This Time It’s Different: The Role of Women’s Employmentin the Great Lockdown

Titan Alon, Matthias Doepke, Jane Olmstead-Rumsey, and Michèle Tertilt

Why Gender?

COVID-19 kills more men than women . . .

. . . but Great Lockdown recession has big economic impact on women.

→ Matters for welfare.

→ Matters for policy.

→ Matters for macroeconomic repercussions.

An Evolving Project

I Step 1: Paper in Covid Economics: Vetted and Real-Time Papers, April 2020:“The Impact of COVID-19 on Gender Equality”

→ Assess likely impacts and consequences based on pre-crisis data.

I Step 2: Data gathering on actual impacts during the crisis in various countries.

I Step 3: Macro model with rich household heterogeneity:“This Time It’s Different: The Role of Women’s Employment in the GreatLockdown”

→ Assess macro implications.

→ Assess long-run impacts.

→ Assess policy options.

Regular Recessions are “Mancessions”

−.0

50

.05

log_

hrs

cycl

ical

com

pone

nt fr

om h

p fil

ter

1960 1980 2000 2015Survey year

Males unmarried Females unmarriedMales married Females married

Regular Recessions are “Mancessions”

All Married SingleTotal Women Men Women Men Women Men

1989–2014Total Volatility 1.15 0.87 1.47 0.79 1.16 1.30 2.25Cyclical Volatility 0.91 0.51 1.23 0.38 0.95 0.70 1.82Hours Share 42.6 57.4 25.9 39.8 16.8 17.5Volatility Share 23.7 76.3 10.8 41.5 12.9 34.8

(Doepke and Tertilt 2016)

Regular Recessions are “Mancessions”

All Married SingleTotal Women Men Women Men Women Men

1989–2014Total Volatility 1.15 0.87 1.47 0.79 1.16 1.30 2.25Cyclical Volatility 0.91 0.51 1.23 0.38 0.95 0.70 1.82Hours Share 42.6 57.4 25.9 39.8 16.8 17.5Volatility Share 23.7 76.3 10.8 41.5 12.9 34.8

(Doepke and Tertilt 2016)

Regular Recessions are “Mancessions”

All Married SingleTotal Women Men Women Men Women Men

1989–2014Total Volatility 1.15 0.87 1.47 0.79 1.16 1.30 2.25Cyclical Volatility 0.91 0.51 1.23 0.38 0.95 0.70 1.82Hours Share 42.6 57.4 25.9 39.8 16.8 17.5Volatility Share 23.7 76.3 10.8 41.5 12.9 34.8

(Doepke and Tertilt 2016)

This Time It’s Different . . .

. . . for two reasons:

I Usual recessions are concentrated in sectors where many men work, such asconstruction and manufacturing. Not so this time.

I People’s ability to work also affected by increased childcare needs during schooland daycare closures. Affects women more than men.

Consequences of the Large Impact on Women’s Employment

I Welfare: Big impact on single parents and couples with little work flexibility.

I Amplification: Diminished scope for intra-household insurance magnifiestransmission of income shocks to consumption.

I Gender gap: Employment losses combined with returns to experience implywidening of the gender gap in the medium term.

I Social norms: Greater work flexibility and shifting social norms in response tomen’s greater involvement in childcare will benefit working mothers in the longterm.

Outline

1. Facts from pre-crisis data.

2. Evidence on actual impact so far.

3. Macro model with heterogeneity in gender, marital status, occupation, andchildcare needs.

4. Short-run, medium-run, and long-run implications (preliminary).

Facts from Pre-Crisis Data

Women versus Men in Critical and Telecommutable Occupations

Critical TelecommutableWomen 17% 22%Men 24% 28%

Critical: transportation, health, food production, maintenance and repair, public safetyand protective services

Classification of “telecommutable” based on ATUS 2017–2018.

Childcare Needs

More than 15 million single moms!

