Content uploaded by Nico Stawarz
Author content
All content in this area was uploaded by Nico Stawarz on Jun 04, 2024
Content may be subject to copyright.
Travel Behaviour and Society 37 (2024) 100839
Available online 3 June 2024
2214-367X/© 2024 The Authors. Published by Elsevier Ltd on behalf of Hong Kong Society for Transportation Studies. This is an open access article under the CC
BY license (http://creativecommons.org/licenses/by/4.0/).
To what extent does working from home lead to savings in commuting
time? A panel analysis using the Australian HILDA Survey
Heiko Rüger
a
,
*
, Inga Laß
a
,
c
, Nico Stawarz
a
, Alexandra Mergener
b
a
Federal Institute for Population Research (BiB), Wiesbaden, Germany
b
Federal Institute for Vocational Education and Training (BIBB), Bonn, Germany
c
Melbourne Institute of Applied Economic and Social Research, The University of Melbourne, Carlton, Victoria, Australia
ARTICLE INFO
Keywords:
Working from home
Telework
Commuting time
Australia
HILDA Survey
Fixed effects regression
ABSTRACT
With growing concern about the climate impact of travel, a central question is the extent to which working from
home (WFH) can reduce commuting. Recently, the question has received even more attention as WFH has
increased sharply with the onset of the COVID-19 pandemic. However, the state of research is marked by mixed
results and lacking longitudinal evidence. We investigate the link between WFH and total weekly commuting
time by applying xed effects regression to panel data from the Australian HILDA Survey, covering the period
2002–2019. We go beyond previous research by examining the moderating roles of the extent of WFH, the
duration of the WFH episode, and gender. Overall, we nd that doing any work from home is associated with a
signicant decrease in employees’ weekly commuting time of about 14% on average. The reduction sets in
immediately with the start of WFH and tends to further increase thereafter. However, only high shares of WFH
are associated with substantial drops in commuting time, and reductions are larger for women than men. Taking
into account Australian workers’ reported WFH preferences, our results suggest maximum potential future
commuting time savings of about 17–25% compared to 2019.
1. Introduction
As early as the 1970s, researchers expressed the hope that working
from home (WFH) could reduce commuting and thus lower congestion,
pollution, and energy consumption (Nilles et al., 1976; Harkness, 1977).
Nevertheless, global CO
2
emissions from the transport sector have risen
sharply over the past two decades, with passenger cars contributing by
far the most in absolute terms (IEA, 2021). At the same time, commuter
trafc accounts for a signicant share of total trafc in many countries
(e.g. U. S. Department of Transportation, 2017; Eurostat, 2021). A
reduction in commuter trafc could therefore contribute to a signicant
reduction in emissions.
More recently, the COVID-19 pandemic has shown impressively that
work lives can change rapidly. Whereas previously WFH was practiced
rather rarely – although there were clear differences between countries
(e.g. Messenger, 2019) – the pandemic led to a sharp increase in the
prevalence of this work mode. For example, based on averaged data
from eight OECD countries, the share of workers WFH increased from
16 % before the pandemic to 37 % in March 2020 (OECD, 2021). As
WFH is expected to remain an important phenomenon beyond the
COVID-19-period (e.g. Buomprisco et al., 2021; Salon et al., 2022), the
long-standing question of whether WFH can reduce commuting has
received even more attention.
The simple assumption is that WFH reduces overall commuting
because it allows workers to reduce the number of trips to the workplace
(i.e. the frequency of commuting). However, the current state of
research is inconclusive. While earlier studies in particular suggest that
WFH leads – at least in the short term – to a reduction in total distance
travelled to work (for a review, see Andreev et al., 2010), other research
shows that home workers commute longer one-way distances than their
non-WFH counterparts (Helminen and Ristim¨
aki, 2007; Zhu, 2013; Melo
and de Abreu e Silva, 2017).
In general, the existing research on this topic has a number of limi-
tations. First, it relies mainly on cross-sectional data and lacks longitu-
dinal studies (Allen et al., 2015), with the exception of the studies by de
Vos et al., (2018; 2019) and Faber et al. (2023). This means that con-
clusions regarding causality are uncertain due to bias from unobserved
heterogeneity, such as selection with respect to ambition, ability, and
* Corresponding author.
E-mail addresses: heiko.rueger@bib.bund.de (H. Rüger), i.lass@unimelb.edu.au (I. Laß), nico.stawarz@bib.bund.de (N. Stawarz), mergener@bibb.de
(A. Mergener).
Contents lists available at ScienceDirect
Travel Behaviour and Society
journal homepage: www.elsevier.com/locate/tbs
https://doi.org/10.1016/j.tbs.2024.100839
Received 6 October 2023; Received in revised form 24 May 2024; Accepted 27 May 2024
Travel Behaviour and Society 37 (2024) 100839
2
residential or mobility preferences (de Graaff, 2004; Wang et al., 2023).
There are arguments that effects of WFH on commuting can be biased
upwards or downwards in simple cross-sectional studies (de Vos et al.,
2018). On the one hand, telecommuters may have stronger preferences
for shorter commutes. On the other hand, there might be a bias of res-
idential sorting, e.g. individuals living in certain areas may have longer
commutes and stronger preferences for telecommuting. Furthermore,
those who are more ambitious may have both a higher willingness to
commute and a higher likelihood of doing additional work from home. A
few studies use instrumental variable approaches to deal with issues of
omitted variables and reverse causality (Jiang, 2008; Zhu, 2012; Zhu
and Mason, 2014); however, these also produce mixed ndings (cf. de
Vos et al., 2018, p. 379).
Second, different measures for WFH and commuting make it difcult
to compare the results across studies (cf. Andreev et al., 2010; Elld´
er,
2020). In this context, the panel study by de Vos et al. (2018) uses one-
way commuting time, which has the drawback that it does not allow
predicting the impact of WFH on total commuting time (e.g. per week).
This is because employees may use WFH to work further away from
home, effectively increasing their daily commuting time, while their
average weekly commuting time remains unchanged or may even
decrease if the frequency of commuting is simultaneously reduced.
Third, few studies focus on whether the extent of WFH and gender
moderate the effect of WFH on commuting times and whether the effect
changes over time, i.e. whether it is short-term or persists over longer
periods (Allen et al., 2015). For example, the panel study by Faber et al.
(2023) examined the linear effect of weekly hours worked from home on
commuting time, the latter measured during a 3-day observation period,
without accounting for potential heterogeneity in the effect with respect
to the extent of WFH and gender. To the best of our knowledge, only one
study (Pabilonia and Vernon, 2022) has investigated the association
between the extent of WFH and commuting for men and women sepa-
rately. Given the marked gender differences in mobility behaviour and
WFH use (e.g. Crane, 2007; Powel and Craig, 2015), this seems an
important shortcoming.
The present paper adds to the literature in several ways. First, our
study draws on panel data from the Household, Income and Labour
Dynamics in Australia (HILDA) Survey, covering the period 2002–2019.
We apply xed effects (FE) panel regression models (e.g. Brüderl and
Ludwig, 2015) to this data to examine how within-person changes in
WFH are associated with changes in commuting time. Compared to cross-
sectional analysis, these models allow for a clearer interpretation in
terms of causal inference by implicitly controlling for all time-constant
unobserved characteristics (e.g. residential and commuting prefer-
ences). Second, we consider WFH as dichotomous indicator as well as
the share of time worked from home per week to take into account
possible effect differences in terms of the extent of WFH. Our outcome
variable represents total weekly commuting time, which has been
described as “ideal” and preferable to one-way commutes (Zhu, 2012;
Melo and de Abreu e Silva, 2017), as it captures other aspects of
commuting, such as frequency and timing, in addition to duration/dis-
tance of the single commute, which is important to assess the impact on
total commuting time. Third, our study examines whether the effect of
WFH on commuting time differs by gender, considering the clear gender
differences in mobility behaviour and WFH use. Fourth, we are the rst
in this eld to estimate FE models with impact functions (e.g. Ludwig
and Brüderl, 2021) that allow investigating how the effect of WFH on
commuting time evolves over the course of the home working episode.
Overall, the HILDA Survey data offers unique advantages in terms of
design, representativeness, sample size, variables on WFH and
commuting, and relevant control variables.
Australia is an interesting case study for several reasons. The sheer
size of the country and the large distances between its major cities
suggests that bridging spatial distances is of crucial importance in
Australia. Pre-pandemic commuting times were long and increasing
over time. Over the 2002–2017 period alone, Australia saw a 23 %
increase in average daily commuting times from 49 min to one hour.
Over the same period, the share of workers with long daily commutes of
two hours or more increased from 12 to 18 %. Commuting times are
particularly long in the most densely populated areas and among men
(Wilkins et al., 2019). Finally, Australia stands out as one of very few
countries that provides specic groups of workers with the right to
request exible work arrangements, including WFH (OECD, 2021). This
right has been established with the Fair Work Act 2009 and relates to
(non-casual) employees who have worked for their employer for at least
12 months and who, for example, have parenting or caring re-
sponsibilities, have a disability, or are older than 54 years. However,
employers can refuse such request on reasonable business grounds.
Overall, about 25 % of Australian workers aged 18 to 64 years reported
usually doing some work from home in 2019 (based on HILDA Survey
data). Even higher levels of WFH can be expected in Australia after
COVID-19. Recent surveys – like the University of Melbourne’s Taking
the Pulse of the Nation Survey (Petrie, 2022) and Boston Consulting
Group’s Workforce Sentiment Survey (Mattey et al., 2020) – show that
many Australian workers would like to continue to work (at least a part
of their time) from home.
2. Literature review, theoretical considerations and hypotheses
In this study, we use the term WFH as a synonym for telework, which
is understood to mean all work performed in paid employment that is
done at home away from the workplace, usually using Information and
Communication Technologies (ICT) (Nilles, 1988). In theoretical terms,
we assume that work life is organised around home and work location,
and the two are linked by commuting distance, where commuting can be
viewed as a cost (e.g. travel expenses, stress, limited time for family and
leisure) (Kalter, 1994; Rouwendal and Rietveld, 1994; Abraham and
Nisic, 2007).
2.1. The effect of working from home on commuting time
Building on Salomon (1985), four theoretical mechanisms in the
relationship between WFH and commuting can be differentiated: Sub-
stitution means that WFH leads to a reduction in commuting; comple-
mentarity implies that WFH generates further commuting; modication
means that WFH leads to a change in commuting behaviour; and
neutrality suggests that WFH does not lead to any effects on commuting
behaviour.
In terms of substitution, WFH can reduce travel between work and
home when the number of trips to the workplace (i.e. the weekly
commute frequency) is reduced (Andreev et al., 2010; Allen et al., 2015;
de Abreu e Silva and Melo, 2018a). Moreover, people with long com-
mutes should be more inclined to WFH, since for them the savings due to
the omitted commutes are particularly high (e.g. Helminen and Ris-
tim¨
aki, 2007; Zhu, 2013; Mergener and Mansfeld, 2021), and WFH is
often considered a strategy for coping with longer and more expensive
commuting trips, at least in the short term (de Abreu e Silva and Melo,
2018b).
