Telecommuting and sustainable travel: Reduction of overall travel time,increases in non-motorised travel and congestion relief?

Abstract

Chinese

Existing research has concluded that shares of telecommuting are low but stable, increase with distance from the workplace and that telecommuting may reduce commuting-related travel. Its effect on work and non-work travel are subject to rebound effects and, thus, still debated. Additionally, telecommuting does not necessarily occur entirely at home. The paper studies telecommuting’s potential as a sustainable mobility tool in Canada to reduce overall travel time and peak hour travel, and to increase non-motorised travel. Do types of telecommuting arrangements have varying relationships with these studied travel patterns? Using time use data from the 2005 Canadian General Social Survey, studied outcomes are regressed on telecommuting arrangements (all day home working, part-day home working and a combination of other locations and home and/or workplace) and other personal characteristics. Depending on telecommuting arrangements and travel outcomes, results vary. Working from home is associated with decreases in overall travel time by 14 minutes and increases in odds of non-motorised travel by 77%. Other forms of telecommuting yield different results. Telecommuters may be more likely to avoid peak hours when they do take trips. Types of telecommuting arrangements have different impacts on sustainable travel outcomes that should be considered depending on policy priorities.

Keywords

active transport coffee shops non-work travel peak hour telework time use

Introduction

Telecommuting has been proposed as a solution to provide employees with more personal time and flexible working conditions as well as to reduce commuting and peak hour congestion. Although telecommuting adoption rates are lower than expected both in Canada and the USA (Alizadeh and Sipe, 2013; Turcotte, 2010), it is still viewed as a Transportation Demand Management (TDM) strategy (Kim et al., 2012). TDM policy packages are being developed in metropolitan areas to address issues of traffic congestion, metropolitan mobility and greenhouse gas emissions (Rietveld, 2011). The sustainable mobility paradigm also supports these objectives as well as the increased use of non-motorised modes of transportation because walking and cycling produce limited travel-related emissions and improve personal health (CSEP, 2012; Transport Canada, 2007).

While telecommuting is supported by organisations advocating for sustainable transportation and employer-based travel policies in Canada, no national-level policy directly targets the transportation-related benefits of telecommuting. Transport Canada (2007) provides a case study on teleworking’s benefits but, as many other countries, has no policy, objectives or reporting requirements with respect to travel. The Treasury Board of Canada Secretariat’s (Secrétariat du Conseil du Trésor du Canada, 1999) policy on telecommuting for Federal employees does not include travel or congestion reduction requirements or monitoring.

In the USA, The National Air Quality Act of 1999 established telecommuting programmes in cities with severe air quality conditions to assess whether a reduction in emissions could be achieved (Atkyns et al., 2002). Designing a rigorous methodology to measure emissions reduction from telecommuting has presented important challenges (Nelson et al., 2007). The more recent 2010 National Air Quality and Telework Enhancement Act renewed provisions for telecommuting policies in US federal agencies (Lari, 2012) even though the literature is still inconclusive as to the transportation benefits of telecommuting.

As is the case with many environmental policies, telecommuting may cause unintended rebound effects (Rietveld, 2011). Understanding these may support policy designs that mitigate negative effects and enhance telecommuting’s travel and emission reduction performance. For example: commute trips of telecommuters may be replaced by other trips during the day; on a telecommuting day, the car usually used to commute can be used by other household members to conduct activities; part-day telecommuting may shift travel off-peak hours but not reduce commute travel; and more telecommuting potentially enables locating farther from the workplace (Kim et al., 2012; Rietveld, 2011).

The objective of this paper is to provide a more detailed classification of teleworking lifestyles and assess whether travel patterns differ between types of telecommuters and non-telecommuters. Can telecommuting be considered a sustainable travel tool to reduce overall travel time, peak hour travel and to increase non-motorised travel in Canada? More specifically, we assess associations between these three outcomes and three telecommuting arrangements on a telecommuting day. While causality is implied theoretically, the cross-sectional analysis provided in this paper cannot assert it.

Research hypotheses

Avoided trips to work may be substituted by spending time travelling to other locations, which leads to modest differences in overall travel time.

Because telecommuters are freed from the longer distance travel that commuting involves, they may undertake more non-motorised travel during their telecommuting day.

Because of their flexible work arrangements, telecommuters may avoid travel at peak times for all trips.

Telecommuters sometimes travel to work in coffee shops, libraries and other locations and their travel outcomes will differ from home-working or part-day home-working telecommuters.

Justification and empirical evidence on the three studied travel outcomes are followed by a presentation of telecommuting trends in Canada. We then explore associations between telecommuting arrangements and the travel outcomes using a national-level time-use survey. This approach can be replicated elsewhere with similar data. The possibility of formalising goal-specific telecommuting policies in Canada and elsewhere for the purpose of TDM and sustainable travel is then discussed.

Telecommuting in the context of travel research

For the purpose of travel research, telecommuting is distinct from teleworking. Gurstein (2001: 31) broadly defines telework as ‘work-related substitution of telecommunications and related information technologies for travel’ and can therefore take place in an office space. However, office-based telework does not reduce commute travel (Mokhtarian, 1991), although it may reduce other work-related travel. Travel reduction (or changes in trip departure times) and remote supervision are thus identified as the two basic criteria for telecommuting.

Telecommuting as a sustainable travel and travel demand management strategy

Telecommuting can be considered a TDM initiative along with carpool and discounted transit pass programmes (Litman, 2011) and has long been identified as a promising strategy for reducing travel demand (Mokhtarian et al., 1995). Effective TDM programmes provide a mutually supportive set of sustainable travel options that offer more flexibility and convenience (Vanoutrive et al., 2010). According to Transport Canada (2011), three key dimensions of sustainability are sought when designing TDM programmes: quality of life, environmental health and economic growth. Quality of life may be improved by reducing congestion through travel options that are convenient, reliable and that increase physical activity levels through walking, cycling and transit use. Environmental health effects include lower emissions of greenhouse gases and air pollutants that can be achieved with telecommuting as it reduces the need to use or even to own a car. Economic benefits include lower long-term costs of transportation infrastructure (e.g. reduced need for highway expansion due to reduced peak hour travel) and operations (e.g. vehicle fuel, repairs, insurance). While these benefits are presented by Transport Canada (2011) as facts, the body of available research is not as conclusive.

