Understanding the intention to use bike-sharing system: A case study in Xi’an, China

Bike-sharing is widely recognized as an eco-friendly mode of transportation and seen as one of the solutions to the problem of air pollution and congestion. With the emphasis on sustainability in transportation, bike-sharing systems is an emerging topic of urban transport and sustainable mobility related research. Existing studies mainly explored the factors affecting individuals’ initial intentions to start using a shared bicycle, but few looked at the likelihood that a user would continue using one This study proposed a structural equation model with bike-sharing purchase decision involvement as independent variable, bike-sharing willingness to use as dependent variable, traveler participation and traveler perceived value as intermediary variables by introducing the concepts of purchase decision involvement, customer participation and perceived value in consumer psychology and behavior. A survey on bike-sharing users in Xi’an was conducted online and offline, and 622 effective responses were collected. The research model was tested by Amos 24.0 and the empirical results showed that All influencing factors including bike-sharing’s purchase decision-making involvement, traveler participation and traveler’s perceived value are found to be significantly and positively associated with usage intention; traveler perceived value play a chain-mediating role between bike-sharing purchase decision involvement and usage intention; bike-sharing purchase decision involvement have indirect effects on traveler perceived value through traveler participation. The results of this study enrich the current research’s in the field of sharing economy, and it is certain guiding significant for how to obtain and maintain stable customers in bicycle-sharing industry.


Introduction
Bike sharing, as a new environmental-friendly travel mode, has attracted large numbers of travelers due to its flexibility and convenience. In today's construction of the low-carbon society, it has been promoted as an effective solution to the "last mile" problem for. At present, the research on bike-sharing mainly focuses on innovation of business management and profit models [1][2][3], legal system [4], influencing factors [5][6][7]and usage behavior [8,9]. However, the mechanism between the influencing factors and usage intention is seldom researched. In a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 their study of Litchi Bay, a historic block in China, that there is a positive correlation between tourist involvement and satisfaction [32].
According to previous studies, the degree of customer involvement can influence their behavioral intention. Under the Relationship Marketing Paradigm, some empirical studies have revealed that the degree of involvement is significantly positively correlated with customers' consumption intention. Namely, customers with high degree of involvement have strong consumption intention, and vice versa [33]. Customers with high degree of network involvement are willing to spend more time and cognitive efforts to search for information, compare differences, and consider about their choices and decisions. Furthermore, customer's loyalty behavior will occur when a certain degree of familiarity is reached [34]. In addition, the high degree of online involvement would lead to high cognition of online shopping usability [35]. Customer participation and perceived value have a significant chain mediating effect between the purchase decision involvement and purchase intention [36]. Similarly, self-driving tourism involvement exerts a significant positive effect on behavioral intention, whereas tourism involvement plays a partially mediating role between self-driving travel motivation and behavioral intention [37]. Travelers with a high degree of involvement tend to make their choice behavior rationally, while travelers with a low degree of involvement often make choice inertially [38].
To sum up, most previous literatures study the relationships between one or two variables of customer's involvement, participation, perceived value and consumer's purchase intention. Yet they seldom combine the three variables to explore the influencing mechanism with intention. In addition, the previous studies mainly focus on the relevant areas of commodity consumption, while the consumption intention of bike-sharing is rarely researched. Thus, this paper introduces the related concepts of purchase decision involvement, customer participation, as well as customer perceived value, and adopts structural equation model to analyze the interactions and influence mechanism between the three variables and the usage intention. It not only expands the application scope of consumer behavior theory, but also provides bikesharing enterprises with management and marketing strategies.
The remainder of this paper is structured as follows. Section 2 introduces the theoretical source of bike-sharing purchase decision involvement, traveler participation, traveler perceived value and the new definitions in this study. Section 3 puts forward the theoretical model and the hypothesis development. Section 4 takes Xi'an, Shaanxi province as an empirical study and presents the results of the descriptive statistics, reliability evaluation and factor analysis. Section 5. To conclude, the findings, implications, limitations and some possible directions for future research are discussed in Section 6.

