香京julia种子在线播放

Nguyen-Phuoc et al. (11) found trust fully mediate the relationships between perceived booking app-related risks and satisfaction and loyalty.

The female passengers are 28.51% higher than that of male passengers who use ride-hailing services (10). Older ride-hailing passengers make more transit trips than others (12). Alemi et al. (13) found that passengers with higher educational and income levels were more likely to utilize these services, due to their familiarity with new technologies and other attributes of vehicle ownership. Murphy and Feigon (14) showed that the most frequent users of on-demand ride-hailing services were those from middle-income families with annual incomes between 50,000 USD and 75,000 USD. Those who have a strong inclination toward using their own vehicle tend to use ride-hailing services less frequently (15).

There is a close relationship between different building environments and passengers' ride-hailing frequency. For example, building density has a significant inhibitory effect on ride-hailing trips (16). Moreover, environmental consciousness plays an important role in ride-hailing frequency of use (17).

However, so far, there have been limited studies on the factors affecting the usage frequency of ride-hailing services in developing countries, especially the quantitative analysis of the potential impact of these factors. In addition, the continuous success of ride-hailing services is accompanied by various challenges, such as passengers' trust and loyalty, which are essential for these companies in terms of market share and profits. The concept of customer loyalty was first proposed by Guest (18), referring to customers' long-term psychological attachment to the company's products. Later, Boulding et al. (19) showed that loyalty is expressed in customers' higher willingness to recommend the company's supplies. Ride-hailing services have the potential to reduce traffic and emissions by reducing car ownership (20, 21). However, the impact of passengers' trust and loyalty on the frequency of ride-hailing usage has not been fully explored (22). Studies have shown that trust can increase people's intention to use such services (23, 24). Previous research has mainly focused on passengers' trust and loyalty prior to adoption (25). There are discrepancies between passengers' initial adoption and their actual usage frequency (26–28). For example, passengers' initial adoption can be altered over time (28). Thus, it is essential for ride-hailing companies to understand the influence of trust and loyalty on the frequency of using ride-hailing services in order to meet passengers' needs.

Additionally, due to the COVID-19 pandemic, another new challenge arises for ride-hailing: the social environment changed dramatically. After the outbreak of the COVID-19 pandemic, the World Health Organization (WHO) encourages people to maintain physical distance to avoid the epidemic spreading. In December 2019, after the outbreak of COVID-19, the Chinese government encouraged residents to minimize travel, for security reasons. The transportation services market has shrunk significantly, and ride-hailing is inevitably affected (29). In addition, safety concerns during the pandemic become the key consideration of transportation services (30). Passengers' behavioral patterns including the usage frequency of ride-hailing changed under the influence of COVID-19 pandemic (31). Therefore, to better understand the passengers' behavioral change, this study conducted a two-stage survey among the same group of passengers in pre and post COVID-19 pandemic.

Overall, this study attempts to addresses the following questions:

In order to answer these questions, this research proposes a theoretical model (as shown in Figure 1) based on Kotler (32) in order to construct and validate a conceptual framework to explain the usage frequency of ride-hailing. Specifically, the exploratory factor analysis (EFA) is conducted to develop the ordered probit and ordered logit models to study the affecting factors of the usage frequency of ride-hailing. Methods section introduces the data and methods. Results section introduces the models' construction details and analysis results. Discussion section discusses the findings. Finally, Conclusion section draws the conclusions.

The questionnaire survey was conducted through Sojump, a popular online survey platform in China, following a pre-test with 30 respondents to ensure its applicability. The survey was divided into two phases: the pre-epidemic phase (from December 2019 to January 2020, Phase-1), where COVID-19 was not yet widespread in China, and the post-epidemic phase (from August 2020 to September 2020, Phase-2), where the number of new cases per day was less than 20 and mostly imported from abroad, indicating the effective control of the epidemic. Figure 2 shows the newly confirmed cases of COVID-19 during the survey period in China.

This two-stage survey was conducted among the same group of participants. Initially, 800 questionnaires were collected in the Phase-1 survey. When the epidemic was under control, the Phase-2 survey was conducted, resulting in 694 questionnaires. After eliminating 24 ineffective questionnaires, 670 valid samples were recruited. The participants were asked to answer questions related to the usage of ride-hailing, including whether they had used ride-hailing, followed by social factors, psychological factors, personal factors and perception of the epidemic. The psychological factors and perception of the epidemic were measured using a five-level Likert scale, with “1” indicating “strongly disagree,” “3” indicating “neutral,” and “5” indicating “strongly agree”.