From a transportation perspective, the main question regarding telecommuting is whether it constitutes a substitute for travel. But if telecommuters potentially use the time saved by not travelling to a workplace to conduct other travel-related activities, the effect of telecommuting on overall travel time would be expected to be modest or null, and favor the substitution viewpoint, our first hypothesis.

Types of telecommuters

Providing an operational definition and set of criteria for telecommuting has been complicated by the fact that telecommuting can take many forms. Time schedule (regular, part-day, overtime or outside workdays), location, frequency or proportion of work time, duration, status (self-employed, full- or part time) can all be used to define a telecommuter (Mokhtarian et al., 2005; Pratt, 2000).

Three types of home-based workers have been distinguished by Schweitzer and Duxbury (2006): employees who work at home instead of at the office (the true potential substitution); employees who perform overtime work at home; and self-employed workers whose business is based in their home. Home-based businesses and overtime work should not be considered telecommuting (Mokhtarian, 1991) because they have no notable incidence on travel. This study also excludes overtime and self-employed workers to focus on those employees that may actually reduce travel or peak time congestion as a result of their telecommuting.

Telecommuting does not necessarily need to take place at home and telecommuting centres (locations closer than a more distant central work location) have been the subject of travel reduction studies (Mokhtarian and Varma, 1998; Pendyala et al., 1991; Saxena and Mokhtarian, 1997). Contemporary anecdotal evidence also suggests that individuals conduct paid work from public locations such as coffee shops, parks and libraries, yet no documented evidence of the rate of use of these alternative locations was found. In one Finnish study, 8.5% of telecommuters worked from summer cottages or from ‘elsewhere’ (Helminen and Ristimäki, 2007).

Part-day home working involves either working at home in the morning to then go to the workplace, or working at home in the afternoon or evening after returning from work (Haddad et al., 2009; Rietveld, 2011). From a travel perspective, part-day home working does not remove workplace access and egress trips, but may rather displace travel outside peak hours. While morning home working likely has a clearer impact on morning peak hour travel, evening home working can potentially have the reverse impact by enabling workers to leave at standard hours, during peak time, to continue working at home. As part-day home working is expected to have lesser impacts on afternoon congestion (Rietveld, 2011), it could also contribute to spreading peak hour congestion across longer periods (Haddad et al., 2009). This study will thus explore trip timing depending on working arrangements.

Overall travel time

Studies on travel time and distance of telecommuters have focused largely on the work trip itself, but some have also included an analysis of non-work trips. Results have been inconsistent, possibly owing to sample sizes and methodologies. One could assume that every day a telecommuter works completely from home, his or her weekly commute-related travel time and distance are decreased by 20%. Yet net trip reductions by employees are much more complicated to compute (Pratt, 2000) because of the potential rebound effects of telecommuting mentioned earlier (Rietveld, 2011). As a result of such an effect, the overall amount of travel (time, distance or trips) dedicated to other activities could likely increase and replace reduced travel. This justifies focusing on overall travel on telecommuting days.

A recent analysis of the National Household Transportation Survey (NHTS) of 2001 and 2009 concluded that telecommuting had a complementary effect on both workers’ one-way commute trips and total non-work trips (Zhu, 2012). That is, for both years, travel distance and time were higher for telecommuters than for non-telecommuters, both for work-related and non-work trips. Other studies have concluded the opposite, leaving the question up for debate. For example, non-commute trips of centre-based telecommuters in California decreased on telecommuting days (Mokhtarian and Varma, 1998). The State of California Telecommuting Pilot Project showed that telecommuters substantially reduced total trips, total distance travelled, and tended to visit non-work destinations that were closer to home (Pendyala et al., 1991). And Saxena and Mokhtarian (1997) revealed that on telecommuting days, the share of activities performed close to home increased for telecommuters, and even on telecommuters’ commuting days, the size of their activity space was smaller than that of non-telecommuters. While Kim (2016) also found that telecommuting reduced commute travel, this household-level analysis revealed that other travel for both the telecommuter and other household members were offsetting this reduction.

When calculating the benefits of the workplace programme in Minnesota, telecommuting generated an estimated travel time saving of 44 hours per year per telecommuter, the equivalent of more than a full week of work (Lari, 2012). For telecommuters, office days generated nearly 28 more miles of overall travel (approx. 45 km), as compared with telecommuting days.

Telecommuting and non-motorised travel

The relationship between telecommuting and non-motorised travel has received little attention. Yet, results on the frequentation of destinations close to home by telecommuters presented above (Pendyala et al., 1991; Saxena and Mokhtarian, 1997) suggest a potentially greater use of non-motorised transportation. The positive health consequences of non-motorised travel have been documented in great details by health researchers: increased prevention of the occurrence of disease, maintenance of good health and promotion of healthy aging have been identified as more likely (CSEP, 2012; TRB-IOM, 2005).

Many employers are developing programmes to help employees be more active in and out of the workplace as such behaviour may generate productivity gains, employee retention and improved quality of life (Lucove et al., 2007). We found no studies focusing on how enabling employees to telecommute may influence non-motorised transportation. Transport Canada (2007) suggested that working from home could actually reduce walking during the day because a person may avoid going out altogether and be confined to a smaller space than most offices where at least minimal walking is required. On the other hand, additional time saved by avoiding the commute may be reinvested in shorter non-motorised errand trips near the home. In Californian centre-based telecommuters, walking and bicycling shares increased modestly but significantly on telecommuting days, as compared with non-telecommuting days (Mokhtarian and Varma, 1998). However, a study on the relationship between commute mode and frequency of walking for other purposes found that individuals working from home were not associated with more walking as compared with car commuters (Lachapelle and Noland, 2012).

