This work was supported by the Smart Lab LABILITY of the University Gustave Eiffel, funded by the Région Île de France under Grant N° 20012741.

Introduction

Context

1Teleworking (i.e. working from home by using information technology) is a mode of organisation in which the employee carries out his usual working hours away from his employer’s office on a regular and voluntary basis, for functions that could have been carried out in the employer’s office.

2Teleworking may be seen as a tool to regulate mobility reducing commuter travel and the associated congestion, air pollution, energy use and carbon emissions. However, teleworking may have various rebound effects on daily and residential mobility practices (Hostettler Macias et al. 2022). For these authors, the rebound effects correspond to the potential increase in the distance and frequency of journey, as well as the possibility of a residential relocation or multilocal dwelling as a result of the reduction in commuting. The availability of teleworking may encourage people to take additional trips for non-work purposes (Asgari et al. 2016, Budnitz et al. 2020, Nikolaeva et al. 2023), encourage them to move to suburban or rural residential locations that are further from their place of work (de Vos et al. 2018), and/or encourage them to take up employments that are further from their place of residence. In addition, teleworking may influence not only the number but also the travel patterns of individuals and other household members (Kim et al. 2015, Asgari et al. 2016, de Abreu e Silva and Melo 2018a, Caldarola and Sorrell 2022), but also switch them to the use of a private car (Wells et al. 2001, Kim et al. 2015, Asgari and Jin 2022). It was by integrating all motives and travel times into the analyses – and not just based on commuting – that the studies concluded about a gain in terms of greenhouse gas emissions and energy savings were very modest if not zero (Cerqueira et al. 2020, O’Brien and Yazdani Aliabadi 2020). These complex interactions make the overall impact of teleworking uncertain and contribute to explaining the contradictory results of the empirical literature (Hook et al. 2020).

3Before the Covid-19 pandemic, home-based teleworking was a small and mainly occasional practice in France. The DARES (Direction de l’Animation de la Recherche, des Études et des Statistiques), a French government agency, estimated that in 2019 only 7% of employees teleworked, including occasional and informal teleworking and only 3% for what was a regular and legally formalised practice. In 2022, the DARES estimated that 29% of employees were teleworking (Gouyon et al. 2022). This shows a strong increase in the practice of teleworking following the Covid-19 pandemic. Therefore, this new context implies reconsidering the links between teleworking and mobility for at least three reasons. Firstly with the pandemic, teleworking has grown significantly in terms of the share of the workforce involved and also in terms of the number of days of teleworking per week. Secondly it has extended to categories of workers previously little or not concerned, on whom we have no studies in terms of changes in mobility behaviour or residential choices. Thirdly the health crisis has prompted many working people to change their mobility (reduced use of public transport, increased use of active modes but also the car) and their lifestyles (increased use of local shops), practices which could be reinforced (or on the contrary reduced) with the development of teleworking.

Framing, interaction and overlap of life’s spheres

4Although the link between teleworking and mobility has been widely studied in transport research in recent years (Mokhtarian 1991, Andreev et al. 2010), most research focused on commuting practices (e.g. traffic congestion, car travel distances, spatial and temporal trips distribution, air pollution or mode choice) by comparing teleworkers with non-teleworkers (Pendyala et al. 1992, Wells et al. 2001, Mokhtarian 2002, Mokhtarian and Salomon 2002, Zhu 2013). Now, the coping strategies of teleworkers and other household members (e.g. reorganisation of activities according to the days of the week, and the associated travel behaviour) remain insufficiently studied. However, to understand the rebound effects of teleworking on these potential changes in practices, it is necessary to study teleworking with a systemic approach to interpersonal relationships and their activities in the different spheres of their lives (i.e. work, family, social and personal). For example, the work sphere, with its location(s) and schedule, has a structuring role in the daily activity schedule. It may be associated with other travel motives (e.g. shopping, leisure or other personal activities; Cerqueira et al. 2020). In addition, the influence of lifestyles and life courses explains the various experiences (e.g family, leisure, friends, associative involvement). Thus, this ‘system of activities’ considers ‘people’s activities form a system because first, they use resources that are limited in time and in energy, and second, every activity is likely to provide other activities with resources that the person derives from their physical and social environment’ (Vayre and Pignault 2014). We extend this analysis by showing how teleworkers perceived, adapted and organised their relationships with others, and their activities in the different spheres of their lives.

5Methodologically, to capture these complex interactions, many studies suggested analysing teleworking with a holistic approach (Mosa 2011, Asgari et al. 2016, Hostettler Macias et al. 2022, Huang et al. 2023). This is why it also seems obvious for us to compare teleworkers with themselves, and not with non-teleworkers. As engagement in teleworking and participation in activities implies an organisation of lifestyles and interactions with the other members of the household, an exploratory quantitative textual analysis seems relevant to better understand these interdependence effects. This approach is promising, since, contrary to qualitative approaches, it allows for the analysis of a large volume of textual data.

Hypothesis and objective

6As travel is the consequence of the need to carry out activities in certain spaces, the activities-based approach of individuals provides a particularly relevant angle of study for understanding the impacts of teleworking. Based on exploratory textual analysis, this article aims to identify how the adoption of teleworking leads teleworkers to modify their daily activities and travel.

7We hypothesise that the adoption of teleworking may be considered a moment of rupture in the organisation of the system of activities of individuals. Teleworking may modify them and integrate new ones which will then have to be reorganised with the other spheres of daily life. This professional transition could therefore change relationships and activities, calling for new choices, aspirations and trade-offs by teleworkers and others household members, to find a balance between the different days of the week on the one hand, and in the different spheres of life on the other hand.

8Thus, the objective is to identify and analyse how teleworking has changed the organisation of daily activities and their relationship to the different spheres of life. This is particularly innovative and important, especially as some changes in travel behaviours are likely to persist after the Covid-19 pandemic (Kaufmann 2021, Nikolaeva et al. 2023). The results produced here could therefore inform planning policies aimed at reducing the use of motorised modes and increasing the use of active modes for example, for health or sustainability purposes.

