Understanding travel mode choice through the lens of COVID-19: a systematic review of pandemic commuters

ABSTRACT The COVID-19 pandemic disrupted travel behaviours for very large numbers of people including those who shifted to teleworking and those without the option to work from home. While there is much valuable transport research that has examined the former category, it is still unknown how certain people such as health sector employees and delivery drivers changed their physical commuting in transport contexts that were radically different from those existing normally in urban areas. Based on a systematic review of 36 scientific publications on commuting during pandemic, this study pursues a dual objective. First, by examining the interrelated institutional, physical, and socio-psychological processes that supported or hindered low-carbon transport the study revealed that (A) public transport (PT) reduced service levels and concerns related to COVID were positively associated with substantial shifts away from PT towards car and active travel; (B) this positive association was found to be even stronger in the existence of pre-pandemic habit of car use for commute and strong negative emotions like fear triggered by environmental changes and health risks. Second, by synthesising the key findings from the literature, this study provides significant implications for how mode choice is modelled through the Theory of Planned Behavior and Norm Activation Model. By questioning whether the pandemic commuters had a “normal” set of travel mode alternatives to choose from, the study draws attention to the nuances of mode “choice” versus mode “use” and moves beyond the assumption that commuting always results from individuals making choices. It also argues that the role of (negative) emotions along with the importance of proximity to, or separation from, other bodies on how people commute should be considered in future research. Finally, the crucial role of COVID-19 in changing travel-related norms and the resulting long-term implications for policy interventions require further investigation by future research.


Introduction
With the growing efforts to decarbonise transport, travel behaviour research has highlighted the importance of analysing transport disruptions and subsequent individual traveller responses.Voluntary or involuntary, planned or unplanned, travel disruptions change stable mobility habits and increase attentiveness to alternative solutions (e.g.teleworking) and transport modes, hence, a "window of opportunity" opens up for introducing and encouraging the use of low-carbon transportation alternatives (Schwanen et al., 2012;Verplanken & Roy, 2016;Zarabi et al., 2019).The recent pandemic and government-mandated restrictions on mobility have constituted a set of important involuntary and unplanned life-changing events, which have disrupted mobility behaviours for very large numbers of people (Lee et al., 2021) at least temporarily (Hensher & Beck, 2022).In this period, teleworking increased and physical commuting reduced on an unprecedented scale.However, certain parts of the population such as health sector employees and delivery drivers had to continue travelling as they did not have the option to work from home (WFH).Since March 2020, travelling to a place of work was defined in various ways as an essential trip and specific employees among those who made these journeys were recognised as key or essential workers.This designation varied in different geographical contexts around the world but mainly included those who worked in sectors such as safety, government, healthcare, food, energy, and transport (Plyushteva, 2022).Additionally, there were individuals who were not classified as key or essential workers but still had to commute to their workplace due to the lack of WFH options.In this study, we collectively refer to both essential and non-essential workers who travelled to work during travel restrictions over the course of the COVID-19 pandemic as "pandemic commuters".
This study offers a systematic review of the published literature on the commuting mode use of workers (both essential and non-essential) during the COVID-19 pandemic.The objectives are two-fold.First, it evaluates how and to what extent the commuting patterns (or future intentions when studied) of different groups of these workers have been affected by (1) changes to the transport context (e.g.reduced public transit services and road congestion) and (2) their perception, interpretation, and emotional experience of those context changes.Within the 36 papers that met our inclusion criteria special attention has been devoted to two topics: commute mode shifts away from Public Transport (PT) to car driving or active travel, and the (negative) emotions triggered by changes in the environment in which their commuting occurs (e.g.changing health risks, changes in popular discourse, situational uncertainty).Second, by synthesising the findings from the review, we derive multiple implications for future studies of commute mode use.For this purpose, we draw insights from two prominent behaviour theories that have been used widely in travel behaviour research, namely the Theory of Planned Behavior (TPB) (Ajzen, 1991) and the Norm Activation Model (NAM) (Schwartz, 1977).Both theories share an expectancy-value structure that aligns with traditional economic analyses, portraying individuals as rational agents driven by utility-oriented preferences (Gardner & Abraham, 2008).Additionally, these theories are grounded in the assumption of conscious awareness and the perception of action consequences.However, a growing body of research has emphasised the influence of habits, which can lead to actions that occur outside of conscious awareness (Donald et al., 2014).Several reviews on the application of the TPB and/or NAM within the domain of travel behaviour are available, offering valuable insights into the extent to which commuting behaviour can reflect an individual's intentions or habits (e.g.Si et al., 2019;Yuriev et al., 2020).Specifically, commute intentions are shaped by individuals' positive or negative attitudes towards their work trip and the different transport modes available, the social pressure (i.e.social norms) and/or personal norms to make a particular choice for commute mode, and perception of the ease or difficulty of executing (i.e.perceived behavioural control (PBC)) a particular commute pattern (e.g.mode, route, time).This study will conclude with a heuristic model informed by the TPB and NAM of how highly disruptive changes in the transport context can differently affect daily travel mode use.
By studying pandemic commuters, we can derive significant implications for future research on commute mode use.The underlying principle is that atypical or extreme situations can provide valuable opportunities for understanding the deeper causes behind a given problem (e.g.here auto-dependency or barriers to low-carbon mode use) and its consequences than one can achieve through usual or "normal" situations where less basic mechanisms are often activated (Flyvbjerg, 2006).In that sense, understanding the complex relationships between the individual, physical, and socio-psychological factors that affected the travel of workers who continued to commute on a regular basis, but in an extremely changed context, can provide important lessons for understanding commute mode use.Finally, the wave of research concerning the mobility impacts of COVID-19 conducted during this period (between January 2020 and 2022) was unique and thus provides an excellent opportunity for analysis of travel mode use during highly disruptive circumstances.
The remainder of this paper is organised into four parts.The second section will summarise the scoping review approach, explain how it has been adapted for the purpose of this study, and describe how the adapted method has been employed.The third section will summarise the findings from the systematic review of pandemic commuters travel to/from work.The fourth section will synthesise the findings into a heuristic model of travel mode use and provide theoretical implications for understanding the traditional theories of mode choice modelling.Finally, the conclusion section will present the study's main results and contributions and discuss priorities for future research deriving from the reported work.

