Micro-Sharing Mobility for Sustainable Cities: Bike or Scooter Sharing?

Abstract

Urban pollution awareness is a subject of widespread debate, particularly regarding the role of future urban transportation. In recent years, local policymakers and private operators have implemented various measures to address the negative impacts of transportation, including promoting micro-shared mobility services. Our research investigates the factors influencing citizens’ decisions to use these services, focusing on e-bike sharing and e-scooter sharing. We collected data on individual mobility patterns in Italian cities and administered hypothetical choice tasks to examine revealed and stated preferences. Our findings highlight the most influential factors guiding users’ decisions and identify the preferred sharing option between e-bikes and e-scooters. The implications of our results could provide valuable insights for local regulators and shared mobility operators in designing effective and sustainable future transportation policies.

1. Introduction

Active shared micro-mobility services have been advocated in urban areas to address the adverse effects of the transportation sector [1,2,3,4,5]. These services also yield positive effects, such as reducing noise and air pollution, easing traffic congestion, and enhancing the overall quality of life and accessibility in urban areas [6,7]. Their importance is expected to grow in the coming years, as the urban population is projected to increase to 70% by 2050, resulting in negative urban externalities [5,8].

In recent years, in Italy, there has been an increase in the number of trips made via sharing services [9,10]. However, according to [11], only 5% of the population uses sharing mobility systems in Italy, and most are young users. Moreover, there is still a geographical imbalance in the use of mobility systems between the centre, north, and south of Italy, as well as between large cities and less populated areas [10]. The underuse and undersupply in southern Italy is due to several factors, including the lack of cycle lanes, higher levels of traffic congestion, and the greater likelihood of vehicles being stolen or damaged [12]. The bike sharing system is the most used in Italian cities. It is provided via bike stations distributed within a definite area and either muscular or electric bicycles [13]. An alternative system emerging in the urban context is scooter sharing, with scooters powered by an electric engine. Many large Italian cities have introduced this additional type of micro-active sharing service, and national and local regulators are issuing rules to promote its diffusion while preserving the safety of both users and pedestrians. Indeed, e-scooters are considered a flexible transport system covering the gap between individual and public transport by providing an extended feeder service, particularly for urban areas characterized by irregular demand [14]. Since 2020, there has been a significant increase in the demand for shared micromobility services, which has led Italian cities to shift urban traffic patterns towards this mode of transportation [11].

Although e-scooters and e-bikes have typically been regarded as substitutes in urban shared mobility systems, they differ in speed and acceleration, which has implications for the behaviour of users on the street [15]. Furthermore, these two systems require different transportation regulations and infrastructures, which vary based on geographical contexts and city sizes. This includes street networks, road design, and different services and investments provided by private transport operators. Several cities, including Luxembourg, Riga, Paris, and Malta, have recently changed urban regulations, promoting bike sharing services and opposing or overregulating e-scooter sharing services [16]. Therefore, it is important to consider how the differences between geographical areas and between small, medium, and large cities impact users’ preferences and mobility choices [2]. This information is essential for private operators to make informed investment decisions based on consumer preferences for future transportation initiatives.

The existing literature tends to group micromobility services together, failing to acknowledge the variables that may influence the selection of one service over another. Additionally, the focus is often limited to large cities. This oversight could result in the inappropriate deployment of sharing supply services in smaller cities and inefficient investment management. A recent paper examines the factors influencing users to select privately owned e-scooters over public bike sharing services [16]. However, to our knowledge, no study has yet examined the preferences between public e-scooters and public e-bike sharing systems in urban areas. To address this gap in the literature, we collected and analyzed revealed and stated preference data to study individual preferences for Italian citizens in 2021. The questionnaire includes two different choice settings. The first one compares two different e-bike sharing services provided by different operators, while the second one compares an e-bike sharing service and an e-scooter sharing service. In the discrete choice experiment, we studied user charges (per minute vs. fixed), service area, the number and range of vehicles, operating system type (station-based vs. free-floating), and vehicle esthetics. We also investigated whether the size of the cities and the geographical area (north, central, and south Italy) where the respondents live influenced their preferences for e-bike and e-scooter sharing services. Last, we looked at how the socio-demographic characteristics of the respondents influenced their preferences.

The results of our analysis will help to identify the investments needed to provide micromobility services that are appropriate to the urban and geographical characteristics of the area under consideration. In fact, providing bike and scooter sharing services via electric vehicles could increase the number of people switching from private motorized vehicles to active mobility, as electric bikes and scooters require less physical effort, are faster, and allow one to reach distant destinations. However, providing the service via electric vehicles implies higher purchase, maintenance, and operating costs of the fleet. It is therefore crucial to determine whether the potential increase in demand guaranteed by the use of electric means would compensate for the expected higher costs. Our results shed light on this issue by showing how the business models of private or public operators should be adapted to the specific needs of the areas served.

The rest of the paper is structured as follows: Section 2 discusses the related literature on bike sharing and scooter sharing systems. Section 3 presents the methodology and the data used. Section 4 discusses the empirical results. Finally, Section 5 provides concluding remarks and future transportation policies.

2. Literature Review

The literature has analyzed the bike sharing system in detail over the years, focusing on several aspects like the availability of bicycle facilities, service design attributes, and parking and traffic facilities [2,17,18,19,20,21,22,23,24,25,26,27,28]. Ref. [26] conducted a comprehensive literature review on bike sharing, focusing on papers published during the COVID-19 pandemic. They analyzed the impacts of the pandemic on actual and potential user preferences, as well as on bike sharing demand compared to other transportation options. Cost saving is one of the most influential elements for bike sharing users [29,30]. Wu et al. [31] found utility, fun, and perceived ease of use as being the most important factors. Ref. [2] pointed out that factors such as environmental awareness, willingness to avoid congestion, the importance of safety for cyclists, and physical activity are the main motivations for users. They found that attitudes and psychological barriers, such as finding biking boring, tiring, or uncomfortable, have a negative impact on the usage of bike sharing systems. The existence of adequate infrastructure, the level of safety for cyclists, and the convenience of using the service are also important factors for users. They also found that respondents would use bike sharing mainly for commuting to work or study. Using a stated preference survey, ref. [32] found that users prefer cycling in non-residential areas to avoid traffic congestion. Ref. [33], using a stated and revealed preference survey, noted that cycle paths provide more incentives for inexperienced cyclists. Similarly, ref. [34] highlighted how unsuitable cycling infrastructure could hinder the adoption of bicycles.

