Introduction

Inhabitants of rural areas tend to be highly dependent on car use for their basic travel needs. Due to the longer distances between travel destinations the use of active modes of transportation, like walking and cycling, is often not a viable option. Furthermore, public transport in low-density rural areas tends to be very limited or unavailable. Electrically-assisted bicycles or e-bikes may provide a useful addition to available transport modes in rural areas by combining the assets of active modes (e.g. health benefits, flexibility, enjoyment) with those of motorized modes (e.g. range, speed, comfort). As such e-bikes may support more sustainable rural mobility in terms of increased accessibility, health, and environmental benefits.

Despite these potential benefits however, e-bike use in rural areas has received limited attention. As yet, little is known about how rural residents can be supported and encouraged to using the e-bike. Therefore, the purpose of this study is to explore the determinants of current and potential e-bike use in rural areas specifically. The role of sociodemographic characteristics, mobility situation and attitudes are considered. The paper aims to contribute to a better understanding of e-bike mobility in rural areas and to inform policies to address rural mobility issues. In pursuit of this objective we first discuss the literature regarding rural mobility and e-bike use in general, and e-bike use in rural contexts specifically. We then present empirical data on current and potential e-bike use from a large-scale mobility survey among rural residents in the northern parts of the Netherlands.

Rural mobility

Contrary to highly mobile and connected urban areas, rural areas in modern western societies have long been thought of as places of tradition, fixity and stability (Bell and Osti 2010). However, it has become clear that mobility is an equally important aspect of rural life, necessary for accessing basic services, facilities and activity locations (Milbourne and Kitchen 2014; Osti 2010). Car travel offers benefits for rural residents in terms of speed, flexibility and range (Gray et al. 2006). In the U.K., people living in remote rural areas consider a car essential to live in their area (Milbourne and Kitchen 2014). Availability of road infrastructure and private car ownership are important determinants of car use in rural areas. In the U.S. for instance, high levels of rural mobility are almost entirely made possible by the extensive road network in rural areas and sheer universal car ownership (Pucher and Renne 2005). Accounts from the Netherlands (Harms 2008) and France (Orfeuil 2010) paint similar pictures of heavy reliance on car use by populations in rural areas.

Although car ownership has improved mobility and accessibility for many rural residents, the consequences are not uniquely positive. A too high dependence on car transport can be problematic for specific social groups. Among these are individuals or households unable to afford a car (Smith et al. 2012), unable to operate a car (e.g. physically impaired individuals, younger populations or elderly people) (Shergold and Parkhurst 2012), or unable to rely on their social networks for lifts. Lucas (2012) describes how transport disadvantage (i.e. not having access to a car, but also high cost of fares, or lack of information) can interact with social disadvantage (i.e. no job, no income) to cause transport poverty, which can lead to a situation of inaccessibility (of goods, services, social networks), ultimately causing transport-related social exclusion. This has attracted considerable attention in the literature in recent decades (Cass et al. 2005; Lucas 2012; Preston and Rajé 2007). Besides affordability, ability and access, increased attention is also paid to negative impacts of high car mobility such as congestion, accidents and environmental damage (Preston and Rajé 2007).

Car mobility is but one aspect of access in rural areas (Gray et al. 2006; Lyons 2003). Public transport can form a high quality and sustainable alternative. Public transport in rural areas is however challenged by the low numbers of passengers and longer distances between stops which lead to cost-inefficient operation of public transport. Resulting cutbacks can decrease demand, which leads to further cutbacks (De Jong et al. 2011). Public transport authorities have different options to deal with such challenges: they can become more actively involved in spatial policy forming in rural areas in order to anticipate and act on community developments and demographic and land-use changes; they may work with local parties to enable more efficient use of empty seats in their vehicles; and they may adopt technology and innovation to improve communications and dispatching, and expand services to match an as broad range of user needs as possible (Rosenbloom 2003).

Active modes like walking and cycling form flexible, sustainable and healthy alternatives to car mobility. These are generally sought-after solutions in the mobility domain due to their low environmental impacts and benefits on physical health and subjective wellbeing (Van Wee and Ettema 2016). It is well established that active transport levels in urban areas can be influenced by the built environment through land use, density and availability of non-motorized linkages (e.g. Cervero, 2002), but knowledge on determinants of active mobility in rural areas is limited (Frank et al. 2006; Saelens et al. 2003). However, the lower densities and longer distances that force high reliance on private car use seem to negatively impact the use of active modes. In the U.S., walking and cycling is less common in rural areas than in urban areas (Pucher and Renne 2005). This is similar to the Netherlands, where the majority of trips in urban areas are made by active modes, but walking and cycling levels are lower in the least urbanized areas (Harms 2008; Harms et al. 2014; KiM 2015). E-bikes can provide a “third” alternative to car use in rural areas that may compensate some of the disadvantages of public transport and active modes.

