THE FUTURE OF WORK: WORK FROM HOME PREPAREDNESS AND FIRM RESILIENCE

DURING THE COVID-19 PANDEMIC

John (Jianqiu) Bai, Erik Brynjolfsson, Wang Jin, Sebastian Steffen, and Chi Wan *

Abstract

We use the universe of U.S. job postings to construct a firm-level index that assesses each

company’s work from home (WFH) suitability, and study its impact on firms’ resilience

during the recent COVID-19 pandemic. Using a difference-in-differences framework, we find

strong evidence that, compared to their low-WFH peers, firms with high WFH index values

have higher stock returns, lower return volatility, and better financial performance during

the pandemic. Our results are robust to stringent empirical specifications that include a

variety of fixed effects. Contrary to conventional wisdom, our results hold primarily in non-

high-tech industries.

Key Words: Work from Home, Performance, COVID-19

JOE: D24, L23, L84, M11, M54, O31

* John Bai is from the D’Amore-McKim School of Business at Northeastern University, Boston, MA, 02115; and
can be reached at j.bai@northeastern.edu; Erik Brynjolfsson is from Stanford University Digital Economy
Labratory and can be reached at erik.brynjolfsson@gmail.com; Wang Jin is from MIT Sloan School of Management,
Initiative on the Digital Economy, Cambridge, MA; and can be reached at jwangjin@mit.edu; Sebastian Steffen is
from MIT Sloan School of Management, Cambridge, MA, and can be reached at ssteffen@mit.edu; Chi Wan is
from the College of Management at the University of Massachusetts Boston; and can be reached at
chi.wan@umb.edu.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 THE FUTURE OF WORK: WORK FROM HOME PREPAREDNESS AND FIRM RESILIENCE

DURING THE COVID-19 PANDEMIC

Abstract

We use the universe of U.S. job postings to construct a firm-level index that assesses each

company’s work from home (WFH) suitability, and study its impact on firms’ resilience

during the recent COVID-19 pandemic. Using a difference-in-differences framework, we find

strong evidence that, compared to their low-WFH peers, firms with high WFH index values

have higher stock returns, lower return volatility, and better financial performance during

the pandemic. Our results are robust to stringent empirical specifications that include a

variety of fixed effects. Contrary to conventional wisdom, our results hold primarily in non-

high-tech industries.

Key Words: Work from Home, Performance, COVID-19

JOE: D24, L23, L84, M11, M54, O31

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 1. Introduction

Working from home (WFH), sometimes referred to as telework, telecommuting, or

remote work, has become an increasingly common practice and constitutes an important

dimension in the future of work.1 A recent survey by Gallup estimates that more than 43%

of the workers reported working remotely at least one day a week in 2017, and this number

is expected to significantly increase in the coming decades.2 While the coronavirus outbreak

puts WFH once again under the spotlight and likely will cause many firms to shift to WFH

as the new work norm, firms vary greatly in their adoption of WFH before the onset of the

outbreak. In this paper, we study the impact of WFH suitability on firms’ resilience of the

COVID-19 pandemic.

The COVID-19 pandemic caused by the SARS-CoV-2 virus is an ongoing outbreak for

which to date there is no proven therapeutics or available vaccine. As of May 27, 2020, the

pandemic has infected more than 5.6 million people and claimed over 350,000 lives around

the globe.3 In the United States, the pandemic has resulted in some of the most significant

disruptions to social and economic activities, as non-essential businesses are ordered to shut

down and individuals are ordered to stay home.4

In theory, the effect of WFH on firms’ resilience during the COVID-19 pandemic is

ambiguous. On one hand, WFH could help firms ensure continued communication with their

stakeholders such as customers and suppliers, thus mitigating the disruptive effect of

COVID-19 on firms’ operations. We term this the Efficiency-Improving Hypothesis. On the

other hand, WFH could result in reduced employee productivity, morale, and even mental

1 See U.S. Census report: https://www.census.gov/library/visualizations/2013/comm/home_based_workers.html

2 https://www.nytimes.com/2017/02/15/us/remote-workers-work-from-home.html
3 https://coronavirus.jhu.edu/map.html
4 As of April 4, 2020, most of the U.S. states have instituted some form of stay-at-home executive order. See

https://www.nytimes.com/interactive/2020/us/coronavirus-stay-at-home-order.html for details.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 well-being, as it blurs the boundary between personal and professional life. We term this the

Morale-Reducing Hypothesis. As a result, the overall impact of WFH on firm performance

during the COVID-19 pandemic is largely an empirical question.

