APRIL 2022
Increasing Equity and
Improving Measurement in the
U.S. Unemployment System:
10 Key Insights from the COVID-19 Pandemic
ALEX BELL, THOMAS J. HEDIN, PETER MANNINO,
ROOZBEH MOGHADAM, CARL ROMER, GEOFFREY SCHNORR
AND TILL VON WACHTER
CONTENTS
Summary ..........................................................3
Key Takeaways......................................................4
Changes to Unemployment Insurance During the Pandemic ...........7
Part I: Measuring Equity in Access and Receipt of ...................9
Unemployment Insurance Benets
1. Layos and Long-Term Unemployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2. Unequal Access to Unemployment Benets ........................12
3. Unequal Weekly Benet Amounts ................................16
4. UI Benet Exhaustions and Extended UI Programs ..................18
5. Extended Benets Triggers .......................................24
6. Reemployment and Recall Patterns ................................29
Part II: Insights into the Shortcomings of Current Measures of
Unemployment Benets and How to Improve Them
7. Measurement Issues in Initial Claims Data ..........................34
8. Churn and Repeat Layos in the UI System ........................39
9. Improvements in Counting Continuing Claims ......................42
10. Partial UI and Denials: An Overlooked Measurement ................46
Conclusion........................................................49
References ........................................................52
Figure Appendix ...................................................55
The California Policy Lab builds better lives through data-driven policy. We are an independent, nonpartisan research
institute at the University of California with sites at the Berkeley and Los Angeles campuses.
This research publication reects the views of the authors and not necessarily the views of our funders, our sta, our
advisory board, the California Employment Development Department, or the Regents of the University of California.
2 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
SUMMARY
The COVID-19 pandemic had an immense impact on the labor market in California and
the U.S., with unemployment reaching highs not seen since the Great Depression. The
regular Unemployment Insurance (UI) system, as well as federal legislation that created
supplemental UI programs and benet extensions, played a fundamental role in the
country’s economic and public health response. Individual-level data on who applied for and
received UI benets can provide crucial insights into understanding how the crisis evolved,
how well the government’s response worked, and what the implications are for future crises.
The pandemic also brought to the forefront more fundamental issues with our nation’s UI
system, such as pervasive inequities in which workers actually receive benets and large
dierences in benet amounts and durations. While some of these issues have persisted for
decades, a lack of access to individual-level UI data has prevented a deeper understanding
of the extent of the inequities. The crisis also revealed that even the most fundamental
statistics that policymakers rely on have important aws, and basic information on which
workers benet from some of the core UI programs is completely missing, despite the data
being collected every day by UI agencies as they pay benets.
The California Policy Lab (CPL) partnered with the California Employment Development
Department (EDD), which manages unemployment insurance in California, to help bring
greater clarity to policymakers about the impact of the crisis in California. CPL’s rst analysis
was published in April 2020, and through this unique relationship, CPL was able to use
anonymized claims data to track the labor market crisis in close to real-time and to provide
in-depth, detailed insights on the federal government’s response. Through a series of 19
reports, CPL generated new ndings about a range of issues in the California UI system.
CPL’s research also shed light on which demographic groups and types of workers beneted
the most from the dierent program extensions during the pandemic and which workers
fell through the cracks and were unable to access vital UI benets.
By working directly with complicated claims data for more than two years, CPL developed
new, more accurate and timely measures of how many people were relying on the UI
system as compared to the measures that policymakers and the media typically have relied
on. This work directly demonstrated for state and federal policymakers how the measures
they had historically relied on provided a distorted view of the labor market during the
crisis. For example, a 2020 GAO report cited CPL’s research, nding that published claims
data were likely inated (Government Accountability Oce, 2020).
This report highlights ten key insights about the UI system and the labor
market that are based on CPL’s unique partnership with EDD and the
unique data access this partnership provided. The report focuses on six insights
on equity and disparities in access to unemployment insurance benets during the pandemic.
It then documents four insights on measurement issues in the publicly available data and
how access to California’s administrative records allows CPL to overcome these issues. We
also share how these improved measurements can inform policy choices to improve equity.
3 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
KEY TAKEAWAYS
Part I: Measuring Equity in Access and Receipt of Unemployment
Insurance Benets
1. Workers experienced unequal impacts from the pandemic as
measured by CPL’s new cumulative measures of layos and long-
term unemployment. CPL’s access to UI administrative data allowed us
to, for the rst time, track the impact of the pandemic across demographic
groups using initial claims data. In the rst year of the pandemic, 31% of
Californians in the labor force applied for regular unemployment benets.
Female workers and workers with a high school degree or less had higher
rates of UI claiming than other groups. Of claimants who received benets,
Black, female, and less educated workers claimed UI for longer periods of
time and had higher long-term unemployment rates than other groups.
2. The recipiency rate (which measures how many unemployed
workers are receiving UI benets) increased during the
pandemic. However, there were large disparities across
communities. Publicly available claims data are typically only published
at the state level, which makes it impossible to analyze access to UI across
communities at more detailed geographic levels. Using EDD claims data,
CPL found that California counties with higher incomes and more access
to broadband internet were associated with high recipiency rates, while
areas with a larger share of Hispanic individuals, people with limited English
prociency, and more COVID-19 cases were associated with lower rates.
3. The $600 and $300 UI supplements provided during the
pandemic temporarily helped raise claimants’ incomes above
the “very low” income threshold for single individuals, but
disparities in weekly benet amounts remained. Because of how UI
benets are calculated, males, more educated workers, and White workers
typically receive higher UI benet amounts than other workers. In contrast,
CPL’s research showed that the individuals most impacted by the pandemic
tended to receive smaller weekly UI benet amounts. By measuring benet
amounts with and without these supplements, CPL demonstrated the role
that federal supplements played in raising these workers’ incomes above the
“very low” threshold.
4 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
4. The Pandemic Emergency Unemployment Compensation
(PEUC) program and the Extended Benets (EB) program
helped reduce the number of people who ran out of UI
benets during the crisis and these programs were especially
important for less advantaged workers. There is no publicly available
information on which workers beneted from PEUC or EB programs, and
only rudimentary information on average exhaustion rates, severely limiting
the public’s understanding of the impact of these key programs during the
pandemic. CPL’s research showed that 25% of Californians who applied for UI
benets during the pandemic exhausted their benets, which is lower than the
32% national cohort exhaustion rate estimated during previous recessions.
Over half of all claimants in the PEUC program, a key UI benet extension
program, had a high school diploma or less just before the program expired.
Female workers and younger workers also accounted for a larger share of
claimants receiving PEUC near the end of the program.
5. Due to a aw in how states determine if Extended Benets (EB)
are available, California and 38 other states ended their EB
programs while many workers still relied on them. State Extended
benet programs provide a crucial backstop for long-term unemployed
workers during recessions, and CPL’s research shows that in California it
mostly benets workers with a high school degree or less, female workers,
and Hispanic workers. If Congress reformed automatic benets triggers to
align with CPL’s improved Insured Unemployment Rate trigger (which also
counts claimants in extension programs), these 38 states would have been
able to provide claimants who had exhausted state UI benets with an
additional 13 weeks of benets.
6. Re-employment rates one year after the start of the pandemic
had steadily increased, but remained lower for less-educated
workers than during the pre-pandemic expansion period. In
contrast, a higher share of workers returned to their previous
employer and remained in their previous industry. The
employment outcomes of workers receiving UI are typically not visible to
policymakers, but can provide important insight into the pace of recovery
and needs for intervention across types of workers and labor markets.
Using linked employment data, CPL’s research found that 57% of claimants
who were fully separated from their employer at the beginning of the crisis
(Q2, 2020) were employed one year later, in Q2, 2021. In comparison,
71% of people who claimed UI before the pandemic (in Q4, 2018) were
re-employed within a year. Out of all the re-employed workers, 60%
returned to their previous employer, which was substantially larger than
before the pandemic, when 31% of re-employed workers returned to their
previous employer.
5 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Part II: Insights into the Shortcomings of Current Measures of
Unemployment Benets and How to Improve Them
7. Initial claims are one of the most watched measures of
unemployment, yet CPL’s research found that this measure
dramatically overstates the number of individuals entering the UI
system each week. Between early March 2021 and mid October 2021,
there were 2.7 million initial claims for regular UI in California — but during
that same time period, there were only 1.9 million individuals who entered
the regular UI system. In other words, the typical interpretation of the initial
claims measure overstates the number of individuals claiming regular UI by
42% during this time period. CPL created a new measure called Entries into
Paid Unemployment that more accurately measures the number of people
entering (or re-entering) the UI system.
8. CPL’s new measure of Entries into Paid Unemployment based
on actual payments also claries the total amount of churn in
and out of the system. In early May 2021, we found that the share of
people entering the UI system that were reentries — i.e., claimants who
were beginning their second (or more) period of unemployment— was
over 90%. A second measure of churn shows that in 2021, between 30%
and 40% of all claimants who left the UI system returned after four weeks.
Across demographic groups, White workers, female workers, and older
workers were more likely to suer repeated job loss and multiple periods of
unemployment.
9. Dierences between when unemployment is experienced and
when UI claims are certied can give a distorted view of the
number of individuals receiving UI. Ocial UI statistics count the
number of claims certied each week, not the number of people experiencing
unemployment each week. If individuals retroactively certify for benets or
state agencies are delayed in certifying benets for unemployment experienced
in the past, then the published UI measures could provide a distorted view of
the timing of unemployment. CPL’s new metric measures the number of UI
claims based on when unemployment was actually experienced and provides
a more accurate and timely measure of the labor market.
10. Partial and denied UI payments can shed light on how the UI
system interacts with people who are still connected to their
employer. Through most of the pandemic, the share of UI claimants
receiving partial payments hovered between 9% and 11% and the share being
denied payments due to excess earnings was between 5% and 6%, indicating
that many UI claimants continued to work (albeit at reduced amounts) and
received lower benets as a result. Older workers, more educated workers,
and White workers were more likely to receive partial UI payments and
more likely to be denied payments due to excess earnings.
6 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Changes to Unemployment Insurance During the
Pandemic
At the beginning of the COVID-19 pandemic, federal and state policy makers
introduced a large number of temporary changes to the UI program in response
to the economic disruption caused by the pandemic. For example, nearly all states
suspended work search requirements at the beginning of the pandemic, but states
re-introduced them at dierent times as the pandemic evolved. Many of these
changes motivated CPL’s research during the pandemic and reviewing the changes
provides important context for several of the insights described in this report.
In the CARES Act (2020), federal policymakers introduced the Pandemic
Emergency Unemployment Compensation (PEUC) program that
provided additional weeks of UI to claimants who used all their regular UI benets.
At its maximum, PEUC provided up to 53 weeks of additional benets. Depending
on local labor market conditions a state program called Extended Benets
(EB) provides additional weeks of benets for those that exhausted PEUC
benets (see Section 5).
Federal policymakers also provided supplemental weekly payments of either $300
or $600 to claimants’ normal Weekly Benet Amounts through the Pandemic
Additional Compensation program and the Federal Pandemic
Unemployment Compensation program (which will be described in more
detail in Section 3).
Finally, a new program called Pandemic Unemployment Assistance (PUA)
provided benets for workers who are normally not eligible for regular UI, such as
self-employed workers.
Figure 1 indicates the extent of UI benet use and the importance of extended
benet programs in the pandemic. The role of these benets especially for more
vulnerable workers is analyzed in several sections of this paper. As reected in
the gure, the nal expiration date for all the temporary programs was in early
September 2021.
7 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 1: Total Number of Individuals Paid Benets by Week of Unemployment
Regular UI
(Continuing)
PUA (Expiring)
PEUC (Expiring)
EB (Expiring)
December 26th:
PEUC/PUA Temporarily Expire
September 4th:
PEUC/PUA Expire
September 11th:
Extended Benefits Expire
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
Feb 8
, 2020
Mar 7
Apr 4
May 2
May 30
Jun 27
Jul 25
Aug 22
Sep 19
Oct 17
Nov 14
Dec 12
Jan 9
Feb 6
Mar 6
Apr 3
May 1
May 29
Jun 26
Jul 24
Aug 21
Sep 18
, 2021
Week of Unemployment
Individuals Receiving Benefits
Notes: X-axis labels correspond to Saturdays. Data has been adjusted to account for delays in processing and retroactive claims. Figure includes all available
programs for the period 2/2/20-10/2/21.
While CPL’s and other research indicates these programs have successfully
supported workers in the pandemic, the PUA program has been the subject of
controversy due to high reported levels of fraud. As a result, this report (with a
few exceptions) does not analyze the PUA program. Ongoing work by CPL will
shed light on this program in the future.
8 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Part I: Measuring Equity in Access and Receipt of
Unemployment Insurance Benets
1. Layos and Long-Term Unemployment: How the Crisis Impacted
Demographic Groups and Workers Dierently
Issue. The weekly initial claims data released by the Department of Labor (DOL)
is an important indicator of changes in the labor market. Unfortunately, none of
the publicly available initial claims data include demographic characteristics, which
makes it impossible to use the indicator to measure labor market changes by
demographic group or industry. Access to the UI microdata in CA has allowed
CPL, for the rst time, to measure dierences in UI claims across groups during
the pandemic and we have released this data in our UI reports throughout the
crisis
1
(Bell, Hedin, Moghadam, et al., 2021b). Throughout much of the crisis,
initial UI claims were particularly high for less advantaged groups in the labor
market, including female and less educated workers, and in communities with a
higher rate of poverty and higher shares of Black individuals.
Access to the microdata has an additional advantage over the published DOL
data, which is that we can calculate deduplicated cumulative measures of UI use
with both the initial claims and continuing claims data. This allows us to measure
the total number of workers who applied for UI during the crisis. By comparison,
in the publicly available data, the new initial claims series could provide a close
estimate, but would still suer from some duplication from claimants who le
multiple claims.
2,3
Additionally, the publicly available data does not provide the
new initial claims series with demographic or detailed geographic breakdowns.
4
Analyzing California claims data also allows CPL to measure the number of weeks
of payments each recipient received. The DOL publishes data on the distribution
of the number of payments received by recipients, but it does not provide the
detailed demographic distribution that is available in the California microdata. We
can use these two metrics to measure the total impact of the pandemic across
demographic groups in California.
1 Figures 8A and 8B in the June 2021 UI report show initial claims by demographic groups over the course of the crisis. See: https://www.capolicylab.org/wp-
content/uploads/2021/12/June-30th-Analysis-of-Unemployment-Insurance-Claims-in-California-During-the-COVID-19-Pandemic.pdf
2 This could be the case if the Employment Development Department website experienced crashes from high usage and users resubmit their claims because
they were unsure if their rst claim was received. Alternatively, it could happen if employers submit claims on an employee’s behalf and an employee submits a
claim themselves (Cajner et al., 2020).