Determinants of e-bike use

Several studies have addressed the role of positive and negative experiences in e-bike use (Fishman and Cherry 2015; Jones et al. 2016). Based on these studies, Simsekoglu and Klöckner (2019) examine the factors predicting intention to buy an e-bike. They find that perceived benefits for health, mobility and personal image, social and descriptive norms, and familiarity with e-bikes are significantly and positively related to the intention to buy an e-bike.

More comprehensive models that attempt to explain actual use are sparse in e-bike literature, but are found in literature on regular bicycle use. Handy et al. (2010) for instance employ the ecological model of physical activity research to conceptualize factors that influence bicycling. Three sets of factors are hypothesized to directly affect bicycling behavior: individual factors, which are socio-demographic characteristics, health, cycling preferences and self-efficacy; social-environmental factors, which relate to a cycling-supportive social environment (e.g. friends and family that cycle); and physical environmental factors, or whether environment, available infrastructure and travel distances are supportive to bicycle use. The same factors are considered in a model proposed by Fernandez-Herredia et al. (2014) who emphasize on attitudes, which are in turn influenced by socio-demographic characteristics and contextual conditions, such as policies and the extent to which there is a cycling culture in place.

In this paper, we study current and potential e-bike use by considering the role of sociodemographic characteristics, mobility situation and attitudinal variables. First, we consider sociodemographic characteristics. Studies from the Netherlands, Austria and the U.S. find that its users are older than the general population, and more often female (Fishman and Cherry 2015). In the Netherlands, e-bike adoption was initially high among physically impaired and elderly people with difficulty using a bicycle (Peine et al. 2016). The relations between e-bike use and levels of income and education are ambiguous: a U.S. study showed that e-bike users have higher incomes and education (Popovich et al. 2014), but an Austrian survey found the opposite (Wolf and Seebauer 2014). A study on e-bike users in the Netherlands found that income levels of e-bike users reflect those of the general population (Lee et al. 2015). This study also found that e-bike use is higher among households without children.

Second, we consider the mobility situation. As described in the preceding paragraph, rural mobility can distinctly differ from mobility in more urbanized areas. This might impact the uptake and use of e-bikes. The level and type of mode substitution brought about by the uptake of e-bike use is an important subject of research in the field. Thus, it is important to study how ownership and use of transport modes and travel distances affect potential or current e-bike use. So far, findings on e-bike mode substitution vary per context. Studies in Australia and the U.S. have suggested that the e-bike can replace car trips (Johnson and Rose 2013; MacArthur et al. 2014) whereas in the Netherlands, a country with high bicycle use, e-biking also substitutes cycling (Kroesen 2017). This suggests that e-bikes are more likely to substitute the dominant modes in a given mobility system (Sun et al. 2020). Apart from geographical variation, substitution also seems to vary over time, with newer studies reporting greater displacement of driving and walking and lower displacement of conventional bicycle trips (Bigazzi and Wong 2020). In terms of travel behavior, e-bike users generally cover longer distances than bicycle users (Cairns et al. 2017). In the Netherlands, around 91% of all bicycle trips are less than 7.5 km long. In the case of e-bikes, 91% of all trips made are less than 15 km long. This difference in distance covered seems particularly strong for e-bike commutes (up to twice as long as bicycle commutes) and recreational e-bike use (up to three times longer than recreational bike use) (KiM 2015).

Third, we consider the role of attitudes in current and potential e-bike use. Attitudes on health, safety, reliability, ease of use, speed, comfort, fun, image, environmental friendliness and suitability for daily use can have an important influence over actual behavior. The health effects of e-bike use have been subject of various studies (Bourne et al. 2018; Castro et al. 2019; Sundfør et al. 2020). These can be an important motivator to use an e-bike, especially when substituting sedentary transportation (Plazier et al. 2017b). In terms of safety, e-bikers seem subject to higher risk (Fishman and Cherry 2015). For instance, a study in the Netherlands found that they are twice as likely to suffer injury from a crash than bicycle users (Poos et al. 2017). However, this may be partly a result of the large share of older, more vulnerable e-bike users. Perception of e-bike safety can influence the decision to use an e-bike. In rural areas, both safety and comfort might be related to the cycling infrastructure. Dedicated bicycle infrastructure is important to encourage bicycle use (Pucher and Buehler, 2008). This is especially true for rural areas where vehicle speeds are higher and heavy goods traffic is more important (Laird et al., 2013). The rural context might also be of influence on attitudes toward reliability, ease of use and suitability for daily use of the e-bike. An e-bike can provide flexibility and independency (Jones et al. 2016), for example from bus and train schedules, which might be important in areas with lower- frequency transit provision. Higher average cycling speed might in turn favor the e-bike use over regular cycling in areas where distances between destinations are longer. In terms of image, the e-bike was long associated with the stigma of being a bike for “older people”, although this seems to be changing rapidly (Jones et al. 2016; Peine et al. 2016). In this study we attempt to uncover whether e-bike image plays a role in rural e-bike adoption and use. The positive experience and enjoyment associated with e-cycling has proven to be an important motivator for its use (Plazier et al. 2017b). Finally, concerning the environmental impact, the net effect of e-bikes on the environment is dependent on the mode they replace. The impact is negative if they replace non-motorized active modes. However, e-bikes are much more energy-efficient than cars, and the trend towards use of lithium ion batteries over lead acid batteries further improves environmental efficiency of e-bikes (Fishman and Cherry 2015). Thus, they can help mitigate expected growth of levels of CO2 and energy use for transportation (Mason et al. 2015). However, concerns have also been expressed regarding the sustainability of moving emissions from tailpipe to power plant (Sandy Thomas 2012). Therefore, in order to be truly sustainable, e-bike charging infrastructure should be powered by renewable energy sources.

