number of 
average weekly UI claims would have 
been nearly 114,000 (or 53 percent) 
higher than it actually was. In contrast, 
had the overall relationship from 2002– 
2019 been held with characteristics 
from 1990–2001, average weekly claims 
would have been about 20,000 (or 9 
percent) higher. Put diferently, average 
weekly UI claims in 2019 would have 
been more than 133,000 higher if both 
the relationship and characteristics of 
2 
   
 
 
EMPLOYMENT RESEARCH  • JULY 2023 W.E. UPJOHN INSTITUTE 
upward pressure on claims, although Table 1  Simulating How Average Weekly Initial UI Claims for 2019 Would Change if Worker 
not enough to ofset the role played Characteristics and/or Relationships with These Characteristics Were Fixed at Earlier Periods 
by changes in the other demographic 
groups. Relationship from 1990–2001 Characteristics from 1990–2001 
Changes in state UI beneft 
generosity also signifcantly reduced UI 
claims (Table 1), a fnding that supports 
the argument made by DeAntonio 
(2018). Immediately following the 
fnancial crisis of 2007–2009, nine 
states cut potential durations of UI 
from the common 26 weeks to shorter 
potential durations.1 To illustrate how 
these cuts afected UI claims, Figure 2 
plots indices (1990 = 1.00) of UI claims 
for this group of nine states and for all 
the other states. While UI claims rates 
declined in all states afer 2011, the 
trend for states that had cut potential 
durations (solid blue line) shows a 
steeper fall. 
We fnd that that a state’s potential 
duration of benefts is positively 
related to claims; however, our 
estimates suggest that declining wage 
replacement rates are a stronger 
factor in reducing UI claims. When 
we consider as a group the nine states 
that cut beneft duration, we estimate 
a larger but less precise efect for 
that group. Estimates are imprecise 
because potential duration and wage 
replacement rates tend to move 
together within states. Tat is, states 
Characteristic 
Predicted 
value 
Change 
from 
baseline 
Percent 
change 
Predicted 
value 
Change 
from 
baseline 
Percent 
change 
UI eligibility 227,055 13,955 6.5 220,738 7,639 3.6 
UI generosity 273,509 60,410 28.3 233,068 19,968 9.4 
Part-time employment 166,240 −46,860 −22.0 215,285 2,185 1.0 
New or re-entrants to labor force 196,321 −16,779 −7.9 213,209 109 0.1 
Industry and occupation 261,182 48,082 22.6 196,321 −16,779 −7.9 
Labor force characteristics 267,960 54,860 25.7 219,659 6,559 3.1 
Total from simulations 113,668 53.3 19,682 9.2 
NOTE: Baseline 2019 prediction = 213,100. This number is based on a model described in more detail in the full paper. 
SOURCE: O’Leary et al. (2023), Table 4. 
vigorous eforts by the Federal Reserve Specifcally, three factors—
to raise interest rates to curb infation. 1) reductions in UI generosity and 
Drawing on state-level data access (both in weekly replacement 
from the past three decades, our rates and in maximum weeks of 
decomposition analysis fnds that eligibility), 2) shifs in employment 
declines in UI claims are best from higher-claims manufacturing to 
explained by interactions between lower-claims services, and 
changes in worker and state-UI- 3) demographic shifs of the 
program characteristics and changes population—have all played important 
in how these factors relate to claims. roles. 
Figure 2  Regular UI Initial Claims Indexed to 1990 for the Nine States That Reduced UI 
Benefts after the Financial Crisis Compared with Those That Did Not, 1990–2001 
that cut potential duration also had 1.75 
lower replacement rates. Tus, cuts 
to multiple dimensions of beneft 1.50 
Nine states 
Remaining states 
W.E.UPJOHN 
INSTITUTE 
generosity have played a role in the 
aggregate decline in UI claims. 
Summary 
In the nearly two decades between 
the “dot-com recession” in 2001 and 
the start of the COVID pandemic, 
In
iti
al
 c
la
im
s 
in
de
x 
(1
99
0 
= 
1.
00
) 
1.25 
1.00 
0.75 
average weekly UI claims dropped by 0.50 
about 80,000 per week, making the 
level about one-third less than the 
average of roughly 300,000 during the 0.25 
1990s. In April 2023, weekly UI claims 
were still only 240,000, despite the 
labor force being 20 percent larger than 
at the turn of the century, and despite 
SOURCE: Initial claims data: U.S. Department of Labor, Employment and Training Administration; downloaded from 
https://oui.doleta.gov/unemploy/claimssum.asp. 
3 
 
