Content uploaded by Xia Jin
Author content
All content in this area was uploaded by Xia Jin on Nov 09, 2019
Content may be subject to copyright.
83
Transportation Research Record: Journal of the Transportation Research Board,
No. 2566, Transportation Research Board, Washington, D.C., 2016, pp. 83–92.
DOI: 10.3141/2566-09
How and to what extent telecommuting engagement affects time alloca-
tion among nonmandatory activities are examined to help understand the
impacts of telecommuting on daily activity–travel patterns. Five catego-
ries of nonmandatory activities are considered: shopping, maintenance,
discretionary, escort, and in-home shopping. The hypothesis is that tele-
commuting relaxes the temporal and spatial constraints related to work
activities at the regular workplace, and telecommuters may allocate some
of the time budget to other nonmandatory activities, which may or may
not lead to additional travel. The structural equations model approach is
applied to capture the impacts of telecommuting as well as the interactions
among the nonmandatory activities. The activity durations by type along
with the number of total daily trips are considered as endogenous (depen-
dent) variables. By incorporating work hours at the regular workplace
and daily telecommuting hours as exogenous variables, the models can
reveal how people may reallocate their time among different nonmanda-
tory activities given different levels of telecommuting engagement (either
part-day or full-day). All types of telecommuting arrangements increased
nonmandatory activity durations (compared with those of nontelecom-
muters). Full-day telecommuters have higher durations of discretionary
activities, while part-day telecommuters have higher durations of mainte-
nance and out-of-home shopping errands. Telecommuting also increased
total daily trip rates for both telecommuters and their household members.
This study used data obtained from the 2010–2011 Regional Household
Travel Survey in the New York metropolitan region.
With recent advances in computer and telecommunication technolo-
gies, workers have benefited from more flexible work arrangements,
such as working from home, known as telecommuting. A recent
report by the U.S. Office of Personnel Management indicated that
a noticeable portion of U.S. workers (approximately 14%) telecom-
mute, either full day or part day (1). From a transportation perspec-
tive, telecommuting can be viewed as a demand management tool that
could help alleviate traffic congestion, particularly during peak hours.
Increasing attention has been given to the impacts of telecommuting
in the past three decades.
A substantial body of literature has examined the overall impacts
of telecommuting on the transportation network (2–7). There is
a consensus that telecommuting benefits the traffic network in a
variety of ways: by alleviating congestion, reducing vehicle miles
traveled and total delay, improving safety, and affecting the envi-
ronment in positive ways such as reductions in vehicle emissions.
Some researchers indicated that telecommuting impacts are not
limited to commute reduction, and relief from commute travel may
result in other changes in travel patterns (8–10). Several research
efforts have compared the trip-making behavior of regular workers
and telecommuters, with a focus on travel dimensions such as com-
mute distance, time–space distribution of trips, residential relocation,
and mode choice (9, 11–15).
While the aforementioned studies opened new horizons in analy-
sis of telecommuting impacts, they focused mainly on trip-based
investigations. In this paper, the impacts of telecommuting on the
individual’s activity patterns are examined. Since trips are derived
from the need to accomplish certain activities in certain locations,
such activity-level analysis is expected to provide a more compre-
hensive and accurate behavioral framework for telecommuting
impact evaluation. For example, telecommuting may relax work-
related time–space constraints through commute replacement or
displacement and may provide workers with more freedom in their
daily activity–travel decision making. Such freedom may increase
an individual’s chances of participating in other nonmandatory activi-
ties and lead to additional trips. Telecommuting may even affect
other household members by shifting some duties between members
in the same household, which may result in changes in their travel
behavior. This activity-based approach provides a better platform
for understanding the impacts of telecommuting on people’s daily
lives. It could help in answering questions such as the following:
Would telecommuters spend more time at home or allocate the time
savings from commute travel to other activities, or would telecom-
muters take on more household responsibilities, such as escorting
children, and therefore allow other household members more time
for other activities? A number of studies have examined the impacts
of telecommunications such as cell phone and Internet usage (not
telecommuting) on activity duration and trip generation (16–19); all
these studies are from other countries.
The major objective of this paper is to investigate the impacts of
telecommuting on individuals’ daily activity scheduling behavior.
The focus is on time use among nonmandatory activities. Since
activity scheduling could be affected by interactions among house-
hold members, the analysis targets both telecommuters and their
household members. Activity durations and total daily trip rates are
compared for commuters and telecommuters (at different levels of
telecommuting engagement) and for nonworkers with and with-
out telecommuters in the household. The intent is to indicate how
and to what extent telecommuting may affect individuals and their
household members in terms of time use patterns.
Examination of the Impacts of
Telecommuting on the Time Use
of Nonmandatory Activities
Hamidreza Asgari, Xia Jin, and Yiman Du
H. Asgari, EC 3725, and X. Jin, EC 3603, Department of Civil and Environmental
Engineering, Florida International University, 10555 West Flagler Street, Miami,
FL 33174. Y. Du, Department of Civil Engineering, Tsinghua University, Haidian
District, Beijing 100084, China. Corresponding author: X. Jin, xjin1@fiu.edu.