To test these two competing hypotheses, we merge the near-universe of U.S. job

postings from Burning Glass Technologies (BGT) with a recently developed occupational

WFH suitability indicator by Dingel and Neiman (2020) to construct, for the first time, a firm-

level measure of the percentage of its workforce that has the WFH option. We then estimate

a difference-in-differences regression model to examine whether a firm’s WFH index prior to

the COVID-19 pandemic influences its stock and financial performance during this crisis.

Our main finding is that, compared to their low-WFH counterparts during the

COVID-19 pandemic, high-WFH firms have higher stock returns, lower return volatility, and

better financial performance (measured by net income, sales, return on equity, and capital

expenditure). These results hold with stringent empirical specifications that control for a host

of firm-level characteristics and fixed effects.

One potential identification challenge is that the firm-level WFH suitability is not

randomly assigned, but rather every company’s endogenous choice. However, our identifying

assumption is that firms’ pre-COVID-19 choice of WFH is reasonably orthogonal to the

pandemic. This condition is likely satisfied given the unanticipated nature of the rapid spread

of coronavirus.

Our study contributes to two strands of literature. By studying the determinants of

firm resilience during the COVID-19 pandemic, our paper joins a recent body of research that

examines contributing factors to firms’ differential performance during the crisis episodes.

Albuquerque, Koskinen, Yang, and Zhang (2020) and Lins, Servaes, and Tamayo (2017) find

that firms with high environmental and social ratings tend to perform better during the

COVID-19 pandemic and the 2007-2008 financial crisis, respectively. Other studies find

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 evidence that firms with access to liquidity (Acharya and Steffen, 2020), high cash holdings

(Ramelli and Wagner, 2020), or strong balance sheet (Ding, Levine, Lin, and Xie, 2020) tend

to perform better in the first quarter of 2020. Our study expands the understanding of the

determinants of firm resilience during the crisis period by taking a more labor-related focus.

Specifically, our evidence suggests that flexible work arrangements such as WFH can

significantly reduce operational disruptions that firms experience in difficult times and help

with their resurgence in the aftermath of the crisis.5

Second, our study contributes to the recent literature that studies the shift of the

workplace norm towards a WFH approach. In particular, Brynjolfsson et al. (2020) surveys a

nationally-representative sample of the U.S. population and finds a strong negative

relationship between the fraction in a state still commuting to work and the fraction working

from home. They also report that the switch to WFH can be predicted by the incidence of

COVID-19. We take a different approach by conditioning on firms’ ex ante preparedness in

WFH and investigating how such differences lead to differential performance during the

pandemic.

The remainder of the paper proceeds as follows. Section 2 develops the testable

hypothesis. Section 3 discusses the data and summary statistics. We present our empirical

findings in Section 4 and conclude in Section 5.

2. Hypothesis Development

Why would WFH affect firm resilience during the COVID-19 pandemic? In theory, the

impact of WFH on firm performance is ambiguous.

5 Other papers that examine the labor aspects during the COVID-19 crisis include the following: Benzell, Collis,
and Nicolaides (2020) compare the actual closures of commercial locations to what they suggest should be closed
first, and generate implications for the optimal sequence of reopenings when policymakers revive the economy.
Arkeson (2020) analyzes the economic consequences of the COVID-19 pandemic and how these correlate with
various assumptions of the ratio between the susceptible – infected – recovered groups in the population. Khan,
Lange, and Wiczer (2020a; 2020b) and Coibion, Gorodnichenko, and Weber (2020) study the labor market
implications of the COVID-19 pandemic.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 First, WFH enables employees to maintain continuous communication with the firm’s

customers and suppliers, thus ensuring a less disrupted operation during the period of a

shutdown. The enhanced sustainability and continuation associated with firms that allow for

WFH naturally imply that these firms should be less affected by the adverse shocks caused

by the COVID-19-induced business closures. We term this effect as the WFH Efficiency-

Improving hypothesis.

Meanwhile, the effect of WFH on worker morale and satisfaction is ambiguous. While

some studies (Bloom, Liang, Roberts, and Ying, 2015) find that WFH improves worker

satisfaction and lowers attribution rate, WFH does entail a significant change in work and

lifestyle and takes a fair amount of getting used to. Furthermore, Hochschild (2000) and

Clark (2000) both suggest that the separation between home and work is an important

division for workers’ mental wellbeing. By blurring this distinction, WFH could create a

detrimental effect that potentially hurts employee morale and productivity. This could in

turn lead to the firm’s worsened financial and stock market performance. We term this effect

as the WFH Morale-Reducing hypothesis.