The demographic variables play a significant role in bike sharing adoption. According to [35], younger individuals with full-time jobs are more likely to use bike sharing systems, while older people and women tend to use them less. Ref. [36] found that men, young people, and those in higher professional positions were more likely to use bike sharing systems, while [37] found that people with an annual income of between €15,000 and €25,000 were more likely to use bike sharing.

There are only few studies on scooter and e-scooter sharing services. Among the most recent, ref. [38] examined the role of safety and convenience in encouraging the use of e-scooter sharing and found that the provision of dedicated e-scooter lanes and real-time travel information significantly increases the potential demand for ridesharing. Only [39] compared preferences for bike sharing, both electric and conventional, and e-scooter sharing services. They focused on the cost of the trip, the type of service, either free-floating or station-based, and the walking time to reach the bike or scooter, and found that all the attributes studied significantly influenced the willingness to use a shared service, and that free-floating services were preferred to station-based services. Ref. [16] studied preferences between privately owned e-scooters and shared bikes. They used a logistic regression model and found that e-scooter users are more satisfied than bike sharing users. Additionally, while those who use bike sharing systems acknowledge the environmental benefits of using bikes, e-scooter users are more interested in the convenience and speed of using the vehicle, as well as the cost savings compared to other modes of transportation. Similar results have been found in [40]. Ref. [41] conducted a study on users’ preferences for cars, public transport, e-scooters, and bike sharing using stated preference techniques. However, the study only focused on two attributes, cost and travel time, without considering other potential attributes that could influence users’ choices. The study found that users prefer public transport, followed by bike sharing and e-scooters, over private cars. Ref. [42] conducted a study comparing scooter sharing, walking, and bus services in Seoul using a stated preference methodology. The respondents were asked to choose their means of transport based on travel time savings and personal income. The study found that people may find it more convenient to combine subway and scooter sharing than to use a car for the whole journey. In their study, ref. [43] used factor analysis and a structural equation model to determine that mobility habits and environmental awareness are important factors influencing the use of scooter sharing among university students in Taiwan. Meanwhile, ref. [44] conducted a stated preference experiment and found that scooter sharing addresses excessive indirectness, multiple transfers, and long-access walking. However, safety, weather conditions, and baggage capacity limit its potential usage [45].

Papers that consider socio-demographic variables have shown that the impact of e-scooters is similar to that of bike sharing systems, as both often replace walking, cycling, and public transport usage [46,47]. Ref. [48] analyzed the scooter sharing system in Spain. They found that young and highly educated people are more likely to use this system, while those with higher incomes are not interested. Also, ref. [49] finds that graduate students, full-time workers, males, and young people are more willing to use shared electric micromobility services. Using a probit model, ref. [50] identified that most scooter sharing users in Zurich are young, educated, fully employed men without children. According to this research, females prefer to use scooters rather than bike sharing services. Furthermore, those who own e-scooters, e-bikes, or public transportation season tickets are more likely to use scooters and bikes for sharing. Similar results have been found by [51], according to which e-scooters are mainly used for short trips by middle-aged males with relatively high education.

While the existing scientific literature mainly focuses on the potential demand for bike or scooter sharing, there is a lack of studies on the choice between the two types of shared micromobility, the only exception being [39]. In particular, there is a lack of studies focusing on electric micromobility, which is potentially more attractive than conventional micromobility, and of those investigating the most important attributes for potential users. In addition, there are no studies analyzing how the importance of the service attributes varies according to the geographical area considered in terms of the availability of dedicated micromobility infrastructure and the size of the cities served. As the provision of shared micromobility services with electric vehicles is much more expensive and technically complex than with conventional vehicles, it is crucial to understand the preferences of potential users for the characteristics of the service. Given the relevance of the missing information both for public administrations that regulate and subsidize micromobility services and for operators that have to design and provide them, we decided to investigate this topic further. Our results will hopefully fill the existing gap in the literature.

3. Methodology

The questionnaire used to collect the data comprises several sections and is reported in Appendix A.

3.1. E-Bike Sharing Services

The first part of the questionnaire aimed to detect if respondents had used an e-bike sharing service (Q1), the factors that influenced their decision to use (Q2) or not use (Q7) such a service, and the key characteristics that respondents believed an e-bike sharing service should possess based on their preferences (Q6 and Q8). Respondents who stated to have used an e-bike sharing service were also asked about their frequency of use per month (Q3), trip purpose (Q4), and whether they primarily used the service on weekdays or weekends (Q5). This section of the questionnaire enabled us to gather revealed preference data on the decision to use an e-bike sharing service and the factors that influence this decision in the real-life setting experienced by the respondents.

To also test the preferences of the respondents who never used the service and to check what would be the most important technical and operational characteristics of the service for all the respondents, in the second part of the questionnaire we administered a Discrete Choice Experiment (DCE). This part was aimed at determining whether respondents would be willing to use an e-bike sharing service, and was characterized by the following attributes and attribute levels:

• Vehicle design: standard vs. trendy;
• Operating system: free-floating vs. station based;
• Operating range: 20 km, 40 km, 60 km;
• Vehicle availability: high, medium, low (each probability level was depicted via a picture describing the number of vehicles available on the city map);
• Extension of the area served: large, medium, small (each area extension was described via a picture like those reported in Figure 1);
• Per minute tariff: €0.02, €0.15, €0.30;
• Fixed tariff to unlock the vehicle: €0, €0.40, €0.80.

