Will Telework Reduce Travel? An Evaluation of Empirical Evidence with Meta-Analysis

Abstract

Telework emerged in the 1970s with the advent of Information and Communication Technologies (ICT) as a potential substitute for commuting trips and an answer to avoid congestion. While early studies supported this substitution effect, subsequent research has presented contradictory findings, with some studies demonstrating complementary effects and increased travel distances, while others show a reduction in travel or mixed results. These discrepancies may arise from methodological differences in study design, sampling, and modeling approaches. To analyze these factors, a systematic literature review complemented by three meta-analyses was developed. OLS and GLS-RE models were built to measure telework impacts on the number of trips (total and by purpose), commuting distance, and total distance traveled. Our research suggests that while telework reduces commuting and business trips, particularly for full-time teleworkers, it may increase commuting distances. Total distance traveled presents mixed results, heavily dependent on research design. By identifying these patterns, we outline methodological directions for future research, including improved sampling strategies, advanced modeling techniques, and rigorous control variable selection.

1. Introduction

Demand management strategies seek to reduce transport externalities associated with urban mobility, such as dependence on private motorized transport, high congestion levels, energy consumption, and pollutant emissions. In the 1970s, Information and Communication Technologies (ICTs) emerged, allowing for new and flexible work arrangements, such as telework [1,2,3]. Work from home (WFH) was seen as an opportunity to reduce, or even eliminate, commuting trips, leading to shorter weekday distances traveled and less congestion during peak hours [4,5]. Since then, researchers highlighted the potential of telework as a travel demand management strategy and studied its effects on travel patterns.

However, while earlier studies analyzing telework effects on travel behavior have generally supported the expectations towards a substitution effect associated with working from home, researchers remain divided on the issue. Over time, studies have indicated that telework might also have complementary or ambiguous effects on travel [6,7,8,9,10,11]. They tend to suggest that teleworkers may increase distances traveled [6,12,13], travel more by car [14,15] or engage in more active modes [13,16,17,18]. Additionally, teleworkers tend to travel more for leisure [13,19,20], non-work, and business-related purposes [9]. The relationship between commuting distance and telework is still controversial. Some researchers claim that telework could incentivize workers to live farther away from their jobs [13], while others point out that it is primarily a response to long commutes [2,12]. Moreover, telework is associated with longer commutes and suburban residence locations [13]. Even so, there are also recent findings supporting that telework may also lead to reductions in travel and congestion levels [17,19,21].

Telework is relatively recent and mainly a niche practice prior to 2020 [22]. However, the COVID-19 pandemic outbreak, along with lockdowns and travel restrictions, made working from home the most feasible and widely adopted alternative at that moment and changed the perceptions about telework [23]. Since then, research evaluating the impact of telework on travel behavior has shown significant decreases on travel with a greater adoption of telework [21]. In addition to substantially reduced mobility (during curfews), modal shares have changed [23]. A portion of the users avoid public transport because of the concentration of people, opting for car trips, carpooling or taxi services, or switching to walking and cycling modes for short trips [23]. Studies that focused on the mode split corroborate previous findings, revealing that car and active modes use increased [19,20], while public transport use saw a marked reduction [16,17,18].

Furthermore, the pandemic has accelerated the widespread acceptance of telework, demonstrating to both companies and employees that working from home is feasible for many, putting to test some barriers to the adoption of telework [23], such as productivity reduction and social interaction [24]. Additionally, some advantages came to the fore, as teleworkers could save time and money by reducing the number of commuting trips [25]. However, this reduction may be offset by longer commutes in the medium and long term [26]. Moreover, the increase in telework frequency could change the location patterns of households and increase sprawl, change the total amount of travel by different transport modes, transform activity spaces, and impact the attainment of more sustainable urban mobility patterns.

Although telework adoption and frequency have declined after the pandemic, current levels remain substantially higher than pre-2020 baselines. Telework, mainly in the hybrid regime, will likely continue to be a common work arrangement in the future. Given that stronger interest and engagement in telework can amplify its impact on travel behavior—and that existing research presents mixed results—it is essential to understand the medium- and long-term effects of telework adoption on travel patterns. Such understanding can offer valuable insights for policymakers and planners aiming to shape future mobility systems. Moreover, while telework has the potential to reduce commuting, it may also lead to unintended consequences such as urban sprawl, increased non-work travel, and reshaping of individuals activity spaces. These changes could hinder progress toward more sustainable urban mobility patterns if not properly managed. In the face of the mixed results of decades of research, this study conducts a systematic literature review paired with meta-analysis to examine how methodological and contextual characteristics of existing studies influence their outcomes. This work aims to establish methodological guidelines for future empirical investigations of telework’s impact on travel behavior by systematically analyzing how study characteristics influence research outcomes. Our research examines several critical dimensions, such as data collection procedures, research design and methods, sampling strategies and sample size, modeling approaches, and control variable selection.