Purchase decision involvement
In 1947, Sherif and Cantril first proposed the concept of involvement theory in the study of social judgement theory. The theory is mainly used to study "the problem of personal attitude in social events". However, the definitions are various in other fields. The definition proposed by Zaichkowsky in 1985 has been widely accepted by scholars, i.e. " Decision involvement means the degree of correlation to the target perception of individual's internal needs, value and interest" [39]. The concept better represents the involvement of an internal psychological state and intention with the resulting consumer behavior. Purchase decision involvement refers to the degree to which consumers are concerned about purchase decisions. It can be divided into product involvement, advertisement involvement and purchase decision involvement according to the object involvement of influence. Mittal clearly distinguished purchase decision involvement from product involvement, and defined purchase decision involvement as consumers' interest and degree of concern. At the same time, four consumer purchase decision involvement scales(PII, CPI, PDI, FCB) were developed through empirical methods [40].
Based on the existing studies, Bike-sharing purchase decision refers primarily to the degree of travelers ' concern to bike-sharing in this paper. It is a temporary concern for decision-making on bike-sharing choice under the combined effect of user's internal demand such as own needs, interests, values and external situations such as social environment. The design scale is mainly considered from two dimensions: personal values and community attributes.

2 Customer participation
Customer participation is usually defined as behavior in the process of service production or delivery [41,42]. Scholars have used different methods to measure the dimension of customer participation based on different perspectives. In 1985, Fisk and Sialkot defined customer participation as consumer effort and involvement (including mental and physical participation). Sialkot believed that customer participation can be divided into three aspects input: mental input, physical input and emotional input. The cognitive ability of customers to obtain and process information is regarded as mental input, customers' consumption of tangible objects and physical fitness are regarded as physical input, and customers' emotional intensity is regarded as emotional input [43]. Kelley et al. (1990) [44] believed that certain service transactions must be completed by providing certain information or making efforts. Cermak, File, and Prince [45] pointed out that customer participation is a behavior related to service delivery, and was related with service specifications and service delivery. It is a customer's personal activity, the importance of single association or commodity to consumers is customer input.
In this paper, customer participation means traveler participation. It refers to the specific behavior of travelers spending time, energy, intelligence and emotion when using bike-sharing. It can be described in detail as: whether to download bike-sharing Apps, buy a bike-sharing rental card, pay attention to preferential activities, and such associated behaviors.

3 Perceived value
Perceived value refers to subjective evaluation of the utility of a product or service, which is obtained by subtracting the cost of acquiring the product or service from the perceived benefit of customer. In a general sense, perceived value is different from the objective value of the product or service, which reflects the specific cognitive value of customer [46]. Customer perceived value (CPV) is usually considered as a multi-dimensional concept. According to the theory of consumption value, SHETH et al. [47] put forward the five customer perceived value dimension, which included the functional value related to the practical function of product, the social value related to social image, the emotional value related to feeling of customer, the cognitive value related to customer's curiosity about new things, and the situational value in a specific environment. The contribution of this theory lies in its practicality and conciseness. It is easier for people to understand the connotation of customer perceived value through dimensionality division. SWEENEY et al. [48] believed that the price and quality are both functional value factors, but their contribution to customer perceived value should be measured separately. Therefore, on the basis of SHETH's theory, the function value is further divided into two sub-dimensions, namely quality factor and price factor. The PERVAL scale of customer perceived value is verified through empirical research on durable goods, including function value quality factor, function value price factor, social value and emotional value. Liu et al. divided consumers' perception value of electric vehicles into functional value, emotional value, social value and cognitive value [49].
To sum up, the perceived value in this study refers to the riders' subjective evaluation of the emotional value, functional value and safety value based on the trade-off between travel gains and losses.