Living arrangements of US children in 2019:

Total children under 18 73,525 100%Two parents 51,561 70%Mother only 15,764 21%Father only 3,234 4%Other relatives 2,319 3%Non-relatives 647 1%

Distribution of Married Couples with Children by Employment Status

WivesNot Employed Part-Time Full-Time

Not Employed 4% 1% 5%Husbands Part-Time 1% 1% 2%

Full-Time 25% 15% 44%

Distribution of Married Couples with Children by Employment Status

WivesNot Employed Part-Time Full-Time

Not Employed 4% 1% 5%Husbands Part-Time 1% 1% 2%

Full-Time 25% 15% 44%

Distribution of Married Couples with Children by Employment Status

WivesNot Employed Part-Time Full-Time

Not Employed 4% 1% 5%Husbands Part-Time 1% 1% 2%

Full-Time 25% 15% 44%

Distribution of Married Couples with Children by Employment Status

WivesNot Employed Part-Time Full-Time

Not Employed 4% 1% 5%Husbands Part-Time 1% 1% 2%

Full-Time 25% 15% 44%

Weekly Childcare Hours for Couples who Both Work Full Time

All Children ≤ 5Women 10.3 16.8Men 7.2 10.6

Impact of Job Flexibility

Today, children and childcare explain much of gender wage gap (Kleven et al. 2019,Gallen 2018).

In ATUS-CPS data, having a telecommutable job is highly predictive of involvement inchildcare.

When married to women who cannot telecommute, men who can telecommute provide50 percent more childcare compared to men who cannot.

Potential for Changing Social Norms

Even small changes in childcare distribution can have long-run impact (e.g., paternityleave; Farré and González 2019).

Other evidence for importance of social norms and beliefs: Fernández, Fogli, andOlivetti (2004), Fernández (2013), Fogli and Veldkamp (2011), Olivetti, Patacchini,and Zenou (2018).

We estimate that in 9–12 percent of married couples with children the father will bethe primary childcare provider during the crisis.

The Great Lockdown is a huge change in childcare!

Potential for Changing Social Norms

”I just wanted to say I was excited to see your paper. My husband is a dentist who wasnever much involved with the kids or domestic activities, and has transformed to astay-at-home dad for the past 5 weeks while I continue to work from home as agovernment contractor. The effect on our family has been profound and life changing,and it would NEVER have happened without a global pandemic.”

The Actual Impact So Far

Employment Impact in the United States, Women versus Men

6 For Whom Did Employment and Hours Fall Most?

6.1 Labor Market Outcomes by Sex

Figure 6: Labor Market Outcomes by Sex, Age 18-64

(a) Employment Rate

77.7

68.261.6

67.8

53.3 50.4

010

2030

4050

6070

8090

100

Empl

oym

ent R

ate

(%)

Male Female

Mar 8 Mar 29 (RPS) Apr 12 (RPS)

(b) Hours Worked per Employed

40.1

35.2 36.3 36.2

31.9 33.2

05

1015

2025

3035

4045

Hou

rs W

orke

d pe

r Em

ploy

ed

Male Female

Mar 8 Mar 29 (RPS) Apr 12 (RPS)

Notes: Data from March 8 are derived from CPS surveys. More recent data are derived from responses to our onlineRPS survey.

Figure 6 displays the employment rate and hours worked per employed by sex from January-April, 2020. At the start of April, employment declined more severely for females, 21.3% (14.5percentage points), than for males, 12.2% (9.5 percentage points). Hours worked per employeddeclined similarly for females (4.3 hours per week) and males (4.9 hours per week). This initialdifference in employment outcomes by sex is consistent with the notion that females are dispro-portionately employed in industries hit hardest by the Coronavirus outbreak (such as services andespecially face to face services).

Since our first wave, declines in employment have shifted somewhat towards men. Between ourfirst and second waves, employment declined 6.6 percentage points for men, versus 2.9 percentagepoints for women. This could be one indication that the economic downturn is spreading beyondthe face to face service sector which saw the largest declines initially.

14

Alexander Bick and Adam Blandin “Real Time Labor Market Estimates During the2020 Coronavirus Outbreak”

State-Level Unemployment by Gender

Fuller Project for International Reporting:

I Women filed majority of unemployment claims in week after school and workplaceclosures.