To select the attributes and levels of attributes to be studied, we conducted a focus group with colleagues from the Universities of Trieste and Bari and with experts in the field of sustainable mobility. We also carried out a pre-test survey with a group of 30 people to validate the questionnaire, check the clarity of the proposed choice tasks and collect feedback on the research topic. On the basis of the feedback received, we extended the range of values proposed to operating range and tariff per minute, shortened the questionnaire, and changed the order of the questions in order to reduce the respondents’ burden.

Figure 1 provides an example of one of the choice tasks administered during this part of the survey. Respondents were asked to indicate their preferred alternative for the services described in the hypothetical possibilities for users, assuming that they were available in their city of residence or work. To collect a larger number of stated preference data, we administered twelve choice tasks to each respondent, each task differing in the attribute levels that characterized the hypothetical service alternatives (Q9–Q20).

Compared to the revealed preference approach, the use of stated preferences, and more specifically of DCEs, enables better handling of multi-dimensional possibilities for users that are not yet available in the market. Furthermore, DCEs are designed to improve internal consistency in respondents’ choices and provide more informative results. This is because respondents have multiple opportunities to express their preferences for a particular good or service across a range of payment amounts [52].

To limit the cognitive difficulty associated with multiple complex choices, we used an orthogonal fractional factorial design to select the combination of attribute levels characterizing each service alternative [53]. We opted for an orthogonal design because it satisfies attribute-level balance, avoids collinearity problems among variables, and guarantees that each parameter is independently estimable. We could not use an efficient design since we were missing the a priori data needed regarding the characteristics of the service we wanted to study.

To minimize hypothetical bias [54] in data collected through stated preference experiments, we anchored attribute levels to real-life settings. The tariff’s fixed and variable components, as well as the operating range values, were defined on the basis of the most commonly used average values in Italy. To ensure data reliability, we also included the ‘no-choice option’ (opt-out alternative). Finally, to enhance the robustness of the collected data, we collected revealed preference data prior to the state preference section, as suggested by [53,55,56,57].

3.2. E-Scooter vs. E-Bike Sharing Services

The third part of the questionnaire aimed to detect if respondents had used an e-scooter sharing service (Q21) and would use it if provided in the city where the respondents lived (Q22). They were also asked to state why they would (Q23) or wouldn’t (Q25 and Q39) use such a service, and the key characteristics that respondents believed an e-scooter sharing service should possess based on their preferences (Q24, Q26, Q40). Respondents who stated that they had used an e-bike sharing service were also asked about their frequency of use per month (Q3), trip purpose (Q4), and whether they primarily used the service on weekdays or weekends (Q5). This section of the questionnaire enabled us to gather revealed preference data on the decision to use an e-scooter sharing service and the factors that influence this decision in the real-life setting experienced by the respondents.

The fourth part of the questionnaire aimed at detecting whether the respondents would prefer an e-bike or an e-scooter sharing service, that is, at collecting stated preference data. Each e-bike sharing service was described with reference to the same attributes and attribute levels described in Section 3.1. As for the e-scooter sharing alternatives, we used the same characteristics in defining the e-bike service alternatives, except for the fixed fee to unlock the scooter, which was set equal to €0, €0.75, or €1.50.

We used a fractional factorial design to select the combination of attribute levels characterizing each service alternative. Twelve choice tasks asking the respondents to choose between a hypothetical e-scooter sharing service, a hypothetical e-bike sharing service, and the ‘no-choice option’ were administered during this part of the survey (Q27–Q38). Figure 2 depicts an example of one of the choice tasks proposed to the respondents.

3.3. Vehicle’s Ownership and Socio-Demographic Characteristics.

With the last part of the questionnaire, we collected data on the number of vehicles (cars, motorbikes, bikes, and scooters) owned by the household (Q41–Q43), the number of scooters and bicycles bought via the subsidy granted by the Italian government in 2020 (Q44), and the frequency of the public transport use (Q46). Last, we collected data on the socio-demographic characteristics of the respondents: age (Q47), gender (Q48), education (Q49), occupational status (Q50), residential location (Q51 and Q53), number of household members (Q45), commuting destination (Q52), and personal disposable income (Q54). We also collected information on the municipality size (small, medium, and large) and the geographical location of the municipality (north, centre, or south of Italy).

3.4. The Sampling Strategy

We administered the questionnaire to a heterogeneous sample via different document sharing platforms and social media, including Moodle (for university students). Due to financial and time constraints, we could not carry out the survey via face-to-face interviews. We collected 843 valid answers.

Our sample includes individuals living in 246 Italian municipalities, which vary considerably in size, population density, geographical location, public transport services, the number and type of shared transport services, and extent of bicycle lanes.

Table 1. The characteristics of the municipalities comprised in the study compared to the rest of the country.

Sample
	Min	Max	Average
N. inhabitants	541	2,748,109	62,274
Density	13	7680	654
km2	3	1287	100
	<20,000 inhab.	20–90,000 inhab.	>90,000 inhab.
Size	36%	25%	39%
	North	Centre	South
Geographical location	24%	8%	68%
Italy
	Min	Max	Average
N. inhabitants	31	2,748,109	7472
Density	1	11,927	298
km2	1	1287	38
	<20,000 inhab.	20–90,000 inhab.	>90,000 inhab.
Size	47%	28%	25%
	North	Centre	South
Geographical location	46%	20%	34%

shows the main characteristics of the individuals’ residence places in our sample and the corresponding values for the Italian population. Compared to the rest of the country, our sample includes a higher percentage of people living in larger cities (>90,000 inhabitants) and in the south of Italy. The Italian map is shown in Figure 3.

The descriptive statistics of the main socio-demographic characteristics of our sample and the Italian population are reported in

Table 2. Descriptive statistics for the sample and the Italian population.

