You are on page 1of 37

Has the Rise of Work-from-Home Reduced the

Motherhood Penalty in the Labor Market?

Emma Harrington1 · Matthew E. Kahn2

October 31, 2023

Abstract

When women become mothers, they often leave the labor force. Will the
recent rise of work-from-home (WFH) reduce the motherhood penalty in the
labor market, particularly in traditionally family-unfriendly careers? In the
decade before the pandemic, technological improvements increased WFH for
workers with some college degrees but not others. In degrees where WFH
increased, motherhood gaps in employment narrowed. On average, a ten per-
cent increase in WFH is associated with an 0.78 percentage point (or 0.94%)
increase in mothers’ employment relative to that of other women. This pattern
holds when defining the change in WFH using only the behavior of men, when
instrumenting for the change in WFH using the in-person nature of work, and
when conditioning on other changes within affected degrees in, e.g., hours
and income. These trends suggest that the rise of WFH transformed a broader
set of jobs into more family-friendly occupations. Traditionally flexible jobs in
sectors like education and pharmacy were relatively unaffected by the rise of
WFH, while motherhood employment gaps narrowed in traditionally family-
unfriendly fields like finance and marketing.

Keywords: Work-from-home, remote work, childhood penalty, motherhood,


gender gaps
JEL: J16, J13, J22, J24, L23, M54

1 University of Virginia (emma.k.harrington4@gmail.com)


2 University of Southern California and NBER (kahnme@usc.edu)
WFH & Motherhood Penalties Harrington & Kahn

When women become mothers, they often leave the labor force (Wood et al., 1993;
Bertrand et al., 2010; Angelov et al., 2016; Kuziemko et al., 2018; Kleven et al., 2019,
2021). Indeed, the career penalty of motherhood now accounts for most of the
gender gap in labor force participation (Kleven, 2022), especially in highly-paid
careers (Wood et al., 1993; Bertrand et al., 2010). Traditionally, family-friendly oc-
cupations required flexibility over when to work (Goldin and Katz, 2016; Cortés
and Pan, 2019; Bolotnyy and Emanuel, 2022). Yet such flexibility is often costly to
provide in jobs involving high degrees of coordination, collaboration, and continu-
ity (Goldin, 2014; Cubas et al., 2023; Noray, 2023). Could flexibility over where to
work reduce motherhood penalties in the labor market, especially in highly-paid
careers where demands on workers’ time are often unstable and inflexible?

Work from home (WFH) may help mothers juggle work and child-care responsi-
bilities. In time diaries, mothers who work from home report that they spend more
than a third of their working time passively taking care of children, compared to
negligible rates on-site.1 WFH may also reduce the logistical costs of working by
making it easier for mothers to, e.g., pick up a sick kid from school or drop off a
healthy kid at soccer practice.2 Consistent with these advantages, prior work has
found that women — and particularly mothers — have a higher willingness to pay
for WFH than men (Mas and Pallais, 2017; Maestas et al., 2018) and a greater dis-
taste for commuting (e.g., Black et al., 2014; Le Barbanchon et al., 2021; Bütikofer
et al., 2022).

The question of whether WFH can reduce the motherhood penalty in the labor
market has become increasingly relevant as new information communication tech-
nologies (ICT) have made it increasingly possible to WFH effectively. Modern
1 Figure A.1(a) shows that mothers who worked from home often coupled work-time with child-

care both before and after COVID-19 using data from the American Time Use Survey.
2 The logistical toll of motherhood may be magnified by school officials’ tendencies to call moth-

ers rather than fathers to, e.g., pick up sick kids — even when mothers report having more con-
straints on their time (Buzard et al., 2023).

1
WFH & Motherhood Penalties Harrington & Kahn

PCs have the capabilities once exclusive to mainframes, thanks to the exponen-
tial decline in the costs of computer memory and processing (Figure 1(a)). Internet
speeds have also precipitously risen, allowing workers to communicate instanta-
neously from afar, work on VPNs, and download and upload large amounts of
data seamlessly (Figure 1(b)).3 As PCs have gotten increasingly powerful and in-
ternet connections have gotten increasingly fast, WFH has likely become a better
substitute for office-based work in jobs that do not require physical presence.

With better technology, more workers started to WFH in the decade before the
pandemic. This increase was concentrated among college-educated workers with
“remotable degrees,” that did not require them to be physically present in places
like schools or hospitals.4 For workers with degrees in fields like marketing and
computer science, a growing minority of workers started to WFH most of the time
(Figure 1(c)). In addition, time-diary data indicate that a more sizable share of
workers started to WFH sporadically if, for example, their kid was sick or they
needed to work late into the evening. For workers who lacked discretion over
when to work, they may have had increasing discretion over where to work when
there was a deadline or other contingency. By contrast, in fields like education and
nursing, where physical presence is crucial, there was little change in WFH.

We ask how these changes in WFH affect the labor-market outcomes of mothers.
The raw time-series are consistent with the rise of WFH mitigating motherhood
penalties in the labor market (Figure 1(d)). We focus on the employment gap be-
tween mothers and other women with the same college degrees. In the decade
3 In addition, new video-conferencing platforms also made meeting online more seamless.
Skype was released in 2003, Zoom in 2013, and Google Meet and Microsoft Teams in 2017.
4 We use the Dingel and Neiman (2020) classification of occupations as amenable to WFH or

requiring in-person contact. Our preferred definition also includes educators other than professors
as in-person jobs since they were almost always required to be in-person prior to COVID-19. We
then take the average classification within each college degree to determine whether degree-holders
are typically required to work in-person. We classify degrees as typically requiring in-person work
if more than half of those with the degree were in an occupation that we classify as being in-person.

2
WFH & Motherhood Penalties Harrington & Kahn

prior to the pandemic, the motherhood employment gap substantially narrowed


among women with degrees in fields, like marketing and computer science, that
could increasingly be done from home. By contrast, there was little change in the
motherhood employment gap among women whose college degrees typically re-
quired in-person work in, for example, schools, hospitals, or pharmacies. These
trends represent expansions in what constituted family-friendly jobs: while tradi-
tionally family-friendly jobs in sectors like education or pharmacy were relatively
unaffected by the rise of WFH (Goldin and Katz, 2016), the motherhood employ-
ment gap narrowed for women with degrees is traditionally family-unfriendly oc-
cupations.

Our analysis then drills down to individual college degrees to investigate whether
the differential rise of WFH is reflected in differential changes in motherhood em-
ployment gaps at a more granular level. To do this, we first compute the change in
WFH in the decade prior to COVID-19 for each college degree. We then calculate
the change in mothers’ employment rate relative to that of other women in each
degree. We then consider how changes in WFH correlate with changes in mothers’
relative employment rates across college degrees.

