Introduction
Using the bicycle as a primary or complementary mode of transportation is acknowledged to have multiple benefits both in terms of people’s health and wellbeing (Götschi et al.
2016; Kelly et al.
2014) as well as achieving cheaper costs both for the individual and the community and less environmental pollution (de Nazelle et al.
2011; Macmillan et al.
2014). Promoting the use of the bicycle is thus seen as an essential strategy to improve citizens’ quality of life and reduce the adverse effects of the car-centric urban mobility plans and policies, which characterised previous decades (Nieuwenhuijsen and Khreis
2016; Mueller et al.
2018).
Various studies focused on identifying cyclists’ segments to discern differences in the needs and preferences regarding such mode of transport. More generally, in recent years, studies using attitude-based segmentation to promote environmentally sustainable transport have increased (Haustein and Hunecke
2007). An insightful literature review from Haustein and Hunecke (
2007) compares attitudinal, sociodemographic, geographical and behavioural segmentation criteria in multiple studies. Authors concluded that none of the different approaches can claim absolute superiority and that attitudinal approaches show advantages in providing starting-points for interventions to increase cycling use (Haustein and Hunecke
2007). Specifically, clustering cyclists has been proved as an effective strategy to identify meaningful differences in patterns and behaviours (Ahmed et al.
2017; Félix et al.
2017) and develop typologies useful to understand variations and target policies towards the identified groups (Chaloux and El-Geneidy
2019). Félix et al. (
2017) also conducted a literature review that appraised cyclists’ categorisation methods and compared the obtained cyclists’ categories. The authors highlighted some limitations in the current literature: firstly, the studies that have been considered in the review focus on one single city or country, thus not allowing to profile segments in terms of countries and identify meaningful cross-boarder differences. This issue is exacerbated by the fact that studies in different countries use different approaches to segmentation as well (i.e. “Top-Down” approaches such as a priori expert judgment or rule-based classification, versus “Bottom-Up” such as using respondents’ data to perform cluster or factorial analysis). Secondly, when profiling segments of cyclists, there is a need to explore motivators, deterrents, and attitudes further. Heinen et al. (
2011b) indicate that attitudes toward cycling have a relatively strong impact on the choice to commute by bicycle and suggested that psychosocial factors should receive more attention. Commonalities and interplay between sociodemographic, spatial and attitudinal variables should be further investigated according to Haustein and Hunecke (
2007).
Profiling clusters of cyclists has been found to provide deeper insights into their behaviour and characteristics (Ahmed et al.
2017; Kroesen and Handy
2014). Cycling patterns (e.g., motives, frequency) can be considered intrinsic lifestyle choices (Gatersleben and Appleton
2007; Heinen et al.
2011a) and can be influenced by a range of factors including socio-demographics and psychosocial ones. Ahmed et al. (
2017) divided their study population into groups according to ‘cycling frequency’, ‘cycling distance’ and ‘travel planning behaviour’ to increase the knowledge of commuters’ cycling behaviour. Authors profiled the obtained groups specifically to understand how various factors relate to different types of commuter cyclists’ riding decisions. Damant-Sirois et al. (
2014) revealed four distinct cyclist types: "dedicated cyclists", "path-using cyclists", "fairweather utilitarians", and "leisure cyclists". The authors discerned which type of cyclists will likely be affected by certain interventions through assessing respondents’ answers about motivators, deterrents, and infrastructure preferences. Their work suggests that building a network that is tailored to different cyclist types and emphasising its convenience, flexibility, and speed, can be an effective strategy to increase cycling mode share and frequency among the various groups.
Furthermore, marketing studies show that addressing the
average consumer, or in the context of the present study the “average cyclist”, is of little use in terms of research impact. Instead, it is more useful to identify different groups of people which can be separately addressed because they are motivated by different factors and are affected in different ways by policies and interventions (Anable
2005).