Methodology
An adapted version of the PRISMA 2020 method for scoping reviews (Page et al., 2021) has been deployed to conduct and report the review of relevant publications in peerreviewed, English-language academic journals across diverse academic disciplines published in 2020-2022.Scoping reviews differ from conventional methods of literature review in that they include a robust research method that reduces bias by being transparent and accurate while providing a complete overview of the literature according to the research aims and objectives.

Search strategy
A search for pertinent articles was conducted with the assistance of a library scientist in September 2022 using five different databases: Web of Science, Scopus, PubMed, Google Scholar, and Transportation Research International Documentation (TRID).
Based on the main research topic, "Influence of COVID-19 on the travel behavior 1 of workers", the research protocol consisted of variations of three sets of keywords.These included travel behaviour (e.g."means of transport" OR "commuting behavior") AND COVID-19 (e.g."pandemic" OR "coronavirus") AND worker (e.g."employee" OR "personnel") which were searched for in the articles' title, abstract, and keywords.A complete description of search terms and Boolean combinations is provided in the Appendix.

Study selection and inclusion/exclusion criteria
The selection process was performed in different steps: (1) search query design (explained above); (2) screening based on title, abstract and keywords; and (3) study selection based on full text.In the identification step, 17 additional records were identified through citation tracking and recommendations among the authors.However, upon further investigation, we discovered that all 17 of these records were already included among the initial set of 1885 papers.Consequently, these duplicates were removed during the elimination process.While we anticipated this outcome from the outset, we opted for a transparent approach and provided this level of detail to ensure that no papers were inadvertently overlooked.
Step Two was conducted in two rounds to efficiently manage the large number of papers identified in the initial search.In the first round, a quick review of the abstracts, titles, and keywords allowed us to exclude papers that were clearly irrelevant to our research question (e.g.transport of blood cells, (bone) joint mobility, healthcare employees travelling for non-work purposes).This reduced the number of papers for the second round which involved a more detailed evaluation of the remaining papers to determine their relevance based on specific inclusion and exclusion criteria.Additionally, the papers were classified based on the predominant commute mode shifts discussed (e.g. car use, active transport, or both) during the screening process for easier data overview.These classifications were for organisational purposes and did not necessarily influence the review and analysis sections of the paper.Citations collected from the five databases mentioned above were managed using Endnote 20.4. Figure 1 illustrates a flowchart of the selection process.
The authors agreed upon the inclusion and exclusion criteria to make sure that only the most relevant publications were collected (Table 1).With respect to the criteria, first, included articles were expected to be focused on travel behaviour.More specifically, we were interested in travel to/from work (i.e.commuting) behaviour of workers during the pandemic.While our initial focus was on key workers' commuting, the literature review revealed challenges in precisely delineating key workers as a distinct segment within the commuting population.Consequently, our findings disproportionally pertain to key workers but also extend to other individuals who did not have an essential job but had to commute while some were able or forced to stay at home.Therefore, we use the term "pandemic commuters" as a broader category that encompasses both essential/frontline/key workers and non-essential non-teleworkers (see Appendix for the complete search query).Regarding the research design or socio-demographic characteristics of the studies population, no restrictions were set.From the 1885 identified records, after removing the duplicates, screening for pertinent titles, abstracts, and full texts, 36 publications that matched the eligibility criteria were selected.Further details about the included studies can be found in Table 2.

