See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/351376324

Work from Home & Productivity: Evidence from Personnel & Analytics Data on
IT Professionals

Preprint · November 2021

CITATIONS READS

0 2,952

3 authors, including:

Michael Gibbs
University of Chicago
60 PUBLICATIONS 2,854 CITATIONS

SEE PROFILE

All content following this page was uploaded by Michael Gibbs on 01 April 2022.

The user has requested enhancement of the downloaded file.

 Work from Home & Productivity: Evidence from
Personnel & Analytics Data on IT Professionals∗

Michael Gibbs† Friederike Mengel‡ Christoph Siemroth§

April 1, 2022

Abstract

We study employee productivity (output per hour worked) before and during the working from
home period of the Covid-19 pandemic, using personnel and analytics data from over 10,000 skilled
professionals at a large Asian IT services company. Hours worked increased, including a rise of
18% outside normal business hours. Average output declined slightly and employee productivity
fell 8-19%. We then analyze determinants of changes in productivity. An important source is
higher communication costs. Time spent on coordination activities and meetings increased, while
uninterrupted work hours shrank considerably. Employees networked with fewer individuals and
business units, both inside and outside the firm. They received less coaching and 1:1 meetings with
supervisors. The findings suggest key issues for firms to address in implementing remote work.

Keywords: Collaboration, Coordination, Productivity, Remote Working, Telecommuting, Work-

ing From Home, Work Hours, Work Time

JEL Classification: D2, M5.

∗
We are grateful to several employees of HCL Technologies who spent a great deal of time helping us collect the data
and understand the firm and context, during a difficult period of time. We thank the Tata Center for Development at
the University of Chicago for funding to help cover the cost of WPA licenses. We appreciate helpful comments from
Anna Dreber Almenberg (the editor), the referees, Sonia Bhalotra, Ron Burt, Matthew Clancy, Hans Peter Grüner,
Emma Harrington, Kathryn Ierulli, Duk Gyoo Kim, Alan Kwan, Adam Ozimek, Tim Perri, Paul Smeets, Wim Van der
Stede, and seminar participants at the Bureau of Labor Statistics, Universities of Mannheim and Essex, and the Latin
American and Caribbean Economics Association annual meeting. Siemroth was supported by the Economic and Social
Research Council [grant ES/T006048/1].
†
University of Chicago & IZA. E-mail: [email protected].
‡
University of Essex, Department of Economics, Wivenhoe Park, Colchester, CO4 3SQ, UK and Department of
Economics, Lund University, Tycho Brahes vaeg 1, Lund, Sweden. E-mail: [email protected].
§
University of Essex, Department of Economics, Wivenhoe Park, Colchester, CO4 3SQ, UK. E-mail:
[email protected].
1 Introduction

Working from Home [WFH] has been rising for years, as more occupations use computers and telecom-
munications, more people have reliable home Internet connections, and more families have both parents
working full time. The Covid-19 pandemic accelerated this process by forcing a large fraction of the
global workforce to switch to WFH at least temporarily. Even if only a fraction of this shift became
permanent, it would have implications for urban design, infrastructure development and reallocation
of investment from inner cities to residential areas. It would also have significant implications for how
businesses organize and manage their workforces. However, little is yet known about some of the more
fundamental consequences of WFH, including its effects on employee productivity.
Compared to Working from the Office [WFO], WFH has the potential to reduce commute time,
provide more flexible working hours, increase job satisfaction, and improve work-life balance. Some
employees may find it easier to concentrate if they have a quiet space at home. On the other hand,
collaboration, innovation and interactions with clients may suffer. That WFH was not used for the
majority of the workforce prior to the pandemic suggests that for many employers there was at least
a perception that WFH productivity is lower compared to WFO productivity. Indeed some firms
question the sustainability of extensive WFH (WSJ, 2020; Financial Times, 2021b).
In this paper we provide a comprehensive analysis of the effect of WFH on employee productivity
at HCL Technologies, a large IT services company based in India. The company abruptly switched
all employees from WFO to WFH in March 2020, in response to the largely unanticipated pandemic
shock. Our study has several novel and interesting features.
First, our sample is of a type not previously studied in the small WFH literature: high-skilled
employees in cognitively demanding jobs involving collaboration and innovation. Previous studies
either use survey self reports or focus on low-skilled, individualistic and narrow jobs. Our sample
therefore sheds light on the potential for WFH among the large fraction of the workforce in previously
unstudied professional occupations.1
Second, the firm provided unusually broad and high quality data. Employee productivity for high
skilled jobs is notoriously hard to measure. The data include a quantitative measure of employee
output, which the firm goes to significant lengths to calculate, and uses to supervise and evaluate
these high-skilled professionals. Employee time use was captured by monitoring applications on work
devices, so we know total hours worked, start and end times, how much time was spent in various
types of meetings and communications, and how much time the employee focused on work without
interruptions. The high quality performance measure and total work time provide a natural and
relatively accurate measure of the employee’s productivity. These outcome measures are some of the
best that modern analytics and monitoring software can provide. The firm also provided information
on the types of meetings and communications in which employees engaged, providing valuable insights
1
The need for research on WFH for these type of professions is highlighted by the fact that the incidence of WFH is
highest for these type of jobs. Professionals, managers, knowledge workers, those in clerical support or data processing,
and those with higher education or income make more use of WFH (Bick et al., 2020; Brynjolfsson et al., 2020; Zimpel-
mann et al., 2021; Gottlieb et al., 2021; Hensvik et al., 2020). The industry and occupations analyzed here are among
those predicted to most effectively switch to WFH (Dingel and Neiman, 2020; Adams-Prassl et al., 2020).

2
into how employees spent their work time in WFH compared to WFO. Finally, we know employee
experience, age, gender, whether there are children at home, and their usual commute time.
Since we can link employee productivity, demographics, and detailed communication logs, our rich
dataset allows for in-depth analysis of the causes of employee productivity differences when working
remotely, which has not been possible in other studies. Analyses of unskilled jobs find that WFH
may improve productivity (Bloom et al. (2014); Emanuel and Harrington (2021)), but it is an open
question whether these effects extend to more complex jobs. Our analyses provide valuable insights
into the types of work for which WFH may be especially challenging.
Our key findings are as follows. Employees significantly increased average hours worked during
WFH. Much of this came from starting work earlier and ending it later in the day. At the same time,
there was a slight decline in output as measured by the employer’s primary performance measure.
Combining these, we estimate that average employee output per hour of work declined by 8-19%.
The effects on work time and productivity materialize immediately after the transition to WFH.
In contrast, there is no change in our outcome measures at the start of the pandemic before WFH
was implemented. Changes in work time and productivity also do not co-move with the evolution of
the pandemic, such as the rate of infections or easing of lockdown measures.2 These show that WFH
and not the pandemic best explain these changes.
The increase in overall working hours and corresponding decrease in productivity are associated
with substantial changes in working patterns during WFH. Employees spent more time participating
in a larger number of shorter, larger group meetings, but less time in personal or small group meetings
with their manager. They had less “focus time”; i.e., work time uninterrupted by meetings or calls.
At the same time, they narrowed the scope of their networks, engaging in fewer contacts with col-
leagues and organizational units inside and outside of the firm. All of these factors were significantly
correlated with changes in employee productivity. These findings are evidence that coordination and
communication is more difficult with remote work.
We also uncover some interesting dimensions of heterogeneity. Employees with lower company
tenure decreased output slightly more during WFH, whereas output remained about the same for
those with longer tenure. This suggests that employees who are more adapted to firm culture and
processes are better able to work remotely, where there is no colleague at the next desk for quick help
or advice. As a separate effect, those with greater career experience increased hours worked during
WFH more than those with lower experience, with no effect on productivity. This suggests that more
senior employees, with greater managerial duties, spent more time coordinating during WFH.
Employees with children at home had a greater decline in productivity than those without, but
even those without suffered significant productivity losses. Women were more negatively affected by
WFH than men, but this gender difference was not due to the presence of children in the home. We
2
Further, as with many information technology firms, the company’s business performance was strong throughout
the pandemic, so employees were did not have increased risk of job loss or decreased promotion prospects. There was a
decline, not an increase, in sick days during WFH.

3
conjecture that it might be due to other demands placed on women in the domestic setting while
working from home.3
Our analysis and insights contribute to an emerging literature on how WFH affects work patterns,
employee productivity and well-being. Most of the early research in this area is based on survey-
data. Etheridge et al. (2020) find that employees who work from home state that they are about as
productive as in the office and those who perceive declines experience lower levels of well-being from
WFH. In Barrero et al. (2021) employees report benefits from lower commute time, more flexible work
hours, and increased productivity, but Bellmann and Hübler (2020) find that working remotely has
no long-run effect on work-life balance, and increases job satisfaction only temporarily. Barrero et al.
(2020) estimate that WFH reduced self reported US commuting time by more than 60 million hours
per work day at the height of the pandemic. Further, employees report that about 35% of this time
saved was reallocated to work. In our data, variation in estimated commute time does not predict the
increase in WFH work hours that we observe.
Survey data may be biased if employees enjoy WFH and hope to retain the practice, conflate output
with productivity, or refer to individual performance without consideration of effects on colleagues. A
few WFH studies use employee performance data. Bloom et al. (2015) showed that both hours worked
and productivity increased during WFH for call-center employees at a Chinese firm. Emanuel and
Harrington (2021) studied call-center workers at a large US company, including some who abruptly
moved to WFH in response to Covid-19. Productivity rose in the switch to remote work. However,
average productivity was lower for remote workers than office workers. This suggests that remote
work has an adverse selection effect, with more productive workers preferring to be at the office.
Choudhury et al. (2019) study examiners for the US Patent and Trademark Office who were allowed
to work from anywhere. Productivity rose by 4.4%, with no decline in the quality of work. However,
they note that these jobs (as with call center workers) do not require significant collaboration and
coordination, and involve relatively simple tasks. Künn et al. (2021) analyze an occupation with
extremely high cognitive demands but no collaboration: professional chess players. The players showed
worse performance when competing online during Covid-19, compared to in-person tournaments. To
our knowledge, no research uses observational data to study WFH productivity for high-skilled work
in which coordination is important. Our paper fills this gap.
Other recent studies document shifts in working patterns in high skilled jobs, with findings that are
very consistent with our evidence. DeFilippis et al. (2020) use communication and email meta-data.
Their estimate that WFH employees work 0.8 more hours per day is in line with our findings. They also
find that employees attend more meetings, with more attendees (as do Teevan et al. (2020) and Yang
et al. (2021)). Flassak et al. (2021) present survey data from a large international company. Under
WFH, the firm used more standardization, planning, and employees spent more time in meetings.
Kwan (2021) presents evidence that proxies for employee interactions and the need for coordination
are negatively associated with a shift to WFH.
3
In the Western context it has often been reported that the burden of childcare and home-schooling disproportionately
affected women during the pandemic (Financial Times, 2021a). In India, extended families often live together, and middle
and upper class families often have domestic staff. Having extended family and staff at home can provide help with
childcare, but may place other demands on women at home whether or not children are present.

