Did Pandemic Unemployment Benefits Reduce Employment? Evidence from Early State-Level Expirations (Working Paper 29575)

Summary

NBER Working Paper 29575, dated December 2021, is by Harry J. Holzer, R. Glenn Hubbard, and Michael R. Strain and is titled Did Pandemic Unemployment Benefits Reduce Employment? Evidence from Early State-Level Expirations in June 2021. The abstract says the FPUC and PUA programs were set to expire in September 2021, but 18 states opted out of both in June 2021. Using Current Population Survey data, the authors present estimates that the flow of unemployed workers into employment increased by around two-thirds following early termination. They say a counterfactual implies the national unemployment rate in each of July and August would have been around 0.3 percentage point lower had all states ended FPUC and PUA in June.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

About this paper. The authors find that ending pandemic unemployment benefits early, in June 2021, was associated with a 14 percentage point increase in the flow from unemployment to employment among unemployed workers aged 25 to 54. They also cite a median statutory replacement rate of 145 percent in 2020. The size of an employment effect cannot be read from replacement rates alone.

Issuer: Source documents
Document type: Did Pandemic Unemployment Benefits Reduce Employment Evidence From Early State Level Expir
Case: Did Pandemic Unemployment Benefits Reduce Employment Evidence From Early State Level Expirations In June 2021

Cite this document

“Did Pandemic Unemployment Benefits Reduce Employment? Evidence from Early State-Level Expirations (Working Paper 29575).” Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/did-pandemic-unemployment-benefits-reduce-employment-evidence-from-early-state-level-expirations-in-june-2021/, SHA-256 fa38d0ca551e12dd2f8b6dab6bec7ca079265829539480e9633df0a1a7a30a72.

Full text

NBER WORKING PAPER SERIES

DID PANDEMIC UNEMPLOYMENT BENEFITS REDUCE EMPLOYMENT? EVIDENCE
FROM EARLY STATE-LEVEL EXPIRATIONS IN JUNE 2021

Harry J. Holzer
R. Glenn Hubbard
Michael R. Strain

Working Paper 29575
http://www.nber.org/papers/w29575

NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
December 2021

We thank Duncan Hobbs for outstanding research assistance and Diane Schanzenbach for helpful
comments. The views expressed herein are those of the authors and do not necessarily reflect the
views of the National Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.

© 2021 by Harry J. Holzer, R. Glenn Hubbard, and Michael R. Strain. All rights reserved. Short
sections of text, not to exceed two paragraphs, may be quoted without explicit permission
provided that full credit, including © notice, is given to the source.
Did Pandemic Unemployment Benefits Reduce Employment? Evidence from Early State-Level
Expirations in June 2021
Harry J. Holzer, R. Glenn Hubbard, and Michael R. Strain
NBER Working Paper No. 29575
December 2021
JEL No. J08,J65

ABSTRACT

The generosity of Unemployment Insurance (UI) benefits was expanded during the pandemic
(FPUC), along with the groups of workers eligible for benefits (PUA). These two programs were
set to expire in September 2021, but 18 states opted out of both in June 2021. Using Current
Population Survey data, we present difference-in-difference and event study estimates that the
flow of unemployed workers into employment increased by around two-thirds following early
termination. We construct a counterfactual scenario that implies the national unemployment rate
in each of July and August would have been around 0.3 percentage point lower than they were,
and the employment-population ratio would have been around 0.1-0.2 percentage point higher
than it was, had all states ended FPUC and PUA in June. Expanded eligibility and generosity of
UI may have both slowed transitions from unemployment to employment. We also present some
suggestive evidence that households with relatively high confidence in their ability to meet
expenses may have been less sensitive to the termination of expanded benefits. Finally, we
present evidence that early termination reduced the share of households that had no difficulty
meeting expenses by five percent. The welfare implications of the early termination of FPUC and
PUA are therefore ambiguous.

Harry J. Holzer Michael R. Strain
McCourt School of Public Policy American Enterprise Institute
Old North 401 1789 Massachusetts Avenue, NW
Georgetown University Washington, DC 20036
37 and O St. NW and IZA
Washington DC 20057 michael.strain@aei.org
hjh4@georgetown.edu

R. Glenn Hubbard
Graduate School of Business
Columbia University, 607 Uris Hall
3022 Broadway
New York, NY 10027
and NBER
rgh1@columbia.edu
1. Introduction

This paper studies whether special pandemic-era unemployment benefits reduced

employment. The American Rescue Plan, enacted in March 2021, extended previous pandemic-

era measures that expanded eligibility for Unemployment Insurance (UI) benefits to workers

who are typically ineligible for state UI programs (Pandemic Unemployment Assistance; PUA)

and that expanded the generosity of standard benefits. The latter program, Federal Pandemic

Unemployment Compensation (FPUC), added a $300 weekly supplement to standard state UI

benefits from the law’s passage in March 2021 until September 6, 2021.

Concerns about the labor-market effects of PUA and FPUC led 26 states to opt out of at

least one of these programs before it was set to expire in September 2021. Of those 26 states, 18

stopped participating in both programs in June 2021.1 These 18 states constitute the main focus

of our analysis. Twenty-four states and the District of Columbia continued participating until the

federal programs expired.

Using difference-and-difference and event-study estimators on Current Population

Survey data, we estimate the effect of early termination of pandemic-era UI benefits on flows

from unemployment to employment. Among unemployed workers ages 25 to 54, we find that

early termination is associated with a 14 percentage point increase in the unemployment-to-

employment flow. This effect is over two-thirds the size of the unemployed-to-employed (U-to-

E) flow among control states during the February-June 2021 “pre period” (21 percent).

1
Tennessee and Louisiana opted out of both FPUC and PUA in July 2021. Alaska, Arizona, Florida, and Ohio
stopped participating in FPUC prior to September 2021, but continued participating in PUA. Maryland and Indiana
opted out, but were ordered by a court to resume participating. We drop all these states from our main analysis. See
Table 1 for details. AZ opted out of FPUC in July. We use the experience of AK, FL, and OH to study whether
ending expanded benefit generosity affected transitions from unemployment to employment in states that continued
expanded eligibility. See Table A5 for our estimates.

2
The magnitude of the effect is slightly smaller when we estimate our models on samples

of unemployed workers ages 25-54 with less than a college degree (12 percentage points) and of

workers ages 16-64 (13 percentage points). Among these samples, the coefficient of interest is

statistically significant. Among workers ages 25-54 who last worked in the leisure and

hospitality and retail industries, the coefficient magnitude is also smaller (11 percentage points)

and the coefficient is statistically insignificant.

The possibility that unemployment-to-employment flows were on separate trajectories

among states that did and did not withdraw early from pandemic-era UI benefit programs is a

potential threat to the validity of our estimates. To explore this possibility, in Figure 1 we present

event study evidence for each of these four sample groups. For each, the monthly coefficients

from February through May are centered around or close to zero, and are imprecisely estimated.

This pattern mitigates concerns about divergent “pre-period” trends and supports a causal

interpretation of our estimates. It is also reassuring that the magnitude of the September

coefficient falls closer to zero and that the coefficient is statistically insignificant. Since both the

“treatment” and “control” states had ended pandemic-era UI programs in September, this finding

strengthens confidence in our results for July and August.

For July and August, the estimated coefficients are positive and for our main sample —

workers ages 25 to 54 — and are precisely estimated. The magnitude is slightly larger in July

than in August (14 and 13 percentage points, respectively). Among prime-wage workers without

a college degree, the coefficient is statistically significant in July (12 percentage points) but not

in August. Among prime-age leisure and hospitality and retail workers, the coefficient is

statistically insignificant for both months. Among workers ages 16 to 64, both coefficients are

3
precisely estimated, with magnitudes in July and August of 14 and 15 percentage points,

respectively.