The paper is structured as follows: Section 2 shows the methodology of the systematic literature review, detailing the scope of the meta-analysis and its advantages and limitations. Section 3 provides a scoping review of the empirical evidence concerning the effects of telework on travel behavior. Section 4 presents the results and discussion of the meta-analyses. Finally, Section 5 presents the main conclusions.

2. Methodology

A systematic literature review was conducted to examine the effects of telework on travel behavior. Systematic literature review makes a comprehensive, methodological, and rigorous synthesis of the literature on telework and travel behavior. Given the heterogeneous and sometimes contradictory results in previous research, this method provides a structured approach to aggregating findings and identifying key factors that may influence reported outcomes, which can guide future studies. PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 protocol, used here, ensures that the process can be replicable, minimize bias, and improve the quality of reviewing conclusions [27]. Nonetheless, there are limitations. First, it is susceptible to publication bias, since researchers tend to publish significant outcomes. Second, the difference in sample methods, modeling approaches, and study design across papers limits the comparability of findings, which will affect our meta-analysis.

The review summarized the factors contributing to the diversity of outcomes found in the studies. This was achieved through two main approaches: (i) conducting a scoping review to analyze the studies in a broad context and (ii) selecting a subgroup of papers to perform a meta-analysis. Both the systematic literature review and the meta-analysis were carried out and reported according to the PRISMA 2020 guidelines [27].

2.1. Literature Review

The systematic literature review was conducted between December 2024 and February 2025 on two of the most relevant databases in transportation literature, SCOPUS and Web of Science, to identify studies evaluating telework adoption’s impacts on travel behavior. The search included the following strings within the title, abstract, and keywords: (“travel behavior” OR “distances traveled” OR “mode choice” OR “travel patterns” OR “trips purpose” OR “modal share” OR “commuting trips” OR “commuting distance” OR “commuting travel” OR “number of trips” OR “kilometers traveled” OR “trip scheduling” OR “total amount of travel” OR “residential location” OR “residential preferences”) AND (“telework” OR “teleworking” OR “telecommuting” OR “work from home” OR “home-based telework” OR “home-based teleworking” OR “home-based telecommuting”). The studies included in the review came from peer-reviewed journals published in English and encompass publications from all time periods.

A total of 982 papers were initially obtained—815 from SCOPUS and 167 from Web of Science, with 247 identified as duplicates. This left 735 papers for initial screening (based on title, abstract, and keyword review), of which 199 were selected for in-depth reading. The inclusion criteria required that studies (1) be empirical investigations of telework’s effects on travel behavior; (2) include data collection and statistical modeling; and (3) define telework specifically as working from home. Additionally, the conclusions had to be reported in terms of travel metrics such as the number of trips, commuting distance, distance traveled, travel frequency, modal splits, or other mobility-related variables. In the end, 85 studies met the criteria and were included in the review. The screening process followed the updated PRISMA 2020 flow diagram, as shown in Figure 1.

The 85 studies included in this systematic literature review use various metrics to report travel, with some studies employing more than one metric. However, not all studies provided quantitative estimates suitable for meta-analysis. Section 3 presents and discusses the results of the systematic literature review to map the existing literature on telework’s effects on travel behavior. Our descriptive analysis examines both how evidence has been generated and explores intrinsic study characteristics that may influence conclusions about whether telework substitutes for, complements, or show no measurable effect on travel. While we do not evaluate the robustness of evidence or quantify effects, we identify key factors to inform our subsequent meta-analysis.

2.2. Meta-Analysis

While a systematic literature review uses a rigor protocol to ensure reproducible study identification, the subsequent meta-analysis quantitatively integrates empirical findings across these studies through statistical methods to generate evidence-based conclusions. This two-stage process combines systematic screening with analytical synthesis to overcome individual studies’ limitations, establish robust literature patterns, and identify factors influencing results—ultimately guiding future research [28,29].

Adherence to a rigorous protocol is essential to ensure methodological transparency and consistency. The systematic literature review and the meta-analysis must include at least two databases to minimize bias. Empirical evidence must include numerical results that are quantitatively comparable, i.e., all measurements must either use comparable units or units that might be converted into a comparable format [28]. The sample size may sometimes be smaller than the number of published papers on the topic of interest because some studies do not provide adequate data or sufficient information to calculate effect sizes, which limits the number of observations [30]. Additionally, the process must include eligibility criteria for selecting studies and exclusion criteria for discarding those not meeting the necessary standards [28].