Regarding the European context, previous authors argued that it is crucial to develop targeted efforts to increase shares of sustainable mobility, in particular, cycling (Haustein and Nielsen
2016). It requires an understanding of peoples mobility patterns, motivations and barriers as well as an understanding of differences and similarities between countries. Authors showed that accounting for cross borders differences and similarities allows identifying evidence that could be used to change people’s behaviour. Sustainable mobility promotion can find leverages according to similarities or differences in individual countries, with comparable patterns (Haustein and Nielsen
2016). Haustein and Nielsen (
2016) focused on identifying mobility patterns clusters across 28 European countries, pinpointing the existence of two clusters for what regards cycling: Practical cyclists and Green cyclists. However, current knowledge on the matter could still be further improved. The present study will thus follow this line of research, comparing people’s cycling patterns, attitudes, and psychosocial factors within Europe. Comparisons could empower European policy and decision-makers with the knowledge to address the diversity of sustainable mobility challenges. We would expect to find commonalities in cycling patterns related to psychosocial factors across countries.
Based on the above considerations, the present study aimed to address the research gap previously highlighted in the literature, explicitly identifying naturally occurring segments in a population of cyclists in six different European countries (U.K., The Netherlands, Sweden, Hungary, Italy, and Spain) basing on cycling patterns and profiling such groups according to demographics and sociopsychological variables. It is possible to identify previous studies that used cluster analysis to highlight natural occurring subgroups in a population of cyclists and to profile segments according to sociodemographic and psychological factors such as attitudes towards cycling (e.g., Nkurunziza et al.
2012; Haustein
2012; Haustein and Nielsen
2016). Very few of them focused exclusively on cycling behaviours and attitudes in Europe. The present study aims to add to this literature branch though providing insights in comparing riding behaviours and attitudes in different European countries. We selected the six countries based on two criteria: cycling mode share and balanced overview of E.U. geographical areas: Northern Europe (U.K. and Sweden); Western Europe (The Netherlands); Eastern Europe (Hungary); Southern Europe (Italy and Spain). According to aggregated data, the E.U. average cycling mode share is 8% while it substantially differs in each selected country (European Commission
2014). Table
1 shows cycling modal share for each country from the lowest to the highest. The next section will provide a review of the variables that have been studied concerning cycling and have been proved to give valuable insight for developing targeted interventions.
Table 1
Cycling Modal Share of the selected countries
U. K | 3% |
Spain | 3% |
Italy | 6% |
Sweden | 17% |
Hungary | 22% |
The Netherlands | 36% |
Materials and methods
Procedure
An online survey was administered between 27th January and 5th February 2018, to a panel of respondents in six countries (Hungary, Italy, Spain, Sweden, The Netherlands, United Kingdom) who had previously agreed to take part in data collection. The online panel was bought from Qa survey company. Survey companies hold a significant amount of demographic data on their panellists and keep them up to date. It allows targeting surveys at specific groups and obtaining a particular sample, such as a nationally representative sample or a female-only sample. The sample within each country was required to meet the following criteria: All respondents must make at least one cycle trip per month (on average); At least 50% of respondents must be regular cyclists (i.e. make on average > 2 cycle trips per week); At least 30% of the sample must be female; At least 10% of the sample must be aged 50 years or more. The inclusion of participants aged 50 or more years old was particularly relevant to assess whether there were significant differences between young and elderly cyclists. The average completion time was around 30 min. A pilot version of the questionnaire was administered to 60 participants, 30 in The Netherlands, and 30 in the United Kingdom. After examining the pilot questionnaire data, the survey was updated with the new wording of items that produced anomalous replies. Pilot survey data was not included in the final dataset due to changes in the post-pilot questionnaire. Then, the questionnaire's finalised version was translated, uploaded to a custom online survey platform, and administered to participants.
Data and participants
A total of 2389 participants completed the questionnaire. Of these, 1171 (49.1%) were male, 1210 (50.6%) were female and 8 (0.3%) identified themselves as transgender. Given that the sample of transgender participants was too small to be comparable with the other two categories, it was not included in the subsequent analyses, leaving a sample of 2,381 cyclists.
Measures
The questionnaire included the following sections: Cycling frequency, Trip purpose and Modal split between bicycle and car which were used as segmentation variables; Sociodemographic characteristics, Attitudes towards cycling, Infrastructure rating, discomfort while cycling in mixed traffic, Motorists’ behaviour rating and Comparative cycling risk perception which were used as profiling variables.