I Women’s share ranges from 53 percent in Wyoming to 67 percent in Alabama.

International Evidence

Adams-Prassl, Boneva, Golin, and Rauh: “Inequality in the Impact of the CoronavirusShock: Evidence from Real Time Surveys”

Real-time survey evidence for UK, US, and Germany.

I Women more likely to lose employment in US and UK, but not in Germany.

I Employment losses much smaller in Germany.

I Labor market institutions matter!

Daily Childcare Responsibilities During the CrisisAmong parents who now work from home

Figure B.14: Hours spent on a “typical” work day during the past week on activechildcare and home schooling

0

1

2

3

4

Hou

rs

Childcare Home schooling

US

0

1

2

3

4

Hou

rs

Childcare Home schooling

UK

0

1

2

3

4

Hou

rs

Childcare Home schooling

Germany

Men Women

Notes: Data from wave 2 of the surveys. The thin black bars represent the 95% confidence intervals.The figure shows average number of hours that men and women reported spending on childcare andhomeschooling. We restrict the sample to individuals with children who report working from home,and whose answers to the time use questions combined do not exceed 24 hours.

43

Many Men Now Take Primary Responsibility

Mannheimer Coronastudie:Abbildung 8: Betreuungssituation nach Geschlecht

0

20

40

60

80

100

Proz

ent

Woche 1:20.-26.3.

Woche 2:27.3.-2.4.

Mehrfachnennung möglich; ohne Alleinerziehende

Nur die Frau Nur der Mann Beide Partner

Anmerkungen zur Abbildung: Fragetexte: Wie werden diese Kinder heute am [WOCHENTAG], dem [DATUM], tagsüber hauptsächlich betreut?Antwortmöglichkeiten: Durch mich selber, Durch die Notfallbetreuung der Kita/Schule, Durch eine andere Person im Haushalt (Alter 16 bis 59),Durch eine Person (Alter 16 bis 59), die nicht im Haushalt lebt, Durch eine andere Person im Haushalt (Alter 60+), Durch eine Person (Alter60+), die nicht im Haushalt lebt, Die Kinder können heute von keiner Person (Alter 16+) betreut werden (Mehrfachnennungen möglich).

Da nahezu alle Eltern die Betreuung ihrer Kinder nun im Haushalt selber leisten, betrachtenwir im Folgenden, wie die Betreuung zwischen den Partnern innerhalb des Haushaltes aufgeteiltist. Hierfür fokussieren wir auf alle Paarhaushalte mit Kindern und schließen jene wenigenHaushalte aus, die die Betreuung ihrer Kinder nicht selber übernehmen. Abbildung 8 zeigtdie Betreuungssituation im Haushalt im Wochenvergleich. Die Angaben beziehen sich immerauf die Betreuungssituation am Befragungstag. In etwa der Hälfte der Haushalte wird dieKinderbetreuung alleine von der Frau übernommen, dieser Wert ist im Wochenvergleich leicht,von 49,0 Prozent auf 52,0 Prozent, gestiegen. In etwa einem Viertel der Haushalte wird dieBetreuung am Befragungstag alleine vom Mann übernommen (Woche 1: 26,5 Prozent; Woche2: 24,0 Prozent). In 24 Prozent der Haushalte teilen sich beide Partner die Kinderbetreuung(Woche 1: 24,5 Prozent; Woche 2: 24,0 Prozent). Durch den Bezug auf den Befragungstag kannes zu einer Untererfassung von gemeinsamer Kinderbetreuung kommen, wenn die Partner sichtageweise abwechseln. Unsere Ergebnisse sind vor dem Hintergrund der Erwerbskonstellationenvon Paaren in Deutschland vor Einsetzen der Corona-Krise zu bewerten. Hier zeigen Zahlen desStatistischen Bundesamtes aus dem Jahr 2017, dass lediglich in knapp 15 Prozent der Paaremit Kindern unter 18 Jahren beide Partner Vollzeit arbeiten, es überwiegt die KonstellationMann Vollzeit / Frau Teilzeit (38 Prozent), in 28 Prozent der Paare mit Kind ist nurMann erwerbstätig. Somit zeigen die aktuellen Werte nicht unbedingt eine Verstärkung derGeschlechterungleichheit im Vergleich zu vor der Corona-Krise.