We find a strong relationship between increases in WFH within a college degree


and a narrowing of the motherhood employment gap. On average, a ten percent
increase in the share of workers working primarily from home (0.6 percentage
point) is associated with a 0.78 percentage point narrowing of the motherhood
employment gap (p-value = 0.0097). The motherhood employment gap narrowed
more than one-for-one with the change in the share of workers primarily working
from home. This out-sized change likely reflects the multiple dimensions along
which WFH increased. Not only did more workers in affected fields WFH primar-
ily but also more of these workers worked from home sporadically on days when
kids were sick or looming deadlines led to late-night hours. Mothers’ employment

3
WFH & Motherhood Penalties Harrington & Kahn

rose in absolute as well as relative terms in affected majors, suggesting that these
changes do not reflect a zero-sum gain across women.

For this relationship to be causal, any changes in the other factors affecting the
motherhood gap in employment must be uncorrelated with changes in WFH across
college degrees. Reverse causality poses one threat to this assumption: mothers’
greater attachment to the labor-market could change the provision of WFH (rather
than WFH changing mothers’ attachment). To sidestep this concern, we estimate
the changes in WFH using only the subset of men. We find nearly identical results
using only the behavior of men to measure the WFH trend, suggesting reverse
causality does not drive the estimated relationship. We also find similar patterns
when we condition on other changes in the nature of work within specific college
degrees, including changes in hours, income, educational attainment, and gender
composition. Finally, instrumenting for the change in WFH using the need for
physical presence yields a similar estimated relationship.

In fields where WFH increased, mothers’ incomes rose relative to other women’s
even conditional on being employed. On average, a ten percent increase in WFH
(or 0.6pp increase) was associated with a 1.3 percent increase in employed moth-
ers’ incomes relative to other employed women’s (p-value = 0.043). This finding
is consistent with the rise of WFH not only expanding the labor-market attach-
ment of mothers but also increasing the opportunities of mothers who would have
persisted in the labor market regardless.

We do not find comparable patterns when analyzing the employment of fathers


relative to other men. These gendered findings are consistent with existing evi-
dence that mothers have stronger preferences for WFH than fathers (Mas and Pal-
lais, 2017; Maestas et al., 2023). These gendered preferences and effects likely re-
flect the gendered division of household labor, which makes WFH more valuable
for mother than fathers who seek to juggle career and family.

4
WFH & Motherhood Penalties Harrington & Kahn

Despite the potential advantages of WFH, relatively few mothers worked from
home before COVID-19. Why did the market provide so few WFH jobs for moth-
ers? Did the recent shock to WFH change this equilibrium? We hypothesize that
the transitory shock to WFH during the pandemic will push the market into a new
equilibrium. In our model, firms decide whether to invest in mothers’ skills after
observing their choice to work in the office or at home. In the initial equilibrium,
choosing WFH signaled low attachment to the workforce, and firms consequently
did not invest in the skills of mothers who chose WFH. In jobs that hinged on
firms’ investments, mothers faced a stark choice to either go to the office or drop
out of the workforce entirely. In jobs that were less reliant on firms’ investments,
mothers could opt for WFH in a lower track of the firm or leave the firm and enter
self-employment. After the rise of WFH, choosing WFH sent less of a signal about
mothers’ attachment to the workforce, firms continued to invest in mothers’ skills
even if they chose WFH, and this investment reinforced high-attachment mothers’
decisions to WFH. Thus, our model predicts that the rise of WFH would increase
mothers’ employment, particularly in high-investment jobs.

Our paper makes two contribution to the literature about WFH. First, we con-
tribute to the ongoing debate about the COVID-19 WFH shock on mothers’ labor-
market attachment. Goldin (2022) and Alon et al. (2023) have argued that, at least
in the long run, the rise of WFH should benefit college-educated mothers. How-
ever, these predictions have not been borne out in early data on the pandemic and
its aftermath. Heggeness et al. (2021) find that the pandemic took a toll on moth-
ers’ labor-market outcomes even for college-educated mothers, many of whom
could work from home. The pandemic’s childcare disruptions — coupled with
the continued unequal division of household labor — may have obscured the re-
turns to WFH for working mothers (Lyttelton et al., 2020; Pabilonia and Vernon,
2022).5 Our paper complements concurrent structural work that tries to estimate
5 Before COVID-19, Adams-Prassl et al. (2023) find a high cost of distractions from children for

5
WFH & Motherhood Penalties Harrington & Kahn

the life-cycle returns to work from home for mothers (Farooqi, 2023) in more nor-
mal times. Our work is also in line with recent experimental evidence in a starkly
different context: among data-entry workers in India, Jalota and Ho (2023) find
that the ability to work from home doubles mothers’ labor force participation.

Our paper also contributes to the research trying to solve puzzle of why WFH
was so rare prior to COVID-19 even in seemingly remotable jobs (Mas and Pallais,
Forthcoming). For fairly autonomous tasks, WFH has been found to increase pro-
ductivity or only slightly reduce productivity (Bloom et al., 2015; Choudhury et
al., 2020; Emanuel and Harrington, 2023). Further, workers were willing to accept
substantially lower wages to have the option to work from home, with particularly
high valuations among mothers (Mas and Pallais, 2017; He et al., 2021; Maestas et
al., 2023). We contribute to this debate by arguing that work from home may have
been less common prior to COVID-19 because those who valued locational flexi-
bility the most also had the most to lose from signaling lower attachment to the
workforce.

The rest of the paper is organized as follows. Section I describes our data sources,
and Section II describes our empirical design. Section III presents our results on
the relationship between increases in WFH and reductions in the motherhood gap
in labor-market outcomes. Section IV develops a simple signaling model of moth-
ers’ decisions about whether or not to work and if so whether to do so from home.
Section ?? discusses the lasting implications of the rise of WFH caused by the pan-
demic on mothers’ labor-market outcomes and investigates the extent to which
those predicted gains have been realized. Section V discusses the findings and
future directions.
mothers working from home, which contributes to a gender gap in earnings in an online labor
market.

6
WFH & Motherhood Penalties Harrington & Kahn

I Data

We use widely-used, publicly-available data from U.S. household surveys (specif-


ically, the American Community Survey) and time-diaries (specifically, the Amer-
ican Time Use Survey). We use both of these data sources to characterize the het-
erogeneous changes in work-from-home. We use the household surveys to charac-
terize changes in the motherhood gaps in labor market outcomes among women
with more and less remotable skill sets.

Measuring WFH in Houshold Survey Data. The American Community Survey


(ACS) asks individuals how they usually got to work last week with one option
of, worked from home. This field largely captures whether individuals primarily
worked from home.6

Measuring WFH in Time Diaries. Most people who WFH do not do so most of
the time. Thus, to capture a more complete picture of the changing landscape of
WFH, we integrate time-diary data from the American Time Use Survey (ATUS).
These time-diaries allow us to capture individuals who report working from home
on part or all of the recorded day but would not necessarily report working from
home most of the time. We focus on weekdays (rather than weekends), in which
individuals worked at least five hours. We then measure the share of total work
hours done at home and the share of evening hours done at home.