3 While the duplication issue will be relatively modest for new initial claims, the duplication with total initial claims will be substantially larger.
4 The DOL does provide demographics for all initial claims, but not the new initial claims, which is what would be needed to account for duplication.
9 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Analysis. Figure 2 shows the rst cumulative measure of UI claiming — the
share of unique people who led for regular UI benets as a share of the
pre-crisis labor force. It shows that through the rst year of the crisis, 31% of
the labor force led for regular UI benets in California. By this measure, less
educated workers have been hit hardest by the pandemic with over half of
workers with a high school degree or less applying for Regular UI benets in the
rst year of the pandemic. The data also illustrate the previously documented
“she-cession” character of the crisis with a higher percentage of female workers
ling for UI benets than male workers (Gupta, 2020).
Racial groups are not included in this gure because of dierences in how racial
and ethnic groups are categorized between the California UI administrative data
and the labor force data from the Current Population Survey (CPS). The CPS
asks about race and ethnicity as separate questions so that respondents can
indicate their race and ethnicity separately. The California UI application asks
about race and ethnicity as a combined question where the racial and ethnic
identities cannot be separated. This means the same individual could identify
dierently between the two sources. Therefore, constructing measures across
these two datasets that assumes individuals would identify the same way in each
source is problematic.
FIGURE 2: Unique Individuals Filing for Regular UI Benets, as a Percentage of Group’s Pre-Crisis Labor Force
Note: This gure shows the number of unique individuals with initial claims for regular UI processed between March 15, 2020, and March, 15, 2021 as a share of
the February 2020 labor force.
10 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Figure 3 illustrates the second measure of the cumulative impact of the pandemic
on UI claimants — the long-term unemployment rate. Overall, about 46% of
UI recipients received over 26 weeks of UI payments during the rst year of
the pandemic. Typically, the long-term unemployment rate is dened as the
share of the unemployed who have been jobless for over 26 weeks (about six
months). CPL’s modied version of the long-term unemployment rate using the
administrative UI data is the share of UI claimants who received at least one
payment who have received more than 26 weeks of unemployment benets
during the rst year of the crisis. The gure also shows that Black claimants, older
claimants, less educated claimants, and female claimants experienced higher rates
of long-term unemployment than other groups over the rst year of the pandemic.
FIGURE 3: The Share of Regular UI Claimants who Received More than 26 weeks of UI Payments in the First Year of
the Crisis.
Note: Long-term unemployment is dened as receiving more than 26 UI payments during the 52 weeks between March 2020 and March 2021.
11 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
2. Access to Unemployment Benets Increased During the Pandemic,
but There Were Substantial Dierences Across Communities
Issue. The process to receive UI is relatively laborious and requires being able
to access, understand, and complete the applications to apply for benets.
Applicants then have to meet certain requirements such as having earned a
minimum amount of income during a base period (prior to being unemployed)
to be approved to receive benets. Finally, they have to verify that they meet
ongoing requirements, such as looking for work each week to receive payments.
The requirements at each stage and other barriers, such as access to an internet
connection, can aect workers’ ability to receive benets that they might be
eligible for.
The recipiency rate measures the rate at which unemployed workers receive
UI benets. It is a summary measure of unemployed workers’ ability to access
benets and can vary over time, by geography, and by demographic group. The
standard recipiency rate is calculated using the publicly available continuing claims
data from the DOL and the unemployed workers estimate from the Current
Population Survey. One issue with the calculation is that continuing claims data
from the DOL measures the number of claims certied each week, not the
number of claims paid for unemployment experienced in a week. Since claims
can be submitted retroactively and there can be administrative delays in certifying
benets, the number of claims paid for unemployment experienced in a week
and the number of claims certied in a week can be very dierent. This could
be particularly true when there are large changes in the labor market like the
beginning of the pandemic.
We improve on this approach by measuring the number of people who received
UI benets for the week in which they actually experienced unemployment
(Hedin et al., 2020b). This allows us to calculate a recipiency rate measure that
provides a more apples-to-apples comparison of the number of people who
received UI benets for unemployment experienced during a week to the total
number of people who were unemployed during the week. Additionally, most
prior measures of the recipiency rate used U3 unemployment, which is the
standard measure of unemployment that includes people actively looking for
work, as the denominator. During the pandemic, work search requirements were
suspended which allowed “discouraged workers” (workers who were looking
for work in the recent past, but stopped because they could not nd any or
felt it was unsafe to work during the pandemic) to qualify for UI benets. This
group is excluded from the U3 unemployed, so using U3 as the denominator
would overstate the recipiency rate. Our measure of the recipiency rate uses
the alternative U6 unemployment rate, which includes discouraged workers and
therefore will not overstate the recipiency rate (Bell, Hedin, Schnoor, et al., 2021).
5
5 For more details of the U3 and U6 unemployment rates, see:: https://www.bls.gov/lau/stalt.htm
12 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Measuring recipiency rates for regions within California is an important but
dicult task. Although we have precise measures of how many Californians
collected benets from a given geographic area, estimating the number of
unemployed workers in a specic region at a specic time is more cumbersome.
In this analysis, we rely on ocial county-level estimates from the Bureau of Labor
Statistics Local Area Unemployment Statistics (US Bureau of Labor Statistics,
2021) for our measure of the unemployed. However, estimating recipiency rates
this way is far from ideal because — due to the small sample size of the Current
Population Survey (108,000 people) — the LAUS estimates for unemployment at
the sub-state level rely on certain measures of UI claims themselves.
6
While we
have contrasted the LAUS county unemployment rates to comparable estimates
based on the CPS microdata and found them to be similar, the fact remains that
for many smaller geographic units the estimates are based on small samples so
are prone to statistical noise. For this reason, the county-level estimates of UI
recipiency rates presented below should be interpreted with caution.
7
Analysis. Figure 4 shows how recipiency rates for Regular UI varied within
California before and during the pandemic.
8
Based on the comparisons of UI
claimants to LAUS unemployment rates (re-scaled to mirror U-6), Los Angeles
County has by far the lowest recipiency rate among large counties in California.
The recipiency rate in California increased steadily from about 20% at the start of
the crisis to almost 90% by the end of 2020. This increase coincided with more
unemployed workers, new federal UI programs, and work requirement rules
being suspended. The gure also demonstrates the large geographic variation in
recipiency rates. Before the crisis, recipiency rates across counties were relatively
similar, but during the crisis, the gaps between counties increased substantially.
Finally, the recipiency rate plummeted back to its pre-crisis level of about 20% as
a result of the federal UI programs expiring in September 2021.
6 For more information, see the LAU.S. methodology note: https://www.bls.gov/lau/laumthd.htm.
7 In our ongoing series of policy briefs, we have compared geographic patterns of recipiency rates using the LAU.S. county-level denition of unemployment to
the tract-level unemployment estimates near the start of the pandemic of (Ghitza & Steitz, 2020). We have not detected meaningful dierences in the spatial
correlations using either measure of unemployment.
8 We do not include claimants on the Pandemic Unemployment Assistance (PUA) program for two reasons. First, high reported levels of fraud in the program
raise questions about the accuracy of the data. Second, we want to compare changes in recipiency from the pre-pandemic period and PUA did not exist
before the pandemic, so excluding PUA makes for a better comparison.
13 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 4: The Share of Unemployed or Underemployed (U-6*) in Each California County Who Are Receiving
Regular UI Benets (Jan. 2019–Sept. 2021)
Figure 5 shows county-level correlations
9
of recipiency rates with socioeconomic
indicators just before the federal UI expirations. It indicates that before the
expirations counties with higher rates of COVID-19 deaths saw lower rates of
recipiency, as did those counties with higher shares of Hispanic residents. We
nd that counties with more broadband access had substantially higher rates of
UI recipiency, which points to the importance of technological gaps in access to
UI during the pandemic. We also nd that counties with more residents with
limited English prociency had lower rates of UI recipiency, which is consistent
with reports that language barriers may also have played a role in limiting access
(Hellerstein, 2020). Many of these correlational ndings corroborate the more
qualitative conclusions (Fields-White et al., 2020) on the role that barriers like
stigma, documentation burdens, and the digital divide have played in widening
racial disparities in access to UI. Although an authoritative dissection of the roots
of these dierences is beyond the scope of the current study, a growing body
of quantitative and qualitative evidence (Fields-White et al., 2020; Gould-Werth,
2016; Shaefer, 2010) suggests that both legal eligibility and more nuanced barriers
to accessibility of UI have played important roles in determining UI recipiency rates.
9 All statistically signicant correlations are signicant at the 95% condence level.
Statewide Average
Los Angeles
County
Recipiency rates jumped suddenly
at the start of the pandemic.
Then steadily increased
throughout the fall.
Topping out at
around 90% by
the end of 2020.
Pandemic UI Programs Expire
0
20
40
60
80
100
120
Jan 2019 May 2019 Sep 2019 Jan 2020 May 2020 Sep 2020 Jan 2021 May 2021
Recipiency Rate in County
Note: Each dot represents the recipiency rate in each month for each of the 58 counties in California. The size of the dot corresponds to the number of U6
unemployed in each county. The line represents the weighted average recipiency rate in California for each month.
14 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 5: Recipiency Rates Within California, County-Level Correlations
Note: Each dot represents the correlation between the county U*6 recipiency rate for August 2021 and county characteristics from the 2019 American Com-
munity Survey. The error bars represent 95% condence intervals.
In our recent report to the US DOL (Bell et al., 2022), we leverage our
experience with California’s claims-level data to construct similar measures of
access in other states during the pandemic. For each state, we used publicly
available data to calculate rates of rst payments, recipiency rates, and exhaustion
rates. Measures of access varied much more across states than across California’s
58 counties, likely reecting the importance of state-level policy dierences.
Although UI has played a critical role in supporting individuals and stabilizing
communities during the pandemic, serious disparities in access exist across the
nation. Policy dierences (such as diering benet amounts, base periods, and
benet durations) across states likely underlie many of these disparities. For
instance, although states with more Black or low-income residents had higher
rates of UI exhaustion during the pandemic, one of the strongest predictors of
a state’s exhaustion rate was the potential benet duration (how long a worker
could claim UI benets) the state extended to claimants. Though more analysis is
needed to isolate the causal eects of particular policies, this descriptive cross-
state analysis suggests there is likely great scope to equalize access to UI in the
country by harmonizing UI policies.
Share Self-Employed
Non-Citizen, percent
Limited English, share aged 5+
Hispanic, percent
COVID Confirmed Cases, % pop
Percent in poverty
Population share aged 20-24
Black, percent
Agricultural employment, percent
Population, ages 65 plus, percent
SNAP recipient, percent
Means of transportation to work, public transit
Retail trade employment, share
Median Household Income
Broadband access of any type, share
-1 -.8 -.6 -.4 -.2 0 .2 .4 .6 .8 1
Correlation With Recipiency Rate
15 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
3. Pre-pandemic labor market inequalities translated to very low UI
benet amounts for more vulnerable workers
Issue. In California most claimants found to be eligible for regular UI are paid
benets equal to 50% of their average weekly earnings in a base period, up to a
maximum of $450 per week. As a result of this benet structure, higher income
workers receive higher absolute weekly benet amounts than lower income
workers. Due to inequalities in labor outcomes across demographic groups, this
also implies that the weekly benets provided by UI will dier across dierent
workers to the detriment of disadvantaged groups. These lower absolute benet
amounts mean that already disadvantaged workers will have a harder time
providing the essentials for their families. The DOL has also recognized the
negative consequences of inadequate benet levels and has made addressing
them a key component of UI reform.
10
Congress mandated that all claimants for regular UI and PUA who were eligible
for benets between March 29th and July 25th, 2020 received an additional $600
weekly benet through the Federal Pandemic Unemployment Compensation
(FPUC) and all claimants for regular UI and PUA eligible for benets between
December 27th, 2020 and September 4th, 2021 received an additional $300
supplement through Pandemic Additional Compensation (PAC). These benet
amounts can be compared to California’s 2020 state income limits, which are
used for eligibility determinations of various government programs.
Analysis. In our UI policy briefs, we have been able to use our unique access
to California UI administrative data to measure how the average weekly benet
amount (WBA) has both changed through the crisis and how it has diered
across demographic groups (Hedin et al., 2020a). The average weekly benet
amount (WBA) for new initial claimants who led for the week ending in
10/16/21 was $330.83 per week. Figure 6 shows how weekly benet amounts
have evolved over the course of the pandemic, and how total benet amounts
(including supplements) relate to the income limits for single person households.
Throughout the whole period, the average WBA without supplements would
have provided an income that was only about half the “Very Low Income”
threshold for a single person household. Due to the $600 supplement and the
later $300 supplement, the average claimant was lifted above the “Very Low
Income” threshold, and for nearly four months at the beginning of the crisis
almost surpassed the “Low Income” threshold. Appendix Figure A1 shows the
evolution of weekly benet amounts across programs for the cohort of claimants
who entered UI in March 2020. On average, PUA recipients had the lowest
and regular UI recipients had the highest benets. Workers receiving extended
benets had lower average benets, especially those that had very long periods of
unemployment..
10 See page 17: https://www.dol.gov/sites/dolgov/les/general/budget/2022/CBJ-2022-V1-07.pdf
16 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 6: Average Weekly Benet Amounts of Initial Claimants for Regular Unemployment Insurance
Notes: Data is from March 1, 2020 to October 16, 2021. X-axis labels correspond to Saturdays. Does not include PUA claimants or transitions to extension
programs. Eligibility for FPUC, LWA, and PAC supplements are determined by the week of unemployment, not the week the initial claim was led
Figure 7 shows how households of dierent demographic groups might fare
under these scenarios. The average WBA for initial claimants over the course
of the pandemic was lower for female workers than it was for male workers.
Lower-educated claimants, younger claimants, claimants in the Retail Trade and
Accommodation & Food Service sectors, and Asian American, Hispanic, and
Black claimants have all seen lower WBAs.
+$600 FPUC Supplement
+$300 LWA
Supplement
+$300 PAC
Supplement
Average WBA (No Supplements)
Single Person HH: Very Low Income Threshold
Single Person HH: Low Income Threshold
0
100
200
300
400
500
600
700
800
900
Mar 14
, 2020
Apr 11
May 9
Jun 6
Jul 4
Aug 1
Aug 29
Sep 26
Oct 24
Nov 21
Dec 19
Jan 16
Feb 13
Mar 13
Apr 10
May 8
Jun 5
Jul 3
Jul 31
Aug 28
Sep 25
Oct 23
, 2021
Processed in Week Ending
1,000
$
17 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
4. Better Measure of UI Benet Exhaustions and Which Workers
Beneted from Extended UI Programs
Issue. Unemployment durations typically increase during recessions because jobs
become more scarce and additional weeks of UI are needed to prevent severe
economic distress among the unemployed. In response, Congress often creates
supplementary UI programs that increase the number of weeks of UI benets
that claimants can receive. Policymakers have to grapple with two questions when
designing these programs: How many additional weeks of benets to provide
and how long to keep the program active. An important input into making these
decisions is to understand who benets from these extended UI programs.