These opposing results call for a closer look at this second outcome. We analyse how telecommuting arrangements were compliant with the Canadian guideline of completing 30 minutes or more of physical activity in a day through non-motorised transportation (CSEP, 2012).

Telecommuting and peak hour travel

Congestion reduction was one of the first objectives identified for increasing the adoption of telecommuting (Downs, 1992). Reducing peak hour travel creates environmental benefits through reduced travel emissions, social benefits through shorter travel times, and economic benefits through reduced need to develop large-scale transportation infrastructures. Telecommuting can either replace or displace a number of car trips, thereby alleviating congestion during peak hours, depending on whether individuals avoid work trips entirely, or use part-day telecommuting to avoid peak hour travel. While the argument of congestion reduction is used to support telecommuting, few empirical studies address this potential benefit.

Through a survey of telecommuters in Minnesota, Lari (2012) found that telecommuting days were mainly associated with reduced peak hour trips on two major Highways of the region, but observations were not corroborated by traffic flow data. A study conducted in Montréal and Québec City estimated that over the long term, a reduction of roughly 6% of daily morning peak work trips could be achieved with a combination of telecommuting, telework and flexible work schedule (Bussiére and Lewis, 2002). In Tokyo, it was estimated that by 2010, telecommuting would result in 7% to 11% reduction in traffic congestion (Mitomo and Jitsuzumi, 1999) but no evidence confirmed this reduction.

Evolution of telecommuting in Canada

Research on Canadian teleworkers, telecommuters and homeworkers is relatively scarce (Schweitzer and Duxbury, 2006) and has mostly focused on their characteristics and distribution across urban regions (Moos and Skaburskis, 2007, 2008; Turcotte, 2010) or on potential environmental impacts related to telework (Moos et al., 2006). There is limited evidence on how telecommuting relates to travel in Canada. Existing research has focused more generally on ‘working from home’ through the use of a survey question from the Canadian General Social Survey (GSS): ‘Some people do all or some of their paid work at home. Excluding overtime, do you usually work any of your scheduled hours at home?’ (Turcotte, 2010). Between 2000 and 2008, a relatively stable 10–11% of employees declared working from home. During the same period, the percentage of self-employed workers working from home increased from roughly 50% to 60%. When combining the two groups, the total percentage of the Canadian workforce working from home was similar to that of the USA, approximately 17% in 2000 (Akyeampong and Nadwodny, 2001; Pratt, 2002; Turcotte, 2010). Thus, the slow growth in working from home is mostly due to self-employed workers who are not likely to reduce their travel as a result of their working arrangements. For home-working employees, living in urban areas, living farther from the workplace, being a women and having children were the factors most positively associated with home-working (Turcotte, 2010). Home-working employees were also more likely to be management and professional staff, as well as irregular or on-call employees. Women home-working was associated with impacts on family activities (Hilbrecht et al., 2008).

Pratt (2000) suggests that household activity surveys that capture 24-hour patterns of work at home, at the office or in other locations as well as work and non work-related travel can be useful tools to capture the nuanced mix of personal and work trips fragmented in time (short episodes) and space (at multiple locations). This study uses the GSS’s time use diary to explore travel outcomes related to three telecommuting working arrangements, recognising that home-based, part-day and non-home-based telecommuting may have distinct transportation, energy, and air quality implications (Mokhtarian, 1991).

Methodology

The one-day time use diary module of the GSS (Cycle 19, 2005, public use files) was used to assess telecommuting and travel behaviour of participants on a survey day.

The 2005 GSS is representative of the non-institutionalised Canadian population of the ten provinces and over 15 years old. It includes 19,597 respondents representing slightly more than 26 million Canadians (Béchard and Marchand, 2006). Statistics Canada provides expansion weights adjusted using a raking ratio calibration (post-stratification) technique ‘to match Census based population estimates for strata and for provincial age-sex groups’ (Béchard, 2011: 5). Sampling was carried out in 27 strata representing provinces and urban and non-urban areas within provinces. The following analysis mainly uses information from the time use diary but compares one-day diary shares to the GSS question on telecommuting mentioned earlier.

The unit of analysis of a time use diary is the activity. Activities can be associated back to respondent in the individual level file. The activity diary records information on activities starting at 4 a.m. on the day prior to the survey with a start and end time, duration, location code and activity code. The 21 location options include fixed places such as an individual’s home, workplace, friend’s home and restaurants, as well a places of transit (automobile, public transit, walking and cycling). Activity codes include information on types of activities and on the purpose of trips when location is a travel mode.

Travel-related dependent variables

Three major travel outcomes were analysed. The first two, the overall travel time (the sum of all travel, by all modes), and whether a person walked or bicycled 30 minutes or more (dichotomous variable) on the survey day were analysed at the individual level. We expect telecommuting to be associated with reduced overall travel time and with an increased probability of participating in sufficient non-motorised transportation from a health perspective. We expect categories of telecommuting to have distinct relationships with these outcomes.

To supplement these individual-level analyses, a trip level analysis explored the distribution and correlates of motorised trip departure time during the day. Non-motorised trips were not included, as they do not contribute to road congestion or public transport overcrowding and traffic congestion likely exerts less influence on the decision to take a non-motorised trip at a specific time. Trip departure times were categorised into morning peak (>07:00h–09:00h), evening peak (>16:00h–18:00h), mid day (>09:00h–16:00h), evening (>18:00h–23:00h) and night and early morning (>23:00h–07:00h). This analysis helps understand whether telecommuting is associated with a decrease in peak hour motorised travel. We expect telecommuters to have a lower probability of travelling on morning and evening peak times.

Finally, descriptive statistics on dichotomous travel mode information help interpret results.