Current evidence on e-bike use in rural areas

E-bikes permit covering longer distances at high and constant average speeds against reduced physical effort. As such, they have also attracted the interest of policy makers, planners and scholars in various countries, for their potential to reduce congestion and environmental damage, and improve physical health and mental wellbeing (Cairns et al. 2017; Plazier et al. 2017a) have previously suggested that e-bike use might hold specific potential in peri-urban and rural areas. However, the evidence from these contexts is to date limited.

Some Chinese studies consider the urban-rural differences in e-bike use, but it should be noted that these studies mostly concern scooter-style electric bicycles. An analysis of the future of electric mobility in China, concluded that “electric two-wheelers [….] can be an energy-efficient, low-emission form of private transport in small and medium size cities where public transit service is limited or the city is geographically disperse” (Weinert et al. 2008, p. 2554). Compared to more urbanized areas, e-bike ownership in Chinese households is higher in rural environments with low density, poorly connected street networks, lower destination accessibility, lower transit accessibility and lower land use diversity (Zhang et al. 2013). Furthermore, e-bike ownership was inversely related to car, bicycle and motorcycle ownership, the availability of bus stops in the nearby environment. Finally, in the Chinese context, no notable differences have been found in safety behavior between urban and rural environments (Yang et al. 2014).

A study in the western context of Sweden specifically focused on urban – rural differences in e-bike use (Winslott Hiselius and Svensson 2017). In both areas, car travel was the most substituted mode, but trip distances replaced by the e-bike were higher in rural than in urban areas. The authors stress that specific attention should be paid to support e-bike use in rural areas where they can play a role in better accessing facilities and shopping centers, but accumulation of knowledge is needed in order to better understand the specific determinants of e-bike use in rural areas.

Method

Study location

Data were collected in the North of the Netherlands. Compared to other European countries, the Netherlands is densely populated and highly urbanized. The North of the Netherlands is the most rural part of the country, both in terms of population density and perception of the Dutch population (Haartsen et al. 2003). The study area is shown in Fig. 1. The area mostly consists of grass- and farmland, and has a flat topography. Like the rest of the Netherlands, it has a temperate oceanic climate influenced by the North Sea, with average temperatures in the coldest months above zero, but regular frost periods. Periods of extended rainfall are common.

Fig. 1
figure 1

Study area, population and public transport connections

Full size image

In the Netherlands, urbanization has been categorized in five classes (Statistics and Netherlands 1992). Based on this classification, 49.5% of the population in the study area lives in areas with very low urbanization rates (< 500 addresses per km2), 39.3% in areas with low urbanization rates (500-1,000 addresses per km2), and 11.2% in areas with moderate urbanization rates (1,000–1,500 addresses per km2). No areas of high or very high urbanization (> 1,500 addresses) are present here. The East of the study area is moderately urbanized, with two towns with populations over 10,000. Overall, the region is subject to population decline (Statistics Netherlands 2022).

Towns in the area are well-connected by roads. Over the past decades, the quality of public transportation in the area has deteriorated due to budget cuts. However, public transportation authorities have been fairly successful in finding new ways to finance and operate transit. As a result, transit ridership has been rising in recent years (OV-bureau Groningen and Drenthe 2016). The available public transport in the area now consists of two train lines and a series of bus lines. Most of the bus lines are full-sized, scheduled buses run by a public transport authority, but some lines are operated by volunteer workers. Both have fixed routes and schedules, running regularly during the day. One line in the area operates as an on-call bus service, with a fixed route but flexible schedule. Additionally, the “RegioTaxi” service is flexible in route and schedule (De Jong et al. 2011, p. 69).