One way of understanding who would benet from these programs is to
measure the rate at which claimants exhaust their benets. These exhaustion
rates are useful in understanding who benets from extended benets programs
because, to a large degree, they reect the degree in which workers were
insured against the length of job loss they experienced (Schmieder et al., 2012). If
exhaustion rates rise during recessions, then a larger number of claimants would
have beneted from programs that increase potential UI durations.
FIGURE 7: Average Weekly Benet Amounts of Initial Claimants for Regular Unemployment Insurance Over the
Course of the COVID-19 Crisis
Notes: Weekly Benet Amounts are the average among all new initial claims (for Regular UI) led between March 15th, 2020 and October 16, 2021. It does not
include new claims led when individuals transition to extension programs. The average benet amount for individuals who either did not self-identify their race
or self-identied as a dierent racial group is not shown.
$405 $705
$285 $585
$281 $581
$267 $567
$359 $659
$313 $613
$309 $609
$334 $634
$321 $621
$311 $611
$292 $592
$344 $644
$298 $598
$320 $620
Single Person HH:
"Very Low Income"
Threshold
0 100 200 300 400 500 600 700 800 900 1,000
Average Weekly Benefit Amount + $300 PAC Supplement
Construction
Education Services
Accommodation & Food Services
Retail Trade
Bachelor's Degree or More
Some College or Associate's
High School Degree or Less
White
Asian American
Hispanic
Black
Male
Female
Statewide
Single Person HH:
'Very Low Income'
Threshold
Single Person HH:
'Low Income'
Threshold
18 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
However, exhaustion rates have proven particularly dicult to measure, especially
in the publicly available data from the DOL. Whereas the term “exhaustion” has
at times been used to refer to claimants who exhausted their (regular non-
extension) state UI benets and moved on to extension programs, in this report
we dene exhaustions as those cases in which a claimant has exhausted state UI
benets as well as all available extensions. This is a more meaningful measure for
understanding which groups of workers would have beneted from additional
benet extensions. The data supplied by DOL indicate only the number of
claimants who have exhausted particular UI programs, and so are not well-suited
for research on exhaustions during periods of benet extensions, particularly
when extension programs were abruptly modied.
A second way of understanding who benets from extended benet programs is
to measure which groups of workers received payments through the programs.
Early in the pandemic, Congress created two new UI programs that extended
the duration of benets for regular UI claimants and also expanded eligibility
to include workers who did not traditionally qualify for UI. The Pandemic
Emergency Unemployment Compensation (PEUC) program added additional
weeks of UI for workers who exhausted their regular UI benets and the
Pandemic Unemployment Assistance (PUA) program expanded UI access to
workers who did not qualify for regular UI benets. Along with these federal
programs, each state has their own Extended Benet (EB) program that provides
additional weeks of UI during periods of high unemployment. While the DOL
provides data on the number of claimants receiving benets through these
programs, it does not provide a demographic breakdown of beneciaries.
Analysis. First, we illustrate how the UI microdata can improve on the
exhaustion measures available in the DOL data. We do this by counting
exhaustions as the co-occurrence of two separate events. The rst event is that a
worker received a nal payment for a particular UI program, and the second is that
another payment does not follow within four weeks.
11
This combination of events
ensures that we are measuring people who received their last payments and then
did not transition to any other programs, overcoming the issue in the publicly
available data where claimants could have transitioned to another program.
12
11 We are considering the expiration of the PEUC and EB programs to be the nal payment for those programs and thus count as exhaustions.
12 The one drawback of this method is that not all claimants who are eligible for the state Extended Benets program are aware that they could actually
transition onto it. This would count as an exhaustion using this method even though the claimant is actually eligible for additional weeks.
19 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
In the California Policy Lab’s recent report to DOL, we put forward two distinct
measures of exhaustion rates (Bell et al., 2022). To mirror the denition of
exhaustion rates that can be operationalized in the DOL data, we rst divide
the number of claimants who exhausted UI in a given week by the total number
of claimants who certied that week. Conceptually, this ratio is dicult to
interpret.
13
A more readily interpretable statistic is what we call the ‘cohort
exhaustion rate,’ dened as the share of UI entrants in a given week who will
eventually exhaust UI.
14
Our cohort measure tends to be higher than the
traditional measure. A potential drawback is that it cannot be implemented
nationally with available data.
Figure 8 plots how these two denitions of exhaustion rates have evolved in
California during the pandemic. Before the pandemic, the share of California’s
claimants exhausting each week was typically 2–3% and then fell below 1% during
most of the pandemic when the extension programs were active (Panel A). This
initial drop in the exhaustion is driven mostly by claimants entering the UI system
at the beginning of the pandemic. Then in early September 2021, the pandemic-
era expanded unemployment insurance benets (PUA, PEUC, and EB) turned o,
at which point 51.2% of claimants exhausted during the week of September 11th.
A dierent story emerges when analyzing exhaustees as a share of the weekly
entry cohort (Panel B). Among Californians who entered UI at the beginning
of the pandemic, 24% of these claimants exhausted benets as of the end of
October, 2021. This is somewhat lower than what prior literature had found
during past recessions. Nicholson and Needels (2006) look at cohort exhaustion
rates during recession years between 1970 and 2003. They show that the
(national) exhaustion rate for the early 2000s recession was on average 32%. This
might suggest that the extension programs during the pandemic have done a
better job of insuring claimants against long unemployment spells than programs
in previous recessions. It could also reect a dierence between numbers for
California and U.S nationwide estimates. Figure A2 in the appendix compares the
traditional exhaustion rate for Regular UI only to the new exhaustion rate for
all programs. Without the extensions the exhaustion rate for the March 2020
cohort would be nearly 50% — about twice the actual rate.
13 Although each claimant can count at most once in the numerator (during the week of exhaustion), the same individual would count toward the denominator
for multiple weeks (during each week claimed).
14 Because this statistic counts each claimant exactly once in the denominator (during the week of entry), it is more accurate.
20 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Notes: The blue line shows
the share of all claimants
who entered UI each week
and who ultimately received
all the benets they were
eligible for before and during
the pandemic. The gray
dashed line shows the max-
imum number of weeks of
UI each cohort had available
to them. The maximum
number of weeks available
during this period was 92.
All claimants qualied for
the Regular 26 weeks of UI.
Claimants whose benet
year began between sum-
mer 2018 and September
4h 2021 were eligible for
PEUC. Claimants whose
benet year began after May
1st, 2019 were eligible for
Extended Benets.
FIGURE 8: Exhaustion Rates Within California, Weekly Resolution, Jan. 2019–Oct. 2021
Notes: The Y-axis is log
scale to improve readability.
Graph shows the share of
all claimants who received
their nal UI payment each
week before and during the
pandemic.
PANEL B: NUMBER OF CLAIMANTS EXHAUSTING AS A SHARE OF WEEKLY ENTRY COHORT
PANEL A: NUMBER OF CLAIMANTS IN CALIFORNIA EXHAUSTING AS A SHARE OF WEEKLY CONTINUING CLAIMANTS
21 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
An important but understudied question is understanding which demographic
groups and what types of workers benet from extension benet programs. One
way to address this question is to measure the share of all PEUC or EB claimants
that each demographic group accounts for. The second is to measure the share
of all exhaustions that each demographic group accounts for. Neither of these
measures are available in the publicly available DOL data, but we can answer
these questions for California using our access to the UI administrative data.
Figure 9 shows the share of PEUC or EB claimants that each demographic group
accounted for just before the program expirations in September 2021. Overall it
shows that claimants with a high school degree or less, millennials, and Hispanic
claimants made up the largest share of PEUC or EB claimants and thus beneted
the most from the PEUC and EB program in California. Appendix Figure A3
shows the tight relationship between the exhaustion share and the extended
benets program share. Either measure is very informative of who benets from
extension programs, but the data in Figures 9 and 10 are in principle more readily
available.
FIGURE 9: The Share of PEUC or EB Claimants by Demographics Group as of 9/4/2021
Notes: The blue bars show the share of all PEUC and EB claimants that each demographic group accounted for during the week of 9/4/2021
22 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Figure 10 shows the industry composition of PEUC and EB claimants.
Accommodation and Food, Retail Trade, Administrative Services, and Healthcare
& Social Assistance made up the largest shares of all extension claimants. These
industries often employ less educated workers, which is consistent with the
previous nding that most PEUC or EB claimants were workers with a high
school degree or less.
Additionally, along with the composition of workers that would benet from
extended benets programs, we can measure the rate of exhaustion within
demographic groups. Appendix Figure A4 reveals the exhaustion rates by
demographic group. Black claimants, older claimants, and less educated claimants
had higher exhaustion rates than other types of claimants.
FIGURE 10: The Share of PEUC or EB Claimants by Industry as of 9/4/2021
Notes: The blue bars show the share of all PEUC and EB claimants that each industry accounted for during the week of 9/4/2021.
23 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Overall, the exhaustion rate analysis shows that the pandemic UI programs did a
reasonably eective job of insuring workers against their increased unemployment
periods as the cohort exhaustion rate during the pandemic recession was lower
than in previous recessions and was either the same as or lower than prior to
the crisis. The programs were also eective in supporting some, albeit not all, of
the most disadvantaged workers.
To aid in further research on measuring exhaustion rates nationally, we suggest
the DOL collect data across all states on claimants who have exhausted
all UI programs, not just particular programs. Publishing this information
by demographic group would also aid in analyzing disparities in access to
UI. Furthermore, DOL should also publish the demographic and industry
breakdowns of workers receiving extended benet programs, something not
currently available.
5. How an Incomplete Measure of Unemployment Led Many States to
“Turn O” their Extended Benets Programs Prematurely During
the Pandemic
Issue. During the COVID-19 crisis, two forms of benets extensions allowed
unemployed workers in California and other states to claim unemployment
beyond the standard 26 weeks. A special extension program created in the
CARES Act of 2020, PEUC, originally provided 13 weeks (which Congress
increased to 24 on December 27th, 2020, and then to 53 on March 11th,
2021).
15
Claimants who rst exhaust their 26 weeks of regular benets, then their
13–24 weeks under PEUC, have been eligible for an additional 13 to 20 weeks
of unemployment insurance benets, called Extended Benets (EB).
16,17
The EB
program is designed to automatically “trigger” on in each state during times of
unusually high unemployment in that state, when additional relief and stimulus is
most needed, and to trigger o when the labor market has recovered.
15 https://edd.ca.gov/unemployment/pdf/unemployment-benets-chart.pdf
16 Not all claimants who are eligible for PEUC benets meet the monetary eligibility restrictions for receiving Federal-State Extended Benets.
17 Due to the timing of the roll-out of additional PEUC extensions, in certain cases claimants exhausting EB may be eligible to move back to PEUC.
24 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
There are two types of economic triggers that can cause a state’s Extended
Benets period to turn on or o (summarized graphically in Figure 11). The rst
trigger, used in all states, is the Insured Unemployment Rate (IUR) trigger. The
IUR is a measure of the share of the covered labor force that is currently claiming
regular unemployment benets. Importantly, the IUR does not count
claimants receiving benets under an extension program. The IUR
was changed in 1980 to no longer count claimants on EB, as part of a package
of changes to EB that some scholars have suggested “very nearly disabled the
program” (Woodbury, 1996).
18
In all states, when at least 5% of the state’s labor
force is claiming regular benets, EB is triggered on, provided that IUR in the
prior years was suciently low. The latter provision is called the “IUR lookback”
provision.
19,20
When the share falls back below 5% (or the lookback provision
no longer holds), the IUR trigger turns o, and Extended Benets expire.
21
In
recessions when unemployment remains elevated, the lookback provision is often
hard to satisfy.
FIGURE 11: Pathways for EB to Trigger on
Notes: This graphic depicts the ways in which an EB period can trigger on for a state. Possible scenarios in which IUR and TUR do not satisfy the EB trigger
conditions are not illustrated.
18 See also page 10 here on the Omnibus Reconciliation Act of 1980: https://fas.org/sgp/crs/misc/RL34340.pdf
19 The 5% IUR trigger is satised if the IUR is at least 120% of the IUR in the same period of the prior two years, averaged over the two years. Accounts by
researchers have suggested that the lookback provision, which references the state’s data in the same seasonal period of prior years, arose “primarily due to
the costliness and diculty of seasonally adjusting the weekly claims data for each state” (Wenger & Walters, 2006).
20 In addition, either IUR trigger requires that “[s]tate extended benet payments have not been paid for at least 12 weeks including the current report week, so
that there will have been at least 13 weeks of nonpayment before the week benet payments begin.” See page 10: https://wdr.doleta.gov/directives/attach/
ETAH/ETHand401_4th_s01.pdf
21 Typically, there is a lag of a few weeks between when the IUR trigger crosses the threshold and the last week of unemployment for which EB can be paid.
EB is off, and has not
been on during last
12 weeks
IUR>=5.00%
State has 6% non-
lookback IUR
trigger law
State has TUR
trigger law
IUR>=120% of
prior year’s on
same day
IUR>=6.00%
TUR>=110% of
either of last 2
years’ on same day
TUR>=6.50%
TUR>=8.00%
13 Weeks of EB Benefits 20 Weeks of EB Benefits
IUR: 13-week
average of the share
of the covered labor
force claiming
regular non-
extension UI
TUR: 3-month
average
unemployment rate
as estimated by the
Bureau of Labor
Statistics
25 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
The second type of EB trigger is the Total Unemployment Rate (TUR) trigger.
The TUR is derived from monthly surveys of the number of unemployed people
conducted by the Current Population Survey (CPS) run by the Bureau of Labor
Statistics. The TUR trigger is optional and before the crisis only 10 states had it.
By a year into the pandemic, half of all states had adopted it,
22
though the number
of states with TUR triggers has since fallen.
23
If a state adopts a TUR law, there is
only one system of thresholds that it can adopt. The TUR trigger has three levels:
O, TUR Light, and TUR Heavy.
24
A state can move from “O” to “TUR Light”
when TUR reaches at least 6.5%, and can move to “TUR Heavy” when TUR rises
above 8%. Both TUR Light and TUR Heavy turn Extended Benets “on,” with the
dierence between the two being that TUR Light provides 13 weeks of extended
benets (the same amount as the IUR trigger), whereas TUR Heavy provides
20 weeks of benets.
25
Because TUR Heavy provides more weeks of benets, it
supersedes TUR Light or either IUR based trigger in instances when multiple
triggers are satised.