hypothesis
3.1.1 Bike-sharing purchase decision involvement and traveler participation. Involvement is a psychological state variable, focusing on the individual's attitude, such as the degree of attention, self-relevance, etc. Engagement is a behavioral variable, focusing on customer behavior. Involvement is the antecedent variable of information search, which belongs to customer participation behavior [50]. Customers with high network involvement are more willing to give more time and cognitive effort to search for information, compare differences and make decisions. Consumer involvement is an antecedent variable of customer participation, which has a positive impact on the four dimensions of customer participation, including preparation, information exchange, cooperative behavior and interpersonal interaction [51]. This paper attempts to analyze and verify the effect of bike-sharing purchase decision involvement on traveler participation and proposes the following hypothesis: H1: Bike-sharing purchase decision involvement has a positive impact on traveler participation.

Bike-sharing purchase decision involvement, traveler perceived value and usage intention.
An empirical study showed that tourists involvement has a positive impact on perceived value of tourist attractions and plays a mediating role in the influence of motivation on tourists' perceived value [52]. Campbell believed that the higher the consumer's product involvement was, the stronger the willingness to purchase was [53]. In the field of organic food consumption in Taiwan, it has been verified that, consumer involvement has positive impact on purchase intention, and involvement played an intermediary role between purchase motivation and purchase intention [54]. The same conclusion had been obtained in the field of new energy vehicles [55]. Therefore, the hypothesis are proposed: H2: Bike-sharing purchase decision involvement has a positive impact on traveler perceived value.
H3: Bike-sharing purchase decision involvement has a positive impact on intention of bikesharing usage.

Traveler participation and traveler perceived value.
The impact of customer participation on perceived value has been confirmed by many scholars. Under the condition of service success, the higher the degree of customer participation is, the greater the perceived value has [56]. Social network users' participation behavior such as browsing, interaction and creation has a significant positive impact on users' perceived value [57]. Therefore, the following hypothesis has been proposed in this paper: H4: Traveler participation has a positive impact on traveler perceived value.

Traveler participation and intention of bike-sharing usage.
An empirical study on real estate marketing, customer participation has a significant positive impact on purchase intention [58]. In a B2B business model of the building materials and design industry, customer participation had a significant positive effect on traveler perceived value and an indirect effect on behavioral intention [59]. A study on consumption behavior of e-coupons behavior verified traveler perceived value plays a mediating effect on customer participation and purchase intention [60]. Therefore, the following hypothesis is proposed: H5: Traveler participation has a positive impact on intention of bike-sharing usage.

Traveler perceived value and intention of bike-sharing usage.
Many studies have shown that perceived value is an important antecedent variable of behavior intention. A study on the Malaysian housing market, it was found that functional value, social value, relationship value, situational value and cognitive value have significant positive effects on purchase intention [61]. In the study of revisiting tourist destinations, when a tourist with a high perceived value in a tourist destinations tend to revisit or actively recommend behaviors in the future [62]. Customer perceived value has a significant impact on consumer behavioral intention in the sharing economy [63]. There is a significant positive correlation between consumer perceived values and purchase intentions in online shopping. Taking clothing brands as the research object, traveler perceived value, network interaction and brand relationship have significant positive effects on consumer behavioral intention [64]. Therefore, the following hypothesis is proposed: H6: Traveler perceived value has a positive impact on the intention of bike-sharing usage.

The chain-mediating role of traveler perceived value and traveler participation.
Traveler purchase decision involvement, participation, perceived value and intention of bikesharing usage conform to the time sequence of the consumer decision-making process model. Involvement is the prerequisite and starting point of the purchase decision-making process, the level of purchase decision involvement will cause differences in a series of subsequent behaviors [65]. That is, the higher involvement in the decision-making of travel mode selection, the clearer the perceived value of travelers have, it can promote the willingness to use shared bicycles. Therefore, four variables of bike-sharing purchase decision involvement, traveler participation, traveler perceived value and intention of bike-sharing usage are included in this study. On the basis of Hypotheses H1 to H6, the following hypothesis are proposed: H7: Bike-sharing purchase decision Involvement affects usage intention through the mediating role of traveler participation and traveler perceived value.
H8: bike-sharing purchase decision involvement have indirect effects on traveler perceived value through traveler participation.