Cycling volumes in Dutch rural areas are lower than in more urbanized parts of the country and have been declining for over a decade (Harms et al. 2014). Yet, a recent bicycle survey by the Province of Groningen revealed that 95% of the inhabitants of the province cycles at least once a week (Provincie Groningen 2016). Bicycle infrastructure is generally available between larger towns in the area. On roads with lower traffic intensity, cyclists share the road with other vehicles, but on roads with higher traffic intensity separate infrastructure is often available, running parallel to the roadway. Despite this availability of infrastructure, respondents in the survey of the Province of Groningen raised concern about the unsafety of narrow bicycle paths, sharing the road with motorized traffic, complex traffic situations at intersections, and the quality of route signing.

Survey

The present survey was part of an effort of the Province of Groningen and University of Groningen to address the mobility situation of populations in rural areas. Therefore, it was designed and advertised as a mobility survey. An advantage of this is that a “mobility survey” is likely to interest a broader range of respondents than an “e-bike survey”, which is probable to engage a disproportionate amount of e-bike owners and/or enthusiasts, thereby introducing bias in the results (Groves et al. 2004). The population targeted consisted of all inhabitants in the study area aged 12 and over. As the same laws and regulations apply to e-bike use as to regular bicycle use, children are legally allowed to ride e-bikes. The aim of the survey was to include an a as wide diversity of people as possible, ranging from students in secondary education to the elderly.

The survey consisted of four parts. The first part covered sociodemographic characteristics, including age, gender, six-digit zip code, occupation, level of education, household composition, physical impairment, income level, and whether income was sufficient to cover household expenses. The second part consisted of questions about the current mobility situation of the respondent, including license ownership, vehicle ownership, and modes of transport used for commuting or going to school, grocery shopping, visiting social contacts and attending sports activities (if applicable). For each of these activities, respondents could indicate the distance traveled to these destinations (what is the distance (one way) to the location of your work or school?), the mode they used the most to attend this activity (what mode of transport did you most use last month to go to work or to school?), and an alternative mode of transport (what alternative mode of transport do you use to go to work or to school?). They were also asked to motivate their mode choice by picking two out of a list of ten mode choice factors (safety, reliability, speed, ease of use, comfort, fun, cost, health, sustainability and image) (What are the two most important reason for using this mode of transport?). The third part of the survey focused specifically on e-bikes, including questions on current and potential use of e-bikes as the study’s two main independent variables. Current e-bike use was measured with the question “Do you currently use an e-bike?” (0 = no, 1 = yes). Respondents who answered “no” to this question were asked about their potential e-bike use with the question “Would you like to use an e-bike?” (0 = no, 1 = yes). Other questions in this part of the survey asked whether respondents had heard of e-bikes (are you familiar with e-bikes?) had tried an e-bike (have you ever tried riding an e-bike?) or knew people with an e-bike (do you know people who own an e-bike?) (all answer options “yes” or “no”). Finally, in the fourth part of the survey, respondents were asked to rate car, public transport, bicycle and e-bike on a 10-point scale ranging from 1 (very bad) to 10 (excellent) for safety, reliability, ease of use, speed, comfort, fun, health, environmental friendliness, image, and suitability for daily use (“How would you rate the following modes of transport on the different aspects mentioned?”).

Throughout the survey respondents were allowed to skip questions if they felt uncomfortable answering, could not accurately recall, or did not know the answer. The survey also covered some questions that are not relevant to the current study and will not be discussed further.

Respondents

A total of 1135 persons (444 men, mean age 48 years) filled in a survey online (1,069 surveys) or returned a survey by mail (66 surveys). Because the study did not use a forced response design, there were many respondents with missing data on key variables. After removal of these respondents, a final dataset of 612 respondents (252 men, mean age 49 years) with complete data on current e-bike use was included in the analyses. Of the 410 non-users in this final dataset, 372 had complete data on potential e-bike use (Fig. 2). Table 1 gives an overview of the characteristics of the sample used in the analysis (n = 612), the initial sample (n = 1,135), and the total population of the study area.

Fig. 2
figure 2

Response overview

Full size image
Table 1 Sociodemographic characteristics of the population in the study area, and the initial and final samples
Full size table

Respondents in the younger age groups (12–24 years) and older age groups (65+) were somewhat underrepresented in the final sample compared to the population in the study area, while women and highly educated respondents were somewhat overrepresented. A relatively large amount of the respondents reported low- or no income, which can be related to the fact that both younger (age > 12) and older people were included in the sample.

The majority of the respondents (88.9%) in the final sample lived in areas with (very) low degrees of urbanization, which matches well with the population statistics for this part of the Netherlands. Compared to national data, e-bike use among respondents in the final sample was fairly high. It is estimated that approximately 1.6 million e-bikes were in use in the Netherlands in 2016 (Stichting BOVAG-RAI Mobiliteit, 2016), which amounts to 10.7% of the total population of 14.9 million people of 12 years and older (Statistics Netherlands 2022). Of the 612 respondents included in the analysis, 33% (n = 202) owned an e-bike. Of the remaining 372 non-users who had complete data on all variables included in the analysis, 62% (n = 232) indicated they would be willing to do so.