Cycling frequency has been considered as the product of the number of days cycled per month and months cycled per year. It was measured by the product of 2 items: “How many months a year do you normally cycle?” with answers from 1 to 12, and “In general, during these months, how often do you cycle?” with answers on a 5-point scale (1 = daily, 2 = 3 or more days per week, 3 = 1–2 days per week, 1 to 3 times a month, 5 = less than once a month).
Trip purpose. Respondents were asked why they were making their cycle journeys using the following three dichotomous (Yes/No) items: commuting/travelling to or from work/college, personal business (e.g., shopping/entertainment, health appointments, or visiting family/friends), and leisure/training. Non-commuting trips have been divided into two categories that are leisure and sports trips (such as bike tour, sightseeing bike ride, competitive cycling training, or bike race) and personal business trips (such as using the bike for shopping, visiting relatives or friends, health appointments). Cycling trip purpose has been used, in the present study, to discern differences and similarities among European cyclists and it is particularly relevant since it can help to identify strategies to increase cycling levels between different segments of the population. In the present study, cyclists have been clustered according to the reason why they cycle, explicitly aiming to identify differences between commuting (i.e., travelling to or from work/college) and non-commuting (i.e., personal business and leisure/training) trips.
Modal split between bicycle and car was assessed using two questions about the car and bicycle usage during months of cycling (i.e., “In general, during the months you cycle, how often do you travel in a car, whether as a driver or passenger?”) which were both measured on a 6-point scale (1 = daily, 2 = 3 or more days per week, 3 = 1–2 days per week, 1 to 3 times a month, 5 = less than once a month, 6 = never). To investigate respondents’ use of car or bicycle as the primary means of transport, answers to these items were combined into a new modal split variable coded as 1 = more car than the bicycle, 2 = same, 3 = more bicycle than the car, and 4 = bicycle only. Clustering cyclists concerning modal split brought insights on their habits and helped understand if the increase in cycling demand could grow (Streit et al.
2014). It becomes particularly relevant to the present study because it can allow discerning which group has higher potential to increase the use of sustainable mobility.
Cycling environment was assessed through a single item: “What type of environment do you make the majority of your cycle trips within?” with four response options (i.e., city, town, village, rural).
Sociodemographic characteristics. The sociodemographic form asked for age, gender, country of residence, and job status. As for the latter, participants were asked to select one of the following options: full time employed, part-time employed, self-employed, not currently working, students, homemaker, full-time caregiver and retired. Responses were recategorised into three groups, those who are actively involved in a job (employed), those who are studying and those who have no contractual employment or have retired (unemployed). As advocated by Félix et al. (
2017), the present study aimed at assessing clusters of cyclists at a European level, exploring differences and similarities in terms of sociodemographic characteristics of different type of cyclists in different countries.
Attitudes towards cycling. To assess hedonic attitudes towards cycling, we used three items that expressed positive feelings towards cycling. Instrumental attitudes were measured through four items measuring the functionality and convenience of the bicycle. Finally, four items were used to evaluate the expected benefits of cycling. We added flexibility (Stinson & Bhat,
2004) as an additional benefit to health and environmental benefits. All items were rated on a 5-point scale (1 = completely disagree to 5 = completely agree). Exploratory factor analyses with principal axis factoring yielded a one-factor solution for all three attitudes towards cycling, with no differences across countries. Cronbach’s alpha reliability coefficient was 0.85 for hedonic attitudes, 0.83 for instrumental attitudes, and 0.75 for benefits. Items are listed in Table
2.
Table 2
Items measuring Attitudes Towards Cycling
How far do you agree that you cycle because… | Hedonic Attitudes |
it is pleasant? |
it is mentally relaxing? |
it is physically relaxing? |
Instrumental Attitudes |
it is comfortable? |
of traffic safety? |
of personal security? |
it is time-saving? |
Benefits |
it has benefits for the environment? |
it has benefits for personal health? |
it has economic benefits? |
it is flexible? |
Infrastructure rating. We asked the participants to evaluate on a 5-point scale (1 = very poor to 5 = excellent) the overall level of the cycling infrastructure with two items referring to the quality and provision of the infrastructure (i.e., “How would you rate the cycling infrastructure in terms of the level of provision of cycling infrastructure?”; How would you rate the cycling infrastructure in terms of quality of the infrastructure?”). The two questions were introduced by the sentence “Thinking about the environment you mainly cycle in…”. Cronbach’s alpha reliability coefficient was 0.92. In the present study, respondents’ evaluation of cycling infrastructure in terms of quality and provision has been included to understand whether it would increase the chances of belonging to one cluster compared to the others.