84 Transportation Research Record 2566
TERMS AND DEFINITIONS
One of the major assumptions in this study is that telecommuters are
not a homogeneous group and that different forms of telecommut-
ing engagement will have different impacts on daily activity–travel
patterns. A worker who adopts telecommuting may not telecommute
every workday, and telecommuters may not all telecommute on a full-
day basis. The presence of day-to-day (full-time versus part-time) and
within-day (full-day versus part-day) variations in telecommuting
engagement is well recognized in the literature (12, 20–23).
In this study, three major dimensions are taken into account to
categorize different forms of telecommuting activity (24, 25):
• Choice, which is an index of telecommuting regularity for any
observed worker;
• Daily engagement, which defines whether a worker participates
in telecommuting on a specific day; and
• Daily commute, which denotes whether telecommuting replaces
the daily commute.
Figure 1 indicates the different levels of telecommuting engagement
on the basis of the three dimensions.
Full-day regular telecommuters are considered as one group, pri-
mary telecommuters, regardless of their telecommuting regularity.
Part-day regular telecommuters are referred to as ancillary telecom-
muters, in the sense that telecommuting does not replace the commute
but rather serves an ancillary role. Part-day nonregular telecommuters
are referred to as passive telecommuters, since they telecommute on an
occasional basis, and the telecommuting captured on this specific day
is not the usual work arrangement and probably plays a passive role.
DATA PREPARATION AND
DESCRIPTIVE STATISTICS
This study uses data obtained from the 2010–2011 Regional House-
hold Travel Survey (RHTS). The 2010–2011 RHTS is a comprehen-
sive study of the demographic and travel behavior characteristics
of residents within 28 counties of New York, New Jersey, and
Connecticut (26).
The cleaned data set includes 12,593 workers and 19,431 non-
workers, and a total of 29 activity types were captured. The activity
types were aggregated into five major nonmandatory activities:
out-of-home shopping, out-of-home maintenance, out-of-home dis-
cretionary, escort, and in-home shopping. Although other in-home
activities were initially considered as part of the analysis, they were
later removed because of remarkably longer durations (and signifi-
cantly larger variance) than other activities, which resulted in inconsis-
tencies in the model estimation. Furthermore, other in-home activities
were considered mainly to include activities such as sleeping or
in-home meals, which are essential and therefore not influenced by
telecommuting activities.
Nonmandatory errands are the main targets of this study. Out-of-
home work (regular work) and telecommuting are considered as two
major mandatory activities that are expected to impose time–space
restrictions on individuals’ daily activity plans. Regular work dura-
tions were obtained directly from the survey data. Telecommute dura-
tions were derived with extra care. One major adjustment was applied
when telecommuting took place during the early morning period.
According to the data dictionary rules, the start point of the day was
fixed at 3:00 a.m. Thus, any telecommuting activity (before depart-
ing from home) was automatically computed from 3:00 a.m. As
a result, total work durations for telecommuters were unreasonably
higher than were those for commuters. To resolve this issue, the earli-
est telecommuting start time was heuristically shifted to 6:00 a.m.,
which was expected to ameliorate some of the unreasonably long
durations. In addition, when telecommuting is reported as a second-
ary purpose, there is no tool for quantifying its duration clearly. All
secondary telecommuting purposes (392 observations, almost 13% of
the telecommuters’ sample) were removed for this time use analysis.
The adjusted work durations are shown in Figure 2.
Figure 2 shows the total daily work durations decomposed into reg-
ular work and telecommute durations for different work arrangement
forms. Results indicate that among the four work arrangements, non-
telecommuters showed the lowest total work duration, with an aver-
FIGURE 1 Identifying different telecommuting patterns.
Asgari, Jin, and Du 85
age of 485 min/day, which is well within the range of the expected
8-h schedule. Primary telecommuters reported an approximate daily
duration of 10 h. This appears reasonable if the fact that in-home work
is often accompanied by other activities in parallel, such as meals or
other personal activities, is taken into account.
However, part-day telecommuting reflects significantly higher
work hours than the others. This may indicate that part-day telecom-
muting is usually accompanied by prolonged work hours for one job
or perhaps multiple jobs. As expected, both part-day forms tend to
reduce regular work durations. On average, the observed reduction
in regular work hours varies from approximately 60 min for passive
telecommuters to 120 min for ancillary telecommuters. The difference
observed in the reduction in regular work hours between ancillary
and passive telecommuters indicates the necessity of differentiating
the two groups. The fact that ancillary telecommuters spend less
time at the regular workplace may arise from the regularity of their
behavior. They have the option of avoiding peak hour congestion by
shifting their commute and higher probabilities of exercising it. A
complete time-of-day analysis of commute departure time to inves-
tigate the impacts of telecommuting on commute displacement is
presented in a separate paper.
Figure 3 shows the durations for the five nonmandatory activities
by telecommuting form. Primary telecommuters showed higher
durations of out-of-home nonmandatory activities than did other
workers. This indicates that the removal of commute travel may lead to
additional travel. Part-day telecommuters (either ancillary or passive)
showed lower durations of shopping, maintenance, and discretionary
activities than did commuters (nontelecommuters). The reason may
be that part-day telecommuting sometimes involves longer working
hours (as shown in Figure 2), which decreases the time budget for
participation in nonmandatory activities.