We use these estimates in a simple counterfactual exercise to examine what the

unemployment rate and employment-population ratio might have been if the 24 states and the

District of Columbia that participated in FPUC and PUA through September had instead opted

out in June. In those states, workers ages 25-54 would have had an unemployment rate 0.8

percentage point lower in July and 0.7 percentage point lower in August. Employment rates

would have been 0.7 percentage point and 0.6 percentage point higher in July and August,

respectively. The differences between the actual and the counterfactual unemployment rates are

larger for workers ages 16-64, while counterfactual employment rates are roughly the same.

We extend this exercise to determine what the national unemployment rate and

employment-population ratios would have been if all states opted out of pandemic-era UI

programs in June. We estimate that the national unemployment rate in July would have been

around 0.3 percentage point lower and the aggregate employment rate in July would have been

0.2 percentage point higher. In August, the unemployment rate would have been around 0.3

percentage point lower and the employment rate about 0.1 percentage point higher. The

differences between the actual and counterfactual unemployment and employment rates using

estimates from workers ages 16-64 are larger.

Characterizing the magnitude of these effects is challenging because of the unusual

circumstances of the pandemic and various policy responses to support household, businesses,

and the economy in the face of it, and because of the unusually heated debate over these

programs.

4
Our counterfactual estimates that FPUC and PUA added 0.3 percentage point to the

national unemployment rate in each of July and August are quite modest when compared to the

likely effect of Covid-19 itself. Data from the September 15-27 Household Pulse Survey (HPS),

administered by the U.S. Census Bureau, report that 4 million people were home sick with Covid

symptoms or caring for someone in the same situation and 3 million weren’t working because

they were worried about Covid. Our estimates may also be modest when compared with the

labor market effect of unstable school and day cares. The HPS reported 5 million people were at

home looking after kids not in school or day care.

But our estimates also suggest a role for FPUC and PUA in increasing unemployment

and reducing employment. This role runs counter to the conventional wisdom that these

programs had a very small or even negligible effect on unemployment and employment (e.g.,

New York Times, Oct. 2021 and Wall Street Journal, Sept. 2021). Consider that it took five

months for the unemployment rate to increase 0.4 percentage point after the Great Recession

begin in 2007 following the onset of the global financial crisis. While the economy was in

recession in 2001, the unemployment rate increased by 1.2 percentage points in total.

The labor market effects of UI have been extensively studied. Research documents that

the duration of unemployment increases with the generosity of unemployment benefits, even

during periods of (historically familiar) labor market weakness. 2 Findings from studies of the

pandemic-era effects of UI diverged from this consensus in the early stages of the pandemic and

associated social distancing policies, through the fall and winter of 2020 (e.g., Ganong et al.,

2021a, and Dube, 2021) and the spring of 2021 (Ganong et al., 2021b).3

2
Krueger and Meyer (2002) and Schmieder and von Wachter (2016) provide reviews of this research.
3
To the best of our knowledge, Coombs et al. (2021) is the only other paper to study the effect of early withdrawal
of FPUC and PUA. Using anonymous bank transaction data from the financial services company Earnin, they find
that employment increased through August 6 by 4.4 percentage points among unemployed workers who were

5
While there are many potential explanations for this divergence, the unusual

circumstances of the pandemic likely play a role. For example, individuals’ labor market

behavior may be less sensitive to the generosity of UI benefits when they are concerned about

contracting a novel and serious illness. Particularly during the early months of the pandemic,

labor demand had collapsed, which also likely attenuated the relationship between UI benefit

generosity and unemployment duration.

In addition, direct cash payments from various government programs, along with less

opportunity to spend money due to social distancing measures, lead to substantially increased

household savings. Savings balances may affect the elasticity of unemployment duration with

respect to UI benefit generosity. We present some weakly suggestive evidence that states with a

relatively high share of households that are confident they can pay their mortgage or rent in the

following month may have been less responsive to the early termination of pandemic-era UI

benefits.

Our analysis suggests that the summer of 2021 could be interpreted as a period of healing

and of returning normalcy for the labor market. As mentioned earlier, the pandemic still had

substantial effects of labor supply. But the familiar relationship between unemployment duration

and UI benefit generosity may have been reasserting itself relative to its apparent relationship in

2020.

receiving benefits at the end of April 2021. They also found that income and spending fell. Our paper makes several
advances. We use CPS data that captures the labor market through August in our difference-and-difference models
and through September in our event studies. Earinin is an app used predominantly by low-income workers with low
access to credit, and Earinin data for California, Maryland, Nevada, Arizona, Oklahoma, and Mississippi have less
coverage of UI receipt because those states’ UI programs do not use direct deposit. The CPS data we use are
nationally representative. In a supplemental blog post, Arindrajit Dube, a coauthor of Coombs et al., uses regression
adjusted CPS data on the same set of states as Coombs et al. Through July, he finds job finding increased by roughly
25 percent.

6
Finally, we note that the welfare implications of the early termination of FPUC and PUA

are ambiguous. Decreases in unemployment and increases in employment may be welfare

enhancing in many cases. But we also present evidence that early termination increased financial

stress. Using HPS data, we find that the share of respondents who report that they had no

difficulty meeting expenses in the past seven days dropped by slightly more than two percentage

points, or about five percent of the average share from February-June 2021.

2. Background on FPUC and PUA

The Coronavirus Aid, Relief, and Economic Security (CARES) Act of March 2020

substantially increased the generosity and expanded the eligibility of UI benefits.4 It created two

new federally-funded benefit programs for unemployed workers. Pandemic Unemployment

Assistance (PUA) extended eligibility for unemployment benefits to workers who are typically

ineligible for standard, state benefits, such as self-employed workers, “gig economy” workers,

workers whose work history was insufficiently lengthy, and workers seeing part-time work,

among others.

In addition, the Federal Pandemic Unemployment Compensation (FPUC) program added

a $600 weekly, federally funded supplement on top of existing, standard UI benefits. In addition

to regular UI recipients, participants in PUA were also eligible.5 Between April and July 2020,

76 percent of eligible workers were eligible for benefits that exceeded their lost wages, with a

median statutory replacement rate of 145 percent (Ganong, Noel, and Vavra, 2020).

4
Other provisions included providing full federal funding to the permanent Extended Benefits (EB) and providing
extra weeks of emergency federal benefits through the Pandemic Emergency Unemployment Compensation (PEUC)
program.
5
PEUC and EB participants were eligible for FPUC, as well.

7
FPUC originally lasted through July 31, 2020. The supplement then expired for one

week. On August 8, the Trump administration created the Lost Wages Assistance program

through an executive order that provided a federally funded $300 weekly UI benefit supplement.

The program was administered haphazardly by the states and lasted for several weeks. In

December 2020, President Trump signed the Consolidated Appropriations Act of 2021 that

revived FPUC and PUA through March 2021, with FPUC supplements of $300.

In March 2021, President Biden signed the American Rescue Plan, which further

extended FPUC (at $300 per week) and PUA through September 6, 2021. Strengthening labor

markets and concern by business about “worker shortages” led 18 states to opt out of both PUA

and FPUC in June. Two states opted out of both programs in July. Four states opted out of FPUC

in June but continued participating in PUA. Indiana and Maryland filed to terminate FPUC and

PUA early, but were required by court order to continue paying benefits for both programs.

Twenty four states and the District of Columbia participated in both programs until September

(Congressional Research Service 2021).

3. Data Sources

In this section, we briefly describe our data on state policies pertaining to FPUC and

PUA, our data on labor market flows, data from the Household Pulse Survey, and our pandemic-

related control variables.

