Worker Heterogeneity, Selection, and Unemployment Dynamics in a Pandemic

• Published date: 01 February 2022
• Author: FEDERICO RAVENNA,CARL E. WALSH
• Date: 01 February 2022
• DOI: http://doi.org/10.1111/jmcb.12899

DOI: 10.1111/jmcb.12899

FEDERICO RAVENNA

CARL E. WALSH

Worker Heterogeneity, Selection, and

Unemployment Dynamics in a Pandemic

Weemploy a new Keynesian model with random search in the labor market

and endogenous selection among heterogeneous workers to investigate the

impact of a pandemic-induced recession on the distribution of unemploy-

ment across workers. In such a recession, workers whose unemployment

spells in normal times are inefciently frequent and long are disproportion-

ately affected. This remains true even when the pandemic initially causes

mass layoffs that affect workers broadly or if many separations represent

temporary layoffs. Monetary policy that responds to labor market variables

affects unemployment for all workers but does relativelylittle for the distri-

bution of unemployment across workers types.

JEL codes E24, E32, E52

Keywords: COVID-19,heterogeneity, monetary policy, selection,

unemployment

T COVID-19   2020 generated a severe eco-

nomic contraction in the global economy,and its impact on unemployment was unlike

anything seen in previous recessions. The unemployment rate in the United States

jumped from 3.5% in February 2020 to 14.8% in April. In contrast, the Great Reces-

sion following the global nancial crisis resulted in a peak U.S. unemployment rate

of 10% in October 2009, which, in turn, was the highest level seen over the previ-

ous quarter of a century.1We employ a heterogeneous worker new Keynesian model

Weare very grateful for comments from participants at the National Bank of Ukraine Annual Research

Conference on “Labor Market and Monetary Policy,” 28–29 May 2020, and from Bart Hobijn, Renato

Faccini, Yuriy Gorodnichenko, Simon Juul Hviid, Oleksiy Kryvtsov, Karl Walentin, Kenneth West, and

two anonymous referees.

F R is Head of Researchat Danmarks Nationalbank, and Afliated Professor at Uni-

versity of Copenhagen, HEC Montreal and CEPR (E-mail: fera@nationalbanken.dk). C E. W

is a Distinguished Professor of Economics Emeritus, University of California, Santa Cruz (E-mail:

walshc@ucsc.edu).

Received July 13, 2020; and accepted in revised form September 10, 2021.

1. In June 1983, the unemployment rate was 10.1%.

Journal of Money, Credit and Banking, Supplement to Vol. 54, No. S1 (February 2022)

© 2021 The Ohio State University

114 :MONEY,CREDIT AND BANKING

with search and matching frictions in the labor market to show how a pandemic re-

cession disproportionately affects those workers who, in normal times, experience

longer and more frequent spells of unemployment. Endogenous selection, in which

rm-worker matches that end are not simply chosen randomly from among exist-

ing matches and rms selectively screen job seekers before making hires, is the key

mechanism that drives this result. In addition, endogenous separations and hiring de-

cisions are inefcient in the competitive market equilibrium; workerswith higher av-

erage unemployment rates experience more unemployment than is socially efcient;

workers with lower average unemploymentrates experience less unemployment than

is socially efcient. Endogenous selection matters even in a COVID-19 pandemic

scenario in which there is a surge in mass layoffs that initially affects all workers

nonselectively. Workers with lower average unemployment rates benet if separa-

tions in a pandemic take the form of temporary layoffs, as layoffs initially did in the

COVID-19 recession, but this benet is not shared by workers with higher average

unemployment rates.

The paper makes four primary contributions. First, we identify a new externality

when selection arises from labor heterogeneity. Individual rms in the market equi-

librium ignore the effects their separation and hiring decisions have on the size and

the composition of the pool of unemployed workers. The rst effect on the size of the

unemployment pool is well-known and gives rise to the Hosios condition for search

efciency.The second effect on the average quality of the unemployment pool distorts

the distribution of unemployment across worker types. Second, we model the pan-

demic as a negativepreference shock - that reduces market consumption and increases

home production - and a spike in mass layoffs, and showhow the latter ends up having

a disproportional effect on the workers with higher average rates of unemployment.