Discomfort while cycling in mixed traffic. Participants were asked to assess their level of discomfort when cycling in the following six scenarios: (1) a path separated from the street, (2) a two-lane (one in each direction) residential commercial shopping street, and no bike lane, (3) the previous scenario with a striped bicycle lane added, (4) a major urban or suburban street with four lanes (2 each direction) and no bike lane, (5) the previous scenario with a striped bike lane added or (6) with a bike lane separated from traffic by parked cars or a kerb. All items were rated on a 5-point scale (1 = very comfortable to 5 = very uncomfortable). Exploratory factor analyses with principal axis factoring and quartimax rotation yielded a two-factor solution. The first factor was composed by items assessing the feeling of discomfort in mixed traffic with no cycling infrastructure (items 2, 4; with a factor loading of 0.91 and 0.74 respectively) while the second factor included items investigating the feeling of discomfort while riding on streets with cycling infrastructure (items 1, 3, 5, 6). Since factor 2 assessed the feeling of discomfort when using cycling infrastructure, providing some protection from motorised traffic, we focused on factor 1 to assess the discomfort of cycling in mixed traffic. Cronbach’s alpha reliability coefficient was 0.81.
Motorists’ behaviour rating. To explore in which way cyclists perceive and evaluate the behaviour of motorists, the participants replied to the following question using a 5-point scale (1 = very poor to 5 = excellent): “How would you rate the driving behaviour of motorists and van/truck drivers within the environment you mainly cycle in?”.
Comparative cycling risk perception. As done in previous studies (Friedman
2011; Martha and Delhomme
2009; Schwarzer
1999), we assessed the comparative perception of risk while cycling by asking participants to estimate their likelihood of being involved in an accident compared to that of their counterparts (i.e. “Compared to other bicycle riders of my age and sex, my risk of being involved in a traffic accident is?). The item was rated on a 5-point scale (1 = much smaller to 5 = much higher).
Statistical analysis
Data were analysed using IBM SPSS version 23. Differences in respondents’ characteristics between countries were examined using analysis of variance (ANOVA) with Welch’s correction and Games-Howell pairwise comparisons for age, and χ2 tests with posthoc z-scores and Bonferroni correction for gender, employment status, bicycle use and evaluation of motorists’ behaviour.
To identify segments of cyclists based on cycling patterns a two-step cluster analysis was performed using both categorical (i.e., cycling trip purpose, and relative use of the car vs bicycle) and continuous (i.e., riding frequency) segmentation variables. Cyclists were clustered according to cycling patterns for two main reasons: first because, following the literature review, it has been proven to be a meaningful way to cluster cyclists and, secondly, to allow comparing cycling behaviours and attitudes in different European countries, addressing research gaps identified by Félix et al. (
2017). To our knowledge, it has never been done with a sample from six different countries. Secondly, using many variables in the segmentation part would have led to overly complex segments, leading to difficulties in discussing results in a meaningful way. We thus decided to use cycling habits (i.e., Cycling frequency, Trip purpose and Modal split) to identify the segments as those variables reflected the actual usage of the bicycle by participants and could allow defining meaningful segments which would later be profiled in terms of sociodemographic and psychosocial characteristics. Two-step cluster analysis first pre-clusters cases into many small sub-clusters, using a sequential clustering algorithm. Nearby sub-clusters are then recursively merged to form the final cluster solution, using an agglomerative hierarchical clustering algorithm. The SPSS two-step clustering component requires only one data pass in the procedure. Automatic selection of the optimal number of clusters was based on the log-likelihood as the distance measure between clusters and Schwarz’s Bayesian information criterion (BIC). The silhouette coefficient, which compares the average within-cluster cohesion with the average between-cluster separation, was examined to assess the goodness-of-fit of the cluster solution. Values between 0.20 and 0.50 indicate a fair fit, and values of 0.50 or more a good fit (Sarstedt and Mooi
2014). For validation and interpretation of the cluster solution, χ
2 tests and ANOVAs were performed on categorical and continuous segmentation variables, respectively. Further between-segment differences were examined with ANOVA for age, attitudes towards cycling, cycling environment, discomfort while cycling in mixed traffic, cycling infrastructure, motorists’ behaviour ratings and cycling risk perception, and with χ
2 tests for the country, gender, employment status and cycling environment.