13

US: 12 percentage point March-April increase in millennial moms reporting they sharechildcare equally with their partner

Evidence from Dutch LISS DataGaudecker, Holler, Janys, Siflinger, and Zimpelmann: “The early impact of theCoViD-19 Pandemic on Gender Inequality”

3.2 Hours worked from home

• Home office overall does not look dramatically different in relative terms3

• Home office shares: Figures 9 and 10 suggest that home office shares movedcloser to each other— HMG:

probably kick out and onlyleave Table in

. Confirmed by regressions in Table 6

• Before the crisis, women without children below 12 had lower home officeshares, early in the crisis the gender pattern flips. Children did not predicthome office shares for men before the crisis, afterwards they do so verystrongly. They did before the crisis for women; afterwards the coefficientis just one fourth of the former despite the variance in the outcome goingup dramatically. So at least in terms of home office shares, the crisis workslike an equaliser across genders. Consistent with technology shaping thisand reluctance to allow home office for clerical/administrative work (e.g.,secretaries, ...— HMG:

need to add profession with atleast 3 categories (manual,lower-rank office,supervisory/academic)

).

before CoViD-19 late March before CoViD-19 late March

male female

0

5

10

15

20

25

30

35

40

wee

kly

wor

king

hou

rs

homeworkplace

Figure 3: Working hours at the workplace and at home, by gender

Numbers and standard errors in Table 16.

12

Women’s hours drop by 17 percent, men’s by 10 percent.

Men: Children predict home office during crisis, but not before.

Women: Children always predict home office.

Evidence from Dutch LISS Data

essential occupation non-essential occupationEmployed in ..

0.0

0.2

0.4

0.6

0.8

Shar

e th

at ta

kes

care

of c

hild

gendermalefemale

Figure 7: Gender differences in childcare

21

Lower gender gap in care conditional on essential occupation.

In 30 percent of couples where only mother is in critical occupation father is now solechildcare provider.

Family Macro Model

(Some) Literature We Build On

I Role of female labor supply in business cycles: Albanesi (2020), Albanesi andSahin (2018), Bardoczy (2020), Fukui, Nakamura, and Steinsson (2019), Doepkeand Tertilt (2016), Mankart and Oikonomou (2016, 2017).

I Family risk sharing: Blundell, Pistaferri, and Saporta-Eksten (2016), Ellieroth(2019), Guler, Guvenen, and Violante (2012), Ortigueira and Sassi (2013), Wuand Krueger (2018).

I The Great Lockdown Recession: Gregory, Menzio, and Wiczer (2020), . . .

Overview of Ingredients

I Women and men, singles and couples, childcare needs.

I Labor supply on the extensive margin, part-time work possible.

I Occupations differ by telecommutability.

I Job destruction shocks and unemployment.

I Endogenous accumulation of experience.

I Division of labor partly shaped by social norm.

No infection model!

Great Lockdown: shock to labor market and childcare needs.

Setting

Continuum of three types of households: single women, single men, couples.

State variables:

I Assets a.

I Human capital h.

I Kids k ∈ {0, s, b} (no kids, small kid, big kid).

I Employment e ∈ {E ,U} (employed, unemployed).

I Occupation o ∈ {TC ,NT} (can telecommute, cannot telecommute).

I Social norm m ∈ {0, 1} (modern, traditional).

Evolution of State Variables

Household type and occupation are permanent types.

Children arrive and leave according to probabilities πg (k ′|k) for singles and Π(k ′|k) forcouples.

Employment opportunities arise and vanish according to probabilities πg (e′|e, k) forsingles and Πg (e′|e, k). Offers can be rejected.

Human capital accumulates and depreciates stochastically as a function of labor supply.

Transition probabilities for social norm depend on fractions of couples with husband asequal or main primary childcare provider.