To validate the observed differences between the work arrange-
ments in terms of statistical significance, pairwise t-test comparisons
were conducted (Table 1).
The computed t-values confirm significant differences between pri-
mary telecommuters and the other three categories in almost all cases.
Ancillary and passive telecommuters reflect similar durations, with
t-values lower than 1.64 at a 90% confidence level. The only statistical
difference between the two is for discretionary activities, for which
ancillary telecommuters show significantly higher average durations.
Compared with nontelecommuters, ancillary arrangements do not
show any significant differences, while passive telecommuters have
significantly lower durations of discretionary errands. In contrast, in-
home shopping activities do not show significant differences among
any categories, which may be due to their relatively short durations.
Table 1 also compares average daily trip rates. Telecommuters
(regardless of their telecommuting form) tend to show higher daily trip
rates than do nontelecommuters. This may confirm the hypothesis
0
100
200
300
400
500
600
700
800
Primary Ancillary Passive Nontelecommuter
Telecommute
Regular work
Min/day
FIGURE 2 Average daily work durations by telecommuting form.
0
10
20
30
40
50
60
70
Out-of-Home Shopping MaintenanceDiscretionaryEscort In-Home Shopping
Primary telecommuter
Ancillary telecommuter
Passive telecommuter
Nontelecommuter
Min/day
FIGURE 3 Average daily durations for nonmandatory activities.
86 Transportation Research Record 2566
that any change in commute travel (either complete removal or
temporal shift) provides individuals with additional freedom to
participate in other out-of-home activities (mainly nonmandatory)
and therefore leads to higher daily trip rates.
METHODOLOGY
Since the advent of the activity-based framework, time use analysis has
been of particular interest to transportation planners. The idea is that
different activities compete for a limited time (or space) budget within
a day. Research has found that the interactions among different activity
durations and their impacts on travel behavior can be well represented
in a structural equations model (SEM) framework (27–33).
SEMs are multivariate (i.e., multiequation) regression struc-
tures. In contrast to the multivariate linear model, the response
variables in one regression equation in an SEM may also appear as
predictors in another equation; variables in an SEM may influence
one another reciprocally, either directly or through other variables
as intermediaries. This aspect of SEMs allows them to capture
relationships between different endogenous variables (34–37 ).
A typical SEM (with G endogenous variables) is defined by a
matrix equation system as shown in Equation 1:
Y
Y
Y
Y
X
X
G
G
GGGG
n
GGnn G
=
ββ
ββ
+
γγ
γγ
+
ε
ε
(1)
1111
1
111 1
1
11
This equation can also be written as
YB
YX
=+Γ+ε
(2)
or
YI
BX
()()
=− Γ+ε
−(3)
1
where
Y = column vector of endogenous variables,
I = identity matrix,
B = matrix of parameters associated with right-hand-side endog-
enous variables,
X = column vector of exogenous variables,
Γ = matrix of parameters associated with exogenous variables, and
ε = column vector of error terms associated with the endogenous
variables.
SEMs are estimated by covariance-based structural analysis, also
called the method of moments, in which the difference between the
sample covariance and the model-implied covariance matrices is mini-
mized (36). Available methods for parameter estimation include maxi-
mum likelihood (ML), unweighted least squares, generalized least
squares, scale-free least squares, and asymptotically distribution free.
While the violation of normality assumptions might be a problematic
issue in using the ML estimation method, some statistical studies have
shown that ML provides the best performance among the various esti-
mation methods and that variable estimates are robust with regard to
nonnormality issues, especially when the sample size is large (38–40).
Therefore, the ML estimation approach was adopted in this research.
Two SEMs, one focusing on workers and the other on nonworkers,
were developed for this study. In each model, the endogenous vari-
ables represent the durations for the five aforementioned nonmanda-
tory activities as well as the total number of daily trips. Exogenous
variables include regular work duration, telecommute durations (classi-
fied by their forms), presence of other telecommuters in the household,
and socioeconomic and demographic attributes. To maintain the mod-
el’s parsimony and avoid overcomplexity, only six socioeconomic–
demographic attributes were tested in the model. These variables were
selected on the basis of recommendations in the literature and include
age, gender, driver’s license, number of children, number of workers,
and vehicle ownership. Figure 4 illustrates the expected interactions
among the variables.
MODEL RESULTS
SEM Results for Workers
The model results for workers are presented in Table 2 and Figure 5.
Since the ML estimation approach only provides t-values for direct
effects, significance levels are shown for direct effects only.
TABLE 1 Pairwise t-Test Comparisons for Activity Duration
and Daily Trip Rates
Primary Ancillary Passive Nontelecommuter
Out-of-Home Shopping
Primary — — — —
Ancillary 5.45*** — — —
Passive 4.31*** 1.12 — —
Nontelecommuter 6.12*** 0.254 1.23
Discretionary
Primary — — — —
Ancillary 3.67*** — — —
Passive 6.33*** 1.89* — —
Nontelecommuter 6.38*** 1.04 1.84*
In-Home Shopping
Primary — — — —
Ancillary 0.178 — — —
Passive 1.35 1.06 — —
Nontelecommuter 0.485 0.09 1.25 —
Maintenance
Primary — — — —
Ancillary 4.66*** — — —
Passive 5.04*** 0.355 — —
Nontelecommuter 5.77*** 0.42 0.07
Escort
Primary — — — —
Ancillary 0.9 — — —
Passive 1.93* 0.61 — —
Nontelecommuter 2.94*** 1.15 0.69 —
Daily Trips
Primary — — — —
Ancillary 1.78* — — —
Passive 3.63*** 1.54 — —
Nontelecommuter 7.51*** 3.8 1.95* —
Note: — = 0 value.