However, based on search theory, individuals can be expected to
constantly search for new jobs and/or dwellings to maximise wages and
place utility (Devine and Kiefer, 1993; van Ommeren et al., 1999). Since
WFH can reduce commuting costs, it enables persons to widen their
search radius, which increases the accessible number of more valuable
jobs and better dwellings (e.g. Sjaastad, 1962; So et al., 2001; Manning,
2003). Thus, WFH can dissolve the strong ties between place of resi-
dence and employer’s premises, so that employees no longer attach as
much importance to living close to their place of work (e.g. DLR, 2021).
Individuals may be willing to accept longer one-way commutes in favour
of a better job or dwelling knowing that they can reduce commuting
times by WFH if necessary (e.g. Muhammed et al., 2007; de Abreu e Silva
and Melo, 2018b).This could indicate a complementarity effect, where
the reduced number of trips is (more than) offset by longer one-way
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
3
commuting distances (e.g. Ravalet and R´
erat, 2019).
WFH could also result in a modication of commuting patterns with
commuting being shifted to off-peak hours (Andreev et al., 2010; Huang
et al., 2023). This, in turn, could lead to a (small) reduction in
commuting time due to lower trafc volumes. In this regard, studies
show that employees who do not WFH on a full-day basis are more likely
to commute off-peak (Glogger et al., 2008; Asgari and Jin, 2018). In
addition, car use and the use of active transportation modes seem to
increase with WFH (de Abreu e Silva and Melo, 2018a; Elld´
er, 2020),
which also represents a modication of commuting patterns that could
affect commuting time.
WFH could also be used for reasons other than to reduce commuting
time. In particular, WFH could be deployed to perform additional work
from home, e.g., in the evenings or on weekends. Telework has repeat-
edly been shown to be associated with longer working hours (Glass and
Noonan, 2016; Gschwind and Vargas, 2019). In terms of neutrality, this
could mean that WFH has no impact on commuting behaviour. If,
however, by doing the extra work from home, commuting is increased at
peak instead of off-peak times, this could have a complementary effect
and increase commuting time (de Graaff, 2004).
Further and updated empirical research is needed to gain insights
into the relationships between WFH and commuting, as previous studies
do not provide a clear picture (cf. Elld´
er, 2020). In their literature review
of fairly early studies, with the majority conducted before the year 2000,
Andreev et al. (2010) concluded that, in the short term, WFH leads to a
(small) reduction in distance travelled to work. For instance, a U.S. study
shows that the reduction in annual vehicle miles travelled due to tele-
commuting was 0.8 % or less (Choo et al., 2005). Mokhtarian et al.
(2004) also nd teleworkers’ total commuting to be shorter compared to
non-teleworkers, even if teleworkers have longer company-home dis-
tances. Likewise, other recent studies nd substitution effects in the
form of reduced travel (Lachapelle et al., 2018; Elld´
er, 2020; Faber et al.,
2023).
By contrast, some recent studies found complementarity effects (i.e.
WFH generates additional commuting) (cf. Elld´
er, 2020). For instance,
de Abreu e Silva and Melo (2018b) showed a positive association of
telework frequency with weekly commuting distance travelled,
although the association between telework and the number of
commuting trips was negative. Using panel data from Dutch workers, de
Vos et al. (2018) found that telecommuting was associated with 5 %
longer one-way commuting times. Similarly, a recent study by Motte-
Baumvol and Schwanen (2024), using cross-sectional data, found that
teleworkers in the UK had longer commuting times despite commuting
fewer days than other workers, which was mainly due to different
commuting patterns (e.g. commuting by train and/or working in a
congested area).
Overall, considering the theoretical arguments and empirical evi-
dence, WFH and commuting appear to be substitutes – although not
perfect ones, and savings in commuting may turn out to be lower than
expected. Therefore, we put forward our rst hypothesis:
H1a: Working from home is associated with a small reduction in weekly
commuting time.
Following the theoretical arguments above, a crucial moderating
factor is the extent of WFH. If only a small share of time is worked from
home, complementarity and neutrality effects should dominate. For
example, a low WFH extent may result from employees simply bringing
additional work home, while still commuting on a daily basis. In addi-
tion, longer one-way commuting distances could be less effectively
compensated if commuting trips are only rarely saved. However, rela-
tively few studies have examined the moderating effect of the extent of
WFH on the relationship between WFH and commuting. Most of these
suggest that overall commuting time savings are larger for those who
work from home to a greater extent (e.g. Melo and de Abreu e Silva,
2017; Lachapelle et al., 2018; Elld´
er, 2020; Caldarola and Sorrell, 2022;
Pabilonia and Vernon, 2022). For instance, according to a British study
(Melo and de Abreu e Silva, 2017), a signicant reduction in weekly
commute time is only achieved with a teleworking extent of 3 +days/
week. Overall, our theoretical considerations and the existing empirical
ndings suggest that the extent of WFH moderates the effect on
commuting time: A small share of total working time done from home
might not yield any savings in commute time, but there might be a
“tipping point” at a greater extent beyond which commute time is
greatly reduced. Therefore, we put forward our second hypothesis:
H2a: Only a large extent of WFH will result in a substantial reduction in
weekly commuting time.
2.2. The role of gender differences in WFH and commuting
Since mobility behaviour, including WFH, is gender-specic, we
expect differences for female and male employees regarding the link
between WFH and commuting. Research has repeatedly shown that
women have shorter commutes than men (e.g. Crane, 2007; Wilkins
et al., 2019; Skora et al., 2020). Following the Household Responsibility
Hypothesis (e.g. Turner and Niemeier, 1997; Gimenez-Nadal and
Molina, 2016), mobility differences are primarily explained by pre-
vailing traditional gender roles. Men are more likely to be main bread-
winners and more focused on their employment careers. Women, by
contrast, more often act as secondary earners on a part-time basis and
take on more responsibility for household tasks and unpaid care work (e.
g. Altintas and Sullivan, 2016; Pailh´
e et al., 2021). Given this gendered
labour division, women are more limited in time and space than men
(Nisic, 2017). This is reected in women’s shorter commutes, as it is
more important for them to choose a job close to home to take care of
other obligations on short notice, such as running errands or dropping
off and picking up children, if needed (McGuckin et al., 2005). More-
over, the cost-benet ratio of commuting is worse for women than for
men, as commuting is typically associated with monetary costs due to
travel expenses (van Ommeren and Fosgerau, 2009), whereas earnings
are usually lower among secondary earners. In addition, long commutes
can lead to travel-related stress, especially for women (e.g. Rüger et al.,
2017), and limit available time that could be used for household re-
sponsibilities, resulting in higher opportunity costs for women (e.g. Van
den Berg and Gorter, 1997; Rouwendal and Nijkamp, 2004).
Thus, WFH to save valuable time by reducing commuting should be
more important for women than men. This notion is corroborated by
previous research elucidating that the use of WFH seems to be charac-
terised by traditional gender roles. While women tend to use WFH for
combining work and private life, or to even spend more time on do-
mestic work and childcare (Powel and Craig, 2015), men do not pri-
marily consider WFH to better meet private demands (Sullivan and
Lewis, 2001). By contrast, studies suggest that men are more likely to use
telework to enhance their work performance, leading to working longer
hours and during leisure time to meet work demands (e.g. Glass and
Noonan, 2016; Eurofound and the International Labour Ofce, 2017).
Thus, men appear to use WFH more often to perform additional work (e.
g. in the evening) rather than to replace regular ofce hours with WFH.
By this, no savings in commuting time can be expected. Moreover, men
with a small extent of WFH commute longer distances than men who do
not WFH at all, while this is not the case for women (Mergener and
Mansfeld, 2021). Thus, only when a large part of the working time is
spent at home, a reduction in the total commuting time can also be
expected for men. These considerations suggest that gender moderates
the effect of (the extent of) WFH on commuting time and lead us to the
following hypothesis:
H1b: Overall, WFH is associated with a stronger reduction in weekly
commuting time among women than men.
H2b: For women, a reduction in weekly commuting time can also be
observed with a smaller extent of WFH compared to men.
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
4
2.3. Variations over time in the effect of working from home on
commuting time
To our knowledge, no research so far has examined how the effect of
WFH on commute time evolves over the course of the home working
episode (for a general argument in favour of more research on the role of
time in relation to WFH using longitudinal data, see Allen et al., 2015).
In principle, the effect could occur immediately with the onset of WFH
or later, and it could be stable, increasing or decreasing over time as the
worker maintains WFH. Therefore, the dynamics of commuting trajec-
tories over the course of the WFH episode could reveal more about the
nature of the relationship between WFH and commuting and the po-
tential underlying mechanisms.
As discussed earlier, WFH may be an incentive for longer one-way
commutes by expanding the search radius for residences and/or work
locations or by reducing residential relocations closer to the workplace
(e.g. Muhammed et al., 2007; de Vos et al., 2018; Ravalet and R´
erat,
2019). Changing residences and/or work locations could therefore be
processes that might offset the commuting time saved by WFH in the
long run. This would suggest that a potential negative effect of WFH on
commuting time decreases or even turns positive over the course of the
home working episode. However, most existing studies suggest no or
only a small counteracting effect by changes of residences and/or work
locations (Ettema, 2010; Kim, 2016; Muhammed et al., 2007; Ory and
Mokhtarian, 2006; but see Ory and Mokhtarian, 2005). Thus, a clear
tendency for workers to trade off lower commuting frequency for longer
commuting distances over time cannot be identied based on their
relocation behaviour.
Following another line of reasoning, commuting time can be ex-
pected to continue to decrease over time as workers (are able to) expand
their share of WFH for several reasons: First, as the duration of the WFH
experience increases, trust builds between employers and employees,
which appears to be a key factor in successful work and good perfor-
mance at home (Kim et al., 2021) and allows employees to increase their
share of time WFH (Golden, 2006). However, building trust takes time,
which is why a common strategy is to initially introduce WFH as a test or
trial for a period of time (Beauregard et al., 2019). If the employer’s trust
and expectations are met, it can be assumed that employees will (be
allowed to) continue to use WFH and also extend their share of time
WFH (Kaplan et al., 2018). Second, WFH can also present challenges for
employees when they start using it, as specic competencies are
required to be productive at home, e.g. self-discipline, self-motivation,
ability to work alone, and time management skills (Baruch, 2000). Such
competencies can be acquired with increasing experience with WFH.
Thus, it can be assumed that WFH becomes more attractive with
increasing experience, which is why employees then also extend their
share of time WFH over time. Third, being productive when WFH also
depends on the physical set-up at home, e.g. the technological infra-
structure and furniture. If employees enjoy WFH and receive support for
it from their employers, it can be assumed that they will equip their
ofces at home better and more professionally over time (Gajendran and
Harrison, 2007; Scott et al., 2012; Carillo et al., 2021). Thus, working in
an increasingly well-equipped and optimised work environment may
lead to an increase in the share of time WFH over time.