The Decision Problem for Unemployed Singles

vgU(a, h, k) = max

a′,c,l ,t

{ug (c, l) + βE

[wg

e′(a′, h′, k ′, o′)]}

subject to:

c + a′ = z + (1 + r)a,t = γ(k),

l + t = T .

Childcare needs: γ(s) > γ(b) > γ(0) = 0.

The Decision Problem for Employed Singles

vgE (a, h, k, o) = max

a′,c,l ,n,t

{ug (c, l) + βE

[wg

e′(a′, h′, k ′, o′)]}

subject to:

c + a′ = wgh n + (1 + r)a,t = γ(k),

l + n + max{t − φ(k) n I(o = TC), 0} = T .

The Decision Problem for Employed Singles

vgE (a, h, k, o) = max

a′,c,l ,n,t

{ug (c, l) + βE

[wg

e′(a′, h′, k ′, o′)]}

subject to:

c + a′ = wgh n + (1 + r)a,t = γ(k),

l + n + max{t − φ(k) n I(o = TC), 0} = T .

Decision Problem for Singles at Start of Period

Job offer:wg

E (a, h, k, o) = max{vg

E (a, h, k, o), vgU(a, h, k)

}.

No job offer:wg

U(a, h, k) = vgU(a, h, k).

Decision Problem for Dual-Earner Couples

VEE (a, hf , hm, k, of , om,m) = max{λuf (c f , l f ) + (1− λ)um(cm, lm)

−mψ(tm − t f ) + βE[W(ef )′,(em)′(a′, (hf )′, (hm)′, k, (of )′, (om)′,m)

]}subject to:

c f + cm + a′ = w f hf nf + wmhmnm + (1 + r)a,t f + tm = γ(k),

l f + nf + max{t f − φ(k) nf I(of = TC), 0} = T ,lm + nm + max{tm − φ(k) nmI(om = TC), 0} = T .

Couples’ Problem at Start of Period

Both have job offer:

WEE (a, hf , hm, k, of , om,m) = max{VEE (a, hf , hm, k, of , om,m),

VEU(a, hf , hm, k, of ,m),VUE (a, hf , hm, k, om,m),VUU(a, hf , hm, k,m)}.

. . . and so on.

Evolution of Human Capital and Social Norm

Human capital h is on a skill ladder H, probability of moving up and down depends onlabor supply, gender, and kids.

→ Returns to experience, skill loss when not employed

Transition probabilities for social norm taken as given by couple, depend on aggregatefraction of couples where man is primary childcare provider.

→ Social externalities

Calibration

Choose initial parameters to match:

I Labor market flows by marital status, gender, and kids.

I Distribution of couples among combinations of full-time work, part-time work, nowork.

I Estimates of returns to experience and skill loss in unemployment.

I Impact of occupation type on time allocation.

The Great Lockdown in the Model:Short-, Medium-, and Long-Run Effects

(in progress)

Recessions in the Model

Regular recession:

1. Large change in men’s job destruction and job finding rates.

2. Smaller change in women’s job destruction and job finding rates.

Great Lockdown recession:

1. Large change in men’s job destruction and job finding rates.

2. Equally large change in women’s job destruction and job finding rates.

3. Rise in childcare needs.

4. Shift in telecommuting ability and social norms.

Quantitative Questions to be Answered Using the Model

1. Welfare: Measure impact on different groups of households.

2. Amplification: Quantify higher MPC due to loss of intra-household insurance.

3. Policy: Effect of measures (e.g. earlier school openings) to mitigate the crisis.

4. Gender gap: Quantify impact on gender gap through employment lossescombined with returns to experience.

5. Division of labor: Quantify potential long-run impact on intra-householddivision of labor due to more telecommuting and shifting social norms.

Women’s vs. Men’s Labor Supply, Great Lockdown vs. Regular Recession

Labor Supply of Singles with and without Children

Gender Wage Gap, Great Lockdown vs. Regular Recession

Summary

Economically, impact on women and childcare needs is biggest distinction betweenGreat Lockdown and regular recession.

→ Matters for short-run policy response.

→ Matters for medium-run economic impact.

→ Matters for long-run changes to social norms and gender equality.