Frequent public transit users views and attitudes toward cycling in Canada in the context of the COVID-19 pandemic

Several Canadian cities observed a shift from public transit use to private cars and active transport modes during the COVID-19 pandemic. At a moment where pre-pandemic public transit users are reconsidering their travel options, studies describing their attitudes toward cycling are lacking. Because most trips in urban areas involve short- and mid-range travel, cycling is seen as a promising environmentally sustainable means of transportation. This study aims to describe how pre-pandemic public transit users in Toronto and Vancouver view cycling, including their comfort with available infrastructure, cycling frequency, and perceived barriers to adoption. Data from the Public Transit and COVID-19 Survey, a web-based panel survey of over 3500 regular transit riders in Toronto and Vancouver administered in May 2020 and April 2021 were analysed. Applying Geller's typology, 70% of participants could be classified as interested but concerned and one fifth as no way no how regarding their comfort levels toward cycling. Women were more likely to be no way no how cyclist type. Weather, lack of safe routes, and having to carry things were the main barriers to cycling in both cities. Our results give insight on who should be targeted by city initiatives aiming to promote changes toward more active modes of transportation. Further studies with a causal design are required to identify possible mitigating strategies to the main barriers to cycling.


Introduction
COVID-19 has transformed transportation mode shares in cities around the world. The lockdowns across Europe and North America in the early phase of the pandemic (March to May 2020) led to drastic declines in road traffic volumes ( Budd and Ison, 2020 ;Katrakazas et al., 2020 ;Vingilis et al., 2020 ;Oguzoglu, 2020 ). In Ontario, Canada, for example, early data suggest a 60% decrease in private car volume from March to mid-May 2020, followed by a rapid increase during the summer of that same year, 45% above prior years ( Apple, 2020 ). When governments started relaxing some public health restrictions, several cities launched local actions such as 'pop-up' bike lanes, shared spaces, and pedestrian streets to support walking and cycling when physical distancing measures were still in place ( Budd and Ison, 2020 ;Pedbikeinfo, 2020 ;European Transport Safety Council, 2020 ;Combs and Pardo, 2021 ;Fischer and Winters, 2021 ). Alongside these changes, a shift from public transit use to private cars and active transport modes such as biking, and walking was observed ( De Vos, 2020 ;Palm et al., 2020 ;APTA 2021 ;Buehler and Pucher, 2021 ).
Cycling increased considerably from 2019 to 2020 and 2021 in most cities in Europe, North America, and Australia ( Heydari et al., 2021 ;Rérat et al., 2022 ;Qiao, 2021 ). That increase was larger when periods of total lockdowns are excluded ( Buehler and Pucher, 2021 ). Many factors suggest that the increase in cycling levels during the COVID-19 pandemic may persist over the coming years. These factors include first the expansion and improvement of cycling infrastructure, both completed and planned. In 2016, the City of Toronto has approved a 10-Year Plan to renew and grow the city's bicycle infrastructure ( City of Toronto 2016 ). Similar plans have been adopted in other Canadian cities ( City of Vancouver 2018 ; 2021 ). Second, a boom in bicycle sales was observed across Canada since the pandemic started ( Mazerolle, 2021 ). Finally, some of the increased cycling in 2020 may at least in part be attributable to former public transport passengers who shifted to cycling because they were afraid to ride potentially crowded trains and buses during the pandemic ( Buehler and Pucher, 2021 ).
Pre-pandemic public transit users represent a large share of the population in major urban areas. Although they represent an important potential bicycle-ridership market, studies describing their view and attitudes toward cycling are lacking. This study adds to a rapidly growing literature of the pandemic effects on public transportation Loa et al., 2021 ), by filling this gap. Because many of these riders are forced to actively reconsider their travel patterns, including their choice of modes, it is important to understand their views about cycling at this point to offer an alternative to the use of a private car.
This study, therefore, aims to describe how frequent public transit users before the pandemic in Toronto and Vancouver view cycling, including their comfort with available infrastructure, cycling frequency, and perceived barriers to adoption. The article has the following structure. Section 2 provides a brief literature review on public transit users views on cycling. Section 3 details the data and methods used for analyses. Section 4 present findings and Section 5 discuss the findings in relation to the existing literature. Section 6 provides the main conclusions and future work to deepen our understandings.