The final sample was highly comparable to the initial sample on most of the socio-demographic characteristics, indicating that characteristics of respondents with missing data did not differ considerably from those with complete data. Response in the final sample was somewhat lower among 12-24-year olds, and somewhat higher among 45–64 and > 65-year olds. Yearly income was somewhat higher in the final sample than in the initial sample, as was the share of respondents who had completed higher education. Household composition, degree of urbanization and household car ownership did not differ much between the samples. E-bike use was slightly higher in the final sample, but the share of non-users willing to use an e-bike was approximately the same.

Procedure

Data collection took place in March and April of 2017. All inhabitants of the study area age 12 and older were eligible for participation. Potential participants were informed about the research topic, confidentiality and anonymity, and the right to withdraw from the study at any time during the data collection. Respondents were invited by means of announcements in local newspapers, on web sites, and via social media channels of participating municipalities, the Province of Groningen and the University of Groningen. Also, Facebook advertisements were used to target potential respondents. Groups of students were sent out to major towns in each of the participating municipalities in the study area to invite potential participants by distributing flyers. Participants could choose between a digital or paper version of the survey. As an incentive, all participants were offered a chance to win one of five gift cards of €100, -

Data analysis

Data were analyzed using SPSS version 24 for Windows (SPSS Inc., Chicago, Illinois). We used binary logistic regression analyses to determine the degree to which sociodemographic characteristics, mobility factors and attitudes toward the e-bike predict the odds of current and future e-bike use. Responses from persons in the same household were treated as independent responses. The predictor variables were added to each regression model in three blocks in a stepwise manner

The first block contains the sociodemographic characteristics including age (continuous), gender (0 = male, 1 = female), education level (0 = no higher education; 1 = higher education), household composition (0 = no children, 1 = children), physical impairment (0 = no, 1 = yes), degree of urbanization of the six position postal code area (derived from Statistics Netherlands (2022), dummy coded as 0 = very low, 1 = low to moderate), and occupation (dummy coded in four groups of students [of secondary schools or higher education], commuters [doing paid work in permanent or freelance jobs or working in charity], retirees, and ‘others’ [mainly stay-at-home or unemployed individuals], with others as the reference category). Instead of annual income level, we included ‘capability to afford an e-bike” (0 = no, 1 = yes) as a predictor, derived from the question whether the income is sufficient to cover household expenses.

The second block contains variables related to respondents’ current mobility situation: car drivers’ license (0 = no, 1 = yes), household bicycle ownership (number of bicycles), household car ownership (number of cars), average distances traveled for varying purposes (less than 15 km or 15 km and more, 15 km being the maximum distance deemed acceptable to travel by e-bike (KiM 2015, p23), mainly car user (0 = no, 1 = yes), mainly bike user (0 = no, 1 = yes), and mainly public transport user (0 = no, 1 = yes). The latter three variables were dummy coded based on the respondents’ most-used mode to commute, do grocery shopping and visit social contacts (if applicable).

The third block consists of ten variables measuring attitude towards the e-bike on a 10-point scale: safety, reliability, ease of use, speed, comfort, fun, health, environmental friendliness, image, and suitability for daily use.

Results

Sample characteristics

Table 2 shows the distribution of the predictor variables included in the models of current and potential e-bike us. There are several differences between e-bike users and non-users. E-bike users are on average older than non-users, and more often female. Less of them have completed higher education, but more e-bike users than non-users state making enough money to cover household expenses. Furthermore, e-bike users more often are part of a household without children, are more often physically impaired, and the proportion of retirees is higher than among non-users. While e-bike users more often have a driver’s license than non-users, they do not more often own a car, and they less often report using the car as main means of transportation. E-bike users also have less bicycles in their household than non-users, and they less often report using a bicycle as means of transportation. The proportion of people traveling an average less than 15 km to daily destinations is higher among e-bike users than among non-users. This suggests that for people with lower distances to activity locations, the e-bike might be a viable alternative to car use. However, it does not provide insight in the distances for which e-bike use competes with bicycle use. In general, e-bikers seem less committed to car or bicycle use than non-e-bikers. Finally, they are also significantly more positive on all aspects of e-bike use compared to non-users.

Among the non-e-bike users (n = 374), there are no significant differences in terms of age, gender, physical impairment or rate of urbanization between those who would like to use an e-bike, and those who would not. However, potential users have less often completed higher education, are more often from a household with children, and less often make enough money to cover their expenses. In terms of occupation, potential users are more often students or commuters. In terms of mobility, potential users are committed more to car use, and less to bicycle use. Finally, potential users respond significantly more positively to all attitudinal aspects than the group that expressed no interest.