4
 A broader agenda in economics studies determinants of employee productivity. Significant work
has focused on incentive pay (e.g., Lazear, 2000; Hamilton et al., 2003; Shearer, 2004; Babcock et al.,
2015; Friebel et al., 2017; Aakvik et al., 2017; Dohmen and Falk, 2011). Other research looked at
effects of other human resource practices (Ichniowski et al., 1995; Ichniowski and Shaw, 2003; Bartel
et al., 2007), ways to engage employees in innovation (Gibbs et al., 2017), and the effects of supervisors
(Lazear et al., 2015) or peers (Bandiera et al., 2005; Arcidiacono et al., 2017; Song et al., 2018). Our
research contributes to this agenda by analyzing productivity in a high-skilled, coordination-intensive
profession, relating it to detailed measures of working patterns, and evaluating how it changes with
WFH.
Our evidence yields important insights into how to implement hybrid WFH / WFO work. Com-
munication, coordination and collaboration are more difficult in a virtual context. This challenge must
be addressed to implement significant WFH in occupations where such considerations are important,
especially for less experienced employees. While WFH will remain a feature of modern workplaces,
some aspects of in-person interactions cannot easily be replicated virtually, including the quality of col-
laboration and coaching, client engagement, and “productive accidents” that arise from spontaneously
meeting people (including those with whom there is not yet a working relationship).

2 Data and Empirical Strategy

The setting for this study is HCL Technologies, one of the world’s largest IT services companies, with
over 150,000 employees who work with clients across the globe. While it employs workers in many
countries, the group studied here are employed at its corporate campuses in India. The company
provides a wide range of technology consulting and outsourcing services for clients, including product
and process improvement and R&D to develop new products and services. Additional details about the
company are in Gibbs et al. (2017), which studies a field experiment by HCL on the use of incentives
to motivate employee ideas.
The workforce is highly skilled and educated. Virtually all have at least a bachelor’s degree, often
in a technology field such as computer engineering or electronics. Most work at the company’s large,
modern corporate campuses in several Indian cities. These campuses look and feel very similar to
what one sees at Microsoft, Apple or Amazon. All authors have visited the company headquarters,
and senior executives spent considerable time explaining practices and business conditions to us.
The company has a semi-annual planning process that culminates in goals for each organizational
level from the CEO down. Ultimately the process leads to goals for each team, based on business unit
objectives and expectations about customer requests during the next six months. The team supervisor
then sets goals for each employee.4 Analytics systems (described below) are used to track progress
against the plan throughout the organization. Each manager is provided with reports for his or her
unit. For example, the CEO reviews the corporate-wide report at least once per month.
4
If an employee is reassigned, or a project is completed and the team begins a new one, this process is refreshed for
affected employees before the next cycle begins.

5
 That employee goal-setting is embedded in a corporate-wide process and tied to each supervisor’s
own goals and evaluation is important. If managers lowered the goals of employees in their team, it
would make fulfilling team goals harder, and is therefore not in their interest. Company executives
confirmed that goals were not changed as a result of the pandemic or the switch to WFH. Indeed,
most of the company’s customers are also information technology firms, and that sector performed
well during the pandemic. This firm and its clients largely continued expected business activities.
These plans and goals form the basis of supervision. Informally, managers monitor employee work
time and performance across key applications and tasks, in order to better supervise and coach them.
This is supported by the analytics platforms described below. Formally, each employee is measured
by performance against monthly goal on a key metric, which we will refer to as “Output.” He or she
also receives quarterly supervisor feedback, and bi-annual subjective performance ratings.
A notable advantage of this study is that the performance measure is relatively rigorous and
objective, despite the fact that the employees are high-skilled professionals for which performance is
notoriously hard to measure.
The supervisor devises the key metric to reflect the most important aspect of the job, and this is
tracked with the analytics systems. For example, for a software engineer it might be the number of
code segments completed. Importantly, the measure does not merely reflect quantity. It is Output
conditional on adequate performance on other dimensions of the job, such as quality or client satisfac-
tion. Code segments will not be counted as complete until they meet company and client standards
for errors, speed, and functionality. Moreover, since the measure is output-based, it reflects various
employee inputs, including time, effort, skill development, client interactions, drawing upon colleagues
for advice, etc. Thus our measure is broader than might initially seem apparent, accounting for key
intangible aspects of each employee’s performance.
The employees in our sample do not receive incentive pay tied directly to this or any other mea-
sure. Compensation is comprised of salary, an annual merit bonus based on overall performance, and
occasional small rewards for activities such as suggestion of valuable new ideas. That the performance
measure is tied to the company’s business plan suggests employees are motivated to try to meet their
goals, and our evidence confirms this. However, motivation is broader than this, due to supervisor
monitoring, feedback, subjective evaluation, merit pay, and the potential to earn a promotion (as is
true at most companies).
Company financial statements reveal that total workforce size and revenue both rose by more than
5% in 2020 compared to 2019, and profit margins rose even more. Promotion rates were higher in
2020 than in 2019. Thus employees did not experience any decline in formal or informal incentives
during the sample period.

2.1 Main outcome variables

HCL uses two systems, Sapience Analytics and IDMS (“Internal Data Management System”), to track
employee activity and performance. Our three key outcome variables, Input, Output, and Produc-
tivity, derive from these. Managers use analytics reports based on these data to support decisions,

6
 Table 1: Summary statistics for outcome variables

Mean SD 1st Quartile 3rd Quartile N

WFO (pre March 2020)

Input 5.08 2.03 3.78 6.35 47387
Output 100.82 9.00 100.00 100.00 47387
Productivity 1.36 2.99 0.75 1.27 47387

WFH (post March 2020)

Input 7.04 2.75 5.38 8.90 22862
Output 100.30 8.80 100.00 100.00 22862
Productivity 1.11 2.41 0.52 0.88 22862

goal-setting, monitoring, and performance evaluation. Because of its corporate-wide significance, the
company has devoted substantial effort to making sure that the data are meaningful and reliable.
Such a “data-driven” approach is common in technology firms (Brynjolfsson and McAfee, 2012).
We obtained data for all employees from the R&D part of the firm who are managed via these
systems. Data cover April 2019 through August 2020, resulting in a panel dataset with 10,384 unique
employees observed over 17 months. The company moved abruptly to WFH in March 2020 as Covid-19
became serious in India.
Sapience Analytics software is installed on employee work devices. It records the time that an
employee is working by tracking applications or websites used, and whether the employee is active
(i.e., using the keyboard or mouse). If an employee procrastinates on a social media platform, and
this is not part of the job, that would not be recorded as work time. If a program from a pre-defined
list of relevant tools (programming, collaboration, document production, communication, etc.) is
active, that is recorded as work time. It includes meetings entered in the Outlook calendar, which is
used throughout the company, even if these are face-to-face. Sapience therefore records effective work
hours. If an employee sits at his or her desk for 8 hours a day, Sapience might only record 5 hours,
because breaks or surfing the web are not effective work time. Employees are aware that the software
is used in this way.
Based on these data, we calculate the outcome variable Input, equal to average working hours
per working day that month. That is, we take the total time worked that month and divide it by
the number of working days (taking into account weekends and local holidays). In section 2.3 below,
we describe other input measures that generate similar qualitative results; our findings are robust to
details of the definition of hours worked.
IDMS is a proprietary system used to track employee performance, including the primary per-
formance measure. That measure is normalized to make different jobs and roles comparable.5 For
example, key measures might be the number of code segments, code reviews, or reports delivered per
month (subject to quality and customer satisfaction standards). These are divided by the monthly
5
Our analysis does not rely on this normalization, as our fixed effects regressions compare the same employee before
and during WFH.

7
 Table 2: Summary statistics for employee variables

Mean SD 1st Quartile 3rd Quartile N

Age (in years) 31.91 5.95 27.10 36.03 7969

HighAge 0.50 0.50 0.00 1.00 7969
Tenure (in years) 4.21 3.90 1.11 5.11 7969
HighTenure 0.52 0.50 0.00 1.00 7969
Experience (in years) 8.10 5.22 4.04 11.10 7969
HighExperience 0.50 0.50 0.00 1.00 7969
Male 0.76 0.43 1.00 1.00 7969
NumChildren 0.52 0.73 0.00 1.00 8934
Children 0.39 0.49 0.00 1.00 8934
CommuteTime 0.65 0.33 0.38 0.85 4323
Rating 2.66 0.88 2.00 3.00 5354

goal, multiplied by 100, to scale as percentage achievement of the goal. As explained above, goals
were not changed due to WFH.
It is possible to complete more tasks than are assigned, so the outcome variable Output can take
values in R+
0 , but is typically between 0 and 100. The most common value is 100, meaning that
employees continue working longer hours until they meet their targets, but we also see employees
falling short of or exceeding their targets.
Finally, our outcome variable Productivity is calculated by dividing Output by Input. Differences
in productivity will often be a consequence of differences in Input needed to reach the goal, but they
can also come from employees falling short of or exceeding targets. Table 1 displays summary statistics
before and during WFH. The number of observations under the two regimes differs, because we have
more pre-WFH months.

2.2 Employee/HR variables

We obtained information on employee characteristics, collected as of March 20, 2020 (roughly the
date on which WFH was implemented). Summary statistics are in Table 2. For some employees some
variables are missing. One reason is that HR data are deleted if an employee leaves the company or
transfers to a branch outside India.
Age, company tenure and industry experience (collected at hiring and updated to the current
date) are all measured in years. For each we generate median splits: HighAge, HighTenure and
HighExperience. Mean age is quite young, which is not unusual in the IT sector. Mean tenure is low
at about 4 years, as is expected since employee turnover is high in the IT sector. As in tech companies
around the world, men are a significant majority.
The variable NumChildren is the number of children up to age 21 who are covered under the
company’s employee health insurance plan. The company believes that the vast majority of employees
who have dependent children insure them via the company, because of its relatively generous health
insurance coverage. However, some might instead be insured through a partner’s employer. Hence, a

8
zero means that there are either no children at home, or there are but they have not been declared.
The dummy Children equals one if and only if NumChildren is positive.
CommuteTime is an estimate of the time in hours needed to get from the home address to the
office (during WFO), one-way. The company calculated this based on home and office addresses,
using the Google Maps API to incorporate factors such as traffic and not merely distance. Thus,
it is an estimate of the usual time taken, assuming that the employee commutes by car.6 Address
data are often incomplete, so there are more missing values than for other variables. We discarded
extreme values (larger than 2 hours). According to the company these are cases where commute
time is unreliable; for example, an employee actually worked at a client office closer to home, not the
company office where his or her team is located.
Rating is the supervisor’s subjective evaluation of the employee on an integer scale of 1 to 5, where
1 is the best rating. We have the most recent rating from May/June 2020. The outcome measures
discussed above are predictive of performance ratings: mean input and mean productivity in months
prior to the rating significantly improve that rating (see Table A.1 in the Appendix). Figure B.1 in the
Appendix plots kernel density estimates of subjective ratings for different levels of Output. Ratings
generally rise with Output, but start to decrease once Output substantially exceeds the target. A
possible interpretation is that such an employee gave too much emphasis to meeting the goal, and the
supervisor gives a lower subjective rating to balance multitask incentives. Another is that the goal
was too easy to achieve. Overall, this is more evidence that the outcome measures introduced above
are meaningful and reflected in subjective performance evaluations.