Variable	Sample		Italian Population
	Mean	Std. Dev.	Mean
Age	27.13	10.69	46.4
Gender
(1 = female)	0.51	0.50	0.51
Education level
(0 = middle school; 1 = high school;
2 = bachelor; 3 = masters; 4 = PhD.)	1.48	0.68	0.77
Income
(0 = EUR 0–499; 1 = EUR 500–899; 2 = 900–1399;
3 = EUR 1400–1799; 4 = EUR 1800–2199; 5 = EUR 2200–2599;
6 = EUR 2600–2999; 7 = ≥ EUR 3000)	1.58	1.91	1.70
Public transport use per month
(0 = never; 1 = rarely;
2 = 1–4 times/1 per week; 3 = almost every day)	1.47	0.96	n.a.
N. of family members	2.71	0.89	2.3
N. of scooters owned	0.13	0.40	n.a.
N. of bikes owned	2.10	1.38	n.a.
N. of cars owned	2.12	0.89	1.51

.

The sample is gender-balanced and reflects the same proportion of men and women as in the Italian population. The number of family members is also in line with that of the Italian population. However, the average age of the respondents is much lower than that of the Italian population (27 years compared to 46 years), resulting in a higher educational level for the sample compared to the population. This is most likely due to the fact that we administered the questionnaire online rather than through face-to-face interviews. This misalignment may reduce the generalizability of our findings to the wider population. However, it should be noted that in Italy, e-bike and e-scooter services are more commonly used by younger individuals. A survey conducted in 2022 by the National Sharing Mobility Observatory in the city of Bologna found that only 6% of the sample population aged over 65 uses bike or scooter sharing services, despite accounting for 27% of the city’s inhabitants. Regarding personal net monthly income, the average value of the sample aligns with that of the Italian population. However, the households of the respondents own a larger number of cars on average compared to the general population.

4. Results

4.1. Revealed Preference on Bike Sharing Use

We separately analyzed the revealed and the stated preference data, also differentiating the analysis with regard to the choice setting. First, we estimated a binary logit model to understand which factors influenced respondents’ choice between using or not using a bike sharing service in the past (

Table 3. Binary logit model of bike sharing use (revealed preference data). Dependent variable: bike sharing use vs. non-use (binary—answer to question Q1).

	Coefficient	S.E.
Constant	0.24	0.47
Age (cardinal) [Q47]	−0.04 ***	0.01
Woman (binary vs. man) [Q48]	−0.41 ***	0.17
Number of family members (cardinal) [Q45]	−0.25 ***	0.10
Public transport frequency use (ordinal-0 = never; 1 = rarely;
2 = 1–4 times/1 per week; 3 = almost every day) [Q46]	0.16 **	0.09
Small municipality (binary-1 = <20,000 inhab.) [Q51]	−0.30 *	0.18
North of Italy (binary vs. South) [Q51]	0.44 **	0.22
Centre of Italy (binary vs. South) [Q51]	0.72 ***	0.27
Number of bicycles available (binary—1 = mentioned as important feature) [Q8]	0.72 ***	0.20
Operating range (binary—1 = mentioned as important feature) [Q8]	2.01 ***	0.56
Chi-square (9) = 71.80 prob Chi-square < 0.001 (N = 843)
AIC = 878.9
Adj R2 = 0.08

Note: the demand code corresponding to the variable specified in the model is reported within the square brackets; (*) p-value ≤ 0.05; (**) p-value ≤ 0.01; (***) p-value ≤ 0.001.

). The dependent variable is the answer to question Q1 in the Appendix A, which is a binary variable, equal to one if the respondent reported having used the service in the past and zero otherwise. We also performed a similar analysis regarding past use of the scooter sharing (Table 4).

The econometric analysis based on the revealed preference data is presented in this and the following subsection. In both cases, the goodness of fit index of the estimated models (Adj R²) is quite low. This is partly due to the fact that there are no repeated observations for each respondent and the number and variability of the independent variables are limited [56]. However, the model is capable of correctly predicting 76% of the observations collected on the current use of bike sharing services. Moreover, as the estimated coefficients are statistically significant and in line with either our expectations or the results reported in the literature, we believe that the results are worth commenting on, all the more so considering the lack of empirical evidence on the factors that influence the real choice of micromobility services in Italy.

Regarding the past choice of using the existing bike sharing services (Table 3), we find that women (−0.41), older people (−0.04), and those living in larger households (−0.25) are less likely to use the existing bike sharing system. Our results are consistent with the findings of [58,59], that women are less likely than men to use bike sharing, especially for commuting. They are consistent also with the findings of [35,36,37,60], that user age is the most important factor influencing bike sharing adoption, along with place of residence. Regarding the role of the number of family members, ref. [61] also find that individuals from single-person households are more likely to use bike sharing services, as the presence of children makes it more difficult for an individual to use a bicycle. According to our results, frequent use of public transport services increases the likelihood of using a bike sharing service (0.16), while living in small, less-densely populated cities decreases the likelihood (−0.30), in line with the findings of [62]. We find significant differences in the propensity to use bike sharing services according to the macro-area of residence of the respondents. Those living in northern (0.44) or central Italy (0.72) are more likely to use bike sharing than those living in the south or on the islands. This may be due to the greater offer of bike sharing services in northern and central Italy but also to the greater extension of bicycle lanes. Last, we find that the characteristics of the bike sharing service that influence the likelihood of using the service are the number of bikes available (0.72) and the operating range (2.01).

4.2. Revealed Preference on Scooter Sharing Use

In Table 4, we report the results analysis of the sample’s revealed preference data in terms of the current use of the scooter sharing service. The dependent variable is a binary variable equal to one if the respondent reported using the service in the past and zero otherwise (answer to question Q21 in the Appendix A).

Table 4. Binary logit model of actual scooter sharing use (revealed preference data). Dependent variable: scooter sharing use vs. non-use.