We summarize the two opposing predictions in the following competing hypotheses.

H1 (WFH Efficiency-Improving Hypothesis): Firms’ financial performance and stock

return during the COVID-19 pandemic are positively correlated with firms’ WFH suitability.

H1a (WFH Morale-Reducing Hypothesis): Firms’ financial performance and stock

return during the COVID-19 pandemic are negatively correlated with firms’ WFH suitability.

3. Variable Construction and Summary Statistics

3.1 Data Source

This section details the construction of the proposed WFH index and various firm

performance metrics.

3.1.1 WFH Index

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 To construct a firm-level proxy of WFH index, we proceed in several steps: First, we

obtain job postings data from Burning Glass Technologies (BGT). Although BGT only starts

its full coverage in 2010,6 it is considered a high-quality data source that keeps a detailed

record of each job posting, and therefore provides coverage for the near-universe of job

postings in the US. For each job posting, BGT labels it with an occupational title (6-digit SOC

code) and records the employer. We aggregate this data to the employer-SOC (6-digit) level

to obtain a measure of how many job postings the employer posts in each year for each SOC

occupation. Second, we merge this data with the WFH suitability data from Dingel and

Neiman (2020). We first aggregate their binary index, which is originally available at the 8-

digit Occupational Information Network (i.e., O*NET) code level,7 to the 6-digit SOC level

and then subsequently to the firm level. This procedure enables us to construct, for each firm,

the percentage of its incoming workers that is likely to have a WFH option. Ideally, one would

like to have such a measure constructed from all existing employees, which would capture

the actual prevalence of WFH within an organization. Since such data is not available, we

take the next best alternative and use the average value of each firm’s annual WFH index

over the 2010-2019 period.8

3.1.2 Accounting Fundamentals & Stock Return Data

We obtain from Compustat quarterly accounting information from 2019 Q1 to 2020

Q1. We also require that all observations have non-missing information for total assets, sales,

6 BGT provides partial coverage in 2007, but does not provide any coverage of job postings in 2008 or 2009. We
therefore use BGT data from 2010.
7 O*Net is a free online database that contains hundreds of occupational definitions to help students, job seekers,

businesses and workforce development professionals to understand today's world of work in the United States. It
was developed under the sponsorship of the US Department of Labor/Employment and Training Administration
(USDOL/ETA) through a grant to the North Carolina Employment Security Commission (now part of the NC
Commerce Department) during the 1990s.
8 Note that although this approach is also not perfect, it does help alleviate the issue of whether or not using the

“flow” vs. “stock” measure to construct WFH is the more appropriate approach.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 capital expenditures, and net income. We obtain stock return data from the Center for

Research in Security Prices (CRSP).

3.1.3 Merging COMPUSTAT and Burning Glass Technologies

Since Compustat and BGT do not share a common firm identifier, we take a multi-

step approach to merge the two databases. Specifically, we use a combination of name and

address fuzzy matching to construct a bridge between Compustat and BGT data. We use

several methods including Soundex and Levenshtein distance to ensure match quality.

In some cases, an employer name in the BGT data is a subsidiary of a Compustat firm

but its name is distinct from its parent, thus the existing algorithm cannot recognize their

connection. To resolve this problem, we follow Campello, Murillo and Gao (2019) and match

the remaining employers to the subsidiaries of Compustat firms using information extracted

from historical Orbis data provided by Bureau van Dijk (BvD). Orbis traces the evolution of

firms’ organizational structure through time, maintaining the parent-subsidiary

correspondence. This historical information is robust to subsidiary opening, closing, and

ownership changes, which is crucial for accurate matching. 9 We manually check the links

identified to ensure the accuracy of our matching.

3.2 Summary Statistics

Following the procedure in 3.1.3, we are able to match over 3800 unique firms in the

Compustat. For instances where a Compustat firm has multiple subsidiaries, the firm-level

WFH index takes a weighted average of all the subsidiaries’ WFH indices, where the weight

is each subsidiary’s number of job postings. Our final analytical sample includes 7457

observations corresponding to 2,121 unique firms. Table 1 Panel A shows that the mean value

9 We refer interested readers to Campello et al for a more detailed description of this part of the matching exercise.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 of the WFH index is 0.596 with a standard deviation of 0.216, which indicates a sizeable

amount of variation in the cross section.