2.3 Workplace Analytics Data

Microsoft Workplace Analytics [WPA] is a tool that many companies use to track and analyze various
aspects of their workforces. For example, it can be used to study collaboration or professional net-
working activity by using data on emails, calendar appointments, amount of time spent in meetings,
etc. WPA data have been used in several organizational studies (Brynjolfsson and McAfee, 2012;
Hoffmann et al., 2012; Levenson, 2018).
The company had been considering adoption of this tool. For the purposes of this study they
purchased 914 licenses to apply to a subset of employees in our full sample. Appendix Table A.2
compares those in the WPA sample to those not in the WPA sample. The WPA group are slightly
younger, have lower tenure and are less productive, but are overall quite similar.
Table 3 summarizes variables obtained from WPA. WPA data were collected at the weekly level (in
contrast to Sapience data, which are collected at the monthly level). We have 10 weeks of data before
WFH (starting January 1, 2020) and 24 weeks of data during WFH (ending September 6, 2020). The
switch to WFH happened in the week starting March 16.
Variables fall into several categories. Working Hours measures overall time worked by the employee.
It is best viewed as capturing time “at work” (at home or in the office), not every minute of which
is necessarily spent working; e.g., it will count small breaks in between emails as work time. It can
6
Some employees use public transport, but we have no data about the mode of travel.

9
 Table 3: Summary statistics for WPA variables

Mean SD 1st Quartile 3rd Quartile N

WFO (before March 15th 2020)

Working Hours 44.71 5.16 43 46.46 6755
After Hours 9.64 9.55 2.33 14.04 6755
Focus Hours 34.49 9.02 30 41.25 6755
Collaboration Hours 10.20 9.24 3.55 13.75 6755
Meetings Manager 3.97 4.35 0.5 5 6755
Meetings 1:1 0.18 1.37 0 0.5 6755
Coaching Meets 0.13 1.03 0 0 6755
MS Teams Calls 0.36 1.63 0 0 6755
Internal NW 18.91 14.21 10 24 6755
External NW 2.58 3.61 0 3 6755
NW EXT 0.98 1.04 0 1 6755
NW ORG 0.05 0.22 0 1 6755
Emails 23.61 23.68 9 30 6755

WFH (after March 15th 2020)

Working Hours 49.03 7.58 45.14 52.49 19220
After Hours 12.98 12.70 3.71 18.44 19220
Focus Hours 32.73 9.99 28 40 19220
Collaboration Hours 11.07 9.97 4.08 15 19220
Meetings Manager 5.48 6.57 1 7.33 19220
Meetings 1:1 0.11 1.07 0 0 19220
Coaching Meets 0.09 0.98 0 0 19220
MS Teams Calls 21.46 25.22 3 30 19220
Internal NW 23.44 19.89 11 30 19220
External NW 3.05 4.36 0 4 19220
NW EXT 0.91 0.89 0 1 19220
NW ORG 0.05 0.23 0 0 19220
Emails 25.26 29.89 8 30 19220

Note: “Working Hours” are weekly hours worked. “After Hours” are weekly hours worked outside regular work time.
“Focus Hours” are hours with two or more hour blocks not spent in meetings, on calls or writing e-mails. “Collaboration
Hours” are hours spent in meetings or MS Teams calls. “Meetings Manager” is the number of meetings involving the
employee’s manager. “Meetings 1:1” is the number of meetings between the employee and their manager. “Coaching
Meets” is the number of meetings the employee attended with the manager and all of the manager’s direct reports. “MS
Teams Calls” is the number of calls in which the employee participated. “Internal NW” is the number of people inside
the company with whom the employee had meaningful contact in the last 28 days. “External NW” is the same measure
for people outside the company. “NW ORG” is the number of distinct company organizational units (outside his or her
own) with which the employee had at least two meaningful interactions in the last four weeks. “NW EXT” is the same
measure for domains outside the company. Emails is the weekly number of emails sent by the employee.

detect longer work hours due to additional email traffic, meetings in the calendar or online, and other
activity involving Microsoft services. In contrast, Sapience Input captures time “effectively working,”
excluding breaks or non-work activities.
During WFO, employees spent on average 44 Working Hours per week (Table 3), which is close to
the contractual 40 hours per week. After Hours measures the number of weekly hours worked outside

10
regular office hours. Working Hours and After Hours are mechanically related as the latter is part of
the former.
A second group of variables (Focus Hours, Collaboration Hours, Meetings Manager, Meetings 1:1,
Coaching Meets and MS Teams Calls) relate to meetings. Focus Hours measures time blocks of 2 hours
or more that is uninterrupted by meetings, calls or emails. It is thus a measure of the amount of time
the employee can concentrate on tasks. Collaboration Hours is the total time spent in various forms
of meetings. The latter four variables measure time in meetings by structure and purpose. Meetings
Manager is the number of meetings the employee attends that involve their manager, and Meetings
1:1 are personal meetings between the employee and manager. Coaching Meets is the number of
meetings involving the employee, their manager and the manager’s direct reports. MS Teams Calls is
the number of calls using MS Teams (which hosts video meetings similar to Zoom or Skype).
Appendix Table A.3 shows (pre WFH) pairwise correlations between these meeting-related vari-
ables. As expected, all correlate negatively with Focus Hours, especially Collaboration Hours and
Meetings Manager. All pairwise correlations are statistically significant at the 1% level. The different
types of meetings are positively related among each other, but with smaller coefficients. These corre-
lations are positive both across employees – some job roles involve more meetings than others – and
across time – some periods involve more meetings of all types.
The third group of variables (Internal NW, External NW, NW EXT, NW ORG, Emails) relate to
networking with colleagues and clients more explicitly. The first two measure the number of individuals
(inside and outside of the company, respectively) with whom the employee had contact. The latter
two measure the number of business units (e.g., teams) involved in those contacts. These measure
the breadth of the employee’s communications and networking contacts. Appendix Table A.4 shows
the (pre WFH) pairwise correlation between these networking-related variables. All correlations are
positive and highly statistically significant, across employees as well as across time.

2.4 Empirical strategy

As a first step, in Section 3.1 we estimate the average WFH effect. Our main specification exploits
differences in outcomes for each employee, when working from home compared to working in the office,
controlling for employee and customer team fixed effects. The unit of observation is the employee-
month. Index the employee by i and the month by t = 1, 2, . . . , 17. For outcome variable yit , we
estimate by OLS:
X X
yit = αi + βWFHt + γj CustomerTeamjit + δs Monthst + εit , (1)
j s

where αi is the employee fixed effect, WFH is a dummy variable indicating months working from
home, and CustomerTeamjit is a dummy variable equal to one if and only if employee i in month t
was part of team j. Monthst is a month (not month-year) dummy variable, so that Month1t = 1 if and
only if t is January, Month2t = 1 if and only if t is February, etc. In addition, we report an alternative

11
specification controlling for a linear rather than seasonal time trend:
X
yit = αi + βWFHt + γj CustomerTeamjit + δt + εit . (2)
j

Once the average WFH effect is established, we analyze variation in this effect. Section 3.2 studies how
effects vary with employee characteristics. We interact the WFH dummy in the previous specifications
with additional explanatory variables X1i , X2i , . . .. Because the Xji variables are employee specific but
time invariant, we do not separately control for them, as that is already achieved by the employee
fixed effects.7
We exclude March 2020 (t = 12) from regressions because our main outcome variables are collected
at the monthly level, and working from home started in mid-March 2020.8 Thus, this month is neither
purely WFO nor WFH. Moreover, it is likely that WFH increased in the days prior to the official WFH
start, so the switch date was not clear-cut. An implication of excluding March 2020 is that teething
problems and short-term adaption effects are not reflected in our estimate.
Section 3.3 analyzes the effects of mechanisms such as time spent in communication, and focus
time. For these we rely on the WPA data described above. Here our empirical strategy is identical to
the one described in equation (2), except that we control for weekly instead of monthly time trends,
as these data are available weekly. Hence, in these regressions t = 1, ..., 34 represents weeks. For all
analyses we cluster standard errors at the employee level.

3 Results

3.1 Average WFH effect

Before proceeding to the regression analysis, Figure 1 plots the three main outcomes by month to get
an intuitive idea about the WFH effect. This will also help us understand which of the econometric
models (1) or (2) seems most appropriate for controlling for time trends.
According to Figure 1a, Input, employees provide about 5-5.5 hours of daily input; i.e., time in
which they are actively using their software or programming tools, or in meetings or communications.
There is relatively little variation in average input pre-WFH, with a slight upward trend. Hence, a
linear time trend as in model (2) might be more appropriate for this outcome measure. From the first
month of WFH, there is a large and persistent jump in input by more than 1.5 hours per day.9
For Output, Figure 1b, there is no visible monotonic or linear trend, so a seasonal time correction
might be more appropriate here. Moreover, average output appears to be slightly lower during WFH.
7
While age, tenure, and experience are not time invariant, our sample window is only 17 months, so there is no
meaningful variation during that window. Hence, to avoid collinearity issues, we use only employee fixed effects.
8
This month is nevertheless plotted in the graphs.
9
Note that a national statistical office, which does not have access to these effective work hours, would not pick up
these changes in input at the intensive margin when using nominal work hours (which did not change during WFH) to
compute productivity.

12
 102
8
Time worked per working day
7

101
Output by month
6

100
5

99
4

−11 −10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 −11 −10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5

Month Month

(a) Input: Time worked per working day (b) Output: Tasks completed relative to target
1.6
1.4
Productivity
1.2
1
.8

−11 −10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5
Month

(c) Productivity

Figure 1: Average outcomes by month without outliers (bottom and top 0.1%). The vertical line (month 0)
indicates the switch to working from home. The vertical bars represent 95% confidence intervals
based on clustered standard errors. Month -17 = April 2019; 0 = March 2020; 5 = August 2020.

Finally, for Productivity, Figure 1c, the graph is more volatile, which is not surprising for a ratio.
There is no clear linear time trend pre-WFH, but some variation from month to month, so a seasonal
correction might be more appropriate. Productivity drops visibly during WFH.
To quantify the WFH effect, and to control for employee and team time-invariant variables (via
employee and team fixed effects), we now turn to the regression analyses. Informally, the estimates
give us average differences in outcomes pre- and during WFH for the same employee, controlling for
team effects (since employees sometimes switch teams) and time trends.
Table 4 reports WFH effect estimates based on OLS regressions for all three outcome variables,
plus the natural logarithm of Productivity, in each case with linear and seasonal time trend corrections.
All estimates are in line with the visible effects in the raw data in Figure 1.10
10
Table A.5 in the appendix estimates the same regressions, but truncates the most extreme observations to account
for outliers. The qualitative results remain the same.