Home workers may not only be able to optimise their work envi-
ronment but also the rhythm of everyday life with continued WFH. They
may increasingly optimise their commuting behaviour, for example, by
commuting at off-peak times, which may lead to further savings in
commuting time. Furthermore, it may take some time for all household
members to adjust to a new pattern of working time, household chores
and family time after someone in the household starts WFH. Again, this
suggests that the home worker may be able to extent the share of WFH
over time, leading to additional commuting time savings.
Based on these considerations, we put forward our third hypothesis:
H3: The negative effect of working from home on weekly commuting time
further increases over time.
3. Data and methods
3.1. Data
We draw on data for the period 2002 to 2019 from Release 19 of the
Household, Income and Labour Dynamics in Australia (HILDA) Survey
(Department of Social Services and Melbourne Institute of Applied
Economic and Social Research, 2020). The HILDA Survey is an annual
household panel study rst conducted in 2001 that captures a repre-
sentative sample of households in Australia (Watson and Wooden,
2021). We restrict the sample to employees aged 16 to 64 years with at
least 10 working hours per week (n =135,365 observations). For rea-
sons of plausibility, we exclude observations where workers report a
commuting time of zero but at the same time work less than 95 % of their
time from home (n =1,791). After also excluding cases with missing
values (n =21,377), our sample comprises 112,197 observations from
18,803 individuals.
3.2. Measures
As dependent variable we use the logarithmised total weekly
commuting time in hours and minutes. This information was collected
via a question asking workers to report the time they spend travelling to
and from a place of paid employment in a typical week. In this regard,
total weekly commuting time has been described as an “ideal” measure
(Zhu, 2012; Melo and de Abreu e Silva, 2017), as it simultaneously
captures several aspects of commuting, such as frequency, timing, and
duration/distance of the single commute. As we used the natural loga-
rithm of weekly commuting time to approximate a normal distribution,
all observations with zero commuting time were assigned a value of
0.001 h to be included in the analysis.
As central independent variables we draw on the question “In your
main job, are any of your usual working hours worked at your home
(that is, the address of your usual place of residence)?” to create a
dichotomous indicator for WFH and the question “Approximately how
many hours each week do you usually work at home?” to operationalise
the extent of WFH. We divide the number of hours usually worked from
home by the total number of working hours in the main job to generate a
measure of the share of time worked from home. In order to test for non-
linear effects, we categorise this share of time as 0 %, 1–19 %, 20–39 %,
40–59 %, 60–79 % and 80–100 %.
In order to adjust for general shifts in commuting times, we control
for survey year in all models. In a second specication, we add worker-
and job-related control variables, broadly following de Vos et al. (2018).
In terms of worker characteristics, we consider the presence of a partner
or children in the household, the weekly wage of the partner (deated
using the wage price index; Australian Bureau of Statistics, 2021), the
educational level, the labour market experience (as measured by years in
paid work in a quadratic specication) as well as the region of residence
(differentiating between major cities, inner regional areas, and more
remote areas). In terms of job characteristics, we account for the daily
wage (deated) and weekly working hours in the main job, occupation,
industry and rm size. Given our commuting time measure relates to
entire working weeks rather than individual trips, we also control for the
number of working days in the main job. Finally, we include an indicator
for multiple job holders as the commuting time measure reects time
spent travelling to and from all jobs, whereas the information on WFH
relates to the main job only. Frequencies for all variables in the sample
(except survey year) are presented in Table A1 in the appendix.
3.3. Methods
To investigate how WFH impacts on weekly commuting time we use
xed-effects (FE) panel regression models (e.g. Brüderl and Ludwig,
2015). These models implicitly control for time-invariant unobserved
individual characteristics, because they use only within-person variance
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
5
over time. Consequently, time-constant characteristics that may affect
both the likelihood of WFH and commuting time, such as ambition,
ability, residential or mobility preferences, are accounted for. This
means we come closer to identifying the causal link between WFH and
commuting time.
A potential limitation of FE regression is that low variance in key
variables may result in limited external validity, as the ndings may
only apply to a selected subset of respondents (Hill et al., 2020). How-
ever, background analyses show that in our case the extent to which
individuals work from home is very variable. The share of workers in the
sample who are observed changing between any of the six (0 % up to
80–100 %) categories over the course of the study is around 27 %
overall. This is because most people never work from home (since most
jobs actually cannot be done from home, e.g. Sostero et al., 2020).
Among those individuals who did work from home at any point in time,
88 % changed categories. These “changers” account for 94 % of the
observations in the WFH categories between 1–19 % and 80–100 %. In
Table A2 we present the characteristics of workers who are observed
changing between WFH categories, separately by WFH category (with
some categories combined to save space).
In addition, we use FE models with impact functions (Ludwig and
Brüderl, 2021) that help us understand how the effect of WFH on
commuting is distributed over time (i.e. changes after the onset of WFH).
We constructed a variable counting the number of years that have
passed since entry into WFH, with periods longer than ten years grouped
together. All time points prior to entry into WFH serve as reference. The
results presented in this study were estimated using the command ‘xtreg’
in Stata 16 with standard errors clustered on the individual level.
To examine the effect of WFH on commuting time, we estimate three
sets of models: First, we investigate the overall effect of WFH on
commuting time using a binary indicator for home working. Second, we
analyse the effect of WFH in more detail by looking at the share of time
worked from home. For these steps, we analyse separate models by
gender and formally test for gender differences by running models with
an interaction term for WFH and being female. Third, we use impact
function FE models to analyse how the effect of WFH on commuting time
varies over time. In addition, we combine the results of our FE models
with the ndings from recent surveys on Australian workers’ stated
preferences regarding future WFH patterns (Mattey et al., 2020; Petrie,
2022) to provide insights for policy on how much commuting time could
potentially be saved in the future through WFH if workers’ stated
preferences were accommodated.
4. Results
4.1. Overall effect
Panel A in Table 1 presents results from the analysis of the binary
measure, indicating whether workers do or do not work any of their
usual hours from home, for the pooled sample and separately for female
and male workers. In a base model, we only controlled for survey year,
and in an extended model, we included all other control variables. In
order to conserve space, only the coefcients of interest are reported;
however, the full models can be found in Appendix Table A3. Almost
across the board, doing any work from home in a usual week was
associated with a signicant decrease in commuting time. For example,
the coefcient of −0.092 in the base model for the pooled sample re-
ects about a 9 % reduction in commuting time (calculated as 100*(exp
(ß)-1)). Furthermore, in all cases, the effect was larger in the extended
compared to the base model, suggesting that individuals working from
home have job and personal characteristics associated with lengthy
commutes. For example, the coefcient of −0.149 in the extended model
for the pooled sample reects about a 14 % reduction in commuting
time. We also formally tested for gender differences by re-running the
models with an interaction term for WFH and being female. The inter-
action term was sizable and statistically signicant (b
base
=-0.138, p ≤
0.01; b
extended
=-0.152, p ≤0.01), suggesting that WFH was associated
with larger reductions in commuting time among female compared to
male home workers.
Panel B of Table 1 presents results from models considering the
extent of WFH in the form of the share of hours worked from home as a
percentage of total working hours (estimates of control variables not
reported but very similar to those presented in Table A3). The base
model showed a non-linear relationship between the share of time
worked from home and weekly commuting time: Relatively low shares
of WFH (below 40 % for the pooled sample as well as for men and below
20 % for women) were associated with signicantly longer commutes
compared to workers not WFH at all, whereas higher shares were
associated with shorter commutes compared to those not WFH. In the
extended model, the positive coefcients for the lower categories
attenuated, but the positive association of working 20–39 % of the time
from home among men remained signicant. Among both genders, the
signicant negative associations for the two highest categories remained
statistically signicant and similar in size in the extended model.
Overall, the reduction in commuting time was much larger in the top
category than in the other categories. For example, the coefcients in the
extended model for all workers designated a 32 % decrease in
commuting time for those working 60–79 % from home but a 98 %
decrease for those working at least 80 % from home. A formal test for
Table 1
Working from home and (ln) weekly commuting time: Results from xed-effects regression (b-coefcients).
(1) (2) (3) (4) (5) (6)
All base model All extended model Men
base model
Men extended model Women base model Women extended model
Panel A: Doing any work from home (ref. ¼no)
Works from home −0.092** −0.149** −0.022 −0.063** −0.158** −0.229**
n(observations) 112,197 112,197 55,603 55,603 56,594 56,594
adj. within-R
2
0.012 0.060 0.014 0.053 0.011 0.073
Panel B: Share of time worked from home (%, ref. ¼none)
1–19 0.065** 0.005 0.058** 0.017 0.074** −0.007
20–39 0.063** 0.015 0.103** 0.058* 0.038 −0.013
40–59 −0.040 −0.064+0.075 0.046 −0.110* −0.128**
60–79 −0.396** −0.388** −0.359** −0.351** −0.396** −0.383**
80–100 −3.909** −3.889** −3.371** −3.358** −4.193** −4.151**
n(observations) 112,134 112,134 55,580 55,580 56,554 56,554
adj. within-R
2
0.105 0.149 0.069 0.107 0.135 0.188
Notes: **, * and
+
denote statistical signicance at the 0.01, 0.05 and 0.10 levels, respectively. The base models are adjusted for survey year, the extended models
additionally control for worker and job-related characteristics (see Methods section and Table A3).
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
6
gender differences in the form of an interaction model revealed that the
effects of a WFH share of 20–39 % (b
extended
=-0.063, p =0.070), 40–59
% (b
extended
=-0.180, p =0.007) and 80–100 % (b
extended
=-0.811, p =
0.039) differed signicantly between men and women, indicating that
commuting time savings were notably larger for women than men even
when WFH to a similar extent.
4.2. Variations in the effect over time
Fig. 1 presents results from FE regression with impact functions. It
shows a marked and signicant drop in commuting time of around 10 %
in the rst year after starting WFH compared to the years before WFH.
For workers who continued WFH, commuting time decreased further
over time.
Separate analyses by gender (results not shown) revealed that this
pattern was particularly marked for women. Furthermore, background
analyses showed that the decrease in commuting time was partly due to
workers extending their share of time WFH over time, with the average
share of time worked from home increasing from 15.2 % in year 1 to
25.1 % in year 9. Another relevant factor appeared to be selection, as
those who eventually ended up WFH for a long period already started
out with slightly higher shares of time WFH. Precisely, those who ended
up WFH for ve years or less started out with an average WFH share of
15.0 % in year 1, whereas those who ended up WFH for six years or more
started out with an average WFH share of 17.9 %. However, in both
groups, the share of time WFH increased over the years.
Additionally, we analysed the effects of exiting WFH on commuting
time (results not shown). The ndings complemented those for entry
into home working: Exit from WFH resulted in a signicant initial in-
crease in commuting time of around 12 % (b
extended
=0.112, p ≤0.01),
followed by very modest further increases over time.