*Significance level = .1; ***significance level = .01.
Asgari, Jin, and Du 87
The modeling procedure involved three steps. First, the core struc-
ture of the model was built on the basis of the causal relationships
among endogenous variables only. No initial assumptions were made
with regard to causality; different combinations of causal effects were
tried and the best combination on the basis of t-values was selected.
Work and telecommute durations were then added to the model.
Finally, socioeconomic variables were incorporated into the paths.
As expected, Table 2 shows that all direct effects among the activi-
ties (including work activities) are negative, which indicates substi-
tution effects. All positive (supplementary) impacts are indirect and
result from a more complicated set of interactions (e.g., two negative
direct effects, between A and B and between B and C, could result in
a positive indirect effect between A and C).
The effects of activities on daily trip generation comply with the
literature and with general expectations. Higher out-of-home non-
mandatory activity durations tend to lead to more trips, and in-home
shopping and work activities discourage trip generation. The high-
est positive impacts are shown by out-of-home shopping and escort
activities, which reflects the prevalence of these two purposes among
workers in a workday. Among the demographic variables tested, only
age and vehicle ownership showed significant impacts.
Since the purpose of this study was to investigate the impacts of
telecommuting, the analysis given here focuses on the comparison
between regular work and telecommuting engagement in terms of
their effects on nonmandatory activity durations. The analysis is
enhanced by t-test comparisons evaluating the statistical significance
of each of the computed differences with regard to total effects. To
obtain the t-values, sample bootstrapping of the data using 2,000 draws
was conducted, which provides the standard errors for the total effects
in the model. These values were then manipulated to obtain the pair-
wise comparison t-values. The absolute values of t-statistics are given
in Table 3.
Regular work hours exhibit negative direct impacts on all five
nonmandatory activities. Telecommuting does not appear to have
significant impacts on escort and in-home shopping activities. All
three forms of telecommuting engagement show smaller negative
impacts than workplace work, with all differences being statistically
significant at the 1% level (10% for in-home shopping). This implies
that telecommuting increases the duration of participating in these
nonmandatory activities when it fully or partially replaces work at
the regular workplace.
For example, in an 8-h workday, a full-day telecommuter increases
out-of-home shopping duration by about 11.52 min compared with
a commuter: [−0.0115 − (−0.0355)] ∗ 60 ∗ 8 = 11.52 min. Similarly,
the time spent in maintenance and discretionary activities increases
by about 16 and 47 min, respectively.
Every hour of telecommuting increases the time spent on out-
of-home shopping, maintenance, and discretionary activities by
approximately 1.30, 3.21, and 3.63 min, respectively, for ancillary
telecommuters and by about 2.14, 2.45, and 3.10 min, respectively,
for passive telecommuters. Comparison among the three forms of
telecommuting indicates that primary telecommuters have the high-
est tendency of spending the time savings from commuting on dis-
cretionary activities, whereas ancillary and passive telecommuters
are more likely to spend time in maintenance and shopping activities,
respectively.
Time use in escort and in-home shopping activities is not affected
by other activities, except for work at the workplace, and the magni-
tude of impacts is relatively small. This may not be surprising, since
escort and in-home shopping activities by nature are independent of
external influences. Escort activities are usually constrained in a fixed
time–space schedule and are expected to have priority over other non-
mandatory activities. The absence of causal effects on escort from
other nonmandatory errands is therefore reasonable. With regard to
in-home shopping, previous studies suggest that any virtual in-home
activities could be done in parallel with other in-home subsistence
errands and may not be affected by other activities (18).
In terms of trip generation, all telecommuting forms have smaller
negative effects than workplace work hours, and the differences are
statistically significant. This indicates that telecommuters produce
Other in-home
activities
Socioeconomic/
demographic
Out-of-home
shopping duration
Out-of-home
discretionary
duration
Out-of-home escort
duration
Out-of-home
maintenance
duration
Work/telework
duration
Telecommuting form
Number of daily trips
• Primary
• Ancillary
• Passive
In-home
shopping
FIGURE 4 Sample path diagram for workers’ model.