Table 2 Descriptive statistics of the independent variables included in the model predicting current and potential e-bike use
Full size table

E-bike use

Table 3 presents the results of the logistic regression analysis estimating the probability of e-bike use. As a set, the sociodemographic characteristics reliably predicted e-bike use, as shown by the significant value for Nagelkerke R2. The likelihood of using an e-bike increases with age. Gender is also a strong predictor, with females being 1.99 times more likely to own an e-bike than men. Higher education is negatively related to e-bike use, but on the other hand, making enough money to cover household expenses increases the odds of using an e-bike with 95.9%. E-bike use is 86.2% more likely in households without children than in households with children. Finally, e-bike use is also more likely among students and commuters than among retirees. Physical disability and degree of urbanization were not significant predictors of e-bike use.

Table 3 Unstandardized coefficients for logistic regression of e-bike use on blocks of sociodemographic, mobility and attitudinal factors (0 = no, 1 = yes), N = 612
Full size table

The mobility factors add further explanatory value to the model, as shown by the substantial increase in Nagelkerke R2. Car ownership is positively associated with using an e-bike, with car owners being 45.7% more likely to use an e-bike than non-owners. This suggests that e-bike use can complement car travel. Furthermore, e-bike users are associated with lower numbers of bicycles in the household. This suggests that e-bike use substitutes rather than complements bicycle use. Daily travel distance also contributes to the model. Respondents traveling less than 15 km on average to their daily activities are 98% more likely to use an e-bike than respondents traveling 15 km or more on average. Finally, e-bike users are much less likely to be primary bicycle users, and less likely to be primary car users, which suggests substitution of both modes at least to a certain extent. Other mobility factors in model 2 do not affect e-bike use, such as license ownership and use of public transport as the main mode of transportation.

The regression weights of the sociodemographic characteristics included in model 1 slightly change after inclusion of the mobility factors in model 2. Age, gender, financial situation, and being a student or commuter remain significant predictors. However, the contribution of higher education and households with children is less significant after adding mobility factors. This might be due to higher educated respondents and households with children being more likely to own a car, or living closer to their work and other daily activities.

Finally, the explanatory value of the model as indicated by the Nagelkerke R2 again increases substantially after adding the attitudinal variables. Four values considerably increase the likeliness of using an e-bike: higher valuation of e-bike safety, reliability, fun, and health benefits. A higher valuation of image was however negatively related to e-bike use. After addition of the attitudinal variables, the explanatory power of car ownership changed to non-significant, although still positive.

Potential e-bike use

The results of the logistic regression analysis estimating the probability of willing to use an e-bike are shown in Table 4. Again, all sets of predictors contributed significantly to the predictive power of the model, although the predictive efficacy, as indicated by Nagelkerke R2, was generally lower. Of the sociodemographic characteristics, completion of higher education and making enough money to cover expenses both significantly contribute to willingness of using an e-bike. These results suggest that e-bike use does appeal to lower educated people with a less favorable financial situation. The other sociodemographic characteristics do not significantly contribute to the model.

Table 4 Unstandardized coefficients for logistic regression of potential e-bike use (0 = not willing to use, 1 = willing to use), on blocks of sociodemographic, mobility and attitudinal factors N = 374
Full size table

Of the mobility factors, bicycle ownership is negatively related to the willingness to use an e-bike. Respondents who do not use a bicycle as their main mode of transportation are almost three times more willing to adopt an e-bike than bicycle users. These findings suggest that future e-bike adoption might be lower among people that currently own a bicycle or have a bicycle at their disposal. Distance also contributes to the model: respondents who travel less than 15 km on average to their different destinations are 51.1% more likely to be potential e-bike users than respondent who travel more than 15 km. Although mainly using the car and public transport to get around are positively related to potential e-bike use, their contributions are non-significant. Similarly, the other mobility factors added in model 2 do not predict potential e-bike use in rural areas. After inclusion of the mobility factors the significant negative impact of income situation on potential e-bike use disappears. This could result from higher household bike ownership and bicycle use among those with a less favorable income situation. Households with children are however positively correlated with potential e-bike use after inclusion of the mobility factors.

Finally, the attitudinal factors positively contribute to potential e-bike use. Respondents who value the fun aspect of e-bikes more, are more willing to use an e-bike. This stresses that, similar to current e-bike use, valuing enjoyment seems to be an important predictor of potential e-bike use. Also, a positive valuation of e-bike image contributes to potential e-bike use. The other attitudinal factors do not predict potential e-bike use. The contribution of sociodemographic characteristics to potential e-bike use does not change after inclusion of the attitudinal factors. However, within the set of mobility factors, the average distance traveled no longer significantly contributes to model 3 after inclusion of the attitudinal factors. A possible explanation for this is that people with a more positive stance on e-bike image and enjoyment are more likely to be willing to use an e-bike regardless of the distances they travel.