Measuring Remotable Skillsets. The American Community Survey asks individ-


uals about their college degree. This is an essential variable because it gives us a
measure of whether or not skills are suitable for WFH regardless of whether peo-
ple are employed. This allows us to characterize the changes in the motherhood
6 Given the specific wording of the questions, this measure may capture individuals who es-
sentially never go into the office. That interpretation would be consistent with others’ validation
efforts of the question (Barrero et al., 2021).

7
WFH & Motherhood Penalties Harrington & Kahn

employment gap for women with more and less remotable skill-sets.7

We use the classification of Dingel and Neiman (2020) to characterize whether col-
lege degrees are associated with occupations that require workers to be physically
present. Our preferred definition augments Dingel and Neiman (2020)’s classi-
fication to include educators, other than professors, as needing to be physically
present to reflect pre-pandemic norms, but our results are not sensitive to this re-
classification. We then take the average occupational classification within each col-
lege degree to determine whether degree-holders are typically required to work in-
person. We classify degrees as typically requiring workers to be physically present
work if more than half of those with the degree were in an occupation that we
classify as having such a requirement.

Labor-market outcomes: We focus on whether individuals were employed in the


week preceding the survey. We also consider total hours and wage/salary income.

Demographics: We focus on the differences in labor-market outcomes across re-


spondents with and without children among respondents who identify as women.
We focus on comparing women whose eldest child is under 15 and women with
no children and exclude women with older children. Our results are not sensitive
to this restriction.

II Empirical Design

Our empirical design leverages the fact that technological changes in the years pre-
ceding the pandemic led to more precipitous increases in work from home (WFH)
for college-educated workers with some degrees than others. For example, faster
internet speeds and better video-conferencing technologies made it easier to WFH
7 The
ATUS time-diaries do not record individuals’ college degrees. To identify comparable
populations across the datasets, we use occupational information to generate a crosswalk across
the datasets.

8
WFH & Motherhood Penalties Harrington & Kahn

for workers with marketing and computer science degrees. These technologies,
however, did not typically obviate the need to be in-person for educators, pharma-
cists, and health-care professionals. We then investigate the association between
increases in WFH in degrees and changes in the motherhood employment gap.

II.A Estimating changes in WFH

We estimate the average annual change in WFH in each college degree in the
decade prior to the pandemic. For each degree d, we limit the sample to indi-
viduals i who have that degree. We then estimate:

WFHit = δd + βWFH
d Yeart + ϵit , (1)

where δd represents a degree-specific intercept and WFHit is an indicator for indi-


vidual i reporting working primarily from home when they were surveyed by the
Census in year t. The coefficient of interest is βWFH
d , which summarises the change
in WFH within the degree prior to the pandemic. Figure 2(a) illustrates this estima-
tion for marketing and education degrees, and Figure 2(b) illustrates the estimated
β̂ d coefficients for the largest majors. The changes in WFH broadly accord with
intuitions about in-person versus potentially remotable work.

To more fully contextualize the changes in WFH, we use time-diaries, which re-
veal how many people are working from home on any given day and at any given
hour. Since time-diary data does not include individuals’ college degrees and in-
stead only records their current occupations, we impute whether individuals likely
have degrees more or less exposed to increases in WFH given their current occupa-
tion. For each occupation, we take a weighted average of the changes in WFH by
degree (from Equation 1). This then gives us a predicted change in WFH for each
occupation that we can then merge into the time-diary data to classify individuals’
likely exposure to changes in WFH based on their current occupation.

9
WFH & Motherhood Penalties Harrington & Kahn

II.B Estimating changes in motherhood penalties

We then estimate the changes in the motherhood employment gap in each degree
d. We focus on the differential changes in employment between women with and
without children among those with a particular degree. Let Motheri be an indica-
tor for identifying as a woman with children. We then estimate:

Motherhood-Gap
Employedit = θd + γd Motheri + ϕd Yeart + αd Motheri · Yeart + ζ it , (2)

where θd represents a degree-specific intercept of baseline employment for women


with degree d; γd reflects the initial difference in employment between mothers
and other women among those with the degree; ϕd represents the time-trend in
Motherhood-Gap
employment for women without children; and αd , our coefficient of
interest, represents the differential time-trend in employment for women versus
other children. Figure 2(a) illustrates this estimation for the illustrative examples
of marketing and education degrees. We are interested in whether and to what
extent increases in mothers’ employment outpaced those of other women among
those with a given degree.

In addition to estimating Equation 2 for employment, we also estimate versions of


this specification for total hours worked and log income, both unconditionally and
conditional on employment.

II.C Estimating the Link Between Changes in WFH & Motherhood Employ-
ment Gaps

We use the estimated trends from Equations 1-2 to estimate how increases in WFH
tend to relate to changes in the motherhood employment gap across college de-
grees:
Motherhood-Gap
α̂d = ψ + ρ β̂WFH
d + ud . (3)

10
WFH & Motherhood Penalties Harrington & Kahn

We weight these regressions by the number of degree-holders. We focus on degrees


with at least five thousand women to reduce measurement error, but our results are
not sensitive to this restriction. By estimating this at the degree level, we respect
the fact that our design leverages degree-level variation in WFH changes.8

We would like to interpret ρ as reflecting the causal effect of increasing WFH on the
motherhood gap in employment. For this causal interpretation to hold, other fac-
tors that shift mothers’ relative employment rates in degrees must be orthogonal
to changes in WFH. Notably, this identifying assumption is about changes rather
than levels: as a result, it does not require that the level of WFH be orthogonal to
the level of other degree-specific factors affecting the motherhood penalty. Because
we focus on mothers’ employment rates relative to those of other women, factors
that equally affect women with and without children do not impact the identify-
ing assumption. Nonetheless, this identifying assumption is a strong one, so it is
valuable to consider potential threats to the design.

One threat to this design is reversal causality: mothers’ greater attachment to the
workforce could cause increases in WFH rather than the other way around. Par-
ticularly, as mothers with specific degrees become more attached to the workforce,
they may advocate for WFH to ease their attempts to juggle career and family. Our
estimated relationship could then reflect reverse causality as mothers’ attachment
increases WFH rates. To sidestep this concern, we estimate versions of Equation
3 where we estimate the trend in WFH solely using the sample of men. This de-
sign yields similar estimates as our baseline specification, suggesting that reverse
causality does not drive our results.

Another threat to the design is that changes in WFH are correlated with other
changes in work that ease mothers’ attempts to juggle career and family or in-
8 Currentlywe report analytical standard errors. Future drafts will bootstrap standard errors to
respect the multiple sources of measurement errors.

11
WFH & Motherhood Penalties Harrington & Kahn

crease mothers’ incentives to persist in the workforce. To account for some of these
factors, we control for changes in average hours, educational-attainment, income,
and gender composition within degrees, which we compute using the analogues of
Equation 1. We find that our results are substantively unchanged by the inclusion
of these controls.