9The following part of this article is organised as follows: Section 2 describes the method of constitution and analysis of the corpus and provides descriptive statistics for the selected variables. Section 3 summarises the results of the changes in activity-travel behaviour following the adoption of teleworking, and of the speeches analysis. Section 4 discusses three new research directions aimed at better understanding the interaction between telework, mobility and rebound effects: (i) reinvesting time saved in commuting in other forms of mobility; (ii) teleworking as a means of changing travel patterns to achieve more satisfying lifestyles; (iii) teleworking as a catalyst, a means of facilitating choices and decisions at certain stages of the life cycle of individuals and their household. Finally, section 5 presents conclusions, limits and policy implications of the obtained results.

Method and data

Data acquisition

10Our corpus includes answers to open-ended and optional questions in an online survey, related to changes in mobility patterns with the adoption of telework. This survey was distributed between March and June 2022 in metropolitan France, when teleworking was no longer imposed by the French government. This questionnaire aimed to better understand the effects of the adoption of teleworking on employees’ mobility (work and non-work). More specifically, it focused on changes in (1) the mobility frequencies, (2) the distribution between the days of the week, (3) territories, and (4) the travel modes. In addition, the survey included a section on the potential impacts of teleworking on employees’ residential and workplace choices in near future.

11The survey was first tested on a small sample and then disseminated as widely as possible via professional and personal networks, mailing lists of associations, political parties, companies and newspapers. Participants were encouraged to share the survey online with their own personal and professional networks. The scope of the survey was employees (public and private) in metropolitan France, whether they were teleworking. 1,412 people contributed to this survey.

Constitution of the corpus

12To address our hypothesis, we focused only on home-based teleworkers who teleworked occasionally and regularly. We excluded non-teleworkers, employees who had always teleworked and those who teleworked 6 and 7 days per week. In total, the sample is 967 participants. Next, we considered only those participants who indicated a change in daily mobility patterns due to teleworking and answered the open-ended and optional questions described above, leading to a final sample of 392 participants. The answers provided by these participants formed our corpus.

13In the absence of official statistics, the representativeness of our sample could not be assessed. However, there is the 2021 employment survey in France, which indicates that teleworking mainly concerns executives (Jauneau 2022), as we observed in our sample. Therefore, since the population we studied was not described elsewhere, it is not possible to establish its composition and to apply any sample selection, correction, or to weight our analyses.

14The questionnaire asked teleworkers whether teleworking had changed their daily mobility patterns (‘With the adoption of teleworking, do you think that the frequency of your activities has:’). Participants responded with a Likert scale from 1 to 5 (‘decreased’ to ‘increased’ and 3 for ‘unchanged’). If participants indicated a change in frequency (i.e. answered 1, 2, 4 or 5), they were asked an open-ended and optional question: ‘In a few words, what explains these changes in your travel habits?’ These two questions were asked in the same way for each of the following activities: food stores (supermarket, grocery and market or farm); leisure activities (e.g. sports, cultural or associative); and personal activities with escort children, appointment (e.g. doctor, hairdresser), and visit friends or family.

Pre-processing of data

15The corpus was manually cleaned and normalised to homogenise several terms (to avoid the analyses considering two synonyms as unrelated) and to add terms in the case of implicit reference (e.g. ‘TP’ or ‘underground’ were replaced by ‘public_transport’). Then, the corpus was loaded into the IRaMuTeQ software (R Interface for Multidimensional Analysis of Texts and Questionnaires, version 0.7–alpha 2), an open-source text mining dedicated tool based on both R and Python languages (Ratinaud 2009). IRaMuTeQ applies text pre-processing in four steps.

1. It divides each text into smaller units on the criteria of size (number of words) and punctuation, called ‘text segments’. This segmentation has the advantage of decreasing the units’ granularity and increasing the precision of the analyses.

2. It applies a lemmatisation, corresponding to modifying each word to its most generic form (i.e. putting each verb in the infinitive and each noun and objective in the masculine singular; e.g. ‘ate’ and ‘eaten’ were changed for ‘eat’). This step makes it possible to group the different uses of the same word to avoid considering them as different, and to reduce the amount of information to be analysed.

3. It then categorises the words into two subcategories: ‘full words’ (e.g. verbs, nouns, and adjectives) and ‘tool words’ (e.g. pronouns and determiners). This step allows only full words to be considered in the analysis.

4. Then, it constructs a matrix with each text segment in the row and each occurrence (i.e. word) in the column. If for a given segment the word is present once, the value ‘1’ is given in the matrix. If for the same segment, the word is not present, the value ‘0’ is given in the matrix.

Data analyses and selected variables

16We used the Reinert method (Top-Down Hierarchical Clustering [TDHC]) originally released by the ALCESTE software (1983, 1990) and implemented in the software IRaMuTeQ (Ratinaud and Déjean 2009). This text-mining approach consists of extracting lexical worlds. It makes it possible to objectify and synthesise the contents of texts, ‘by bringing out themes and comparing them according to the social and demographic characteristics of the respondents’ (Garnier and Guérin-Pace 2010:9). In detail, this method groups the words statistically associated with each other in different clusters (here, the five main clusters are presented in the results). It also allows us to see the modalities of several co-variables significantly associated with each cluster by calculating a Chi-square test of association (the list of co-variables included in the model is available in Table 1).