Analysis. During the COVID-19 crisis a large share of unemployed individuals
(particularly those who began claiming benets at the beginning of the crisis)
have remained unemployed long-term, and thus have transitioned from Regular
UI programs to extension programs.
26
As a result, even as regular UI claims
captured by IUR started to fall, a large number of individuals were still depending
on EB. Minnesota provides a clear example of this issue. Figure 12 shows that EB
triggered o in December 2020 because the ocial IUR (not counting claimants
on extension programs) fell below the IUR threshold. However, if claimants who
had transitioned from receiving benets through regular UI to receiving benets
through an extension program were included in the IUR, then the EB program
would have remained on for an additional 9 months (Bell, Hedin, Moghadam, et
al., 2021a).
22 See this article for list of all states who adopted the TUR trigger: https://www.peoplespolicyproject.org/2021/03/09/states-should-activate-extended-benet-
triggers/
23 https://oui.doleta.gov/unemploy/trigger/2022/trig_011622.html
24 In some states, the period we refer to as TUR Heavy is termed High Unemployment Period, or HUP.
25 Both TUR triggers also have “lookback” provisions stipulating the TUR must exceed 110% of either of the prior two years. During most of the Great
Recession period, the lookback period was extended to 3 years. For further information on TUR triggers, see Appendix 2 here https://georeyschnorr.com/
pdf/ui_worker_productivity_oct2020.pdf as well as https://oui.doleta.gov/unemploy/pdf/uilaws_extended.pdf. The requirement that EB cannot trigger on
again if it was recently triggered on and o also applies to the TUR trigger.
26 CPL’s March report found more than half of all CA UI claimants during the crisis had claimed more than 6 months of benets. See: https://www.capolicylab.
org/wp-content/uploads/2021/03/March-18th-Analysis-of-CA-UI-Claims-during-the-COVID-19-Pandemic.pdf
26 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Claimants receiving EB have to rst exhaust both their regular UI and PEUC
benets and would therefore be among the workers who have spent the most
time unemployed. This current policy design is most harmful to the people who
have struggled the most to nd employment during the pandemic. Appendix
Figure A5 shows the demographic groups that made up the EB program in
California and thus reveals the workers that are the most impacted by the
premature trigger turn o. It shows that already disadvantaged groups such as
workers with a high school degree or less, females, and Hispanic workers make
up a larger share of EB claimants than other groups and are the most impacted by
prematurely triggering o the EB program.
Figure 13 below shows a map of all the states that were impacted by the awed
IUR trigger. EB would have remained triggered on longer for all the states colored
in red. Table A1 in the appendix shows the number of weeks that each state lost
in EB benets due to the design and shows the number of claimants on EB leading
up to the ocial turn o date. Overall, 39 states (plus the District of Columbia
and Puerto Rico) ended their EB programs earlier than they would have if their
version of IUR had instead included extension claimants. Workers in California
lost out on 15 weeks of EB due to the awed IUR measure.
Official IUR
(Excludes Extensions)
Replicated IUR
(Includes Extensions)
EB Weeks Claimed
(Right axis)
IUR Threshold
Official EB Turn-Off:
December 19, 2020
Official PEUC Turn-Off:
September 4, 2021
If the IUR measure included
claimants on extension programs,
EB would have remained
available for another
9 months in MN.
0
5,000
10,000
15,000
20,000
25,000
0
5
10
15
Feb 1, 2020
Feb 25
Mar 20
Apr 13
May 7
May 31
Jun 24
Jul 18
Aug 11
Sep 4
Sep 28
Oct 22
Nov 15
Dec 9
Jan 2
Jan 26
Feb 19
Mar 15
Apr 8
May 2
May 26
Jun 19
Jul 13
Aug 6
Aug 30
Sep 23
Oct 17
Nov 10, 2021
|
FIGURE 12: Extended Benets Triggers O Early in Minnesota (01/26/20-11/10/21)
Notes: This gure shows the divergence between Minnesota’s ocial Insured Unemployment Rate (IUR) and that which includes workers reporting unemploy-
ment but paid through extensions. Had the IUR counted extensions, it would not yet have fallen below the 5% threshold used to determine the state’s EB period
until October 2021. The Ocial IUR (dark blue) is as reported in Weekly Trigger Reports. The Replicated IUR with Extensions (orange) is based on authors’
calculations to additionally include PEUC and EB weeks claimed. EB Weeks Claimed (light blue, right axis) is reported on AR539 report from DOL, and indicates
the number of weeks of benets that were claimed that week to be paid under the EB program.
27 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 13: States Where Extended Benets Would Have Remained on Longer If Claimants on Extension Programs
Were Counted Toward a State’s IUR Trigger
Notes: The red states in the map had their EB benets turned o early due to the design of the IUR trigger.
To include all workers who are relying on UI benets (including extended
benets), Congress could pass legislation to include extension claimants in
IUR. Such a reform would update the EB program to be more eective at
automatically providing additional benets in times of elevated unemployment.
■
Aected
D
Not Aected
•
•
.
, .
28 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
6. Reemployment Patterns and Recall to Former Employers has been
Uneven During the Recovery
Issue. In March 2020, as the COVID-19 pandemic hit California, a wave of new
public health measures closed down restaurants, retail stores, and other outlets
and caused the unemployment rate to rise higher than at any point since the
Great Depression. In Q2 2020, over 2 million workers entered California’s UI
system and were fully separated from their employers (meaning they did not
report any earnings in their rst week of certifying for benets). Using earnings
data from the EDD, CPL measured the quarterly employment outcomes of all
workers covered by UI (Bell, Hedin, Mannino, et al., 2021). Using this data, we
followed the cohort of two million individuals who were unemployed at the
beginning of the pandemic to see how they fared in the labor market a year later
(in Q2 2021).
Analysis. Figure 14 shows that of the two million claimants who entered the UI
system during the second quarter of 2020 and who were “fully separated” from
their employers (meaning they did not report any earnings in their rst week
of certifying for benets), 57% were employed four quarters later in Q2 2021,
and 34% had been recalled to a prior employer. Among those that had found
reemployment, the percent recalled to a prior employer was 60%. Compared
to before the pandemic, reemployment after four quarters is substantially lower
but the rate of recall conditional on reemployment is slightly higher. For claimants
who entered the UI system in Q4 2018, 71% were reemployed four quarters later
in Q4 2019 and 31% had been recalled to a prior employer. Among those who
were reemployed only 44% were recalled to a previous employer. Recall played
a much larger role in overall reemployment during the pandemic than during
normal times. Overall, the relatively high rate of recall suggests that workers and
rms have maintained some ties during the job separations of the pandemic.
Figure 15 shows how, to some extent, the high rate of recall during the pandemic
was expected by claimants. At the beginning of the pandemic, the share of new
initial claimants reporting that they expected to be recalled to their previous
employer increased substantially and remained elevated through most of the
pandemic. The increase and gradual decrease in claimants expecting recall is
similar to results found using temporary layos from the CPS to measure recall
expectations (Bartik et al., 2020). This rate varied moderately by race and
ethnicity with Hispanic and Asian American claimants expecting recall at higher
rates than other groups. Reemployment rates varied only moderately between
those expecting recall and others (60% vs. 50%). The nal column in Appendix
Table A2 shows how these expectations may have been too optimistic with 40%
of claimants who expected to be recalled actually being recalled (65% among
those reemployed). In contrast, among those not expecting recall, only 20%
returned to their previous employer (39% among those reemployed).
29 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 14: Recall and Re-employment Rates by Demographic Groups
Notes: The denominator for each bar includes all regular UI claimants from that demographic group who led a UI claim during the 2nd quarter of 2020. The
numerator consists of the subset of those claimants who had found any type of employment a year later, and those who became re-employed by their separating
employer.
While 34% of all claimants had been recalled in Q2 2021, substantial
heterogeneity exists across demographic groups, industries, and geographies.
Figure 14 contrasts how our measures of recall and re-employment vary across
demographic groups. Although older workers have had among the lowest rates
of reemployment, their likelihood of recall was among the highest, suggesting
that recall has been more common among longer-tenured workers. Female,
more educated workers, and workers identifying as Hispanic have each seen
relatively high rates of both recall and overall re-employment. These patterns
for older workers and Hispanic workers are also observed before the pandemic,
but prior to the pandemic, workers with a high school degree or less and male
workers had the highest rates of both reemployment and recall. CPL’s December
2021 UI report also examined claimant transitions across industries and found
that a higher share of claimants returned to their previous industry during the
pandemic than they did before the pandemic (Bell, Hedin, Mannino, et al., 2021).
The increased rate at which workers returned to their previous industry is
likely driven, in part, by the increase in workers being recalled to their previous
employers as described above. Appendix Figure A6 shows how this rate of
return varied by demographic group.
30 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 15: Recall Expectations Among New Initial Claimants
Notes: X-axis Labels Correspond to Saturdays. This gure excludes PUA claims. New initial claims exclude additional and transitional claims
Figure 16 compares rates of recall and re-employment across industries. Workers
in the Education, Manufacturing, Arts & Entertainment, Health Care and Social
Assistance, Management, and Mining, Oil, and Gas industries have all seen rates
of recall of at least 45%. Education, Management, and Mining, Oil, and Gas also
had the highest overall reemployment rates at over 70%. Information also had
a reemployment rate of 70% even though it had a lower recall rate. The three
industries with the lowest rates of both re-employment (<= 55 %) and recall (<=
23%) are Finance and Insurance, Administrative Support and Waste Management,
and Agriculture, Forestry, and Fishing.
Before the pandemic, workers from the Construction industry and Transportation
& Warehousing industries had some of the highest reemployment rates, but
during the pandemic workers from those industries saw much lower rates of
reemployment relative to other industries. Conversely, workers from the Mining,
Oil, and Gas industries had some of the highest reemployment rates during the
pandemic, but had relatively lower reemployment rates before the pandemic.
31 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 16: Recall and Re-employment Rates by Major Industry
20% 43%
23% 53%
23% 55%
36% 58%
34% 58%
31% 60%
32% 61%
38% 61%
37% 62%
40% 65%
42% 65%
46% 66%
43% 67%
44% 67%
47% 68%
49% 69%
43% 70%
45% 70%
46% 70%
53% 71%
0 20 40 60 80
Percent of Claimants Who Experienced Full Separations
Agriculture, Forestry, Fishing (a)
Admin. Support, Waste Man. (a)
Finance and Insurance
Other Services
Real Estate and Leasing
Accommodation Food Svc
Prof., Scientific, Techn. Services (a)
Transp and Warehousing
Construction
Retail Trade
Wholesale Trade
Health Care and Social Assistance
Utilities
Public Admin
Arts, Entertainment, Recreation
Manufacturing
Information
Mining, Oil and Gas
Management
Education Services
Percent Recalled Percent Re-Employed (By any Employer)
Notes: The denominator for each bar includes all regular UI claimants from that industry who led a UI claim during the 2nd quarter of 2020 and who were
“fully separated” from their employer (i.e., not just working reduced hours). The numerator consists of the subset of those claimants who had found any type of
employment a year later, and those who became re-employed by their separating employer. Graph excludes claimants whose industry could not be identied.
a) Full Names of Sectors: Administrative Support, Waste Management, and Remediation. Agriculture, Forestry, Fishing, and Hunting. Professional, Scientic, and
Technical Services
As noted previously, a higher share of re-employed workers returned to the
previous industry during the pandemic than before it. Figure 17 reveals which
industries re-employed workers transitioned to when they did not return to
their previous employers. During the pandemic, Admin, Support, and Waste
Management, Retail Trade, and Healthcare & Social Assistance were the industries
that received the highest percent of transitioning workers. Compared to the
pre-pandemic period, Retail Trade, and Healthcare and Social Assistance became
more important destination industries for transitioning workers, while Admin,
Support, and Waste Management became less important. Admin, Support, and
Waste Management includes jobs like janitors, oce clerks, and security guards
that might typically be employed in corporate oces (or retail settings), but the
closing of oces and the increasing prevalence of work from home arrangements
could account for the industry becoming a less popular (though still important)
destination. Construction, which normally contracts during recessions, was also a
less popular destination for transitioning workers during the pandemic.
32 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 17: Industries that Claimants Transitioned to Before and During the Pandemic
Notes: The denominator for each bar includes all regular UI claimants who led a UI claim during the 2nd quarter of 2020 (pandemic) or the 4th quarter of 2018
(pre-pandemic) and were re-employed a year later in a dierent industry. The numerator consists of the subset of those claimants who were employed in that industry.
Geographically, we nd that more auent counties had higher rates of
re-employment. Appendix Figure A7 presents spatial correlations of these
county-level re-employment rates with a variety of socioeconomic factors.
Some of the strongest county-level predictors of re-employment are measures
of economic well-being, including poverty, median household income, fraction
receiving SNAP/CalFresh benets, and access to broadband internet. We also
see the importance of measures of urbanicity, namely the share of workers who
take public transit to work. CPL’s December 2021 report found that before
the pandemic, the fraction receiving SNAP/CalFresh benets was still highly
correlated with reemployment, but median household income, broadband access,
and public transportation were not. On the other hand, the share of employment
in agriculture, the percent non-citizen, and the percent of people with limited
English prociency were more highly correlated with reemployment rates before
the pandemic.
Taken together, these results suggest that by mid-2021, jobless workers had
continued to maintain some ties with their former employers, but pandemic
reemployment rates are lower than pre-pandemic rates.
33 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Part II: Insights into the Shortcomings of Current
Measures of Unemployment Benets and How to
Improve Them
7. Initial Claims Dramatically Overstated the Number of Workers
Entering the UI System During the Pandemic
FIGURE 18: Weekly Initial UI Claims (including PUA) During the COVID-19 Crisis in California. Initial Claims for
regular UI include New Initial and Additional Claims.
Issue. Initial claims are an important leading indicator of changes in the labor
market, however the weekly published numbers provide a misleading view of the
number of newly unemployed workers. For example, despite the gradual decline
in initial claims both in California and nationwide during the course of 2021, initial
claims have not fallen by as much as one might expect, and at the end of August
2021 were still near the peak of the Great Recession. Initial claims for regular
UI are composed of two major components, as illustrated in Figure 18: New
initial claims and additional claims. It shows that the high level of initial claims in
California has been driven by a persistently elevated number of additional claims
processed in each week, while new initial claims have dropped substantially and
are now in line with the levels seen prior to the pandemic.
Notes: X-axis labels correspond to Saturdays. New Initial Regular Claims includes new initial claims for regular state UI. Additional Regular Claims
includes additional claims for regular state UI and additional claims for extension programs. (DOL does not include additional claims for claimants on
extension programs in their initial claims numbers, only additional claims for regular UI.) This gure does not include transitional claims, as DOL does
not include them in their headline initial claim number nor do they represent ows into the UI system.