FPUC and PUA

Data on state policies pertaining to FPUC and PUA come from state government

websites and the Congressional Research Service (2021). We include only states ending both

PUA and FPUC in June or states ending neither PUA nor FPUC before September in our main

8
analysis. We do not include Alaska, Arizona, Florida, or Ohio, since these four states ended

FPUC but not PUA before September. We also do not include Maryland or Indiana, since these

states initially stopped providing FPUC and PUA, but were later required to resume paying these

benefits during our sample period. Finally, we do not include Louisiana and Tennessee because

these states withdrew from both FPUC and PUA in July of 2021. We present a list of states in

our treatment and control groups in Table 1.

Current Population Survey

Our data on unemployment to employment transitions come from the Current Population

Survey (CPS) provided by the Integrated Public Use Microdata Series. Respondents are included

in the CPS for four consecutive months, then are not interviewed for the next eight months.

Respondents are then included in the CPS again for the next four months. We exploit this 4-8-4

panel structure of the CPS to identify individual transitions from unemployment to employment

in consecutive months.

A concern with the matched CPS is that mismeasurement of labor force status may lead

to spurious transitions out of unemployment (Abowd and Zellner, 1985). To address this, we

implement a recoding procedure used, for example, in Rothstein (2011), Farber, Rothstein, and

Valletta (2015), and Petrosky-Nadau and Valleta (2021). For individuals who transition out of

unemployment and into employment or nonparticipation in one month, but then return to

unemployment in the following month (i.e., U-E-U or U-N-U), we consider the transition

spurious, and recode the respondent as having been unemployed for each of the three months.

This requires us to observe respondents in three consecutive months of the four-month CPS

rotation, reducing the size of the matched sample by approximately one-third.

9
Household Pulse Survey

Data on difficulty paying expenses come from the weekly Household Pulse Survey,

administered by the U.S. Census Bureau. The survey is designed to provide real-time estimates

of impacts of the Covid-19 pandemic and government response programs on households.

The HPS asks households, “In the last 7 days, how difficult has it been for your

household to pay for usual household expenses, including but not limited to food, rent or

mortgage, car payments, medical expenses, student loans, and so on?” Households are given a

variety of responses. We create a variable, “not difficult to pay expenses,” that equals one if the

respondent selects “not difficult at all” and equal to zero if they select “a little difficult,”

“somewhat difficult,” or very difficult.”

We match survey periods in the HPS to survey months in the CPS depending on whether

the survey period in the HPS overlaps the CPS reference week (containing the 12th day of each

month) or extends into multiple months. If the survey period overlaps multiple months, but does

not extend into the reference week for the following month, we assign it to the previous month.

If the survey period overlaps multiple months and extends into the reference week for the

following month, we assign it to the following month. We then calculate the share of people in

each state-month assigned a value of one for our “not difficult to pay expenses” variable.

Pandemic-related control variables

We incorporate data on the severity of restrictions designed to combat Covid-19 and the

number of new cumulative Covid-19 cases that may be relevant as control variables. These data

come from the Oxford Covid-19 Government Response Tracker (OxCGRT), which have been

used in recent papers studying Covid-19 and the labor market (e.g., Lee et al., 2021, and Agrawal

et al., 2021). An index records the strictness of social distancing policies that primarily restrict

10
people’s behavior including restrictions on gatherings, canceling public events, closing

workplaces, restrictions on public transport, and school closures.6 OxCGRT calculates this index

for all 50 states and the District of Columbia daily. We average the daily values of the index for

each state to create a monthly stringency index and calculate the total number of new Covid-19

cases each month from February – August 2021.

4. Empirical Strategy

In this section, we briefly discuss our summary statistics of unadjusted data, and briefly

describe our difference-in-difference and event studies models, and our simple counterfactual

exercise.

Unadjusted differences

To provide a transparent overview of the relationships we explore, we first present a table

(Table 2) of unadjusted, raw means from our various data sources. We calculate mean values by

“pre” and “post” period and “treatment” and “control” group. For example, we present the mean

unemployment-to-employment flow for states that did and did not withdraw early from FPUC

and PUA, both for February-June 2021 and for July-August 2021. From this table, one can

calculate a simple, unadjusted difference-in-difference-style estimate of the effect of early

withdrawal. (For unemployed workers ages 25-54, this effect is 14.2 percentage points; see

Column 4.)

6
The construction of the stringency index is described in Hallas et al. (2021). The specific indicators in the
stringency index include school, workplace, and public transportation closing, canceled public events, restrictions on
gathering, shelter-in-place orders, restrictions on movement between cities or regions, restrictions on international
travel, and public information campaigns. More details on how the index is calculated are available here. The data
are frequently revised, and the data used in our analysis were last updated November 24, 2021.

11
Difference-in-differences

We then estimate transparent difference-in-differences models:

𝑦𝑖,𝑠,𝑚 = 𝛼 + 𝛽1 ∗ 𝑒𝑎𝑟𝑙𝑦 𝑒𝑥𝑖𝑡𝑠 + 𝛽2 ∗ 𝐽𝑢𝑙𝑦 𝑜𝑟 𝐴𝑢𝑔𝑢𝑠𝑡𝑚

+ 𝛽3 ∗ (𝑒𝑎𝑟𝑙𝑦 𝑒𝑥𝑖𝑡𝑠 ∗ 𝐽𝑢𝑙𝑦 𝑜𝑟 𝐴𝑢𝑔𝑢𝑠𝑡𝑚 ) + 𝛿1 ∗ ln(𝐶𝑜𝑣𝑖𝑑 𝑐𝑎𝑠𝑒𝑠𝑠,𝑚 ) (1)

+ 𝛿2 ∗ 𝑠𝑜𝑐𝑖𝑎𝑙 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑖𝑛𝑔 𝑠𝑡𝑟𝑖𝑛𝑔𝑒𝑛𝑐𝑦𝑠,𝑚 + 𝑋𝑖,𝑠,𝑚 𝛾 + 𝜖𝑖,𝑠,𝑚 ,

where 𝑦𝑖,𝑠,𝑚 is an indicator variable for whether an unemployed individual, in month 𝑚 − 1,

living in state 𝑠 enters into employment in month 𝑚. 𝐸𝑎𝑟𝑙𝑦 𝑒𝑥𝑖𝑡𝑠 is an indicator variable for

whether the state ended participation in both FPUC and PUA in June 2021; 𝐽𝑢𝑙𝑦 𝑜𝑟 𝐴𝑢𝑔𝑢𝑠𝑡𝑚 is

an indicator equal to zero from February – June 2021 and equal to one in July and August 2021.

The vector 𝑋𝑖 includes state, month, age, and education fixed effects. We describe

𝐶𝑜𝑣𝑖𝑑 𝑐𝑎𝑠𝑒𝑠𝑠,𝑚 and 𝑠𝑜𝑐𝑖𝑎𝑙 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑖𝑛𝑔 𝑠𝑡𝑟𝑖𝑛𝑔𝑒𝑛𝑐𝑦𝑠,𝑚 in Section 3. The state fixed effects

control for time-invariant state-level factors. Given the short estimation window, these state

effects probably go some distance toward controlling for attitudes about Covid and behavior in

the face of Covid.