[Table 1 about here]

Panel B breaks down the key variables into pre- and post-COVID-19 periods and

contrasts their values between subsamples with 25% highest pre-COVID-19 WFH score

(HighWFH=1) and the rest (HighWFH=0). We observe a differential impact of COVID-19 on

high- and low-WFH firms. For instance, while the two groups of firms have similar average

quarterly stock returns before the pandemic (8.4% vs 8.0%), the high-WFH firms experience

a relatively smaller decline during the recent market crash than low-WFH firms (-24% vs -

30%). Similarly, while the average net income of high-WFH firms increased modestly from

$365.9 million in 2019 Q4 to $393.7 million in 2020 Q1, the measure for low-WFH firms has

declined from $196.1 million to $183.0 million over the same period.

4. Empirical Methodology & Results

4.1 Empirical Methodology

To test our hypothesis, we employ a difference-in-differences (DID) research design.

Specifically, we estimate the following multi-variate fixed effects regression:

𝑌𝑌𝑖𝑖𝑖𝑖 = 𝛽𝛽(𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝑊𝑊𝑊𝑊𝑊𝑊𝑖𝑖 × 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶19𝑡𝑡 ) + 𝛾𝛾𝛾𝛾𝑖𝑖,𝑡𝑡−1 + 𝑣𝑣𝑖𝑖 + 𝜏𝜏𝑡𝑡 + 𝜀𝜀𝑖𝑖𝑖𝑖 , (1)

where subscript 𝑖𝑖 and 𝑡𝑡 index firm and time (i.e., quarter), respectively. 𝑌𝑌𝑖𝑖𝑖𝑖 is the firm-level

outcome variables such as stock returns and various performance measures including net

income, sales, return on equity, and capital investment. 𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝑊𝑊𝑊𝑊𝑊𝑊𝑖𝑖 is an indicator variable,

which takes the value of one if firm 𝑖𝑖’s average WFH score calculated based on its annual job

posting data during the pre-COVID-19 period (2010-2019) falls into the top quartile of the

sample distribution and zero otherwise. 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶19𝑡𝑡 is an indicator variable that is set equal to

one for 2020 Q1, and zero otherwise. The parameter of interest is 𝛽𝛽, which captures the

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 differential impact of 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶19 on firms with high versus low values of pre-COVID-19 WFH

score. 𝑋𝑋 is an array of firm-level controls including firm size, cash holdings, leverage ratio,

R&D indicator, and the dividend payout dummy. 𝑣𝑣𝑖𝑖 is firm fixed effects, which controls for

time-invariant firm-level characteristics. And 𝜏𝜏𝑡𝑡 is the time fixed effects. Note that the

inclusion of both firm fixed effects and time fixed effects absorb the main effect of 𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝑊𝑊𝑊𝑊𝑊𝑊

and 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶19. Throughout our empirical analyses, standard errors are clustered at the firm

level.10

4.2 Empirical Results

In this section, we present regression results on the impact of pre-COVID-19 WFH on

firm performance during the first quarter of 2020. We focus our attention primarily on firms’

stock performance and financial outcomes.

4.2.1 Stock Performance

In this subsection, we examine whether firms’ stock market performance during the

COVID-19 pandemic is correlated with their ex ante WFH suitability. In March 2020, the

COVID-19 pandemic and the oil price war between Russia and the Organization of the

Petroleum Exporting Counties (OPEC) resulted in the most significant stock market crash.

In particular, between February 12, 2020 and March 23, 2020, all three major stock indices

(i.e., the Dow Jones Industrial Average, the NASDAQ Composite, and S&P 500 Index)

experienced larger than 20% declines.

We examine, as outcome variables, the stocks’ raw returns, abnormal returns,

volatility, and idiosyncratic volatility, and report the results in Table 2. We find that

compared to low-WFH firms, firms with high levels of WFH prior to the COVID-19-induced

10In an alternative specification, we augment the time fixed effect with the state-quarter fixed effects, which is
motivated by the recent findings in Brynjolfsson et al. (2020) that the COVID-19-induced switch to WFH is highly
correlated with the incidence of the pandemic in each state. Our baseline results are fully retained after
controlling for the interactions of time and state (based on firm headquarters) fixed effects.

10

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 shutdown have higher returns and abnormal returns, and meanwhile, lower volatility and

idiosyncratic volatility. 11 Concretely, our estimates in Column (2) suggest that, ceteris

paribus, firms with the top quartile pre-COVID-19 WFH suitability (i.e., HighWFH=1) earn

a 5.2% higher abnormal return than the rest firms in 2020 Q1. These results together suggest

that high values of WFH are correlated with firms’ resilience in the recent market turmoil

sparked by coronavirus fears, and are largely consistent with the Efficiency-Improving

Hypothesis (Hypothesis 1).