Structural equation model
The Structural Equation Model(SEM) is a quantitative research method in contemporary social, psychological, and behavioral sciences. It is a method of establishing, estimating and testing causality. It includes methods such as multiple regression, factor analysis, and path analysis. SEM can clearly analyze the effect of individual indicators on the overall and the interrelationship between individual indicators. SEM includes two parts: measurement model and structural model. Measurement model can be expressed as: Formula (1) is the measurement model of exogenous variables. Formula (2) is the measurement model of endogenous variables.
Structural model can be expressed as: Formula (3) means the structural relationship of potential exogenous variables and potential endogenous variables.
Where, X represents the observable variable vector, Λ X and Λ y represent the factor load matrix, ξ, η represent the vector of latent variables (common factors), and δ ε represent the vector of errors (unique factors).

Conceptual model
Based on the above hypotheses, this paper proposes a bike-sharing usage intention model based on SEM S1 Fig including the independent variables bike-sharing purchase decision involvement, the dependent variable usage intention, and the mediating variables traveler participation and traveler perceived value.

Participants and procedure
A 36-item questionnaire was designed to collect data from a sample of residents of Xi'an. The questionnaire consisted of two sections: the first part collected sociodemographic of the interviewees, including gender, age, occupation, income, education, private car ownership and other basic characteristics information. The second part measured residents' willingness to use bike-sharing by 5-point Likert scale, which includes some descriptive views of residents' personal values, community influence, perceived value, etc. while the measurement of personal values, community influence and usage intention was adapted from the literature [10,66,67], traveler participation and perceived value was based on the scales by Liang m et al [68][69][70].
Before the formal investigation, we conducted an online survey with written informed consent, obtaining 203 questionnaires in the pre-survey. After conducting a content analysis of the questionnaires and consulting with survey experts, certain items was modified and improved within the questionnaire, a formal questionnaire comprised 6 variables and 24 measurement items. The questionnaire statements for measuring usage factors are shown in Table 1.
It should be noted that, in the first page of the questionnaire, we introduced our research purpose and secured their written informed consent to participate in this academic research. Non-adult participants were allowed to answer this questionnaire only after getting their parent or guardian's written informed consent. The Chang'an University Ethics Committee approved the protocol and informed consent forms for this study.

Data collection
The questionnaires were collected though the combination of online and offline survey. The online survey questionnaires are distributed by posting a microtask on the www.wjx.cn, one of the largest free research platforms in China. The offline paper-based survey questionnaires, we randomly distributed them to the residents in Xi'an, For the specific survey site, we divided the Xi'an study region into seven urban districts(Beilin, Lianhu, Baqiao, Weiyang, Yanta, Xincheng, Chang'an) and 35 collection blocks. Both the online and offline respondents received 2RMB each for participation.
A total of 650 questionnaires were collected, 500 field survey questionnaires and 150 online questionnaires, yielding 622 valid questionnaires, with an effective response rate of 95.7% (see Table 1). The demographic characteristics of the sample, including gender, age, occupation, education, and income, were recorded. Of the respondents, 52.1% (n = 324) were male and the others female (n = 298). The age grouping with the greatest number of responses was 19-30years of age(47.4%),followedby31-45years of age(44.8%). As for education and income, the greater part of the participants hold an undergraduate degree (65.9%) and 13.5% were graduate students. Further, the largest response group reported a monthly income below CNY3000 (41.6%), and the next group represented 26.4% reported an income 3001-6000 yuan, high income group(monthly income above CNY10000) accounts for 10.0%. Our sample's age distribution is consistent with the results of 2018 China shared Bicycle Industry Research Report [71]. Thus, our sample can be representative of the population of interest. The respondents' characteristics can be seen in Table 2.