34 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Other researchers have examined the impact of the Pandemic Unemployment
Assistance program on the elevated level of claims (Cajner et al., 2020; Price,
2021). Here and in our June 2021 report, we explain the measurement issues
that can cause initial claims to overstate the number of people entering the
unemployment system, and later we introduce a new metric that overcomes
these measurement issues (Bell, Hedin, Moghadam, et al., 2021b). To start, we
rst explain what triggers each type of claim. New initial claims correspond to
“an application for the establishment of a benet year and a computation of the
maximum benets payable and the weekly rate.” Everyone attempting to collect
unemployment benets must le a new initial claim. Additional claims are a subset
of initial claims dened by EDD as a claim which is re-opened after a “break of
one or more weeks in the claims series with intervening employment.”
27
Both new initial claims and additional claims overstate the amount of people
entering (or re-entering) the UI system, but the ination to additional claims
is much larger. New initial claims can overstate entries into the UI system if
individuals le multiple claims within a single week, if claimants le initial claims
which are denied, or if claimants led initial claims which are accepted, but the
claimant never certies for benets, perhaps because they found a job (after
a person les a claim, they then need to certify every two weeks in order
to receive benets). While this denition may seem to imply that additional
claims represent individuals beginning a second (or third, or fourth) period of
unemployment, this is not necessarily the case. Additional claims can also be
triggered by a claimant who works reduced hours for multiple weeks in a row,
certies for partial UI benets, and is either denied payment on one or more
of these certications because their (reduced) earnings are above the partial UI
threshold. In addition, an additional claim can be triggered if the claimant delays
one or more certications (even without a denied payment). These “breaks in the
claim cycle” trigger additional claims because the partial UI claimant has ongoing
employment. However, because these individuals are not actually exiting and
re-entering the UI system each week, the additional claims measure will overstate
the number of people owing into the UI system.
27 For denitions of new initial claims and additional claims, see: https://www.edd.ca.gov/uibdg/Miscellaneous_MI_5.htm. There are also transitional claims which
are led when a claimants benet year expires, they worked enough hours to qualify for a new benet year, and they immediately le a claim with no breaks
between the weeks (https://www.edd.ca.gov/uibdg/Miscellaneous_MI_20.htm).
35 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Because additional claims were responsible for 59% of initial claims for regular
UI in the 12 months ending October 16, 2021, the degree by which initial claims
numbers overstate entry (or re-entry) has been substantial during the COVID-
19 crisis. Moreover, the number of additional claims each week is directly tied to
the total number of continuing claimants. Since the total number of continuing
claimants declined very gradually over the course of the COVID-19 crisis, it is
not surprising that the level of additional claims remained elevated. An important
implication of this discussion is that a large recession with a high number of UI
recipients will partly mechanically have elevated initial claims. If (or when) the
number of continuing claimants is reduced, either due to increased job nding
rates or due to administrative reasons (such as benet exhaustions), this will
simultaneously decrease the number of additional claims, and therefore the total
number of initial claims.
Alternative Measure. To better understand how well initial claims correspond
to new people claiming unemployment, CPL has created a measure analogous
to initial claims, which measures the actual ow of individuals into the UI system:
Entries into Paid Unemployment (Bell, Hedin, Moghadam, et al., 2021b).
This measure is dened as the number of individuals entering a new period of
paid unemployment each week. We dene a claimant as entering a new period
of unemployment if he or she experiences a gap of one or more weeks between
two weeks of compensated unemployment or if they entered UI for the rst
time since January 2020.
If a claimant certies retroactively, the claimant is determined to have entered or
reentered the UI system in the week in which his or her unemployment actually
began, regardless of any administrative irregularities that may have aected the
claim processing date or the date in which they certied for benets. Figure 19
compares CPL’s new measure, Entries into Paid Unemployment, with the
traditional measure of initial claims for regular UI.
36 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 19: Initial UI Claims Substantially Over-State Entry into the UI System
Notes: Bars are not stacked, but instead compare the dierent measures within the same week. The initial claims bars do not include initial claims for PUA, new
initial claims for extension programs, or transitional claims. They only include new initial claims for regular UI, additional claims for regular UI, and additional claims
for claimants on extension programs.
Figure 19 shows that initial claims substantially overstated the number of entries
into paid unemployment during the crisis. From the start of March 2021 to Oct
2021, there were 2,696,793 initial claims for regular UI (blue bars). During that
same time period, there were just 1,894,809 individuals who entered the regular
UI system (orange bars). This shows that the standard initial claims measure
overstates the number of individuals entering regular UI by 42%. Our June 2021
report provides evidence that the gap between initial claims and entries into the
UI system is driven largely by additional claims that do not result in a payment
(Bell et al. 2021). Appendix Figure A9 replicates Figure 18 for 2019, and shows
that initial claims overstated entries into the UI system by 45%, about as much
as in 2021. This indicates that the ination to initial claims is not just a pandemic
phenomenon.
Additionally, it is possible that the relationship between initial claims and people
actually re-entering paid unemployment could dier by demographic group. For
some groups, initial claims might substantially overstate the number of people
re-entering UI, while for other groups initial claims and new entries into paid
unemployment might be closer. Appendix Figure A10 shows the amount by
which initial claims overstate entry into paid unemployment by demographic
group. The standard interpretation of the initial claims data overstates entry
by younger workers, more educated workers, and Black and Hispanic workers
relative to other workers.
Entries into
Paid Unemployment
Initial Claims
De-Duplicated
Initial Claims
0
200,000
400,000
600,000
800,000
1,000,000
1,200,000
Jan 25
, 2020
Feb 22
Mar 21
Apr 18
May 16
Jun 13
Jul 11
Aug 8
Sep 5
Oct 3
Oct 31
Nov 28
Dec 26
Jan 23
Feb 20
Mar 20
Apr 17
May 15
Jun 12
Jul 10
Aug 7
Sep 4
Oct 2
, 2021
Week Initial Claim Processed / First Paid Week of Unemployment
37 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Figure 20 shows CPL’s new measure of entries into the UI system, Entries into
Paid Unemployment, broken down by the number of the UI period. There
are key dierences relative to the analogous version of initial claims (Figure 18).
As already discussed in Figure 19, the level is lower (entries for the 1st spell
corresponds to “Entries into Paid Unemployment”). In addition, Figure 20 shows
clearly that while the number of people re-entering the UI system is lower than
the number of additional claims, the share of all entries which are re-entries
(individuals entering paid unemployment at a second or higher spell) is higher than
the share of initial claims which are additional. In other words, our measure of
churn is lower in absolute terms (relative to that implied by initial claims), but higher
in relative terms. This implies that using additional claims to measure the rate of
churn in the labor market will tend to underestimate the true rate since the share
of actual people re-entering UI is higher than the share of claims that are additional.
FIGURE 20: CPL’s Preferred Measure of Entries into Paid Unemployment, by Spell of Unemployment
Notes: Bars are stacked and mutually exclusive. An individual is determined to have begun their rst period of unemployment in the week which corresponds to
their rst paid week of unemployment. If an individual then experiences a gap in payments (one or more unpaid weeks) but then returns to paid unemployment,
they are determined to have begun another “spell” of unemployment. Only includes Regular UI claims and excludes PUA.
1st
Period
2nd+ Period
0
200,000
400,000
600,000
800,000
Jan 25
Feb 22
Mar 21
Apr 18
May 16
Jun 13
Jul 11
Aug 8
Sep 5
Oct 3
Oct 31
Nov 28
Dec 26
Jan 23
Feb 20
Mar 20
Apr 17
May 15
Jun 12
Jul 10
Aug 7
Sep 4
Oct 2
Oct 30
Nov 27
Dec 25
Jan 22
First Week of Unemployment
Individuals Entering Paid Unemployment
38 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
8. New Measure of Churn Indicates Substantial Repeat Layos and
Returns to UI that Diered Across Workers and Industries
Issue. The previous section illustrated how the initial claims measures overstate
the number of people entering the UI system. An implication of this result is
that using the raw additional claims data would provide a biased view of churn
in the labor market. This section uses our new measure of Entries into Paid
Unemployment to describe the amount of churn in the UI system. First, it
describes how the share of entries into the UI system that are re-entries has
evolved over the pandemic, and second it measures what share of people who
leave the UI system return within a short period of time. The pandemic began
with a large shock in early 2020 as the scale of the pandemic grew, and has since
been followed by periods of uncertainty as new virus variants emerge and public
health measures have changed. Our measure of churn can help shed light on how
stable new employment relationships have been throughout the pandemic (Bell,
Hedin, Moghadam, et al., 2021b).
Analysis. The rst measure of churn is again presented in Figure 20. It shows
that at the beginning of the crisis most people entering the UI system were
entering for the rst time. The share of people re-entering increased steadily
throughout 2020 and was over 50% of all entries by the summer of 2020.
Throughout most of 2021, the share of entries that were re-entries rarely
dipped below 90%. This means that throughout 2021, most of the people who
were becoming newly unemployed were people who had only recently found
work. On the other hand, the level of people re-entering paid unemployment
fell through 2021, which could mean that claimants were nding their new jobs
increasingly stable throughout 2021.
Figure 21 illustrates the second measure of churn. It takes the number of
claimants who transition from a week of paid unemployment to an unpaid week,
and then computes the share that return to paid unemployment within either
two, four, or ten weeks. Throughout most of the crisis, the amount of reentry
after four weeks was stable at 30–40% of claimants who had exited the UI
system, only declining slightly in April 2021.
If the percentage returning to UI increases, it implies that claimants who exit UI
are nding their newfound employment to be more unstable (i.e., they are more
likely to be laid o again at their new job or face another reduction in work
hours, bringing them back into the UI system). Alternatively, if the percentage
returning decreases, it could mean that more claimants are not returning to the
UI system for other reasons, for example, they exhausted their benets.
39 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Claimants who exited the UI system just before the pandemic were especially likely to return
once the pandemic hit. However, the amount of “churn” (people entering and exiting UI)
declined in April and May of 2020, partly because EDD automatically certied individuals for
benets from March 14th, 2020 to May 9th, 2020. Between June 2020 and January 2021,
the amount of churn was elevated, but remained remarkably stable, with about 30–40% of
claimants who exited the UI system returning within the next two weeks. During 2021, it
appears claimants who exited the UI system found employment to be increasingly stable,
and the level of churn has slowly declined between February and August. For claimants
who transitioned out of paid unemployment in the week ending August 7th, 2021, 30% had
returned to the UI system within the following four weeks. Overall, both measures of churn
nd that employment was likely becoming more stable through 2021.
As mentioned in the previous section, we can also use our new measure of entries into paid
unemployment to measure churn in the UI system across demographic groups. By using our
new measure, we do not have to worry about the possibility that there are dierences in the
amount of ination due to additional claims across demographic groups. In Figure 22, we
compare dierences in the level of churn between demographic groups, focusing on
claimants who exited the UI system between January and October of 2021. Thirty-two
percent of all claimants who exited during this period returned within four weeks. Female,
White, and older workers who exited were especially likely to return to the UI system, while
Black workers, Hispanic workers and younger workers were less likely to return. Figure 23
shows dierences in churn across industries with Information; Arts, Entertainment,
Recreation; and Healthcare and Social Services having the highest levels of churn.
Notes: This gure takes the number of people who transition from being paid regular UI benets in one week to not being paid in the next, and then calculates
the share who are paid for any unemployment experienced in the next 2, 4, or 10 weeks. It does not include PUA claimants. It also does not include claimants
who transition to unpaid in the same week that their benet year ends.
2020 2021
0
10
20
30
40
50
60
70
80
90
100
Jan 4
Feb 1
Feb 29
Mar 28
Apr 25
May 23
Jun 20
Jul 18
Aug 15
Sep 12
Oct 10
Nov 7
Dec 5
Jan 2
Jan 30
Feb 27
Mar 27
Apr 24
May 22
Jun 19
Jul 17
Aug 14
Sep 11
Last Week Collecting Benefits Before an Exit
2 weeks
4 weeks
10 Weeks
% Returning Within
Percent Returning
FIGURE 21: Share of Claimants Who Return to the UI System Within 2, 4, and 10 Weeks After Experiencing a Gap
in Payments
40 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
13%
15%
15%
16%
21%
21%
21%
23%
24%
25%
28%
28%
29%
29%
29%
30%
32%
32%
33%
37%
0 5 10 15 20 25 30 35 40 45
Percent of Exiters Who Returned to UI System Within 4 Weeks
Mining, Oil and Gas
Agriculture, Forestry, Fishing (a)
Utilities
Finance and Insurance
Management
Manufacturing
Public Admin
Wholesale Trade
Transp and Warehousing
Real Estate and Leasing
Admin. Support, Waste Man. (a)
Other Services
Construction
Education Services
Prof., Scientific, Techn. Services (a)
Retail Trade
Accommodation Food Svc
Health Care and Social Assistance
Arts, Entertainment, Recreation
Information
Notes: This gure takes the number
of people who, for any given week
of 2021, transition from being paid
UI benets in one week to not
being paid in the next, and then
calculates the share who are paid
for any unemployment experienced
in the next 4 weeks. It includes Reg-
ular UI claimants only. It does not
include claimants who transition to
unpaid in the same week that their
benet year ends.
FIGURE 23: Share of Claimants Who Return to the UI System Within 2, 4, and 10 Weeks After Experiencing a Gap
in Payments by Industry
Notes: This gure takes the number
of people who, for any given week
of 2021, transition from being paid
UI benets in one week to not
being paid in the next, and then
calculates the share who are paid
for any unemployment experienced
in the next 4 weeks. It includes Reg-
ular UI claimants only. It does not
include claimants who transition to
unpaid in the same week that their
benet year ends.
33%
30%
30%
31%
31%
32%
30%
34%
33%
31%
28%
31%
33%
32%
0 5 10 15 20 25 30 35 40 45
Percent of Exiters Who Returned to UI System Within 4 Weeks
White
Hispanic
Black
Asian American
Bachelor's or More
Some College/Associate's
HS Deg. or Less
Baby Boomers (56-85)
Gen X (40-55)
Millennials (24-38)
Gen Z (16-23)
Male
Female
Statewide
FIGURE 22: Share of Claimants Who Return to the UI System Within 4 Weeks After Experiencing a Gap in
Payments by Demographic Group
41 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
9. Retroactive Certications Can Distort Standard UI Benet Statistics:
How CPL’s Alternative Measure Provides a More Accurate Picture
Issue. While the previous sections and related work have focused on initial
claims, the other work-horse statistics of the UI system — continuing claims and
insured unemployment — are aected by measurement issues as well. Published
UI statistics typically show the total number of UI payments that were “certied”
in a given week, not the number of UI recipients who were actually unemployed in
a given week. Since individuals can retroactively certify for payments for multiple
weeks, both the level and the timing of this measure (usually called “continuing
claims”) may not accurately reect the number of individuals actually receiving
benets in that timeframe. Since the insured unemployment rate is simply the
ratio of continuing claims and number of individuals in covered employment, it is
also aected.