II.D Instrumenting for Changes in WFH

To bolster our design, we also instrument for changes in WFH based on whether
the work associated with each college degree tends to require physical presence.

First Stage: β̂WFH


d = κ + τ · Requires Physical Presenced + vd (4)

= ψIV + ρIV β̂ˆ WFH


Motherhood-Gap
Second Stage: α̂d d + νd , (5)

where the second stages uses the first-stage predictions. This design solely relies on
the fact that technological changes in ICT had little impact on WFH for those with
college degrees that typically required physical presence at work, while allowing
for increases in WFH for those with other degrees. The identifying assumption is
then that other factors influencing the motherhood penalty did not differentially
change across degrees that required physical presence and those that did not.

III Results

Section III.A analyze the heterogeneous changes in WFH in the decade prior to
COVID-19 across college-educated workers with different degrees. Section ?? then
turns to how increases in flexibility over where to work related to changes in moth-
ers’ employment rates vis-á-vis those of other women with the same degrees.

12
WFH & Motherhood Penalties Harrington & Kahn

III.A Changes in WFH

Work from home increased in the decade prior to COVID-19, particularly among
workers with degrees that did not typically require being physically present with,
for example, patients or students (Figure 1(b)). College degrees like education and
nursing was little increase in WFH, while marketing and finance saw relatively
large increases (Figure 2(b)).

While a growing minority of workers worked primarily from home, a larger share
of workers used work from home in more limited ways. Figure 3 splits individuals
into terciles based on the extent to which they likely saw increases in the suitability
of their skills to WFH. Figure 3(a) focuses on workers primarily working from
home (as in (Figure 1(b) and Figure 2(a)-(b)). Figure 3(a) shows there were larger
changes and higher levels of people working from home in any given hour of a
workday in the time-diary data. By 2019, over a fifth of work hours were spent at
home among the tercile of workers whose skills had become most remotable.

In addition, late-night hours in fields like finance were increasingly done from
home rather than from the office, among both workers and managers. Figures
3(c)-(d) show this. By 2019, over half of workers’ work hours after 6pm were done
at home among the tercile of workers whose skills had become most remotable.
Fully 80 percent of managers’ time after 6pm was done at home in this group. This
suggests that the returns to working late in the office for promotions and pay raises
may have declined as WFH became increasingly technologically feasible. To the
extent to which mothers found late nights at the office particularly challenging in
balancing career and family, these changes may have been particularly impactful
for working mothers.

13
WFH & Motherhood Penalties Harrington & Kahn

III.B Changes in Motherhood Employment Gaps

In college degrees in which WFH increased, the motherhood gap in employment


narrowed. Figure 4 illustrates this relationship. Among college degrees like mar-
keting that saw large increases in WFH (on the x-axis), there also tended to be large
increases in mothers’ employment rates relative to those of other women (on the
y-axis). By contrast, in fields like education that saw less of an increase in WFH,
there also tended to be less movement in the motherhood employment gap.

On average, a one percentage point increase in the share of workers working pri-
marily from home is associated with a 1.2 percentage point narrowing of the moth-
erhood employment gap (p-value = 0.0097, Column 1 of Table 1). On its face,
this one-for-one relationship seems too large and suggests that most of mothers
who newly work from home would have otherwise dropped out of the workforce.
However, as discussed above, increases in the prevalence of workers working pri-
marily from home are accompanied by larger increases in more limited uses of
WFH on a day here or there or in the evenings. Considering the percentage in-
crease in the prevalence of workers working primarily from home may better re-
flect the magnitude of the shift in locational flexibility. Taking this view suggests
that a ten percent increase in WFH would close the motherhood employment gap
by 0.78 percentage points (or 0.94 percent).9

Robustness: We find similar relationships after controlling for other changes in


the nature of work in affected degrees, specifically the average hours, educational
attainment, income, and gender composition (Columns 2-4 in 1). We also find sim-
ilar patterns when characterizing changes in WFH within each degree using only
9 Tosee this, note that a 0.6 percentage point (p.p.) increase in WFH represents a 10 percent
increase off the base of 6 percent. Multiplying this by the 1.2 p.p. coefficient yields the estimated
change of 0.78 p.p. This approach is most reasonable if other dimensions of WFH — such as work-
ing from home in the evenings — are increasing proportionately to the share of workers primarily
working from home.

14
WFH & Motherhood Penalties Harrington & Kahn

the sample of men (Table A.1), suggesting that the observed pattern is not driven
by reverse causality (e.g., because increasingly attached mothers increasingly ad-
vocate for WFH in certain degrees).

We also find qualitatively similar patterns when instrumenting for the change in
WFH using information on whether or not the college degree typically involves
work that typically requires physical presence at, for example, a school or hospital
(Table 2).

Additional Labor-Market Outcomes: We see similar patterns for other labor mar-
ket outcomes in Figure 5. In college degrees where WFH increased, mothers started
to work more total hours relative to other women and earn more income, even con-
ditional on employment.

Mothers vs. Fathers: Figure 6 considers the change in mothers’ relative employ-
ment rates (versus other women’s) versus fathers’ relative employment rates (ver-
sus other men’s). This figure is almost identical to the baseline analysis in Figure 4,
suggesting that the increasing prevalence of WFH in certain degrees was uniquely
valuable for mothers. This pattern is consistent with the continued gendered divi-
sion of household labor, which makes WFH more valuable for mothers attempting
to balance career and family.

IV Model

We analyze the equilibrium impacts of the rise of remote work on mothers’ labor-
market outcomes. To do so, we build a stylized two-period model, in which moth-
ers choose whether and where to work, and firms decide whether to invest in
workers’ skills. We show how COVID-19’s massive WFH experiment could have
shifted the market’s equilibrium provision of WFH for mothers.

15
WFH & Motherhood Penalties Harrington & Kahn

IV.A Setup

Timing of the Model. In the first period, each mother chooses whether to work
from home or in the office, and then her firm chooses whether to invest in her
skills. In the second period, the mother chooses whether or not to persist in the
workforce after a shock to her opportunity cost of work has been realized. When
deciding whether to invest in the mother’s skills, the firm can observe her work-
location choice but not her opportunity cost of work.

WFH & mothers’ opportunity cost of work. We assume that mothers’ utility is
quasi-linear in her wage and her opportunity cost of work for home production.
In the first period, each mother’s opportunity cost of work is high, c H , or low, c L .10
WFH reduces work’s opportunity cost from c to γc where γ < 1 because, at home,
mothers can, e.g., passively care for napping children or cart kids around without
first commuting home from the office.11 To benefit from WFH, mothers must bear
a fixed cost F, from, e.g., setting up a home office.