17Table 1 shows descriptive data of our final sample for all the co-variables which were inserted in the text mining analysis. The catchment area of a city is a statistical area used by France’s national statistics office, which consists of a densely populated urban agglomeration and the surrounding exurbs, towns and intervening rural areas that are socioeconomically tied to the central urban agglomeration, as measured by commuting patterns (INSEE, 2022). The survey sample has an over-representation of women (64%). While the distribution of the household’s life course is fairly well distributed, we have an over-representation of highly skilled teleworkers (96%), concentrated in large municipalities (Aguilera et al. 2016, Vilhelmson and Thulin 2016). The sample of teleworkers is indeed very urban with 72% living in a city centre and only 3% in isolated municipalities. This is consistent with a high use of public transport for commuting. These unbalanced representations of our sample are obvious. Teleworking is strongly influenced by individual and household demographics at different stages of family life (Zhang et al. 2020), job characteristics (Walls et al. 2007, Sener and Bhat 2011, Singh et al. 2013, Asgari et al. 2014), residential/job location and commuting distance/duration (Ellen and Hempstead 2002, Mokhtarian et al. 2004, Ory and Mokhtarian 2006, Muhammad et al. 2007, Kim et al. 2012, 2015, Elldér 2014, de Vos et al. 2018, Ravalet and Rérat 2019, Rüger et al. 2021). Teleworkers were asked about their telework frequency: 13% teleworked occasionally, 22% one day per week, and 48% two days per week. 61% started to telework with Covid-19. Finally, 70% telework with a formal agreement signed with their employer. This may be due to the specificity of the studied population with workers having high flexibility in work, especially executives.

Table 1: Descriptive statistics of the illustrative variables introduced in the TDHC

Variable	Definition	N=392 (%)
Individual and household characteristics
Gender
Female	Female (1 = Yes, 0 = No)	251 (64%)
Life course
Under_45	20-45 years old, no children at home	138 (35.20%)
Parents_0_10 years	Parents with children under 10	93 (23.72%)
Parents_11_18	Parents with children between 11 and 18	61 (15.56%)
Over_45	45 years and over, without children at home	100 (25.51%)
Education level
No_bac	Primary school, junior high school	5 (1.27%)
Bac	General baccalaureate, vocational certificate	15 (3.93%
Bac+2/3/4	Bac +2 (two-year Higher Education diploma), Bac +3 (vocational degree), Bac +4 (first year of a master's degree)	101 (25.76%)
Bac+5/8	Bac +5 (master’s degree 2), Bac +8 (PhD)	271 (69.13%)
Residential localisation
Catchment area of a city
Centre	Centre municipality	142 (36.22%)
Main pole	Other municipality of a main centre	143 (36.48%)
Secondary pole	Municipality of a secondary centre	5 (1.28%)
Suburban	Suburban municipality	88 (22.45%)
Isolated	Isolated municipality	14 (3.57%)
Commuting
Distance
0_5	Less than 5 km	64 (16.32%)
5_9	Between 5 and 9 km	85 (21.68%)
10_14	Between 10 and 14 km	61 (15.56%)
15_19	Between 15 and 19 km	33 (8.42%)
20_29	Between 20 and 29 km	58 (14.80%)
30_39	Between 30 and 39 km	28 (7.14%)
40_49	Between 40 and 49 km	20 (5.10%)
50_99	Between 50 and 99 km	30 (7.65%)
100_plus	100 km or more	13 (3.32%)
Duration
Less_15	Less than 15 minutes	43 (10.97%)
15_29	Between 15 and 29 minutes	95 (24.23%)
30_44	Between 30 and 44 minutes	101 (25.76%)
45_59	Between 45 and 59 minutes	69 (17.60%)
60_89	Between 60 and 89 minutes	56 (14.29%)
90min_plus	90 minutes or more	28 (7.14%)
Mode of travel
No	Exclusive teleworker	5 (1.30%)
Walk	Walking	15 (3.82%)
Bicycle	Bicycle, scooter	95 (24.23%)
Motorbike	Scooter, motorbike	4 (1.02%)
Car	Car (alone or with someone from your family)	121 (30.87%)
Carpooling	Carpooling (with a colleague, neighbour, etc.)	5 (1.27%)
Public transport	Bus, coach, tram, metro	60 (15.31%)
Train	Train, RER, TER	87 (22.19%)
Satisfaction
1	Very uncomfortable	15 (3.83%)
2	Uncomfortable	94 (23.98%)
3	Indifferent	82 (20.92%)
4	Comfortable	144 (36.73%)
5	Very comfortable	52 (13.26%)
Not relevant	Exclusive teleworker	5 (1.27%)
Job characteristics & telework
Socio-professional category
Head	Public and private employee: Managing director, direct assistant, salaried company manager, CEO, minority manager, partner	13 (3.32%)
Private_executive	Private employee: Engineer, executive (except for general managers or their direct assistants)	128 (32.65%)
Public_executive	Public employee: Category A staff	157 (40.05%)
Private_technician	Private employee: Technician, supervisor, administrative or commercial supervisor, sales representative (non-executive)	27 (6.89%)
Public_technician	Public employee: category B, Technician	35 (8.93%)
Private_worker	Private employee: Office, commercial or service employee, Labourer or skilled worker, Skilled or highly skilled worker, Workshop technician	23 (5.87%)
Public_worker	Public employee: category C or D, skilled or highly skilled worker	9 (2.30%)
Telework frequency
0.5	Occasional or a few half-days per month	51 (13.01%)
1	1 day per week	87 (22.19%)
2	2 days a week	190 (48.47%)
3	3 days a week	43 (10.97%)
4	4 days a week	7 (1.79%)
5	5 days a week	14 (3.57%)
Seniority in telework
0	Less than 2 years (with Covid-19) = 0 , More than 2 years = 1	238 (60.71%)
Agreement with employer
Formal	With a formal agreement signed with my employer (contract, rider, charter)	277 (70.66%)
Informal	With an informal agreement with my employer	77 (19.64%)
No need	No need for an agreement	38 (9.96%)

Results

Changes in travel frequency with teleworking

18Before describing the TDHC results, we highlight the proportion of changes in teleworkers’ mobility patterns since they started teleworking, for each of the activities: grocery shopping, leisure and personal activities (based on all teleworkers, 967 participants).