Once a UI claim is deemed eligible, the claimant must meet separate eligibility
criteria in each week of unemployment to receive payment for that week. These
eligibility criteria are veried through a process known as certication, which
claimants in California complete bi-weekly. We call individuals that complete
certication and are either paid UI benets for a given week, or who could have
received benets if not for excess earnings in that week, “potentially eligible
claimants” (Hedin et al., 2020b). Two key characteristics of this measure are worth
noting. First, at the time of certication these weeks are in the past. This means
that measures of UI receipt that count certications in each week (i.e., “continued
claims”) reect unemployment experienced for various time periods that are
at least 1–2 weeks prior to those certications. It is not possible to accurately
deduce from only counts of certications processed in a given week (the more
commonly reported measure) when that unemployment was experienced. Second,
due to processing lags, the date on which we observe a certication sometimes
comes later than the date that the certication was submitted by the claimant.
An alternate measure of continuing claims could overcome these problems
by focusing directly on the number of individuals receiving UI benets for
unemployment experienced in any given week, providing a more accurate
measure of the evolving status of the labor market. This measure is more directly
comparable to the number of unemployed individuals or the number of workers
in the labor force reported from Current Population Survey data than existing UI
statistics (Hedin et al., 2020b).
Figure 24 illustrates our key ndings about the complex and evolving relationship
between certications processed in a week and the number of Californians
who experienced unemployment that week for Regular UI and PUA claims. The
dashed dark blue line shows the number of payments certied each week, and is
analogous to “continued claims” measures often reported by the DOL.
42 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 24A: Comparison of Four Dierent Methods for Measuring Continuing Claims in Regular UI
Notes: The “Number of Payments Certied” refers to the number of payments that were certied during a given week (the common denition of continued UI
claims). The “Number of Individuals Certifying” refers to the number of people that certify for UI benets in a given week (which is roughly half of the number
of payments because certication is bi-weekly in California). The “Individuals Paid Benets by Week of Unemployment” refers to the number of individuals paid
benets for the week they experienced unemployment either adjusting for historical lags in claiming behavior (“Adjusted”) or not (“Unadjusted”). This gure
excludes PUA.
FIGURE 24B: Comparison of Four Dierent Methods for Measuring Continuing Claims in PUA
Notes: The “Number of Payments Certied” refers to the number of payments that were certied during a given week (the common denition of continued UI
claims). The “Number of Individuals Certifying” refers to the number of people that certify for UI benets in a given week (which is roughly half of the number
of payments because certication is bi-weekly in California). The “Individuals Paid Benets by Week of Unemployment” refers to the number of individuals paid
benets for the week they experienced unemployment either adjusting for historical lags in claiming behavior (“Adjusted”) or not (“Unadjusted”). This gure
includes only PUA.
Number of Payments Certied
(By Week Certication Processed)
Individuals Certifying
(By Week Certication Processed)
Individuals Paid Benets (Unadjusted)
(By Week of Unemployment)
Individuals Paid Benets (Adjusted)
(By Week of Unemployment)
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
Feb 15, 2020
Mar 14
Apr 11
May 9
Jun 6
Jul 4
Aug 1
Aug 29
Sep 26
Oct 24
Nov 21
Dec 19
Jan 16
Feb 13
Mar 13
Apr 10
May 8
Jun 5
Jul 3
Jul 31
Aug 28
Sep 25
Oct 23, 2021
Week Ending
Number of Payments Certied
(By Week Certication Processed)
Individuals Certifying
(By Week Certication Processed)
Individuals Paid Benets (Unadjusted)
(By Week of Unemployment)
Individuals Paid Benets (Adjusted)
(By Week of Unemployment)
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
7,000,000
Feb 15, 2020
Mar 14
Apr 11
May 9
Jun 6
Jul 4
Aug 1
Aug 29
Sep 26
Oct 24
Nov 21
Dec 19
Jan 16
Feb 13
Mar 13
Apr 10
May 8
Jun 5
Jul 3
Jul 31
Aug 28
Sep 25
Oct 23, 2021
Week Ending
43 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Figure 24A for regular UI shows a fairly consistent relationship between the
number of payments certied and the number of people who experienced
unemployment each week. Figure 24B, however, shows that for PUA claims,
the traditional measure of the number of payments certied each week grew
gradually during the pandemic until skyrocketing in August 2020. While this
traditional measure of payment certications may seem to indicate that there
were millions of new lings in August, our September 2020 analysis suggested
this was not the case (Bell et al., 2020a). There was an increase in initial claims
in late August, but the spike in payment certications was driven by the fact that
many of the individuals who led claims during that period (and certied for the
rst time) had been certifying for multiple weeks of benets, often all the way
back to the early stages of the crisis.
28
One thing that all the measures agree on is
that claims paid fell substantially after the September federal UI expirations.
29
Alternate Measure. In this section, we report original estimates of the total
number of individuals receiving UI benets based on the week in which they
experienced unemployment. This measure is shown for all programs in Figure
25 (Regular UI, PUA, PEUC Extensions, and EB extensions). Because we do
not observe certications until they are processed, our December 2020 report
developed a censoring adjustment for this series based on recent lag patterns
(Bell et al., 2020b). Intuitively, we cannot directly count the number of claimants
who were unemployed in recent weeks because many certications for these
weeks have yet to be processed (or potentially even submitted). The censoring
adjustment inates recent weeks’ counts of unemployed claimants by the percent
of processed certications that have typically trickled in at later dates.
30
However,
our censoring adjustment does not attempt to adjust for irregular delays in the
processing of claims.
28 Appendix Figure A11 also illustrates the lag when a claim was processed and the week of UI it applied to by plotting the average number of weeks between
when a claim was processed and the week of unemployment it applied to for PUA and Regular UI claims.
29 Appendix Figure A12 shows the relationship between week of unemployment and certication week before the pandemic (and does not include PUA claims
because there was no PUA program) and indicates the relationship was even more consistent before the pandemic.
30 Appendix Figure A13 shows the censuring adjustment used to inate recent weeks to their estimated nal total. Figure A13 also contains detailed notes about
how the gure was created and applied to the claims data.
44 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 25: Total Number of Individuals Paid Benets by Week of Unemployment
Notes: X-axis labels correspond to Saturdays. Data has been adjusted to account for delays in processing and retroactive claims. Figure includes all available
programs for the period 2/2/20–10/2/21.
Figure 25 also illustrates the impact of the federal UI program expiring in early
September, 2021. The pandemic UI extension programs (PEUC and PUA)
expired on September 4th, and the Extended Benets program (EB) expired
in California on September 11th, so the only program currently available is
the regular state UI program. As a result, the total number of claimants fell
substantially from 2.69 million during the week before September 4th to 406,000
the week after September 11th. The gure also highlights the eorts that the CA
Employment Development Department took to automatically transition eligible
claimants from the PEUC program, which ended September 4th, to the EB
program, which ended September 11th.
31
In total, we estimate that these eorts
provided over 400,000 Californians with an additional week of UI benets.
31 See news release on EDDs eorts: https://edd.ca.gov/About_EDD/pdf/news-21-55.pdf
Regular UI
(Continuing)
PUA (Expiring)
PEUC (Expiring)
EB (Expiring)
December 26th:
PEUC/PUA Temporarily Expire
September 4th:
PEUC/PUA Expire
September 11th:
Extended Benefits Expire
0
1,000,000
2,000,000
3,000,000
4,000,000
5,000,000
6,000,000
Feb 8
, 2020
Mar 7
Apr 4
May 2
May 30
Jun 27
Jul 25
Aug 22
Sep 19
Oct 17
Nov 14
Dec 12
Jan 9
Feb 6
Mar 6
Apr 3
May 1
May 29
Jun 26
Jul 24
Aug 21
Sep 18
, 2021
Week of Unemployment
Individuals Receiving Benefits
45 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
10. Partial UI and Denials Due to Excess Earnings Show Many Claimants
Remain Connected to Work, Reducing their Benets
Issue. Workers receiving UI benets are allowed to also earn partial wages up to a
threshold before becoming ineligible for UI in that week. If a claimant’s earnings are
over 133% of their weekly benet amount (WBA), then they are not considered
unemployed by EDD, and do not receive a payment. If they earn less than that,
the rst 25% of earnings in a week, or $25 (whichever is less) is disregarded, to
incentivize part-time work. Every dollar earned beyond this disregard amount is
deducted 1 for 1 from the claimant’s WBA. If the claimant earns $25 or less per
week, they receive the full WBA. This system is often referred to as “partial UI.”
Since partial UI benets are determined at the payment level, a partial UI claimant
may later receive higher UI payments (up to their full WBA) if their earnings
decrease in subsequent weeks. Similarly, a claimant whose payment is denied in a
given week due to excessive earnings can later receive partial UI or full benets if
their earnings decrease in subsequent weeks.
If the share of claimants reporting earnings increases, that could suggest a change
in economic conditions. One must, however, be cautious about interpreting these
changes in the rate of partial UI receipt. An increase in partial UI could result
from full-time workers experiencing a reduction in hours, and would indicate a
deterioration of economic conditions. However, an increase in partial UI could
also be the result of previously unemployed workers receiving part-time hours and
would indicate an improvement in economic conditions.
Additionally, an implication of the current structure for determining partial UI
payments is that changes in the “earnings disregard” (currently $25 or 25% of
weekly earnings in California) will inuence the number of claimants who receive
partial UI benets and the amount of benets they receive. Organizations like the
National Employment Law Project have advocated for larger earnings disregards
(Stettner et al., 2004). This would provide an increase in partial UI benets for
claimants who are working part-time, and would encourage claimants to take
additional part-time work since less of the income they earned would be “taxed”
away through lower UI benets.
Analysis. Between late February and August 2021, the share of paid claimants
who received partial UI benets fell from 11% to 9% (Figure 26). The share of
UI claimants who certied for benets but were denied payment due to excess
earnings or full-time work hovered between 5% and 6% from January through
April, before falling to just 4% in August. The share of claimants denied benets and
receiving partial benets increased substantially after the September 2021 federal UI
expirations. Appendix Figure A14 shows that regular UI claimants received partial
payments at much higher rates than PEUC or EB claimants and explains the large
spike in partial and denied payments when the extension programs expired.
46 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE 26: Partial UI and Denials Due to Excess Earnings
Notes: X-axis labels correspond to Saturdays. Partial UI refers to those reporting earnings during that week as a percent of all paid claims. Denied UI Payment
is as a percent of Potentially Eligible Claims, which is the sum of the number of paid claimants and the number of individuals denied payment because of excess
weekly earnings or full time work (see text). Does not include PUA Claims.
Appendix Table A3 shows the rate of partial UI and denied UI across
demographic groups in California during the week of October 2nd, 2021. The
table shows that older claimants, more educated claimants, and Asian American
and White workers were more likely to receive partial UI payments and be
denied payments due to excess earnings or full-time work. While female
claimants received partial UI at substantially higher rates than male claimants,
male claimants were slightly more likely to be denied payments.
To better understand how partial UI and denials due to excess earnings have
been inuenced by the pandemic (and policy responses to it), Appendix Figure
A15 plots these measures by industry. The Accommodation and Food Services
Sector has seen consistently high rates of partial UI during the pandemic, and has
seen less of a decline in the rate of partial benets during 2021 than that of other
industries. Similarly, the Accommodation and Food Services industry has seen a
larger share of claimants with payment denied relative to other industries, but
still well below the levels seen during the summer of 2020. Other industries have
seen a steady decline in the rate of payment denials over the course of 2021.
8.9%
15.0%
Denied UI Payment:
Excess Earnings
or Full-Time Work
Partial UI
2020 2021
0
2
4
6
8
10
12
14
16
18
Feb 29, 2020
Mar 28
Apr 25
May 23
Jun 20
Jul 18
Aug 15
Sep 12
Oct 10
Nov 7
Dec 5
Jan 2
Jan 30
Feb 27
Mar 27
Apr 24
May 22
Jun 19
Jul 17
Aug 14
Sep 11
Oct 9, 2021
Week of Unemployment
Percent of Individuals
47 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Finally, as described earlier, the number of recipients who receive partial
UI is inuenced by the size of the earnings disregard. Figure 27 illustrates
how increasing the earnings disregard from $25 to $300 would impact the
unemployment benets that claimants with diering WBA’s would receive. It
shows that increasing the earnings disregard increases benet payments on both
the extensive (more people would be able to receive some UI benets while
working) and intensive margins (people already receiving UI benets would
receive a higher amount). CPL’s policy brief on the earnings disregard examined
the potential impact of increasing the disregard to $150 and $300. It found that
early in the pandemic increasing the disregard to $150 could have resulted in a
13% increase in the number of claimants receiving partial UI payments. It would
have also resulted in an additional $51 million in benets paid out per week, and
would have disproportionately beneted female, younger, less educated, and
Hispanic workers (Hedin et al., 2020c).
FIGURE 27: Partial UI Benets under a $25 Disregard and a $300 Disregard
Notes: This gure simulates the increase in the amount of partial UI payments for claimants with dierent levels of WBA and dierent levels of weekly earn-
ings. For example, for claimants with a WBA of $450, if they earned $600 in a week, they would receive no UI payment under current law. If the disregard was
increased to $300, they could receive $150 dollars in partial UI.
48 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
CONCLUSION
Better Data can Lead to a More Equitable UI System and More
Accurate and Timely Measures of Labor Market Crises and
Recessions
Despite policymakers’ eorts to support workers during the pandemic through
various supplemental programs and extensions to regular UI benets, the
pandemic had a disproportionately negative impact on low-wage and less
advantaged workers. Together with pre-existing inequities in access to UI
benets, this exacerbated existing racial and economic disparities. Through the
ongoing relationship with the EDD, CPL was able to use California data to better
understand how the COVID-19 pandemic impacted demographic and economic
groups dierently, to improve our understanding of what policy changes could
lead to a more eective and equitable Unemployment Insurance System, and to
generate more timely, accurate, and inclusive national statistics on unemployment.
This paper summarized ten key insights from CPL’s 19 policy briefs and UI
reports based on anonymized administrative UI data published in the course
of the pandemic. Information on initial claims for UI allowed us to track the
cumulative impact of the pandemic on layos and number of individuals
experiencing long-term unemployment by demographic and industry groups and
across communities. Our access to the continuing claims data allowed us to not
just see who applied for UI, but to understand which workers and communities
actually received benets during the crisis and how the generosity of those
benets diered by group. Additionally, access to the “Base Wage” le containing
earnings information for all UI covered workers allows us to understand how
quickly dierent groups of workers were able to nd re-employment during and
after claiming their UI benets, and which groups of workers struggled the most.