[Table 2 about here]

4.2.2 Financial Performance

Using a similar regression framework, we investigate whether WFH correlates with

firms’ financial performance during the first quarter of 2020. As is shown in Table 3: after

controlling for firm-level characteristics such as firm size, cash holdings, leverage ratio, R&D,

and dividend payout dummy, firms that have high WFH levels prior to the COVID-19

pandemic have higher net income, sales, return on equity, as well as capital investment

relative their low-WFH peers. These results corroborates Hypothesis 1 and substantiate that

WFH helps firms minimize their operational disruptions in the pandemic.

[Table 3 about here]

4.2.3 Industry Heterogeneity: High-tech vs. Non-high-tech

We now follow Haltiwanger et al. (2017) to separate our sample into high-tech and

non-high-tech industries and re-estimate our main regressions separately in these

subsamples. Conventional wisdom suggests that firms in high-tech industries often adopt

11Return volatility and idiosyncratic volatility are calculated as the standard deviation of daily returns over a
quarter and that of residuals obtained from fitting a CAPM using daily returns in a quarter. Our construction of
idiosyncratic volatility is in line with Ang, Hodrick, Xing, and Zhang (2006). Table 1 details variable definitions.

11

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 WFH due to the nature of their work, but firms in other industries generally do not have such

flexibility. The results of this exercise are presented in Table 4.

[Table 4 about here]

Panels A1 and A2 present the stock performance results for the high-tech and non-

high-tech industries, respectively. Overall, we find that our results are insignificant in the

high-tech industries, but very pronounced in the non-high-tech industries. Panels B1 and B2

display a similar pattern for firms’ financial performance. While these results may appear

surprising at first glance, they are consistent with the notion in Brynjolfsson and Milgrom

(2013): once the adoption of WFH is already prevalent to begin with, which is likely the case

for the high-tech industry, further improving WFH suitability might not generate additional

comparative advantage in firm performance.

4.2.3 Additional Evidence

We conduct several robustness checks and obtain similar results as the main findings.

First, we examine the effect of WFH suitability before the COVID-19 pandemic to make sure

that there is no pre-trend exist. That is, the performance differences between high- and low-

WFH firms should only be evident in the first quarter of 2020 and not in the quarters

preceding that. Second, we recognize that, although our identification strategy exploits the

fact that a firm’s pre-COVID-19 WFH suitability is unlikely to be correlated with its exposure

to the coronavirus, there is still the possibility that they are driven by some other firm-level

unobserved characteristics. We thus perform a matched sample analysis (using both the

propensity score matching and coarsened exact matching) to identify firms that have a

similar propensity to adopt WFH before the COVID-19 outbreak. Within this matched

subsample, we still find results similar to our main findings. Third, we consider an

alternative WFH proxy (WFH score) and find that our baseline results are fully retained. All

these results are presented in an online appendix, accompanying this paper.

12

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 5. Conclusion

As part of a new norm for the future of work in the aftermath of the COVID-19

pandemic, WFH has and will continue to gain unprecedented popularity. Large-scale

evidence on the effectiveness of WFH, however, is scarce.

In this paper, we exploit a novel dataset of the near-universe of U.S. job postings to

construct a novel firm-level WFH index. We find that firms of high WFH index values perform

better compared to low-WFH firms during the COVID-19 crisis on a number of dimensions

ranging from stock returns and return volatility to financial performance.

Our study provides some of the first glances at how WFH helps firms cope with major

adverse social and economic shocks. Further investigation into the exact mechanisms

through which WFH are linked to performance is a fruitful area for future research. A deep

understanding of these underlying issues is particularly valuable to ensure a more efficient

and smooth adoption of and transition into WFH as employers and employees alike embrace

the new reality in the aftermath of the crisis.