	Coefficient	S.E.
Constant	−0.36	0.67
Age (cardinal) [Q47]	−0.04 **	0.02
Woman (binary vs. man) [Q48]	−0.48 **	0.20
Student (binary vs. other occupational status) [Q50]	0.63 **	0.32
Centre of Italy (binary vs. north and south) [Q51]	0.57 *	0.32
Large municipality (binary—1 = >90,000 inhab) [Q51]	0.33 *	0.21
Scooter parking easiness (binary—1 = mentioned as important feature) [Q40]	0.45 **	0.21
Trendy transport choice (binary—1 = mentioned as important feature) [Q40]	0.74 ***	0.30
Operating range (binary—1 = mentioned as important feature) [Q40]	1.62 ***	0.65
Chi-square (8) = 53.34 prob Chi-square < 0.001 (N = 475)
AIC = 595
Adj R2 = 0.08

Note: the demand code corresponding to the variable specified in the model is reported within the square brackets; (*) p-value ≤ 0.05; (**) p-value ≤ 0.01; (***) p-value ≤ 0.001.

Table 4. Binary logit model of actual scooter sharing use (revealed preference data). Dependent variable: scooter sharing use vs. non-use.

	Coefficient	S.E.
Constant	−0.36	0.67
Age (cardinal) [Q47]	−0.04 **	0.02
Woman (binary vs. man) [Q48]	−0.48 **	0.20
Student (binary vs. other occupational status) [Q50]	0.63 **	0.32
Centre of Italy (binary vs. north and south) [Q51]	0.57 *	0.32
Large municipality (binary—1 = >90,000 inhab) [Q51]	0.33 *	0.21
Scooter parking easiness (binary—1 = mentioned as important feature) [Q40]	0.45 **	0.21
Trendy transport choice (binary—1 = mentioned as important feature) [Q40]	0.74 ***	0.30
Operating range (binary—1 = mentioned as important feature) [Q40]	1.62 ***	0.65
Chi-square (8) = 53.34 prob Chi-square < 0.001 (N = 475)
AIC = 595
Adj R2 = 0.08

Note: the demand code corresponding to the variable specified in the model is reported within the square brackets; (*) p-value ≤ 0.05; (**) p-value ≤ 0.01; (***) p-value ≤ 0.001.

Similar to what we found for the bike sharing service use, gender and age are also relevant for the use of scooter sharing. In fact, women (−0.48) and older people (−0.04) are less likely to use the scooter sharing services. Our results are consistent with the findings of [63], who found that only 37.5% of scooter sharing users were women, and that e-scooter users are on average younger than e-bike users. We also find that students (0.63) are more likely to use the scooter sharing service compared to people who have a different occupational status and that large municipalities (0.33), being better equipped with cycling infrastructures, attract more scooter sharing users, in line with the findings of [64]. People living in central Italy are more likely to use the scooter sharing service (0.57), probably due to the larger and denser supply of services that can be found there. The characteristics of services that favour the likelihood of using the scooter sharing service include the user-friendliness of the parking facilities (0.45), the fact that it is an innovative, trendy means of transport (0.74), and, above all, a long operating range (1.62). They are in line with the results found by [65] on the importance of the perceived usefulness, attitude, and perceived ease of use as well as the personal innovativeness for the intended use of a scooter sharing service. Also in this case, the goodness of fit of the estimated model (Adj R²) is quite low, but the model is able to correctly predict 67% of the observations collected on the current use of scooter sharing services, which is still a remarkable result given the small amount of data we were able to use for the econometric analysis.

4.3. Stated Preference on E-Bike Sharing Use

In order to explore what is the relative importance of the characteristics of an e-bike sharing service, we collected and analyzed the respondents’ stated preferences between two hypothetical e-bike sharing services and not using either of them (answers to questions Q9–Q20 in the Appendix A). We estimated a random-parameter logit model, which is better able to capture the heterogeneity of individuals’ preferences than the fixed-parameter logit model. To this aim, we also specified some interaction terms between the random coefficients and the socio-demographic variables of the sample. We report the results obtained in

Table 5. Random-parameter logit model: choice between two hypothetical e-bike sharing services and not using either of them (binary—answers to questions Q9–Q20).