Variables listed in the previous paragraph were then entered as independent variables in a multinomial logistic regression analysis aimed to profile the segments. Segment membership was set as the outcome variable. The model χ
2, Pearson and deviance tests, and Nagelkerke Pseudo-R
2 were considered to assess the goodness-of-fit of the logistic regression model. A significant model χ
2 indicates that the model with its independent variables fits the data better than a model without those variables; nonsignificant Pearson and deviance tests indicate nonsignificant differences between the observed and the predicted probabilities; and Nagelkerke R2 values larger than 0.20 represent an acceptable approximate amount of explained variability (Hosmer and Lemeshow
2000). Odds ratios and 95% confidence intervals (C.I.s) were reported for each independent variable.
Interpretation of results was based on both statistical significance (
p < 0.05) and measures of effect size: Cramer’s V of 0.10 was considered small, 0.30 medium, and 0.50 large; η2 of 0.01 was considered small, 0.06 medium and 0.14 large; Cohen’s d of 0.20 was considered small, 0.50 medium, and 0.80 large; and O.R.s of 1.5 were considered small, 3.5 medium, and 9 large (Cohen
1988).
Discussion
The present study contributed to describing common patterns, attitudes and psychosocial characteristics of segments of cyclists in Europe, which according to other authors is one of the less investigated aspects if compared to other modes of transport (Handy et al.
2014). It was possible to identify three clusters that substantially differ according to cycling trip purpose, modal split and cycling frequency. We want to emphasise that labelling participants according to a transport mode (i.e., cyclists) is, in most cases, simplistic. A more appropriate approach in psychosocial studies would be to use labels that address human behaviour’s nuances and complexities. This study aimed to address participants’ cycling-related behaviours; thus, we believe that labelling clusters according to the different cycling use patterns would help compare results with other studies and ease the reader.
The first cluster, Leisure-Time cyclists, is composed of people using the bicycle exclusively for sport or recreational activity, mostly in rural areas, with low cycling frequency and a clear preference for using the car for their daily travel. Living in Sweden decrease the probability of belonging to this cluster, especially if compared to the U.K., Spain and Italy. Results suggest that cyclists belonging to this cluster perceive cycling as a pleasant and relaxing activity, and they cycle mainly in rural areas for sport or leisure activities. However, they do not perceive it as a convenient mode of transport nor are interested in the benefits that everyday cycling could bring for themselves and society. Age and income play a role. Like older adults, unemployed people are more likely to be Leisure-time Cyclists.
The second cluster,
Resolute Cyclists, is composed of cyclists that prefer using the bicycle instead of the private car and cycle often for their everyday trips, especially for commuting. Living in The Netherlands, Hungary or Sweden increases the chance of belonging to this cluster, as well as being younger, student or being employed, which increase the probability of belonging to this cluster. Cyclists in this cluster give importance to the several benefits of cycling while they tend not to perceive it as a pleasant and relaxing activity.
Resolute Cyclists are well aware of the benefits of cycling as a healthy and flexible means of transport. Older cyclists are less likely to be Resolute Cyclists; a possible explanation is that they may have health-related issues or reduced mobility. This result is only partially in line with results from Dill and Voros (
2007) which showed that younger adults were more likely to be regular cyclists and that the significant drop-off in regular cycling occurred at age 55 years and above. In addition, recent studies have shown that the younger generation (Delbosc and Currie
2013; Kuhnimhof et al.
2012) are more willing to use other modes of transport than the car. As stated by Prillwitz and Bar (
2011), younger generations are often defined as “green travellers”, suggesting a mode preference for active means of transport, possibly dictated by income constraints or delays in adult life transition (Delbosc and Currie
2014), shifts in attitudes (Vij et al.
2017), differences in their daily activities or even by changes in the local transport or land-use systems (Delbosc et al.