The descriptive statistics about the sample are presented in Table 1. Participants included both students and the working population, which accounted for 39.48 and 57.49%, respectively. Males constitute 61.38% and females represent 38.62%. According to the report released by Aurora Big Data, in China, most of the users of independent ride-hailing apps were male, and the proportion of male users of Shenzhou Special Car and Shouqi's ride-hailing apps has been close to 70%. Didi Chuxing has a relatively high proportion of female users of 41.1%. In addition, according to the report released by Aurora Big Data, Beijing, Hangzhou, Wuhan, Guangzhou, Dalian, Tianjin, Chengdu, Shenyang, Shenzhen, and Hefei have been the top ten most popular cities for ride-hailing in China. In our survey, participants from these ten cities accounting for 67% which indicate that the participants in our survey have a certain degree of representativeness.

In order to reduce the influence of individual choice differences on the results, subjects whose current city and urban residential location remained constant between the two phases of the study were chosen. It was also determined that the time gap between the two questionnaires was too short for personal preferences to change significantly. Additionally, Cronbach's alpha coefficient of the questionnaires was calculated to be 0.953 and 0.954 for the first and second phases respectively, indicating a high reliability. Furthermore, the KMO values of the first and second phases were 0.949 and 0.950 respectively, both of which were larger than 0.7, thus demonstrating the suitability of the questionnaires for further analysis using EFA.

Subsequently, EFA was employed to identify independent regression variables and calculate factor scores. After data cleaning and preprocessing, the number of factors influencing personal attitudes before the epidemic was reduced from 49 to 23, which were organized into three groups: Attitudes to Environmental Protection and Commuting (ATT), Social Life and Technology Use (SOC), and Passenger Trust and Loyalty (TRU). The latter refers to the passengers' loyalty to the ride-hailing service, not the third-party platforms providing the service. Similarly, the number of factors influencing personal attitudes after the epidemic was reduced from 53 to 29, which were organized into four groups: Attitudes to Environmental Protection and Commuting (ATT), Social Life and Technology Use (SOC), Passenger Trust and Loyalty (TRU), and Epidemic Impact (EPI). The results of the rotated component matrix before and after the epidemic are presented in Supplementary Tables A2, A3 in the Appendix, respectively, with factor loading coefficient values all greater than 0.5.

The ordered probit regression and an ordered logit regression were conducted on all the factors in the above analysis that affected the usage frequency of ride-hailing services in the pre-epidemic and post-epidemic stages. The regression results showed that in the pre-epidemic and post-epidemic stages, seven factors had significant effects on the frequency of use of the dependent variable. It is worth noting that in the post-epidemic stage, the EPI factor did not have a significant effect on the dependent variable. The comparison results of regression models before and after the epidemic are presented in Table 4.

The regression model can be explained based on the results of the ordered logit and ordered probit regression, which are shown in Table 2. The results showed that people's usage frequency of ride-hailing services was significantly affected by the following seven independent variables: ATT, SOC, TRU, AGE, INC, CAR, and LOC. The comparison of the regression coefficients before and after the epidemic shows that the influences of various factors on the dependent variable before and after the epidemic did not change significantly in the coefficients. However, after the epidemic, the influence of passengers' trust and loyalty on the usage frequency of ride-hailing significantly increased, while the influences of social life and technology significantly decreased.

The main reason why individuals' perception of the epidemic was not a significant affecting factor was that, in the second stage of the questionnaire, the epidemic situation in China had been effectively controlled, work and production had been resumed in an orderly manner in various regions, and travel restrictions had gradually weakened. In the long run, the authors believe that the usage frequency of ride-hailing services may no longer be affected by the epidemic in the future. However, the emergence of the epidemic has brought higher requirements for the safety of ride-hailing services, such as requiring daily disinfection and requiring passengers and drivers to wear masks. Therefore, in the second phase of the questionnaire, the influence of passengers' trust and loyalty on the usage frequency of ride-hailing increased significantly. Due to restrictions on travel during the epidemic, people's social and entertainment activities were reduced during the investigation phase, so the impact of social life and technology use on the usage frequency of ride-hailing in the post-epidemic phase was significantly weakened.

The goodness of fit of the model was tested. The test results of the two models before and after the epidemic are presented in Table 3. The result of the chi-square test shows that P(Sig.) was less than 0.001, which was statistically significant, illustrating that our model was meaningful as a whole. The values of R² and Log likelihood of the two models were similar, indicating that there was no obvious difference between the advantages and disadvantages of the two models.