Using meta-regression is one possibility for conducting a meta-analysis, which employs a set of explanatory variables (categorical or discrete) to measure the observed variation in the effect sizes [28]. The choice of adequate explanatory variables is based on the literature review [29]. A critical decision involves choosing between including either a single estimate per study—ensuring equal weighting but limiting sample size—or incorporating all available estimates, which increases sample size but may bias results toward studies with more observations.

Three meta-analyses were developed from papers selected from the total sample of 85 studies. The first meta-analysis aims to evaluate the effects of teleworking adoption on the total number of trips and/or the number of trips by purpose (commuting trips, business trips, non-work trips or others). A total of 21 papers were found, but just 13 studies had reported the outcome properly for the measurement of the effect size. In total, the first meta-analysis has 42 estimates. The second meta-analysis, focused on commuting distance, found only 18 papers, resulting in a small sample of 8 papers and 11 observations. For the third meta-analysis, 22 studies about total distance traveled were found in the literature review. However, only 10 papers were eligible for the meta-analysis, resulting in a sample of 16 estimates.

3. Systematic Literature Review Results and Discussion

Telework emerged as a potential tool for managing travel demand with the rise of ICTs in the 1970s [4]. Only a few empirical studies were conducted in the initial decades following the introduction of the concept. The first empirically based papers date back to 1990. Figure 2 presents the chronological distribution of the reviewed studies. The data show the limited empirical research on telework effects on travel behavior prior to 2005, followed by gradual growth after this date. A particularly substantial increase in publications becomes evident after 2020.

Initially, it was believed that telework could serve as a substitute for travel [1,2,31]. Early studies reported a significant reduction in the number of commuting trips and total distance traveled by teleworkers, particularly during peak hours [5,31,32], and a decrease in household travel [31]. However, since around 2010, research has begun to highlight some nuances in the effects of telework on travel patterns, with several studies concluding that telework could increase travel or that its effect was neutral at best [7,8,11]. Some studies indicate that teleworkers exhibit higher accumulated distances traveled when compared to non-teleworkers [9,13]. Teleworkers make more non-work and business-related trips [9]. The negative impacts on travel behavior are also related to the fact that teleworkers travel more by car [14,15] and less by public transport [15,18], although they use active modes more frequently [16,18,33]. However, active modes are mainly used for leisure trips [13,19,20]. On the other hand, several studies also concluded that telework reduces travel [17,21]. Particularly during the pandemic, many studies indicated that telework reduced the number of trips and distances traveled [24,34,35,36]. However, these conclusions may have been biased by the travel restrictions and increased frequency of working from home during that period. Recent studies suggest that full-day telework can reduce travel by all modes (especially by motorized ones), while part-day telework may increase travel [37]. Additionally, some findings point to the spatial implications of telework: teleworkers may choose to live farther from their workplace, potentially improving overall distance travel for frequent teleworkers (at least three or more days per week) [38].

Figure 3 shows the conclusions of studies published in peer-reviewed journals about the effects of telework on travel behavior since 1990. In more than 43% of these studies (37 articles), telework adoption has no impact on travel behavior, or the measured metrics yield ambiguous results, making it difficult to determine the direction of the total effects. A neutral impact is observed when telework adoption does not significantly change travel behavior. Ambiguous conclusions refer to studies that found a substitution effect in one travel aspect while observing a complementary effect in another. Nevertheless, approximately 19% of the studies (16 articles) concluded that telework has a complementarity effect on travel (i.e., leading to an increase in travel). In contrast, over 37% of the publications (32 articles) suggest that telework can indeed substitute travel by eliminating or reducing trips.

There is a clear temporal relationship between the number of publications over time, the outbreak of the COVID-19 pandemic, and the results of the studies, as we can see in

Table 1. Difference in results before and after the COVID-19 pandemic.

	Before 2020	After 2020	Pre-Pandemic Data	During Pandemic Data
Decreases travel	8 (21.1%)	24 (51.1%)	11 (20.8%)	21 (65.6%)
Increases travel	9 (23.7%)	7 (14.9%)	14 (26.4%)	2 (6.3%)
Ambiguous/Neutral	21 (55.3%)	16 (34.0%)	28 (52.8%)	9 (28.1%)
Total	38 (44.7%)	47 (55.3%)	53 (62.3%)	32 (37.7%)

. Most papers published since 2020 concluded that telework adoption decreases travel (51.1%). However, studies that analyzed travel behavior with pre-pandemic data tend to indicate more ambiguous or neutral findings (52.8%). In contrast, those using “during pandemic data” mainly indicate that telework decreases travel (65.6%), which may reflect the measures implemented during the COVID-19 pandemic. Despite these temporary effects, some studies suggest limited long-term changes in modal split and emphasize the relevance of professional category education level in continuing telework after the pandemic [21]. Moreover, there are concerns regarding equity and environmental justice, since lower-income groups had fewer opportunities to telework and experienced increased transport costs and emissions during the pandemic [39].