19For teleworkers who shop or used to shop in supermarkets – 33% have changed their travel frequency (26% decreased, 7% increased); for grocery (e.g. local shops, butcher, greengrocer), 50% have changed their frequency (32% decreased, 18% increased); for the market or farm, the values are similar to those of the supermarkets: 34% have changed their travel frequency (26% decreased, 7% increased). For leisure activities, 31% of teleworkers have changed their travel frequency since teleworking (9% decreased, 22% increased). Regarding personal activities, 30% have changed their frequency of escorting children (6% decreased, 25% increased); only 22% have changed their frequency of personal appointments (4% decreased, 18% increased); and 15% have changed the frequency of visits to friends or family with teleworking (6% decreased, 9% increased). In total, while teleworkers have reduced the number of commuting trips, they have changed their daily activity patterns and thus their mobility for 40.5% of them.

20In the next section, we will look in more detail at the spill-over effects of telework, to see how the time saved in commuting is reinvested since employees teleworking.

Impact of teleworking on spatial-temporal travel patterns

21Our corpus consists of 392 texts, 693 text segments and 9,281 occurrences. The TDHC divided the corpus into five clusters, representing 90.91% of the text segments (i.e. 630/693; Figure 1). Cluster 1, called ‘Commuting’, is different from the four others. It refers to the time saved by teleworking through the reduction of commuting, described as long, congested and tiring. The four others clusters highlight changes in daily activities and non-work mobility patterns. A second branch stands out, contrasting cluster 2, called ‘life course’, which focuses on the various explanations for the change in travel frequency, with clusters 3 ‘temporal patterns of personal activities’, 4 ‘spatial proximity’ and 5 ‘grocery shopping behaviour’. These three clusters are more interdependent with different subsystems of daily life and allow for the characterisation of activity-mobility patterns. The most representative quotes from the clusters are reported in Figure 2.

The ten more significant words are shown in decreasing order for each cluster, as well as the significant co-variables modalities. The significant threshold was set at p <0.001. The grey values correspond to the χ 2 association score.

Less commuting, less fatigue and more time saved

22Cluster 1 ‘commuting’ involves 176 text segments (27.94% of the corpus). With the adoption of teleworking, teleworkers travel less often to their workplace, hence the words associated with saving time, eliminating the long commute and losing less time with traffic jams. These speeches are associated with a decreased feeling of tiredness. Teleworkers highlight that the time benefits obtained are reinvested in the possibility of doing other activities, and having more freedom and time available through teleworking.

23This cluster 1 ‘commuting’ is representative of home-based teleworkers described in the literature, with a high commuting duration and distance (more than 60 minutes [χ ² = 23.15], more than 50 km [χ ² = 18.92]), and who are rather unsatisfied with the comfort of their commute (χ ² = 21.27). They are more likely to live in suburban municipalities (χ ² = 9.42) and and they use the train more than in the others clusters (χ ² = 7.62). However, it is not representative of occasional teleworkers (χ 2 = -7.63) and bicycle commuters (χ 2 = -11.38).

Lifestyles and life-course stages of teleworkers and their households

24Cluster 2 ‘life course’ involves 154 text segments (24.44% of the corpus). It highlights the complex interaction of the impacts of teleworking on change in the activity and mobility patterns of teleworkers and other household members. The results show that regardless of the activity mentioned (e.g. shopping at the farm, producers, short food chain, in drive, modal shift to walking, ‘stopping’ social life and leisure activities), all these changes are associated with many factors such as the Covid-19 pandemic, a new child in the household, a new job, residential mobility, but are relatively little or not associated with the adoption of telework. The positive impacts of teleworking are mentioned thanks to the time allowed for escorting children to school and extracurricular activities.

25This cluster 2 ‘life course’ is a characteristic of occasional teleworkers (χ ² = 9.74). Regarding commuting, they are rather indifferent to the question of satisfaction (χ ² = 12.16) and live less than 5 km from their workplace (χ ² = 9.61).

More flexible time slots on teleworking days to avoid peak hours

26Cluster 3 ‘temporal patterns of personal activities’ involves 73 text segments (11.59% of the corpus). With the adoption of teleworking, teleworkers report having more flexibility and ease in planning their personal activities during the teleworking day as opposed to a day at the usual workplace. They plan activities before starting their usual working hours, during the lunch break or earlier in the evening instead of commuting. Teleworking thus allows them to organise administrative and medical obligations in their schedule, but also to access shops and services during their opening hours.

27This cluster 3 ‘temporal patterns of personal activities’ is a characteristic of teleworkers living in the municipalities of a main centre (χ ² = 7.16). No telework modality is significantly associated with it.

Daily activities close to home/workplace

28Cluster 4 ‘spatial proximity’ involves 100 text segments (15.87% of the corpus). It focuses on mobility patterns, proximity to geographical locations (home and work), and non-work activities, especially leisure (e.g. sports, lunch with friends). An important result is highlighted. In contrast to cluster 5 ‘grocery shopping behaviour’ described below, teleworkers do not mention a change of location for their usual activities but talked about accessibility and spatial proximity. With the adoption of teleworking, they can maximise, rationalise and refocus their trips close to where they already are some activities are carried out next to the home, others next to the workplace.

29This cluster 4 ‘spatial proximity’ is illustrative of highly educated teleworkers (χ 2 = 6.68). Such as cluster 3, no teleworking modalities are not significant.

Food consumption and modes patterns differ between weekend-weekdays

30Cluster 5 ‘grocery shopping behaviour’ involves 127 text segments (20.16% of the corpus). With the adoption of teleworking, teleworkers are more likely to go to grocery shops, to market, and to buy directly from producers and the short food chain. Consuming fresh food requires travelling more. As they have more time, they change their food consumption places, but also their travel modes for those who can. Indeed, teleworkers shop in more groceries shops, walking or cycling, instead of going to the large hypermarkets by car. In the same sense, they have also substituted the usual food supply days, from weekends to weekdays, when it is less crowded. In addition, these changes in consumption behaviour are associated with positive evaluations (e.g. ‘I choose better quality products’, ‘I buy healthier’).