	Coefficient	S.E.
Likelihood of using e-bike sharing services
Per minute tariff (hypothetical attribute, EUR, triangular mean)	−8.09 ***	0.43
Ts—per minute tariff (triangular spread)	8.09 ***	0.43
Per minute tariff * north of Italy (binary vs. centre or south) [Q51]	−1.62 ***	0.58
Per minute tariff * large municipality (binary—1 = >90,000 inhab.) [Q51]	1.64 ***	0.40
Per minute tariff * woman (binary vs. man) [Q48]	1.87 ***	0.39
Per minute tariff * income (ordinal, 1 = 0–499 €; ….8 = >3000 €) [Q54]	0.60 ***	0.10
Fixed tariff to unlock the vehicle (hypothetical attribute, EUR, triangular mean)	−1.81 ***	0.09
Ts—fixed tariff to unlock the vehicle (triangular spread)	1.81 ***	0.09
Fixed tariff to unlock the vehicle * large municipality (binary—1 = >90,000 inhab.) [Q51]	0.38 ***	0.12
Extension of the area served (hypothetical attribute, ordinal, 1 = small, 2 = medium, 3 = large, triangular mean)	1.35 ***	0.06
Ts—extension of the area served (triangular spread)	1.35 ***	0.06
Extension of the area served * centre of Italy (binary vs. north or south) [Q51]	−0.34 ***	0.10
Extension of the area served * woman (binary vs. man) [Q48]	−0.19 ***	0.06
Extension of the area served * workplace (binary, 1 = same municipality of residence and workplace) [Q51 and Q52]	−0.14 **	0.06
Vehicle availability (hypothetical attribute, ordinal, 1 = low, 2 = medium, 3 = high; triangular mean)	0.81 ***	0.05
Ts—vehicle availability (triangular spread)	0.81 ***	0.05
Vehicle availability * centre of Italy (binary vs. north or south) [Q51]	0.49 ***	0.12
Operating range (hypothetical attribute, km; triangular mean)	0.06 ***	0.00
Ts—operating range (triangular spread)	0.06 ***	0.00
Operating range (km) * centre of Italy (binary vs. north or south) [Q51]	−0.01 **	0.01
Operating range * large municipality (binary—1 = >90,000 inhab.) [Q51]	−0.01 ***	0.00
Operating range * medium municipality (binary—1 = 20,000–90,000 inhab.) [Q51]	−0.01 ***	0.00
Operating range * age (cardinal) [Q47]	0.0003 ***	0.00
Operating system (hypothetical attribute, 1 = station-based, 0 = free-floating; triangular mean)	0.04	0.08
Ts—operating system (triangular spread)	3.52 ***	0.16
Operating system (station-based) * north of Italy (binary vs. centre or south) [Q51]	0.71 ***	0.19
Vehicle design (hypothetical attribute, 1 = trendy, 0 = standard; triangular mean)	−0.04	0.06
Ts—vehicle design (triangular spread)	1.66 ***	0.21
Actual bike sharing user (binary, triangular mean) [Q1]	0.71 ***	0.21
Ts—actual bike sharing user (triangular spread)	0.71 ***	0.21
ASC hypothetical e-bike sharing on the right	−0.10 **	0.04
Likelihood of not using bike sharing services
Woman (binary vs. man) [Q48]	0.03	0.14
Age (cardinal) [Q47]	0.01	0.01
Student (binary vs. other occupational status) [Q50]	−0.94 ***	0.18
Income (ordinal, 1 = EUR 0–499; ….8 = >EUR 3000) [Q54]	−0.11 ***	0.04
Bike owner (binary) [Q41]	−0.27 ***	0.05
Residence area (cardinal, 1 = city centre, 2 = periphery, 3 = rural area) [Q53]	0.30 ***	0.10
Workplace (binary, 1 = same municipality of residence and workplace) [Q51 and Q52]	−0.21	0.14
North of Italy (binary vs. centre or south) [Q51]	−0.43 **	0.22
Centre of Italy (binary vs. north or south) [Q51]	−0.43	0.29
ASC opt-out option	−1.76 ***	0.37
Chi-square (35) = 6418.11 prob Chi-square < 0.001 (N = 9768)
AIC = 15114
Adj R2 = 0.299

Note: the demand code corresponding to the variable specified in the model is reported within the square brackets; (*) p-value ≤ 0.05; (**) p-value ≤ 0.01; (***) p-value ≤ 0.001.

.

According to our analysis, the most important factor influencing the likelihood of using an e-bike sharing service is the price per minute (−8.09). Still, respondents’ preferences are quite heterogeneous concerning the importance of this characteristic, as shown by the spread of the distribution of the parameter (8.09), which is statistically significant. In fact, while people living in the north of Italy are more sensitive to the price per minute (−1.62), women (1.87), people living in larger towns (1.64), and those with a higher income level (0.60) are less sensitive to the cost of the service. Our results are not directly comparable with those reported in the literature because, to the best of our knowledge, none of the previous studies analyzed the two components of the tariff separately: variable and fixed [30,41].

In line with our expectations, the second most important factor influencing the willingness to use an e-bike sharing service is the fixed part of the tariff (−1.81), although, as expected, it is much less relevant than the variable part. Again, respondents’ preferences are not homogeneous; in fact, the spread of the triangular distribution of the parameter is statistically significant (1.81) and, similarly to the results obtained for the variable part of the tariff, the less sensitive are people living in larger cities (0.38). This result is probably due to the fact that they use sharing services more extensively and frequently than those living in smaller municipalities and because sharing services are relatively more convenient for their mobility needs.

The third most important factor influencing the decision to use an e-bike sharing service is the extension of the area served (1.35), but respondents’ preferences are also very heterogeneous in this respect. A larger coverage area is more important for people living in the north and south of Italy than for those living in the centre of Italy (−0.34). This is probably due to the greater availability of complementary and substitute mobility services offered in the central macro area of the country. The extension of the area served is more important for men than for women (−0.19) and for those whose place of work or study is different from their place of residence (−0.14), which is in line with our expectations.

The availability of a large number of bicycles has a significant impact on the likelihood of using an e-bike sharing service (0.81), although the importance of this characteristic is lower than that of the service tariff and the extent of the area covered. This factor also shows that respondents’ preferences are not homogeneous, with people living in the centre of Italy being the most sensitive (0.49).

Operating range is an additional factor that influences the choice of e-bike sharing (0.06). Respondents’ preferences vary considerably according to age and place of residence. Older people are significantly more sensitive to the range of the bikes (0.0003). According to our results, increasing the range would significantly increase the number of older people using the service, improving their mobility options and social inclusion. In line with our expectations, range is more important for people living in small towns than in medium (−0.01) and large cities (−0.01). This is probably due to the longer distances that must be covered to reach the destinations, which are often located in places other than the place of residence. Respondents living in the central part of Italy (−0.01) are also less sensitive to this characteristic, probably due to the denser network of educational, professional, commercial, and tertiary activities located in this area.

In terms of the operating system, on the other hand, we found a marked heterogeneity, with half of the sample preferring the station-based system and the other half preferring the free-floating system. This is evidenced by the mean of the random parameter, which is not statistically significant, and the spread of the triangular distribution of the parameter, which is large and statistically significant (3.52). However, according to our results, people living in northern Italy prefer the station-based system (0.71). Our result is quite unexpected, since the free-floating system is more flexible. This result may be due to the possibility of knowing where to find the bicycles in advance, which reduces the perception of needing help finding the bicycle when needed. A second explanation could be related to the fact that in Italy there has been a significant increase in the number of incidents caused by shared vehicles abandoned in side streets and in the damage to urban amenities caused by abandoned shared vehicles [66]. Organizing the service through a station-based system could be perceived by residents as a way of reducing both phenomena. It is, however, a factor that needs to be studied further, as many operators are moving from station-based to free-floating services, which require much higher investments in terms of the number of bicycles and cost of repositioning, while users may be perfectly happy with the existing station-based services.