Lack of consistent definition
Among the reviewed literature, there is a lack of consistent criteria or definition for identifying pandemic commuters, especially in relation to workers with essential jobs.Some studies provided limited information, making it challenging to determine whether the participants were essential or non-essential workers.However, the majority of pandemic commuters were observed to be younger, low-income individuals from minority ethnic/ racial groups who did not own cars and typically held jobs that did not offer the option to work from home (WFH).Examples of such include non-white, Hispanic/Latino in U.S. (Sy et al., 2021), young low-income frequent bus commuters in Indonesia (Bria et al., 2021), and transit-dependent keyworkers in Boston, U.S. (Basu & Ferreira, 2021).Generally, the significance of some socio-economic variables such as income was consistent across the studies whereas for some others context-specific outcomes were observed such as gender, with more female keyworkers in Belgium (Assoumou Ella, 2021) versus more male keyworkers in Bangladesh (Anwari et al., 2021).Concerning the keyworkers' job categories, limited information is provided by the studies.One study in Belgium (Assoumou Ella, 2021) described that the majority of essential workers were female and held occupations with the highest exposure to infection including health-care professionals (e.g.nurses and medical secretaries), supermarket cashier, childhood educators, and cleaners.Other studies that categorised their participants into professional versus non-professional roles (Angell & Potoglou, 2022) or high-skilled white-collar Information and Communication Technology (ICT)-related versus low-skilled blue-collar jobs (defined as technicians and trades, machine operators/drivers and labourer) (Anwari et al., 2021;Balbontin et al., 2022) found that the latter category had limited capacity to WFH than the former.

Quadruple selectivity
There is a quadruple selectivity regarding spatial, sampling, behavioural, and temporal aspects in the published literature on pandemic commuters.First, studies from around the world have been included, although they are dominantly from the global North.The majority of the papers come from Europe: Poland and U.K. (three each), the Netherlands (two), Germany, Belgium, Ireland, Italy, and Portugal (one each); North America: U.S.A. (eight) and Canada (two); and Australia (two).However, studies from Asia (eight), the Middle East and Africa (one each) were also retained.
Second, 25 out of 36 studies applied an online sampling method by distributing their questionnaires either through survey companies or social media channels such as Facebook, Twitter, and WhatsApp while targeting a general population of workers belonging       to different unknown workplaces, and without verification of sampling accuracy.This method not only reduces the accuracy of screening participants for relevancy but also results in a higher percentage of younger participants with higher education levels due to their active presence in social media (e.g.Dias et al., 2021) as well as an under-representation of extremely low-income households (Schaefer et al., 2021) and minorities (e.g.Basu & Ferreira, 2021).Third, studies have focused more on pandemic commuters' mode use than on other aspects of commuting behaviour such as frequency, route, distance, and time of day.There have been many studies on telecommuting by workers in general, but few have looked at how pandemic commuters have changed the frequency with which they travelled to work.Only in one study, among 572 individuals in Bangladesh in May-June 2020, Anwari et al. (2021) found that while certain occupations (e.g.service-holders and businessmen) shifted from trip frequencies of more than 5 times a week to WFH by 10%, day labourers did not have such option especially in rural areas with weaker ICT access.
Fourth, published studies have mostly focused on the first wave, or at least early stages, of the pandemic.At present, there is very little research on commuting in later waves/ stages.This is significant due to (1) the difference between the pandemic waves regarding travel restrictions and breaking news and (2) the effects pursuing beyond a given wave.The latter point suggests that, participants who are studied during the second or later waves of the pandemic, are different from those studied during the first wave (i.e. between March and May 2020 in most countries) due to the cognitive, affective, and behavioural mechanisms they involved in over time (van Wee et al., 2019).It is argued that because of being exposed to new knowledge (e.g. about risks of different ways of transmission), experiencing new emotions or engaging in new behaviours, individuals' mobility conditions in subsequent waves can hardly be considered similar to those in the initial wave.For example, a person comes to believe that PT use increases COVID transmission risks, may then start driving (or walking or cycling) for their commute and come to prefer the new mode.Subsequently, this affective outcome is likely to strengthen the behaviour and/or cognitive beliefs (Kroesen et al., 2023).As a result, the person is no more avoiding PT because of the risks of infection but because of a preference to the new mode.
Among the 3 studies that applied a longitudinal panel method, in a study among thousands of commuters in the Netherlands, Kroesen et al. (2023) discovered that train users with a stronger fear of infection in one pandemic wave tend to use that mode less in a subsequent wave.This study emphasised the bidirectional effect between emotions and behaviour by also showing that higher use of the train led to a decline in the perceived fear of infection.In another study in the U.S., Kim and Kwan (2021) also used mobility data obtained from people's mobile phone signals in 2639 counties from wave 1 (March-June) and wave 2 (June-September).The authors found that there was very little change in mobility levels in wave 2 compared to the V-shaped (sharp decline and then return to the pre-pandemic) trend in wave 1.This was due to social distancing inertia and quarantine fatigue despite the existence of mobility restrictions and COVID cases climbed.Javadinasr et al. (2022) also targeted nearly 3000 respondents across various U.S. states in wave 1 (April-October 2020) and wave 2 (November 2020-May 2021).The authors not only reported a 31% increase in transit use from wave 1 to wave 2, but also a 10% decline in the share of participants who stated, "I no longer feel safe/comfortable sharing space with strangers" (73% versus 63% in wave 1 and wave 2, respectively).It is not clear what the starting point of that response was, but the trends all point to a reduction in fear over time, though where that will plateau is not yet clear.