The methodological approaches across studies exhibit substantial variation in survey design. While some employ generic questions about telework engagement, others specifically investigate the telework–travel relationship. The adoption of travel diaries emerged in 1990s publications, though with notable variability in duration, ranging from one to seven days. Sample sizes also range from 30 respondents to about 300,000 observations. The statistical approaches include simple comparison of means, regression-based models, path analysis and Structural Equation Models, and before and after analysis. More than 90% of the studies include socioeconomic control variables (e.g., gender, age, and occupation are generally considered), but few considered attitudinal variables and preferences about telework in the analysis. The relationship between the number of trips and whether teleworkers work from home part- or full-day is rarely explored [37].

Table 2. Difference in conclusions and main factors.

Factors		Decreases Travel	Increases Travel	Ambiguous/Neutral
Focus of the study	Individual-level impacts	24 (75.0%)	6 (37.5%)	26 (70.3%)
	Household-level impacts	6 (18.8%)	10 (62.5%)	11 (29.7%)
	Telework impacts (general)	2 (6.2%)	0 (0.0%)	0 (0.0%)
Survey	Specific to the study	13 (40.6%)	3 (18.8%)	8 (21.6%)
	Regional/National	19 (59.4%)	13 (81.2%)	29 (78.4%)
Travel diary	No/general question	21 (65.6%)	7 (43.8%)	16 (43.2%)
	One day	1 (3.1%)	5 (31.2%)	10 (27.0%)
	More than one day	10 (31.3%)	4 (25.0%)	11 (29.8%)
Telework measure	General question	6 (18.8%)	2 (12.5%)	4 (10.8%)
	Binary variable	9 (28.1%)	6 (37.5%)	17 (46.0%)
	Ordinal variable	17 (53.1%)	8 (50%)	16 (43.2%)
Telework regime	Does not specify	29 (90.6%)	15 (93.8%)	31 (83.8%)
	Full and part-day telework	3 (9.4%)	1 (6.2%)	6 (16.2%)
Telework directly related to the day of travel diary	Yes	5 (15.6%)	1 (6.2%)	3 (8.1%)
	No	27 (84.4%)	15 (93.8%)	34 (91.9%)
Self-selection control	Yes	6 (18.8%)	6 (37.5%)	5 (13.5%)
	No	26 (81.2%)	10 (62.5%)	32 (86.5%)
Includes attitudes about telework	Yes	13 (40.6%)	3 (18.8%)	8 (21.6%)
	No	19 (59.4%)	13 (81.2%)	29 (78.4%)

shows the main factors considered in the studies and how they vary and indicate different conclusions.

Studies showing that telework adoption decreases travel are mainly focused on individual-level impacts (75%). This effect (i.e., telework adoption decreases travel) is also more frequently reported in more extensive questionnaires (59.4%), in surveys explicitly conducted for the study (40.6%), and in travel diaries including no question/general question about travel patterns (65.6%). Additionally, within the studies reporting telework adoption decreases travel, 90.6% did not specify the telework regime (vs. reporting full- and part-day telework), and in 84.4%, telework frequency was not directly related to the day trip reported. Finally, 81.2% did not consider self-selection control, and 59.4% did not include attitudes about telework.

Studies that reported travel increases related to telework adoption mainly considered household-level effects (62.5%), assessing whether telework adoption influenced not just the teleworker’s trips but also those of household members. Research examining the relationship between household travel and telework suggests that working from home might lead to an overall increase in household trips, which could outweigh or compensate for the decrease in travel by the teleworkers themselves [7]. Most studies that reported that travel increases with telework adoption (81.2%) used comprehensive surveys, such as national surveys, while roughly half considered a question about telework frequency. In all, 93.8% did not ask about the regime of telework adoption (full- and part-day). Self-selection (62.5%) and attitude variables about telework (81.2%) are also not common in these papers, and the same applies to the relationship between the day of telework practice and the trips made on the respective day (93.8% of the studies do not consider it).

The findings on ambiguous or neutral effects of telework are primarily focused on teleworkers’ trips (70.3%) and used regional or national surveys (78.4%). Including travel diaries was dominant—27% applied a one-day travel diary, and 29.8% more than one day. Most surveys had specific questions about telework, with questions about whether or not individuals teleworked (46%) and/or how often they teleworked (43.2%). However, they did not specify if it was full- or part-day telework (91.9%). Self-selection controls (86.5%) or attitudinal variables about telework (78.4%) were generally not included.

The use of attitudinal factors and self-selection control has gained importance in recent years.

Table 3. Difference in conclusions of papers using attitudinal variables and self-control.