In the second period, there is a probability π that a mother’s opportunity costs of


work discretely increases. For example, she might have another child or her par-
ents may be no longer able to help with her kid(s). For simplicity, we assume that a
mother’s response to an adverse shock to her opportunity cost is fully determined
by her initial cost: a mother with a high initial cost drops out, and a mother with a
low initial cost persists.

Mothers have private information about their initial opportunity costs of working.
Firms know that a share, θ, of mothers have a high opportunity cost of working,
but do not know which mothers have high and low costs. Since WFH is more
10 Thisopportunity cost might depend on whether, e.g., grandparents can take care of children.
11 These commuting costs may be especially binding for women who only want to work part-
time, for whom the fixed costs of a daily commute are particularly large as noted by Oi (1962);
Cogan (1981).

16
WFH & Motherhood Penalties Harrington & Kahn

valuable to mothers with high opportunity costs, mothers’ WFH choices can reveal
information about their types and likely attachment to the workforce.

Firms’ production functions & investment choices. The firm can invest in the
worker’s skills. In job j, investing in the worker reduces first-period output from
y j to y j − i j but increases second-period output from y j to to y j + t j . We assume that
t j > i j , so the firm would invest in the worker if it knew that the worker would
persist in the workforce. We model firms as having linear production functions,
yielding perfect substitutability across efficiency units of labor.12

For simplicity, we assume that WFH does not directly impact mothers’ productiv-
ity: if, instead, a mother’s opportunity cost of work not only determined her at-
tachment to the workforce but also her likely distractions at home, then this would
amplify the mechanisms in our model.13

In the model, mothers do not invest in their own human capital. If such invest-
ments were substitutes with firm-side investment, then this would temper the
model’s mechanism. If, instead, such investments were complementary with firm
investments, then this would amplify the model’s mechanism because firms would
not only learn about mothers’ likely persistence from their WFH choices but also
about their likely investments in their own skills.

We assume that mothers cannot fully defray the costs of training in the form of
lower initial wages, either because of regulatory limitations (e.g., the minimum
wage) or credibility problems (e.g., mothers cannot trust firms to increase pay if
12 This assumption helps us abstract from men in our model. The firm would view men as similar
to the subset of women who always persist in the workforce.
13 Suppose mothers with high opportunity costs of working were most distracted at home (or

perceived to be most distracted). In this case, firms would be more reluctant to let mothers WFH
if only high-cost mothers chose WFH, since these mothers would be exactly the mothers whose
productivity would be most negatively impacted by WFH. Indeed, Adams-Prassl et al. (2023) found
distractions at home reduced mothers’ productivity on MTurk, particularly for mothers with small
children who might have the highest opportunity cost of going to the office.

17
WFH & Motherhood Penalties Harrington & Kahn

they persist).14 For illustrative purposes, we assume that wages are constant across
periods and bid up to the workers’ expected, average product.

IV.B Equilibrium

We consider the necessary conditions for optimizing firms and rational mothers to
arrive at a separating equilibrium in which low opportunity cost (L) mothers work
on-site and high opportunity cost (H) mothers either work from home or drop out
of the workforce. We then consider the conditions under which a WFH pooling
equilibrium could arise after an exogenous experiment with WFH.

Low-WFH, Separating Equilibrium. Let’s consider whether it would be an equi-


librium, pre-COVID, for firms to offer a WFH wage of y j with no investments in
workers and an on-site wage of y j + 21 (t j − i j ) with investments in workers.

For this to be an equilibrium, three conditions must hold. First, low-opportunity-


cost (L) mothers must choose on-site jobs. This will be satisfied if the benefits of
WFH in reducing work’s opportunity costs [(1 − γ)c L ] fall short of the fixed costs
of adopting WFH and the loss of firm investment [ 12 (t j − i j + F )]:

1
(1 − γ ) c L < ( t j − i j + F ).
2

Second, high-cost (H) mothers must either choose to WFH or dropout:

1 1
WFH: (1 − γ ) c H > ( t j − i j + F ) and y j ≥ γc H + F or (6)
2 2
1 1
Dropout: y j < γc H + F and y j + (t j − i j ) < c H . (7)
2 2

Third, a firm must not be able to profit by deviating and offering WFH with in-
14 For example, Haegele (2022) finds that firms do not promote talented women, muting rewards
for training. If we had curvature in the workers’ utility functions, it could also be that credit con-
straints make deferred consumption costly.

18
WFH & Motherhood Penalties Harrington & Kahn

vestment. Such a job would, at best, attract both types of workers so could at most
offer a wage of w j = y j − 12 i j + 21 t j [1 − θπ ]. This job would not attract L mothers if:

1 1
F + t j θπ > (1 − γ)c L . (8)
2 2

In this equation, 21 t j θπ represents the costs to L mothers of pooling with H moth-


ers. This pooling cost is higher when the returns to training (t j ) are high, when H
mothers are more likely to drop out (high π), and when H mothers make up more
of the workforce (high θ). Conversely, L mothers are more willing to pool with H
mothers in jobs with limited investment in workers’ skills (low t j ), making it easier
to sustain a pooling WFH equilibrium.

Let’s consider H mothers’ decisions to drop out in Equation 7. Drop out is more
likely in jobs where baseline productivity (y j ) is low without investment. Thus, in
jobs that hinge on firm investments in workers, it’s more likely that a separating
equilibrium arises in which L mothers stay on-site and H mothers drop out.

High-WFH, Pooling Equilibrium. We suppose an exogenous shock (like COVID-


19) caused all working women with remotable skills to bear the fixed cost F of
adapting to WFH. If L mothers choose WFH jobs after bearing this fixed cost, the
market can sustain a pooling equilibrium after this shock if:

1
t θπ < (1 − γ)c L . (9)
2 j

For the subset of jobs where both Equations 8 and 9 held, the market would switch
from an equilibrium in which WFH meant forgoing firm-side investments to one
in which mothers did not face such a steep penalty for WFH.15
15 Itis also possible that an exogenous experiment with WFH could make it easier for low op-
portunity cost (L) mothers to differentiate themselves. If the initial equilibrium featured pooling
on-site because of high fixed costs of WFH adoption (F), then it would be hard for L mothers to
distinguish themselves and the firm might not invest in any mothers even if they stayed on-site.

19
WFH & Motherhood Penalties Harrington & Kahn

Welfare Implications. So long as the fixed costs of adopting WFH were not too
high, an exogenous experiment with WFH will make mothers better off if it creates
a pooling equilibrium. The biggest beneficiaries are mothers who would otherwise
exit the workforce entirely or work from home but forego any investment from
firms in their skills. For mothers who initially worked on-site but can now advance
their careers at home, they defray the opportunity costs of working by (1 − γ)c L .

IV.C Testable Predictions

• In the initial equilibrium before COVID-19’s shock to work-from-home. . .

1. Mothers were more likely to work from home in jobs with low returns
to firm-side investments in their human capital.

2. Mothers were more likely to exit the labor force in jobs with high returns
to firm-side investment in their human capital.