31This cluster 5 ‘grocery shopping behaviour’ is characteristic of teleworkers living in centre municipalities with high density (χ ² = 9.59), and who work less than 15 minutes from home (χ ² = 6.69). As for clusters 3 and 4, no teleworking modality is significantly associated with this cluster.

Discussion

32The objective of this article was to identify and analyse the representation and reorganisation of teleworkers’ daily activities and mobility patterns. With the adoption of teleworking, 40.5% changed their travel frequency. The results of the TDHC highlighted that through the flexibility of schedules, teleworking allows reorganising daily activities both spatially and temporally, solving problems of fatigue and work-life imbalance (Vittersø et al. 2003, Hartig et al. 2007, Peters and van der Lippe 2007, Thulin et al. 2019), and allocating more time to activities (Asgari et al. 2016). Thus, our results show the complexity of the interactions involved in life course decision-making, and the difficulty of studying the ‘pure’ and ‘direct’ impacts of teleworking. We point out the interest in analysing the speeches through the activity approach of individuals, as the rebound effects of teleworking depend on the lifestyles and life course of the teleworkers and their household members.

‘Time saved’ is at the heart of teleworkers’ speeches

Reduced commuting frequency

33Reducing the time spent commuting to the workplace is one of the main motivations of teleworkers (cluster 1 ‘commuting’), and their associated modalities corroborate the numerous studies that focused on the high distance/duration of commuting (dissatisfaction [Scott et al. 2012]; user of public transport [Budnitz et al. 2020, Caldarola and Sorrell 2022, Ravalet and Rérat 2019]; teleworking frequency 1 day or more per week [Caldarola and Sorrell 2022]; better work-life balance and decreased tiredness [Haddad et al. 2009, Hartig et al. 2007, Mokhtarian and Salomon 1997, Peters and van der Lippe 2007, Thulin et al. 2019, Vittersø et al. 2003]). Like money and energy, time is a limited resource and a feeling of time deprivation can lead to individual and peer suffering (Hobfoll 1989, Drevon et al. 2020). Although we expected differences in discourse, the seniority in teleworking was not significantly related to this cluster (Asgari et al. 2022a).

34Regarding the relationship between teleworking and commuting frequency, studies do not corroborate it. Some of them concluded that regular teleworkers travel more than other employees (Zhu 2012, non-teleworkers and non-regular teleworkers being included in the same category; de Abreu e Silva and Melo 2018a), others showed the opposite (Lim et al. 2003), and still, others suggested that they travel further each week than non-teleworkers, but make fewer commuting (Caldarola and Sorrell 2022).

35At the household level, the results do not show a significant impact of teleworking on the teleworker’s partner’s commuting (Zhu 2013, Melo and de Abreu e Silva 2017). Nevertheless, with Covid-19, this relationship needs to be updated with the influence of new practices (e.g. videoconferencing) and the desire of companies to reduce their carbon footprint and transport costs.

Teleworking, commuting distance and urban sprawl

36Corroborating the literature, living further away from one’s job is one of the characteristics of teleworkers (cluster 1 ‘commuting’). This association between teleworking and commuting distance is not new (Mokhtarian 1991, Nilles 1991). However, the direction of causality is not established. It is therefore not clear whether the possibility of teleworking encourages people to move away from their workplace, or whether teleworking is more attractive for people who already live further away from their workplace (Mokhtarian et al. 2004, Kim et al. 2012, Hostettler Macias et al. 2022). In any case, it would seem that telework, if it leads to residential changes, will mostly move them to peri-urban areas rather than to significant distances (Habib and Anik 2021). In fact, by reducing the number of home-to-work trips, teleworking allows teleworkers to extend their residential search areas (Janelle 1986, Nilles 1991, Lund and Mokhtarian 1994, Ory and Mokhtarian 2006, Muhammad et al. 2007, de Vos et al. 2018, Cerqueira et al. 2020, Hostettler Macias et al. 2022) even if it means increasing the distances to be covered (de Abreu e Silva and Melo 2018a, Lachapelle et al. 2018, Ravalet and Rérat 2019, Wang and Ozbilen 2020, de Abreu e Silva 2022, Caldarola and Sorrell 2022). However, this is a matter of debate. Some research showed a link between teleworking and urban sprawl (Larson and Zhao 2017) while others argued that the two are not related (Ellen and Hempstead 2002, Ory and Mokhtarian 2006). Moreover, teleworking may further hide other more important motives for residential change. These residential preferences may be subject to socioeconomic, demographic and environmental characteristics. Therefore, the impact of teleworking on residential location choice would vary by life course (Muhammad et al. 2007, Coulter and Scott 2015, Coulter et al. 2016, Rau and Manton 2016), making teleworking a less important factor in relocation decisions than initially estimated (Muhammad et al. 2007, Ettema 2010).

Teleworking stimulates changes in spatiotemporal mobility

37The results of the teleworkers’ speeches analysis highlight that the reduction in commuting and fatigue allows them to reinvest the time saved in other activities (cluster 1 ‘commuting’). Thanks to teleworking, employees report having more control over their choice of activities, locations, days and travel modes (clusters 3 ‘temporal patterns’, 4 ‘spatial proximity’ and 5 ‘grocery shopping behaviour’). By sharing many words, these three clusters illustrate the interdependence with different daily life subsystems.

38Employees reported reorganising their activity-travel behaviour including teleworking days. The two main motivations are the physiological need for mobility and the stimulation to make other trips related to personal and leisure activities (e.g. going for a run, going out to relax, visiting family or friends), escorting children or grocery shopping. All these non-work activities are associated with proximity and accessibility. This suggests changing travel patterns at the individual and household levels. According to the employees’ speeches, teleworking allows for reducing spatiotemporal constraints to undertake activities in different locations (cluster 4 ‘spatial proximity’) and at different times (cluster 3 ‘temporal patterns’). Thus, our results suggest an important relation between telework, the number of trips and accessibility to non-work activities.