Travelled on day

Travelled by car

Travelled by public transit

Travelled by non-motorised modes

Main independent variable: Working arrangements

Work-related activities (‘Work for Pay at Main Job’) was used to classify respondents. ‘Overtime’ was not selected to be consistent with other studies (only four respondents reported overtime). Table 1 presents weighted percentages and frequencies of work episodes on the survey day using the location codes. Because working elsewhere than the workplace or home was much less frequent, and because individuals sometimes worked at multiple locations during a survey day (home, workplace, other locations or a combination of these reported locations: library, coffee shops, restaurants, etc.), an individual-level variable representing combinations of non-overtime working arrangements was created. The four categories include working only from the workplace, only from home, combining home and workplace (part-day homeworking), and combining elsewhere, home and/or workplace.

Table 1.

Work episodes location.

Episode location	Entire survey sample of employed workers (n = 9601 respondents)				Final sample used in analyses (n = 5060 respondents)
	Unweighted		Weighted		Unweighted		Weighted
	Freq.	Percent	Freq.	Percent	Freq.	Percent	Freq.	Percent
Workplace	14,447	85.7	20,932,538	86.1	10,282	91.55	14,697,288	91.57
Home	2030	12.1	2,825,338	11.6	791	7.04	1,114,863	6.95
Other private domicile	114	0.7	161,817	0.7	63	0.56	96,231	0.60
School	33	0.2	74,372	0.3	13	0.12	33,327	0.21
Outdoors	41	0.2	65,338	0.3	13	0.12	17,728	0.11
Coffee shop, restaurants and library	188	1.1	254,984	1.1	69	0.61	91,096	0.57
Total episodes	16,853	100.0	24,314,388	100.0	11,231	100.0	16,050,533	100

Other independent variables

Based on their relevance in other studies, household income (five categories), age (five categories), sex (women), having at least one child, survey day (weekday, Saturday or Sunday) were selected as independent variables in the models. Models were also tested to exclude weekend. The trip level analysis of peak period travel additionally includes trip purpose (seven categories, see Table 5).

While location of an individuals’ home from major employment centres, and the population density and transportation infrastructures near homes may influence travel patterns, this information was not available. The only geography-related information provided in the GSS public use files identifies respondents living in Census Metropolitan Areas (CMA) or Census Agglomerations (CA) and was used to create a dichotomous variable of living in urban versus rural areas (including small towns). We supplemented this information with a survey item on whether a person lives in a single family home, which provides a proxy for population density (Turcotte, 2008) as those living in single family homes are more likely to live in suburban areas, further from the centre.

Case exclusion

To be consistent with other studies and definitions of telecommuting, we restricted the analyses to respondents that worked on the survey day, aged less than 65 years old, not self-employed, and whose main dependent variable of interest, the overall time spent travelling on the survey day, was lower than three standard deviations above the mean value. Outlying values above this cutoff point of five hours are mostly associated with interurban and air travel. The final sample size included 5060 respondents. Weighting expands this analytical sample to represent just over 7 million Canadians carrying out approximately 16 million work-related episodes. For all inferential analyses, weights were rescaled to the sample size (redistribution weights) by dividing weights by the mean weight (Béchard, 2011) to avoid falsely inflating the standard errors of the coefficient estimators by suggesting a sample size as large as the population.

Statistical approach

All analyses were conducted in STATA 14 using weighted and unweighted descriptive statistics. We first compared the analytical sample with the entire survey population (Table 1). Chi squared statistics assess bivariate relationships between working arrangement, socio-demographic characteristics and travel indicators.

Model 1: Based on the variable’s distribution, a cluster of people that did not travel on the survey day, and a distribution of positive values that approximate a normal distribution, we used a Tobit regression to model the censored variable. While Stewart (2013) suggests that OLS may provide unbiased estimates in research using time use diaries, non-travellers are particularly of interest in this research. Removing them from the sample would inevitably create an upward bias on time travelled for telecommuters who are theoretically less likely to be travelling. Because we are also interested in the effect of working arrangements on non-travellers we report on model coefficients and on the unconditional expected value of the dependent variable. Because a very small share of observations was below the limit, the expected value of y over all cases was only slightly different from those above the limit. The unconditional expected value can be interpreted as the linear effect of an independent variable on the dependent variable when the entire population is of interest. To account for the typically longer duration of transit trips, a binary variable of transit use during the day was added to this model (Turcotte, 2010).

Model 2: Non-motorised transportation was modelled using a logistic regression of the probability of achieving 30 minutes or more of physical activity, a threshold used in public health research to match physical activity guidelines (CSEP, 2012). The same independent variables are used in this model, but the analysis is restricted to residents of urban areas (n = 3550), because walking and bicycling for travel are unlikely outside major urban areas. A binary variable of transit use during the day was added because transit users report walking to access transit and are more likely to walk as part of their daily life (Lachapelle et al., 2011). Activity limitation (‘Respondent is limited in the amount or kind of activity he/she can do … because of a physical condition or health problem’) was also used to exclude cases.

Model 3: The final analysis concerns the departure times of all motorised trips. Participants took 16,706 motorised trips during the survey day. Five categories of trip departure time were modelled in a multinomial logit regression with working arrangement, and variables used in previous models as independent variables and with midday trips as the reference category. Trip purposes were included as they were expected to be associated with trip departure times. We expect that the probability of taking a trip decreased during peak travel periods for telecommuter groups. Because the same individuals took multiple trips, clustered variance-covariance estimators were used to allow the standard errors and variance-covariance matrix of the estimators to account for intragroup correlation.

Results

Individuals worked at different locations on the survey day. Of the more than 24 million work episodes made by Canadians on the survey day, more than 20 million were made at a workplace (Table 1). More than 2.8 million were made from home and much fewer work episodes were reported in various locations such as coffee shops, schools, restaurants, libraries and outdoors. We conduct the rest of the analysis using a condensed variable of the working arrangements of surveyed individuals.