13
 Table 4: Average Working-From-Home effect

(1) (2) (3) (4) (5) (6) (7) (8)

Dependent variable Input Input Output Output Productivity Productivity LogProd LogProd

WFH 2.117*** 1.592*** -0.530*** -0.098 -0.256*** -0.138 -0.334*** -0.288***

(0.034) (0.038) (0.125) (0.155) (0.031) (0.074) (0.008) (0.009)
Linear month trend 0.040*** -0.035* -0.010 -0.004***
(0.003) (0.015) (0.007) (0.001)

Employee FE Yes Yes Yes Yes Yes Yes Yes Yes

Team FE Yes Yes Yes Yes Yes Yes Yes Yes

Month FE Yes No Yes No Yes No Yes No

R2 0.24 0.22 0.02 0.02 0.01 0.01 0.12 0.11
Observations 70249 70249 70249 70249 70249 70249 70239 70239
Clusters 10312 10312 10312 10312 10312 10312 10312 10312

Note: Input is the individual time in hours that the employee worked per working day in a month. Output is the
normalized output of the employee relative to the target in a month. Productivity is output divided time worked.
LogProd is the natural logarithm of Productivity. The unit of observation is the employee-month. Standard errors
are shown in brackets below the point estimates, and are clustered on employee level. ***Significant at the 0.1% level;
**significant at the 1% level; *significant at the 5% level.

According to the estimates, WFH increased the time worked per day by roughly 2.1 hours (based
on a seasonal time trend) or by 1.6 hours (with a linear time trend). Both estimates are economically
very meaningful, and statistically significant at all conventional levels.11 Since both Figure 1a and
the regression indicate a linear time trend, we prefer column (2). The estimate of the WFH effect in
column (1) is larger, because it does not take the pre-WFH time trend into account in the same way.
Columns (3) and (4) estimate that Output changed by -0.5 or -0.1 percentage points, depending
on time controls (recall that fulfilling all monthly tasks implies an output of 100%). Both point
estimates are slightly negative, but only the estimate with the seasonal time trend is significantly
different from zero at all conventional levels. Hence, we conclude that WFH had no or only a small
negative effect on Output. While the regression indicates a significantly negative linear time trend,
we prefer specification (3) since a linear trend does not reflect the raw data very well.
Columns (5) and (6) estimate that Productivity decreased by 0.26 or 0.14 output percentage points
per hour worked, depending on the time controls. Given an average WFO productivity of 1.36, these
effects correspond to a 19% or 10% drop in output per hour worked. Both are economically significant:
if employees worked a fixed 40 hours per week, this would imply a drop in output of 10.2 or 5.5 output
percentage points in a week. In other words, if employees had not increased time worked during
WFH, on average they would have completed only 90 to 95 of 100 assigned tasks. The WFH effect in
specification (6) is not significant at the 5% level, whereas the effect in specification (5) is statistically
11
For comparison, the contractual working day at HCL (pre-WFH) is 9 hours, which includes a 1 hour lunch break.

14
significant at all conventional levels. We prefer specification (5), since both the plot and the linear
time trend coefficient indicate that a linear trend is not as appropriate.12
Columns (7) and (8) explain the log of Productivity, which strongly increases the fit of the re-
gression. The WFH effect is negative and significantly different from zero at all significance levels,
independent of time controls.
In summary, this evidence indicates that employees worked longer but less productively, with
output remaining about the same or dropping slightly. Our interpretation of these patterns is that
employees were less productive during WFH, but still aimed to reach the same output or goals, and
hence worked longer until the same output was reached. In the next sections, additional results will
suggest that productivity decreased due to increased distractions and coordination costs.
A potential alternative explanation for the jump in work time during WFH might be that employees
“gamed” the Input numbers. This is unlikely for several reasons. First, Sapience time measurement
is sophisticated and designed to be resilient to manipulation attempts. Merely keeping the computer
on for longer or watching videos instead of working does not increase Input. Rather, it would require
having the relevant work software as the active window, and giving continuous mouse or keyboard
input. Second, gaming time measurement in Sapience would not translate into increases in WPA’s
time measurement.13 WPA time recording is from activity in MS Outlook, MS Teams etc., rather than
programming tools or similar, and is not dependent on mouse / keyboard activity. Yet the WFH effect
we see with this alternative time measurement is also positive; see section 3.3. Third, employees are
not paid by the hour. To impress superiors, time is better spent generating output. Fourth, Sapience
was in use long before the switch to WFH, so this potential concern cannot explain the WFH effect
well. Fifth, the additional WFH effects we find from WPA activity (section 3.3), such as more time
spent on conference calls and fewer focus hours, cannot be explained by “gaming.”
Another potential concern might be that the increase in Input is due to measurement error, as
spontaneous (unscheduled) face-to-face meetings during WFO might not be counted as working time,
while spontaneous online meetings during WFH are counted. This is not backed up by the data. We
are able to reproduce the findings in this section using other measures of working time, available for
a sub-sample of employees (Section 3.3). We also find adverse effects of WFH on direct measures
of networking and training; in particular the number of some types of meetings decreases during
WFH. Moreover, employees with children, and those working in roles where networking with people
outside the company is more important, are more adversely affected by the switch to WFH. This is
difficult to explain with unscheduled offline meetings, but is consistent with increased disruptions and
communication costs. Last, we observe when an employee starts and ends their working day. The
length of the workday measured in this way increases from 7.64 hours to 9.17 hours during the WFH
period. Taken together, these findings suggest that spontaneous offline meetings explain at most a
small part of our overall results.
12
We also estimated the productivity regression without time controls. The WFH estimate is -0.2 output percentage
points per hour worked with a t-statistic of -6.8; i.e., a highly significant effect consistent with the other specifications.
13
Unlike Sapience, employees were not aware of the use of WPA analytics. WPA licenses were purchased for the first
time and for this study specifically. Very few people at the company knew about it and had access to these data.

15
 If employees used their own devices to work from home, we would not track that activity (in either
period). However, HCL did not allow employees to do so, to insure integrity of confidential client
information. Any employees who worked at home in both periods had to use a company-owned laptop
computer and phone, both of which included the tracking software used to collect our variables, with
one small exception. During the transition to WFH, a small percentage of employees who did not
yet have a company-owned laptop computer were briefly allowed to use personal devices to perform
some work. This lasted only until the company could provide them with laptops. In such cases
Sapience would not track their work, so our Input variable has missing values. This affects a very
small percentage of our data.
Another possible explanation for the increase in time worked is that pandemic lockdown measures
closed restaurants, cinemas, etc., thereby reducing the value of leisure time. Under this explanation,
however, we would expect Output to increase and Productivity to remain approximately constant,
which is not what we observe. Appendix Figure B.2 shows that while we see a slight dip in working
hours after every stage of lockdown easing, the effect is small and, more importantly, only temporary.
We also do not find evidence that productivity or other outcomes co-vary with national or regional
indicators of the severity of Covid, such as deaths or case rates.
Finally, we are able to measure the number of days the employee worked relative to the number
of work days in a month. According to Table A.6 in the appendix, the number of sick days decreased
significantly. This suggests that absences or sickness were not a driver of the decrease in productivity.

3.2 Who copes better with WFH? Heterogeneous WFH effects

We now explore what drives the WFH effect in more depth, and which subgroups are most affected
by the shift to WFH. Table 5 displays estimates for all outcome variables, separately by whether
employees have children at home and by gender, using our preferred time control from the previous
section. The number of observations is slightly reduced since the additional explanatory variables
are missing for some employees. We converted the explanatory variables (except commute time) into
dummy variables for easier interpretation.
In India, all schools closed in March 2020 during the Covid-19 pandemic, so working from home
was presumably an even greater challenge for some parents, as children needed to be supervised and
perhaps taught. Hence, we investigate whether having children at home changed an employee’s WFH
effect. An important qualification is that under normal conditions, children would attend school and
adverse effects of WFH on productivity of parents might be lower.
Column (1) shows that employees who have at least one child at home (as measured by company
health insurance coverage) increased work time more during WFH than did their counterparts without
children. Possibly, this is due to the fact that employees with children get distracted more often
during WFH and compensate by working longer hours. Employees with children at home work almost
a third of an hour more per working day during WFH than those without children, who themselves
still work 1.4 hours more during WFH. These effects are highly significant. Column (3) reveals no
significant change in the WFH effect on output with children at home. However, column (5) shows

16
 Table 5: Working-From-Home: Children at home and gender differences

(1) (2) (3) (4) (5) (6)

Dependent variable Input Input Output Output Productivity Productivity

WFH 1.416*** 1.599*** -0.528*** -1.176*** -0.206*** -0.361***

(0.046) (0.086) (0.150) (0.267) (0.042) (0.075)
WFH × Children 0.307*** 0.091 0.061 0.673 -0.169** -0.027
(0.059) (0.128) (0.205) (0.431) (0.058) (0.111)
WFH × Male -0.211* 0.895** 0.149
(0.094) (0.300) (0.093)
WFH × Male × Children 0.252 -0.822 -0.157
(0.145) (0.495) (0.133)

Employee FE Yes Yes Yes Yes Yes Yes

Team FE Yes Yes Yes Yes Yes Yes
Month FE No No Yes Yes Yes Yes
Linear Month Trend Yes Yes No No No No
R2 0.23 0.25 0.02 0.02 0.01 0.01
Observations 64392 58644 64392 58644 64392 58644
Clusters 8865 7911 8865 7911 8865 7911

Note: Input is the individual time in hours that the employee worked per working day in a month. Output is the
normalized output of the employee relative to the target in a month. Productivity is output divided time worked. The
unit of observation is the employee-month. Standard errors are shown in brackets below the point estimates, and are
clustered on employee level. ***Significant at the 0.1% level; **significant at the 1% level; *significant at the 5% level.

that the increased working time implies a larger drop in productivity when there are children at home.
Consequently, the patterns seen for the average employee are exacerbated for employees with children.
The even columns in Table 5 investigate whether there was a gender difference in how outcomes
changed during WFH, conditional on whether there were children at home.
The WFH × Male interaction represents the difference in the WFH effect between male and female
employees without children. Male employees without children increased working time by about 0.2
hours less per day than did female employees without children, a significant effect at the 5% level.
They also suffered a significantly smaller decline in output.
The WFH × Children interaction represents the difference in the WFH effect between female
employees with and without children. Female employees with children did not significantly increase
working time during WFH compared to female employees without children, nor did their output or
productivity significantly differ.
Finally, the sum of WFH × Male and WFH × Male × Children is the difference in the WFH effect
between male and female employees with children. This difference is roughly zero for all outcome
measures, so there is no gender difference among employees with children at home.
Thus, female employees are more adversely affected by WFH, but this is not due to child care
responsibilities. The latter finding contrasts with much of the narrative in western countries, where

17
 Table 6: Working-From-Home: Age, experience, tenure, commute times

(1) (2) (3) (4) (5) (6)

Dependent variable Input Input Output Output Productivity Productivity

WFH 1.361*** 1.516*** -0.743*** -0.603 -0.245*** -0.286***

(0.058) (0.095) (0.194) (0.334) (0.048) (0.083)
WFH × HighTenure 0.036 0.519* 0.003
(0.067) (0.229) (0.090)
WFH × HighAge 0.057 0.097 -0.084
(0.092) (0.366) (0.073)
WFH × HighExperience 0.270** -0.138 -0.038
(0.094) (0.390) (0.084)
WFH × CommuteTime 0.107 0.316 -0.030
(0.122) (0.405) (0.093)