13

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 References
Acharya, V.V., and Steffen, S (2020) The risk of being a fallen angel and the corporate dash
for cash in the midst of COVID. CEPR COVID Economics, 10.
Albuquerque, R.A., Koskinen, Y., Yang, S. and Zhang, C (2020) Resiliency of Environmental
and Social Stocks: An Analysis of the Exogenous COVID-19 Market Crash, Working
paper.
Ang A, Hodrick R J, Xing Y, and Zhang X (2006) The Cross-Section of Volatility and Expected
Returns, Journal of Finance, 61, 259–299.
Apgar M IV (1998) The alternative workplace: Changing where and how people
work. Harvard business review, 76(3), 121-137.
Atkeson, A (2020) What Will be the Economic Impact of COVID-19 in the US? Rough
Estimates of Disease Scenarios. Working paper, National Bureau of Economic Research
(No. w26867).
Benzell, S., Collis, A. and Nicolaides, C (2020) Rationing Social Contact During the COVID-
19 Pandemic: Transmission Risk and Social Benefits of US Locations. Working paper
Bloom, N., Liang, J., Roberts, J. and Ying, Z.J (2015) Does working from home work?
Evidence from a Chinese experiment. The Quarterly Journal of Economics, 130(1),
pp.165-218.
Brynjolfsson E, Horton J, Ozimek A, Rock D, Sharma G, Ye T H (2020) COVID-19 and Remote
Work: An Early Look at US Data. Working Paper.
Brynjolfsson E and Milgrom P (2013) Complementarity in organizations. The handbook of
organizational economics, 11-55.
Campello, M., Gao, J. and Xu, Q (2019) Personal Income Taxes and Labor Downskilling:
Evidence from 27 Million Job Postings.
Clark SC (2000) Work/family border theory: A new theory of work/family balance. Human
relations, 53(6), 747-770.
Coibion, O., Gorodnichenko, Y. and Weber, M (2020) Labor Markets During the COVID-19
Crisis: A Preliminary View, Working paper, National Bureau of Economic Research (No.
w27017).
Ding, W., Levine, R., Lin, C. and Xie, W (2020) Corporate immunity to the covid-19
pandemic (No. w27055). National Bureau of Economic Research.
Dingel, J.I. and Neiman, B (2020) How many jobs can be done at home? (No. w26948).
National Bureau of Economic Research.
Haltiwanger, J., Hathaway, I. and Miranda, J (2014) Declining business dynamism in the
US high-technology sector. Available at SSRN 2397310.
Hochschild A R (2001) The time bind: When work becomes home and home becomes work (Vol.
2). Macmillan.
Kahn, L. B., Lange, F., and Wiczer, D (2020a) Labor Demand in the Time of COVID-19:
Evidence from Vacancy Postings and UI Claims. Working paper.
Kahn, L. B., Lange, F., and Wiczer, D (2020b) Labor Supply in the Time of COVID19. Working
paper.
Lins, K.V., Servaes, H. and Tamayo, A (2017) Social capital, trust, and firm performance: The
value of corporate social responsibility during the financial crisis. The Journal of
Finance, 72(4), pp.1785-1824.
Ramelli, S. and Wagner, A.F (2020) Feverish stock price reactions to covid-19.

14

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 Table 1. Variable Definitions and Summary Statistics
Panel A reports variable definitions and data sources. The rightmost column shows the
sample mean and median (in parentheses). The sample period is 2019Q1-2020Q1. Panel B
contrasts the mean and median (in parentheses) of several key variables in pre- and post-
COVID-19 subsamples (i.e., 2019Q4 and 2020Q1, respectively); and in subsamples with 25%
highest pre-COVID-19 WFH score (HighWFH=1) and the rest (HighWFH=0). All the
continuous variables are winsorized at the 1st and 99th percentiles.

Panel A. Full sample

Mean
Variable Definition and data source
(Median)
Average pre-COVID-19 WFH suitability score (based on the
WFH score universal job posting from the Burning Glass Technology 0.596
database) (0.601)

A dummy variable that takes the value of one if the firm’s

WFH score calculated based on its annual job posting data
HighWFH during the pre-COVID-19 period (2010-2019) falls into the 0.25
top quartile of the sample distribution and zero otherwise. (0)
(based on the universal job postings from the Burning Glass
Technology database)

Proxies of firm performance

Sum of monthly returns in a quarter (Compustat security
Return 0.020
monthly)
(0.043)
Sum of monthly abnormal returns in a quarter. Monthly
abnormal return is the difference between excess return
and the CAPM beta times the market excess return. The
Abn Return -0.013
CAPM beta is estimated using the past 36 monthly returns
(-0.005)
(Compustat security monthly; Kenneth R. French’s Data
Library)

Standard deviation of daily returns in a quarter (Compustat

security daily). The return standard deviation is then
Return Volatility 0.049
multiplied by 2 [i.e., �4 (quarters)] to convert to the annual
(0.037)
basis.