Description of sample

Figure 1 compares the percentage distribution of each category of the working arrangements variable developed using the time use diary to the share of respondents reporting home-working in the survey question. It also compares the final analytical sample to the entire sample of employed workers. The shares of the final sample are very similar to those of the entire sample of workers. While a higher share of telecommuters is computed when using the time use survey (12.8% versus 10.3%), removing individuals that worked from elsewhere (2.2%) makes survey question and time use diary estimates quite similar. However, as the GSS survey question asks about ‘usual’ working arrangements, on any given time use survey day, we would expect fewer telecommuters. Misunderstanding of question or combined other location, home and/or workplace days may account for this discrepancy.

Figure 1.

Comparison of survey question on paid work at home and work arrangement categorisation created from time use diary, weighted percentages.

In Table 2, the weighted distribution of working arrangement types for the analytical sample is explored with respect to other independent variables. There are significant differences across working arrangements in income, age, living in urban environment (versus rural) and day of survey. With respect to travel, there were significant differences in percentage travelling at all and by car. Those only working only from home had smaller shares of users of all three modes. Achieving 30 minutes of physical activity through the active transportation modes was higher but not significantly for those only working from home and those that had worked elsewhere.

Table 2.

Description of sample covariates and mode use by working arrangements (weighted percentages).

	Only from workplace	Only from home	Worked from home and workplace	Worked from elsewhere and home or workplace	Total	Chi square test (for unweighted estimates)
Sample size (% row)	4437 (87.3)	283 (5.6)	239 (5.0)	101 (2.2)	5060 (100)
Weighted sample	4,625,191	295,163	257,131	86,183	5,263,667
	%	%	%	%	%
Household income						0.001
Less than Can$20,000	4.39	4.97	2.14	8.51	4.4
Can$20,000–$39,999	15.57	13.01	15.71	7.13	15.25
Can$40,000–$59,999	22.28	20.01	16.65	20.34	21.84
Can$60,000 to $99,999	33.81	32.3	35.01	38.81	33.9
Can$100,000 or more	23.95	29.71	30.48	25.2	24.62
Women	46.45	42.71	44.29	44.65	46.09	0.756
Age						0.006
15–24	15.3	8.06	8.32	20.36	14.66
25–34	24.01	18.88	22.12	23.98	23.63
35–44	26.4	30.05	35.05	19.3	26.88
45–54	24.16	32.41	26.95	27.2	24.82
55–64	10.13	10.6	7.56	9.16	10.01
Has children	30.78	36.61	37.85	34.22	31.53	0.063
Lives in single family home	68.8	72.24	69.6	67.16	69	0.612
Urban (vs) rural	82.9	86.3	79.19	71.4	82.66	0.034
Survey day						0.001
Weekday	89.22	69.52	95.31	78.54	88.19
Saturday	5.61	15.9	3.64	9.17	6.16
Sunday	5.17	14.58	1.05	12.29	5.64
Travelled at all	99.7	89.1	98.9	90.3	98.8	0.001
Travelled by car	88.1	76.3	81.0	80.8	86.9	0.001
Travelled by public transport	12.8	6.9	14.8	13.7	12.6	0.059
Travelled by active modes	18.0	15.0	19.0	25.7	18.1	0.088
Travelled at least 30 min. by active modes	6.9	8.1	6.0	9.7	7.0	0.126

Overall travel time

The Tobit model of overall travel time is presented in Table 3. When compared with the reference category (workplace only), working only from home was associated with less overall travel time by an average of 13 minutes, and working from a combination of location was associated with increased overall travel time by nearly 10 minutes.

Table 3.

Tobit model of total time spent travelling on survey day (weighted estimates).

Overall travel	Coef.	Std err.	P > t	Unconditional expected value
Working arrangements
Only from workplace	(base)
Only from home	−14.217	3.033	0.000	−13.244
Worked from home and workplace	−5.313	3.146	0.091	−5.005
Worked from elsewhere and home or workplace	9.963	4.736	0.035	9.536
Household income
Less than Can$20,000	(base)
Can$20,000–$39,999	0.403	3.703	0.913	0.382
Can$40,000–$59,999	3.962	3.597	0.271	3.764
Can$60,000–$99,999	9.998	3.529	0.005	9.507
Can$100,000 or more	19.060	3.649	0.000	18.232
Women	0.786	1.381	0.569	0.745
Age
15–24	(base)
25–34	1.154	2.333	0.621	1.094
35–44	0.093	2.404	0.969	0.088
45–54	0.820	2.298	0.721	0.778
55–64	−0.963	2.819	0.733	−0.912
Used public transit during day	46.248	2.117	0.000	44.951
Has children	10.381	1.642	0.000	9.878
Lives in single family home	−1.126	1.615	0.486	−1.067
Urban (vs) rural	−6.976	1.856	0.000	−6.574
Survey day
Weekday	(base)
Saturday	−3.876	2.876	0.178	−3.659
Sunday	−9.156	3.000	0.002	−8.587
_cons	62.412	4.313	0.000	59.149
/sigma	48.347	0.485
Total observations	5060
Left-censored observations	59
Uncensored observations	5001
Log likelihood(equal likely)	−30,581.991
Log likelihood(constant only)	−27,604.15
Log likelihood(full model)	−27,258.44
LR chi²(17)	235.44
Prob > chi²	0.000
McKelvey & Zavoina’s R²	0.128

The highest income group, transit users and households with children travelled more, while households living in single-family homes or in urban environments travelled less time. Weekend day were also associated with shorter travel time.

Meeting physical activity guidelines through non-motorised travel

Non-motorised trips advantageously yield no impacts on energy use, emissions and traffic congestion, and additionally provide health benefits to an employee (Lachapelle and Noland, 2012; TRB-IOM, 2005). The logistic regression in Table 4 shows that working only from home was associated with significantly greater odds of achieving 30+ minute of non-motorised travel. Other telecommuting arrangements were associated with higher odds of achieving public health guidelines, but not significantly.

Table 4.

Logistic regression of spending 30 minutes or more in non-motorised transportation (walking and bicycling), restricted to urban areas (weighted estimates).