TABLE 2 SEM Results for Workers: Nonstandardized Coefficients
Out-of-Home Activities
In-Home
Activities
Workplace
Work Duration
Telecommuting Duration Socioeconomics and Demographics
Shopping Maintenance Discretionary Escort Shopping Primary Ancillary Passive Age Household Vehicles
Out-of-home activities
Shopping
Total −0.0182 0.0005 −0.0001 −0.0355 −0.0115 −0.0138 0.0002 0.1539 0.0273
Direct —−0.0182*** — — — −0.0364*** −0.0118*** −0.0138*** — 0.1583*** —
Indirect — 0.0005 −0.0001 0.001 0.0004 — 0.0002 −0.0045 0.0273
Maintenance
Total 0.0022 −0.0288 0.0032 0.0016 −0.0524 −0.0194 0.0011 −0.0116 0.2452 −1.5007
Direct — — −0.0288*** — — −0.0552*** −0.0194*** — −0.0129 0.2388*** −1.4993***
Indirect 0.0022 — 0.0032 0.0016 0.0028 — 0.0011 0.0013 0.0064 −0.0014
Discretionary
Total −0.0759 0.0014 −0.1096 −0.055 −0.0976 0.0009 −0.0371 −0.046 −0.2214 0.0485
Direct −0.0759*** — — −0.1096*** −0.055*** −0.1014*** — −0.0382*** −0.046*** −0.2133*** —
Indirect — 0.0014 — — 0.0039 0.0009 0.001 — −0.0081 0.0485
Escort
Total −0.0091 −0.0325
Direct — — — — — −0.0091*** — — — −0.0325* —
Indirect — —
In-home activities
Shopping
Total −0.0031 −0.919
Direct — — — — — −0.0031* — — — — −0.919***
Indirect — —
Total daily trips
Total 0.0091 0.0021 0.0045 0.0086 −0.0014 −0.0027 −0.0007 −0.0003 −0.0002 0.0059 0.0551
Direct 0.0095*** 0.0022*** 0.0045*** 0.0091*** −0.0011*** −0.0017*** −0.0006*** — — 0.0052*** 0.0569***
Indirect −0.0003 −0.0002 −0.0001 −0.0005 −0.0002 −0.001 −0.0001 −0.0003 −0.0002 0.0007 −0.0018
Note: — = 0 value. Goodness-of-fit information: for absolute fit, minimum chi-square = 113.3273, degrees of freedom = 32, minimum chi-square/degrees of freedom = 3.5415, goodness-of-fit index = 0.9988, adjusted
goodness-of-fit index = 0.997, and root mean square error of approximation = 0.0128; for relative fit, normed fit index = 0.9811, relative fit index = 0.961, incremental fit index = 0.9864, Tucker–Lewis index = 0.9717,
comparative fit index = 0.9863, and number of observations = 12,593; for parsimony of fit, parsimonious normed fit index = 0.4757 and parsimonious comparative fit index = 0.4782.
*Significance level = .1; ***significance level = .01.
Asgari, Jin, and Du 89
more trips than do commuters, even when commute travel is com-
pletely removed. The higher trip rates for passive telecommuters
(compared with ancillary telecommuters) may result from their irreg-
ular decision-making pattern, which encourages them to make the
most of their telecommuting opportunity and accomplish more out-
of-home activities and thus leads to higher daily trip rates. However,
the significance of this comparison is rejected by the corresponding
t-test comparison.
To capture interactions among workers in the same household, the
presence of other telecommuters and their telecommuting durations
were tested in the model. The outcomes were poor in terms of the t-test
(statistical significance) and the overall goodness of fit of the model
(false causal effects). Therefore, those variables were excluded.
SEM Results for Nonworkers
The model results for nonworkers are given in Table 4. Significance
levels are shown for direct effects only. The influence of telecom-
muting is reflected by two variables: the presence of telecommuters
in the household and their telecommuting durations. Only the first
variable shows reasonable results in terms of significance and overall
model performance. In terms of demographic variables, only gender
shows significant impacts.
The presence of telecommuters in the household has remark-
able impacts on time use in maintenance activities for nonworking
household members. The model shows that primary telecommut-
ing increases the time spent on maintenance activities by 14 min
Shopping duration—
out of home
Shopping duration—
in home
HHVEH
AGE
Maintenance duration—
out of home
Discretionary duration—
out of home
Escort duration—
out of home Work duration—
out of home
Linked trips—sum
Passive duration
Ancillary duration
Primary duration
FIGURE 5 SEM path diagram for workers (HHVEH 5 household vehicle).
TABLE 3 Pairwise Comparison t-Values for Different Telecommuting Arrangements
Primary Versus
Passive
Primary Versus
Ancillary
Ancillary Versus
Passive
Passive Versus
Nontelecommuter
Ancillary Versus
Nontelecommuter
Primary Versus
Nontelecommuter
Out-of-home shopping 4.30*** 4.07*** 0.67 17.97*** 7.50*** 7.12***
Maintenance 1.42 5.11*** 3.34*** 8.38*** 17.46*** 6.55***
Discretionary 8.38*** 5.42*** 0.99 7.39*** 7.41*** 23.49***
Escort — — — 7.46*** 7.46*** 7.46***
In-home shopping — — — 1.83* 1.83* 1.83*
Total daily trips 2.92*** 2.52** 1.29 26.88*** 25.15*** 10.17***
Note: — = 0 value.
*Significance level = .1; **significance level = .05; ***significance level = .01.