The design of the bicycles could be more relevant when choosing whether to use an e-bike sharing service. The respondents’ preferences are highly heterogeneous since the spread of the parameter distribution is statistically significant and large (1.66). Still, none of the two designs proposed are preferred on average by the sample. Indeed, the expected value of the parameter in not statistically significant. Being an actual bike sharing user (0.71) increases the likelihood of choosing a bike service over the opt-out alternative. Our result highlights that awareness of the service increases the likelihood of using it, so providers should invest in promotional campaigns, especially when launching a new service in a city that has never had a bike sharing system before. It also shows that those who have used the service before are satisfied with their experience and that this form of mobility successfully meets users’ needs.

The likelihood of not using e-bike sharing services, that is, of choosing the opt-out alternative, is not influenced, ceteris paribus, by gender, age, and the fact that the respondent’s place of work is the same as their place of residence, as these parameters are not statistically significant. Students (−0.94), more affluent people (−0.11), bicycle owners (−0.27), and people living in the north of Italy (−0.43) are less likely to choose the opt-out alternative, i.e., not to choose the hypothetical e-bike sharing options proposed. On the contrary, people living in rural areas (0.30) are more likely not to choose the e-bike sharing alternatives. These results are particularly important for defining the segment of potential users that operators should target when offering a service on the market.

4.4. Stated Preference on E-Bike Sharing vs. E-Scooter Sharing Use

In

Table 6. Random-parameter logit model: choice between e-bike sharing service, e-scooter sharing service, and not using sharing service (binary—answers to questions Q9–Q20).

	Coefficient	S.E.
Likelihood of choosing an e-scooter sharing service
Per minute tariff (hypothetical attribute, EUR, triangular mean)	−6.09 ***	0.48
Ts—per minute tariff (triangular spread)	6.09 ***	0.48
Per minute tariff * income (ordinal, 1 = EUR 0–499; ….8 = >EUR 3000) [Q54]	0.51 ***	0.16
Fixed tariff to unlock the vehicle (hypothetical attribute, EUR, triangular mean)	−0.89 ***	0.09
Ts—fixed tariff to unlock the vehicle (triangular spread)	0.89 ***	0.09
Extension of the area served (hypothetical attribute, ordinal, 1 = small, 2 = medium, 3 = large, triangular mean)	0.82 ***	0.06
Ts—extension of the area served (triangular spread)	0.82 ***	0.06
Vehicle availability (hypothetical attribute, ordinal, 1 = low, 2 = medium, 3 = high; triangular mean)	1.16 ***	0.10
Ts—vehicle availability (triangular spread)	1.16 ***	0.10
Vehicle availability * age (cardinal) [Q47]	−0.02 ***	0.004
Operating range (hypothetical attribute, km; triangular mean)	0.04 ***	0.004
Ts—operating range (triangular spread))	0.04 ***	0.004
Operating range * age (cardinal) [Q47]	−0.0009 ***	0.0002
Operating system (hypothetical attribute, 1 = station-based, 0 = free-floating; triangular mean)	0.45 ***	0.12
Ts—operating system (triangular spread)	2.86 ***	0.19
Actual scooter sharing users (binary, triangular mean) [Q21]	0.53 ***	0.15
Likelihood of choosing ane-bike sharing service
Per minute tariff (hypothetical attribute, EUR, triangular mean)	−1.71 ***	0.73
Ts—per minute tariff (triangular spread)	1.71 ***	0.73
Fixed tariff to unlock the vehicle (hypothetical attribute, EUR, triangular mean)	−0.45 ***	0.17
Ts—fixed tariff to unlock the vehicle (triangular spread)	0.45 ***	0.17
Fixed tariff to unlock the vehicle * income	0.13 ***	0.05
Extension of the area served (hypothetical attribute, ordinal, 1 = small, 2 = medium, 3 = large, triangular mean)	0.71 ***	0.06
Ts—extension of the area served (triangular spread)	0.71 ***	0.06
Vehicle availability (hypothetical attribute, ordinal, 1 = low, 2 = medium, 3 = high; triangular mean)	1.25 ***	0.07
Ts—vehicle availability (triangular spread)	1.25 ***	0.07
Operating range (hypothetical attribute, km; triangular mean)	0.01 ***	0.005
Ts—operating range (triangular spread)	0.01 ***	0.005
Operating system (hypothetical attribute, 1 = station-based, 0 = free-floating; triangular mean)	0.13	0.11
Ts—operating system (triangular spread)	2.30 ***	0.30
Likelihood of not using any sharing service
Woman (binary vs. man) [Q48]	−0.68 ***	0.10
Student (binary vs. other occupational status) [Q50]	−0.82 ***	0.12
Income (ordinal, 1 = EUR 0–499; ….8 = >EUR 3000) [Q54]	−0.08 **	0.04
N. of cars owned (cardinal) [Q41]	−0.13 ***	0.06
Bike owner (binary) [Q42]	−0.15 ***	0.04
Public transport frequency use (ordinal-0 = never; 1 = rarely; 2 = 1–4 times/1 per week; 3 = almost every day) [Q46]	−0.14 **	0.05
Residence area (cardinal, 1 = city centre, 2 = periphery, 3 = rural area) [Q53]	0.24 ***	0.08
North of Italy (binary vs. centre or south) [Q51]	0.41 ***	0.11
Centre of Italy (binary vs. north or south) [Q51]	0.46 **	0.21
ASC opt-out option	−1.15 ***	0.24
Chi-square (29) = 3100 prob Chi-square < 0.001 (N = 6364)
AIC = 10,852
Adj R2 = 0.22

Note: the demand code corresponding to the variable specified in the model is reported within the square brackets; (*) p-value ≤ 0.05; (**) p-value ≤ 0.01; (***) p-value ≤ 0.001.