Literature review
Previous work on factors that may influence cycling uptake in Canada generally focused on frequent cyclists ( Damant-Sirois et al., 2014 ;Winters et al., 2019 ), university students ( Moniruzzaman and Farber, 2018 ;Mitra and Nash, 2019 ), and the general population ( Young et al., 2021 ). Therefore, they were limited in terms of providing insights for people who might be willing to adopt cycling as a mode of transportation but have some concerns. In the context of combating climate change and reducing greenhouse gas emissions, cycling is seen as a promising alternative means of transport in urban areas ( Pucher and Buehler, 2005 ). This is because cycling is an effective mode for short-and mid-range travel, as is the case for most trips in urban areas ( Houde et al., 2018 ;Fishman, 2016 ;Pucher et al., 2010 ). These potentials highlight the need to identify ways to encourage this form of transportation.
Social norms, the importance that individuals attach to health, environmental benefits, and positive attitudes toward cycling have been associated with a higher propensity to cycle ( Willis, 2015 ;Fernández-Heredia et al., 2014 ;Emond and Handy, 2012 ;Heinen et al., 2011 ;Pooley et al., 2012 ). In Canada, east/west regional differences in support of cycling as a common good have been described ( Scott, 2020 ). Ontario (east) residents were found to have higher odds of agreeing that bike lanes make a community a better place to live than participants from British Columbia (west). Similar results were found when participants were asked if separated bike lanes are a good thing and the authors also found very similar results when only metropolitan areas were considered.
Perceptions also play an important role, with a study suggesting that the perception of the built environment may have a higher impact on cycling behaviour compared to objective measures of the bicycling environment ( Ma et al., 2014 ). Furthermore, non-cyclists tend to perceive more barriers to cycling than utility cyclists and vice versa ( Gatersleben and Appleton, 2007 ). Perception may vary by socio-demographic characteristics ( Ma et al., 2014 ;Branion-Calles et al., 2019 ;Sottile et al., 2019 ). Age, gender, income, ethnicity, car and bike possession have been identified as determinants of bicycle use ( Houde et al., 2018 ;Fernández-Heredia et al., 2016 ). In fact, there is a gender gap in cycling across urban regions with low rates of cycling among women in countries such as Canada and the United States ( Mitra and Nash, 2019 ;El-Geneidy, 2015 , Garrard et al., 2012 ;Pucher and Buehler, 2012 ). For example, in the Greater Toronto and Hamilton Area, only 29% of current cyclists are women ( Mitra et al., 2016 ). However, in countries with high cycling rates and better infrastructures, such as the Netherlands and Denmark, cycling is more evenly spread across the two genders ( Sottile et al., 2019 ;Heinen et al., 2013 ).
Other factors, such as travel distance, has been mentioned in previous studies. In general, shorter distances ( < 10 km) are associated with a higher propensity to cycle to work or campus ( Moniruzzaman and Farber, 2018 ;Mitra and Nash, 2019 ;Sears et al., 2012 ;Nahal and Mitra, 2018 ). Day-to-day weather conditions such as precipitation and freezing temperatures in countries with northern climatic conditions are known barriers that discourage cycling, especially during the winter months ( Winters et al., 2007 ;Bergström and Magnusson, 2003 ). A study investigating travel behaviours of current cyclists who commute to the Toronto Metropolitan University found that the density of bicycle infrastructure within 500 meters of the shortest route was positively associated with all-season cycling ( Nahal and Mitra, 2018 ).