Trip-chaining patterns

39The employees’ speeches highlight that teleworking has changed the frequency and patterns of non-work travel by eliminating non-work trips related to commuting. For example, unlike on normal working days, they no longer go out for a drink after work with their colleagues or pass by the hypermarket or grocery shop. While some teleworkers say they stay at home, others say it is an opportunity to change their consumption patterns (which may explain the frequency decreased of all consumption patterns [hypermarkets, grocery shops and markets]), favouring proximity (Wells et al. 2001). This is paradoxical because, in the teleworkers’ speeches, they say having increased the frequency of grocery shopping with the adoption of teleworking. On the one hand, they eat lunch at home on teleworking days. On the other hand, they enjoy preparing their meals with fresh products, requiring more time and more travel to get supply fresh products. A second hypothesis could support that this decreased frequency of consumption patterns could be due to the increase in online shopping (Mouratidis and Papagiannakis 2021, Mouratidis and Peters 2022, Shah et al. 2022). More broadly, previous studies suggested that virtual activities at home (online shopping, online banking, online restaurant/hotel/movie booking and online administrative services) increased the propensity to travel (Mosa 2011). In addition, it is important to recall the post-pandemic context of Covid-19. These changes in consumption patterns may be more a Covid-19 effect rather than a teleworking one (cluster 2 ‘life course’).

40Thus, teleworking changes spatiotemporal behaviours in terms of trip chaining patterns by allowing a multitasking system of activities, which may therefore be more or less efficient: it may create several one-stop trips instead of one multi-stop trip (Mokhtarian 1991). The results of Wells et al. showed that telecommuters who alternated days at home and in the office tended to handle errands on the way home from the workplace when they were already out driving. ‘Employees from departments mandating remote work on a full-time basis also tended to handle errands after work, but errand-running locale changed’ (2001).

Proximity and teleworking days

41Previous studies have already stated that teleworkers tend to visit non-work destinations closer to home (Pendyala et al. 1991), including teleworking days and that their activity space is more restricted than that of non-teleworkers (Saxena and Mokhtarian 1997). The Singh et al. study also showed that high accessibility to leisure, dining, religious, car repair, personal business and medical centres implies a higher probability of teleworking and that those who lived closer to non-work and leisure opportunities attached more value to teleworking and were, therefore, more likely to adopt it (2013). This could be explained by their residential location and the accessibility of activities more grouped in one place. Our clusters 3 ‘temporal patterns of personal activities’ and 5 ‘grocery shopping behaviour’ are mainly characterised by teleworkers residing in central municipalities or main centres with high density. Budnitz et al. make the same point: ‘the density of local shopping and leisure options, as well as the proximity of schools and other escort and business destinations, is more relevant to the travel behaviour of telecommuters than of non-telecommuters. This means that accessible, mixed use areas could enable frequent telecommuters, who have the temporal flexibility to make more trips for purposes other than commuting, to do so more sustainably’ (2020).

42However, while our results highlight a relocation of activities close to home (notably grocery shopping and personal activities), this is not the case for leisure activities (Saxena and Mokhtarian 1997). They seem to be more difficult to relocate or to change the type of activity than, for example, changing grocery consumption habits.

43This result illustrates the willingness of individuals to change their lifestyles and consumption patterns in favour of proximity.

Weekends/weekdays and daytime hours distribution patterns

44The analysis of the teleworkers’ speeches highlights the possibility of shifting trips to off-peak periods (different times of the day) to avoid congestion delays, and/or to different days of the week (cluster 3 ‘temporal patterns of personal activities’). For example, thanks to the temporal flexibility of teleworking, employees have substituted or added trips, from weekends to weekdays, on days when there are many people in the shops or on the roads: going shopping on Wednesdays because it is market day instead of going to the hypermarket on Saturdays when it is too busy; leaving work earlier on Fridays to access leisure activities or going on weekends instead of Saturdays to avoid congestion. The specific days that stand out in the teleworkers’ speeches are Wednesday and Friday. In the literature, Fridays are the most popular, followed by Mondays (Hostettler Macias et al. 2022) and Wednesdays, a situation similar to that before the pandemic. Nevertheless, the proportion of teleworkers varies between days of the week, leading to a variable reduction in congestion (Windeler et al. 2017). Thus, teleworking improves the well-being of those who telework by limiting long commutes and also improves the situation of those who do not telework by reducing congestion on certain days (Larson and Zhao 2017).

45Regarding the time of day, thanks to the temporal flexibility of teleworking, employees report doing their activities on teleworking days (Saxena and Mokhtarian 1997) following the same schedule as their usual working hours (Wells et al. 2001). Indeed, they spread their trips over the day. For example, they shop earlier in the morning, during the lunch break and in the late afternoon and evening, and avoid peak travel on teleworking days (Pendyala et al. 1992). Thus, in line with Pendyala et al.’s findings, our results suggest that no major changes in weekly activity reorganisation are to be expected, as non-work trips show relatively similar temporal distribution patterns between teleworking days and working days (1992).

So, what about the effects of teleworking on the frequency of non-work trips?

46Although few studies address the link between teleworking and non-work travel, the increase in non-work trips offsets and often cancels out the reduction in commuting trips (Kim et al. 2015, Caldarola and Sorrell 2022, Wöhner 2022), and the demand for non-work activities may influence their choice to telework (van Wee et al. 2013, Asgari and Jin 2017, Loo and Wang 2018). Teleworkers not only make more non-work trips on average than non-teleworkers (Zhu 2012, Wang and Ozbilen 2020, Caldarola and Sorrell 2022) but also more trips for all purposes (He and Hu 2015, de Abreu e Silva and Melo 2018a, Cerqueira et al. 2020, Hook et al. 2020). For Budnitz et al. teleworkers also do more shopping than non-teleworkers, but also more leisure-related trips, holidays/day trips and more ‘other’ non-utility trips than non-teleworkers (2020). But this increase in non-work trips does not concern all trips and depends on the teleworking arrangements (frequency, full/part-day; Asgari et al. 2016). Some studies, however, find the opposite or more measured results, which are difficult to determine if they are only related to the context and/or the methodology used. Elldér shows that, on days when they telework, Swedish employees who telework the whole day make fewer trips (for all purposes) than those who telework only part of the day, and those who did not telework at all on that day (but who may telework on another day; 2020).