Employee FE Yes Yes Yes Yes Yes Yes

Team FE Yes Yes Yes Yes Yes Yes
Month FE No No Yes Yes Yes Yes
Linear Month Trend Yes Yes No No No No
R2 0.25 0.26 0.02 0.03 0.01 0.01
Observations 58644 31848 58644 31848 58644 31848
Clusters 7911 4295 7911 4295 7911 4295

Note: Input is the individual time in hours that the employee worked per working day in a month. Output is the
normalized output of the employee relative to the target in a month. Productivity is output divided time worked. The
unit of observation is the employee-month. Standard errors are shown in brackets below the point estimates, and are
clustered on employee level. ***Significant at the 0.1% level; **significant at the 1% level; *significant at the 5% level.

childcare responsibilities are given as a main reason why women are more adversely affected by WFH
(Financial Times, 2021a). We conjecture that this is due to greater expectations placed on women by
parents and parents-in-law in the Indian domestic setting.
Next, we investigate whether employees with more industry experience or company tenure were
affected differently by WFH. One reason this could be the case is that they have greater institutional
knowledge and social capital, and are less reliant on help from colleagues or find it relatively easier to
obtain during WFH. To investigate the effects of age, company tenure, and industry experience (at
the company or elsewhere), we generate dummy variables with a median split. Since these variables
are highly correlated, we estimate their effect on the WFH estimate jointly in Table 6.14
Column (1) shows that work experience has the largest and only significant impact on the WFH
effect for time worked. During WFH, experienced employees worked roughly a quarter hour more
per day compared to less experienced employees, holding age and company tenure constant.15 Our
14
Appendix Figure B.3 illustrates individual employee productivity during WFH as a function of productivity before
WFH. Two patterns emerge. First, most employees are more productive in WFO than WFH. Second, those who are
more productive in WFO tend to also be more productive WFH. However, there is considerable heterogeneity.
15
When estimating the regression with one interaction for age, tenure, and experience at a time (not displayed),
all interactions show significantly positive point estimates due to their positive correlation. That is, older employees

18
Figure 2: Heterogeneity of WFH Effect on Productivity Across Teams. The vertical bars represent 95%
confidence intervals.

interpretation is that more experienced employees have more managerial duties. The increased costs of
coordination (also see next section) during WFH are therefore borne by these experienced employees,
who have to put in more time to coordinate their team. Additionally, it is likely that the lion’s share
of managing the WFH transition falls on experienced employees with more responsibility.16
The change in Output during WFH is roughly 0.5 percentage points larger per hour for employees
with longer company tenure, holding age and experience constant. Thus, while low tenure employees’
Output falls, high tenure employees keep meeting their goals. The other characteristics do not show
a significant effect. It appears that employees who had worked at the company for longer were able
to adapt more effectively to the WFH-shock, and that this was more important than general industry
experience. This finding suggests that greater firm-specific human capital in the form of familiarity
with company procedures, or more fully-developed networks and working relationships with colleagues
and clients, were helpful during WFH. Alternatively, those with greater experience or tenure might
be in positions with more responsibility, and so responded more to the shift to WFH. For the last
outcome measure, productivity, there is no significant difference in the WFH effect among these
employee groups.
increased WFH work hours more compared to younger employees, but this is no longer true when holding tenure and
experience constant, see Table 6.
16
While more experienced employees might be more likely to have children at home, this experience effect is unrelated
to having children. When adding the interaction with Children to regression (1) in Table 6 (not displayed), the interaction
with HighExperience remains significantly positive.

19
 The even columns in Table 6 estimate the WFH effect by employee commute time (when working
from the office). The effect does not significantly differ by commute time for any outcome measure.
Hence, the finding that WFH increased hours worked is not merely due to more available time.
Rather, it supports the interpretation that productivity fell during WFH, and employees worked more
to compensate.
Figure 2 shows the mean and 95% confidence intervals of the estimated effect of WFH on produc-
tivity separately for different customer teams. Those are teams of employees who work for the same
customer, but not necessarily on the same project. In the figure teams are sorted by size (number of
team members) in increasing order from left to right. The effect is negative for almost all teams. This
illustrates that the overall negative effect on productivity (illustrated by the dashed line in Figure 2)
is not driven by a few customers or teams. Rather, WFH productivity declined broadly across the
group studied.
For 43 teams the negative effect is statistically significant. Five teams, by contrast, show a statis-
tically significant positive effect on productivity. With the exception of one team, those are all teams
in the bottom half of the size distribution. In smaller teams it might be easier to solve coordination
problems during WFH. Still, even among the smaller teams, the effect on productivity is negative for
the vast majority.

3.3 Mechanisms: What contributes to lower productivity?

To better understand the mechanisms behind the decrease in productivity, we study the sub-sample
of 914 employees for which WPA data were obtained (see Table A.2). Using these data we document
three patterns: an overall increase in working time; a shift away from performing work tasks and
towards spending time on meetings, calls, or answering e-mails; and reduced time networking with
others or meeting 1:1 with one’s manager.

Working Hours

Figure 3 illustrates the shift in working patterns after the start of WFH. In line with the evidence
in Section 3.1, the WPA data show that overall hours worked increased, including those after regular
office hours. Panels (c) and (d) show an interesting pattern. Employees spend more time in meetings
or calls and have less “focus time” (as described above, this measures the amount of work time that
was not interrupted by various types of meetings or communications). The increased time spent in
meetings, and its persistence after the initial WFH transition phase, suggest substantial and ongoing
coordination costs with WFH, which negatively impact time available to work in a productive manner.
The fact that employees have less time to work in a focused or uninterrupted manner could also explain
why they have been found to multitask more during meetings (Cao et al., 2021).
The technological shift is illustrated in Appendix Figure B.4. In particular there is a drastic
increase in the number of hours spent on virtual meetings using MS Teams. Interestingly, the number
of such meetings continues rising almost six months after the switch to WFH. The increase in number

20
 (a) Working Hours (b) After Hours

(c) Focus Hours (d) Collaboration Hours

Figure 3: Working patterns pre- and post WFH. Panel (a): mean weekly working hours. Panel (b): hours
worked outside regular working hours. Panel (c): number of hours with two or more hour blocks
not spent in meetings, on calls or writing e-mails. Panel (d): hours spent in meetings or in calls.
Week= 0 is 9th-15th March 2020.

of meetings is clearly much larger than the increase in total time spent in meetings. Therefore,
employees are interrupted more frequently by having to attend more meetings of shorter duration.
Table 7 shows regressions estimating the WFH effect on these outcomes. Both overall working
hours and working hours outside of regular office time increase during WFH. In fact, comparing the
size of coefficients in columns (1) and (3) we see that the increase in overall working hours takes place
almost entirely outside of normal working hours. The table also confirms the increase in time spent in
meetings and on calls, with a corresponding decrease in uninterrupted work time (focus hours). In all
cases, the WFH effect persists and remains highly statistically significant when we include a linear time
trend. The estimated effect size is smaller in these cases but remains substantial. After controlling
for time trends, employees work 2.7 hours more per week, out of which 1.8 are spent working outside
regular office hours. However, they also spend 1.4 hours less working in a focused or uninterrupted
manner. Since working hours are the sum of focus hours, collaboration hours, and “unfocused hours”
(i.e., hours working but getting interrupted), the table also shows that WFH replaced focus hours by

21
 Table 7: Shift in Working Patterns due to WFH

(1) (2) (3) (4) (5) (6) (7) (8)

Working Hours After Hours Focus Hours Collaboration Hours

WFH 4.431*** 2.743*** 4.212*** 1.822*** -2.665*** -1.417*** 1.251*** 0.704***

(0.161) (0.218) (0.288) (0.320) (0.245) (0.297) (0.122) (0.137)
Linear weekly trend 0.096*** 0.137*** -0.071*** 0.031***
(0.009) (0.014) (0.012) (0.005)

Employee FE Y Y Y Y Y Y Y Y
Team FE Y Y Y Y Y Y Y Y
R-squared 0.703 0.709 0.743 0.747 0.717 0.719 0.745 0.746
Observations 25,893 25,893 25,893 25,893 25,893 25,893 25,893 25,893
Clusters 914 914 914 914 914 914 914 914

Note: Working hours are weekly hours worked. After hours are weekly hours worked outside regular work time. Focus
Hours are hours with two or more hour blocks not spent in meetings, on calls or writing e-mails. Collaboration Hours are
hours spent in meetings or in calls. The unit of observation is the employee-week. Standard errors are shown in brackets
below the point estimates, and are clustered at the employee level. ***Significant at the 0.1% level; **significant at the
1% level; *significant at the 5% level.

a lot more unfocused work hours. These shifts in working patterns could explain why productivity
decreases under the WFH regime.17
We conjectured that some of the increase in working hours and decrease in productivity is due
to increased costs of communication and coordination within teams. If this is the case, then roles
which are characterized by more interaction and networking prior to WFH are more affected by the
switch to WFH. If we median split by how many internal networking contacts (variable Internal NW,
see Table 3) an employee had prior to WFH, working hours in specification (1) of Table 7 increase
by 6.01∗∗∗ hours for those with above median contacts and by 3.515∗∗∗ hours for those with below
median contacts. The difference is highly statistically significant (t-test p < 0.0001) suggesting that
coordination costs are indeed an important factor behind the decline in productivity during WFH.

Networking and Collaboration

We now focus on networking and collaboration patterns in more detail. Understanding changes in
networking and collaboration can tell us something about the value of additional time spent in meet-
ings. Shifts in networking patterns can also impact productivity in different ways, for example, by
affecting the exchange of knowledge and ideas.
Table 8 shows how networking and collaboration patterns change with WFH. Columns (1) and (2)
focus on the number of people inside and outside the company, respectively, with whom an employee
had a meaningful contact (defined as an email, meeting, call, or at least 3 instant messages) during
the last 28 days. There is a generally positive time trend, possibly reflecting the fact that networking
17
Additional analysis in Appendix A shows that for overall working hours the time trend is even stronger during WFH.
In this case the pre-WFH trend is only about 60% of the overall trend. For the other outcomes the trend is mitigated
after the initial shift in levels, in line with ceiling effects (Appendix Table A.7).