Discussion

This study examined the determinants of current and potential e-bike use in rural areas by means of a survey in the North of the Netherlands. The findings suggest that the e-bike is already used among a broad population of varied ages and backgrounds and for different purposes here. Among respondents who did not own an e-bike, especially those with low education and income levels showed more willingness to use an e-bike in the future. Number of cars in the household was positively related to current e-bike use, and number of bicycles in the household was negatively related to both current and potential e-bike use. This suggests that the e-bike complements car ownership and substitutes bicycle ownership. However, current e-bike users less likely use a car or regular bicycle as their primary mode of transport. Those who are willing to use an e-bike are less likely to currently use a regular bicycle as their main mode of transport. This suggests that the e-bike can substitute both car and bicycle use to some extent. However, bicycle users seem somewhat more reluctant towards owning or adopting an e-bike than regular car users, suggesting a greater potential for a shift from car travel. Average travelling distances over 15 km had a negative effect on e-bike use. This suggests that the e-bike is most suitable for shorter distances up to 15 km, distances beyond this range are traveled by motorized modes. Respondents who currently use an e-bike, or are willing to do so, generally have a more positive attitude towards the e-bike than non-users, especially regarding the enjoyment provided by the e-bike. However, current e-bike users rated the image of the e-bike less favorably than non-users.

Findings of the present study regarding differences between those who currently use an e-bike and non-users are largely in line with previous studies, which have also found e-bike use to be higher among women, older age groups and those with a lower income and education level (Fishman and Cherry 2015; Fyhri and Fearnley 2015; KiM 2016; Rose 2012), although the results regarding age and education level have thus far been somewhat ambivalent (Johnson and Rose 2015; Popovich et al. 2014). The positive relation between being a student or commuter and e-bike use seems in line with earlier studies that proposed that potential for e-bike use in these groups is high (Plazier et al. 2017a, b). Unlike previous research, the present study did not find any support for higher e-bike use among the physically impaired (Peine et al. 2016). E-bike users less often reported mainly using a car or a bicycle to access activity locations, which is in line with previous findings that e-bikes can substitute both car and bicycle travel (Lee et al. 2015). In this particular Dutch context, no evidence was found for substitution of using public transport by e-bikes, which is consistent with previous findings indicating that substitution of a mode depends on its level of use in that context (considering that use of public transport in Dutch rural areas is relatively low, while both car and bicycle use are high) (Kroesen 2017).

The differences in attitudes towards e-bikes between users and non-users revolve around aspects of safety, fun and health benefits which have previously been found to constitute recurring themes in the e-bike literature (Berntsen et al. 2017; Haustein and Møller 2016; Langford et al. 2017; Plazier et al. 2017b; Poos et al. 2017; Schepers et al. 2014; Vlakveld et al. 2015). The present study provides further support for the importance of these aspects in individual decisions of rural residents to use an e-bike. The importance of reliability as a determinant of e-bike use is less well-studied. However, it might relate to the predictability of travel time, which has been found to be an important motivation for daily commuting by e-bike over use of automobile and public transport (Plazier et al. 2017a). In rural areas, travel times by e-bike may be more predictable than those by public transport, which is often infrequent and irregular. E-biking may also be more reliable than cycling, which is more sensitive to weather circumstances such as wind. An unexpected result is that current e-bike users have a more negative image of e-bikes than non-users, where one would expect them to have a more positive image. A possible explanation for this finding is that e-bike users are aware of the negative stigma of e-bikes as “old people’s bikes”, perhaps because they have been reminded of this by their peers (Jones et al. 2016; Plazier et al. 2017b).

The present study also sheds some more light on the determinants of willingness to use an e-bike among those who currently do not use an e-bike. In general, there were less differences between potential e-bike users and non-potential users, than between current users and non-users. This finding may be partly due to methodological issues. For instance, due to the lack of a forced response design in the survey, there were many respondents with missing data on key variables. However, the final sample proved comparable to the initial sample on key variables. The subsample of respondents who do not currently use an e-bike was less varied in terms of socio-demographic variables (e.g., they were generally younger, and more often male), which reduces the power to discriminate between groups within the sample. Potential e-bike use was correlated with a relatively low level of income and education, while current e-bike use was correlated with a relatively high level of income. This suggests that there is a high potential for e-bikes among rural residents with a low socio-economic status, which could be tapped by making e-bikes more affordable or available otherwise. Current e-bike use is more likely in households without children, but on the contrary, potential e-bike use is more likely in households with children. Compared to current e-bike users, potential users are more often car users, and less often bicycle users, suggesting a higher potential for e-bikes among car users. Finally, the positive association between image of e-bikes and willingness to use an e-bike suggests that the branding of e-bikes may be an effective strategy to attract new users.