4.3. Possible post-COVID-19 commuting scenarios: Tentative projections
Several recent surveys showed that many Australian workers would
like to continue to work (at least part of their time) from home (e.g.
Mattey et al., 2020; Beck and Hensher, 2021; Petrie, 2022). Combining
our estimates with the ndings from these surveys can provide tentative
insights for policy into how much commuting time could be saved in the
future through WFH on a national level if employees’ preferences were
accommodated. Using workers’ stated preferences for our projections
means that the scenarios presented likely show the maximum potential
savings in weekly commuting time in a ’what if’ scenario. In this context,
it should be noted that preferences which are stated in hypothetical form
in surveys may differ from revealed preferences. That is, even when
given the choice to work from home as much as they want, some of those
workers who stated a preference for extensive WFH may still end up
WFH less often. This could be, for example, because they miss the face-
to-face contact with co-workers, are disturbed by others when WFH, or
are less productive when WFH. Overall, the realisation of preferences
depends on a number of factors (e.g. Mokhtarian and Salomon, 1994).
Table 2 provides information from two representative surveys, the
University of Melbourne’s Taking the Pulse of the Nation Survey (TTPN,
Petrie, 2022) and the Boston Consulting Group’s Workforce Sentiment
Survey (WSS, Mattey et al., 2020), and links these to our estimates.
In the data from three periods of the TTPN Survey (April 2021,
September/October 2021 and January 2022, N =7,200 from a total of
six survey rounds), close to half (49 %) of employed respondents could
not or would not want to work from home at all. Among the remaining
workers, a large share (24 %) responded that they would like to work
from home between 80 and 100 % of the time, while smaller shares
would like to work some of their time from home but less than 80 %.
Compared to the actual distribution among employees in the HILDA
Survey in 2019, the realisation of this scenario would entail a decline of
the share of workers not WFH (–33 percentage points) and WFH less
than 20 % of their time (−6 percentage points), and an increase in the
categories with a higher WFH share, particularly the 80–100 % category
(+22 percentage points). We have then combined (i.e., multiplied) these
increases/decreases with our estimated differences in weekly
commuting time between home workers in these categories and workers
not WFH from the extended model for all workers in Table 1 to gauge
overall changes in weekly commuting time on a national level. For
example, combining the 22 percentage points inux into the WFH
category of 80–100 % with the estimated reduction in commuting time
of 98 % for every worker who changes into this category from not WFH
would result in about a 21.9 % reduction in overall commuting time
([22.4*-98.0]/100). Evidently, the validity of this extrapolation is based
on the assumption that future home workers would behave similarly in
terms of commuting patterns to the home workers that our FE regression
analysis is based on. Under this assumption, we would see an overall
decline in commuting times of close to 25 %, if future employees were to
work according to the pattern indicated in the TTPN. This decline would
largely be driven by the large increase in the share of workers in the
80–100 % category. A similar conclusion would be reached if we based
Fig. 1. Changes in (ln) weekly commuting time since starting WFH: Estimates from impact function FE model, 95 % condence intervals. Notes: N =104,912.
Control variables as in the extended model in Table 1 (except survey year grouped in ve periods instead of single years). Each “year category” consists of at least 119
cases (in year 9) up to 6,085 cases (in the 1-year category).
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
7
our assumptions on the survey data presented in Beck and Hensher
(2021), where the overall pattern of desired post-COVID home working
patterns is similar to TTPN.
By contrast, respondents in the WSS (conducted in May 2020, N =
1,002) did not show an equally strong preference towards exclusive
WFH as the TTPN or Beck & Hensher survey respondents. Only 15 %
regard it as ideal to work from home for more than 80 % of their working
time. Instead, larger shares chose the middle categories of working be-
tween 1 and 80 % of the time from home. Compared to the actual dis-
tribution in 2019, realisation of this scenario would yield the biggest
increase of workers in the 40–60 % category (+16 percentage points).
However, based on our estimates, this category would only be associated
with a 6 % reduction in commuting time compared to the “none” cate-
gory. Consequently, a future working pattern along the lines of the
preferences stated in the WSS would result in much smaller commuting
time savings, amounting to about 17 % overall.
4.4. Further analyses
We re-ran the extended models of Table 1 with several modications
(key results can be found in Appendix Table A4). First, we excluded
multiple job holders given the information on WFH related to the main
job, whereas information on commuting time related to all jobs. The
results were similar to the main models, but the negative coefcients
pertaining to large WFH shares became larger.
Second, we sought potential explanations for the nding that the
association between working 20–39 % of the time from home and
commuting time remained positive in the extended model for men (b =
0.058, see Table 1, model 4). Further analyses showed that there were
ve industry subdivisions (based on the 2-digit Australian and New
Zealand Standard Industrial Classication) in particular that contrib-
uted to this effect: Finance, Public Administration, Administrative Ser-
vices (e.g., Travel Agency and Tour Arrangement Services), Food and
Beverage Service, and Building Construction. Upon exclusion of these
industries, the weak positive effect disappeared almost entirely. We can
only speculate as to why some workers in these industries experience
longer commutes when working in hybrid arrangements. What these
industries could potentially have in common is a relatively high share of
employees with intensive work-related travel (e.g. business trips, on-site
customer service). Since in the HILDA Survey, commuting time was
collected as “time spent travelling to and from a place of paid employ-
ment” (see Data and methods section), some employees in these in-
dustries may have understood this to refer (also) to their work-related
travel (cf. Caldarola and Sorrell, 2022). Studies have shown a strong
positive association between WFH and work-related travel (e.g. Zhu and
Mason, 2014; Caldarola and Sorrell, 2022). However, if we recalculate
our predictions from Table 2 excluding these industries, the results are
practically identical. All in all, despite some workers potentially
including work-related travel in their estimate of commuting time, this
issue appears to be minor and the measurement of commuting in the
HILDA Survey can be considered valid: It provides results that are
comparable to other surveys, for example in terms of commuting trip
duration (Bureau of Infrastructure, Transport and Regional Economics,
2016).
Third, we analysed the association between WFH changes and
commuting time changes that happen while working for the same
employer. This way, we cancelled out the effect of changes in WFH
patterns and commuting time that co-occur with a job change. Overall,
the results are very similar to the main model, although WFH appeared
to be associated with slightly greater commuting time savings for hybrid
work arrangements. For example, in the model for all workers, the co-
efcient of the 40–59 % WFH category (b =-0.109) is more negative
than in the corresponding main model (b =-0.064), and among men, the
positive coefcient for the 20–39 % category from the main model (b =
0.058) was practically reduced to zero (b =0.013). One explanation
could be that the mechanism of switching to a new job that allows
(more) WFH and at the same time is further away from home is no longer
in play. This mechanism would represent a complementarity effect in
which the reduced number of trips is offset by longer one-way
commuting distances.
Fourth, we attempted to gauge the relevance of unobserved hetero-
geneity by comparing the results from our FE models to pooled OLS
models, where the sample in the OLS models was restricted to workers
changing WFH category to ensure comparability. The results showed
that for low or medium shares of WFH (up to 79 %), positive coefcients
were larger and negative coefcients were smaller in the OLS model
compared to the FE model (with most of these differences being quite
substantial), suggesting these workers have unobserved traits connected
with longer commutes. For example, while the coefcient in the
extended FE model for all workers indicated a 32 % reduction in
commuting time for those working 60–79 % from home (b =-0.388),
this reduction amounted to only 17 % in the pooled OLS model (b =
-0.189). By contrast, the negative coefcient of the topmost category
was slightly larger in the pooled OLS model, indicating this group has
unobserved traits related to shorter commutes. Use of the pooled OLS
estimates instead of the FE estimates for the tentative projections in
Table 2 would therefore result in markedly smaller overall commuting
time savings, e.g. 14 % instead of 17 % when using the WSS data.
Fifth, we used the number of hours usually worked from home
(rather than the share) as the measure for the extent of WFH. The results
were similar to those from the models using the share of time worked
from home. However, the link between the extent of WFH and
commuting time was more linear for men, that is, there was no positive
association between WFH relatively few hours and commuting time.
Sixth, we used the daily commuting time (rather than the weekly
commuting time) as dependent variable (Appendix Table A5). Following
the approach of Laß et al. (2023) and Botha et al. (2023), we calculated
the daily commuting time by dividing the weekly commuting time by
the usual number of working days per week in the main job (resulting in
the average commuting time per working day). As we have no
Table 2
Projections of future weekly commuting time savings based on worker preferences for post-COVID WFH.
Actual
2019
Desired
post-COVID-19
Difference actual and desired Estimates
a
Savings in
commuting time
Share WFH (%) HILDA TTPN WSS TTPN WSS HILDA TTPN WSS
None 81.3 48.5 40.0 –32.8 −41.3 0.0 0.0
<20 9.8 4.0 8.0 −5.8 −1.8 0.5 0.0 0.0
20–39 6.0 7.1 11.0 1.2 5.0 1.5 0.0 0.1
40–59 1.4 9.4 17.0 8.1 15.6 −6.2 −0.5 −1.0
60–79 0.5 7.5 9.0 7.0 8.5 –32.1 −2.2 −2.7
80–100 1.1 23.5 15.0 22.4 13.9 −98.0 −21.9 −13.6
Total 100.0 100.0 100.0 −24.7 −17.3
Notes: Actual distribution in 2019 based on weighted data for employees aged 16–64 years in the HILDA Survey. TTPN Survey and WSS data based on all adult workers.
Original WSS categories were slightly different (1–20 %, 21–40 %, 41–60 %, 61–80 %, 81–100 %).
a
Estimates taken from the extended model for all workers in Panel B,
Table 1, calculated as 100*(exp(ß)-1).
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
8
information in the data on the number of WFH days or ofce days (i.e.
the frequency of commuting), we cannot determine the daily commuting
time per ofce day. Overall, the results were very similar to those from
the models using the weekly commuting time as dependent variable.
5. Discussion and conclusion
This paper complements previous literature on the association be-
tween WFH and commuting behaviour in several ways. First, we
deployed 18 waves of panel data from the HILDA Survey and applied
xed effects (FE) panel regression models to examine how changes in
WFH are associated with changes in commuting time. These models
brought us closer to identifying the causal effect by implicitly control-
ling for all time-constant unobserved characteristics. The comparison
with pooled OLS models suggested that cross-sectional models sub-
stantially underestimate the reductions in commuting associated with
low and medium shares of WFH and slightly overestimate the reductions
associated with high shares of WFH. Second, we measured WFH using
both a dichotomous indicator as well as the share of time worked from
home per week, thereby taking into account possible effect heteroge-
neity in terms of the extent of WFH. Commuting was measured using
weekly commuting time, which has been described as preferable to one-
way commutes in the literature. Third, we examined the potential
moderating role of gender. Fourth, we estimated impact function FE
models that allow to investigate how the effect of WFH on commuting
time evolves over the course of the home working episode.