Data and methods
Data from the Public Transit and COVID-19 Survey ( Zhang et al., 2020 ) are used. In May 2020, over 3500 regular transit riders in Toronto and Vancouver were surveyed on how COVID-19 influenced their trips and their ability to reach essential destinations like healthcare facilities and grocery stores. Toronto and Vancouver are respectively the first and the 8th city in Canada in terms of population, although Toronto population is four times larger than Vancouver (2.8 compared to 0.7 million) ( Statistics Canada, 2022 ). The Weather (7) During each of the following seasons below, how often did you bicycle for non-recreational purposes (i.e., to go somewhere): Summer 2020 (June, July, August), Fall 2020 (September, October, November), Winter 2020-1 (December, January, February, March)?
Could select only one answer Did not cycle at all Less than weekly 1-3 times/week 4-7 times/week (8) In the last 30 days, how many times have you experienced a crash or a 'near crash' (when you felt you were almost hit or injured) while riding a bicycle?
Only numbers were allowed Both cities have similar population densities, and adopted the Vision-Zero approach to road safety and have plans to support safe cycling ( Verlinden, 2019 ;Canada Population, 2022 ). Recruitment took place primarily via Facebook, using its spatial targeting tool to distribute ads only to users who lived or worked within the city limits of Toronto or Vancouver. Community listservs were also used, and eligibility to participate was limited to adults who were regular transit users, defined as those who rode transit more than once per week prior to the COVID-19 outbreak. Details on the data collection methods are documented in prior publications ( Zhang et al., 2020 ;. As the survey used Facebook for recruitment, women, people in their 30s, and lower-income households are over represented in this sample ( Zhang et al., 2020 ).
From March 16th to April 6, 2021, a follow-up survey of the same respondents was performed, adding questions on cycling attitudes ( Table 1 ). The survey also asked about the presence of any disability that prevents the use of walking/cycling. Just over half ( n = 1941, 55%) of first survey wave participants agreed to participate in the second wave and they constitute the study sample.
Several sensitivity analyses were performed. First, censoring weights were applied to adjust for the fact that only 55% of participants agreed to participate in the second wave ( Mansournia and Altman, 2016 ). Second, given the non-representativeness of the study  ( Palm et al., 2020 ;Ye et al., 2009 ). Similar external data representative of Vancouver public transit users was not available. The IPU is a methodology generally used to match the distribution of some characteristics in a survey to those of a census-based marginal distribution. The complete questionnaire as well as program codes used to produce the tables and figures are available here: https://osf.io/gx7wm/?view_only = 1c7d6b297da14319b81a9ac50a00f1d5 .
Based on Geller's typology , responses to questions 1 and 2.a were used to classify participants into four cyclist types, on the basis of their current and potential attitudes toward bicycle riding ( Table 2 ) ( Geller, 2006 ;Dill and McNeil, 2016 ): strong and fearless, enthused and confident, interested but concerned, and no way no how. Geller's typology is useful to distinguish potential markets for cycling ( Dill and McNeil, 2013 ). Questions 2.b to 8 were used to complement our understanding of participant views on cycling. For questions 2 and 3, the six-point Likert scale was converted into a categorical variable (Agree; neither; disagree and not applicable).
The Geller's typology has been criticized because it does not indicate whether a respondent currently bikes ( Damant-Sirois et al., 2014 ;McNeil, 2016 , 2013 ). That is why questions about cycling frequencies for non-recreational purposes were also asked. Another critique of Geller's typology consists of the fact that no other type of potential interventions besides cycle paths can be recommended using this typology ( Damant-Sirois et al., 2014 ). To address that issue, we asked questions about barriers to cycling. Despite these limitations, the Geller typology can help identify potential cyclists, allowing for a better understanding of travel behaviour and mobility styles specific to each cyclist type ( Dill and McNeil, 2013 ). This could serve as the basis for formulating plans that aim to change commuter preferences in urban mobility choices, more especially on the interested but concerned who may be more likely to adopt the bicycle or increase its use compared to other cyclist types ( Sottile et al., 2019 ).
This study received the Social Sciences, Humanities and Education Research Ethics Board approval from the University of Toronto and all participants provided online informed consent.