Variations in commute mode shift
At the aggregate level, a significant modal shift from PT to car and active transport use was observed.Studies reported a range of 9% to 90% decline in PT use, and 5% to 42% and 3% to 33% increase in car and active transport use respectively. 2The main reasons for changing commute mode can be divided into two: concerns related to COVID-19 (e.g.cleanliness and crowdedness) and reduced service levels due to changes in PT capacity and schedules.The two main factors can also relate to each other as reductions in service levels could result in having to queue longer (Caulfield et al., 2021) or concerns related to crowded bus routes (e.g. when non-essential workers returned to regular commute) (Basu & Ferreira, 2021).One study showed that a greater ease of parking (due to fewer cars) at the workplace was one important reason for increased car commuting while reduced congestion and road closures led to 23% increase in cycling (Cusack, 2021).

Car use during the pandemic
The positive association of COVID concerns and less use of PT was found to be more significant in the case of car ownership and pre-pandemic habit of car use for commute.For instance, in a sample of 725 commuters in Bristol, U.K., in May-June 2020, Harrington and Hadjiconstantinou (2022) indicated that 50% of pre-pandemic car commuters (72% of the sample) continued using their car whereas only 3% of the pre-pandemic PT users (28% of the sample) remained on PT during the pandemic.Among the former group, 82% of employees expressed an intention to continue using cars even after the restrictions are lifted, while only 18% of the latter group did so.Data from 1580 commuters in Montreal, Canada, also revealed that car-only mode share that had almost tripled from 9 to 26 percent in the early pandemic (from January to April 2020) declined to 20% in the later pandemic (September) due to lower perceived risk of infection (DeWeese et al., 2022).
Nine studies found that those who already owned and/or used a car for commute were much more critical of PT and indicated higher negative perception of the mode (e.g.transit being unsafe) compared to regular transit users.In a study among 2200 individuals in Boston, U.S. from April to October 2020, Basu and Ferreira (2021) showed that car commuters (73% of the sample) were quite confident (77% agreement) that a private car possibly reduces the virus transmission risks whereas non-car commuters (27% of the sample) displayed a statistically significant lower level of agreement (69%) with this statement.In another study with 787 participants from Australia and New Zealand in July-September 2020, Thomas et al. (2021) found that non-regular PT users (78% of the sample) were more negative towards PT than their regular counterparts (mean attitude of 2.7 versus 1.9 where 1 = extremely positive and 5 = extremely negative).Data from 430 commuters in Jakarta, Indonesia revealed that car and motorcycle ownership had the strongest influence (after income) on choosing these modes for commute, regardless of the workers' perception about health protocols in PT (whether inadequate, poorly adequate, or sufficiently adequate) (Bria et al., 2021).These findings highlight the possibility of scepticism among those who do not use a mode (often PT), and hence, develop an inaccurate negative perception of it (Zarabi et al., 2022).
Furthermore, in their Boston study, Basu and Ferreira (2021) found that the negative perception about mass transit increased the intention to purchase a car in 20% of zerocar households, especially active transport commuters.In a follow up interview, ten households that had purchased a car subsequently named the following reasons for their decisions: uncertainty regarding the transit service frequency, lack of trust in safety measures by the transit agency (e.g.overcrowded buses), and fear of other passengers not following the guidelines.A study from August 2020 in the Philippines similarly suggested the pandemic enhanced intentions to purchase a car, by showing that 6% of the 255 participants intended to purchase one within a year (Garcia et al., 2022).

Cycling and walking during the pandemic
People also chose active transport modes to replace PT trips, although cycling and walking witnessed a smaller boom compared to car.Bike sharing systems (BSS), were also demonstrated to be more resilient than subway with a lower decrease in ridership (71% versus 90%) (Teixeira & Lopes, 2020).This study which was conducted in New York, U.S. during March 2020 reported a 33% increase in the BSS ridership on the day that the subway use started to decline.In a study of the staff and students of the University of Dublin, Ireland (n = 2653) in June-July 2020, the preferred mode of commute after reopening of the campus was active modes with 55% of the sample stating preference for cycling (28%) or walking (27%) compared to 26% prior to the pandemic (Caulfield et al., 2021).In this study, in response to an open-ended question, many participants stated that they had previously considered active transport for their commute, but the inadequate PT during the pandemic gave them the final "push" to make that change.In Bristol, U.K., 22% of 725 survey participants who did not use cycling or walking for commute before the pandemic (only car: 72% and PT: 28%) switched to these modes during the lockdown (May-June 2020) and stated intentions to remain active commuters after the restriction are lifted, albeit at a slightly lower level (16%).
Two studies investigated behavioural beliefs related to active transportation.In a study among 700 commuters in Italy, one month after the summer 2020 lockdown, 90% of the respondents considered walking as a stress reliever and more than 40% of essential workers chose to walk for commute several times per week (compared to a 30% before the outbreak) (Campisi, Tesoriere, et al., 2022).The authors observed an absence of walking frequency beyond three times a day during the lockdown period.This was possibly a consequence of teleworking or loss of jobs as well as keyworkers' modal shift to private car due to feelings of stress associated with the risks of virus transmission while walking for commute.However, the participants believed that the lower walking frequency could negatively influence their mental health.Similar observations were reported in a study on a sample of 213 commuters in Pennsylvania, U.S., in June-August 2020 (Cusack, 2021).Compared to their non-active counterparts, active transport commuters reported greater confidence bicycling in urban areas and the belief that active modes relieve stress and improve mental health.