47While escorting the children came up in the teleworkers’ speeches, the spouse was absent. This is surprising, as we had expected a potential transfer of travel between household members (Asgari et al. 2016). The teleworker who stays at home to work would travel less than his/her spouse. This is because the teleworker’s reduced commute may reduce the risk of chaining or stopping trips, and thus transferring some activities to the spouse who now stops on the way to work. Conversely, the teleworker who stays at home to work would travel more than his/her spouse, as s/he is more available.

48In the literature, there is still much discussion on this topic. The findings seem to point towards an increase in non-work trips for teleworkers and their household members (Kim et al. 2015, Asgari et al. 2016), and for single-worker households (de Abreu e Silva and Melo 2018b, Caldarola and Sorrell 2022), but not for households with two teleworkers (de Abreu e Silva and Melo 2018b). For single-worker households, this main source of additional travel could be the ‘freeing’ of the household car for use by other household members (Kim et al. 2015). While teleworkers travel closer to home, they travel more kilometres (Asgari and Jin 2022), but this does not mean that they necessarily travel by car. While households with two teleworkers, this may reflect the substitution of travel in the latter and a more efficient allocation of household tasks between household members.

Teleworking and travel mode

49The analysis of the teleworkers’ speeches highlights a strong link between teleworking and the choice of transport modes, depending on the activity and residential location. The travel patterns on changes in activity-travel behaviour are common to all our clusters: cluster 1 ‘Commuting’ with significant use of the train and not the bicycle (negatively associated with the car and also positively associated with public transport in Caldarola and Sorrell 2022); cluster 2 ‘life course’ with the escort of children; and clusters 3 ‘temporal patterns of personal activities’, 4 ‘spatial proximity’ and 5 ‘grocery shopping behaviour’ on active modes (cycling and walking; for Caldarola and Sorrell [2022], it is also associated with active modes and public transport and less with the car). Furthermore, one result is very interesting from our textual analysis. The discourse on non-work trips focuses on the possibility of a modal shift. Thanks to the temporal flexibility of teleworking, employees do their activities closer to home, allowing them to walk rather than use public transport or cycle rather than drive, or to take the time to walk/cycle with their children on teleworking days rather than dropping them off by car as on work days. Our results also highlight the positive satisfaction of teleworkers with these active modal shifts by becoming ‘the pleasant outing of the day’ (Nikolaeva et al. 2023).

50Thus, although teleworkers travel more distances and kilometres in general, use active modes (non-work trips) and public transport (commuting). Teleworking can potentially be associated with lower energy consumption and greenhouse gas emissions.

51Our results are consistent with some studies, although the causal relationship between teleworking and active modes is not established (Chakrabarti 2018, Elldér 2022, Huang et al. 2023). The time and energy saved by not commuting every day seem to be spent instead on slower means of travel, with a strong appetite for proximity (Pendyala et al. 1992, Saxena and Mokhtarian 1997), especially on teleworking days compared to working days (Mokhtarian and Varma 1998, Chakrabarti 2018). Chakrabarti indicates that teleworking can promote non-motorised transport and daily physical activity (2018). Lachapelle et al. showed that teleworking was correlated with a higher probability of using more than 30 minutes of non-motorised modes (2018). In contrast, a study of the relationship between travel mode and frequency of non-work walking found that teleworkers were not associated with more walking compared to car-based commuters (Lachapelle and Noland 2012). Another study indicates that as the number of daily trips increases, the individual may be more willing to use other mobility alternatives such as public transport or carpooling options, confirming that a high number of daily trips does not necessarily lead to increased car use (Asgari et al. 2022b).

52However, this point does not mean demotorisation, especially as previous studies have shown the opposite results: trips on teleworking days are more likely to be made by car than on non-teleworking days (Wells et al. 2001). Yearly, teleworking is associated with higher vehicle kilometres due to longer commuting distances (Chakrabarti 2018). Recent findings by Asgari and Jin suggest that teleworkers are unlikely to give up private vehicle ownership simply because of their option to telecommute and that while public transport plays a substitution role for the car, non-motorised modes are more of a complementary effect for private vehicles (2022).

53At the household level, the results differ. De Abreu e Silva and Melo show that for one-worker households with only one member, a higher frequency of teleworking (3 or more days per week) was positively correlated with vehicle kilometres travelled and a higher number of trips for all modes, especially for car trips, followed by active travel (2018b). For single-worker households with more than one member, a higher frequency of teleworking was only correlated with more vehicle kilometres travelled. Teleworkers in two-worker households also made more weekly car trips, without significantly increasing vehicle kilometres travelled. The authors suggest that for these two-worker households, total weekly trips may be compensated for by a more efficient division of labour between household members. In contrast to single-earner households, which by definition have fewer people to share them, resulting in more weekly trips (2018b). Thus, further research is needed on the availability of the vehicle for another household member on teleworking days (Kim et al. 2015), and in particular, the possibility that a different vehicle is used to make the trip, to understand the total energy and air quality impacts of teleworking (Mokhtarian 1991).

54Our results also show that residential location and accessibility must also be considered (characteristics of teleworkers residing in central municipalities and the main pole). According to Asgari and Jin, people living in metropolitan areas with less than 250,000 inhabitants are more likely to telecommute (leading to less commuting and more car dependency), while living in metropolitan areas with less than one million inhabitants is likely to increase commuting (2022). While our results show changes in travel patterns associated with the adoption of teleworking, we have not demonstrated a causal relationship (Caldarola and Sorrell 2022). There may be many reasons for this and the effect of teleworking may be relatively small or even indirect. For example, it may be residential self-selection, where it is unclear whether factors such as residential density, the built environment or urban design lead to sustainable urban travel (Ewing and Cervero 2010), or whether people with sustainable travel preferences choose to live in high-density urban areas (Cao et al. 2009). In both cases, causal mechanisms can work in both directions.