22
 Table 8: Shift in Networking Patterns and types of meetings due to WFH

(1) (2) (3) (4) (5) (6) (7) (8)

Internal External NW NW Meetings Meetings Coaching E-mails
NW NW ORG EXT Manager 1:1 Meets

WFH -7.621*** -0.532*** -0.009 -0.150*** 1.723*** -0.089 -0.060 4.282***

(0.369) (0.103) (0.005) (0.027) (0.205) (0.048) (0.033) (0.680)
Linear weekly trend 0.787*** 0.078*** 0.001** 0.008*** 0.016* 0.002 0.002 -0.020
(0.026) (0.006) (0.000) (0.001) (0.008) (0.002) (0.002) (0.029)

Employee FE Y Y Y Y Y Y Y Y
Team FE Y Y Y Y Y Y Y Y
R-squared 0.801 0.758 0.665 0.624 0.757 0.320 0.386 0.766
Observations 25,893 25,893 25,893 25,893 25,893 25,893 25,893 25,893
Clusters 914 914 914 914 914 914 914 914

Note: “Internal NW” is the number of people inside the company with whom the employee had meaningful contact in
the last 28 days. “External NW” the same measure for contacts outside the company. “NW ORG” is the number of
distinct organizational units within the company with whom the employee had at least two meaningful interactions in
the last four weeks. “NW EXT” is the same measure for domains outside the company. “Meetings Manager” is the
number of meetings involving the manager. “Meetings 1:1” is the number of meetings between the employee and their
manager. “Coaching Meets” is the number of meetings by the manager with all direct reports including the employee.
“E-mails” is the number of emails sent. The unit of observation is the employee-week. Standard errors are shown in
brackets below the point estimates, and are clustered at the employee level. ***Significant at the 0.1% level; **significant
at the 1% level; *significant at the 5% level.

is becoming more important at this company. However, there is a clear negative impact of WFH on
the number of individuals with whom employees share meaningful interactions.18
Columns (3) and (4) contain results for similar measures, focused on the number of organizational
units inside and outside the company at which an employee interacted with someone (column (4)).
Here we also see a decline in contacts caused by WFH, despite a general upward trend, though in the
case of internal organizational units the decline is not statistically significant.
Columns (5)-(7) focus on collaboration patterns. In line with our earlier analysis, the number of
meetings involving the manager increases. By contrast, the number of both 1:1 supervisor meetings and
coaching meetings decrease during WFH. Employees seem to receive less mentoring and coaching, even
though these effects are only statistically significant at the 10% level. Moreover, in WFH managers
are spending more time managing groups rather than individuals.
Last, column (8) shows that the number of emails sent increased substantially during WFH with
about 4 more emails being sent on average. This corresponds to a 17% increase over the baseline (see
Table 3).
Overall, these patterns highlight a detrimental impact of WFH on networking. Employees attend
more meetings, which tended to involve larger groups, but have shorter duration. This reduces their
focus time. Furthermore, they communicate with fewer individuals and organizational units inside
18
Appendix Table A.8 shows specifications where we allow the time trend to interact with WFH. By and large, there
are no significant differences in trend pre- and post WFH. For internal networking the positive time trend is somewhat
bigger during WFH suggesting some potential catch up, but for all other outcomes there is no significant (or even a
negative) difference before and after the introduction of WFH.

23
 Table 9: Productivity in the WPA sample

(1) (2) (3)

Sapience WPA

WFH -0.149** -0.056*** -0.049**

(0.072) (0.008) (0.020)
Mean pre WFH 1.080 0.599 0.599
(1.702) (0.531) (0.531)

Employee FE Y Y Y
Team FE Y Y Y
Linear Trend N N Y
R-squared 0.356 0.783 0.784
Observations 4,359 4,359 4,359
Clusters 888 888 888

Note: Productivity is as in previous sections, Output divided by Sapience monthly work hours. Productivity WPA
divides Output by the WPA Input measure “working hours” (aggregated to monthly level). The unit of observation is
the employee-month. For 888 employees we have all three sources of information: Sapience, IDMS and WPA. Standard
errors are shown in brackets below the point estimates and are clustered at the employee level. ***Significant at the
0.1% level; **significant at the 1% level; *significant at the 5% level.

and outside the company. They have fewer 1:1 meetings with superiors, and receive less coaching.
The next sections that these observation help explain why WFH lowered productivity.

Productivity

Last, we study productivity in this sub-sample. We first ask whether the drop in productivity from
our main sample can also be found in this smaller WPA sample of employees.
As we have additional variables for this smaller sample, we consider two measures of productivity:
(i) the Sapience based measure used above, and (ii) where we replace the Sapience variable Input by
WPA working hours to compute productivity. Because we only have two months of pre-WFH data, we
do not include time trends for the monthly Sapience data here. Using the Sapience time measure, we
estimate an approximately 15% drop in productivity after the introduction of WFH. With the WPA
time measure, we find an approximately 10% drop in productivity, or 8% when controlling for a linear
time trend. Both effects are statistically significant (p < 0.040, see Table 9). The drop in productivity
from the full sample can thus be reproduced in this smaller sample, using different measures of hours
worked. Appendix Figure B.5 plots the change in productivity after the introduction of WFH in this
sample.
We next ask how these changes in work patterns are linked to productivity. This will help us
understand whether the documented changes can explain the decrease in productivity. We would
further like to know which variables are the most important predictors of productivity. For the
remainder of the section we use the Sapience measure of productivity from Section 3.1.
To address these questions, we first estimate adaptive Lasso regressions (Zou, 2006) in which
the dependent variable is productivity and prediction variables are dummies that identify weeks in

24
 Table 10: Variables selected by Lasso and Elasticity of Productivity with respect to selected variable.

pre WFH during WFH

Lasso Elasticity Lasso Elasticity

Hours
Working Hours X 0.014∗∗∗
After Hours

Meetings
Focus Hours X 0.033∗∗∗ X 0.072∗∗∗
Collaboration Hours
Meetings Manager
Meetings 1:1 X 0.002∗∗∗
Coaching 1:1 X 0.003
MS Teams Calls

Networking
Internal NW X 0.040∗∗∗
External NW X −0.000 X 0.029∗∗∗
NW EXT X 0.010∗ X 0.011∗∗
NW ORG X 0.000 X −0.001

Emails
Emails X 0.053∗∗∗

Note: Adaptive Lasso linear regression results (X indicates a variable was selected) and mean elasticity of productivity
with respect to an increase of 1 percentage point in the variables selected by Lasso. The left two columns show the
results restricted to the period before WFH and the right columns the results restricted to the period during WFH. For
all variables in the table Lasso regressions include a dummy identifying weeks in which the variable is above average for
a given employee.

which a variable is above average for a given employee. Lasso regressions select a set of variables
that best explain variation in productivity by minimizing an estimate of the out-of-sample prediction
error.19 We use dummies identifying the weeks where a variable is above average for a given employee
to focus on variation in productivity within employees.20 We conduct this regression separately for
WFO and WFH periods in order to see whether productivity determinants changed between the two
environments. Table 10 presents these results. An X in the table indicates variables that the Lasso
regression includes in the prediction model.
The variables selected by the Lasso include working hours, focus hours, and most networking vari-
ables. Working After Hours and attending many meetings does not seem to contribute substantially to
productivity, nor does spending time on MS Teams calls. The set of selected variables is quite consis-
tent before and during WFH, with focus hours and the networking measures being crucial indicators
of productivity. Interestingly, overall working hours is selected before WFH but not afterwards.
To assess the economic significance of these associations, we compute the elasticity of productivity
with respect to a 1 percentage point increase in the variables selected by Lasso. Table 10 shows
19
Lasso models have a free parameter λ which is the weight on the penalty term. Adaptive Lasso performs multiple
Lassos, where in each the λ is selected that minimizes an estimate of the out-of-sample prediction error. After each
Lasso, variables with zero coefficients are removed and remaining variables are given a penalty weight designed to drive
small coefficients to zero. Zou (2006) has shown that adaptive Lasso enjoys oracle properties: they perform as well as if
the true underlying model were known ex ante.
20
An alternative would be to force Lasso to select employee fixed effects. That is not possible here as in the merged
dataset containing both productivity and WPA variables we do not have enough pre-WFH observations.

25
the mean elasticities. Before WFH, the most important variables associated with productivity are
working hours, focus hours and NW EXT. A 1 percentage point increase in overall working hours
is associated with a 0.014 percentage point average increase in productivity. A 1 percentage point
increase in focus hours is associated with a 0.033 percentage point average increase in productivity,
and a 1 percentage point increase in network contacts outside the company is associated with a 0.01
percentage point average increase in productivity. These elasticities show that these variables are
important correlates of productivity. During WFH, the most important variables are focus hours,
and internal and external networking. As before, both internal and external networking are positively
associated with productivity. Focus Hours are now more than twice as important in terms of their
average marginal effect on productivity, with a 1 percentage point increase in focus hours associated
with a 0.072 percentage point average increase in productivity. There seems to be a broadly stable
relationship between working patterns and productivity. The increased importance of focus hours
during WFH might be explained by the fact that employees have less of it during WFH.
In summary, in this section we showed that WFH induced a significant shift in working patterns.
Employees worked more, including outside regular office hours, but had less uninterrupted time to
focus on task completion as they spent more time in meetings. They networked less and spent less time
being evaluated, trained and coached. We further showed evidence that these reductions, especially
in focus hours and networking, were detrimental to productivity.

4 Conclusion

In this paper we have presented the most detailed analysis of WFH productivity changes for knowl-
edge workers available to date. The paper makes a number of significant contributions. We study an
occupation that is expected to be amenable to WFH, but involves significant cognitive, collaborative
and innovation tasks. The data provide an unusually high quality measure of employee productivity
for knowledge workers. The breadth of the data allow for the first thorough analysis of determinants
of WFH productivity. We provide evidence on how WFH productivity varies with employee char-
acteristics, presence of children at home, and WFO commute time. We also use detailed data on
how employees spend their work time to study the effects of job characteristics on WFH productivity.
These latter results are important, since they provide insights into how the effectiveness of WFH may
vary across different types of jobs, and thus key issues for firms to consider in deploying WFH.
In our sample, employees were able to maintain similar or just slightly lower levels of output
during WFH. In order to do so, they worked longer hours.21 For this reason, productivity, measured
by output per hour worked, fell by 8-19%.
Our main explanation for the decline in productivity is that some aspects of work are more difficult
to perform in a virtual environment. We provide clear evidence that this is the case. Employees spent
more total time attending more meetings of shorter duration. This reduced their focus time. Those
21
It would be interesting to see if this change is sustainable, especially in light of evidence of the adverse effect of long
work hours on employee well-being, and mental and physical health (Sparks et al., 1997; Sokejima and Kagamimori,
1998; Sparks et al., 2001).