2019). Our analysis highlighted that respondents with higher instrumental and benefit-oriented attitudes were more likely to be members of the
Resolute-cyclists segment. This is in line with results from Dill and Voros (
2007), which found that respondents with positive attitudes towards cycling were more likely to be regular cyclists.
Higher ratings of discomfort while cycling in mixed traffic decreased the odds of belonging to the
Resolute Cyclists cluster. These cyclists are in fact, more likely to cycle in city or town environments where mixed traffic situations are frequent. Sharing the road with motorised vehicles can influence cycling patterns and bicycle usage exhibiting avoidance behaviours. While it would be hasty to infer a causal relationship, in this case, it is still important to acknowledge that to increase cycling levels and prompt citizens to increase bicycle use in urban areas it is essential to address negative feelings arising in the interaction with motorised vehicles’ drivers. A study from Chataway et al. (
2014) showed that cyclists in an emerging cycling city tend to be more concerned regarding the interaction with motorised vehicles than those in an established cycling city. Our findings are partially in line with our expectations. Even if, contrarily to our expectations, cyclists from the U.K. tend to report higher evaluations of motorists’ behaviuour, results show that cyclists from Italy and Spain (which can be considered emerging cycling countries) tend to report lower evaluations of motorists behaviour. This is particularly relevant, as it has been argued that to cope with negative feelings arising from the exposure to motorists’ behaviour, cyclists tend to avoid cycling in mixed traffic conditions, and sometimes avoid cycling in general (Prati et al.
2020; Puchades et al.
2018; O’Connor and Brown
2010). This implies that, in low cycling culture countries, promoting cycling becomes particularly difficult if motorised vehicles behaviour is not addressed. In countries such as Italy and Spain, national effort should be targeted at improving interactions between traffic participants. Future studies should address this relationship aiming to investigate to what extent behaviours of traffic participants could actually influence sustainable mobility choices.
The third cluster,
Convenience Cyclists, is composed of cyclists that use the bicycle for personal business or recreational activities, have moderate riding frequency and choose the bike in the daily trips similarly to cars. Convenience cyclists mainly cycle in urban areas (city and town). The results suggest that participants in this cluster are more likely to live in a strong cycling culture country (i.e., The Netherlands, Hungary) and give importance to the several benefits that cycling holds, in particular environmental-related benefits. At the same time, they tend not to perceive cycling as a pleasant and relaxing activity. Similarly, Dill and Voros (
2007) found a significant relationship between environmental values and utilitarian cycling. Respondents who thought air quality was a problem try to limit their driving to help improve air quality, and those who believed that the region did not need to build more highways were more likely to be utilitarian cyclists. Discomfort while cycling in mixed traffic seems not to influence belonging to the
Convenience Cyclists cluster. This is an interesting result and suggests that
Convenience Cyclists tend to cycle because of necessity or simply out of convenience, regardless feelings of fear or discomfort towards sharing the road with motorised vehicles. This adds to results mentioned in the previous paragraph, stressing that, in urban areas, is also essential to make cycling more advantageous in comparison to other motorised mode of transport. Infrastructural solutions addressing this issue are, for instance, green waves, preferential crossings or “all green” crossings for cyclists at busy intersections.
In line with our findings, an aspect worth discussing is the country of residence and its related societal and infrastructural aspects. According to the Eurobarometer 422 survey (European Commission
2014), the countries studied in the present work differ significantly in terms of the use of the bicycle as the primary means of transport. Present results suggest that in countries where cycling is a well-established means of transport (i.e. The Netherlands, Sweden and Hungary), there is a lower probability of people cycling for leisure as well as a higher chance of belonging to the category of
Resolute Cyclists than in countries where the percentage of people who reported cycling on a daily basis is low (i.e. Italy, Spain and the United Kingdom). This could reflect national cycling policies and mobility culture. There are studies in literature exploring the role played by national policies and investments in making cycling safe and popular (Hull and O’Holleran
2014; Kosztin et al.