Event studies

We use event study models to capture the dynamic effect of the early withdrawal from

FPUC and PUA, as well as to test whether the two groups of states were on similar trends prior

to their divergent UI policies. In particular, we estimate the following equation:

12
𝑦𝑖,𝑠,𝑚 = 𝛼 + ∑ 𝜗𝑚 ∗ (𝑒𝑎𝑟𝑙𝑦 𝑒𝑥𝑖𝑡𝑠 ∗ 𝜑𝑚 ) + 𝛿1 ∗ ln(𝐶𝑜𝑣𝑖𝑑 𝑐𝑎𝑠𝑒𝑠𝑠,𝑚 )
𝑚≠𝐽𝑢𝑛𝑒
(2)
+ 𝛿2 ∗ 𝑠𝑜𝑐𝑖𝑎𝑙 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑖𝑛𝑔 𝑠𝑡𝑟𝑖𝑛𝑔𝑒𝑛𝑐𝑦𝑠,𝑚 + 𝑋𝑖,𝑠,𝑚 𝛾 + 𝜀𝑖,𝑠,𝑚 ,

where 𝜗𝑚 is a parameter that captures the dynamic effect of the policy, and 𝜑𝑚 is a indicator

variable for each of the eight months in our estimation sample. We define other variables as in

equation (1).

Counterfactual

We estimate the effect on U-to-E transitions of early withdrawal of FPUC and PUA for

July and August, and we use those estimates in a simple counterfactual exercise to study what

the unemployment rate and employment-population ratio might have been in the states that did

not withdraw early had they instead done so, and for the economy as a whole.

To accomplish this exercise, we first estimate a counterfactual employment level

according to the following formula:

𝑐
𝑒𝑚 𝑎
= 𝑒𝑚 + 𝛽̂ ∗ 𝑢𝑚−1
𝑎
. (3)

In this formula, 𝑒𝑚
𝑐
is the counterfactual level of employment in month 𝑚, 𝑒𝑚
𝑎
is the

actual level of employment in month 𝑚, 𝑢𝑚−1
𝑎
is the actual level of unemployment in the

previous month, and 𝛽̂ is the estimate of the effect on U-to-E transitions of early withdrawal

from FPUC and PUA.

13
After having constructed a counterfactual employment level for July and for August, we

then recalculate the unemployment and employment rates, assuming the levels of the labor force

and the population are unaffected by the policy change. This simplifying assumption raises a

potential concern: a priori, one might expect that this policy change affected other labor market

flows. In Table A3, we present estimates of the effect of early withdrawal from FPUC and PUA

on all six flows, using equation (1). Only two of the six models have precisely estimated

coefficients, transitions from unemployment to employment and from unemployment to

nonparticipation. We constructed counterfactual estimates using all six estimates, and the results

are qualitatively similar.

A second simplifying assumption we employ is not to allow the effect to compound from

July and into August. We make this assumption because allowing compounding would require

using imprecisely estimated coefficients for all other flows, as we discussed earlier. The decision

not to allow for compounding — i.e., to treat July and August counterfactuals as independent

events — biases our counterfactual estimates downward.

Finally, our estimates are biased downward because we implicitly assume there are no

effects from ending only FPUC but not PUA as we drop the states with this policy mix from our

analysis and use their actual employment and unemployment levels in the national

counterfactual. In Table A5, we present evidence that suggests ending FPUC while continuing

PUA led to a smaller increase in transitions from unemployment to employment than opting out

of both programs.

14
5. Results

To begin, which states terminated FPUC and PUA early, and which did not? Table 1 lists

the states in each category. A total of 18 states eliminated both programs fully beginning in June

2021 – mostly in the South and the Great Plains. In contrast, 24 states plus the District of

Columbia eliminated neither program.

Table A1 presents summary data on the relevant characteristics of these two sets of

states. This table shows means and standard deviations on the variables we use in the analysis for

states that did and did not terminate FPUC and PUA, and during time periods of February-June

and July-August 2021, both before and after the early terminations went into effect respectively.

These data indicate a sharp rise in transitions from unemployment to employment in the

states ending extended benefits after the early withdrawal was implemented, while no such rise is

observed in the states that maintained such benefits until September. We also note several other

differences between the two sets of states that could likely affect these transition rates, including:

higher stringency levels in the earlier period in states not ending benefits, but also declining

stringency of restrictions in both sets of states (and especially those not ending benefits);

increases in new Covid cases in July and August, especially in states ending extended benefits;

and higher education levels in the states not ending FPUC and PUA.

We offer further evidence on transitions from unemployment to employment and their

determinants, in these states and time periods, in Table 2. Here, we again present the mean

transitions in these states and time periods among the unemployed, as well as the means of

social-distancing stringency, new Covid cases, and the share of respondents in the HPS reporting

having no difficulty meeting expenses in the previous week. We also present the changes over

time in these measures (Column 3), the “unadjusted differences in differences” between these

15
states and time periods in such transitions (Column 4), and the percent changes relative to

baseline in these differences (Column 5). The goal of Table 2 is to be as transparent as possible

with the data.

The summary data indicate a fairly sharp increase of over 13 percentage points in

transitions from unemployment to employment in the states ending extended benefits, but not in

those allowing them to remain in place. Relative to baseline transitions, the increase is about 60

percent in the states ending benefits, and a four percent decrease those not ending them.

The other data in Table 2 also indicate the relatively greater declines in stringency of

Covid rules in the states not ending benefits — which (all else equal) should have raised their

transition rates — and the relatively worse increases in Covid cases in the states ending benefits,

which should have dampened the increasing employment transitions observed there. In other

words, both of these factors might have limited the observed relative increases in transition rates

observed in states ending extended benefits.

Finally, we show a 2.4 percentage point “unadjusted difference in difference” decrease in

the share of households reporting that they do not have difficulty meeting their expenses. This

effect is unsurprising, since the expiration of FPUC and PUA reduced household income.

Difference-in-differences results

We present difference-in-difference estimates of the impacts of ending extending benefits

on transitions from unemployment to employment in Tables 3. The top row of Table 3 presents

the impact of ending extended UI benefits for four different groups of unemployed workers: (1)

those ages 25-54; (2) those ages 25-54 who do not have college degrees; (3) those in this age

group whose previous jobs were in leisure/hospitality or retail trade (4) those ages 16-64.

16
Focusing on workers ages 25-54 limits the confounding effects of changes in enrollment

in higher education or in retirements in response to the pandemic — early retirements are likely a

major pandemic-era labor market development (Faria e Castro, 2021), so this group is especially

useful. Studying workers without college degrees or from the leisure and retail sectors enable us

to focus on the workers hardest hit by the pandemic, and somewhat slower to recover from it (in

terms of employment).

For each category of workers, we also present two estimates — one with no controls for

social distancing stringency or Covid caseload and one that includes them. All equations control

for month, state, age, and education.

Our results indicate a 13.8 percentage point increase in transitions from unemployment to

employment among prime-age workers before controlling for stringency and new cases, rising to

14.4 percentage points after controlling for them. We estimate an 11.6 percentage point increase

in transitions among prime-age workers without college degrees (and with controls), and an 11.4

percentage point increase among prime-age workers who last worked in the leisure and

hospitality and retail industries. Among workers ages 16-64, the results indicate a 13.3

percentage point increase in transitions, again including control variables. Six of the eight

estimates are statistically significant; both estimates on the leisure/hospitality and retail sample

are not.7

7
These results are estimated on CPS samples in which we attempt to correct for potentially spurious labor force
status transitions using the procedure discussed in Section 3. Our results for prime-age workers are similar in a
sample in which we do not incorporate this correction (14 vs. 12 percentage points in the specification with control
variables), as well as among prime-age workers without college degrees (12 vs. 11 percentage points). Among
prime-age workers last employed in the leisure and hospitality sector, the correction renders the coefficient smaller
(11 vs. 16 percentage points) and statistically insignificant. Among workers ages 16-64, the correction increases the
magnitude of the coefficient (13 vs. 8 percentage points).