Reliability and validity
SPSS version26.0 software was used to test the quality of the data. The entire questionnaire's Cronbach's α reliability coefficient was 0.927, While the four domains, including bike-sharing  [72]. The result showed a high degree of internal consistency and test-retest reliability, as shown in Table 3. An Exploratory factor analysis (EFA) was performed with the results showing clear evidence of a 6-factor structure with eigen values over Kaiser's criterion of 1.00. The newly generated 6-factor model collectively explained 70.0% of the variance, the load of each item on its factor is much more than 0.5, while the load on other factors is much less than 0.5, this shows each factor is unique [73]. The Kaiser-Meyer-Olkin (KMO) value was 0.916 and this measure verified the sampling adequacy for the analysis. Bartlett's test of sphericity indicated that the correlations between items were sufficiently large for the EFA analysis (p < 0.001). The scree plot test was used to determine the number of factors to retain and rotate with the test results

PLOS ONE
confirming the appropriateness of a 6-factor solution. The EFA yielded an 24-item measure with a 6-factor structure. The 6 factors were labeled as Bike-sharing purchase decision involvement, Traveler participation, Perceived emotional value, Perceived functional value, Perceived security value and Usage intention. Reliability of the items in each of the four TRAS subscales was tested using reliability analysis. Table 4 shows the factor loadings after rotation using the Promax method.

Direct effect test
AMOS structural equation model was used to verify the hypothesis. The common parameter estimation method of structural equation model is maximum likelihood or generalized least square method [74]. Both methods require the data is subject to multivariate normal distribution. Therefore, the normal distribution of scale data should be checked before model detection. The results of normal distribution test showed that the multivariate kurtosis value of the sample was 68.293, as shown in Table 5. This indicates that the sample data are significantly non-normal. One of the methods to solve the non-normal distribution of sample data is bootstrap method [75]. Bollen-stine bootstrapping procedure can be used to estimate the standard error and correct the deviation of model fitting data [76]. In addition, Nevitt and Hancock suggest that boot value analysis is more suitable for samples with a sample size of over 200 [77]. The empirical study focused on 622 samples tested by the model and the non-normal distribution of samples. Therefore, Bollen-Stine bootstrapping procedure was adopted in this study. The Table 6  It can be seen from Table 7 the chi square ratio is 3.866, which is higher than the desired cutoff value of 3.000. The root mean squared error of approximation (RMSEA) is 0.068, which is lower than 0.080. The Normed Fit Index (NFI) is 0.891, and the Comparative Fit Index (CFI) is 0.876, which is lower than 0.9. The Tucker-Lewis Index(TLI) is 0.905, which are greater than 0.900. To summarize, the structural model should be improved.

PLOS ONE
According to the MI index, four measuring items, including Q2-4, Q3-4, Q3-5 and Q4-4 were deleted. The optimized chi square ratio is 2.894, which is below the desired cutoff value of 3.000. The root mean squared error of approximation (RMSEA) is 0.055, which is lower than 0.080. The Normed Fit Index (NFI) is 0.932, and the Comparative Fit Index (CFI) is 0.954, and the Tucker-Lewis Index(TLI) is 0.946, both of which are greater than 0.900. To summarize, the structural model has a good fit. The final model path diagram and relevant index are shown in S3 Fig and Table 8.
The Table 9 depicts the results of hypotheses test. all of the path coefficients had acceptable statistical significance level. The p-values of the effects of bike-sharing purchase decision involvement on traveler participation, traveler perceived value, usage intention, the effects of traveler participation on Traveler perceived value and the effects of traveler perceived value on usage intention are all less than 0.001. Hence, H1, H2, H3,H4 and H6 were supported. Traveler participation showed lower significance on the Usage intention (p<0.05), but it also can be said that this effect was significant. That is to say, H5 was supported.