Standard deviation of the residuals obtained from fitting a

daily CAPM in every month for a given firm. The standard 0.039
Idio Volatility
deviation is then multiplied by 2 [i.e., �4 (quarters) ] to (0.031)
convert to the annual basis.
216.9
Net Income Firms’ net income (Compustat quarterly data)
(29.10)
2093
Sales Total sales (Compustat quarterly data)
(369.2)
Return on Equity Net income divided by lagged value of equity (Compustat 0.018
quarterly data) (0.014)

15

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 Firms’ capital expenditure divided by total assets 0.018
Cap Exp / Asset
(Compustat quarterly data) (0.008)

Control variables
8.094
Log Total Asset Logarithm of total assets (Compustat quarterly data)
(8.073)
Book value of debt over total assets (Compustat quarterly 0.279
Leverage
data) (0.251)
Cash holdings divided by total assets (Compustat quarterly 0.106
Cash
data) (0.051)
R&D expenditures over total assets (Compustat quarterly 0.006
R&D
data) (0.000)
Dividend per share scaled by stock price (Compustat 0.004
Dividend
quarterly data) (0.003)

Panel B. Key variables in subsamples

High pre-COVID-19 Low pre-COVID-19
WFH Suitability WFH Suitability
(HighWFH = 1) (HighWFH = 0)
Before After Before After
COVID-19 COVID-19 COVID-19 COVID-19
0.850 0.842 0.517 0.534
WFH score
(0.878) (0.878) (0.536) (0.564)

0.084 -0.243 0.080 -0.300

Return
(0.074) (-0.233) (0.075) (-0.316)
-0.010 -0.031 -0.014 -0.086
Abn Return
(-0.012) (-0.032) (-0.012) (-0.093)
0.040 0.103 0.037 0.103
Return Volatility
(0.031) (0.094) (0.029) (0.097)
0.035 0.063 0.032 0.065
Idio Volatility
(0.027) (0.056) (0.026) (0.058)

365.9 393.7 196.1 183.0

Net Income
(35.25) (46.85) (28.2) (28.4)
2249.2 2813.0 2200.3 2094.1
Sales
(335.1) (439.3) (393.4) (439)
0.020 0.005 0.028 0.007
Return on Equity
(0.011) (0.003) (0.017) (0.004)
0.017 0.014 0.018 0.013
Cap Exp / Asset × 100
(0.012) (0.009) (0.015) (0.010)

16

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 Table 2. WFH Suitability and Stock Market Reactions
This table implements a difference-in-differences (DID) research design to examine the
differential impact of COVID-19 on stock market reactions between firms with high and low
WFH suitability. The dependent variables are total return and abnormal return in Columns
(1) and (2); and return volatility and idiosyncratic volatility in Columns (3) and (4),
respectively. The regression specification is specified in Equation (1), and all variables are
defined in Table 1. We further include Tobin’s Q and return on equity (ROE) in the regression
analysis. Firm and time fixed effects are also included in the estimation. Standard errors are
clustered at the firm level. *, **, and *** indicate statistical significance at the 10%, 5%, and
1% levels, respectively.

Dependent Variable Return Abn Return Volatility Idio Volatility

(1) (2) (3) (4)
HighWFH × COVID 0.057*** 0.052*** -0.005** -0.005***
(3.40) (3.16) (-2.40) (-3.14)
Size -0.143*** -0.109*** -0.005 -0.005*
(-3.83) (-2.99) (-1.28) (-1.69)
Cash 0.133* 0.106 -0.009 -0.003
(1.68) (1.50) (-0.82) (-0.44)
Leverage 0.056 0.015 0.015** 0.014***
(0.87) (0.23) (2.46) (2.96)
R&D 1.405*** 1.214** -0.064 -0.091*
(2.61) (2.32) (-1.06) (-1.87)
Dividend 0.649 0.493 0.034 0.031
(0.93) (0.72) (0.44) (0.54)
Tobin’s q -0.140*** -0.121*** -0.001 -0.001*
(-12.88) (-10.26) (-0.77) (-1.86)
ROE 0.059 0.032 0.003 -0.002
(1.44) (0.83) (0.62) (-0.66)
Firm FE Y Y Y Y
Quarter FE Y Y Y Y
Observations 7431 7431 7431 7431
Adj. R2 0.495 0.148 0.752 0.744

17

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 Table 3. WFH Suitability and Financial Performance
This table implements a difference-in-differences (DID) research design to examine the
differential impact of COVID-19 on various performance metrics between firms with high
and low WFH suitability. The dependent variables across columns 1 to 4 are the logarithm
of net income, logarithm of sales, return on equity, and capital expenditure over total asset,
respectively. The regression specification is specified in Equation (1), and all variables are
defined in Table 1. Firm and time fixed effects are also included in the estimation. Standard
errors are clustered at the firm level. *, **, and *** indicate statistical significance at the
10%, 5%, and 1% levels, respectively.