Experienced emotions by pandemic commuters
In addition to the modal use and changes discussed above, the objective of this research is to assess how pandemic commuters perceived, interpreted, and emotionally experienced external changes and how this affected their commuting.
Negative emotions were a dominant finding.The majority of the reviewed papers (26 out of 36) referred to the pandemic commuters' fear due to a perceived risk of infection as a barrier to use PT, especially in the case of a lack of cleanliness, and other passengers' violating the guidelines (Dias et al., 2021).In a study among 1203 keyworkers in different countries across the world, respondents placed a high priority on COVID-related factors such as "infection concerns"(71%), "safety and security" (66%), "social distance" (64%) "cleanliness" (62%), and "passengers with face masks" (60%) (Abdullah et al., 2020).These concerns led to a 23% decline in PT ridership and 6 and 8 percent rise in car and active transport use, respectively.On the contrary, authors reported that factors such as cost, comfort and travel time saving lost their priority compared to the pre-pandemic situation.Data from a sample of 650 workers in Australia, in September-October 2020, also revealed that biosecurity concerns (fear of virus transmission risk) related to crowdedness and long waiting queues of PT decreased the likelihood of commuters choosing mass transit by 12%.In a study from Poland, Przybylowski et al. (2021) found that 42% of the 302 respondents switched from PT to car, 40% of whom highlighted the following reasons for switching from PT: "number of passengers", "fear of becoming infected", "fear of insufficient disinfection", and "fear of other passengers not following the hygienic regime".
In a sample of 4,359 individuals from Germany's Hanover in June 2020, Schaefer et al. (2021) explored the link between the fear of infection, commute mode and socio-economic variables.Fear was found to have a significant negative effect on PT use and to be a strong predictor of the increase in bike and car use.Fear was also proved to have a mediator effect in the strong association of gender and PT use as women showed a higher level of fear compared to men and therefore reduced PT more.
In Italy, Campisi, Basbas, et al. (2022) investigated the psycho-attitudinal variables of fear, stress, and anxiety of 700 bus commuters during three periods: pre-pandemic, post-lockdown, and wave two.The dominant emotions experienced by bus commuters were stress (∼90% of respondents) and fear (∼80%) during the pandemic's post-lockdown and wave two with the former period associated with a slightly higher percentage.The anxiety level which was much lower than fear and stress during the post-lockdown period (∼53%), reached its peak (∼70%) during the second wave.Despite the existence of such negative emotions, people continued to use PT.Various explanations were found including having no other alternatives (captive riders), the provision of monetary incentives for PT by the government, and the unfavourable driving environment during the pandemic (e.g.travel delays due to police checkpoints of private vehicles).