2017) as well as by the level of infrastructure provided to make cycling attractive (Pucher and Buehler
2008). For example, The Netherlands has been at the forefront of policies to make cycling safe and attractive and, as such, can be considered an exemplary “strong cycling culture” (Hull and O’Holleran
2014). The same reasoning can be applied to Sweden and Hungary, where in recent years efforts have been made to increase the share of cycling in its modal split (Bastian and Börjesson
2018; Koglin
2017; Haustein and Nielsen
2016), a strategy which in turn has been seen to encourage the adoption of this means of transport (Pucher and Buehler
2008).
When discussing infrastructural and societal aspects, it is valuable to mention the concept of human infrastructure as defined by Lugo (
2013). Infrastructures are essential to promote bicycle use but also identities and behaviour of traffic participants. It is thus vital to consider the close interplay between the built and social environment. Lugo (
2013) argued that social networks, identities and cultural practices could act as human infrastructure, upholding bicycle use growth. For instance, bicyclists’ preferred routes are influenced not just by infrastructure, but also by attitudes and participation in particular social networks. Human infrastructure can influence cycling behaviour, both encouraging or discouraging it. According to Lugo (
2013) exchange of specialised pieces of information (e.g. when cyclists give each other suggestions on preferred routes based on personal knowledge and not only on municipal maps) or the enactment of an expectation (e.g. aggressive behaviour of motorists) can constitute human infrastructure. Our results somewhat suggest that human infrastructure and cycling patterns are related. For instance, the cluster with higher cycling levels,
Resolute Cyclists, also reports the highest ratings of motorists’ behaviour and higher ratings of provision and infrastructure quality. The study's design does not allow making strong causal inferences, but it is reasonable to expect that people who experience aggressive behaviours from motorists could be discouraged to cycle and ultimately use the bicycle in fewer occasions.
On the notion of human infrastructure, it is relevant to cite a study by Nello-Deakin and Nikolaeva (
2020), which identified seven main factors encouraging foreigner newcomers to cycle in Amsterdam. Authors highlighted the critical role of cyclists themselves in encouraging others to cycle. Specifically, echoing a study by Larsen (
2017), authors suggested that it is useful to consider cyclists as a form of human infrastructure that plays a crucial role in reproducing the city’s cycling culture and fostering cycling uptake by other newcomers. Cycling practices are partly influenced by environmental factors and mostly by interactions between different time and space factors and how these are interlinked into people’s lives (Nello-Deakin and Nikolaeva
2020). Cycling cultures can thus be considered intricate systems dependent on a series of causal feedback loops related to social, political, and spatial processes (Macmillan and Woodcock
2017). Simultaneously, traffic participants’ human infrastructure also plays an important role in influencing policymakers on the topic of urban mobility, thereby contributing to shaping streets, laws and behaviour (Nello-Deakin and Nikolaeva
2020). As suggested by other authors (e.g. Lugo
2013), we argue that bringing together different cyclists is of utmost importance for cycling advocacy and planning. The present study contributes to highlighting which type of cyclists can be expected in European countries and can guide practitioners and policymakers in involving even bicycle users with lower cycling levels or who could not self-identify as cyclists (e.g.
Leisure-time Cyclists or
Convenience Cyclists). Their use of street space should also be considered part of the human infrastructure, thus influencing bicycle culture and ultimately use. This is equally useful for countries with strong (e.g. the Netherlands) and emerging cycling cultures (e.g. U.K). Emerging cycling culture countries should leverage on human infrastructure and networks of
Resolute Cyclists to ultimately increase population cycling levels. In the study by Nello-Deakin and Nikolaeva (
2020), it has been shown that people who already decided to adopt cycling plays a crucial role in fostering cycling uptake in others.
Results regarding cycling frequency imply possible consideration about participant’ level of physical activity. Active transport contributes to alleviating the adverse health effects of inactivity and cycling can help reach the WHO’s physical activity recommendation, bringing beneficial effects for people’s health (Raser et al.
2018). While acquiring precise data on time spent cycling and travelled distance was out of the scope of the present study, an estimation of the days spent cycling per year can provide a rough picture of the amount of physical activity carried out.
Resolute Cyclists are the ones achieving the higher level of physical activity with, on average, 173 days spent cycling per year and a preference for using the bicycle instead of the car. Several studies revealed that reducing car usage for the active mode of transport provides benefits to people’s health as it counteracts sedentary lifestyles (e. g. De Geus et al.
2008; Basset et al.