17
Event study results

In Figure 1, we present event studies for each of our four sample groups, which trace out

the dynamic effect of early withdrawal from FPUC and PUA on U-to-E transitions. For each, the

monthly coefficients from February through May (the “pre period”) are centered on or near zero,

and are imprecisely estimated. This estimated effect mitigates concerns about divergent “pre-

period” trends and supports a causal interpretation of our estimates.

It is also reassuring that the magnitude of the September coefficient falls closer to zero

and that the coefficient estimate is statistically insignificant. Since both the “treatment” and

“control” states had ended pandemic-era UI programs in September, this finding strengthens

confidence in our results for July and August.

The event studies confirm what we observe in the difference-in-difference models —

namely, that transitions to employment increased in the states ending extended benefits, relative

to those that did not, and only in July and August of 2021. Table A2 presents the estimated

coefficients and standard errors. For July and August, the coefficients are positive and for our

main sample, prime-age workers, are precisely estimated. Among prime-age workers without

college degrees, the coefficient for July is statistically significant and of qualitatively similar

magnitude (11.8 percentage points), but the coefficient for August is statistically insignificant.

The coefficients in the leisure and hospitality and retail sectors are statistically insignificant.

Among unemployed workers ages 16-64, the magnitude of the effect is similar in August (14.9

percentage points) and July (14.1 percentage points), and both coefficients are statistically

significant.

18
Counterfactual exercise

While the magnitudes of the estimated impacts on transitions to employment are large

(relative to the baseline of about 21 percent per month), by how much would ending extended

benefits earlier in the 24 states and D.C. that did not do so have reduced unemployment rates and

raised employment-population ratios? To answer this question, we conduct a simple

counterfactual exercise. We described the exercise in Section 4. Here, we present and discuss the

results.

We provide estimates of these counterfactuals in Tables 4 and 5, for each of the four

groups considered. In Table 4, we estimate these counterfactuals only in the 24 states and D.C.

that did not terminate extended benefits and among each of our four samples; in Table 5, we do

so for the United States economy overall.

Table 4 indicates that the average unemployment rates in the 24 states and D.C. would

have declined from 5.8 to 5.0 percent in July, and from 5.7 to 5.0 percent in August, among

workers ages 25-54. As expected from the earlier point estimates of impacts, the implied

estimates show slightly larger decreases in unemployment rates for workers ages 16-64 of

roughly one percentage point, and larger and more variable changes for non-college educated

workers or workers in the leisure hospitality and retail sectors.

The estimated counterfactuals in employment-population ratios are a bit smaller for all

four groups. Among our main sample of workers ages 25-54, employment rates would have risen

in these states from 77.0 to 77.7 percent in July and from 77.3 to 77.9 percent in August. Among

the broadest group of workers (ages 16-64), employment rates would have risen in these states

from 69.8 to 70.5 percent in July and from 69.4 to 70.1 percent in August. The estimates of

19
counterfactuals for the other two subgroups are again a bit larger for non-college-educated

workers and smaller for workers previously in leisure and hospitality and retail.

By how much would unemployment rates in the United States overall have fallen, had all

states ended extended benefits early, and by how much would employment-population ratios

have risen? Here the implied changes are smaller, as expected, and we present them in Table 5.

Using estimates from the sample of prime-age workers, the national unemployment rate

would have declined in July, from 5.7 to 5.4 percent, or 0.3 percentage point, and in August it

would have also declined by 0.3 percentage point (from 5.3 to 5.0). Similarly, the national

employment rate in the population would have risen from 58.7 to 58.9 percent in July (an

increase of 0.2 percentage point) and from 58.6 to 58.7 in August.8

Household financial conditions

In Table 6, we present estimates of equation (1), but instead of studying U-to-E

transitions these regressions examine the effect of early withdrawal of FPUC and PUA on our

“not difficult to pay expenses” variable. For all four samples, the expiration of generous UI

benefits reduced the share of households who reported not having trouble meeting their

expenses. The coefficient magnitudes range from 2.2 to 2.5 percentage points, or about five

percent of the average value of the variable from February-June 2021.

Household financial conditions and U-to-E transitions

In Table A4, present results from our investigation into whether a household’s financial

health affects its sensitivity to the expiration of benefits. Using data from the HPS, we compute

8
These results are estimated on CPS samples in which we attempt to correct for potentially spurious labor force
status transitions using the procedure discussed in Section 3. The results using the coefficient estimated on a sample
of prime-age workers are very similar when not correcting for potentially spurious transitions. Similar to estimates
of equation (1), the sample with labor-force-status transitions leads to larger counterfactual vs. actual differences
when using the coefficient estimated on workers ages 16-64.

20
the share of respondents in each state-month that report that they are “very confident” that they

can meet their next month’s rent or mortgage payment or know that their payment will be

deferred. In each month, we rank states based on these shares, and we create an indicator

variable equal to one if a state is in the top quintile of the distribution of the variable (in each

month) and equal to zero otherwise. We then interact this variable with the treatment variable

from equation (1). The estimated coefficient on each of the four samples is negative, but is only

statistically significantly different from zero among prime-age workers without college degrees.

This provides suggestive — but not strong — evidence that the unemployment duration

of households in relatively stronger financial shape may be relatively less sensitive to the

generosity of unemployment benefits. This pattern is relevant to the current U.S. economic

context because households have accumulated over $2 trillion of excess savings (Holtz-Eakin,

2021). It is possible that these savings have muted to at least some degree the velocity of labor

market flows following the expiration of pandemic-era UI benefits.

Expanded eligibility, or expanded generosity?

Alaska, Arizona, Florida, and Ohio each ended participation in FPUC, which expanded

the generosity of UI benefits, while continuing to participate in PUA, which expanded eligibility

for benefits. Arizona ended participation in FPUC in July, so we focus on the remaining three

states to study the effect of ending expanded generosity while maintaining expanded eligibility

on transitions from unemployment to employment.

We report results in Table A5. Among prime-age workers, ending only FPUC increased

transitions from unemployment to employment by 8.3 percentage points (in the model with

controls). The magnitude is substantially less than the 14.4 percentage point effect of ending both

programs in June, reported in Table 3. We find qualitatively similar results among workers ages

21
16-64. Unlike the effect of ending participation in both programs, prime-age workers without

college degrees do not seem to have had transitions from unemployment to employment affected

by the expiration of only FPUC. It may be that expanded eligibility is a larger factor for these

workers than expanded generosity. Among prime-age leisure and hospitality and retail workers,

the coefficients are statistically insignificant (as they were in the results presented in Table 3)

with negative signs (unlike in Table 3).

Taken as a whole, these results suggest that both expanded eligibility and expanded

generosity may have slowed transitions from unemployment to employment.

6. Conclusion

In this paper, we provide estimates of the impacts of the early termination of pandemic-

era UI benefits that expanded their generosity (FPUC) and the groups of workers eligible for

benefits (PUA). Using Current Population Survey data, we present difference-in-difference

estimates that the flow of prime-age unemployed workers into employment increased by around

14 percentage points following early termination, or about two-thirds the size of the

unemployed-to-employed flow among control states during the February-June 2021 period. We

construct a counterfactual scenario for the labor market under the assumption that all states

ended FPUC and PUA in June. In this scenario, the unemployment rates in each of July and

August would have been around 0.3 percentage point lower, and the employment-population

ratio would have been 0.1-0.2 percentage point higher. We also present some weakly suggestive

evidence that states with households with greater confidence in their ability to meet key expenses

may have been less sensitive to the expiration of benefits. We also present evidence suggesting

that both expanded eligibility and expanded generosity may have slowed transitions from

22
unemployment to employment. Finally, we present evidence that early termination reduced the

share of households that had no difficulty meeting expenses by five percent. The welfare

implications of the early termination of FPUC and PUA are therefore ambiguous.