Despite its potential role in realizing more healthy and sustainable rural transport systems, e-bike use in rural areas has received little attention so far. A main strength of the current research is that we evaluated current and potential e-bike use in a rural environment. We find that many of the determinants of e-bike use in rural areas correspond with determinants found in earlier studies, even though these did not specifically address rural e-bike use. Another strength of our research is that we were able to identify and distinguish between the factors contributing to current e-bike use and potential e-bike use in these areas.

However, we also identified some limitations. First of all, participants were not randomly selected, and the study sample is not be representative for the study area. Self-selection of participants is probable, because the sample contained a high share of e-bike users compared to available data from the Netherlands (Table 1). This limits the generalizability of the findings. Also, the generalizability of study findings to other countries is limited. Compared to rural areas in other countries, the North of the Netherlands are characterized by relatively high population densities. Thus, findings might not easily translate to rural areas in other countries, although they might also apply to low-density (peri-) urban areas, such as sprawling suburban neighborhoods. In line with this, Dutch rural areas are characterized by high levels of bicycle use and relatively good quality of public transport and cycling infrastructure. The cycling culture already in place, and the flat topography in the study area may further limit the generalizability of findings. A second limitation is the large number of missing values which were omitted from the logistic regression analyses. This reduces the power of the analyses, and may have led to response bias in case the missing cases are non-random. However, a comparison between the initial and final sample suggests that bias effects are limited. Third, the cross-sectional nature of the study does not allow for conclusions on causal relationships between the factors studied and current and potential e-bike use. Fourth, and in line with the previous limitation, the exploratory nature of this study prompted the use of a simple causal model to study the effect of sociodemographic variables, mobility factors and attitudes on travel behavior. Some studies have suggested that more intricate relations exist between attitudes, built environment and travel behavior. For instance, travel attitudes and travel behavior might change over time in response to built environment characteristics (van de Coevering et al. 2016). Fifth, the effect of built environment on travel behavior was measured by considering the effect of urbanization rate on actual and possible e-bike use. Here, urbanization rate is expressed in terms of population density, and used as a proxy for built environment characteristics such as road infrastructure density. Studies have suggested that population density can be a good determinant of infrastructure density (Glover and Simon 1975). However, relations between built environment and travel behavior are better determined by directly addressing these aspects of built environment characteristics in survey questions or through spatial analyses. And finally, we stress the limitations of quantitative assessment of self-reported mobility behavior through surveys. Reported results are based on aggregated data and not necessarily form accurate representations of current individual travel and decision-making behavior.

Future quantitative studies may address rural e-bike mobility using more representative study samples. These studies could focus on minimizing the number of missing responses, and look for possibilities of including missing responses in their analyses. Also, more insight in current and potential e-bike use in rural areas is needed from other geographical contexts, including areas with less bicycle infrastructure, lower acquaintance with cycling in general, and different climatic circumstances and topography. Longitudinal studies on e-bike use might provide more insight in the factors that contribute to a shift from potential to actual e-bike use and participant experience with everyday use over time. Studies may also consider the intricate relations between attitudes, built environment and travel behavior, and more directly address the role of built environment characteristics on travel behavior in the process. Finally, we also stress the importance of evaluative field research, for instance through pilot-testing of e-bikes. These studies can provide further insight in travel behavior prior to and after e-bike adoption, and the variety of motivations concerned with mode choices and uses.

Results imply that current e-bike use in rural areas to some extent ‘fits the picture’ in terms of the determinants identified in previous studies. However, this study has also revealed that use by students and commuters is more common than perhaps thought beforehand. Thus, there may be more potential to realize e-bike use in these demographic groups. The results also suggest a high potential for e-bike use among those with lower socioeconomic status, among households with children, and among those who currently mainly get around by car. These findings offer opportunities for social marketing campaigns and targeted subsidizing and branding of e-bike use to improve the mobility situation for rural inhabitants who might benefit from an additional mode of transport in their everyday lives, be it through substitution of current mode use or through complementation.

Conclusions

Electrically assisted cycling or e-biking is growing in popularity among an increasingly diverse range of users. Results from a survey show that e-bike use is already widespread among people of different ages and occupations in a rural area in the Netherlands. This indicates that for a growing number of people in rural areas the e-bike is a potential addition to the current array of active and motorized modes. A further shift to e-bike use from the motorized modes can benefit health, flexibility, enjoyment and environment, provided it does not replace the active modes. Findings show that rural bicycle users are more reluctant towards adopting an e-bike, while car users are more susceptible to start using an e-bike. Strategies involving the development of e-bike use in rural areas can aim at reducing car use, thereby alleviating possible dependence on automobile transport. This can include financial incentives for persons of lower socio-economic status or living in households with children to make it more attractive to purchase an e-bike. Alternatively, the organization of e-bike pilot-projects in rural areas can help change attitudes by enabling first-hand experiences. These findings can support the development of sustainable transport systems that support active and healthy lifestyles.