Sections 1, 2, 3, and 6 examine these issues in California while also illustrating
how they could be better tracked and addressed by policymakers if these data
were available in every state.
During recessions, the generosity and duration of UI benets are often extended
either through federally created or state programs. However, measurement and
data access problems often adversely impact the evaluation and operation of
these programs. Section 4 uses our administrative data to analyze the importance
of the agship PEUC federal extended benets program for preventing benet
exhaustions across demographic groups. It also highlights the limitations of federal
exhaustion and extension program data and recommends improvements to their
data collection that would assist in better evaluating which workers benet from
these programs. Section 5 describes how a awed statistic adversely impacts the
operation of state Extended Benets programs and causes vulnerable claimants
49 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
to prematurely lose access to UI benets. We propose a simple solution that
Congress could implement to help workers in future recessions to better
weather extended periods of unemployment.
Finally, we were able to use our access to the administrative UI records to
analyze shortcomings in commonly used UI data. Section 7 illustrates how the
weekly initial claims data published by the DOL overstates the number of people
entering the UI system and proposes a new metric to accurately count claimants
entering paid unemployment. Section 8 then uses this new metric to accurately
measure how often people churn in and out of California’s UI system as they
nd jobs and are laid o again. Section 9 then demonstrates how the continuing
claims data can misstate the timing and number of people who experience
unemployment each week. Section 10 highlights how a substantial share of
individuals receiving UI benets still have ongoing employment.
CPL’s fruitful partnership with the California EDD has borne important new
research ndings about the UI system in California and will continue to generate
new insights into the future. Analyzing new topics such as the earnings of UI
claimants, gaining access to new administrative data sources, and linking these to
UI claims data will ensure that CPL will continue to advance our understanding of
the UI system and how the system can be improved to become more eective
and equitable.
50 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
ACKNOWLEDGMENTS
We gratefully acknowledge the Labor Market Information Division of the
California Employment Development Department for their partnership in
producing this analysis. This research was made possible through support from
Arnold Ventures, The James Irvine Foundation, the Smith Richardson Foundation,
the Russell Sage Foundation, the Alfred P. Sloan Foundation, the University of
California Oce of the President Multicampus Research Programs and Initiatives,
MRP-19-600774 and M21PR3278 and the Bylo Chacon Foundation. We also
thank the UCLA Social Science Division, the UCLA Vice Chancellor for Research
and Creative Activities, the Luskin School of Public Aairs and the California
Center for Population Research for their support. All errors should be attributed
to the authors.
Background on the data in this report
The size and richness of the administrative data we use allows us to analyze how
the crisis in the labor market has aected workers by gender, age, education, race,
and ethnic groups, as well as by detailed regions and industries. These analyses
complement both traditional survey-based measures of labor market outcomes,
which are very detailed but suer from large lags and low frequency, and weekly
publications of total UI claims, which are timely but lack the detail available here.
These data allow us to track the fast-moving nature of the crisis and to help
inform assistance for workers and rms aected by the upheaval in the labor
market. For inquiries about the denitions, methodology, and ndings of this
policy brief, please contact Till von Wachter. Email: tv[email protected].edu. To
obtain the data tabulations used in this report, please contact: Dr. Muhammad
Akhtar, Chief, Labor Market Information Division, California Employment
Development Department. Email: Muhammad.Ak[email protected].gov.
51 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
REFERENCES
Bartik, A. W., Bertrand, M., Lin, F., Rothstein, J., & Unrath, M. (2020). Measuring
the Labor Market at the Onset of the COVID-19 Crisis. Brookings Papers on
Economic Activity, 2020(2), 239–268. https://doi.org/10.1353/eca.2020.0010
Bell, A., Hedin, T., Mannino, P., Moghadam, R., Schnoor, G., & Von Wachter,
T. (2021). Re-employment, Recall, and Industry Transitions During the COVID-
19 Pandemic (UI Report). California Policy Lab. https://www.capolicylab.org/
wp-content/uploads/2021/12/Dec-2021-Analysis-of-Unemployment-Insurance-
Claims-in-California.pdf
Bell, A., Hedin, T., Mannino, P., Moghadam, R., Schnoor, G., & Von Wachter, T.
(2022). Disparities in Access to Unemployment Insurance During the COVID-19
Pandemic: Lessons from U.S. and California Claims Data. California Policy Lab.
Bell, A., Hedin, T., Moghadam, R., Schnoor, G., & Von Wachter, T. (2021a).
Why Extended UI Benets were Turned O Prematurely for Workers in 33 States
(UI Report). California Policy Lab. https://www.capolicylab.org/wp-content/
uploads/2021/04/Why-Extended-UI-Benets-Turned-O-Early-in-33-States.pdf
Bell, A., Hedin, T., Moghadam, R., Schnoor, G., & Von Wachter, T. (2021b).
An Analysis of Unemployment Insurance Claims in California During the COVID-
19 Pandemic (UI Report). California Policy Lab. https://www.capolicylab.org/
wp-content/uploads/2021/12/June-30th-Analysis-of-Unemployment-Insurance-
Claims-in-California-During-the-COVID-19-Pandemic.pdf
Bell, A., Hedin, T., Schnoor, G., & Von Wachter, T. (2020a). An Analysis of
Unemployment Insurance Claims in California During the COVID-19 Pandemic
(UI Report). California Policy Lab. https://www.capolicylab.org/wp-content/
uploads/2020/10/Sept-15th-Analysis-of-UI-Claims-in-CA-During-the-COVID-19-
Pandemic.pdf
Bell, A., Hedin, T., Schnoor, G., & Von Wachter, T. (2020b). An Analysis of
Unemployment Insurance Claims in California During the COVID-19 Pandemic
(UI Report). California Policy Lab. https://www.capolicylab.org/wp-content/
uploads/2021/02/Dec-21st-Analysis-of-CA-UI-Claims-during-the-COVID-19-
Pandemic.pdf
Bell, A., Hedin, T., Schnoor, G., & von Wachter, T. (2021). As Crisis Continues, More
Unemployed Californians are Receiving UI Benets (UI Report). California Policy Lab.
https://www.capolicylab.org/wp-content/uploads/2021/02/CPL-Data-Point-on-
Recipiency-in-California.pdf
52 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Cajner, T., Figura, A., Price, B. M., Ratner, D., & Weingarden, A. (2020). Reconciling
Unemployment Claims with Job Losses in the First Months of the COVID-19 Crisis.
https://www.federalreserve.gov/econres/feds/reconciling-unemployment-claims-
with-job-losses-in-the-rst-months-of-the-covid-19-crisis.htm
Fields-White, M., Graubard, V., Rodriguez, A., Zeichner, N., & Robertson, C.
(2020, September 17). Unpacking Inequities in Unemployment Insurance. New
America. http://newamerica.org/pit/reports/unpacking-inequities-unemployment-
insurance/
Ghitza, Y., & Steitz, M. (2020). DEEP-MAPS — Model of the Labor Force. Catalist.
https://catalist.us/deep-maps/
Gould-Werth, A. (2016). Workplace Experiences and Unemployment Insurance
Claims: How Personal Relationships and the Structure of Work Shape
Access to Public Benets. Social Service Review, 90(2), 305–352. https://doi.
org/10.1086/687298
Government Accountability Oce. (2020). COVID-19: Urgent Actions Needed to
Better Ensure an Eective Federal Response (Report to Congressional Committees
GAO-21-191). Government Accountability Oce.
Gupta, A. H. (2020, May 9). Why Some Women Call This Recession a
‘Shecession.’ The New York Times. https://www.nytimes.com/2020/05/09/us/
unemployment-coronavirus-women.html
Hedin, T., Schnoor, G., & Von Wachter, T. (2020a). An Analysis of Unemployment
Insurance Claims in California During the COVID-19 Pandemic (UI Report). California
Policy Lab. https://www.capolicylab.org/wp-content/uploads/2020/06/June-11th-
Analysis-of-CA-UI-Claims-During-the-COVID-19-Pandemic.pdf
Hedin, T., Schnoor, G., & Von Wachter, T. (2020b). An Analysis of Unemployment
Insurance Claims in California During the COVID-19 Pandemic (UI Report). California
Policy Lab. https://www.capolicylab.org/wp-content/uploads/2020/07/July-2nd-
Analysis-of-UI-Claims-in-California-During-the-COVID-19-Pandemic.pdf
Hedin, T., Schnoor, G., & Von Wachter, T. (2020c). The Benet Implications of
Adjusting the Earnings Disregard for Partial UI Benets in California (UI Report).
California Policy Lab. https://www.capolicylab.org/wp-content/uploads/2020/07/
CPL-Analysis-on-Earnings-Disregard-for-Partial-UI-in-California.pdf
Hellerstein, E. (2020, April 2). Non-English speakers struggle to le coronavirus
unemployment claims. CalMatters. http://calmatters.org/california-divide/2020/04/
non-english-speakers-struggle-unemployment-applications/
53 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
Nicholson, W., & Needels, K. (2006). Unemployment Insurance: Strengthening
the Relationship between Theory and Policy. Journal of Economic Perspectives,
20(3), 47–70. https://doi.org/10.1257/jep.20.3.47
Price, B. M. (2021, July 2). Why Have Initial Unemployment Claims Stayed So High
for So Long? https://www.federalreserve.gov/econres/notes/feds-notes/why-have-
initial-unemployment-claims-stayed-so-high-for-so-long-20120702.htm
Schmieder, J. F., von Wachter, T., & Bender, S. (2012). The Eects of Extended
Unemployment Insurance Over the Business Cycle: Evidence from Regression
Discontinuity Estimates Over 20 Years *. The Quarterly Journal of Economics,
127(2), 701–752. https://doi.org/10.1093/qje/qjs010
Shaefer, H. L. (2010). Identifying Key Barriers to Unemployment Insurance for
Disadvantaged Workers in the United States. Journal of Social Policy, 39(3), 439–
460. https://doi.org/10.1017/S0047279410000218
Stettner, A., Smith, R., & McHugh, R. (2004). Changing Workforce, Changing
Economy: State Unemployment Insurance Reforms for the 21st Century. National
Employment Law Project. https://www.nelp.org/wp-content/uploads/2015/03/
ChangingWorkforce.pdf
US Bureau of Labor Statistics. (2021). Local Area Unemployment Statistics Home
Page. https://www.bls.gov/lau/
Wenger, J. B., & Walters, M. J. (2006). Why Triggers Fail (and What to Do about
It): An Examination of the Unemployment Insurance Extended Benets Program.
Journal of Policy Analysis and Management, 25(3), 553–575.
Woodbury, S. (1996). Emergency Extensions of Unemployment Insurance: A
Critical Review and Some New Empirical Findings. Book Chapters. https://research.
upjohn.org/cgi/viewcontent.cgi?article=1099&context=bookchapters
54 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE APPENDIX
FIGURE A1: Weekly Benet Amount by Program for the Cohort of Claimants Who Entered March 2020
Notes: The lines represent the mean weekly benet amounts and the bar represent the number of claims paid each week. Weekly Benet Amounts for the
cohort of claimants who entered the UI system in March 2020. Weekly Benet Amount is from the initial claims paid and is not the average weekly amount
actually paid. The evolution of mean WBAs partly reects the transition of a large cohort of low-wage workers from regular UI to PEUC, partly the selection of
who remains on PEUC over time. Those receiving state extended benets (EB) had higher WBAs than PEUC claimants, partly reecting the elevated earnings
requirements that are necessary to enter the program. The long tail of Regular UI claimants between September 2020 and March 2021 represent claimants who
were either receiving partial UI or who churned in and out of the UI system and were still collecting their 26 weeks of UI through during that period. During that
period about half of the claims paid were partial UI payments suggesting that claimants during this long tail were roughly split between partial UI recipients and
claimants re-entering UI.
55 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A2: Exhaustion Rate of All Programs and Exhaustion Rate of Regular UI by Entry Cohort
Notes: The blue line shows the share of all claimants who entered UI each week and who ultimately received all the benets they were eligible for during the
pandemic. The orange line shows the share of all claimants who entered UI each week and who ultimately received all their Regular (non-extension) benets
during the pandemic.
FIGURE A3: The Relationship Between the Share of PEUC and EB Claimants in Each Demographic Group and the
Share of Exhaustion by Each Demographic Group.
Notes: The x-axis represents the share of all PEUC or EB claimants that each demographic group accounts for during the week of 9/4/21. The y-axis represents
the share of all exhaustions each demographic group accounts for through the end of Q3 2021. The dots are colored by demographic category (i.e. gender, race,
etc.).
56 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A4: Exhaustion Rates within Demographic Groups for Claimants who Entered UI in March 2020
Notes: Each bar represents the share of all claimants of each demographic group who entered the UI system in March 2020 that had exhausted their benets by
the end of Q3 2021.
FIGURE A5: Demographic Groups as a Share of all Claimants on the Extended Benets (EB) Program on 9/4/21
Notes: Each blue bar represents the share of all EB claimants that each demographic group accounts for during the week ending 9/4/21.
57 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
TABLE A1: State Level Impacts of IUR Design
Notes: Table is sorted by the share of claimants on EB when EB triggered o. Number and share of claimants on EB in each state are constructed by taking the
maximum weekly values within a month of the ocial turn-o date. The analysis includes both states and territories (Puerto Rico was the only territory aected).
STATE
WEEKS OF
EB LOST TO
IUR DESIGN
NUMBER
OF EB
CLAIMANTS
SHARE OF
CLAIMANTS
ON EB
Minnesota 44 25,302 10.8
Puerto Rico 39 29,246 20.5
Oregon 39 21,582 13.3
Louisiana 37 5,075 3.9
Delaware 37 1,482 5.8
South Carolina 36 27,687 23.8
Hawaii 35 124 0.2
Vermont 34 1,117 4.9
Maryland 33 8,733 6
West Virginia 32 37 0.1
Michigan 31 61,758 13.2
North Carolina 28 29,554 13.7
Oklahoma 28 0 0
Mississippi 26 8,306 16.3
Pennsylvania 23 9,563 1.8
Massachusetts 20 2,887 1.1
Ohio 20 29,113 10.6
Rhode Island 19 184 0.5
Arizona 17 10,389 7
Washington 15 27,923 16.1
STATE
WEEKS OF
EB LOST TO
IUR DESIGN
NUMBER
OF EB
CLAIMANTS
SHARE OF
CLAIMANTS
ON EB
California 15 236,507 18.5
New York 13 10,441 1
Illinois 12 69,542 22.7
Nevada 12 29,909 42.6
Alabama 11 30,547 29.4
Virginia 8 21,297 12.3
Maine 7 2,388 8.6
Montana 6 929 4.5
Iowa 6 5,300 8
Indiana 6 3,069 2.2
Arkansas 5 3,883 10.2
Tennessee 5 3,394 3.1
North Dakota 4 389 2.6
Missouri 4 4,613 4
New Hampshire 3 500 1.7
Wisconsin 3 10,805 7.1
Idaho 2 232 1.1
Utah 2 1,116 3.3
Wyoming 2 478 4.7
Alaska 1 1,991 21.2
Nebraska 1 232 1
58 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
TABLE A2: Re-Employment and Recall by Demographic group
Notes: All claimants include everyone who led for UI and received at least one payment in Q2 2020. The claimants in-sample exclude anyone who received
income in their rst week of UI (did not experience full separations). A claimant is considered re-employed if they received income in the fourth quarter after
separation. They are considered recalled if they received income from the employer they were separated from. Claimants who identied as Native American or
Native Alaskan and claimants who declined to self-identify are not included.