 
 
 
 
 
 
 
   
 
 
EMPLOYMENT RESEARCH  • JULY 2023 W.E. UPJOHN INSTITUTE 
Why Are Unemployment Insurance
Claims So Low? 
Additional factors may also be 
important. Future research should 
investigate the roles of declining 
unionization rates and the possibly 
growing shif from wage and salary to 
contract employment. 
Note 
1. Nine states permanently cut 
potential durations in 2011 and 2012. 
Te nine states, with their 2019 average 
potential UI durations in parentheses, 
are as follows: Arkansas (17), Florida 
(12), Georgia (14), Idaho (16), Kansas 
(15), Michigan (20), Missouri (20), North 
Carolina (12), and South Carolina (20).  
Many of these states also let weekly beneft 
replacement rates decline since 2012. 
References 
DeAntonio, Dante. 2018. “Why Are 
U.S. Unemployment Insurance Claims 
So Low?” Economic View Report No. 
348474. New York: Moody’s Analytics. 
https://www.economy.com/economicview/ 
analysis/348474. 
O’Leary, Christopher J., Kenneth J. 
Kline, Tomas A. Stengle, and Stephen A. 
Wandner. 2023. “Why Are Unemployment 
Insurance Claims So Low?” Upjohn 
Working Paper No. 2023-383. 
Kalamazoo, MI: W.E. Upjohn Institute for 
Employment Research. 
For additional details, see the full working paper at
http://research.upjohn.org/up_workingpapers/383. 
Christopher J. O’Leary is a senior economist,
Kenneth J. Kline is a senior research analyst, and
Stephen A. Wandner is a research fellow, all at
the Upjohn Institute. Thomas A. Stengle retired as
the lead actuary for the Ofce of Unemployment
Insurance in the U.S. Department of Labor. 
How the Manufacturing
Extension Partnership
Can Anchor U.S.
Workforce Development 
Matthew D. Wilson, Nichola Lowe, Greg Schrock, Rumana Rabbani, and Allison Forbes 
As the U.S. economy rebounds 
from the COVID-19 pandemic, 
strategies that promote long-term 
transformation toward high-quality 
jobs will be critical. If not sufciently 
addressed, long-simmering workforce 
shortages, which intensifed during 
the pandemic, could undermine job 
growth. Manufacturing is a case in 
point. Leading up to the pandemic, 
an estimated 500,000 manufacturing 
job openings went unflled. For 
some legacy manufacturing regions, 
especially those with many older 
businesses that have not modernized 
their technology, difculties with 
worker recruitment and retention are 
especially pronounced. Te “Great 
Resignation” of 2021–2022 has only 
worsened matters, with workers 
voluntarily quitting jobs at historically 
high rates.  Although this phenomenon 
is widespread, it has added to the 
particular challenges facing the 
manufacturing sector. 
In a related paper, we ofer a 
promising institutional fx: centering 
workforce development within the U.S. 
Manufacturing Extension Partnership 
(MEP), a program based at the National 
Institute of Standards and Technology 
(NIST), within the U.S. Department of 
Commerce. In this brief, we summarize 
how this could work by documenting 
the workforce- and workplace-
enhancing strategies that MEP centers 
have adopted since their inception 
in the mid-1990s. While workforce 
development is unevenly implemented 
across today’s MEP network, leading 
centers within the network are devising 
strategies to transform business 
practices to improve the quality of 
frontline manufacturing jobs. Our 
discussions with MEP network leaders 
and center directors point to three 
concurrent approaches that MEP 
centers have adopted to better integrate 
workforce solutions in their service 
delivery models: 
ARTICLE HIGHLIGHTS 
n By coupling workforce services with customized business assistance, the Manufac-
turing Extension Partnership (MEP) provides a promising model to address a range 
of operational and competitive challenges faced by manufacturers.
n Between 2011 and 2019, MEP centers grew their number of workforce projects 
nearly sevenfold, and the share of all MEP projects in workforce jumped from 3 to 12 
percent.
n MEP leaders viewed the COVID-19 pandemic as an opportunity to scale workforce 
development services intended to address preexisting workforce concerns that were 
exacerbated by the pandemic.
n MEP centers have increasingly shifted their emphasis from skill development meant 
to help individual workers find jobs to preparing the workplace to attract, train, and 
nurture the manufacturing workforce. 
4 


Why Are Unemployment Insurance Claims So Low?

Not a member yet