Teleworking, adjustment variable and decision-making over the life course

55The speeches analysis of teleworkers highlights a double discourse: childcare and residential mobility (cluster 2 ‘life course’); Two topics are not dissociated. This cluster is relevant because it raises the question of the ‘direct’ effect of teleworking on changes in activity-travel behaviour.

56Our results suggest that these changes in activity-travel behaviour (frequency, residential/job location, trips chaining patterns, proximity, time-of-day/day-of-week, vehicle ownership, mode) are not only linked to teleworking but also to the lockdown of the Covid-19 pandemic (Kaufmann 2021), the arrival of a child or the change of job, etc. For example, the birth of a child generates more care-related trips, or may even prompt households to move to a larger home with a garden. By the way, Zhang et al. concluded that ‘children are the most important feature in family-life stages for an individual’s telework behaviour’ (2020). In the same sense, the departure of a household member may increase or decrease the frequency of the teleworker’s trips. Government measures during the Covid-19 pandemic (closure of non-essential places and shops, general lockdown compared to a ‘health war’, permission to travel within a 100 km radius, etc.), also contributed to changing our consumption, activity-travel patterns as well as our representations (Atkinson-Clement and Pigalle 2021, Duque-Calvache et al. 2021, Colomb and Gallent 2022, Nikolaeva et al. 2023, Pigalle and Atkinson-Clement 2023). Planning policies in favour of sustainable mobility with the development of carpooling and cycling facilities can also have an impact on teleworking and changes in activity-travel behaviour. Thus, several factors lead to changes in lifestyles and reorganise patterns of daily and residential mobility.

57Trade-offs are made according to the experiences and life courses of the teleworkers and their households. This suggests that the activity-travel patterns are evolving, not fixed in time, and depend on life cycle stages (Muhammad et al. 2007, Rau and Manton 2016, Cailly et al. 2020). Indeed, and our results are consistent with this, practices are evolving, ‘priorities are changing’ and life spheres are intertwining (Sheller and Urry 2006). Teleworking appears to be an adjustment variable to facilitate individual and household choices and decisions, and thus indirectly influences changes in activity-travel behaviour.

58Studying the effects of teleworking from the ‘system of activities’ of teleworkers and their households to understand their mobility patterns appears essential, but methodologically difficult. It is within this framework that more and more authors recommend studying teleworking through a life course approach to better grasp ‘the variety of teleworkers’ biographies, life arrangements and appropriation of mobility’ (Hostettler Macias et al. 2022). According to these authors, the techniques of biographical interviews and longitudinal analyses are a solution.

Conclusion, limitations, implications for future research and practice

59Our study characterised the reorganisation of employees’ travel-activity behaviour following the adoption of teleworking. As travel is a consequence of the need to perform certain activities in certain places, the activity-based approach of individuals has shown the complexity of analysing the rebound effects of teleworking, as it impacts all spheres of daily life (work, family, social), both at the individual and household levels.

60Although large changes in displacement are not expected, teleworking offers an opportunity to rethink trade-offs in the choice of where individuals and households live and work, and additional flexibility in the scheduling of daily activities (to do new activities [or suppress some], and/or to reorganise activities across the week). Our results do not allow us to measure the nature of the relationship between teleworking and travel frequency (substitution, complementarity, neutrality or induction; Andreev et al. 2010), but they suggest that the longer the teleworkers’ commute, the more likely their travel behaviour will change. Therefore, teleworking could be seen as a catalyst for decision-making at certain stages in the life course of teleworkers who accept longer commutes (de Vos et al. 2018). Beyond the reorganisation of usual sub-family travel patterns, the link between telework and proximity to amenities would encourage non-work trips by active modes (Caldarola and Sorrell 2022). This could maximise the environmental benefits of telework and more broadly facilitate the transition to more sustainable travel (Budnitz et al. 2021), without leading to a demotorisation (Chakrabarti 2018, Asgari et al. 2022b).

61However, we have to acknowledge some limitations of our study. Since teleworking is part of a broad network, many confounding variables are related to its effects. This is therefore difficult to extract the direct influence of teleworking on activity-travel patterns, especially in this still unstable context following the Covid-19 pandemic (Nikolaeva et al. 2023). Even if at the time of the survey, teleworking was no longer compulsory in France, we are still in a moving context where the geography of workplaces and company agreements are still under discussion. The second and most important limitation is methodological. Identifying and measuring the rebound effects of teleworking is complex. Teleworking is part of interdependent systems of activity and subsystems where all crossings are possible, and where each individual has his/her pattern of spatiotemporal mobility.

62Quantifying travel frequencies is therefore not obvious, as it is not a matter of measuring a single trip, but of considering the whole. Daily coping strategies, changes and reorganisations of activities and therefore of teleworkers’ trips depend on other variables which should be integrated, such as the availability and price of parking at home and at work, the conditions of the commute (hardship), the type of establishment (private, public), etc., which all influence travel behaviour (Wells et al. 2001). The cases studied are numerous and difficult to compare: they are based on contrasting mobility data (for a single day, a week, etc.), different definitions of teleworkers (occasional and regular workers, self-employed, employees), trips involving only some modes of transport or merging all together. In addition, different forms of telework (full-day or half-day) also have a different impact on the use of time for activities and additional travel (Asgari et al. 2016). Thus, the nature of teleworking is crucial for understanding changes in individual travel behaviour, in terms of effects on overall transport demand, and congestion (van der Loop et al. 2019), for transport authorities or space management for employers.

63Future research should not only focus on policies to regulate commuting but should integrate the different lifestyles and life courses of teleworkers and their households.

64To change mobility behaviour, public policies cannot simply promote teleworking for sustainable mobility they must take into account spatial accessibility and uses, and activities-travel other than commuting.