	Includes Attitudes About Telework	Does Not Include Attitudes About Telework	Controls For Self-Selection	Does Not Control For Self-Selection
Decreases travel	13 (54.2%)	19 (31.1%)	6 (35.3%)	26 (38.2%)
Increases travel	3 (12.5%)	13 (21.3%)	6 (35.3%)	10 (14.7%)
Ambiguous/Neutral	8 (33.3%)	29 (47.5%)	5 (29.4%)	32 (47.1%)
Total	24 (100%)	61 (100%)	17 (100%)	68 (100%)

presents the studies that included attitudinal variables about telework in their analysis (24 out of 85) and those that examined whether self-selection played a role in telework-related travel changes (17 of 85 studies).

Most studies that included attitudes have indicated travel reduction (54.2%). The analysis of attitudinal variables could prove critical for evaluating perceptions regarding telework’s benefits and drawbacks [39] while also elucidating the role of teleworker satisfaction in adoption patterns [40,41]. Among the studies that do not control for self-selection, 47.1% reported ambiguous or neutral findings, supporting the idea that accounting for self-selection is critical. Self-selection bias can directly affect whether observed behavioral changes reflect actual telework effects or pre-existing differences among teleworkers [17].

4. Meta-Analyses Results and Discussion

This work presents three meta-analyses using meta-regressions: first, we evaluated the effect size of the number of trips (total and by purpose); second, the commuting distance; and third, the total distance traveled, all as a function of telework engagement. The inclusion criteria prioritize pre-pandemic studies to avoid potential confounding effects from COVID-19 containment measures—including travel restrictions, curfews, and atypical telework adoption patterns—which may distort the observed relationship between telework and travel behavior.

Additionally, the results must be quantitative and to be able to be convertible in a percentage variation to be comparable. The definition of telework must be limited to home-based telework. Only individual-level travel impacts were included (i.e., household trips were excluded in this analysis to avoid introducing heterogeneity). We also included both significant and non-significant estimates.

The empirical evidence included in these meta-regression models, presented in Equation 1, contains one or more estimates per study, leading us to consider within-study correlation using OLS (Ordinary Least Squares) and GLS-RE (Generalized Least Squares with Random Effects) models. OLS estimator assumes that the observations are independent and identically distributed, not considering correlated observations in the same study. Otherwise, GLS-RE uses random effects to account for the heterogeneity across studies and correlations between observations by the same study. This model is a particularly interesting meta-analysis, which includes more than one estimate per study.

$${\widehat{\eta }}_{ij}={\eta }_0+{\displaystyle {\sum }_{k=1}^K}{\beta }_k{D}_{ij,k}+{\mu }_j+{\epsilon }_{ij}$$

(1)

where i and j are the estimate and its study, respectively. ${\widehat{\eta }}_{ij}$ is the dependent variable (percentage change in the number of trips or commuting distance or total distance traveled due to telework adoption), ${\eta }_0$ is the model constant, $D_{ij,k}$ is a meta-regressor k, ${\beta }_k$ measures its effect on the dependent variable, ${\mu }_j$ is a study-specific term, and ${\epsilon }_{ij}$ is the error term.

The literature review presented in Section 3 guided our choice of the study design characteristics, which enter the meta-regressions as meta-regressors k. We tested the following to understand if they contribute to the model:

• Period of the data collection: we hypothesized that earlier studies tended to conclude that telework reduces travel while more recent ones tend to conclude otherwise;
• Sample size: we hypothesized that studies based on larger samples tend to conclude that telework increases travel since larger datasets may detect marginal behavioral changes across more diverse populations;
• Type of data: whether the questionnaire was applied to volunteers from some organization, whether the questionnaire was developed specifically for the study, or whether it involved more overall data but was not specific, like regional and national studies. We hypothesized that studies including volunteers tend to see telework as more advantageous and tend to be related to travel reduction;
• The study used a control group: to distinguish between studies that considered only teleworkers and those with a control group. We hypothesized that studies that consider control groups are more related to travel increasing;
• Duration of the travel diary: the hypothesis is that more extended travel diaries can be related to travel decreasing since they may capture more variability in travel behavior;
• Measure of telework: whether the questionnaire considered telework frequency or simply asked if respondents work from home. The hypothesis is that telework frequency can better capture the variation of travel behavior, which is related to travel less;
• Telework regime: a small portion of the studies considered part-time telework. We hypothesized that full-day teleworkers are associated with travel decreasing and part-day teleworkers are associated with travel increasing since part-day teleworkers still have to commute to work;
• Telework related directly to the day of travel: to investigate if the small portion of the studies that asked about telework adoption on the specific day of the travel diary had different results from the others but without hypothesizing any effect direction for now;
• Measurement method: more robust methods may capture variabilities in the sample and are related to travel more;
• Self-selection control: studies that considered self-selection may be related to travel increasing;
• Bias control: studies that have bias controls may be related to ambiguous/neutral findings;
• Quality of the estimation: this variable encompasses several other factors, including the study method, whether the study accounted for self-selection and/or control variables, and whether it included a control group (non-teleworkers) to compare the estimates between teleworkers and non-teleworkers.