	Odds ratio	Std err.	P > z
Percent with positive outcome (weighted) : 7.31%
Working arrangements
Only from workplace	(base)
Only from home	1.771	0.436	0.020
Worked from home and workplace	1.255	0.351	0.417
Worked from elsewhere and home or workplace	1.469	0.601	0.347
Household income
Less than Can$20,000	(base)
Can$20,000–$39,999	0.415	0.105	0.000
Can$40,000–$59,999	0.325	0.083	0.000
Can$60,000–$99,999	0.410	0.098	0.000
Can$100,000 or more	0.511	0.129	0.008
Women	1.090	0.138	0.496
Age
15–24	(base)
25–34	0.673	0.121	0.028
35–44	0.555	0.111	0.003
45–54	0.485	0.097	0.000
55–64	0.293	0.091	0.000
Used public transit during day	2.231	0.327	0.000
Has children	0.850	0.135	0.309
Lives in single family home	0.484	0.067	0.000
Survey day
Weekday	(base)
Saturday	1.346	0.308	0.194
Sunday	0.907	0.246	0.719
Constant	0.354	0.088	0.000
Total observations	3550
Log likelihood(equal likely)	−2739.893
Log likelihood(constant only)	−1034.346
Log likelihood(full model)	−962.1909
LR chi²(17)	144.31
Prob > chi²	0.000
McFadden’s R²	0.070

Using public transit during the day was also associated with achieving public health guidelines. Higher income and age were associated with lower odds of achieving 30 or more minutes of non-motorised travel. The higher automobile ownership of wealthier individuals is typically used to explain their lower use of non-motorised modes (TRB-IOM, 2005). Living in a single family home significantly reduced the odds of meeting the physical activity recommendation, as would be expected with low-density suburban locations.

Peak hour motorised travel

This section explores whether motorised trips taken during a travel day were taken at different time periods depending on working arrangements and other covariates. We estimate a multinomial logit regression of departure time to assess associations between telecommuting and peak hour travel (Table 5).

Table 5.

Multinomial logit model of motorised peak hour travel (weighted estimates).

Base: Midday(09:00h–16:00h)	Morning peak (07:00h–09:00h)	Afternoon peak (16:00h–18:00h)	Evening (18:00–23:00h)	Night and early morning (23:00h–07:00h)
Percentage of trips: 22.1	17.7	22.4	24.2	13.6
	Coef. and Sig.	Coef. and Sig.	Coef. and Sig.	Coef. and Sig.
Working arrangements
Only from workplace (ref.)
Only from home	−0.242	−0.549***	−0.968***	−0.794***
Worked from home and workplace	−0.352**	−0.431**	−0.315*	−0.623***
Worked from elsewhere and home or workplace	−0.646**	−0.849***	−0.818***	−1.319***
Household income
Less than Can$20,000 (ref.)
Can$20,000–$39,999	0.15	0.248	0.107	0.348
Can$40,000–$59,999	0.23	0.335*	0.067	0.322
Can$60,000–$99 999	0.473*	0.611***	0.349	0.562**
Can$100,000 or more	0.618**	0.654***	0.431*	0.536**
Women	0.214**	−0.06	−0.194**	−0.487***
Age
15–24 (ref.)
25–34	0.658***	0.516***	0.251*	0.164
35–44	0.583***	0.455***	0.104	0.165
45–54	0.457***	0.379**	−0.047	0.083
55–64	0.282	0.348*	−0.143	0.13
Trip purpose
Work (ref.)
Household tasks	−3.300**	0.441	1.804***	−0.323
Child related	−0.135	0.205	0.930***	−0.787***
Purchases and services	−2.321***	−0.069	0.857***	−2.186***
Restaurants	−2.275***	−0.928***	0.717***	−0.861***
Education	0.083	−0.712**	0.505	−1.660*
Others	−1.590***	0.027	1.916***	−0.089
Visit family and friends	−2.002***	0.192	2.100***	0.481*
Has children	0.079	0.052	−0.064	−0.061
Lives in single family home	−0.077	−0.087	−0.067	−0.260**
Urban (vs) rural	−0.167	−0.083	−0.279**	−0.097
Survey day
Weekday (ref.)
Saturday	−0.634***	−0.664***	−0.685***	−0.238
Sunday	−0.931***	−0.799***	−0.701***	−0.581***
Constant	−0.568**	−0.543**	−0.540**	−0.145
Observations	16,706
Log likelihood(equal likely)	−37,201,427
Log likelihood(constant only)	−36,746,541
Log likelihood(full model)	−33,543,702
Chi square	2017.41
Significance	0.000
McFadden’s Pseudo R²	0.0872

Notes: *p < 0.05, **p < 0.01, ***p < 0.001

Because a reference category is found both in the dependent and independent variables, results should be interpreted following these examples: the relative log odds of motorised travel during morning peak versus midday will decrease by a factor of 0.451 when changing arrangement status from working from the workplace only to working from both home and the workplace. Another example would be that the relative log odds of travelling during afternoon peak versus midday will decrease by a factor of 0.523 if changing arrangement status from working from the workplace only to working from home only. With the important exception of home-working and morning peak trips (discussed below), all other telecommuting arrangements were associated with reduced trips at all periods of the day, as compared with mid-day trips for those working from the workplace.

A number of results give credence to this modelling strategy. For example, child related, purchase and service trips as well as trips to restaurants were more likely to take place during the evening and less likely to take place at night, as compared with work trips. On weekend days, respondents were less likely to take trips than on weekdays and less likely to take trips at all times as compared with mid-day trips.

To facilitate interpretation of the results, predicted probabilities of travelling at both peak hour times are graphed for working arrangements in Figure 2. All three telecommuting working arrangements were associated with lower probabilities of afternoon peak travel and generally of morning peak travel. Working only from home, however, was associated with the highest probability of morning peak travel, which we explain with three features of this category: First a higher percentage of them have children that may need to be driven to school, 36.6% versus 30.8% for office workers (results similar to those of Turcotte, 2010). Examination of trips by purpose confirms that home workers had higher shares of child related trips and education trips during morning peaks than other groups. Second, a higher percentage of home workers were surveyed on weekend days (15.9% and 14.6% on Saturdays and Sundays versus 5.6% and 5.2% for office workers), for which peak hour travel is typically less of an issue. Third, the more flexible schedule of homeworkers may translate in the ability to schedule personal appointments on working days, but may still require peak period travel (e.g. medical, banking).