23
References

Abowd, John M .and Arnold Zellner, “Estimating Gross Labor-Force Flows,” Journal of
Business and Economic Statistics, vol. 3, no. 3, 1985.

Agrawal, Virat, Jonathan H. Cantor, Neeraj Sood, and Christopher M. Whaley, “The Impact of
the Covid-19 Pandemic and Policy Responses on Excess Mortality,” No. w28930,
National Bureau of Economic Research, June 2021.

Coombs, Kyle, Arindrajit Dube, Suresh Naidu, Calvin Jahnke, Raymond Kluender, and Michael
Stepner, “Early Withdrawal of Pandemic Unemployment Insurance: Effects on Earnings,
Employment and Consumption,” unpublished manuscript, August 20, 2021.

Congressional Research Service (2021) States Opting Out of COVID-19 Unemployment
Insurance (UI) Agreements IN11679. Last updated August 20, 2021.

Dube, Arindrajit. 2021. “Aggregate Employment Effects of Unemployment Benefits During
Deep Downturns: Evidence from the Expiration of the Federal Pandemic Unemployment
Compensation.” No. w28470. National Bureau of Economic Research. February 2021.

Farber, Henry S., Jesse Rothstein, and Robert G. Valletta. "The Effect of Extended
Unemployment Insurance Benefits: Evidence from the 2012-2013 Phase-Out." American
Economic Review 105, no. 5 (2015): 171-76.

Faria e Castro, Miguel, “The COVID Retirement Boom,” Economic Synopses, No. 25, 2021.

Ganong, Peter, Fiona Greig, Max Liebeskind, Pascal Noel, Daniel Sullivan, and Joseph Vavra.
2021. "Spending and Job Search Impacts of Expanded Unemployment Benefits:
Evidence from Administrative Micro Data." University of Chicago, Becker Friedman
Institute for Economics Working Paper 2021-19. February 2021.

Ganong, Peter, Fiona Greig, Pascal Noel, Daniel M. Sullivan, and Joseph Vavra, “Micro and
Macro Disincentive Effects of Expanded Unemployment Benefits,” July 29, 2021.

Ganong, Peter, Pascal Noel, and Joseph Vavra, “US Unemployment Insurance Replacement
Rates During the Pandemic,” Journal of Public Economics, November 2020.

Hallas, Laura, Ariq Hatibie, Saptarshi Majumdar, Monika Pyarali, Rachelle Koch, Andrew
Wood and Thomas Hale. “Variation in US States’ Responses to COVID-19 3.0.”
Blavatnik School of Government. Working Paper. May 2021.

Holtz-Eakin, Douglas, “Should You Fear Fiscal Drag?” The Daily Dish, American Action
Forum, November 23, 2021.

Krueger, Alan B. and Bruce D. Meyer, “Labor Supply Effects of Social Insurance,” Handbook of
Public Economics, vol. 4, 2002.

24
Lee, Sang Yoon Tim, Minsung Park, and Yongseok Shin. Hit Harder, Recover Slower? Unequal
Employment Effects of the Covid-19 Shock.” No. w28354. National Bureau of Economic
Research, January 2021.

Petrosky-Nadeau, Nicolas, and Robert G. Valletta. "UI Generosity and Job Acceptance: Effects
of the 2020 CARES Act." No. 14454. IZA Discussion Paper Series. June 2021.

Rothstein, Jesse. “Unemployment Insurance and Job Search in the Great Recession.” Brookings
Papers on Economic Activity, Fall 2011.

Schmieder, Johannes F. and Till von Wachter, “The Effects of Unemployment Insurance
Benefits: New Evidence and Interpretation,” Annual Review of Economics, vol. 8,
October 2016.

25
Tables and Figures

Table 1: List of States Ending Both FPUC and PUA In June 2021 and Ending
Neither Program Before September 2021
States ending both FPUC and PUA States ending neither FPUC nor PUA
Alabama California
Arkansas Colorado
Georgia Connecticut
Idaho Delaware
Iowa District of Columbia
Mississippi Hawaii
Missouri Illinois
Montana Kansas
Nebraska Kentucky
New Hampshire Maine
North Dakota Massachusetts
Oklahoma Michigan
South Carolina Minnesota
South Dakota Nevada
Texas New Jersey
Utah New Mexico
West Virginia New York
Wyoming North Carolina
Oregon
Pennsylvania
Rhode Island
Vermont
Virginia
Washington
Wisconsin
Notes: States ending both FPUC and PUA participation in June 2021 do not include Tennessee and
Louisiana, which terminated payments in July 2021, or Maryland and Indiana which terminated
benefits but where a court order reinstated payments. Alaska, Arizona, Florida, and Ohio terminated
participation in FPUC early, but not PUA.

26
Table 2: Unadjusted Differences Across Sets of States
(1) (2) (3) (4) (5)
Change Relative
Change Relative
Feb-Jun 2021 Jul-Aug 2021 Change to Non-Ending
to Baseline
States
Unemployed to Employed
Ending Neither PUA or FPUC 21.02 20.27 -0.75 -3.57
Ending Both PUA and FPUC 22.43 35.83 13.40 14.15 59.74
Stringency Index
Ending Neither PUA or FPUC 46.51 20.41 -26.10 -56.12
Ending Both PUA and FPUC 31.45 21.92 -9.53 16.57 -30.30
New Covid Cases (1000s)
Ending Neither PUA or FPUC 73.61 105.7 32.09 43.59
Ending Both PUA and FPUC 70.37 159.8 89.43 57.34 127.09
Share Not Having Any Difficulty
Paying Expenses in the Past Week
Ending Neither PUA or FPUC 47.75 51.03 3.28 6.87
Ending Both PUA and FPUC 44.46 45.30 0.84 -2.44 1.89
Notes: This table reports simple differences for our sample of individuals ages 25-54 living in states that ended participation in both PUA
and FPUC in June 2021 or ended participation in neither PUA nor FPUC before September 2021. Entries for unemployed to employed
summarize data from the Current Population Survey (CPS). The stringency index and monthly new COVID-19 case measures come from
OxCGRT. Data on the share not having difficulty meeting expenses in the last seven days come from the Household Pulse Survey.
Column 1 reports the average value between February and June 2021 for each row, column 2 reports the average value between July and
August 2021, and column 3 reports the difference between the two. Column 4 reports the change in the average value for each row
relative to the relevant non-ending value. Column 5 reports the change in the average value from February - June to July - August.
Averages are weighted by state population.