Implications for understanding the use of commute modes
The review of the papers indicates that the pandemic commuting had various health, financial 3 , and emotional implications for pandemic commuters.These implications differed with respect to their level of influence and how they were dealt with.With pandemic commuters facing the challenges of the new mobility context, they had to adapt and if possible, reconsider their everyday commute patterns.However, the magnitude of this adaption, particularly regarding mode use depends on workers' "adaptive capacity" which is a function of, among others, their travel-related choice sets which are conditioned by socio-demographic status and psychological characteristics, built environment features including transport infrastructure and geographical context (Rahimi et al., 2019), and the government mandated regulations discussed earlier (Angell & Potoglou, 2022;Kim & Kwan, 2021).Figure 2 illustrates the three main factors and provides examples of the emotional, physical, and financial implications discussed in the reviewed papers.
Based on synthesis of findings from the reviewed literature, the subsequent sections will elaborate the following lessons for further research on commute mode use: (1) the assumption that all commuting results from individuals making choices is best avoided; (2) the role of (negative) emotions in commuting should not be overlooked; (3) the importance of proximity to, or separation from, other bodies on how people commute should be considered.Finally, we will discuss the implications for the use of the existing travel behaviour theories for understanding commute mode use.Based on the findings from the review, pandemic commuters primarily belonged to groups whose options were limited in some form or another.This was related to individual factors (female and/or minority racial groups) and/or financial constraints (car-free, low-income households mainly with non-ICT-related jobs) which were further exacerbated by the extreme conditions imposed by the pandemic.Consequently, these individuals did not have the luxury to shift away from PT and were constrained to select from limited alternatives, often compromising their health or taking the risk of losing their job.For instance, a delivery person who lives in a low-density environment and works in a peripheral location like a distribution warehouse and does not own a vehicle is forced to use a bus (or other PT modes) which is onlyif everoperating on an irregular schedule with highly reduced capacity due to government-mandated restrictions during lockdowns.Additionally, the person must confront the fear of infection as they are obligated to opt for in-person versus remote work in the first place, risking job loss otherwise.
These findings raise questions about whether all commuters have a true choiceunderstood as the ability to do otherwise if all other factors and conditions are the samewhen it comes to the transport modes they use.Most contemporary transport research tends to assume so, despite scholars in the past already expressing doubts about this (e.g.Burnett, 1980;Swait & Ben-Akiva, 1987).However, a distinction is in order between the availability of choice alternatives and neurological processes responsible for choice.
The above results for pandemic commuters reinforce the importance of considering the availability of alternatives in travel behaviour research and adapting analysis and frameworks accordingly.This point has direct relevance in the application of the TPB and the NAM, because both assume that individuals have a degree of choice over whether to perform a given behaviour, such as driving a car or using PT for commuting.Neither model directly analyses the choice between two alternatives, say driving versus using PT, but both are still based on the assumption that people can do something or refrain from this if all other factors and conditions are the same.Based on our findings we submit that use of models like the TPB and NAM is best restricted to people who (have the luxury of) choice.Alternatively, these models need to be extended to consider the varying level of choice individuals are able to exercise over commute mode use (as in Section 4.4).Therefore, we carefully differentiate between the terms "mode choice" and "mode use" throughout this paper, using the most appropriate of the two where relevant.Beyond conceptual reasons, considering choice alternatives is vital from a social inequality perspective, as our results indicate that marginalised populations are more likely to have limited access to alternative commute modes.
In the random utility theory (RUT) approach to travel behaviour, the need to specify choice sets correctly has received some attention (Li et al., 2015;Swait & Ben-Akiva, 1987).It is argued that ignoring or inaccurate specification of choice sets, for instance because a certain proportion of the population is captive to a particular mode, can lead to a biased understanding of the associated behaviour.Work along these lines needs to be advanced such that considering choice set generation becomes standard practice in RUT-based analyses of commute mode use.We are similarly sympathetic to work by De Vos (2022) who has recently hypothesised that travellers are more likely to exhibit attitude-congruent travel behaviour if they enjoy large freedom of choice while attitude-incongruous behaviour is more likely if choices are constrained.It remains unclear, however, how freedom of choice can be assessed reliably in future studies.
In the longer term, transportation researchers should also pay greater attention to the neurological processes implicated in commute mode use.Since the first experimental studies showed brain activity to precede the time when human beings (appear to) make a conscious decision (Libet et al., 1983), there has been multi-disciplinary debate on whether human decisions are subjectively "free", and whether intention is the antecedent of behaviour and choice or rather the "backward-looking" follow-up to the brain signals that initiate behaviour.While unresolved (Delnatte et al., 2023), this debate matters to transportation studies.Not only does it raise questions about the theoretical adequacy of models like the TPB and conventional understandings of travel choice, but also brings questions about the interplay of contextual changes like those during the COVID-19 pandemic with human brain processes: in how far was that interplay different for pandemic commuters than for other workers and why?This particular question can no longer be answered, but research on mode use, brain processes, and the social and spatial context in which these occur is a particularly fertile area for future research (Eagly & Chaiken, 1993;van Wee et al., 2019).