GROUP
ALL CLAIMANTS
(INCLUDING
THOSE NOT
EXPERIENCING
FULL
SEPARATIONS)”
ALL CLAIMANTS
IN SAMPLE
(EXPERIENCING
FULL
SEPARATIONS)
NUMBER
EMPLOYED
WITHIN 2
QUARTERS
NUMBER
RECALLED
TO PRIOR
EMPLOYER
WITHIN 2
QUARTERS
PERCENT OF
SAMPLE RE-
EMPLOYED
WITHIN 2
QUARTERS
(OBSERVED
EARNINGS IN
BASE WAGE FILE)
PERCENT
OF SAMPLE
RECALLED
PERCENT OF
RE-EMPLOYED
WHO WERE
RECALLED
TO PRIOR
EMPLOYER
PERCENT
OF THOSE
EXPECTING
RECALL
ACTUALLY
EXPERIENCING
RECALL
Statewide 2,739,400 2,088,140 1,200,206 709,440 57.5 34.0 59.1 39.1
GENDER
Female 1,389,947 1,050,828 613,984 373,830 58.4 35.6 60.9 40.6
Male 1,343,329 1,032,959 583,928 334,983 56.5 32.4 57.4 37.6
GENERATION
Gen Z
(16-23)
485,744 381,691 241,320 97,810 63.2 25.6 40.5 31.7
Millennials
(24-38)
1,085,783 824,638 483,415 270,955 58.6 32.9 56.1 38.8
Gen X
(40-55)
701,873 531,827 305,865 210,199 57.5 39.5 68.7 43.7
Baby Boomers
(56+)
455,083 342,684 166,009 129,124 48.4 37.7 77.8 40.5
EDUCATION
HS or Less 1,243,316 956,729 512,346 289,432 53.6 30.3 56.5 34.0
Some College 802,561 619,924 375,366 227,171 60.6 36.6 60.5 43.8
Bachelor’s or
More
489,858 374,809 236,296 139,871 63.0 37.3 59.2 46.4
RACE/ETHNICITY
Asian American 462,192 356,679 198,064 122,588 55.5 34.4 61.9 39.4
Black 224,397 170,474 88,002 43,310 51.6 25.4 49.2 31.0
Hispanic 971,383 742,391 466,020 272,966 62.8 36.8 58.6 42.6
White 849,445 649,082 350,159 209,647 53.9 32.3 59.9 37.8
59 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A6: Percent of Re-employed UI Claimants who Returned to their Previous Industry by Demographic Group
Notes: The denominator for each bar includes all regular UI claimants from that demographic group who led a UI claim during the 2nd quarter of 2020 and
were re-employed a year later. The numerator consists of the subset of those claimants who had returned to the previous industry a year later.
FIGURE A7: Spatial Correlations with Re-employment Rate
Notes: This gure illustrates the correlation between various county-level characteristics and the rate at which individuals who led for UI benets in that county
during quarter 2 of 2020 were re-employed by any employer (a year later). The county-level characteristics are constructed from ACS 5-year estimates from
2015-2019. The information on COVID conrmed cases is sourced from the New York Times.
Percent in poverty
Means of transportation to work, public transit
SNAP recipient, percent
Black, percent
Share Self-Employed
Agricultural employment, percent
Population, ages 65 plus, percent
COVID Confirmed Cases, % pop
Limited English, share aged 5+
Hispanic, percent
Retail trade employment, share
Population share aged 20-24
noncitizen_share
Median Household Income
Broadband access of any type, share
-1 -.8 -.6 -.4 -.2 0 .2 .4 .6 .8 1
Correlation With Re-Employment
60 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A8: Recall among the Re-employed by Industry Before and During the Pandemic.
Notes: The denominator in each period includes all regular UI claimants in each industry who led a UI claim during the 2nd quarter of 2020 or the 4th quarter
of 2018 and were reemployed a year later. The numerator consists of the subset of those claimants who had been reemployed by their most recent employer.
The size of the point represents the share of claimants that each industry accounts for in the Q2 2020 cohort.
FIGURE A9: Initial UI Claims Substantially Over-State Entry into the UI System (Regular UI Only) (2019)
Notes: Bars are not stacked, but instead compare the dierent measures within the same week. The initial claims bars do not include initial claims for PUA, new
initial claims for extension programs, or transitional claims. They only include new initial claims for regular UI, additional claims for regular UI, and additional claims
for claimants on extension programs.
61 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A10: Initial UI Claims Over-State Entry into the UI System by Demographic Group (Regular UI Only)
(3/1/2021–9/28/2021)
Notes: Bars represent the percent by which initial claims are higher than entries into paid unemployment. Excludes PUA claims. Includes the period March 2021
to September 2021. Initial claims include new initial claims, additional initial claims for regular UI and additional initial claims for extension programs.
FIGURE A11: Average Number of Weeks Between When a Claim Was Processed and the Week of Unemployment
It Applied to (By Regular and PUA claims)
Notes: The gure shows the average (mean) number of weeks that elapsed between when a claim was processed and when the unemployment that claim
applied to was actually experienced. The increases at the end of the graph are artifacts of the PUA and PEUC expirations. After the expirations, claimants could
only certify if they had weeks of eligible unemployment before the expiration. These retroactive claims are inating these series near the end.
62 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A12: Total Number of Individuals Paid Benets by Week of Unemployment, Total Number of Individuals
Certifying for Benets by Week of Certication, and Total Number Payments Certied by Week of Certication (2019)
Notes: X-axis labels correspond to Saturdays. The “Number of Payments Certied” refers to the number of payments that were certied during a given week
(the common denition of continued UI claims). The “Number of Individuals Certifying” refers to the number of people that certify for UI benets in a given
week. This gure includes claimants receiving benets for regular UI, EB, and PEUC.
Number of Payments Certified
(By Week Certification Processed)
Number of Individuals Certifying
(By Week Certification Processed)
Individuals Paid Benefits (Adjusted)
(By Week of Unemployment)
0
100,000
200,000
300,000
400,000
500,000
600,000
Jan 2
Jan 30
Feb 27
Mar 27
Apr 24
May 22
Jun 19
Jul 17
Aug 14
Sep 11
Oct 9
Nov 6
Dec 4
Jan 1
Week Ending
63 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A13: Lag Structure of Regular UI Claims by Week of Unemployment
PANEL A: DURING THE PANDEMIC
Note: This gure displays the percent of claims processed for unemployment experienced a variable number of weeks in the past. To construct this gure, we
limit the sample to payments processed between May 16 and November 14 of 2020. We also limit the sample to payments for regular UI (not including PUA)
processed within 26 weeks of unemployment. The line represents the estimated share of nal claims paid by week in the past. For example, the claims paid for
unemployment experienced two weeks before the data is published is estimated to be 76% of total claims that will ever (within 26 weeks) be paid for unemploy-
ment experienced in that week. We have used this series in our reports to construct a censoring-corrected number of Californians who experienced unemploy-
ment in each week. To illustrate how the censoring correction works, if one million Californians had led a UI claim for unemployment experienced two weeks
prior, we would expect that 1 million / .76 = approximately 1.3 million Californians would eventually le a UI claim for that week.
PANEL B: PRE-PANDEMIC
Note: This gure displays the percent of claims processed for unemployment experienced a variable number of weeks in the past. To construct this gure, we limit
the sample to payments processed in 2019 within 26 weeks of unemployment. The line represents the estimated share of nal claims paid by week in the past.
0.6
0.7
0.8
0.9
1.0
0
2
4
6
8
10
12
14
16
18
20
22
24
26
Weeks in the Past
Percent of Final Value
0.6
0.7
0.8
0.9
1.0
0
2
4
6
8
10
12
14
16
18
20
22
24
26
Weeks in the Past
Percent of Final Value
64 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A14: Partial Payments by Regular vs PEUC & EB Claimants
Notes: X-axis labels correspond to Saturdays. Partial UI refers to those reporting earnings during that week as a percent of all paid claims. Does not include PUA
Claims. The blue line represents Regular UI claimants receiving partial UI and the orange represents PEUC or EB claimants receiving partial UI.
65 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
FIGURE A15: Denied Payments and Partial Payments by Industry
Notes: X-axis labels correspond to Saturdays. Does not include PUA claims. Partial UI is as a percent of all paid claimants. Denied UI payment is a percent of potentially
eligible claimants, which is the sum of the number of paid claimants and the number of claimants denied because of excess weekly earnings or full time work.
Accommodation and Food Services
Retail Trade
Health Care and Social Assistance
Admin. Support, Waste Man.
Manufacturing
0
5
10
15
20
Feb 15
Mar 14
Apr 11
May 9
Jun 6
Jul 4
Aug 1
Aug 29
Sep 26
Oct 24
Nov 21
Dec 19
Jan 16
Feb 13
Mar 13
Apr 10
May 8
Jun 5
Jul 3
Jul 31
Aug 28
Sep 25
Week of Unemployment
Percent of All Eligible Claimants
With Payment Denied Due To Excess Earnings or FT Work
Accommodation and Food Services
Retail Trade
Health Care and Social Assistance
Admin. Support, Waste Man.
Manufacturing
0
5
10
15
20
25
Feb 15
Mar 14
Apr 11
May 9
Jun 6
Jul 4
Aug 1
Aug 29
Sep 26
Oct 24
Nov 21
Dec 19
Jan 16
Feb 13
Mar 13
Apr 10
May 8
Jun 5
Jul 3
Jul 31
Aug 28
Sep 25
Week of Unemployment
Percent of All Paid Claimants Receiving Partial UI
X-axis labels correspond to Saturdays. Does not include PUA claims. Partial UI is as a percent of all paid claimants. Denied UI payment is a percent of potentially eligible claimants,
which is the sum of the number of paid claimants and the number of claimants denied because of excess weekly earnings or full time work (see text).
Accommodation and Food Services
Retail Trade
Health Care and Social Assistance
Admin. Support, Waste Man.
Manufacturing
0
5
10
15
20
Feb 15
Mar 14
Apr 11
May 9
Jun 6
Jul 4
Aug 1
Aug 29
Sep 26
Oct 24
Nov 21
Dec 19
Jan 16
Feb 13
Mar 13
Apr 10
May 8
Jun 5
Jul 3
Jul 31
Aug 28
Sep 25
Week of Unemployment
Percent of All Eligible Claimants
With Payment Denied Due To Excess Earnings or FT Work
Accommodation and Food Services
Retail Trade
Health Care and Social Assistance
Admin. Support, Waste Man.
Manufacturing
0
5
10
15
20
25
Feb 15
Mar 14
Apr 11
May 9
Jun 6
Jul 4
Aug 1
Aug 29
Sep 26
Oct 24
Nov 21
Dec 19
Jan 16
Feb 13
Mar 13
Apr 10
May 8
Jun 5
Jul 3
Jul 31
Aug 28
Sep 25
Week of Unemployment
Percent of All Paid Claimants Receiving Partial UI
X-axis labels correspond to Saturdays. Does not include PUA claims. Partial UI is as a percent of all paid claimants. Denied UI payment is a percent of potentially eligible claimants,
which is the sum of the number of paid claimants and the number of claimants denied because of excess weekly earnings or full time work (see text).
66 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
TABLE A3: Individuals Potentially Eligible for Regular UI Benets and Receiving Regular UI Benets, Total and as
Fraction of the Labor Force and the Unemployed, and Share with Reduced UI Benets, for Unemployment in the
Week Ending October 2nd, 2021
Notes: Numbers are adjusted for delays in processing and expected retroactive claims which have not yet been processed. “Potentially Eligible” includes claims
which are either paid or have payment denied due to excess weekly earnings or full-time work. This table does not include PUA claimants. Table does not show
information on claimants for whose race is specied as Native American or Alaskan Natives. Claimants who identify as other race categories or choose not to
self-identify are included in the “Unknown” category.
GROUP
INDIVIDUALS WITH
POTENTIALLY
ELIGIBLE CLAIMS
INDIVIDUALS WITH
CLAIMS PAID
INDIVIDUALS
WITH PARTIAL UI
PAYMENTS AS A
PERCENT OF ALL
PAID CLAIMS
PERCENT OF
POTENTIALLY
ELIGIBLE
INDIVIDUALS WITH
PAYMENT DENIED
INDIVIDUALS WITH
POTENTIALLY
ELIGIBLE CLAIMS AS
A PERCENT OF FEB
LABOR FORCE
INDIVIDUALS
RECEIVING FULL
WBA AS A PERCENT
OF UNEMPLOYED IN
DECEMBER
Statewide 361,979 329,951 15.0 8.9 1.9 25.4
GENDER
Female 186,651 170,429 18.6 8.8 2.1 29.5
Male 174,214 158,564 11.2 9.0 1.6 22.3
GENERATION
16-19 4,888 4,676 6.8 4.3 0.9 5.0
20-24 40,018 37,873 11.6 5.4 2.3 22.4
25-34 103,122 96,235 13.2 6.7 2.2 32.9
35-44 75,460 68,361 14.9 9.5 1.8 29.3
45-54 61,084 54,074 17.0 11.6 1.6 24.0
55-64 55,821 49,414 17.9 11.6 1.8 27.5
65-85 20,649 18,521 20.7 10.4 1.8 18.8
RACE/ETHNICITY
White 109,667 97,911 15.7 10.8
Hispanic 135,056 124,664 13.0 7.8
Asian American 45,109 40,901 23.1 9.4
Black 36,674 34,533 11.5 5.9
EDUCATION
HS or Less 165,206 153,075 13.3 7.4 2.5 26.8
Some College/
Associate's
105,238 95,782 16.9 9.1 2.1 24.6
Bachelor's or
More
62,171 54,458 16.2 12.4 0.8 16.1
67 10 KEY INSIGHTS FROM THE COVID-19 PANDEMICcapolicylab.org