27
Table 3: Effects of Early Expiration of FPUC/PUA on the Probability of Employment in the Current
Month
(1) (2) (3) (4) (5) (6) (7) (8)

Ages 25-54; Not Ages 25-54; LH
Sample Ages 25-54 Ages 16-64
Graduated College & Retail

Ending Both FPUC and PUA In 13.810*** 14.385*** 10.242** 11.561** 10.757 11.374 11.766*** 13.347***
June X Post June 2021 (3.428) (4.523) (4.336) (5.196) (8.162) (8.912) (2.388) (3.488)
State Level Stringency Index -0.163 -0.353** -0.044 -0.160
from OxCGRT (0.136) (0.149) (0.297) (0.116)
2.797* 5.760*** 0.132 1.310
Ln(New Covid Cases)
(1.480) (1.264) (2.434) (1.092)

Observations 4,419 4,419 3,237 3,237 920 920 7,219 7,219
Notes: This table reports regression results measuring the effect of ending participation in both FPUC and PUA early on
unemployment to employment transitions. The sample is from the Basic Monthly CPS from February - August 2021. Columns 1
and 2 include all individuals ages 25-54. Columns 3 and 4 include all individuals ages 25-54 who have not graduated college.
Columns 5 and 6 include all individuals ages 25-54 who last worked in the leisure and hospitality or retail industries. Columns 7
and 8 include all individuals ages 16-64. All specifications include month, state, age, and education fixed effects. Standard errors
are clustered at the state level. *** p<0.01, ** p<0.05, * p<0.1

28
Table 4: Counterfactual Analysis of Ending FPUC/PUA: Estimated Effect on States Not Ending
Programs
(1) (2) (3) (4) (5)
July August
June Actual July Actual August Actual
Counterfactual Counterfactual

Panel A: Among Workers Ages 25-54
Employment-Population Ratio 76.9 77.0 77.7 77.3 77.9
Unemployment Rate 6.0 5.8 5.0 5.7 5.0

Panel B: Among Workers Ages 25-54. Not Graduated College
Employment-Population Ratio 70.4 70.9 71.7 71.6 71.8
Unemployment Rate 8.2 7.5 6.6 7.4 7.1

Panel C: Among Workers Ages 25-54. Leisure Hospitality and Retail
Employment-Population Ratio 89.3 90.9 90.0 90.2 91.8
Unemployment Rate 10.1 8.7 9.6 9.2 7.6

Panel D: Among Workers Ages 16-64
Employment-Population Ratio 69.3 69.8 70.5 69.4 70.1
Unemployment Rate 6.7 6.6 5.6 6.1 5.1

Notes: This table displays results of a counterfactual exercise where we estimate the effects on the level of employment
and unemployment, the employment population ratio and the unemployment rate if FPUC and PUA were ended in all
states in June 2021. The "Actual" series come from the Basic Monthly CPS. We construct July counterfactual employment
and unemployment levels by applying the transition rate from unemployment to employment calculated from the event
study regressions to the actual employment data for June. We construct August counterfactual employment and
unemployment levels by applying the transition rate from unemployment to employment calculated from the event study
regressions to the actual employment data for July. We then use these to calculate the employment-population ratio and
unemployment rate.

29
Table 5: Counterfactual Analysis of Ending FPUC/PUA: National Estimates
(1) (2) (3) (4) (5)
July August
June Actual July Actual August Actual
Counterfactual Counterfactual

Panel A: Using Effect Estimated on a Sample of Workers Ages 25-54
Employment-Population Ratio 58.3 58.7 58.9 58.6 58.7
Unemployment Rate 6.1 5.7 5.4 5.3 5.0

Panel B: Using Effect Estimated on a Sample of Workers Ages 25-54. Not Graduated
College
Employment-Population Ratio 58.3 58.7 58.8 58.6 58.6
Unemployment Rate 6.1 5.7 5.5 5.3 5.2

Panel C: Using Effect Estimated on a Sample of Workers Ages 25-54. Leisure
Hospitality and Retail
Employment-Population Ratio 58.3 58.7 58.7 58.6 58.6
Unemployment Rate 6.1 5.7 5.7 5.3 5.2

Panel D: Using Effect Estimated on a Sample of Workers Ages 16-64
Employment-Population Ratio 58.3 58.7 59.0 58.6 58.9
Unemployment Rate 6.1 5.7 5.2 5.3 4.8

30
Table 6: Effects of Early Expiration of FPUC/PUA on the Share of HPS Respondents Who Report No
Difficulty Paying Expenses in the Past Seven Days
(1) (2) (3) (4) (5) (6) (7) (8)

Ages 25-54; Not Ages 25-54; LH &
Sample Ages 25-54 Ages 16-64
Graduated College Retail

Ending Both FPUC and PUA In -2.535*** -2.231** -2.532*** -2.218** -2.545*** -2.299*** -2.522*** -2.221**
June X Post June 2021 (0.690) (0.837) (0.684) (0.827) (0.685) (0.835) (0.696) (0.845)
State Level Stringency Index from -0.029 -0.031 -0.024 -0.029
OxCGRT (0.025) (0.025) (0.026) (0.026)
0.215 0.244 0.181 0.212
Ln(New Covid Cases)
(0.368) (0.361) (0.377) (0.370)

Observations 103,822 103,822 60,206 60,206 14,009 14,009 169,933 169,933
Notes: This table reports regression results measuring the effect of ending participation in both FPUC and PUA early on the share of
Household Pulse Survey respondents who say they have no difficulty meeting expenses in the past seven days. The sample is from the Basic
Monthly CPS from February - August 2021 who are not in rotations 1, 4, 5, or 8. Columns 1 and 2 include all individuals ages 25-54.
Columns 3 and 4 include all individuals ages 25-54 who have not graduated college. Columns 5 and 6 include all individuals ages 25-54 who
last worked in the leisure and hospitality or retail industries. Columns 7 and 8 include all individuals ages 16-64. All specifications include
month, state, age, and education fixed effects. Standard errors are clustered at the state level. *** p<0.01, ** p<0.05, * p<0.1

31
Figure 1: Event Studies of Changes in Monthly Transitions into Employment from Unemployment Following the June 2021 Expiration of FPUC and
PUA UI Benefits Extended to September 2021. This figure displays estimated coefficients from event study regressions measuring the effect of expiring UI
programs on individual transitions from unemployment to employment extended to September 2021 when the programs expired for all states. The samples are
individuals from the CPS who were unemployed in the previous month. Panel A includes all individuals ages 25-54. Panel B includes all individuals ages 25-54
who have not graduated college. Panel C includes all individuals ages 25-54 who last worked in the leisure and hospitality or retail industries. Panel D includes
all individuals ages 16-64. Regressions include state, month, age, and education fixed effects, as well as controls for new monthly state Covid-19 cases and an
index measuring the stringency of Covid-19-related restrictions. Error bars denote 95 percent confidence intervals around each estimated coefficient, which
measures the change in the transition rate relative to June 2021. Standard errors are clustered by state.

32
Appendix Tables

Table A1: Sample Summary Statistics: CPS Respondents Ages 25-54 Unemployed in the Prior
Month, Employed in the Current Month, and Not Unemployed the Following Month
(1) (2) (3) (4)
Years Feb-Jun 2021 Jul-Aug 2021 Feb-Jun 2021 Jul-Aug 2021
Sample Ending Both FPUC and PUA Ending Neither FPUC nor PUA
22.43 35.83 21.02 20.27
Unemployed to Employed Transition
(41.73) (48.02) (40.75) (40.23)

Stringency Index 31.45 21.92 46.51 20.41
(9.059) (6.070) (12.51) (6.532)

Thousands of New Covid Cases 70.37 159.8 73.61 105.7
(80.49) (169.5) (70.47) (132.7)

Age 37.99 37.02 38.37 37.86
(8.828) (8.591) (8.742) (8.788)

High School Degree 0.359 0.355 0.317 0.292
(0.480) (0.479) (0.466) (0.455)

Some College Education 0.322 0.217 0.284 0.310
(0.467) (0.413) (0.451) (0.463)

4-Year College Education or Greater 0.186 0.273 0.283 0.298
(0.390) (0.446) (0.451) (0.458)
Observations 1,081 341 2,228 769
Notes: This table reports summary statistics for our two sample groups. Columns 1 and 2 report averages and standard
deviations (in parenthesis) of each of the variables for a sample of individuals ages 25-54, who were unemployed in the previous
month, and can be linked across adjacent months in the CPS in states that ended both FPUC and PUA in June. Columns 3 and 4
report averages and standard deviations (in parenthesis) of each of the variables for individuals ages 25-54, who were
unemployed in the previous month, and can be linked across adjacent months in the CPS living in states that did not end
participation in FPUC or PUA before September. Entries for transitions into employment, age, and education summarize data
from the Current Population Survey (CPS). The stringency index and monthly new COVID-19 case measures come from the
Oxford Covid-19 Government Response Tracker (OxCGRT).