, we report the results of the analysis of the stated preferences between an e-bike sharing service, an e-scooter sharing service, and the opt-out option (answers to questions Q27–Q38 in Appendix A).

When analyzing the preferences for the e-scooter sharing service, we find that the most important factor is the variable component of the tariff (−6.09). However, the preferences are heterogonous (the spread of the parameter distribution is statistically significant) and the sensitivity is lower for more affluent respondents (0.51). The fixed component of the tariff (−0.89), similar to what we found previously and in line with our expectations, is relatively less important in influencing the choice, although the cost of the service is one of the most important characteristics taken into account by respondents.

The second most important factor increasing the likelihood of choosing the e-scooter sharing service is the large availability of bicycles (1.16). However, preferences for this characteristic are not homogeneous and older respondents are less sensitive to it (0.02). Area coverage (0.82) and operating range (0.04) are also relevant factors in conditioning the likelihood of using the e-scooter sharing service, with older respondents being slightly less sensitive to operating range (−0.0009) than the rest of the sample. However, contrary to our expectations, but similarly to what was already found for e-bike sharing, the station-based operating system is preferred to the free-floating one (0.45), although the preferences are very heterogeneous, as evidenced by the large statistically significant spread of the parameter distribution.

Furthermore, respondents who had previously used an e-scooter sharing service were more likely, ceteris paribus, to use the service (0.53), in line with the results found by [47].

Regarding the likelihood of choosing the e-bike sharing service, we find the same ranking of factors that we observed for the e-scooter sharing service and that we have already illustrated in the comments to Table 5. The only difference is that there is no clear preference for the type of operating system. The preferences for all the factors analyzed are very heterogeneous since the spread of the triangular distribution of the random parameters is statistically significant.

Focusing on the factors that increase the likelihood of not using any of the proposed sharing services, we find that men (−0.68), people who are not students (−0.82), less affluent people (−0.08), respondents who do not own a car (−0.13) or a bicycle (−0.15), and people who do not use public transport (−0.14) are more likely to choose the opt-out alternative. Our results are partly consistent with those of [48,50], who found that sharing services can substitute for private cars and that people who use e-scooters, e-bikes, or public transport and young and highly educated people are more likely to use e-scooter sharing services. We also found that, ceteris paribus, people living in rural areas and in northern and central Italy are less likely to use both sharing services.

5. Conclusions

This paper examines citizens’ preferences for e-scooter and e-bike sharing systems. While traditional bike sharing services are essential for urban sustainability and are prevalent in European cities, there has recently been a significant increase in e-scooter sharing systems. Our study presents new empirical evidence on the current and potential demand for these shared micromobility services, highlighting the factors that influence their usage and the individuals’ preferences, including the characteristics of the service, the geographical area, and the town size.

To this aim, we collected and analyzed both revealed and stated preference data. Revealed preference results show that women, older individuals, and those living in larger households are less likely to use the existing bike sharing systems. However, the frequent use of public transport increases the likelihood of using a bike sharing service. Other important factors include the number of bikes available and the operating range. As for e-scooter sharing services, user-friendly parking facilities, the innovative and trendy nature of the service, and an extended operating range are key factors that increase the likelihood of use.

Stated preference results show that the most crucial factor affecting the likelihood of using an e-bike sharing service is the value of the per minute tariff. Other factors include the fixed part of the tariff, the extension of the area served, and the operating range. We also observed variations in these results across Italy, such as between the north and south or in large versus small urban areas. These differences underline the importance of considering the unique characteristics of these areas in empirical analyses, as findings from specific case studies may not apply to cities with different socio-economic and infrastructure profiles. The stated preference results for the e-scooter sharing services depict a similar structure of the preferences although the second most important factor beside the cost of the service is the number of vehicles followed by the size of the area served.

Our findings will assist local regulators in implementing specific urban policies and addressing effective and efficient private future transport operators’ investments and management. This is crucial because preferences for these systems vary across different urban contexts and population groups. In particular, both the revealed and stated preference results show that women and older people are less likely to use active shared transport. In order to increase the uptake among these population groups, it would be useful in the future to provide electric bicycles instead of conventional ones, which would reduce the physical effort required to travel and allow for the transport of shopping bags and children if necessary. An information campaign on the convenience, health benefits, and safety of using both sharing systems should be targeted at older people and women. The size of the urban area served should be extended from the main city centre to the more peripheral areas to better meet the needs of users and potential users. Preferences for the type of operating system are very heterogeneous. On the one hand, the free-floating system is more flexible than the station-based one; on the other hand, it is more prone to misuse of the vehicles, which are often abandoned on side streets, causing obstructions to pedestrians and affecting the amenity of the urban context. Residents of large and medium-sized cities show a higher propensity to use e-bike and e-scooter sharing services, so special efforts should be made in these urban contexts to increase the currently undersized supply. Much more future transportation investment should be made at the urban level, especially in the southern part of the country, to increase the density and extension of cycle lanes and to enforce traffic regulation that is more respectful of riders.

Our research has some limitations. The first is that it is based on the preferences of a sample that is relatively younger than the Italian population. This means that the findings we report should be extrapolated with caution to all segments of the Italian population. However, the bias related to the age of the respondents is less problematic than in other research contexts, since, according to the literature, the most promising part of the population that could realistically increase the use of active transport is made up of younger people who are physically better suited to riding a bicycle or scooter. The second limitation is that the data were collected during the COVID-19 pandemic, which may have changed respondents’ preferences over time, particularly for shared transport. Finally, the data were collected unevenly across Italian regions, with Lazio, Lombardia, and Friuli-Venezia Giulia over-represented compared to the others. Future lines of research therefore should include the development of a sampling strategy that will allow us to better analyze the preferences of the entire Italian population and of all Italian regions. A second wave of data collection could also be organized, as this would allow us to see if and how the COVID-19 pandemic might have affected preferences for active shared mobility.