Treating emotions differently
Our systematic review highlights the powerful role of negative emotions, particularly fear (and related stress and anxiety), in both motivating and inhibiting behaviour change for commute mode use over the course of the pandemic.These negative emotions were linked to the perceived risk of infection and the potential life-threatening consequences associated with exposure to infected air and/or surfaces as well as proximity to other bodies (and in some cases fear of empty PT spaces).The corporal proximity during the COVID-19 pandemic raised serious concerns about the safety of PT as individuals are normally close to each other in such settings and older systems may have poor ventilation, facilitating the spread of infectious diseases (Fadaei, 2021).Additionally, the asymptomatic nature of the coronavirus added to the fear of proximity, as infected individuals can spread the virus unknowingly (Wang et al., 2020).It could take even some weeks before a PT passenger could ascertain whether there had been asymptomatic virus carriers around them.This prolonged exposure to uncertainty and worrisome situations has contributed to a decline in interpersonal trust termed "social scarring" (Fang et al., 2023), particularly in crowded public spaces like PT (Navarrete-Hernandez et al., 2023).Therefore, from the emergence of the COVID-19 onwards, proximity to other bodies that may sneeze, cough, or exhibit signs of illness has become a vital concern among some PT passengers. 4 As such, these emotions potentially have longer-term consequences in the form of a stronger intention to obtain and use a private vehicle and a reluctance to (continue to) commute by PT.
We interpret this as a need for treating emotions in travel behaviour research in a manner that gives full and equal consideration to their full range, from positive to negative.A range of negative emotions has been considered in studies, including stress during the commute (e.g.Sannasi et al., 2022), fear of crime (e.g.Heinen, 2023;Truong & Currie, 2019) and gender-based violence (e.g.Infante-Vargas & Boyer, 2022).Regarding the fear of infection, one study in Taipei, Taiwan found that fear of infection caused a 50% decline in underground transit ridership during the peak of the epidemic (e.g.Wang, 2014).Nonetheless, it seems that positive emotions such as (mode-specific) travel liking (e.g.Morris & Guerra, 2015) and positive affects during travel and commute, have received greater attention.Many studies have linked positive affects to various factors associated with car use, including perceived reliability (Friman et al., 2017;Mokhtarian et al., 2004), a sense of control and freedom (Anable & Gatersleben, 2005), prestige, the driver's ability to project positive qualities like mastery and skill onto themselves, and feelings of protection (ontological security) (Hiscock et al., 2002).
The need to treat emotions symmetrically extends to studies of commute mode use deploying the TPB and/or NAM. 5 These frameworks often tend to over-rationalize behaviour, overlooking two important points regarding the significant role of emotions.First, the sense of protection offered by using a car or bicycle, fuelled by the heightened fears discussed earlier, can become a priority for commuters, surpassing factors such as travel time, comfort, and cost.This emphasis on protection may lead to decisions that may be perceived as "irrational" or avoidance of confronting the problem.Zhao and Gao (2022) also found that high levels of anxiety can damage individual's peace of mind and lead them to an overestimate the consequences of their behaviour.These observations aligns with previous research suggesting that people's inclination to seek positive emotions and avoid negative ones (Frijda, 2017), influences their behaviour.Individuals are generally found to avoid danger more than they seek safety, and this disproportion is stronger for safety than for finance (Tversky & Kahneman, 1991).As such, the desire to avoid a potentially risky situation on PT (due to infection or the absence of safety in numbers) was likely a stronger influence than financial costs due to using more expensive means of travel.Second, studies of commute mode use that utilise the TPB should explore the possibility of a direct influence of emotions on behavioural intentions, independent of attitudes.While traditionally emotions are considered to influence behavioural intentions through attitudes, the findings from the reviewed studies challenge this perspective.They indicate that emotions, particularly when they are intense and not yet long-lasting enough to shape attitudes, can directly and immediately drive intentions, regardless of individuals' pre-existing attitudes or beliefs about a particular behaviour.
Furthermore, among the reviewed studies, some argued that repeated engagement in potentially risky behaviour (e.g.taking the commuter train) can lead to a decline in the feeling of fear, but not attitudes, over time, even in the persistence of the external sources of fear.This can occur through an internalisation process where, individuals must convince themselves that their actions (even those forced upon them) are reasonable and necessary, or simply that fear is reduced through repeated action without negative consequence.A decline in negative emotions can also occur through two other mechanisms.The first of these occurs when individuals possess sufficient awareness and perceived knowledge regarding the consequences of a given behaviour.Knowledgeable passengers are found to experience lower levels of fear, hence, they outweigh the benefits of using PT and the associated socio-economic benefits over a small possibility of catching coronavirus caused by using PT (Zhao & Gao, 2022).The second relates to the individuals' perception of the psychological distance (Liberman et al., 2007) from the threatening event (McGraw et al., 2012).A study conducted in eight cities of China, found a positive association between PT passengers' anxiety and being closer to the pandemic (Dong et al., 2021).Therefore, with the pandemic's severity largely contained and an increased psychological distance, reduced fear and anxiety can positively influence the inclination of pandemic commuters to revert to using PT, a result that would be appreciated by PT practitioners.
Future research on commute mode use should not only treat emotions symmetrically.It should also recognise that models like the TPBand other approaches that rely on people's self-reporting of emotions, as in much research on subjective well-being (e.g.Ettema et al., 2011;Friman et al., 2017) treat emotions as something of which people are consciously aware.Emotions and affect can, and have been, conceptualised in other, more embodied ways which involve automatic physiological responses and bodily sensations such as changes in heart rate, breathing, and muscle tension in response to external stimuli.For instance, studies have shown that embodied stress can operate below the threshold of consciousness, resulting in different manifestations compared to what individuals are cognitively aware of (Teixeira et al., 2020).The existing studies have mainly focused on examining the emotions that travellers experience during or after their trips and how these trips influence emotions.To further advance the field, future research could explore the reverse relationship, investigating how emotions impact modal use and other aspects of travel behaviour, such as travel frequency and time of departure.One approach could involve real-time phone surveys, where participants report their emotions before starting their daily travel.Additionally, integrating physiological measurements, such as skin conductivity and temperature, using wearable sensors worn by participants, could offer valuable perspectives.Given the integral role of both cognitive and embodied aspects in emotional experiences, a combination of these methods is recommended for assessing individual emotions in future studies.