33
Table A2: Event Study Estimates of the Effect of Early Expiration of FPUC and PUA
on Transitions From Unemployment Into Employment Using June 2021 as the Base
Period
(1) (2) (3) (4)

Ages 25-54; Ages 25-54;
Sample Ages 25-54 Ages 16-64
No College LH & Retail

Ending Both FPUC and PUA In June X -4.876 -9.647 -18.152 -1.362
February (6.083) (6.069) (15.184) (5.156)
4.054 -0.309 -0.516 8.604*
Ending Both FPUC and PUA In June X March
(3.931) (4.280) (13.519) (4.930)
-4.684 -5.403 6.499 -0.451
Ending Both FPUC and PUA In June X April
(6.561) (6.431) (11.494) (3.405)
-1.142 -3.898 0.666 -0.682
Ending Both FPUC and PUA In June X May
(6.432) (6.266) (12.479) (4.911)
13.737*** 11.754** -9.443 14.083***
Ending Both FPUC and PUA In June X July
(4.863) (5.139) (12.456) (4.305)
12.759** 4.093 19.011 14.911***
Ending Both FPUC and PUA In June X August
(4.824) (7.999) (11.703) (4.254)
-0.176 -0.392** -0.101 -0.151
State Level Stringency Index from OxCGRT
(0.144) (0.151) (0.308) (0.123)
2.784* 5.718*** 1.000 1.471
Ln(New Covid Cases)
(1.562) (1.414) (2.179) (1.115)
Observations 4,419 3,237 920 7,219
Notes: This table reports regression results measuring the effect of ending participation in both FPUC and PUA
early on unemployment to employment transitions using the event study regression specifications in equation 2.
Column 1 includes all individuals ages 25-54. Column 2 includes individuals ages 25-54 who have not graduated
college. Column 3 includes individuals ages 25-54 who last worked in the leisure and hospitality or retail
industries. Column 4 includes all individuals ages 16-64. All specifications include month, state, age, and
education fixed effects. Standard errors are clustered at the state level. *** p<0.01, ** p<0.05, * p<0.1

34
Table A3: Effect of Early Expiration of FPUC and PUA on the Probability that Individuals Ages
25-54 Have Labor Market Transitions
(1) (2) (3) (4) (5) (6)
Dependent Variable UEtoE NILFtoE EtoUE EtoNILF UEtoNILF NILFtoUE
Ending Both FPUC and PUA Before In June 14.385*** 0.307 -0.283 -0.339 -9.399** -1.253
X Post June 2021 (4.523) (1.402) (0.185) (0.332) (4.567) (0.921)
-0.163 0.038 -0.002 -0.001 0.216* 0.020
State Level Stringency Index from OxCGRT
(0.136) (0.057) (0.009) (0.008) (0.121) (0.025)
2.797* -0.045 0.123 0.275*** -0.643 0.286
Ln(New Covid Cases)
(1.480) (0.333) (0.096) (0.074) (1.153) (0.361)

Observations 4,419 18,734 80,669 80,669 4,419 18,734
Notes: This table reports regression results measuring the effect of ending participation in both FPUC and PUA early on
labor market transitions. The sample is from the Basic Monthly CPS from February - August 2021, and consists of all
individuals ages 25-54. All specifications include month, state, age, and education fixed effects. Standard errors are clustered
at the state level. *** p<0.01, ** p<0.05, * p<0.1

35
Table A4: Effects of Early Expiration of FPUC and PUA on Transitions into Employment:
Controlling for Confidence in Ability to Pay Rent or Mortgage Next Month
(1) (2) (3) (4)

Ages 25-54; Ages 25-54;
Sample Ages 25-54 Ages 16-64
No College LH & Retail

Top Quintile Confident Mortgage or Rent
Ending Both FPUC and PUA In June X Post June 2021 -10.158 -19.901** -28.834 -7.385
X Top Quintile Confident Mortgage or Rent (6.901) (7.947) (27.286) (7.071)
15.538*** 13.118** 12.457 13.623***
Ending Both FPUC and PUA In June X Post June 2021
(4.876) (5.479) (8.934) (3.675)
Ending Both FPUC and PUA In June X Top Quintile 6.679 7.321 2.460 0.069
Confident Mortgage or Rent (4.838) (4.546) (20.881) (4.034)
Post June 2021 X Top Quintile Confident Mortgage or -1.941 0.410 -29.326* -5.322
Rent (7.439) (7.714) (15.504) (5.543)
2.927* 5.878*** 0.842 1.433
State Level Stringency Index from OxCGRT
(1.615) (1.348) (2.613) (1.168)
-0.183 -0.366** -0.019 -0.151
Ln(New Covid Cases)
(0.145) (0.157) (0.301) (0.122)
-1.647 -1.156 17.146 2.314
Top Quintile Confident Mortgage or Rent
(5.166) (6.401) (11.313) (3.325)
Observations 4,419 3,237 920 7,219
Notes: This table reports regression results measuring how the effect of ending participation in both FPUC and PUA
early on unemployment to employment transitions varies with the state share of individuals each month very confident
they can make mortgage or rent payments next month or who know their payments will be deferred. Column 1 includes
all individuals ages 25-54. Column 2 includes individuals ages 25-54 who have not graduated college. Column 3 includes
individuals ages 25-54 who last worked in the leisure and hospitality or retail industries. Column 4 includes all
individuals ages 16-64. All specifications include month, state, age, and education fixed effects. Standard errors are
clustered at the state level. *** p<0.01, ** p<0.05, * p<0.1

36
Table A5: Effects of Early Expiration Only FPUC on the Probability of Employment in the
Current Month
(1) (2) (3) (4) (5) (6) (7) (8)

Ages 25-54; Not Ages 25-54; LH
Sample Ages 25-54 Ages 16-64
Graduated College & Retail

Ending Only FPUC In June X Post 9.050*** 8.280*** 2.398 1.572 -7.068 -8.766 9.994* 9.406*
June (1.807) (2.085) (3.101) (2.117) (5.587) (5.342) (5.290) (4.986)
State Level Stringency Index from -0.087 -0.250 0.066 -0.069
OxCGRT (0.126) (0.163) (0.381) (0.114)
2.710 5.536*** 2.873 2.121
Ln(New Covid Cases)
(1.726) (1.577) (2.058) (1.337)

Observations 3,461 3,461 2,479 2,479 782 782 5,560 5,560
Notes: This table reports regression results measuring the effect of ending participation in FPUC but not PUA early on
unemployment to employment transitions. The sample is from the Basic Monthly CPS from February - August 2021.
Columns 1 and 2 include all individuals ages 25-54. Columns 3 and 4 include all individuals ages 25-54 who have not
graduated college. Columns 5 and 6 include all individuals ages 25-54 who last worked in the leisure and hospitality or retail
industries. Columns 7 and 8 include all individuals ages 16-64. All specifications include month, state, age, and education
fixed effects. Standard errors are clustered at the state level. *** p<0.01, ** p<0.05, * p<0.1

File and source

File: did-pandemic-unemployment-benefits-reduce-employment-evidence-from-early-state-level-expirations-in-june-2021.pdf
Size: 676,906 bytes
SHA-256: fa38d0ca551e12dd2f8b6dab6bec7ca079265829539480e9633df0a1a7a30a72
Our copy: did-pandemic-unemployment-benefits-reduce-employment-evidence-from-early-state-level-expirations-in-june-2021.pdf
Original: www.nber.org