Figure 2.

Predicted probability of morning and afternoon peak hour travel by working arrangements, with all other variables at their means (weighted estimates).

Limitations

A few limitations are worth mentioning. First, the time use module of the GSS provides information on a single day. While it allows directly linking telecommuting to travel activities during that day, there is a possibility that telecommuting days may be used to conduct more leisure activities or required errands than office-based workdays would be. This in fact makes the results of this analysis more conservative because it makes a telecommuting-caused reduction in travel less likely on the survey day, as compared with an assessment of change over a full week. Nonetheless, authors have suggested that short term studies may fail to account for indirect and long term relationships (Choo and Mokhtarian, 2007). Second, telecommuting may lead to a reallocation of household members’ responsibilities (Kim, 2016; Mokhtarian et al., 1995) but this survey only assessed the trips of one household member. Third, motorised travel time can be influenced by traffic congestion. The same trip taken in congested conditions will usually take longer. We were not able to account for this. Fourth, the analysis enabled the identification of the variety of locations used for work and the prevalence of combining these locations into out of home telecommuting arrangements, but sample sizes were too small to explore these individually. The variety of potential work situations encountered limits our ability to provide a detailed categorisation and reflects the complexity of modern working arrangements. The criteria identified to develop the work arrangements categorisation have been explored in considerable details and various sensitivity analyses have been conducted (including overtime, excluding weekend cases, exploring minimal thresholds for duration of work episodes). Fifth, the peak hour travel model was sensitive to the categorisation of time periods. We selected conservative estimates of peak periods to fit with the variety of locations but it may not accurately represent peak periods everywhere. Sixth, with the limited amount of information explained by the models, it is possible that included variables in the models are partly capturing the influence of unobserved variables with which they are correlated. Adding these variables could impact the currently estimated relationships with working arrangements. Finally, the absence of location data precludes an analysis of variations within and between metropolitan areas. While the data was more than 11 years old at the time of publication, the modest changes in telecommuting rates before 2008 (Turcotte, 2010) suggest that this is likely not problematic.

Conclusion

While telecommuting is still often presented as a sustainable travel policy, its use remains limited and its effects on reducing overall travel time and peak period travel is dependent on the specific working arrangement of individuals. Our data showed that working from home only was the most popular (5.6% of participants), followed by part day telecommuting from home (5.0%) and working from another location with or without home or office work (2.2%). Less than 2.3% of all non-overtime work episodes were conducted at other locations than the home or workplace.

The analyses showed, without ascertaining causality, that for the three sustainable transportation outcomes explored in this paper – reduction of overall time spent travelling, use of non-motorised transportation for at least 30 minutes, and reduction in peak hour travel – most of the relationships were identified when a person worked only from home. This working arrangement was associated with a reduction in overall travel time (by 13 minutes on average), higher odds (77%) of meeting physical activity recommendations through non-motorised travel, and a reduced probability of taking trips during afternoon peak travel periods. Morning peak travel was however not avoided, partly because of child related travel. Assuming that most work trips typically take more than 6.5 minutes each way, results suggest that a rebound effect is likely occurring owing to more non-work travel, but it does not fully outweigh reduced work-related travel. While a higher share of respondents surveyed on weekend days reported working only from home, the relationships were identified on weekdays for 69.5% of them. Save for the appeal of reduced peak hour travel, part day home working seems to have a weak relationship with the three sustainable travel outcomes. For telecommuting policies to be consistent with sustainable travel, the promotion of full day home working in telecommuting arrangements seems warranted and working in other locations may generate more travel, albeit outside of peak periods.

Footnotes

Funding

Funding was received from the Social Science and Humanities Research Council of Canada (SSHRC) 430-2012-0399.

References

Akyeampong

Nadwodny

(2001) Evolution of the Canadian workplace: Work from home. Perspectives on Labour and Income 2(9): 30–36.

Alizadeh

Sipe

(2013) Impediments to teleworking in live/work communities: Local planning regulations and tax policies. Urban Policy and Research 31(2): 208–224.

Atkyns

Blazek

Roitz

(2002) Measurement of environmental impacts of telework adoption amidst change in complex organizations: AT&T survey methodology and results. Resources Conservation and Recycling 36(3): 267–285.

Béchard

(2011) General social survey cycle 24: Time-stress and well-being - public use microdata file documentation and user’s guide. Product No. 12M0024X. Social and Aboriginal Statistics Division. Ottawa, ON: Statistics Canada, 778 pp.

Béchard

Marchand

(2006) General Social Survey Cycle 19: Time Use (2005) – Public Use Microdata File Documentation and User’s Guide. Product No.12M0019-GPE: 970. Ottawa, ON: Statistics Canada.

Bussière

Lewis

(2002) Impact of telework and flexitime on reducing future urban travel demand: The case of Montreal and Quebec (Canada), 1996–2016. Urban Transport Viii: Transport and The Environment in the 21st Century. 12: 279–288.

Canadian Society for Exercise Physiology (CSEP) (2012) Canadian Physical Activity Guidelines and Canadian Sedentary Behaviour Guidelines: Your Plan to Get Active Every Day. Ottawa, ON: Canadian Society for Exercise Physiology (CSEP), 29 pp.

Choo

Mokhtarian

(2007) Telecommunications and travel demand and supply: Aggregate structural equation models for the US. Transportation Research Part A: Policy and Practice 41(1): 4–18.

Downs

(1992) Stuck in Traffic Coping with Peak-Hour Traffic Congestion. Washington, DC: Brookings Institution; Cambridge, MA: Lincoln Institute of Land Policy.

10.

Gurstein

(2001) Wired to the World Chained to the Home: Telework in Daily Life. Vancouver, BC: UBC Press.