26
meetings tended to involve larger groups. Less time was spent in direct interactions with the supervisor
or close colleagues. Employees also narrowed their spheres of communication, interacting with fewer
people and business units, both inside and outside the firm. Collectively these indicate that costs of
communication, collaboration and coordination are higher when done virtually.
Alternative explanations could be related to the stresses of the pandemic. However, the data show
an immediate and persistent jump in work patterns and productivity at the shift to WFH, which is
inconsistent with the gradual evolution of the pandemic. We found no evidence that employee work
behavior varied with changes in lockdown restrictions in India. Employees actually had a decline in
sick days. Finally, company economic performance was quite good during this period of time, and
promotion rates rose, so career concerns were not an issue.
Another possible explanation is the presence of children at home, exacerbated by the closing
of schools. Indeed, WFH productivity was lower for employees who had children at home, so this
is a partial explanation. However, those without children at home also suffered a large decline in
productivity, with similar patterns for reduced focus time, increased time spent in large meetings, and
decreased 1:1 communications and meetings.
It is likely that WFH also resulted in a decline in intangibles that are valuable to the employee and
the company. Working relationships, professional networks and corporate culture may have suffered.
More subtly, when people work in the same location, they experience unplanned interactions. That can
lead to new working relationships, and “productive accidents” that spur innovation. It is not easy to
generate similar unplanned interactions on teleconferences. Finally, employees had fewer opportunities
for coaching, and meeting directly with supervisors. This undoubtedly slowed their development of
human capital.
HCL executives were not surprised by the findings in this paper, which accord with their percep-
tions of the experience with WFH. As of mid 2021, they estimated that perhaps half of their employees
had moved from their original locations, primarily to live with extended family (there were still signif-
icant pandemic-related restrictions). According to one senior executive, “bringing them back to our
base locations will not be easy.” The company expects to make significant use of WFH in the future,
with perhaps 30-40% of employees in WFO on any given day. No doubt productivity will improve
as the firm refines implementation of WFH and moves to a blended model with WFO. Moreover,
employees will enjoy greater flexibility and lower commute times, which compensates to some extent
for lower productivity.
The findings in this paper will be helpful well beyond this firm. We have presented evidence on
some of the challenges of implementing WFH. In particular, WFH may be more difficult for employees
who are less experienced, have lower tenure, and for jobs that involve significant communication,
collaboration and coordination. Firms will have to develop tools, training and policies to give greater
emphasis to interpersonal interactions during WFO, improve effectiveness of virtual communication,
and train supervisors and employees to schedule work time at home more efficiently.

27
References
Aakvik, A., F. Hansen, and G. Torsvik (2017): “Productivity dynamics, performance feedback and group incentives
in a sales organization,” Labour Economics, 46, 110 – 117.

Adams-Prassl, A., T. Boneva, M. Golin, and C. Rauh (2020): “Work tasks that can be done from home: Evidence
on variation within & across occupations and industries,” CEPR Discussion Paper No. DP14901, CEPR Discussion
Paper No. DP14901.

Arcidiacono, P., J. Kinsler, and J. Price (2017): “Productivity Spillovers in Team Production: Evidence from
Professional Basketball,” Journal of Labor Economics, 35, 191–225.

Babcock, P., K. Bedard, G. Charness, J. Hartman, and H. Royer (2015): “Letting Down the Team? Social
Effects of Team Incentives,” Journal of the European Economic Association, 13, 841–870.

Bandiera, O., I. Barankay, and I. Rasul (2005): “Social Preferences and the Response to Incentives: Evidence from
Personnel Data,” The Quarterly Journal of Economics, 120, 917–962.

Barrero, J., N. Bloom, and S. Davis (2021): “Why Working From Home Will Stick,” Working Paper, Department
of Economics, Stanford University.

Barrero, J. M., N. Bloom, and S. J. Davis (2020): “60 million fewer commuting hours per day: How Americans use
time saved by working from home,” University of Chicago, Becker Friedman Institute for Economics Working Paper.

Bartel, A., C. Ichniowski, and K. Shaw (2007): “How does information technology affect productivity? Plant-level
comparisons of product innovation, process improvement, and worker skills,” The Quarterly Journal of Economics,
122, 1721–1758.

Bellmann, L. and O. Hübler (2020): “Job Satisfaction and Work-Life Balance: Differences between Homework and
Work at the Workplace of the Company,” IZA Discussion Paper.

Bick, A., A. Blandin, and K. Mertens (2020): “Work from home after the COVID-19 Outbreak,” CEPR Discussion
Paper No. DP15000.

Bloom, N., J. Liang, J. Roberts, and Z. J. Ying (2014): “ Does Working from Home Work? Evidence from a
Chinese Experiment *,” The Quarterly Journal of Economics, 130, 165–218.

——— (2015): “Does working from home work? Evidence from a Chinese experiment,” The Quarterly Journal of
Economics, 130, 165–218.

Brynjolfsson, E., J. J. Horton, A. Ozimek, D. Rock, G. Sharma, and H.-Y. TuYe (2020): “COVID-19 and
remote work: An early look at US data,” Tech. rep.

Brynjolfsson, E. and A. McAfee (2012): “Big Data’s Management Revolution.” Harvard Business Review, 9.

Cao, H., C.-J. Lee, S. Igqbal, M. Czerwinski, P. Wong, S. Rintel, B. Hecht, J. Teevan, and L. Yang (2021):
“Large Scale Analysis of Multitasking Behavior During Remote Meetings,” Proceedings of the 2021 CHI Conference
on Human Factors in Computing Systems.

Choudhury, P., C. Foroughi, and B. Laron (2019): “Work-from-anywhere: The productivity effects of geographic
flexibility,” Strategic Management Journal, 42, 655–683.

DeFilippis, E., S. M. Impink, M. Singell, J. Polzer, and R. Sadun (2020): “Collaborating during coronavirus:
The impact of Covid-19 on the nature of work,” NBER Working Paper 27612.

28
Dingel, J. I. and B. Neiman (2020): “How many jobs can be done at home?” Journal of Public Economics, 189,
104235.

Dohmen, T. and A. Falk (2011): “Performance Pay and Multidimensional Sorting: Productivity, Preferences, and
Gender,” American Economic Review, 101, 556–90.

Emanuel, N. and E. Harrington (2021): ““Working” Remotely?” Working Paper, Department of Economics, Har-
vard University.

Etheridge, B., Y. Wang, and L. Tang (2020): “Worker Productivity during Lockdown and Working from Home:
Evidence from Self-Reports,” Covid Economics, 52, 118–151.

Financial Times (2021a): ““I am close to quitting my career”: Mothers step back at work to cope with pandemic
parenting,” https://www.ft.com/content/d5d01f06-9f7c-4cdc-9fee-225e15b5750b.

——— (2021b): “Where’s the Spark? How Lockdown Caused a Creativity Crisis.” https://www.ft.com/content/
27364b27-6c0c-4dec-b109-17c054b49465.

Flassak, K., J. Haag, C. Hofmann, C. Lechner, N. Schwaiger, and R. Zacherl (2021): “Working from home
and management controls,” Working paper, LMU Munich School of Management.

Friebel, G., M. Heinz, M. Krueger, and N. Zubanov (2017): “Team Incentives and Performance: Evidence from
a Retail Chain,” American Economic Review, 107, 2168–2203.

Gibbs, M., S. Neckermann, and C. Siemroth (2017): “A field experiment in motivating employee ideas,” Review of
Economics and Statistics, 99, 577–590.

Gottlieb, C., J. Grobovšek, M. Poschke, and F. Saltiel (2021): “Working from home in developing countries,”
European Economic Review, 103679.

Hamilton, B. H., J. A. Nickerson, and H. Owan (2003): “Team Incentives and Worker Heterogeneity: An Empirical
Analysis of the Impact of Teams on Productivity and Participation,” Journal of Political Economy, 111, 465–497.

Hensvik, L., T. Le Barbanchon, and R. Rathelot (2020): “Which jobs are done from home? Evidence from the
American Time Use Survey,” CEPR Discussion Paper No. DP14611, CEPR Discussion Paper No. DP14611.

Hoffmann, C., E. Lesser, and T. Ringo (2012): “Calculating Success.” Harvard Business Press.

Ichniowski, C. and K. Shaw (2003): “Beyond incentive pay: Insiders’ estimates of the value of complementary human
resource management practices,” Journal of Economic Perspectives, 17, 155–180.

Ichniowski, C., K. Shaw, and G. Prennushi (1995): “The effects of human resource management practices on
productivity,” Tech. rep.

Künn, S., C. Seel, and D. Zegners (2021): “Cognitive Performance in Remote Work - Evidence from Professional
Chess,” Economic Journal.

Kwan, A. (2021): “Measuring Remote Knowledge Work Using Big Data,” Working Paper, Hong Kong University
Business School.

Lazear, E. P. (2000): “Performance Pay and Productivity,” The American Economic Review, 90, 1346–1361.

Lazear, E. P., K. L. Shaw, and C. T. Stanton (2015): “The value of bosses,” Journal of Labor Economics, 33,
823–861.

Levenson, A. (2018): “Using workforce analytics to improve strategy execution.” Human Resource Management, 57,
685–700.

29
Shearer, B. (2004): “Piece Rates, Fixed Wages and Incentives: Evidence from a Field Experiment,” The Review of
Economic Studies, 71, 513–534.

Sokejima, S. and S. Kagamimori (1998): “Working hours as a risk factor for acute myocardiol infarction in Japan:
Case-control study,” British Medical Journal, 317, 775–780.

Song, H., A. L. Tucker, K. L. Murrell, and D. R. Vinson (2018): “Closing the Productivity Gap: Improving
Worker Productivity Through Public Relative Performance Feedback and Validation of Best Practices,” Management
Science, 64, 2628–2649.

Sparks, K., C. Cooper, Y. Fried, and A. Shirom (1997): “The effects of hours of work on health: A meta-analytic
review,” Journal of Occupational and Organizational Psychology, 70, 391–408.

Sparks, K., B. Fragher, and C. Cooper (2001): “Well-being and occupational health in the 21st century workplace,”
Journal of Occupational and Organizational Psychology, 74, 489–509.

Teevan, J., B. Hecht, and S. Jaffe (2020): “The new future of work,” Microsoft internal report.

WSJ (2020): “Companies Start to Think Remote Work Isn’t So Great After All,” https://www.wsj.com/articles/
companies-start-to-think-remote-work-isnt-so-great-after-all-11595603397.

Yang, L., D. Holtz, S. Jaffe, S. Suri, S. Sinha, J. Weston, C. Joyce, N. Shah, K. Sherman, B. Hecht,
and J. Teevan (2021): “The effects of remote work on collaboration among information workers,” Nature: Human
Behaviour.

Zimpelmann, C., H.-M. von Gaudecker, R. Holler, L. Janys, and B. Siflinger (2021): “Hours and income
dynamics during the Covid-19 pandemic: The case of the Netherlands,” Labour Economics, 73, 102055.

Zou, H. (2006): “The adaptive lasso and its oracle properties.” Journal of the American Statistical Association, 101,
1418–1429.

30
Online Appendix “Productivity, Collaboration & Networking During
Work From Home: Evidence from IT Professionals”

M. Gibbs, F. Mengel, C. Siemroth

For Online Publication

Contents

A Additional Tables 2

B Additional Figures 7

1
A Additional Tables

Table A.1 presents OLS and ordered Logit regressions, explaining the performance rating with that
employee’s mean time worked, output, and productivity over the most recent 5 or 10 months. All
regressions show that both time worked and productivity significantly improve an employee’s perfor-
mance rating (a lower score is better). The coefficient on mean output goes in that direction, but is
not statistically significant, because the effect is not monotonic (see also Figure B.1 below). The time
horizon of 5 or 10 months does not noticeably change coefficients. Hence, Sapience input and outcome
measures are meaningful in explaining performance ratings, which the company uses for promotion
decisions, among other things.

Table A.1: Do the Sapience outcome measures predict peformance evaluations?