Results
Sociodemographic characteristics in the first and second waves were similar except for the proportion of women participants, which was higher in the second wave ( Table 3 ).

Cyclist types
Applying Geller's typology to the sample, the distribution of cyclist types was similar between Toronto and Vancouver ( Table 4 ). The strong and fearless represents 3% of participants; enthused and confident, 5%; interested but concerned 71% and the no way no how account for 19%. Some patterns emerged according to socio-demographic characteristics ( Fig. 1 a). Compared to men, women participants in both cities were less likely to be in the strong and fearless group (2% versus 5%), and more likely to be in the no way no how group (22% versus 13%). Older age participant categories were related to an increase in proportions of no way no how and decrease in interested but concerned cyclist types. Although the proportion of no way no how decreases as participants' income increased, the inverse was observed for the proportion of the interested but concerned cyclist type. Finally, in Toronto, the proportion of no way no how cyclist type was higher among blacks and indigenous compared to other ethnicities ( Fig. 1 a).

Transport habits
Differences between Toronto and Vancouver participants were observed in terms of transport habits ( Fig. 1 b). Participants in Vancouver had more access to a bike and a car and were more likely to have boarded a transit vehicle since May 2020 compared to those in Toronto. However, in both cities, the no way no how cyclist type had lower access to a bike, and to a car (bike and car-sharing  services were considered). Other differences between the two cities were observed for questions 4 and 5 (purchasing and access to a bike). While in Toronto the enthused and confident had the highest proportion for purchasing a bicycle in the last 12 months (15%), it was the strong and fearless who had the highest proportion (21%) in Vancouver. In addition, the strong and fearless in Toronto had the highest proportion (17%) for purchasing a car, while in Vancouver they had the lowest proportion (4%). Surprisingly, in both cities, the proportion of participants who had boarded a public transit vehicle since May 2020 was similar across cyclist types, except for strong and fearless in Toronto who had a slightly higher proportion.

Attitudes and perceptions related to cycling
Regarding statements about cycling, overall, there was no difference between the two cities in terms of people biking more because of COVID-19 (question 2a). However, as expected, the enthused and confident and interested but concerned cyclist types were more likely to cycle more because of COVID-19 ( Fig. 2 ). More participants in Toronto agreed that they do not feel safe cycling because of traffic in their area (statement 2c) compared to participants in Vancouver (61% versus 44%). In both cities the interested but concerned and no way no how cyclist types were more likely to agree to that statement. The inverse pattern was observed for the statement 2dthe city has invested in more bike infrastructure in my area: 51% of participants in Vancouver (versus 42% in Toronto) agreed with the statement. In addition, in Vancouver, an increase in the proportion of people agreeing to the statement 2d across cyclist types (44% for the no way no how to 71% for the strong and fearless) was also observed. As expected, the strong and fearless were more likely to disagree with the statement 3a -there are not enough safe cycling routes in my part of the city . Finally, most participants agreed that giving more street space for walking and cycling made them feel more comfortable during COVID-19, and this was true for all cyclist types.

Barriers to cycling
Several barriers to cycling were listed by participants, and those cited by at least 10% of participants are presented in a chart ( Fig. 3 ). In Toronto, weather was the main barrier to cycling for the strong and fearless (45%), enthused and confident (78%) , and the second barrier for the interested but concerned (61%). The lack of safe routes was the main barrier for the interested but concerned   (64%). Feeling uncomfortable was listed by 35% of the no way no how, and the strong and fearless ( Fig. 3 a). In Vancouver, weather was the main barrier for all cyclist types: strong and fearless (50%), enthused and confident (56%), interested but concerned (62%) and no way no how (43%). However, in both cities, having to carry things, fear of theft, and hills were also major barriers to cycling for the interested but concerned and no way no how ( Fig. 3 b) .