3. Mothers often coupled WFH with part-time work because there was lit-
tle incentive to signal high-attachment when firms were already not in-
vesting in their skills.16

4. Mothers often coupled WFH with self-employment because there was


little incentive to work at firms that did not invest in them.

• In the equilibrium after COVID-19’s shock to WFH. . .

1. Mothers were more likely to choose WFH in a wider variety of jobs, even
those featuring high returns to firm-side investment.
Once all working women overcame the fixed costs of remote work, then H mothers would choose
WFH and L mothers would choose to be on-site. Firms could then invest in L mothers when they
previously had not invested in any mothers.
16 While our model does not explicitly feature an intensive-margin choice over how many hours

to work, such an extension would yield this result for a couple of reasons. In the initial equilibrium,
mothers forgo firm-side investments in their human capital if they work from home. Thus, they
have less to lose from working part-time in terms of signaling attachment to the workforce. They
also have lower returns to long hours without investments in their skills.

20
WFH & Motherhood Penalties Harrington & Kahn

2. Mothers were more likely to persist in the labor force, particularly in


jobs with high returns to firm-side investment.

3. Mothers were more likely to pair remote work with full-time work and
less likely to pair it with self-employment.

V Conclusion

We study how changes in work-from-home in the decade prior to the pandemic


impacted motherhood penalties in the labor market. We ask whether flexibility
over where to work can ease mothers’ attempts to juggle career and family even in
jobs where flexibility over when to work is costly to provide.

Among college-educated workers with certain degrees — like marketing or finance


— improvements in ICT made working from home an increasingly good substi-
tute to working from the office. A growing minority of workers with these degrees
worked from home primarily, and a much larger share used worked from home
more sparingly. They might work from home if, e.g., their kids were sick or an
approaching deadline required them to work late. These jobs were increasingly of-
fering workers some flexibility over where to work if not when to work. By contrast,
workers with degrees in fields like education or pharmacy that required physical
presence were largely unaffected by these technological changes.

We ask how changes in flexibility over where to work translate into changes in
the motherhood penalty. We find that in fields where remote work became more
prevalent, the gap in employment between mothers and other women narrowed.
This shift in WFH expanded the set of family-friendly occupations, transforming
a set of high-paid, high-growth careers into occupations that are more reconcilable
with mothers’ home-lives. This expansion of family-friendly occupations promises
to reduce gender inequality in the labor market.

21
WFH & Motherhood Penalties Harrington & Kahn

We then investigate why more women were not working from home before the
pandemic. To do this, we build a signaling model of mothers’ choices to work from
home and firms’ decisions to invest in workers’ skills. In the initial equilibrium, too
few mothers may have worked from home for fear of signaling a low attachment
to the workforce and effectively ending firms’ investments in their skills. Faced
with the stark options of continuing a career on-site or effectively ending their
advancement at home, many women may have opted out of the workforce entirely.
Once WFH becomes a more viable option for their careers, some of these mothers
persist in the workforce in high-wage, high-growth occupations.

We focus on the interlinked decisions of mothers and firms. The rise of WFH may
have even more far-reaching implications for gender gaps in labor markets. The
expectation that women can work from home after becoming mothers may in-
crease young women’s incentives to invest in their educations, choose degrees
that lead to high-growth careers, and invest in firm-specific human capital once
at a given firm. By delinking women’s career choices from their expected fertil-
ity outcomes, work-from-home may allow women to optimize more like men.17
On the firm-side, the eventual option of WFH increases the incentive to invest in
the training of women because of the worker-firm match will persist even if the
woman becomes a mother. These longer-term implications of the rise of WFH are
an exciting avenue of future work.

17 Indeed,the concern that women may optimize quite differently from men when choosing col-
lege majors may be part of the reason that women have often been excluded from analyses of
college-major choices (Arcidiacono et al., 2020).

22
WFH & Motherhood Penalties Harrington & Kahn

References

Adams-Prassl, Abi, Kotaro Hara, Kristy Milland, and Chris Callison-Burch,


“The Gender Wage Gap on an Online Labour Market: The Cost of Interrup-
tions,” The Review of Economics & Statistics, 2023, 23 (1).

Akamai, “Average internet connection speed in the United States from 2007 to
2017 (in Mbps), by quarter,” Statista, 2018.

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


“This time it’s different: the role of women’s employment in a pandemic re-
cession,” Technical Report, National Bureau of Economic Research 2023.

Angelov, Nikolay, Per Johansson, and Erica Lindahl, “Parenthood and the gender
gap in pay,” Journal of labor economics, 2016, 34 (3), 545–579.

Arcidiacono, Peter, V Joseph Hotz, Arnaud Maurel, and Teresa Romano, “Ex
ante returns and occupational choice,” Journal of Political Economy, 2020, 128 (12),
4475–4522.

Barbanchon, Thomas Le, Roland Rathelot, and Alexandra Roulet, “Gender dif-
ferences in job search: Trading off commute against wage,” The Quarterly Journal
of Economics, 2021, 136 (1), 381–426.

Barrero, Jose Maria, Nicholas Bloom, and Steven J Davis, “Why working from
home will stick,” Technical Report, National Bureau of Economic Research 2021.

Bertrand, Marianne, Claudia Goldin, and Lawrence F Katz, “Dynamics of the


gender gap for young professionals in the financial and corporate sectors,”
American economic journal: applied economics, 2010, 2 (3), 228–55.

Black, Dan A, Natalia Kolesnikova, and Lowell J Taylor, “Why do so few women
work in New York (and so many in Minneapolis)? Labor supply of married
women across US cities,” Journal of Urban Economics, 2014, 79, 59–71.

Bloom, Nicholas, James Liang, John Roberts, and Zhichun Jenny Ying, “Does
working from home work? Evidence from a Chinese experiment,” The Quarterly
Journal of Economics, 2015, 130 (1), 165–218.

Bolotnyy, Valentin and Natalia Emanuel, “Why do women earn less than men?
Evidence from bus and train operators,” Journal of Labor Economics, 2022, 40 (2),
283–323.

Bütikofer, Aline, Katrine V Løken, and Alexander Willén, “Building Bridges and
Widening Gaps,” The Review of Economics and Statistics, 2022, pp. 1–44.

23
WFH & Motherhood Penalties Harrington & Kahn

Buzard, Kristy, Laura Gee, and Olga Stoddard, “Who You Gonna Call? Gender
Inequality in External Demands for Parental Involvement,” SSRN Working Paper,
2023.

Choudhury, Prithwiraj, Cirrus Foroughi, and Barbara Zepp Larson, “Work-from-


anywhere: The productivity effects of geographic flexibility,” in “Academy of
Management Proceedings” number 1. In ‘2020.’ Academy of Management Bri-
arcliff Manor, NY 10510 2020, p. 21199.

Cogan, John, “Fixed costs and labor supply,” Econometrica, 1981, 49 (4), 945–963.