The following sections contain the results and discussion of the meta-regression models. Appendix A presents tables reporting the studies included and excluded in each meta-analysis, indicating the reason for exclusion.

4.1. Number of Trips Meta-Analysis

This section shows the meta-analysis results of the studies that considered the number of trips (total and/or by purpose) as a function of telework adoption. Of the 21 papers measuring the number of trips, 13 studies reported the outcome properly for estimating the effect size, resulting in 42 observations. In this analysis, we consider the following factors: trip purpose, quality of the estimation (whether the study uses regression models or not), telework regime (namely, whether the study considered full or part-day telework or not), and if telework was related to the same day of trip reported.

Most estimates of the studies are related to the total number of trips (38.1%) and commuting trips (35.7%). Almost half of the observations (45.2%) applied regression models. More than half of the estimates are from studies since 2010 (54.8%). Controlling by partial or full-day telework is not very recurrent in the studies (almost 70% of the studies did not collect data about that). Finally, in 38.1% of the estimates, telework was related to the same day of the trip reported.

Figure 4 presents the forest plot of the estimates of the number of trips by study.

Table 4. Meta-regression results of telework effects on the number of trips.

	OLS	GLS-RE
Variable	Coefficient	p-Value	Coefficient	p-Value
Constant	3.158	0.665	3.165	0.700
Ref.: Total number of trips
Commuting/work trips	−17.547	0.023	17.553	0.033
Business trips	−22.489	0.034	−22.485	0.030
Non-work trips	−4.449	0.631	−4.444	0.645
Ref.: Non-regression models
Regression models	15.494	0.034	15.484	0.076
Ref.: No specification about telework
Part-day teleworkers	7.082	0.460	7.086	0.467
Full-day teleworkers	−37.809	0.000	−37.809	0.000
Ref.: No direct relationship between telework and travel
Telework related to trip day	−19.222	0.013	−19.227	0.014

shows the meta-regression results, the coefficient and p-value from each dummy variable according to its reference. The model has a good fit with a R-squared of 0.644 and an adjusted R-squared of 0.541. The GLS-RE model has a reasonable adjustment: R-squared of 0.644 and adjusted R-squared of 0.588. However, there is a small group variability of 0.199 in the model; this may be a consequence of the small number of groups considered.

The results of the OLS and GLS-RE models are very similar. Teleworkers tend to reduce commuting, work, and business trips compared to the total number of trips. These results are in accordance with some literature conclusions, which indicate that telework adoption has a substitution effect on commuting trips (reducing or eliminating them, especially during peak hours) [5,24,31,32,34,35,36] and has a neutral or complementary effect on business-related or non-work trips, pointing an increasing in the number of trips [47]. Studies indicate mixing results for household travel behavior: a decrease [31] and an overall increase on the number of trips [7]. More trips related to household members can be partly explained by trying to compensate for trips eliminated by teleworkers [7].

The quality of the estimations does influence the conclusions. Using regression models leads to an increase in the number of trips. This may partly reflect the regression-based methods’ ability to adjust for study-level covariates that influence telework adoption and travel behavior (e.g., sample characteristics and methodology), which can reveal higher adjusted estimates compared to unweighted or non-adjusted summaries. This is in accordance with some research that shows that eliminating commuting trips could compensate for other purposes and increase travel [43,49], while other research supports that telework increases commuting trips, daily total work trips, and total non-work trips [6]. It may be an effect that teleworkers tend to distribute their trips over the day to avoid congestion on commuting trips, maintaining non-work trips on the days they commute to chaining trips [32].

An essential effect is found in controlling full-day telework since teleworkers who work from home the full day might reduce the number of trips. These conclusions follow the recent literature, which has separated teleworkers into full and part-day teleworkers [17,44]. Finally, studies that directly link trip diaries and telework adoption indicate a decrease in the number of trips.

4.2. Commuting Distance Meta-Analysis

This section shows the meta-analysis results of the studies that considered the commuting distance as a function of telework adoption. Of the 18 papers, 8 studies reported the outcome properly for estimating the effect size, resulting in 11 observations. The meta-regression model considered the quality of the estimation (i.e., whether the study uses regression models or not), sample size (until 20,000 observations or more), and measure of telework (binary or ordinal variable). The 20,000 observations threshold was selected based on the distribution of available studies and their ability to differentiate between smaller-scale and larger-scale datasets. A total of 72.7% of the estimates indicate an increase in commuting distance. Also, 72.7% of the observations had a sample size of under 20,000 respondents. Since this meta-analysis included more recent studies, we cannot see a correlation between the period and the quality of estimation, which is remarkably diverse. In all, 63.6% of the estimation involves regression models, and 54.5% of the observations consider a binary variable for telework measurement.