The marginal effect of a previously fitted model was calculated by fixing the values of certain covariates and integrating over the remaining covariates. This effect demonstrated how the dependent variable would alter when a particular independent variable changed, with all other covariates held constant (35). To calculate the Average Marginal Effect (AME), the marginal effect of each variable was calculated for each observation, whilst taking into consideration any covariates, and then the average was determined. Based on the introduction to the regression model method given in results of ordered probit and ordered logit models section, the coefficient estimation of a variable in the regression-result table reflects the impact of the variable on the explained variables. However, the regression models considered in this study are non-linear. The line graphs of the marginal effects of all the factors are presented in Figures 3, 4. The predicted marginal values of seven independent variables before and after the epidemic are presented in Tables 4, 5. These values can be used to explore the influence of changes in independent variables on the changes in dependent variables and analyze and compare the size of the predicted marginal value of the dependent variable in different situations.

The positive marginal effect indicates that as the value of a factor increases, the possibility of consumers choosing to use the ride-hailing will also increase, and the greater the value is, the greater the possibility will be. As shown in Table 5 before the epidemic, the factor TRU “passengers' trust and loyalty” had a negative marginal effect on the frequency of use, i.e., less than once a month, 1–3 times a month, and 1–2 times a week, and had a positive marginal effect on the usage frequency, i.e., 3–4 times a week and more than five times a week, and the marginal effect for more than five times a week was the largest at 0.0647, indicating that when the other factors remained unchanged, for a one-unit increase in the passengers' trust, the possibility of consumers choosing to use the ride-hailing five or more times a week increased by 6.47%, and the possibility of choosing less than once a month decreased by 4.76%. After the epidemic, when other factors remained unchanged, for a one-unit increase in the passengers' trust, the probability that consumers would choose to use the ride-hailing five or more times a week increased by 7.12%, while the probability to use it less than once a month decreased by 5.07%. Thus, the more passengers trusted the service, the more frequently it was used, and vice versa.

Similarly, before and after the epidemic, the four factors, including the ATT, SOC, INC, and CAR, had a negative marginal effect on the frequency of use, i.e., less than once a month, 1–3 per month, and 1–2 times a week, and had a positive marginal effect on the usage frequency, i.e., 3–4 times a week and more than five times a week, and the factor “attitudes to environmental protection and commuting” after the epidemic had the largest marginal effect at 0.1093, indicating that when the other factors remained unchanged, for a one-unit increase in the attitudes to environmental protection and commuting, the possibility of consumers choosing to use the ride-hailing five times or more per week increased by 10.93%. The results also show that the more private cars people owned, the higher the usage frequency of ride-hailing would be. This may be due to the fact that people who own private cars are more inclined to travel by cars, and thus they have a higher acceptance of ride-hailing services. Also, when it is not possible to use a private car to travel, they are more willing to choose ride-hailing services rather than other traveling methods, such as public transportation.

Contrary to the above factors, both before and after the epidemic, the marginal effect of the age and residential location in the urban area on the frequency of use, i.e., less than once a month, 1–3 times a month, and 1–2 times a week, was positive, the marginal effect on the usage frequency, i.e., 3–4 times a week and more than five times a week, was negative. When the other factors remained unchanged, for a one-unit increase in the residential location in the urban area before the epidemic, the possibility of consumers choosing to use the ride-hailing five or more times a week decreased by 5.50%, and the possibility of choosing it less than once a month increased by 4.05%.

After the epidemic, when the other factors remained unchanged, for a one-unit increase in the location and convenience of car usage before the epidemic, the probability that consumers would choose to use the ride-hailing five or more times per week decreased by 5.87%, and the probability of choosing it once a month increased by 4.18%. This result shows that people who live outside the urban area and are farther away from the urban area use ride-hailing services less frequently.

The possible reason for this result can be that the farther away from the urban area people are, the less the supply of ride-hailing services may be. Also, compared to the urban areas, the convenience of obtaining the ride-hailing services in rural and suburban areas can be slightly worse, which can make people in that areas be more inclined not to use the ride-hailing or to use them less often. Furthermore, the older the person is, the lower the usage frequency of ride-hailing will be. Based on the descriptive statistics, users who had used ride-hailing services were mainly between the ages of 18 and 34, and the younger generation was the majority. This age group also accounted for the highest proportion of participants who had used the ride-hailing services. The authors believe that there are two possible reasons for this result: (1) people in this age group have higher travel needs, and (2) they are highly adaptable to technological changes and are more likely to accept new technologies.