TABLE 4 SEM Results for Nonworkers: Nonstandardized Coefficients
Out-of-Home Activities
In-Home
Activities Presence of Telecommuters in Household Socioeconomics
and Demographics:
MaleShopping Maintenance Discretionary Escort Shopping Primary Ancillary Passive
Out-of-home activities
Shopping
Total −0.7649 −0.009 −0.1533 −0.0061 −0.1269 −0.0564 0.1366 −6.0252
Direct —−0.0383** −0.6521*** — −0.0266** — — — −5.7029***
Indirect −0.7649 0.0293 0.4988 0.0205 −0.1269 −0.0564 0.1366 −0.3223
Maintenance
Total −0.0168 14.0732 −0.1546 −15.1525 −3.464
Direct — — — — −0.0168 14.0732* — −15.1525* −3.4967**
Indirect — — −0.1546 — 0.0327
Discretionary
Total 1.173 −0.045 −0.7649 −0.0305 −0.633 −0.2813 0.6815 0.7777
Direct 4.9888*** — — — — — — — 30.8363**
Indirect −3.8158 −0.045 −0.7649 −0.0305 −0.633 −0.2813 0.6815 −30.0586
Escort
Total −0.0021 0.0001 −0.0004 0.0001 0.0011 0.0005 −0.0012 −0.6847
Direct — — −0.0018 — — — — — −0.6834*
Indirect −0.0021 0.0001 0.0013 0.0001 0.0011 0.0005 −0.0012 −0.0014
In-home activities
Shopping
Total 9.2209 −1.9495
Direct — — — — — — 9.2209 — −1.9495*
Indirect — —
Total daily trips
Total 0.006 0.0009 −0.0011 0.014 −0.0011 0.4087 0.5108 −0.0136 −0.0802
Direct 0.0117*** 0.0011*** 0.0028*** 0.014*** −0.0009*** 0.3961** 0.5208** — —
Indirect −0.0057 −0.0002 −0.004 — −0.0002 0.0126 −0.0099 −0.0136 −0.0802
Note: — = 0 value. Goodness-of-fit information: for absolute fit, minimum chi-square = 39.3573, degrees of freedom = 24, minimum chi-square/degrees of freedom = 1.6399, goodness-of-fit index =
0.9996, adjusted goodness-of-fit index = 0.9991, and root mean square error of approximation = 0.0057; for relative fit, normed fit index = 0.9829, relative fit index = 0.9678, incremental fit index =
0.9932, Tucker–Lewis index = 0.9872, comparative fit index = 0.9932, and number of observations = 19,431; for parsimony of fit, parsimonious normed fit index = 0.5242 and parsimonious comparative
fit index = 0.5297.
*Significance level = .1; **significance level = .05; ***significance level = .01.
Asgari, Jin, and Du 91
for nonworking household members. One reason could be that the
removal of commute travel reduces the chances of trip chaining or
stops and therefore shifts maintenance activities to other house-
hold members. However, passive telecommuting shows exactly the
opposite effect by decreasing maintenance duration for nonworking
household members by 15 min. This may imply a shift of mainte-
nance duties from the other household member to the telecommuter
due to the partially relaxed schedule.
Overall, the presence of regular telecommuters, primary and ancil-
lary, increases nonworkers’ daily trip rates by 0.4 and 0.5, respectively.
This may be attributable to the role of telecommuting in increasing
the chances for joint out-of-home activities; any relaxation in the
worker’s schedule will also provide more freedom for the nonwork-
ing members. Another reason may be that the increased time spent at
home by telecommuters provides opportunities for nonworking mem-
bers to replace some in-home activities with out-of-home errands.
Passive telecommuters appear to have a slight indirect impact on trip
generation of nonworking household members. In view of the ran-
dom and short-term nature of part-day nonregular telecommuting, the
impacts on the trip making of other household members are expected
to be trivial. Since both telecommuting engagement and activity par-
ticipation involve midterm lifestyle arrangement and household inter-
actions, processing of the intertwining effects becomes even more
complicated. Capture of all the underlying forces and interactions
would require additional efforts to account fully for the household
interactions and telecommuting arrangement.
To examine whether there are significant differences among the
telecommuting forms in terms of their impacts, bootstrapping tests
were again conducted. However, the results led to remarkable infla-
tion of the standard errors, which would reduce the significance of
total effects. As suggested in the literature (41), this might be due to
the presence of a categorical variable (gender) in the model. There-
fore, the bootstrapping and comparison results are not presented for
the nonworkers’ model.
DISCUSSION AND CONCLUSIONS
The impacts of telecommuting engagement on individuals’ time
use patterns for nonmandatory activities were studied. The results
indicate that although total work hours are higher for telecommuters
(either full-day or part-day), they spend more time on nonmandatory
activities and produce more trips than do commuters. This confirms
the hypothesis that when telecommuting relaxes the constraints
related to commute travel, even partially, workers tend to allocate
their time–space budget to accomplish nonmandatory assignments,
which often lead to additional trips. Furthermore, different forms of
telecommuting have different impacts on time use and trip genera-
tion. Complete replacement of the commute increases total daily
travel by one trip, while displacement of the commute leads to about
a one-half trip increase.
This research effort also explored the impacts of telecommut-
ing on other household members, including both workers and non-
workers. No significant impacts were captured for other workers in
the household. The model shows strong impacts of telecommuting
on the time use pattern of nonworking household members. Since
both telecommuting engagement and activity participation involve
midterm lifestyle arrangement and household interactions, process-
ing of the intertwining effects becomes even more complicated.