Figure 5 presents the forest plot of the estimates of the commuting distance by study.

Table 5. Meta-regression results of telework effects on commuting distance.

	OLS	GLS-RE
Variable	Coefficient	p-Value	Coefficient	p-Value
Constant	2.250	0.935	9.368	0.839
Ref.: Non-regression models
Regression models	82.632	0.143	67.478	0.483
Ref.: Sample size up to 20,000 observations
Sample size upper 20,000 observations	167.235	0.004	174.823	0.004
Ref.: Binary telework variable
Ordinal telework variable	−106.176	0.048	−98.290	0.125

shows the commuting distance meta-regression results: the coefficient and p-value from each dummy variable according to its reference. The meta-regression OLS model has an R-squared of 0.812 and an adjusted R-squared of 0.713, indicating the models are well-adjusted. Also, the GLS-RE model is well-adjusted, with an R-squared of 0.876 and 0.793. The GLS-RE model shows a group variability of 695.47.

The results of the OLS and GLS-RE models for travel distance are similar, with group variability, and the constant not being statistically significant. The results show that teleworkers’ commuting travel distances increase using regression models and larger sample sizes. These findings suggest that teleworkers engage in longer commuting trips—a pattern consistent with prior research since telework is often linked with suburban residential areas, which may influence the longer distances teleworkers travel [10]. On the other hand, more recent findings indicate a strong relationship between working from home and commuting travel in the post-pandemic era, with consistent reductions in commuting time over time [54], which may change teleworkers’ travel behavior in the long-term.

Using ordinal variables to measure telework leads to a decrease in commuting distance. This difference is not significant in the GLS-RE model. Using ordinal measures of telework frequency helps distinguish between frequent and infrequent teleworkers, as the former tend to travel less [9]. The quality of the estimation, although not statistically significant in the models, made the model more parsimonious and converged with the GLS-RE model. This variable can be a sound indicator to consider for future studies.

4.3. Total Distance Traveled Meta-Analysis

This section shows the meta-analysis results of the studies that considered the total distance traveled as a function of telework adoption. Of the 22 papers, 10 studies reported the outcome properly for estimating the effect size, resulting in 16 observations.

The meta-regression model considered the following variables: the quality of the estimation, the presence of a control group, the period in which the study was conducted, and the sample size (up to 15,000 observations or more). In this model, we considered the quality of the estimation to have four levels: (1) if the studies made a simple comparison of means, (2) if studies used econometric models controlled by individuals’ characteristics, (3) if mean comparison between teleworkers was used, and (4) if mean comparison between teleworkers and non-teleworkers was used.

For the period, we also used four levels: (1) publications that used data until 1995, (2) from 1996–2005, (3) from 2006–2015, and (4) from 2016–2025. A total of 62.5% of the estimations were related to data from 2006. The 15,000 observations threshold was selected based on the distribution of available studies and their ability to differentiate between smaller-scale and larger-scale datasets. In total, 68.8% of the estimates had a sample size of less than 15,000 respondents. Finally, one of the variables was related to whether studies used a control group for the analysis and whether the estimate is only about teleworkers’ data was also considered. 62.5% of the observations considered a control group to compare the results with teleworkers.

Figure 6 presents the forest plot of the estimates of the total distance traveled by study.

Table 6. Meta-regression results of telework effects on total distance traveled.

	OLS		GLS-RE
Variable	Coefficient	p-Value	Coefficient	p-Value
Constant	6.500	0.871	6.563	0.869
Ref.: Simple comparison between means
Econometric models	−85.000	0.154	−85.000	0.122
Comparison between teleworkers means	−39.167	0.381	−39.230	0.358
Comparison between teleworkers and control group	−113.000	0.035	−112.446	0.009
Ref.: Only considers teleworkers
Considers Teleworkers and control group	73.000	0.037	73.000	0.009
Ref.: Period between 1988–1995
Period between 1996–2005	89.500	0.037	88.883	0.010
Period between 2006–2015	23.500	0.396	23.437	0.383
Period between 2016–2025	43.500	0.350	43.437	0.335
Ref.: Sample size up to 15,000 observations
Sample size upper 15,000 observations	−92.000	0.033	−92.000	0.011

presents the meta-regressions results for the total distance traveled, specifically the coefficient and p-value from each dummy variable regarding its reference. The meta-regression OLS model has an R-squared of 0.769 and an adjusted R-squared of 0.505, indicating that the model is well-adjusted. Also, the GLS-RE model is well-adjusted, with an R-squared of 0.779 and 0.449. There is a group variability of 63.171 in the GLS-RE model.