(1) OLS (2) Ordered Logit (3) OLS (4) Ordered Logit

Dependent variable Rating Rating Rating Rating

main
MeanInput5 -0.097*** -0.219***
(0.007) (0.015)
MeanOutput5 -0.003 -0.006
(0.002) (0.004)
MeanProductivity5 -0.020** -0.041**
(0.006) (0.015)
MeanInput10 -0.093*** -0.205***
(0.007) (0.016)
MeanOutput10 -0.002 -0.004
(0.002) (0.004)
MeanProductivity10 -0.023*** -0.047**
(0.007) (0.015)
Constant 3.498*** 3.365***
(0.173) (0.183)

R2 0.06 0.04
Observations 4220 4220 4930 4930

Note: MeanInputX is the average of Input (hours worked) over the most recent X months prior to the performance
rating. Similarly, MeanOutputX and MeanProductivityX are the averages of Output and Productivity, respectively,
over the most recent X months prior to the performance rating. Rating takes integer values 1 to 5, with 1 being the
best. Each observation is one employee. Heteroskedasticity-robust standard errors are shown in brackets below the point
estimates. ***Significant at the 0.1% level; **significant at the 1% level; *significant at the 5% level.

Table A.2 provides summary statistics for employees in the WPA sub-sample, compared with other
employees for whom we do not have WPA data. Employees covered by WPA are somewhat younger
and more junior in the company. They are also somewhat less productive. However, these differences
are generally small.
Tables A.3 and A.4 show pairwise raw correlations among meeting variables and networking vari-
ables, respectively. As expected focus hours is negatively related to all meeting variables, while the

2
 Table A.2: Summary Characteristics of Non WPA and WPA sample.

Non WPA WPA p-value

Age (yrs) 31.98 31.29 0.001

Male 0.76 0.78 0.191
Children 0.52 0.50 0.366
Tenure (yrs) 4.49 4.17 0.023
Productivity 1.66 1.49 0.034

N 9398 914

meeting variables have positive correlation among themselves. Networking variables are positively
related among each other, as expected.

Table A.3: Pairwise Correlation of Meeting-Related Variables.

Focus Hours Collab. Hours Meetings Mgr Meetings 1:1 Coaching Meets Calls

Focus Hours 1
Collab. Hours −0.4710∗∗∗ 1
Meetings Mgr −0.4223∗∗∗ 0.4365∗∗∗ 1
Meetings 1:1 −0.0857∗∗∗ 0.0327∗∗∗ 0.1871∗∗∗ 1
Coaching Meets −0.1115∗∗∗ 0.0814∗∗∗ 0.0261∗∗ 0.0059 1
MS Teams Calls −0.0883∗∗∗ 0.1633∗∗∗ 0.0180∗∗∗ −0.0121 0.0317∗∗∗ 1

Note: ***Significant at the 0.1% level; **significant at the 1% level; *significant at the 5% level.

Table A.4: Pairwise Correlation of Networking-Related Variables.

Internal NW External NW NW ORG NW EXT Emails

Internal NW 1
External NW 0.5553∗∗∗ 1
NW ORG 0.3060∗∗∗ 0.1770∗∗∗ 1
NW EXT 0.4314∗∗∗ 0.7186∗∗∗ 0.1183∗∗∗ 1
Emails 0.6385∗∗∗ 0.5342∗∗∗ 0.2316∗∗∗ 0.3854 1

Note: ***Significant at the 0.1% level; **significant at the 1% level; *significant at the 5% level.

Table A.5 estimates the same average WFH effect as Table 4, except with the most extreme 1%
of observations truncated to assess the impact of outliers. The comparison between the two tables
shows that qualitative results are all the same, except that the WFH estimate in specification 4 turns
statistically significant. However, as explained in the text, this is not our preferred specification, as
the linear time trend does not represent the raw data well, and the effect in specification 3 remains
not significantly different from zero.
Table A.6 estimates the average effect of WFH on the share of non-sick days. Based on both time
controls, the effect is significantly positive, i.e., there were fewer sick days during WFH.
Table A.7 analyzes the same outcomes as Table 7 in the main text, but now also includes an
interaction between the time trend and the WFH dummy. The table shows that in the case of

3
 Table A.5: Average Working-From-Home effect (top and bottom 1% of outcomes truncated)

(1) (2) (3) (4) (5) (6) (7) (8)

Dependent variable Input Input Output Output Productivity Productivity LogProd LogProd

WFH 2.045*** 1.547*** -0.384*** -0.388*** -0.300*** -0.226*** -0.333*** -0.307***

(0.031) (0.035) (0.084) (0.097) (0.013) (0.015) (0.007) (0.008)
Linear month trend 0.038*** 0.000 -0.006*** -0.002*
(0.003) (0.010) (0.002) (0.001)

Employee FE Yes Yes Yes Yes Yes Yes Yes Yes

Team FE Yes Yes Yes Yes Yes Yes Yes Yes
Month FE Yes No Yes No Yes No Yes No
R2 0.24 0.23 0.01 0.01 0.06 0.06 0.16 0.14
Observations 68845 68845 68911 68911 68846 68846 68836 68836
Clusters 10217 10217 10258 10258 10217 10217 10217 10217

Note: Input is the individual time in hours that the employee worked per working day in a month. Output is the
normalized output of the employee relative to the target in a month. Productivity is output divided time worked.
LogProd is the natural logarithm of Productivity. The unit of observation is the employee-month. Standard errors
are shown in brackets below the point estimates, and are clustered on employee level. The top 1% and bottom 1% of
outcomes are discarded before running the regression to deal with potential outliers. ***Significant at the 0.1% level;
**significant at the 1% level; *significant at the 5% level.

working hours the increasing trend is even steeper after WFH. In all other cases the trend is mitigated
during WFH.
Table A.8 analyzes the same outcomes as Table 8 in the main text, but now also includes an
interaction between the time trend and the WFH dummy. The table shows that in the case of
Internal NW the increasing trend is even steeper after WFH. In all other cases the trend is mitigated
during WFH or not statistically different from the WFO period.

4
 Table A.6: WFH effect on (non-)sick days

(1) (2)

Dependent variable NonSickDayShare NonSickDayShare

WFH 0.069*** 0.007*

(0.003) (0.004)
Linear month trend 0.005***
(0.000)

Employee FE Yes Yes

Team FE Yes Yes
Month FE Yes No
R2 0.08 0.05
Observations 70168 70168
Clusters 10247 10247

Note: NonSickDayShare is the share of days in a month where the employee worked at least two hours, based on
Sapience measurement, relative to the number of work days in that month. The unit of observation is the employee-
month. Standard errors are shown in brackets below the point estimates, and are clustered on employee level. The top
1% and bottom 1% of outcomes are discarded before running the regression to deal with potential outliers. ***Significant
at the 0.1% level; **significant at the 1% level; *significant at the 5% level.

Table A.7: Shift in Working Patterns due to WFH with Change in Trend

(1) (2) (3) (4)

Working Hours After Hours Focus Hours Collaboration Hours

WFH 2.919*** 1.219*** -0.819** 0.560***

(0.194) (0.305) (0.287) (0.130)
Linear weekly trend 0.058*** 0.270*** -0.203*** 0.063***
(0.012) (0.029) (0.028) (0.011)
WFH× Time 0.041* -0.142*** 0.140*** -0.033**
(0.016) (0.033) (0.031) (0.013)

Employee FE Y Y Y Y
Team FE Y Y Y Y
R-squared 0.709 0.748 0.719 0.747
Observations 25,893 25,893 25,893 25,893
Clusters 914 914 914 914

Note: Working hours are weekly hours worked. After hours are weekly hours worked after regular work time. Focus
Hours are hours with two or more hour blocks not spent in meetings, on calls or writing e-mails. Collaboration Hours are
hours spent in meetings or in calls. The unit of observation is the employee-week. Standard errors are shown in brackets
below the point estimates, and are clustered at the employee level. ***Significant at the 0.1% level; **significant at the
1% level; *significant at the 5% level.

5
 Table A.8: Networking with Change in Trend

(1) (2) (3) (4) (5) (6) (7) (8)

Internal External NW NW Meetings Meetings Coaching E-mails
NW NW ORG EXT Manager 1:1 Meets

WFH -6.537*** -0.585*** -0.011* -0.164*** 1.165*** -0.094 -0.082 1.285*

(0.347) (0.104) (0.005) (0.034) (0.201) (0.059) (0.042) (0.580)
Linear weekly trend 0.549*** 0.089*** 0.001* 0.011* 0.139*** 0.003 0.007 0.639***
(0.040) (0.011) (0.000) (0.004) (0.020) (0.005) (0.004) (0.073)
WFH × Time 0.254*** -0.012 -0.000 -0.003 -0.131*** -0.001 -0.005 -0.703***
(0.043) (0.013) (0.000) (0.004) (0.023) (0.005) (0.004) (0.081)

Employee FE Y Y Y Y Y Y Y Y
Team FE Y Y Y Y Y Y Y Y
R-squared 0.801 0.758 0.665 0.624 0.758 0.320 0.386 0.768
Observations 25,893 25,893 25,893 25,893 25,893 25,893 25,893 25, 893
Clusters 914 914 914 914 914 914 914 914

Note: “Internal NW” is the number of people inside the company with who employee had meaningful contact in last
28 days. “External NW” the same measure for people outside the company. “NW ORG” is the number of distinct
organizational units within the company that the employee had at least two meaningful interactions in the last four
weeks. “NW EXT” the same measure for external domains outside the company. “Meetings Manager” is the number
of meetings involving the manager. “Meetings 1:1” is the number of meetings between the employee and their manager.
“Coaching Meets” is the number between the employee, the manager and all their direct reports. “E-mails” is the
number of emails sent. The unit of observation is the employee-week. Standard errors are shown in brackets below the
point estimates, and are clustered at the employee level. ***Significant at the 0.1% level; **significant at the 1% level;
*significant at the 5% level.

6
B Additional Figures

Figure B.1: Kernel density estimates of subjective Ratings for different levels of Output.

Figure B.2: Working Hours over time with different stages of lockdown and removal of lockdown restrictions
in India. The leftmost dashed line indicates the time at which a national lockdown was imposed.
The first stage of unlocking (“Unlock 1”) allowed e.g. restaurants and shopping malls to reopen,
the second stage (“Unlock 2”) allowed limited travel and at the third stage (“Unlock 3”) gyms
for example reopened.

7
Figure B.3: Productivity during WFH depending on pre-WFH productivity. The solid line is the 45-degree
line.

(a) Number of Emails (b) Number of MS Teams Calls

Figure B.4: Technological shift pre- and post WFH. Panel (a): number of emails sent per week. Panel (b):
weakly number of calls a person joined through MS Teams. Time= 0 is the week 9th-15th March
2020.

8
 (a) Productivity Sapience (b) Productivity WPA

Figure B.5: Productivity pre- and post WFH in the WPA sample. To make both measures comparable, we
standardize them to mean 0 and standard deviation 1. Panel (a): standardized productivity using
Sapience measure of Input. Panel(b): standardized productivity using the WPA variable working
hours as input. Week= 0 is the week 9th-15th March 2020.

View publication stats