11.

Haddad

Lyons

Chatterjee

(2009) An examination of determinants influencing the desire for and frequency of part-day and whole-day homeworking. Journal of Transport Geography 17(2): 124–133.

12.

Helminen

Ristimäki

(2007) Relationships between commuting distance, frequency and telework in Finland. Journal of Transport Geography 15(5): 331–342.

13.

Hilbrecht

Shaw

Johnson

. (2008) ‘I’m home for the kids’: Contradictory implications for work–life balance of teleworking mother. Gender, Work and Organization 15(5): 454–476.

14.

Kim

S-N

(2016) Two traditional questions on the relationships between telecommuting, job and residential location, and household travel: Revisited using a path analysis. The Annals of Regional Science 56: 537–563.

15.

Kim

Mokhtarian

Ahn

(2012) The Seoul of Alonso: New perspectives on telecommuting and residential location from South Korea. Urban Geography 33(8): 1163–1191.

16.

Lachapelle

Noland

(2012) Does the commute mode affect the frequency of walking behavior? The public transit link. Transport Policy 21(0): 26–36.

17.

Lachapelle

Frank

Saelens

. (2011) Commuting by public transit and physical activity: Where you live, where you work, and how you get there. Journal of Physical Activity & Health 8(Suppl. 1): S72–82.

18.

Lari

(2012) Telework/workforce flexibility to reduce congestion and environmental degradation? Procedia – Social and Behavioral Sciences 48: 712–721.

19.

Litman

(2011) If Health Matters: Integrating Health Objectives in Transportation Planning. Victoria, BC: Victoria Transport Policy Institute, Victoria, BC, Canada, pp. 1–42.

20.

Lucove

Huston

Evenson

(2007) Workers’ perceptions about worksite policies and environments and their association with leisure-time physical activity. American Journal of Health Promotion 21(3): 196–200.

21.

Mitomo

Jitsuzumi

(1999) Impact of telecommuting on mass transit congestion: The Tokyo case. Telecommunications Policy 23(10/11): 741–751.

22.

Mokhtarian

(1991) Telecommuting and travel – State of the practice, state-of-the-art. Transportation 18(4): 319–342.

23.

Mokhtarian

Varma

(1998) The trade-off between trips and distance traveled in analyzing the emissions impacts of center-based telecommuting. Transportation Research Part D: Transport and Environment 3(6): 419–428.

24.

Mokhtarian

Handy

Salomon

(1995) Methodological issues in the estimation of the travel, energy, and air-quality impacts of telecommuting. Transportation Research Part A – Policy Practice 29(4): 283–302.

25.

Mokhtarian

Salomon

. (2005) Measuring the measurable: Why can’t we agree on the number of telecommuters in the U.S.? Quality & Quantity 39: 423–452.

26.

Moos

Skaburskis

(2007) The characteristics and location of home workers in Montreal, Toronto and Vancouver. Urban Studies 44(9): 1781–1808.

27.

Moos

Skaburskis

(2008) The probability of single-family dwelling occupancy comparing home workers and commuters in Canadian cities. Journal of Planning Education and Research 27(3): 319–340.

28.

Moos

Andrey

Johnson

(2006) The sustainability of telework: An ecological-footprinting approach. Sustainability: Science, Practice, & Policy 2(1): 3–14.

29.

Nelson

Safirova

Walls

(2007) Telecommuting and environmental policy: Lessons from the ecommute program. Transportation Research Part D 12(3): 195–207.

30.

Pendyala

Goulias

Kitamura

(1991) Impact of telecommuting on spatial and temporal patterns of household travel. Transportation 18(4): 383–409.

31.

Pratt

(2000) Asking the right questions about telecommuting: Avoiding pitfalls in surveying homebased work. Transportation 27(1): 99–116.

32.

Pratt

(2002) Teleworkers, trips, and telecommunications – Technology drives telework – But does it reduce trips? Transportation Planning and Analysis Transportation Research Record 1817: 58–66.

33.

Rietveld

(2011) Telework and the transition to lower energy use in transport: On the relevance of rebound effects. Environmental Innovation and Societal Transitions 1(1): 146–151.

34.

Saxena

Mokhtarian

(1997) The impact of telecommuting on the activity spaces of participants. Geographical Analysis 29(2): 124–144.

35.

Schweitzer

Duxbury

(2006) Benchmarking the use of telework arrangements in Canada. Canadian Journal of Administrative Sciences 23(2): 105–117.

36.

Secrétariat du Conseil du Trésor du Canada (1999) Politique de Télétravail. Available at: http://www.tbs-sct.gc.ca/pol/doc-fra.aspx?section=text&id=12559.

37.

Stewart

(2013) Tobit or not Tobit? Journal of Economic and Social Measurement 38(3): 263–290.

38.

Transport Canada (2007) Telework in Canada. Case Studies in Sustainable Transportation. Urban Transportation Showcase Program. Ottawa: Transport Canada, Case study 52: 5.

39.

Transport Canada (2011) Transportation Demand Management for Canadian Communities: A Guide to Understanding and Delivering TDM Programs. Ottawa: Transport Canada.

40.

Transportation Research Board and Institute of Medicine of the National Academies (TRB-IOM) (2005) Does the built environment influence physical activity? Examining the evidence. TRB Special Report 282. Washington, DC: National Academy of Sciences, 248 pp.

41.

Turcotte

(2008) L’opposition ville/banlieue: Comment la mesurer? Tendances sociales canadiennes 85: 2–20.

42.

Turcotte

(2010) Working at home: An update. Tendances sociales canadiennes 91: 3–11.

43.

Vanoutrive

Van Malderen

Jourquin

. (2010) Mobility management measures by employers: Overview and exploratory analysis for Belgium. European Journal of Transport and Infrastructure Research 10(2): 121–141.

44.

Zhu

(2012) Are telecommuting and personal travel complements or substitutes? The Annals of Regional Science 48(2): 619–639.