Capture of all the underlying forces and interactions would require
additional efforts to account fully for the household interactions and
midterm lifestyle arrangement.
This study provides a foundation for advancing the understand-
ing of the impacts of telecommuting on daily activity time use and
travel demand. From a policy-making perspective, this study might
help shed light on some of the major questions regarding secondary
(long-term) impacts of telecommuting. While telecommuting is gen-
erally believed to reduce commute trips, it has other implications in
terms of time use and trip generation, not only for the telecommuters
but also for their household members. This study indicates the need
for a refined approach to gauging the impacts of telecommuting on
congestion relief and overall travel outcomes.
REFERENCES
1. 2013 Status of Telework in the Federal Government. Report to the
Congress. U.S. Office of Personnel Management, 2013.
2. Schintler, L. A. Trip Reduction Study: The Impact of Telecommuting
Delay in the Washington, D.C. Area. School of Public Policy, George
Mason University, Fairfax, Va., 2001.
3. Vu, S. T., and U. Vandebona. A Model for Analysis of Impacts of Tele-
commuting on Network Travel Time. In 30th Australasian Transport
Research Forum, Melbourne, Victoria, Australia, 2007.
4. Vu, S. T., and U. Vandebona. Telecommuting and Its Impacts on
Vehicle-km Travelled. In Proceedings of International Congress on
Modelling and Simulation, University of Canterbury, Christchurch,
New Zealand, 2007, pp. 253–260.
5. Vu, S. T., and U. Vandebona. Evaluation of Impacts of Telecommut-
ing in Traffic Assignment. Presented at 30th Conference of Australian
Institutes of Transport Research, Perth, Australia, 2008.
6. Vu, S. T. Analysis of Impacts of Telecommuting for Reduction of
Environmental Pollution. Presented at 29th Conference of Australian
Institutes of Transport Research, Adelaide, Australia, 2007.
7. Pirdavani, A., T. Brijs, T. Bellemans, B. Kochan, and G. Wets. Assess-
ing the Impacts of a Teleworking Policy on Crash Occurrence: The
Case of Flanders, Belgium. Presented at 92nd Annual Meeting of the
Transportation Research Board, Washington, D.C., 2013.
8. Niles, J. Beyond Telecommuting: A New Paradigm for the Effect of Tele-
communications on Travel. Report No. DOE/ER-0626. Offices of Energy
Research and Scientific Computing, U.S. Department of Energy, 1994.
9. Mokhtarian, P. L., and I. Salomon. Emerging Travel Patterns: Do Tele-
communications Make a Difference? In In Perpetual Motion: Travel
Behaviour Research Opportunities and Application Challenges (H. S.
Mahmassani, ed.), Pergamon Press/Elsevier, Oxford, United Kingdom,
2002, pp. 143–182.
10. Mokhtarian, P. L. A Synthetic Approach to Estimating the Impacts
of Telecommuting on Travel. Urban Studies, Vol. 35, No. 2, 1998,
pp. 215–241.
11. Pendyala, R. M., G. G. Konstadinos, and R. Kitamura. Impacts of
Telecommuting on Spatial and Temporal Patterns of Household
Travel. Transportation, Vol. 18, 1991, pp. 383–409.
12. Wells, K., F. Douma, H. Loimer, L. Olson, and C. Pansing. Telecommut-
ing Implications for Travel Behavior: Case Studies from Minnesota. In
Transportation Research Record: Journal of the Transportation Research
Board, No. 1752, TRB, National Research Council, Washington, D.C.,
2001, pp. 148–156.
13. Mokhtarian, P. L. Telecommunications and Travel: The Case for Com-
plementarity. Journal of Industrial Ecology, Vol. 6, No. 2, Special Issue
on E-Commerce, the Internet, and the Environment, 2003, pp. 43–57.
14. Jiang, Y. The Impact of Telecommuting on the Journey to Work: A
Two-Sample Instrumental Variable Approach. Dissertation chapter,
University of Maryland, 2008.
15. Zhu, P. Telecommuting Travel Behavior and Residential Location Choice:
Can Telecommuting Be an Effective Policy to Reduce Travel Demand?
PhD dissertation. University of Southern California, Los Angeles, 2011.
16. Ren, F., and M. P. Kwan. The Impact of the Internet on Human Activity–
Travel Patterns: Analysis of Gender Differences Using Multi-Group
Structural Equation Models. Journal of Transport Geography, Vol. 17,
2009, pp. 440–450.
17. van den Berg, P., T. Arentze, and H. Timmermans. A Path Analysis of
Social Networks, ICT Use, and Social Activity–Travel Patterns in the
Netherlands. Presented at 12th World Conference on Transport Research,
Lisbon, Portugal, 2010.
92 Transportation Research Record 2566
18. Mosa, A., N. Harata, and N. Ohmori. Exploring the Impacts of In-Home
Virtual Activities on Daily Activity and Travel Behavior: An Analysis
Using Cairo Activity and Telecommunication Diary Data 2006. Presented
at 12th World Conference on Transport Research, Lisbon, Portugal, 2010.