Log Net Return on Cap Exp /

Dependent Variable Log Sales
Income Equity Total Asset
(1) (2) (3) (4)
HighWFH × COVID 0.152** 0.026** 0.003* 0.003***
(0.072) (0.011) (0.001) (0.001)
Size -0.076 0.336*** -0.007* -0.020
(0.160) (0.033) (0.004) (0.013)
Cash -0.909** -0.338*** -0.020** 0.003
(0.365) (0.081) (0.010) (0.015)
Leverage 0.833** -0.096* 0.036*** 0.052*
(0.333) (0.056) (0.008) (0.031)
R&D 0.557*** 0.010 -0.002 -0.004
(0.191) (0.060) (0.010) (0.007)
Dividend -2.313 -0.671 0.013 0.151
(3.725) (0.659) (0.121) (0.141)
Firm FE Y Y Y Y
Quarter FE Y Y Y Y
Observations 7,457 7,457 7,457 7,457
Adj. R2 0.924 0.998 0.660 0.742

18

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 Table 4. WFH Suitability and Firm Performance: High-tech vs Other Industries
This table re-assesses the results reported in Tables 2 and 3 in two subsamples: high tech vs
other industries. The high-tech industries are defined as in Haltiwanger et al (2017),
including: 3341 Computer and peripheral equipment manufacturing; 3342 Communications
equipment manufacturing; 3344 Semiconductor and other electronic component
manufacturing; 3345 Navigational, measuring, electromedical, and control instruments
manufacturing; 3254 Pharmaceutical and medicine manufacturing; 3364 Aerospace product
and parts manufacturing; 5112 Software publishers; 5161 Internet publishing and
broadcasting; 5179 Other telecommunications; 5181 Internet service providers and Web
search portals; 5182 Data processing, hosting, and related services; 5413 Architectural,
engineering, and related services; 5415 Computer systems design and related services; 5417
Scientific research-and-development services. Panels A1 and A2 report the results with
return and volatility as the dependent variables. Panels B1 and B2 repeat the exercise with
financial performance proxies as the dependent variables. Firm and time fixed effects are
also included in the estimation. Standard errors are clustered at the firm level. *, **, and ***
indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Panel A1. Stock Market Reactions: High-tech Industries

Dependent Variable Return Abn Return Volatility Idio Volatility
(1) (2) (3) (4)
HighWFH × COVID -0.003 -0.031 0.004 0.000
(-0.09) (-0.96) (0.94) (0.07)
Other controls Y Y Y Y
Firm FE Y Y Y Y
Quarter FE Y Y Y Y
Observations 1092 1092 1092 1092
Adj. R2 0.321 0.080 0.710 0.763

Panel A2. Stock Market Reactions: Other Industries

Dependent Variable Return Abn Return Volatility Idio Volatility
(1) (2) (3) (4)
HighWFH × COVID 0.039** 0.031** -0.002 -0.003**
(2.54) (2.14) (-1.15) (-2.22)
Other controls Y Y Y Y
Firm FE Y Y Y Y
Quarter FE Y Y Y Y
Observations 6361 6361 6361 6361
Adj. R2 0.479 0.102 0.754 0.744

Panel B1. Financial Performance: High-tech Industries

Return on Cap Exp /
Dependent Variable Log Net Income Log Sales
Equity Total Asset
(1) (2) (3) (4)
HighWFH × COVID -0.048 -0.008 0.004 0.003
(0.163) (0.021) (0.003) (0.002)

19

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893
 Other controls Y Y Y Y
Firm FE Y Y Y Y
Quarter FE Y Y Y Y
Observations 1,071 1,071 1,071 1,071
Adj. R2 0.953 0.999 0.675 0.872

Panel B2. Financial Performance: Other Industries

Return on Cap Exp /
Dependent Variable Log Net Income Log Sales
Equity Total Asset
(1) (2) (3) (4)
HighWFH × COVID 0.175** 0.028** 0.002 0.002*
(0.089) (0.014) (0.001) (0.001)
Other controls Y Y Y Y
Firm FE Y Y Y Y
Quarter FE Y Y Y Y
Observations 6,386 6,386 6,386 6,386
Adj. R2 0.924 0.998 0.681 0.758

20

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=3616893