Intermediary effect test
Next, the mediation effect of traveler perceived value was analyzed in the SEM model and present the result in Table 10. Among several methods of testing the mediation effect, the causal steps approach and the Sobel test have been used widely but have also been criticized by researchers [78,79]. So, we adopt a bootstrapping method that results in more accurate confidence intervals for indirect effects [80]. We use AMOS 24 to calculate 5000 bootstrap samples. If zero is not between the lower and upper bound of the bias-corrected percentile and percentile, the mediation effect is shown to be significant at a confidence interval of 95%. As presented in Table 9, bike-sharing purchase decision involvement have indirect effects on usage intention through traveler perceived value, meanwhile, bike-sharing purchase decision involvement have indirect effects on traveler perceived value through traveler participation, and their indirect effect were 0.286 and 0.141 respectively, the 95% confidence interval for the bias-corrected percentile and percentile does not contain zero. So, hypotheses H7 and H8 were

PLOS ONE
supported. The direct effect of bike-sharing purchase decision involvement have on usage intention is 0.262 and the direct effect of bike-sharing purchase decision involvement have on traveler perceived value is 0.271, both of the 95% confidence interval for the bias-corrected percentile and percentile does not contain zero. In summary, the above results indicated that traveler perceived value and traveler participation played partial mediating roles.

Discussion
The empirical results showed that the twenty main questionnaire items from bike-sharing purchase decision involvement, traveler participation and traveler perceived value had the influence on usage intention to different degrees. The specific conclusions were obtained as follows: All influencing factors including bike-sharing's purchase decision-making involvement, traveler participation and traveler's perceived value are found to be significantly and positively associated with usage intention. That is, bike-sharing's purchase decision-making involvement, traveler participation and traveler's perceived value are significant predictors of usage intention. Hence, Hypotheses H1, H2, H3, H4, H5 and H6 are supported. Seen from the degree of influence, traveler perceived value had the greatest effect on usage intention (in Table 9), it reflected the fitness, leisure, comfort and safety aspects of bike-sharing had played a great role in intention of bike-sharing usage. It is also one of the reasons for the popularity of bike-sharing and is consistent with existing research conclusions [81]. Then, Bicycle-sharing purchase decision involvement not only has a direct impact on willingness to use, but also has an indirect impact on willingness to use through intermediary variables such as traveler participation and traveler perceived value. Among the dual influences, the indirect effect is dominant, followed by the direct effect, which reflects the usage intention of bike-sharing is positively influenced by personal values, reference groups and others, and the social impact of bike-sharing that conforms to mainstream values of modern energy conservation, environmental protection and low-carbon travel will result in herd consumption. This finding is consistent with previous studies on personal values and community influence [82]. Finally,

PLOS ONE
the effect of traveler participation on usage intention reflects the bike-sharing as a new business model innovation have attraction with different levels for travelers due to different operating procedures, membership systems, preferential information and other activities of operators, which is consistent with the study of the willingness to use e-coupons [83].
The main contributions of this study come from three aspects: first of all, the influential factors of bike-sharing usage intention have been revealed from the three-dimensional perspective of bike-sharing purchase decision involvement, traveler participation and traveler perceived value, which provides a more comprehensive perspective on psychology and behavior for bike-sharing usage. Secondly, this paper proposes a theoretical model of bike-sharing usage intention based on the consumer decision-making process model that enriches empirical research on the bicycle-sharing industry and broadens the scope of application of consumer behavior theory. Finally, these results can guide the management practices of bikesharing enterprises to improve the sustainable utilization rate by improving the user experience through convenience, environmental protection, low price, etc.
This study also has some limitations, and other factors such as residents' travel characteristics, weather, as well as the built environment should be considered in future research. Furthermore, the sample size can be expanded and the groups can be divided, which will help bike-sharing enterprises to understand intention differences between different groups and put forward feasible strategy and targeted tactics.