The results of the extended FE models showed that doing any work
from home was associated with a signicant decrease in weekly
commuting time of 14 %, supporting hypothesis H1a. This nding is
consistent with the notion of a substitution effect and supports the
literature which concludes that WFH leads to a decrease in overall
commuting (e.g. Andreev et al., 2010; Lachapelle et al., 2018; Elld´
er,
2020). It also provides an insightful contrast to the panel study by de Vos
et al. (2018), who used daily instead of weekly commuting time
(measured as the time it takes to travel from home to work) and found
that teleworking was associated with 5 % longer daily commuting times.
However, as the weekly commuting time indicator takes into account
commuting frequency as well as daily commuting time, we argue that it
constitutes a more accurate measure of the overall changes in
commuting.
Additionally, considering the extent of WFH revealed that low shares
of WFH were associated with similar commutes compared to workers
who do not work from home at all, whereas higher shares were associ-
ated with shorter commutes. Overall, the reduction in commuting time
was much larger in the top category than in the other categories, sup-
porting hypothesis H2a as well as existing ndings (e.g. de Abreu e Silva
and Melo, 2018b; Elld´
er, 2020). A large share of time worked from home
likely represents several entire working days spent at home rather than
in the ofce. By contrast, smaller shares of time worked from home may
for many workers simply reect work brought home in addition to a full
day in the ofce.
Our study also highlights important gender differences in the effect
of WFH on commuting. Overall, we found larger reductions in
commuting time among female home workers, supporting hypothesis
H1b. This is in line with our expectation that, due to traditional gender
roles, commuting is associated with higher costs for women, and the
time-saving potential of WFH is therefore more important for them.
Focusing on the extent of home working, the gender-differentiated
analysis revealed that the effects of a share of WFH between 20–39 %,
40–59 % and 80–100 % differed signicantly between the genders, with
reductions in commuting time again signicantly larger for women. In
fact, for men, we observed that commuting times were slightly longer
among those WFH 20–39 % of their time compared to those not WFH at
all. As further analyses showed, this positive association was driven by a
few select industries that may potentially be characterised by a high
share of employees with intensive work-related travel (e.g. business
trips, on-site customer service). Overall, however, these results conrm
hypothesis H2b as well as ndings from the literature (e.g. Sullivan and
Lewis, 2001; Powel and Craig, 2015) suggesting that men are more
likely to use WFH to perform additional work, e.g. in the evenings or on
weekends, rather than to replace regular ofce hours with WFH. One
could also speculate about a possible selection effect in that a tempo-
rarily high career orientation leads men to commute long hours and to
additionally work at home in the evenings.
The results of FE models with impact functions (analysing
commuting trajectories over time) revealed that the reduction in
commuting time occurred immediately with the start of WFH. Moreover,
the time savings tended to increase over time (supporting H3). The latter
nding suggests little or no counteracting effects from potential changes
in residence and work location, and is in line with existing empirical
studies, which were, however, based on cross-sectional, retrospective or
intentional data (e.g. Muhammed et al., 2007; Ettema, 2010; Kim,
2016). Further analyses suggested that the observed commuting tra-
jectory was due to employees being able to expand their share of WFH
over time. This may be due to workers optimizing their work environ-
ment at home, becoming more competent in managing their work at
home, or employers becoming more trusting in workers performing
work at home and thus allowing them to spend a greater share of their
working time at home.
In terms of policy implications, to provide some – tentative – insights
into how much com-muting time may be saved in the future through
WFH, we combined our estimates with the ndings from recent surveys
among Australian workers on their stated preferences regarding the
share of WFH (Mattey et al., 2020; Beck and Hensher, 2021; Petrie,
2022). If future home workers behaved similarly to the home workers
that our FE regression analysis was based on, and the preferences
expressed in the surveys were fully accommodated, we would nd an
overall decline in commuting times compared to 2019 of 17 % to 25 %.
However, it should be noted that using workers’ stated preferences
means that the scenarios presented likely show the maximum potential
savings in weekly commuting time, since the realisation of these pref-
erences depends on many factors (e.g. Mokhtarian and Salomon, 1994).
Preferences may also change over time (Kogus et al., 2023), although a
recent study for Australia showed relatively stable preferences for WFH
between the COVID and the post-COVID period (Petrie, 2023). More-
over, the extent to which WFH actually reduces daily trafc also de-
pends on its effects on non-work-related trips (e.g., to go shopping) as
well as on the mobility behaviour of other household members (e.g.
Allen et al., 2015; Kim, 2016; Caldarola and Sorrell, 2022).
We acknowledge that our study has limitations. First, our study could
not account for the possibility that workers may also change their mode
of transport when starting to work from home. The literature suggests
that car use and the use of active transportation modes increase with
WFH (de Abreu e Silva and Melo, 2018a; Elld´
er, 2020). Assuming that
the use of a car increases commuting speed, our results might be over-
estimated, while in case of an increased use of active modes our results
might be underestimated. Second, since our data did not contain infor-
mation about commuting frequency, it was not possible to determine the
extent to which commute trips are actually saved through WFH or the
extent to which WFH is an incentive for longer daily commutes, i.e. the
extent to which commuting time per ofce day (or distance between
home and work) may increase when people start WFH. Future research
could quantify the relevance of this mechanism for Australia. Third,
although we used panel data and FE models, no conclusive statements
on causality were possible. Nevertheless, our results have demonstrated
the usefulness of such a methodological approach, and we therefore
recommend it for future research. Fourth, it is up to future research to
conrm that our ndings based on the pre-pandemic period hold true in
a post-pandemic world, in which WFH has become accessible to a much
larger proportion of the workforce. However, we gain some condence
in the future relevance of our ndings from the fact that persons who
started or extended WFH during the pandemic resemble those who were
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
9
already WFH in terms of socio-demographic characteristics (Reiffer
et al., 2022; Salon et al., 2022; Richards et al., 2024) and key motiva-
tions for WFH (Thompson et al., 2022). Moreover, Faber et al. (2023) in
their panel study using Dutch data from before and during COVID-19
found that the negative effect of WFH on commuting time was stable
over time when looking at the unstandardised coefcients. However, the
standardised effect was more negative during the pandemic than before
the pandemic, which the authors attributed to changes in the variances
of WFH and commute time. Furthermore, it can be argued that the
predictive power of COVID data for the future relationship between
WFH and commuting time may also be subject to limitations, as the
pandemic was an exceptional period characterised by severe travel re-
strictions, e.g. due to lockdowns and ofce closures (e.g. Jain et al.,
2022).
Overall, our ndings suggest that WFH has the potential for signi-
cant savings on commuting time, supporting the hopes expressed by
scholars that WFH can reduce congestion, pollution, and energy con-
sumption (e.g. Nilles et al., 1976; Harkness, 1977). Yet these savings are
largely contingent on the future share of workers using WFH exten-
sively, which was the case for only a minority before the pandemic.
More extensive shares of WFH have also been shown to be benecial for
family life by reducing work-family conict (Gajendran and Harrison,
2007; Laß and Wooden, 2023). By contrast, with respect to organiza-
tional outcomes, moderate rather than extensive WFH shares appear to
be most benecial (Allen et al., 2015). Moreover, high shares of WFH
have been associated with more health complaints (e.g. W¨
ohrmann and
Ebner, 2021). These mixed ndings call for further research on all
relevant outcomes at the individual, organisational, economic, envi-
ronmental, and societal level to identify the optimal extent of WFH.
CRediT authorship contribution statement
Heiko Rüger: Conceptualization, Methodology, Supervision,
Writing – original draft, Writing – review & editing. Inga Laß:
Conceptualization, Data curation, Methodology, Formal analysis,
Writing – original draft, Writing – review & editing. Nico Stawarz:
Conceptualization, Methodology, Writing – original draft, Writing –
review & editing. Alexandra Mergener: Conceptualization, Writing –
original draft, Writing – review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Acknowledgements
This paper uses unit record data from Release 19 of the Household,
Income and Labour Dynamics in Australia Survey conducted by the
Melbourne Institute of Applied Economic and Social Research on behalf
of the Australian Government Department of Social Services (DSS) (doi:
10.26193/3QRFMZ). The ndings and views reported in this paper,
however, are those of the authors and should not be attributed to the
Australian Government, DSS, or the Melbourne Institute.
This research did not receive any specic grant from funding
agencies in the public, commercial, or not-for-prot sectors.
Appendix
Table A1
Distribution of characteristics in the sample (in % unless stated otherwise).
All Men Women Works from home On-site only
Weekly commuting time (mean hrs) 4.42 4.87 3.98 4.92 4.31
Share of time worked from home (%)
None 82.22 82.70 81.76 0.00 100.00
1–19 11.42 12.15 10.70 64.25 0.00
20–39 4.30 3.54 5.04 24.19 0.00
40–59 0.98 0.78 1.18 5.53 0.00
60–79 0.34 0.32 0.36 1.91 0.00
80–100 0.73 0.50 0.96 4.12 0.00
Years in paid work (mean) 18.46 19.56 17.38 21.44 17.81
Partner in household 66.56 68.88 64.28 78.48 63.97
Partner’s weekly wage (mean in 100A$) 5.68 4.26 7.09 7.43 5.30
Children in household 44.53 44.23 44.81 55.86 42.07
Educational level
Postgraduate 6.08 5.98 6.18 14.08 4.34
Graduate diploma/certicate 7.17 5.68 8.63 15.39 5.38
Bachelor or honours 18.22 15.79 20.60 28.91 15.90
Advanced diploma, diploma 9.95 8.96 10.92 12.49 9.40
Certicate III or IV 22.90 28.38 17.53 13.00 25.05
Year 12 17.50 17.34 17.65 9.47 19.24
Year 11 and below 18.19 17.88 18.49 6.66 20.69
Remoteness area
Major cities 70.02 69.98 70.05 75.42 68.85
Inner regional 19.85 19.82 19.89 16.65 20.55
Outer regional, (very) remote 10.13 10.20 10.06 7.93 10.61
Working hours/week (mean) 37.01 41.38 32.71 42.08 35.91
Working days/week (mean) 4.68 4.92 4.44 4.85 4.64
Occupation
Managers 11.54 14.41 8.72 23.51 8.95
Professionals 25.55 21.27 29.75 51.58 19.90
Technicians and trades workers 12.52 21.16 4.02 4.77 14.20
Community and personal service workers 11.67 7.22 16.03 4.88 13.14
Clerical and administrative service workers 15.92 8.31 23.39 9.15 17.38
(continued on next page)
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
10
Table A1 (continued )
All Men Women Works from home On-site only
Sales workers 8.73 6.02 11.39 4.33 9.68
Machinery operators and drivers 6.08 11.20 1.05 0.77 7.24
Labourers 8.00 10.40 5.64 1.01 9.51
Industry
Agriculture, forestry, shing 1.30 1.96 0.66 0.89 1.39
Mining 2.01 3.43 0.60 1.48 2.12
Manufacturing 9.06 13.89 4.31 5.92 9.74
Electricity, gas, water, waste services 1.20 1.89 0.51 0.83 1.28
Construction 6.03 10.82 1.34 3.43 6.60
Wholesale trade 3.31 4.50 2.14 4.24 3.11
Retail trade 9.85 8.12 11.56 3.43 11.24
Accommodation, food services 5.75 4.67 6.81 1.49 6.67
Transport, postal, warehousing 4.37 6.79 2.00 2.30 4.82
Information media, telecommunications 2.16 2.31 2.02 3.13 1.95
Financial and insurance services 4.03 3.56 4.50 5.46 3.73
Rental, hiring, real estate services 1.34 1.12 1.56 2.10 1.17
Professional, scientic, technical services 6.86 7.21 6.52 11.15 5.93
Administrative and support services 2.45 2.11 2.78 1.75 2.60
Public administration and safety 8.39 9.86 6.95 6.66 8.77
Education and training 11.53 6.39 16.58 32.84 6.91
Health care and social assistance 15.53 5.70 25.20 9.11 16.93
Arts and recreation services 1.65 1.87 1.44 1.37 1.71
Other services 3.16 3.81 2.53 2.45 3.32
Firm size
Less than 20 employees 18.47 19.56 17.41 13.80 19.49
20–99 employees 14.72 15.47 13.99 13.81 14.92
100–499 employees 12.76 13.61 11.92 14.08 12.47
500 +employees 48.73 47.16 50.26 55.44 47.27
Unknown/implausible 5.32 4.19 6.43 2.87 5.85
Daily wage (mean in 10 A$) 21.45 24.57 18.38 28.27 19.97
Multiple job holder 8.06 6.77 9.32 7.95 8.08
n(observations) 112,197 55,603 56,594 19,995 92,202
Table A2
Distribution of characteristics in the sample (only workers who change working from home category; in % unless stated otherwise).