The results of the OLS and GLS-RE models for the total distance traveled are also similar, with group variability, and the constant not being statistically significant. The model results show a tendency for increasing travel distance in more recent studies, even though only the period between 1996 and 2005 has a statistically significant effect. Using larger sample sizes is linked with the decrease in total distance traveled. Studies comparing teleworkers to a control group report reduced total travel distances compared to simple mean comparisons. However, using a control group increases reported distances compared to analyses focusing solely on teleworkers. This discrepancy may reflect the methodological rigor of control-group designs, which go beyond basic mean comparisons.

The variability is also seen in the literature. Previous research indicated a reduction in total distance traveled by teleworkers, particularly during peak hours [5,31,32], and a decrease in household travel [31]. Over the years, studies have suggested that teleworkers are actually likely to have longer accumulated distances traveled than non-teleworkers [9,13]. However, more recent studies during the pandemic concluded that teleworkers reduced distances traveled in this period [24,34,35,36]. In addition, full-day teleworkers are more likely to reduce trip frequency, despite teleworkers having longer commuting distances, making fewer trips can decrease the overall travel distance [57]. This is a topic that may impact CO2 emissions [57], making it a particular gap to be studied more deeply in the future.

5. Conclusions

Since the 1990s, research on telework’s impact on travel has produced mixed findings, reflecting its complex relationship with mobility patterns. While some studies suggest telework increases travel distances and leisure trips, others indicate potential reductions in travel and congestion. Before the COVID-19 pandemic, telework adoption was minimal. Its widespread adoption since 2020 has led to shifts in travel behavior, including a decline in public transport ridership. Although telework frequency has decreased post-pandemic, it remains higher than pre-2020 levels, potentially contributing to changes in mobility patterns that could challenge sustainable mobility transitions.

Given this context and the mixed findings found in the literature, this study conducts a systematic literature review and meta-analyses to assess how methodological and contextual factors influence research outcomes. By analyzing crucial study characteristics—such as data collection methods, research design, sample sizes, modeling approaches, and control variables, this work aims to establish methodological guidelines for future investigations into telework’s effects on travel behavior.

A systematic literature review, complemented by three meta-analyses, was conducted for this purpose. The main findings highlight that telework adoption tends to reduce the number of commuting, work, and business trips, particularly for full-day teleworkers, following other studies that found that telework engagement may have a substitution effect on the number of trips by full-day teleworkers while increasing trips by part-day teleworkers, who still have to travel to work [37]. Commuting distance tends to increase with telework adoption because some teleworkers may choose to live in suburban areas, thereby extending the distance between home and work [7,9]. Total distance traveled shows mixed results, being more influenced by research design factors. Some studies measuring the impact of telework on travel distances find an increase among teleworkers [9,13,51], while others report a decrease [5,31,32], including a reduction in overall household travel [31]. The reliability of these estimations depends on factors such as the measurement of telework and travel, sample size, and overall quality of the estimation.

Using a systematic literature review and a reproducible meta-analysis framework, this study employed a robust, transparent, and reliable approach to reviewing existing research and drawing conclusions about the effects of telework on travel patterns. However, this analysis has limitations: since we chose to prioritize pre-pandemic studies in our meta-analysis to avoid potential confounding effects from COVID-19, we did not include more recent studies that, in many cases, use more robust methods and include attitudinal variables and extended travel diaries. Future work will look at the long-term effects of telework on travel behavior, considering the post-pandemic period.

The findings of this study offer several important implications for policy and practice. Since telework is pointed out as a travel demand management tool, and the main findings support that it reduces commuting trips, developing telework policies can significantly contribute to urban mobility planning. Full-day teleworkers tend to reduce the number of trips and use active modes of transport, including using activity spaces closer to their homes. It may alleviate congestion and reduce emissions, supporting the case for institutionalizing hybrid work regimes as part of long-term workforce planning.

Moreover, teleworkers tend to live in suburbs—increasing commuting distance and contributing to urban sprawl—use car more often, and less public transport. For that reason, it is critical to consider investments in sustainable transport infrastructure. Policymakers might anticipate and address potential increases in car dependency by prioritizing investments in suburban public transport and active modes of infrastructure, as well as land-use policies that support transit-oriented development. Also, investing in policies to increase suburbs and neighborhoods, as well as local commerce, can help reduce transport externalities and alleviate traffic in the city.

Furthermore, telework adoption is an opportunity to rethink peak-hour transit service demand, road capacity planning, and investment priorities. As travel patterns become more flexible, transport authorities may need to shift focus from peak congestion mitigation to all-day service reliability and accessibility, especially for non-work-related trips, creating a more sustainable environment.