Cortés, Patricia and Jessica Pan, “When time binds: substitutes for household
production, returns to working long hours, and the skilled gender wage gap,”
Journal of Labor Economics, 2019, 37 (2), 351–398.

Cubas, German, Chinhui Juhn, and Pedro Silos, “Coordinated work schedules
and the gender wage gap,” The Economic Journal, 2023, 133 (651), 1036–1066.

Dingel, Jonathan I and Brent Neiman, “How many jobs can be done at home?,”
Journal of Public Economics, 2020, 189, 104235.

Emanuel, Natalia and Emma Harrington, “Working remotely? Selection, treat-


ment, and the market for remote work,” FRB of New York Staff Report, 2023,
(1061).

Farooqi, Hira, “Telework availability, women’s labor market outcomes and fertil-
ity choices,” Working Paper, 2023.

Goldin, Claudia, “A grand gender convergence: Its last chapter,” American Eco-
nomic Review, 2014, 104 (4), 1091–1119.

, “Understanding the Economic Impact of COVID-19 on Women,” Brookings Pa-


pers on Economic Activity, 2022, 2022 (1), 65–140.

and Lawrence F Katz, “A most egalitarian profession: pharmacy and the evo-
lution of a family-friendly occupation,” Journal of Labor Economics, 2016, 34 (3),
705–746.

Haegele, Ingrid, “Talent hoarding in organizations,” arXiv preprint


arXiv:2206.15098, 2022.

He, Haoran, David Neumark, and Qian Weng, “Do workers value flexible jobs?
A field experiment,” Journal of Labor Economics, 2021, 39 (3), 709–738.

Heggeness, Misty, Palak Suri et al., “Telework, childcare, and mothers’ labor sup-
ply,” Opportunity and Inclusive Growth Institute Working Paper, 2021, 52.

24
WFH & Motherhood Penalties Harrington & Kahn

Jalota, Suhani and Lisa Ho, “What Works for Her? How Work-from-Home Digital
Jobs Affect Female Labor Force Participation,” Working Paper, 2023.

Kleven, Henrik, “The geography of child penalties and gender norms: Evidence
from the United States,” Technical Report, National Bureau of Economic Re-
search 2022.

, Camille Landais, and Jakob Egholt Søgaard, “Children and gender inequality:
Evidence from Denmark,” American Economic Journal: Applied Economics, 2019, 11
(4), 181–209.

, , and , “Does biology drive child penalties? evidence from biological and
adoptive families,” American Economic Review: Insights, 2021, 3 (2), 183–98.

Kuziemko, Ilyana, Jessica Pan, Jenny Shen, and Ebonya Washington, “The
mommy effect: Do women anticipate the employment effects of motherhood?,”
Technical Report, National Bureau of Economic Research 2018.

Lyttelton, Thomas, Emma Zang, and Kelly Musick, “Gender differences in


telecommuting and implications for inequality at home and work,” Available at
SSRN 3645561, 2020.

Maestas, Nicole, Kathleen J Mullen, David Powell, Till Von Wachter, and Jef-
frey B Wenger, “The value of working conditions in the united states and im-
plications for the structure of wages,” Technical Report, National Bureau of Eco-
nomic Research 2018.

, , , , and , “The value of working conditions in the United States and


implications for the structure of wages,” American Economic Review, 2023, 113 (7),
2007–2047.

Mas, Alexandre and Amanda Pallais, “Valuing alternative work arrangements,”


American Economic Review, 2017, 107 (12), 3722–59.

and , “Alternative work arrangements,” Technical Report Forthcoming.

McCallum, John, “Historical cost of computer memory and storage,” Our World
in Data, 2022.

Noray, Savannah, “Inflexibility, Social Skills, and the Allocation of Talent,” Work-
ing Paper, 2023.

Oi, Walter Y, “Labor as a quasi-fixed factor,” Journal of political economy, 1962, 70


(6), 538–555.

Pabilonia, Sabrina Wulff and Victoria Vernon, “Telework and time use,” 2022.

25
WFH & Motherhood Penalties Harrington & Kahn

Richards, Mark A and Gary A Shaw, “Chips, architectures and algorithms: Re-
flections on the exponential growth of digital signal processing capability,”
Working Paper, 2016.

Wood, Robert G, Mary E Corcoran, and Paul N Courant, “Pay differences among
the highly paid: The male-female earnings gap in lawyers’ salaries,” Journal of
Labor Economics, 1993, 11 (3), 417–441.

26
WFH & Motherhood Penalties Harrington & Kahn

Figure 1: Changes in ICT, Work from Home (WFH), & Motherhood


Gaps in Employment
(a) Cheaper Computing (b) Faster Internet

(c) Work from Home (d) Motherhood Employment Gap

Notes: Panels (a)-(b) illustrate recent changes in information communication technologies (ICT)
that may affect the feasibility of work from home. Panel (a) presents information on the costs of
one terabyte of computer memory (McCallum, 2022) and the cost of executing one trillion instruc-
tions from Richards and Shaw (2016) (assuming a price of a kilowatt of power of 7 cents). Panel
(b) presents internet speeds from Akamai (2018). Panel (c) shows changes in work from home
for college-educated workers with degrees that tend to require in-person work (e.g., education or
nursing) and those that do not (e.g., marketing or computer science). Panel (d) shows changes in
mothers’ employment rates versus those of other women for individuals with college degrees that
differentially require in-person work. Both series are based on the American Community Survey.
For more information on the data and definitions, see Section I.

27
WFH & Motherhood Penalties Harrington & Kahn

Figure 2: Changes in WFH & Motherhood Employment Gaps in Spe-


cific College Degrees
(a) Example WFH changes (b) Changes across large degrees

(c): Example Changes in Motherhood Employment Gaps

Notes: This figure illustrates changes in work from home and motherhood employment gaps in
specific college degrees. Data comes from the American Community Survey. The sample is lim-
ited to college-educated workers between the ages of 20 and 55. Panel (a) illustrates the changes in
WFH for the specific examples of marketing (which is amenable to work from home) and education
(which has a substantial in-person component). The fit lines and annotated coefficients reflect Equa-
tion 1. Panel (b) illustrates the estimated time-trends ( β̂WFH
d ) for each of the largest college degrees.
Error bars represent 95% confidence intervals. The colors represent degrees divided into terciles
based on the change in WFH. Panel (c) juxtaposes the trends in mothers’ employment rates and
Motherhood-Gap
those of other women for education and marketing. The annotated coefficients (α̂d )
reflect the differential change in employment for mothers versus other women as in Equation 2.
Standard errors are robust. ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01.
28
WFH & Motherhood Penalties Harrington & Kahn

Figure 3: Changes in Dimensions of WFH


(a) Primarily Work from Home (b) % WFH in Time Diary

(c) WFH after 6pm (d) Manager WFH after 6pm

Notes: This figure shows trends in different dimensions of WFH. Panel (a) uses responses to the
American Community Survey’s question about whether individuals primarily worked from home.
Panels (b)-(d) use time-diary data from the American Time Use Survey (ATUS) that capture where
people worked on any given day. We limit the sample to weekdays in which individuals worked
at least 5 hours. Since college degrees are not recorded in the ATUS, the degree information is
imputed from the respondent’s current occupation (see Section II.A for details). Panel (b) measures
the percent of work hours that are done at home. Panel (c) focuses on work hours after 6pm. Panel
(d) focuses on where managers work from home after 6pm.