On-site only Works from home
1–19 % 20–59 % 60–100 %
Weekly commuting time (mean hrs) 4.72 5.11 5.13 2.52
Years in paid work (mean) 19.17 21.27 21.92 20.94
Partner in household 71.95 78.25 78.75 79.60
Partner’s weekly wage (mean in 100A$) 6.63 7.17 7.90 8.07
Children in household 48.92 54.60 60.51 56.02
Educational level
Postgraduate 8.16 12.62 18.05 13.76
Graduate diploma/certicate 9.56 14.43 18.96 10.68
Bachelor or honours 23.62 28.35 31.64 18.86
Advanced diploma, diploma 12.23 12.83 11.40 15.21
Certicate III or IV 20.03 14.73 8.85 15.11
Year 12 15.52 9.84 7.65 14.24
Year 11 and below 10.88 7.20 3.45 12.13
Remoteness area
Major cities 72.90 75.93 75.23 73.53
Inner regional 18.16 16.14 17.78 17.61
Outer regional, (very) remote 8.94 7.93 6.98 8.85
Working hours/week (mean) 37.76 43.29 41.26 32.24
Working days/week 4.70 4.90 4.76 4.70
Occupation
Managers 16.06 26.01 19.23 15.30
Professionals 32.18 47.19 64.50 42.25
Technicians and trades workers 10.15 5.82 2.32 4.52
Community and personal service workers 11.02 5.43 3.17 7.31
Clerical and administrative service workers 16.95 8.71 6.86 24.64
Sales workers 6.86 4.92 3.06 2.79
Machinery operators and drivers 3.10 0.94 0.33 0.38
Labourers 3.66 0.98 0.53 2.79
Industry
Agriculture, forestry, shing 1.06 0.68 0.53 4.04
Mining 1.80 1.91 0.84 0.48
Manufacturing 7.01 6.76 3.91 5.29
Electricity, gas, water, waste services 1.16 1.02 0.56 0.48
Construction 4.63 4.03 1.57 4.43
Wholesale trade 3.14 4.57 3.57 4.04
(continued on next page)
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
11
Table A2 (continued )
On-site only Works from home
1–19 % 20–59 % 60–100 %
Retail trade 7.92 4.22 1.78 2.31
Accommodation, food services 4.25 1.75 0.72 1.73
Transport, postal, warehousing 4.09 2.63 1.51 2.98
Information media, telecommunications 2.95 3.06 3.25 3.37
Financial and insurance services 5.39 5.18 6.46 6.35
Rental, hiring, real estate services 1.51 1.92 1.97 2.98
Professional, scientic, technical services 9.43 10.52 10.40 19.44
Administrative and support services 2.44 1.48 1.39 6.26
Public administration and safety 10.96 8.20 4.26 3.46
Education and training 11.69 27.92 47.25 17.81
Health care and social assistance 16.29 10.39 6.72 7.99
Arts and recreation services 1.79 1.55 0.95 1.06
Other services 2.49 2.22 2.37 5.49
Firm size
Less than 20 employees 15.52 11.88 10.92 37.44
20–99 employees 15.28 14.40 11.98 14.63
100–499 employees 14.20 15.14 13.05 11.26
500 +employees 51.16 56.02 60.69 34.26
Unknown/implausible 3.84 2.56 3.36 2.41
Daily wage (mean in 10 A$) 24.14 29.13 28.14 21.19
Multiple job holder 8.99 6.84 8.23 18.38
n(observations) 28,591 11,980 5685 1039
Table A3
Doing any work from home and (ln) weekly commuting time: Results from xed-effects regression (b-coefcients).
All base model All extended model Men
base model
Men extended model Women base model Women extended model
Works from home (ref. =no) −0.092
**
−0.149
**
−0.022 −0.063
**
−0.158
**
−0.229
**
Survey year (ref. =2002)
2003 0.022 0.010 0.044* 0.079
**
−0.001 −0.028
2004 0.055
**
0.032 0.066
**
0.134
**
0.044 −0.004
2005 0.086
**
0.055
+
0.073
**
0.178
**
0.098
**
0.031
2006 0.109
**
0.061 0.098
**
0.237
**
0.119
**
0.017
2007 0.144
**
0.086
+
0.151
**
0.316
**
0.135
**
0.023
2008 0.140
**
0.068 0.172
**
0.364
**
0.106
**
−0.025
2009 0.169
**
0.087 0.183
**
0.416
**
0.151
**
−0.005
2010 0.178
**
0.085 0.217
**
0.481
**
0.137
**
−0.038
2011 0.205
**
0.096 0.225
**
0.515
**
0.182
**
−0.014
2012 0.232
**
0.110 0.264
**
0.588
**
0.199
**
−0.026
2013 0.228
**
0.098 0.250
**
0.613
**
0.204
**
−0.044
2014 0.254
**
0.114 0.279
**
0.673
**
0.226
**
−0.035
2015 0.283
**
0.127 0.291
**
0.715
**
0.272
**
−0.016
2016 0.329
**
0.159 0.355
**
0.811
**
0.301
**
−0.014
2017 0.349
**
0.162 0.371
**
0.857
**
0.326
**
−0.022
2018 0.368
**
0.161 0.407
**
0.918
**
0.329
**
−0.047
2019 0.386
**
0.151 0.383
**
0.918
**
0.387
**
−0.033
Years in paid work 0.010 −0.030* 0.015
Years in paid work squared −0.000* −0.000
**
−0.000
Partner in household 0.070
**
0.068
**
0.045
+
Partner’s weekly wage (in 100A$) −0.002
+
0.001 −0.001
Children in household −0.008 0.047* −0.042
+
Educational level (ref. =Postgraduate)
Graduate diploma/certicate −0.011 −0.112 0.056
Bachelor or honours 0.047 −0.023 0.091
Advanced diploma, diploma −0.052 −0.122 −0.000
Certicate III or IV −0.159
**
−0.243
**
−0.107
Year 12 −0.112* −0.141* −0.092
Year 11 and below −0.332
**
−0.391
**
−0.286
**
Remoteness area (ref. =major cities)
Inner regional −0.279
**
−0.214
**
−0.345
**
Outer regional, (very) remote −0.499
**
−0.522
**
−0.473
**
Working hours/week 0.010
**
0.005
**
0.014
**
Working days/week 0.069
**
0.055
**
0.066
**
Occupation (ref. =managers)
Professionals 0.049
**
0.051* 0.043
Technicians and trades workers 0.011 0.025 −0.026
Community and personal service workers −0.032 −0.051 −0.028
Clerical and administrative service workers 0.025 0.054* 0.010
Sales workers −0.081
**
−0.087* −0.072*
Machinery operators and drivers −0.056
+
−0.055 −0.049
Labourers −0.059* −0.066* −0.052
(continued on next page)
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
12
Table A3 (continued )
All base model All extended model Men
base model
Men extended model Women base model Women extended model
Industry (ref. =agriculture, forestry, shing)
Mining 0.087 0.119 0.015
Manufacturing −0.029 −0.034 −0.044
Electricity, gas, water, waste services 0.119 0.046 0.272
Construction 0.117
+
0.131
+
−0.016
Wholesale trade 0.011 −0.006 0.015
Retail trade −0.150* −0.163* −0.152
Accommodation, food services −0.210
**
−0.305
**
−0.161
Transport, postal, warehousing −0.023 −0.024 −0.065
Information media, telecommunications 0.073 0.074 0.047
Financial and insurance services 0.158* 0.119 0.163
Rental, hiring, real estate services 0.017 −0.022 0.044
Professional, scientic, technical services 0.115
+
0.080 0.137
Administrative and support services 0.085 0.079 0.077
Public administration and safety 0.104 0.089 0.104
Education and training 0.037 −0.080 0.107
Health care and social assistance 0.040 −0.079 0.089
Arts and recreation services 0.025 −0.012 0.042
Other services 0.013 0.003 0.017
Firm size (ref. =less than 20 employees)
20–99 employees 0.079
**
0.067
**
0.090
**
100–499 employees 0.151
**
0.137
**
0.167
**
500 +employees 0.154
**
0.139
**
0.169
**
Unknown/implausible 0.136
**
0.101
**
0.163
**
Daily wage (in 10 A$) 0.004
**
0.003
**
0.005
**
Multiple job holder 0.212
**
0.135
**
0.267
**
Constant 0.820
**
0.084 0.928
**
0.903
**
0.716
**
−0.151
n(observations) 112,197 112,197 55,603 55,603 56,594 56,594
adj. within-R
2
0.012 0.060 0.014 0.053 0.011 0.073
Note: **, * and
+
denote statistical signicance at the 0.01, 0.05 and 0.10 levels, respectively.
Table A4
Time worked from home and (ln) weekly commuting time: Results from xed-effects and pooled OLS regression (b-coefcients).