The potential norm-changing consequences of the COVID-19 pandemic
According to the NAM, personal norms determine what individuals believe and how they behave through awareness of consequences and ascription of responsibility (Schwartz, 1977).According to Bouman et al. (2021), if it was not because of strong personal norms and the feelings of being morally responsible, no rapid and powerful prosocial actions were taken during the pandemic just like many other global environmental crises where public responses are less forceful.Building on NAM, we hypothesise that: (1) The perceived knowledge of the coronavirus's ways of spreading, such as through saliva droplets and nasal discharge (WHO, 2020) and the ensuing health threats, likely heightened commuters' awareness of the negative outcomes associated with choosing PT as a "risky" mode.(2) Consequently, this heightened awareness can trigger feelings of personal responsibility and/or fear regarding potential negative outcomes, such as the risk of contracting or transmitting the virus to others, particularly in close contact with immunocompromised individuals (variable of "proximity" in Figure 3).(3) These mechanisms can evoke or strengthen personal norms and guide decisionmaking regarding commute mode use, potentially leading to the avoidance of PT and a shift towards alternatives modes of transportation or WFH.
The caring about hygiene and health of otherswhether known or unknownas a determinant of travel mode use reflects an altruistic and prosocial behaviour that was not pronounced prior to the emergence of the COVID-19 pandemic.If this trend continues, commuters (who have a choice) are likely to continue avoiding PT and instead opt for individual modes such as cars, bikes, or WFH as soon as they feel sick or display COVID/flu-like symptoms.This avoidance behaviour may also extend to caregivers when their children or care-recipients become ill, necessitating time off from work to reduce the risk of disease transmission (whereas before it might only have been acceptable if the child was very sick).These changes in social norms can potentially lead to policy adjustments, such as employers allowing keyworkers the option to WFH or providing caregiver-related vacation days.In that respect, the existing literature on commuting behaviour during COVID has primarily focused on individuals themselves rather than considering the wider network of people with whom the traveller interacts, shares space, and cares about protecting.Furthermore, it is worthwhile that future studies examine whether this altruistic behaviour will persist beyond the pandemic, given the possibility of forgetfulness and selfishness among some cultures/individuals.
The long-term effects of these changes and the social scarring mentioned earlier can have significant implications for commuting behaviour and addressing overcrowding on PT.They also have longer-term consequences for policy attempts aimed at promoting PT usage (Rothengatter et al., 2021).Rebuilding trust and instilling a sense of security in PT may require time and concerted efforts, considering that these changes vary among individuals and communities (Fang et al., 2023).With the shifting norms influenced by the pandemic and their likely continuation, it is advisable for future research to place more emphasis on filling this knowledge gap in our understanding of commute mode use.Understanding and addressing these social and psychological impacts is crucial for designing effective strategies and interventions to support individuals' well-being and facilitate the recovery of PT usage.

Lessons learnt for how commute mode use is theorised
The above reflections on choice, (negative) emotions and norms prompted by corporeal proximity to other human beings reinforce previous statements about the importance of extending the TPB and NAM and their synthesis (Figure 3).By recognising that commuting is not always an outcome of making choices we can refine the TPB and NAM by acknowledging that assessing intentions based on attitudes, PBC, and social/personal norms is relevant only in the presence of a choice alternatives, encompassing the decision of whether to commute and which mode to select.It is therefore suggested that future research that deploy these models account for the varying levels of choice individuals have in their commute mode decisions.As Figure 3 illustrates, the absence of choice alternatives may lead to either the "use" of whatever alternative exists or fundamental changes in one's commute or employment situation.Conversely, the availability of alternatives allows for the commonly understood decision-making process where (both positive and negative) emotions and experienced proximity become crucial factors in commute mode decisions.

Conclusion
This study has investigated the influences of the COVID-19 on pandemic commuters' mode use.Through a systematic review of 36 scientific publications, the study has identified two significant findings related to the intertwined institutional, physical, and sociopsychological processes that influenced the adoption of different transport modes.Firstly, reduced PT services and concerns surrounding COVID-19 have led to significant shifts away from PT towards private car use and active travel among pandemic commuters.Secondly, PT avoidance is more likely pronounced among those with pre-existing car use habits and heightened negative emotions like fear triggered by the potential health risks resulting mainly from proximity to other bodies.Additionally, the study revealed a significant link between commute mode use challenges and low socio-economic status groups during the pandemic, underscoring the importance of addressing inequality in transportation access and resilience during crises.
Additionally, the study offered new insights into the traditional understanding of the TPB and NAM.Firstly, it questioned the assumption of a standard set of travel mode alternatives for commuters, shedding light on the complexities of mode "choice" versus "use".Secondly, it emphasised the significant role of (negative) emotions, triggered by physical proximity to others, in commuting mode use, regardless of attitudes.It underscored the importance of treating emotions differently, considering the full range from positive to negative in travel behaviour research, and paying greater attention to embodied affect exerting influences below the threshold of consciousness.Lastly, the study highlighted the potential impact of the COVID-19 pandemic on transforming modeuse-related norms and the necessity for future research to explore its long-term policy implications.Thus, employing more rigorous longitudinal research methods to study post-pandemic waves is highly recommended.

Figure 1 .
Figure 1.Flowchart of the article selection process.

Figure 2 .
Figure 2. Factors that influence mode use for pandemic commuting.

Figure 3 .
Figure 3. Theoretical model for understanding commute mode use, inspired by the TPB and NAM.The blue colours, including the arrows and the boxes, represent the variables that were emphasised in the literature of pandemic commuters.

Table 1 .
Eligibility criteria to selection of publications.

Table 2 .
Characteristics of included studies and overview of the results.
TRANSPORT REVIEWS