29
WFH & Motherhood Penalties Harrington & Kahn

Figure 4: Rise of WFH & Narrowing of Motherhood Employment


Gaps

Notes: This figure shows the relationship between the increase in WFH in college degrees (on the
x-axis) and narrowing in the motherhood employment gap among women with that degree (on the
y-axis). Each point represents a different college degree. The size of the point reflects the number
of people with the college degree. The fit line comes from Equation 3. Standard errors are robust.
∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01.

30
WFH & Motherhood Penalties Harrington & Kahn

Figure 5: Rise of WFH & Narrowing of Motherhood Gaps in Other


Labor Market Outcomes

Notes: This replicates Figure 4 for alternative labor-market outcomes. The fit lines and annotated
coefficients come from Equation 3. Standard errors are robust. ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01.

31
WFH & Motherhood Penalties Harrington & Kahn

Figure 6: Triple Difference with Motherhood Employment Gap vs.


Fatherhood Employment Gap

Notes: This replicates Figure 4 but compares the change in the motherhood employment gap (be-
tween mothers and other women) and the change in the fatherhood employment gap (between
fathers and other men). The fit lines and annotated coefficients come from versions of Equation 3.
Standard errors are robust. ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01.

32
WFH & Motherhood Penalties Harrington & Kahn

Table 1: Rise of WFH & Narrowing of Motherhood Employment


Gaps

∆ Motherhood Gap in Employment (2009-2019)


(1) (2) (3) (4) (5)
∆ WFH 1.29∗∗∗ 1.33∗∗ 1.50∗∗∗ 1.35∗∗ 1.35∗∗
(0.49) (0.52) (0.52) (0.56) (0.56)

∆ Hours 0.16 0.64 0.63 1.20


(0.67) (0.72) (0.72) (0.80)

∆ % Advanced Degree −0.22 −0.23∗ −0.22


(0.13) (0.13) (0.13)

∆ Income ($1,000s) 0.08 0.10


(0.11) (0.11)

∆ % Female in Degree 0.26


(0.16)

Constant 0.10 0.09 0.13 0.14 0.12


(0.10) (0.11) (0.11) (0.11) (0.11)

Observations 75 75 75 75 75
R2 0.09 0.09 0.12 0.13 0.16

Notes: This table presents the relationship between the increase in WFH in college degrees and
narrowing in the motherhood employment gap among women with that degree. Each column
estimates Equation 3. Columns 2−5 add additional controls for other changes in the nature of work
in college degrees (each estimated according to Equation 1). Each observation is a different college
degree and regressions are weighted to put more weight on larger majors for women. Standard
errors are robust. ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01.

33
WFH & Motherhood Penalties Harrington & Kahn

Table 2: Instrumenting Rise of WFH

∆ WFH ∆ Motherhood Gap in Employment (2009-2019)


(1) (2) (3) (4) (5) (6)
Doesn’t Require Physical Presence 0.10∗∗∗
(0.01)

∆WFH
\ 1.97∗∗∗ 2.20∗∗ 2.47∗∗∗ 2.46∗∗ 2.14∗∗
(0.74) (0.85) (0.88) (1.05) (1.00)

∆ Hours 0.53 1.15 1.15 1.57∗


(0.74) (0.82) (0.85) (0.90)

∆ % Advanced Degree −0.26∗ −0.26∗ −0.24∗


(0.14) (0.14) (0.14)

∆ Income (1, 000s) 0.004 0.05


(0.12) (0.12)

∆ % Female in Degree 0.27


(0.17)

Constant 0.14∗∗∗ −0.04 −0.07 −0.04 −0.04 −0.01


(0.01) (0.15) (0.16) (0.16) (0.18) (0.18)

First Stage F-Stat 60.1 30.8 20.8 16.4 13.4


Observations 75 75 75 75 75 75
R2 0.45 0.06 0.05 0.08 0.08 0.14

Notes: This table instruments the rise of WFH based on the need for physical presence in the college
degree. This is based on Dingel and Neiman (2020)’s classification, where we have also included
educators as needing to be in-person. Column 1 presents the first stage of Equation 4. Columns 2−6
estimate the two-stage least squares of Equation 5. Each observation is a different college degree
and regressions are weighted to put more weight on larger majors for women. Standard errors are
robust. ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01.

34
WFH & Motherhood Penalties Harrington & Kahn

A Appendix Figures and Tables

Figure A.1: Work-from-Home (WFH) and Secondary Child-care


Panel (a): Secondary Child-care in WFH and Work Away from Home

Panel (b): Time-Series of Secondary Child-care

Note: This figure illustrates how work from home is often paired with secondary child-care. Data
comes from the American Time Use Survey (ATUS). The sample is limited to respondents aged 20
to 55 with an eldest child under ten who were surveyed on a weekday and reported working at
least two hours that day. We consider time in which the respondents report their primary activity
as working but the secondary activity as looking after children. Panel (a) shows the share of WFH
work-time and not WFH work-time that involves secondary child-care for mothers and fathers.
Panel (b) shows how the share of all work-time devoted to secondary child-care changed around
the rise of WFH caused by COVID-19.

35
WFH & Motherhood Penalties Harrington & Kahn

Table A.1: Rise of WFH Among Men & Narrowing of Motherhood


Employment Gaps

∆ Motherhood Gap in Employment (2009-2019)


(1) (2) (3) (4) (5)
∆ WFH for Men 1.09∗∗∗ 1.09∗∗ 1.09∗∗ 0.95∗∗ 1.24∗∗∗
(0.41) (0.43) (0.42) (0.46) (0.46)

∆ Hours 0.05 0.32 0.34 1.26


(0.65) (0.70) (0.70) (0.79)

∆ % Advanced Degree −0.15 −0.17 −0.14


(0.13) (0.13) (0.13)

∆ Income ($1,000s) 0.09 0.10


(0.11) (0.10)

∆ % Female in Degree 0.38∗∗


(0.17)

Constant 0.19∗∗ 0.19∗∗ 0.23∗∗∗ 0.24∗∗∗ 0.17∗


(0.07) (0.07) (0.08) (0.08) (0.09)

Observations 75 75 75 75 75
R2 0.09 0.09 0.10 0.11 0.18

Notes: This table replicates Table 1 but estimates the change in WFH using only the behavior of
men. Standard errors are robust. ∗ p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01.

36

You might also like