All extended model Men extended model Women extended model
Excluding multiple job holders (FE regression)
Share of time worked from home (%, ref. =none)
1–19 0.004 0.015 −0.007
20–39 0.014 0.052* −0.011
40–59 −0.064
+
0.037 −0.119
**
60–79 −0.430
**
−0.440
**
−0.358
**
80–100 −4.250
**
−3.535
**
−4.647
**
n(observations) 103,103 51,819 51,284
adj. within-R
2
0.152 0.106 0.197
Excluding specic industries (FE regression)
Share of time worked from home (%, ref. =none)
1–19 −0.003 −0.000 −0.007
20–39 −0.004 0.016 −0.014
40–59 −0.067
+
0.006 −0.106*
60–79 −0.397
**
−0.424
**
−0.342
**
80–100 −3.765
**
−3.390
**
−3.954
**
n(observations) 95,626 47,962 47,664
adj. within-R
2
0.140 0.098 0.179
Changes within same job (FE regression)
Share of time worked from home (%, ref. =none)
1–19 −0.009 −0.004 −0.013
20–39 −0.016 0.013 −0.035
40–59 −0.109** −0.044 −0.150**
60–79 −0.391** −0.405** −0.355**
80–100 −3.389** −2.931** −3.647**
n(observations) 112,134 55,580 56,554
adj. within-R
2
0.104 0.066 0.140
Pooled OLS regression
Share of time worked from home (%, ref. =none)
1–19 0.008 0.021 0.000
20–39 0.047* 0.089
**
0.026
40–59 0.034 0.094
+
0.010
60–79 −0.189
**
−0.214* −0.168*
80–100 −4.376
**
−3.782
**
−4.644
**
n(observations) 47,276 22,755 24,521
adj. R
2
0.303 0.238 0.335
Hours worked from home (FE regression)
Number of hours usually worked from home per week (ref. =none)
(continued on next page)
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
13
Table A4 (continued )
All extended model Men extended model Women extended model
1–7 −0.020
+
0.005 −0.041*
8–15 −0.163
**
−0.007 −0.299
**
16–23 −0.490
**
−0.139* −0.773
**
24–31 −1.167
**
−0.456
**
−1.772
**
32–39 −3.153
**
−2.190
**
−4.070
**
>=40 −3.532
**
−2.970
**
−4.207
**
n(observations) 112,134 55,580 56,554
adj. within-R
2
0.103 0.085 0.131
Notes: **, * and
+
denote statistical signicance at the 0.01, 0.05 and 0.10 levels, respectively. Control variables as in the extended models in Table A3, pooled OLS
models additional account for country of birth/origin.
Table A5
Time worked from home and (ln) daily commuting time: Results from xed-effects regression (b-coefcients).
All extended model Men extended model Women extended model
Share of time worked from home (%, ref. =none)
1–19 0.013 0.023 0.001
20–39 0.029 0.060* 0.010
40–59 −0.061+0.044 −0.125**
60–79 −0.404** −0.357** −0.411**
80–100 −3.924** −3.377** −4.202**
n(observations) 112,134 55,580 56,554
adj. within-R
2
0.129 0.093 0.162
Notes: **, * and
+
denote statistical signicance at the 0.01, 0.05 and 0.10 levels, respectively. Control variables as in the extended models in Table A3, but with the
number of working days per week omitted.
References
Abraham, M., Nisic, N., 2007. Regionale Bindung, r¨
aumliche Mobilit¨
at und Arbeitsmarkt
- Analysen für die Schweiz und Deutschland. Swiss Journal of Sociology 33, 69–87.
Allen, T.D., Golden, T.D., Shockley, K.M., 2015. How Effective Is Telecommuting?
Assessing the Status of Our Scientic Findings. Psychol. Sci. Public Interest 16,
40–68.
Altintas, E., Sullivan, O., 2016. Fifty years of change updated: Cross-national gender
convergence in housework. Demogr. Res. 35, 455–470.
Andreev, P., Salomon, I., Pliskin, N., 2010. State of teleactivities. Transp. Res. C 18, 3–10.
Asgari, H., Jin, X., 2018. An evaluation of part-day telecommute impacts on work trip
departure times. Travel Behav. Soc. 12, 84–92.
Australian Bureau of Statistics, 2021. Wage Price Index, Australia, Catalogue no. 6345.0,
Table 1. https://www.abs.gov.au/statistics/economy/price-indexes-and-ination/
wage-price-index-australia/latest-release#data-download.
Baruch, Y., 2000. Teleworking: benets and pitfalls as perceived by professionals and
managers. N. Technol. Work. Employ. 15, 34–49.
Beauregard, T.A., Basile, K.A., Canonico, E., 2019. Telework: Outcomes and Facilitators
for Employees. In: Landers, R.N. (Ed.), The Cambridge Handbook of Technology and
Employee Behavior. Cambridge University Press, Cambridge, UK, pp. 511–543.
Beck, M.J., Hensher, D.A., 2021. Insights into Working from Home in Australia in 2020:
Positives, Negatives and the Potential for Future Benets to Transport and Society.
Working Paper ITLS.
Botha, F., Kabatek, J., Meekes, J., Wilkins, R., 2023. The effects of commuting and
working from home arrangements on mental health. SSRN Journal. https://doi.org/
10.2139/ssrn.4642013.
Brüderl, J., Ludwig, V., 2015. Fixed-effects panel regression. In: Best, H., Wolf, C. (Eds.),
The SAGE Handbook of Regression Analysis and Causal Inference. Sage, Los Angeles,
CA, pp. 327–357.
Buomprisco, G., Ricci, S., Perri, R., de Sio, S., 2021. Health and Telework: New
Challenges after COVID-19 Pandemic. European Journal of Environment and Public
Health 5, em0073.
Bureau of Infrastructure, Transport and Regional Economics, 2016. Lengthy commutes in
Australia. Report 144, Canberra ACT.
Caldarola, B., Sorrell, S., 2022. Do teleworkers travel less? Evidence from the English
National Travel Survey. Transp. Res. A Policy Pract. 159, 282–303.
Carillo, K., Cachat-Rosset, G., Marsan, J., Saba, T., Klarsfeld, A., 2021. Adjusting to
epidemic-induced telework: empirical insights from teleworkers in France. Eur. J.
Inf. Syst. 30, 69–88.
Choo, S., Mokhtarian, P.L., Salomon, I., 2005. Does telecommuting reduce vehicle-miles
traveled? An aggregate time series analysis for the U.S. Transportation 32, 37–64.
Crane, R., 2007. Is there a quiet revolution in women’s travel? revisiting the gender gap
in commuting. J. Am. Plann. Assoc. 73, 298–316.
de Abreu e Silva, J., Melo, P., 2018a. Does home-based telework reduce household total
travel? A path analysis using single and two worker British households. J. Transp.
Geogr. 73, 148–162.
de Abreu e Silva, J., Melo, P., 2018b. Home telework, travel behavior, and land-use
patterns: A path analysis of British single-worker households. The Journal of
Transport and Land Use 11, 419–441.
de Graaff, T., 2004. On the Substitution and Complementarity between Telework and
Travel: A Review and Application.Serie Research Memoranda 0016, VU University
Amsterdam, Faculty of Economics, Business Administration and Econometrics.
de Vos, D., van Ham, M., Meijers, E., 2019. Working from Home and Commuting:
Heterogeneity Over Time, Space, and Occupations. IZA Discussion Paper 12578.
de Vos, D., Meijers, E., van Ham, M., 2018. Working from home and the willingness to
accept a longer commute. Ann. Reg. Sci. 61, 375–398.
Department of Social Services, Melbourne Institute of Applied Economic and Social
Research, 2020. The Household, Income and Labour Dynamics in Australia (HILDA)
Survey, GENERAL RELEASE 19 (Waves 1-19). https://doi.org/10.26193/3QRFMZ.
Devine, T.J., Kiefer, N.M., 1993. The empirical status of job search theory. Labour Econ.
1, 3–24.
DLR, 2021. Vierte DLR-Erhebung zu Mobilit¨
at & Corona: Hintergrundpapier. DLR. http
s://verkehrsforschung.dlr.de/public/documents/2021/DLR-Befragung4_Hinte
rgrundpapier.pdf.
Elld´
er, E., 2020. Telework and daily travel: New evidence from Sweden. J. Transp.
Geogr. 86, 102777.
Ettema, D., 2010. The impact of telecommuting on residential relocation and residential
preferences. A latent class modeling approach. Journal of Transport and Land Use 3,
7–24.
Eurofound, the International Labour Ofce, 2017. Working anytime, anywhere: the
effects on the world of work. Publications Ofce of the EU and the ILO, Luxembourg
and Geneva.
Eurostat,, 2021. Passenger Mobility Statistics. Eurostat. https://ec.europa.eu/eurostat/st
atistics-explai-ned/index.php?title=Passenger_mobility_statistics#Mobility_data_for
_thirteen_Member_States_with_different_characteristics.
Faber, R.M., Hamersma, M., Brimaire, J., Kroesen, M., Molin, E.J., 2023. The relations
between working from home and travel behaviour: a panel analysis. Transportation.
https://doi.org/10.1007/s11116-023-10401-4.
Gajendran, R.S., Harrison, D.A., 2007. The good, the bad, and the unknown about
telecommuting: meta-analysis of psychological mediators and individual
consequences. J. Appl. Psychol. 97, 1524–1541.
Gimenez-Nadal, J.I., Molina, J.A., 2016. Commuting time and household responsibilities:
Evidence using propensity score matching. J. Reg. Sci. 56, 332–359.
Glass, J.L., Noonan, M.C., 2016. Telecommuting and earnings trajectories among
american women and men 1989–2008. Soc. Forces 95, 217–250.
Glogger, A.F., Z¨
angler, T.W., Karg, G., 2008. The Impact of Telecommuting on
Households’ Travel Behaviour, Expenditures and Emissions. In: Jensen-Butler, C.,
Sloth, B., Larsen, M.M., Madsen, B., Nielsen, O.A. (Eds.), Road Pricing, the Economy
and the Environment. Springer, Berlin/Heidelberg, pp. 411–425.
Golden, T.D., 2006. The role of relationships in understanding telecommuter satisfaction.
J. Organ. Behav. 27, 319–340.
Gschwind, L., Vargas, O., 2019. Telework and its effects in Europe. In: Messenger, J.C.
(Ed.), Telework in the 21st Century. An Evolutionary Perspective. The ILO Future of
Work series, Edward Elgar, Cheltenham/Northhampton, pp. 36–75.
Harkness, R.C., 1977. Selected results from a technology assessment of
telecommunication-transportation interactions. Habitat Int. 2, 37–48.
Helminen, V., Ristim¨
aki, M., 2007. Relationships between commuting distance,
frequency and telework in Finland. J. Transp. Geogr. 15, 331–342.
Hill, T.D., Davis, A.P., Roos, J.M., French, M.T., 2020. Limitations of xed-effects models
for panel data. Sociol. Perspect. 63, 357–369.
H. Rüger et al.
Travel Behaviour and Society 37 (2024) 100839
14
Huang, Z., Loo, B.P.Y., Axhausen, K.W., 2023. Travel behaviour changes under Work-
from-home (WFH) arrangements during COVID-19. Travel Behav. Soc. 30, 202–211.
IEA, 2021. Global CO2 emissions in transport by mode in the Sustainable Development
Scenario, 2000-2070. https://www.iea.org/data-and-statistics/charts/global-co2-
emissions-in-transport-by-mode-in-the-sustainable-development-scenario-2000-
2070.