19. Van Wee, B., K. Geurs, and C. Chorus. Information, Communication,
Travel Behavior, and Accessibility. Journal of Transport and Land Use,
Vol. 6, No. 3, 2013, pp. 1–16.
20. Handy, S. L., and P. L. Mokhtarian. Planning for Telecommuting: Plan-
ning and Policy Issues. Journal of the American Planning Association,
Vol. 61, No. 1, 1995, pp. 99–111.
21. Handy, S. L., and P. L. Mokhtarian. Forecasting Telecommuting: An
Exploration of Methodologies and Research Needs. Transportation,
Vol. 23, 1996, pp. 163–190.
22. Lyons, G., and H. Haddad. Commute Replacement and Commute
Displacement: The Rise of Part-Day Home Working. In Transporta-
tion Research Record: Journal of the Transportation Research Board,
No. 2082, Transportation Research Board of the National Academies,
Washington, D.C., 2008, pp. 1–7.
23. Haddad, H., G. Lyons, and K. Chatterjee. An Examination of Determi-
nants Influencing the Desire for and Frequency of Part-Day and Whole-
Day Homeworking. Journal of Transport Geography, Vol. 17, 2009,
pp. 124–133.
24. Asgari, H., and X. Jin. Toward a Comprehensive Telecommuting Analysis
Framework: Setting the Conceptual Outline. In Transportation Research
Record: Journal of the Transportation Research Board, No. 2496, Trans-
portation Research Board, Washington, D.C., 2015, pp. 1–9.
25. Asgari, H., X. Jin, and A. Mohseni. Choice, Frequency, and Engage-
ment: Framework for Telecommuting Behavior Analysis and Model-
ing. In Transportation Research Record: Journal of the Transportation
Research Board, No. 2413, Transportation Research Board of the
National Academies, Washington, D.C., 2014, pp. 101–109.
26. New York Metropolitan Transportation and North Jersey Transporta-
tion Planning Authority. 2010–2011 Regional Household Travel Survey
Data User’s Manual. 2013.
27. Golob, T. F., and M. G. McNally. A Model of Activity Participation
Between Household Heads. Transportation Research Part B, Vol. 31,
No. 3, 1997, pp. 177–194.
28. Golob, T. F., and H. Meurs. A Structural Model of Temporal Change in
Multi-Modal Travel Demand. Transportation Research Part A, Vol. 21,
No. 6, 1987, pp. 391–400.
29. Golob, T. F., M. A. Bradley, and J. W. Polak. Travel and Activity Par-
ticipation as Influenced by Car Availability and Use. UCTC 286. Uni-
versity of California Transportation Center, University of California,
Berkeley, 1995.
30. Golob, T. F., D. S. Bunch, and D. Brownstone. Forecasting Future Vehicle
Usage Using a Jointly Estimated Revealed- and Stated-Preference Model.
Presented at 75th Annual Meeting of the Transportation Research Board,
Washington, D.C., 1996.
31. Kuppam, A. R., and R. M. Pendyala. A Structural Equations Analysis
of Commuters’ Activity and Travel Patterns. Transportation, Vol. 28,
2001, pp. 33–54.
32. Lu, X., and E. I. Pas. Socio-Demographics, Activity Participation
and Travel Behavior. Transportation Research Part A, Vol. 33, No. 1,
1999, pp. 1–18.
33. Wu, W., and Z. Juan. Activity-Travel Behavior Analysis Based on
Structural Equation Model. In Proceedings of the 2009 International
Conference on Electronic Commerce and Business Intelligence, IEEE,
Los Alamitos, Calif., 2009, pp. 486–490.
34. Koutsoyiannis, A. Theory of Econometrics: An Introductory Exposition
of Econometric Methods, 2nd ed. Barnes and Noble Books, New York,
1972.
35. Judge, G. G., W. E. Griffiths, R. C. Hill, H. Lutkepohl, and T. C. Lee.
The Theory and Practice of Econometrics, 2nd ed. John Wiley & Sons,
New York, 1985.
36. Bollen, K. A. Structural Equations with Latent Variables. John Wiley &
Sons, New York, 1989.
37. Johnston, J., and J. DiNardo. Econometric Methods, 4th ed. McGraw-Hill,
New York, 1997.
38. Iacobucci, D. Structural Equations Modeling: Fit Indices, Sample Size,
and Advanced Topics. Journal of Consumer Psychology, Vol. 20, No. 1,
2010, pp. 90–98.
39. Olsson, U. H., T. Foss, S. V. Troye, and R. D. Howell. The Performance
of ML, GLS, and WLS Estimation in Structural Equation Modeling
Under Conditions of Misspecification and Nonnormality. Structural
Equation Modeling, Vol. 7, No. 4, 2000, pp. 557–595.
40. Hu, L., and P. M. Bentler. Fit Indexes in Covariance Structure Modeling:
Sensitivity to Underparameterized Model Misspecification. Psychological
Methods, Vol. 3, No. 4, 1998, pp. 424–453.
41. Hoyle, R. H. Handbook of Structural Equation Modeling. Guilford
Press, New York, 2012.
The Standing Committee on Effects of Information and Communication Technologies
on Travel Choices peer-reviewed this paper.