2008; Celis-Morales et al.
2017;). Thus, efforts should be focused on
Leisure-Time Cyclists, who prefer using the car instead of bicycle and cycle only 56 days per year on average.
Convenience Cyclists holds the potential for switching towards daily active mobility since they have no clear preference between the use of the car or the bicycle and cycle around 107 days per year.
Our results suggest that each type of cyclists could be differently responsive to campaign leveraging on different messages. Campaign and intervention targeting
Leisure-Time Cyclists could leverage on the hedonic attitudes held by such a group and prompt them to use the bicycle for purposes other than recreation or sport, for instance, messages stressing the pleasure of cycling not only as a leisure activity but as an everyday means of transport. It would be advisable to change their bicycle patterns fostering some kind of halo effect (Nisbett and Wilson
1977) since they already report high scores on hedonic attitudes.
Leisure-Time Cyclists could be influenced by messages stressing the contextual opportunities and the positive consequences of the cycling mode. Care must be taken though when choosing environmental concerns as core messages in campaign and interventions as other findings claim that information about the adverse environmental effects of the car raises public awareness but is usually insufficient to change behaviour (Tertoolen et al.
1998). In any case, it is essential to assume bidirectional effects between cycling behaviour and its determinants, such as attitudes, preferences, and habits (Handy et al.
2014). Travel behaviour studies established that bidirectional effects exist between attitudes and behaviour (Dobson et al.
1978; Golob et al.
1979; Tardiff
1977).
Moreover, results of the present study can be useful to hone the European Cycling Strategy (ECF,
2017) and provide insights for a future version of the document, specifically regarding policy and behavioural change. Market segmentation has, in fact, proven to be beneficial for tailoring communication and intervention for a specific purpose which is in this case, making cycling safer and more attractive for European Citizens. It is as well useful for directing funds to targeted objectives, maximising the expected impact of interventions and making expenses more efficient. Anable (
2005) remarked that the clustering approach illustrates that policy interventions need to be responsive to the different motivations and constraints of the sub-groups. However, such responses may be less about ‘harder’ infrastructural changes and more about ‘softer’ interventions with an emphasis on management and marketing activities. As it has been argued by Handy et al. (
2014) policymakers can benefit from guidance on which of the possible strategies are likely to increase cycling and to what degree. Further studies can help to identify critical factors associated with cyclist preferences, attitudes and behaviours, pointing to a potentially effective strategy.
The current study has some limitations which should be recognised. First, its cross-sectional nature limits our possibility to make strong causal inferences about present results. In addition, social desirability and recall bias may have affected the results. The study population is self-selective (i.e. online panel) and, therefore, the generalizability of the findings is limited. Central tendency bias, which inclines participants to avoid the endpoints of a response scale and to prefer responses closer to the midpoint (Stevens
1971), might have occurred due to the usage of five-point Likert type scales, which is the kind of scale particularly giving rise to such bias (Douven
2018). Furthermore, respondents' perceptions (e.g. regarding risks, evaluation of motorists’ behaviour and infrastructure) are most likely not consistent across countries. It is reasonable to assume that assessments of safety and other factors are strictly dependent on each country’s dominant culture and specificities. However, to allow comparisons between clusters and countries, responses are treated equally. Although the present research considers the environment in which participants mainly cycle in, there are many relevant variables related to the built environment that have not been investigated here (e.g., land use mix, population density, where the participants reside). Finally, the present study focuses on cyclists only and lacks insight into other modes of transport (e.g. public transport) and non-cyclists. Regarding future development of the present study, it is worth considering that comparing current results with data from non-cyclists would allow developing better guidance for policymakers. Future research should aim at filling these gaps.
The present study identified clusters of cyclists based on cycling motivation and patterns at a European level and profiled the membership to each cluster according to socio-demographics and psychosocial variables. The value of the present study is to highlight commonalities in patterns, characteristics and attitudes of cyclists in six different European countries and it could be useful for policymakers, urban planners and transport experts for developing targeted and tailored measures to increase cycling levels. Our study supports that cycling patterns and habits are linked to culture as well as attitudes and evaluation of the cycling environment, highlighting the importance of the feeling of discomfort in mixed traffic and the area in which people mainly cycle in.