Jain, T., Currie, G., Aston, L., 2022. COVID and working from home: Long-term impacts
and psycho-social determinants. Transp. Res. A Policy Pract. 156, 52–68.
Jiang, Y., 2008. The Impact of Telecommuting on the Journey to Work: A Two-Sample
Instrumental Variables Approach. University of Maryland.
Kalter, F., 1994. Pendeln statt Migration? Die Wahl und Stabilit¨
at von Wohn-Arbeitsort-
Kombinationen. ZFS 23, 460–476.
Kaplan, S., Engelsted, L., Lei, X., Lockwood, K., 2018. Unpackaging manager mistrust in
allowing telework: comparing and integrating theoretical perspectives. J. Bus.
Psychol. 33, 365–382.
Kim, S.-N., 2016. Two traditional questions on the relationships between telecommuting,
job and residential location, and household travel: Revisited using a path analysis.
Ann. Reg. Sci. 56, 537–563.
Kim, T., Mullins, L.B., Yoon, T., 2021. Supervision of telework: A key to organizational
performance. Am. Rev. Public Adm. 51, 263–277.
Kogus, A., Gal-Tzur, A., Shiftan, Y., 2023. Modelling the Long-Term Expected Impact of
the Covid-19 Crisis on Commute and Telecommute. Transportation Research Record:
Journal of the Transportation Research Board 036119812311601.
Lachapelle, U., Tanguay, G.A., Neumark-Gaudet, L., 2018. Telecommuting and
sustainable travel: Reduction of overall travel time, increases in non-motorised
travel and congestion relief? Urban Stud. 55, 2226–2244.
Laß, I., Skora, T., Rüger, H., Wooden, M., Bujard, M., 2023. Does temporary employment
increase length of commuting? Longitudinal evidence from Australia and Germany.
Transportation. https://doi.org/10.1007/s11116-023-10374-4.
Laß, I., Wooden, M., 2023. Working from Home and Work-Family Conict. Work Employ
Soc. 37, 176–195.
Ludwig, V., Brüderl, J., 2021. What You Need to Know When Estimating Impact
Functions with Panel Data for Demographic Research. Comp. Popul. Stud. 46,
453–486.
Manning, A., 2003. The real thin theory: monopsony in modern labour markets. Labour
Econ. 10, 105–131.
Mattey, C., Hilberath, C., Sibilio, N., 2020. Personalisation for Your People. How COVID
is Reshaping the Race for Talent, Boston Consulting Group.
McGuckin, N., Zmud, J., Nakamoto, Y., 2005. Trip-chaining trends in the united states:
understanding travel behavior for policy making. Transp. Res. Rec. 1917, 199–204.
Melo, P., de Abreu e Silva, J., 2017. Home telework and household commuting patterns
in Great Britain. Transp. Res. A 103, 1–24.
Mergener, A., Mansfeld, L., 2021. Being spatially mobile without daily commuting? How
Working from Home patterns relate to company-home distances, BIBB, Bonn.
Messenger, J.C., 2019. Conclusions and recommendations for policy and practice. In:
Messenger, J.C. (Ed.), Telework in the 21st Century. An Evolutionary Perspective.
The ILO Future of Work series, Edward Elgar, Cheltenham/Northhampton,
pp. 286–316.
Mokhtarian, P.L., Collantes, G.O., Gertz, C., 2004. Telecommuting, residential location,
and commute-distance traveled: Evidence from State of California employees.
Environ Plan A 36, 1877–1897.
Mokhtarian, P.L., Salomon, I., 1994. Modeling the Choice of Telecommuting: Setting the
Context. Environment and Planning A: Economy and Space 26, 749–766.
Motte-Baumvol, B., Schwanen, T., 2024. Telework, travel times, and peak hour
avoidance in England: An overview using travel times across ve weekdays. Travel
Behav. Soc. 34, 100668.
Muhammed, S., Ottens, H.F., de Jong, T., 2007. Telecommuting and residential
locational preferences: a case study of the Netherlands. J. Hous. Built Environ. 22,
339–358.
Nilles, J.M., 1988. Trafc reduction by telecommuting: A status review and selected
bibliography. Transp. Res. A 22, 307–317.
Nilles, J.M., Carlson, F.R., Gray, P., Hanneman, G.J., 1976. The Telecommunications-
Transportation Tradeoff: Options for tomorrow. Wiley, New York.
Nisic, N., 2017. Smaller Differences in Bigger Cities? Assessing the Regional Dimension
of the Gender Wage Gap. Eur. Sociol. Rev. 33, 292–304.
OECD, 2021. OECD Employment Outlook 2021: Navigating the COVID-19 Crisis and
Recovery. OECD Publishing, Paris.
Ory, D.T., Mokhtarian, P.L., 2005. An Empirical Analysis of Causality in the Relationship
Between Telecommuting and Residential and Job Relocation. Institute of
Transportation Studies, UC Davis.
Ory, D.T., Mokhtarian, P.L., 2006. Which came rst, the telecommuting or the residential
relocation? an empirical analysis of causality. Urban Geogr. 27, 590–609.
Pabilonia, S.W., Vernon, V., 2022. Telework, wages, and time use in the United States.
Rev. Econ. Househ. https://doi.org/10.1007/s11150-022-09601-1.
Pailh´
e, A., Solaz, A., Stanfors, M., 2021. The great convergence: gender and unpaid work
in Europe and the United States. Popul. Dev. Rev. 47, 181–217.
Petrie, R., 2022. Workers prefer a balance of home and ofce, but many employers have
different expectations. https://melbourneinstitute.unimelb.edu.au/data/taking-the-
pulse-of-the-nation-2022/wave-48-49.
Petrie, R., 2023. Ongoing work-from-home negotiation converges (somewhat). https://
melbourneinstitute.unimelb.edu.au/data/taking-the-pulse-of-the-nation-2022/
2023/ttpn-july-2023.
Powel, A., Craig, L., 2015. Gender differences in working at home and time use patterns.
Work Employ Soc. 29, 571–589.
Ravalet, E., R´
erat, P., 2019. Teleworking: Decreasing mobility or increasing tolerance of
commuting distances? Built Environ. 45, 582–602.
Reiffer, A., Magdolen, M., Ecke, L., Vortisch, P., 2022. Effects of COVID-19 on Telework
and Commuting Behavior: Evidence from 3 Years of Panel Data. Transp. Res. Rec.
036119812210899.
Richards, A., Convery, S., O’Mahony, M., Cauleld, B., 2024. Pre and post Covid
preferences for working from home. Travel Behav. Soc. 34, 100679.
Rouwendal, J., Nijkamp, P., 2004. Living in two worlds: A review of home-to-work
decisions. Growth Chang. 35, 287–303.
Rouwendal, J., Rietveld, P., 1994. Changes in commuting distances of dutch households.
Urban Stud. 31, 1545–1557.
Rüger, H., Pfaff, S., Weishaar, H., Wiernik, B.M., 2017. Does perceived stress mediate the
relationship between commuting and health-related quality of life? Transport. Res.
F: Trafc Psychol. Behav. 50, 100–108.
Salomon, I., 1985. Telecommunications and travel: substitution or modied mobility?
JTEP 19, 219–235.
Salon, D., Mirtich, L., Bhagat-Conway, M.W., Costello, A., Rahimi, E., Mohammadian, A.
K., Chauhan, R.S., Derrible, S., Da Silva Baker, D., Pendyala, R.M., 2022. The COVID-
19 pandemic and the future of telecommuting in the United States. Transp. Res. Part
D: Transp. Environ. 112, 103473.
Scott, D.M., Dam, I., P´
aez, A., Wilton, R.D., 2012. Investigating the effects of social
inuence on the choice to telework. Environment and Planning A: Economy and
Space 44, 1016–1031.
Sjaastad, L.A., 1962. The costs and returns of human migration. J. Polit. Econ. 70, 80–93.
Skora, T., Rüger, H., Stawarz, N., 2020. Commuting and the motherhood wage gap:
Evidence from Germany. Sustainability 12, 5692.
So, K.S., Orazem, P.F., Otto, D.M., 2001. The effects of housing prices, wages and
commuting time on joint residential job location choices. Am. J. Agric. Econ. 83,
1036–1048.
Sostero, M., Milasi, S., Hurely, J., Fern´
andez-Macías, E., Bisello, M., 2020.
Teleworkability and the COVID-19 crisis: a new digital divide? JRC Working Papers
Series on Labour, Education and Technology No. 2020/05.
Sullivan, C., Lewis, S., 2001. Home-based telework, gender, and the synchronization of
work and family: perspectives of teleworkers and their co-residents. Gend. Work.
Organ. 8, 123–145.
Thompson, R.J., Payne, S.C., Alexander, A.L., Gaskins, V.A., Henning, J.B., 2022.
A taxonomy of employee motives for telework. Occupational Health Science 6,
149–178.
Turner, T., Niemeier, D., 1997. Travel to work and household responsibility: New
evidence. Transportation 24, 397–419.
U. S. Department of Transportation, 2017. Summary of Travel Trends. 2017 National
Household Travel Survey. https://nhts.ornl.gov/assets/2017_nhts_summary_travel_
trends.pdf.
Van den Berg, G.J., Gorter, C., 1997. Job search and commuting time. J. Bus. Econ. Stat.
15, 269–281.
van Ommeren, J.N., Fosgerau, M., 2009. Workers’ marginal costs of commuting.
J. Urban Econ. 65, 38–47.
van Ommeren, J.N., Rietveld, P., Nijkamp, P., 1999. Job Moving, residential moving, and
commuting: a search perspective. J. Urban Econ. 46, 230–253.
Wang, X., Kim, S.H., Mokhtarian, P.L., 2023. Teleworking behavior pre-, during, and
expected post-COVID: Identication and empirical description of trajectory types.
Travel Behav. Soc. 33, 100628.
Watson, N., Wooden, M., 2021. The household, income and labour dynamics in Australia
(HILDA) survey. Journal of Economics and Statistics 241, 131–141.
Wilkins, R., Laß, I., Butterworth, P., Vera-Toscano, E., 2019. The Household, Income and
Labour Dynamics in Australia Survey: Selected Findings from Waves 1 to 17. Applied
Economic & Social Research, University of Melbourne, Melbourne Institute.
W¨
ohrmann, A.M., Ebner, C., 2021. Understanding the bright side and the dark side of
telework: An empirical analysis of working conditions and psychosomatic health
complaints. New Technology, Work and Employment, pp. 1–23.
Zhu, P., 2012. Are telecommuting and personal travel complements or substitutes? Ann.
Reg. Sci. 48, 619–639.
Zhu, P., 2013. Telecommuting, Household Commute and Location Choice. Urban Stud.
50, 2441–2459.
Zhu, P., Mason, S.G., 2014. The impact of telecommuting on personal vehicle usage and
environmental sustainability. Int. J. Environ. Sci. Technol. 11, 2185–2200.
H. Rüger et al.