Cycling frequency
In terms of cycling frequency for non-recreational purposes (i.e., to go somewhere), most participants did not cycle at all during the last three seasons (summer 2020, fall 2020 and winter 2020/2021). In Toronto, 53% of the interested but concerned did not cycle at all, and just 29% biked at least once per week during the summer of 2020. The corresponding proportions were 94% and 2% for the no way no how ; 49% and 38% for the strong and fearless and 22% and 65% for the enthused and confident ( Fig. 4 a) . Although the proportion of participants who did not cycle at all increase during the following fall and winter, differences in cycling frequency among the four-cyclist types remained. A similar pattern was observed in Vancouver, except for the strong and fearless for whom only a third did not cycle at all (compared to 49% in Toronto) and half of them, biked at least once per week during the summer of 2020 ( Fig. 4 b). Finally, given that most participants did not cycle at all during the winter 2021 season, 93% in Toronto and 88% in Vancouver did not answer question 8 -the number of crashes/near crashes during the last 30 days . Although in the early phase of the pandemic (spring 2020) Toronto had a higher COVID-19 weekly admission rates than Vancouver, both cities observed an increase in hospitalization rates during the fall 2020 and winter 2021 seasons ( Supplemental Fig. 1 ).

Sensitivity analyses
Applying censoring weights and the IPU algorithm give similar results ( Supplemental Tables and Figures ). The few noticeable changes compared to primary results were the fact that a higher proportion of enthused and confident as well as strong and fearless have bought a car. Also, the enthused and confident in the weighted analyses were less in agreement with statement 2d -the city has invested in more bike infrastructure in my area (48% in the primary results versus 35%).

Discussion
As our results, a national survey of adults living in the 50 largest United States (U.S.) Census Bureau-designated metropolitan statistical areas, using the Geller's typology to classify people's views on cycling, also found that the interested but concerned group makes up the largest group of cyclists, although observed proportions were lower (60% versus 71% in our results) ( Dill and McNeil, 2016 ). That U.S. survey also found a higher proportion of the no way no how, close to one third compared to the 20% obtained in this survey. These differences might be explained by the fact that this study over represents people in their thirties, which are millennial and tend to be more in the interested but concerned group ( Dill and McNeil, 2016 ). However, sensitivity analyses applying IPU to the sample of Toronto participants yield similar results. Therefore, the more plausible explanation might be that only frequent public transit users were included in this study, compared to the general adult population for the U.S. Survey.
Our results demonstrate a clear difference by gender, with women being more likely to be in the no way no how cyclist type. This is an accordance with the gender gap in cycling mentioned in the literature review section above. Similarly, black and Indigenous respondents were overrepresented in the no way no how cyclist type. These results are not surprising since low-income people, women and racialized groups have often been overlooked in terms of their inclusion in the planning and decision-making process related to mobility ( Doran et al., 2021 ). For example, a systematic review found good evidence that women express stronger preferences for greater segregation from motor vehicles than men ( Aldred et al., 2017 ). In order to create a mass cycling culture, specific actions targeting infrastructure and policies toward increasing bicycle use by underrepresented groups are necessary ( Aldred et al., 2016 ).
The fact that weather conditions were respectively the first and second barriers to cycling for the interested but concerned in Vancouver and Toronto is also in accordance with previous findings. Better snow and gravel removal and greater density of bicycle infrastructure have been listed as mitigating strategies to increase cycling during the winter in Canada ( Nahal and Mitra, 2018 ;Amiri and Sadeghpour, 2015 ). Even if countries like the Netherlands, Denmark, and Sweden have done better in enabling winter cycling through well-organized snow clearance, the influence of good bicycle infrastructure in mitigating barriers to winter cycling requires systematic examination, especially in a North American context with longer and colder winters ( Nahal and Mitra, 2018 ). Having to carry things and hills were also found to be important barriers to cycling in this study. This suggests that initiatives promoting e-bikes and cargo bikes might be helpful as they have been associated with reducing problems of hilliness and physical strain as barriers to cycling ( Fyhri et al., 2017 ). However, it is worth nothing that distance was not added among the potential barriers to cycling in our questionnaire as we do not focus on a particular travel purpose.
There was no clear difference between Toronto and Vancouver participants, except for three main observations:(1) participants in Vancouver had more access to a bike, a car, and have taken public transit more since May 2020; (2) more Toronto participants agree that they do not feel safe cycling because of traffic in their area; (3) More Vancouver participants agree that their city has invested in more bike infrastructure in their area. This last observation seems to be in line with data suggesting a higher level of cycling to work in Vancouver (6.1%) compared to Toronto (2.8%), but a more rapid growth in levels of cycling to work during the 1996-2016 period in Toronto (1.1% to 2.7%) than in Vancouver (3.3% to 6.1%) ( Verlinden et al., 2019 ). However, that same observation seems to contradict the east/west regional differences in support of cycling as a common good in Canada described in the literature review section ( Scott, 2020 ). This apparent contradiction could be the result of the fact that giving the probable lower support for bike lanes in Vancouver, even a small investment in bike infrastructure might be sufficient to achieve a high level of agreement to the statement 2d -the city has invested in more bike infrastructure in my area.
Despite being original and informative, this study has limitations. By design, it is non-representative of all pre-COVID-19 public transit users in both cities. A comparison of Toronto participants with the Transportation Tomorrow Survey , the region's household travel survey which covers daily public transit users suggests that people living in the north-eastern part of the city were underrepresented while those living downtown were overrepresented . However, applying the Iterative Proportional Updating algorithm to match the Transportation Tomorrow Survey marginal distribution of gender, age and income yields similar results. Our results might therefore generalize to other North American cities with comparable population densities and cycling infrastructure. Data on cycling comfort questions for 52 (3%) participants (27 in Toronto and 25 in Vancouver) were missing, which preluded us to classify these participants within a cyclist type. In addition, data on safety outcomes such as the number of crashes and near-crashes during the last 30 days were also missing, as most participants did not cycle at all during the 2021 winter. Finally, our estimates of cycling frequencies might be biased. A recent study comparing user surveys and trip record data suggest that travel frequencies based on the former tend to underestimate the level of cycling in older people and overestimate the level in other age groups ( Hua et al., 2022 ).