Third, we show that if layoffs in a pandemic are predominately temporary,the reces-

sion still induces a rise in endogenous separations, and the employment benets of

temporary layoffs accrue primarily to the workers with lower averageunemployment

rates. Fourth, we nd that if the central bank responds to labor market variables, it

can limit the volatility of unemployment, but the ability of monetary policy to affect

the distribution of unemployment across worker types is limited.

The COVID-19 recession has been the result of a variety of underlying shocks

that do not easily map into the parsimonious number of shocks typically included in

a macro model, and the recent literature has employed different strategies for mod-

eling the causes of the recession. Combinations of supply and demand shocks are

employed by Baqaee and Farhi (2020), Fornaro and Wolf (2020), and Kocherlakota

(2020). The nature of the supply shock has been treated differently in the literature.

Kocherlakota models it as restrictions on labor supply, while Gregory, Menzio, and

Wiczer (2020) assume a temporary decline in worker productivity, and Bernstein,

Richter, and Throckmortan (2020) employ a largeand persistent job separation shock.

Models based on infection shocks include Kapiˇ

cka and Rupert (2020), Jackson and

Ortego-Marti (2020) who combine an infection shock with a skill loss shock that

hits the unemployed, and Eichenbaum, Rebelo, and Trabandt (2020) and Lepetit and

FEDERICO RAVENNAAND CARL E. WALSH :115

Fuentes-Albero (2021) who imbed an infection shock into a new Keynesian frame-

work.

While it is clear that there is not yet a consensus on how to fully replicate the shocks

that generated the COVID-19 recession, we choose to combine a preference shock

that shifts demand away from market goods and toward leisure and home production

with a shock to exogenous separations that affects all workers as a means of capturing

many aspects observed in the pandemic.2To produce the COVID-19 scenario, we

build a hypothetical scenario based on forecasts for observable variables, specically

total separations and output, that was produced at the beginning of the COVID-19

recession. That is, rather than simulate the model’s response to exogenous shocks, as

in a standard impulse response exercise, we let the model back out the preference and

separation shocks that drive the dynamics, conditional on the forecasts.

A preference shock is a standard means of generating an aggregate demand shock

in new Keynesian models, while a separation shock helps capture the sharp rise in

layoffs during the initial stages of the pandemic when stay-at-home measures caused

a collapse in demand while lockdowns forced businesses to temporarily close. The

result was a surge in unemployment across the entire economy.3In contrast to Bern-

stein, Richter, and Throckmortan (2020) in which all separations are exogenous, we

emphasize endogenous separations that amplify the effects of a shock to exogenous

separations. We stress how endogenous separations differentially affect those work-

ers who take longer to nd new jobs, thus helping to account for the persistence of

the rise in unemployment.4

We focus on heterogeneity across workers rather than other dimensions of hetero-

geneity such as differential effects across sectors or industries. This is motivated in

part by recent discussions by policymakers who have displayed interest in the la-

bor market and distributional consequences across workers of monetary policy. For

example, Federal Reserve Chair Powell has stressed the gains to those who may ben-

et from a strong labor market (Powell 2020).5Our framework allows us to explore

2. Our choice of a separation shock is discussed further in Section 3.

3. Aum, Kee, and Shin (2020) estimate that up to half of job losses in the United States and the United

Kingdom may have been due to lockdowns, and the evidencein Kahn, Lange, and Wiczer (2020) suggests

employment losses in April 2020 as measured by unemployment insurance claims were common across

U.S. industries and occupations, whether the industry was considered essential or work-from-home capa-

ble.

4. Cheng et al. (2020) nds that “the groups that had the highest unemployment rates in April also

tended to have thelowest reemployment rates, potentially making churn harmful to people and groups with

more and/or longer job losses.” And Gregory, Menzio, and Wiczer(2020) conclude that “...the lockdown

instituted to prevent the spread of the novel coronavirus is shownto have long-lasting negative effects on

unemployment. This is so because the lockdown disproportionately disrupts the employment of workers

who need years to nd stable jobs.”

5. Bergman, Matsa, and Weber (2021) use a model of unobserved heterogeneity across workers to

study the effects of a monetary policy shock. They show that tight labor marketsdisproportionally benet

low-skill workers relative to high-skill workers.