Conclusion
At a moment where public transit users are reconsidering their travel options, studies describing their attitudes toward cycling are needed to help ensure these former riders embrace sustainable alternatives instead of driving. In this survey of public transit users in two Canadian cities, Toronto and Vancouver, it was found that around 70% of participants could be classified as interested but concerned and almost one fifth as no way no how regarding their comfort levels toward cycling. These proportions suggest that planners have ample opportunities to steer transit riders' weary of public transit to cycling more as an alternative. This study adds to the current literature by exploring pre-pandemic public transit users' comfort, attitudes, and barriers toward cycling in Canada. The fact that the enthused and confident and the interested but concerned cyclist types were more likely to cycle more because of COVID-19 might suggest these cities' efforts to promote cycling during physical distancing are working. However, other than lack of safe routes, weather, having to carry things and hills were also main barriers to cycling. These findings require further study with a causal design to assess if e-bikes and cargo bikes sharing could be possible mitigating strategies. Applying machine learning approaches to classify potential cyclists could be explored too. Finally, additional data from trips records are necessary to assess how recent city initiatives to promote active modes of transportation are impacting each cyclist type. Transportation changes consequent to the COVID-19 pandemic represents a very rare opportunity to learn about people's preferences during a moment where they are actively reconsidering all their options. With the expected increase in cycling infrastructure over the next few years in many Canadian cities, understanding how to reach these potential cyclists is essential.

Funding
BB received a Postdoctoral fellowship from the Canadian Institutes of Health Research and the Fonds de recherche du Québec-Santé (no grant number).

Disclosure statement
None of the authors has competing interests regarding this manuscript

Data availability statement
The data that support the findings of this study are available from the author, MP, upon reasonable request. The complete questionnaire as well as the program code used to produce the tables and figures are available at: https://osf.io/gx7wm/ .