The Impact of Travel Scenarios and Perceptions on Choice Behavior towards Multi-Forms of Ride-Hailing Services: Case of Nanjing, China

Abstract

The travel behavior of urban residents has gradually changed in response to the widespread adoption of ride-hailing services. This paper explores the travel mode choices made by individuals utilizing multiple forms of ride-hailing services. Eight scenarios were established, which considered combinations of activity types (commute or recreation), travel periods (peak or off-peak), and price levels (discounted or normal rates for ride-hailing). Moreover, socio-psychological variables such as perceived value, behavioral intention, and subjective norm were integrated into the analysis. The findings reveal that consumers of ride-hailing services generally exhibit characteristics such as being younger in age, having higher income, lack of car ownership, and having greater experience in using ride-hailing services. Furthermore, the inclusion of socio-psychological variables significantly improved the model’s fitness. Travelers exhibit a preference for ride-hailing services in scenarios involving recreational activities, normal travel periods, and discounted ride-hailing prices. In conclusion, this study sheds light on the evolving travel behavior of urban residents in light of the widespread availability of ride-hailing services. The incorporation of socio-psychological factors is essential in comprehending and predicting travel mode choices. The insights derived from this research contribute to a nuanced understanding of the factors influencing the adoption of and preference for ride-hailing services among urban commuters.

1. Introduction

With the rapid development of mobile Internet, many kinds of sharing economies have emerged into the lives of urban residents. As a significant supplement of electronic commerce, ride-hailing platforms connect travelers with drivers to provide transportation services through information networks [1,2]. Unlike taxi services, ride-hailing platforms allow private car owners or drivers to provide services for travelers. Currently, Uber is one of the most popular and largest ride-hailing platforms, and it has already promoted its services around the world [3]. There are some regional companies in the market as well, for instance, DiDi Chuxing in China. In past few years, ride-hailing services have gradually become an indispensable element of urban transportation systems [4,5,6,7,8].

With the support of mobile Internet, new features related to ride-hailing services have a significant effect on travelers’ travel behavior. Consequently, travelers’ behavior regarding ride-hailing services may differ from traditional transportation services [9,10,11]. Specifically, travelers interact with each other via online social networks more often nowadays owing to the expansion of mobile Internet. In addition, extant research demonstrates that ride-hailing services have already changed travelers’ behavior, especially regarding the choice of taxi services [12,13,14]. The popularization of such new travel modes has improved the travel experience of urban residents and demanded lower prices than traditional taxi services. However, it has aroused some controversies simultaneously, such as traffic congestion, crime, legitimacy issues [15,16], etc. Specifically, in the context of newly emerging transportation services, it is important to gain insight into the mode choice behavior of ride-hailing services.

There is no doubt that ride-hailing services have aroused increasing attention in many research areas, such as consumer characteristics, competition with the taxi industry [14,17], impact on urban transportation [18], carbon emission [19,20], etc. However, the understanding of consumer behavior and the impact of ride-hailing services on urban transportation is still insufficient. In existing research, consumers’ behavior related to ride-hailing services mainly includes their behavioral intentions towards such services [10,21,22]. Furthermore, some studies have further examined travelers’ choices behavior regarding ride-hailing services [13,14,23]. Nevertheless, little research has investigated travelers’ choice behavior based on multiple forms of ride-hailing services, which may lead to a misunderstanding on consumer behavior related to ride-hailing services. As for the factors considered, more variables related to travel scenarios, which can reflect travelers’ decision behavior, need to be introduced. Although the choice behavior of ride-hailing services has been studied based on specific situations, the combination of more scenario factors should be further investigated. Furthermore, the travel modes of ride-hailing services have been studied separately in most research, while the interactions between ride-hailing services and other travel modes have been overlooked. Hence, it is still necessary to examine travelers’ choice behavior regarding ride-hailing services and its impact on public transportation. Therefore, the aim of this study is to answer the following questions.

(1)

What is the difference in travelers’ choice behavior towards multi-forms of ride-hailing services?

(2)

What is the pattern of travelers’ choice behavior under various scenarios?

(3)

How do socio-psychological factors affect travelers’ choice behavior towards ride-hailing services?

In this paper, the travel modes for ride-hailing services include fast rides, tailored rides, and carpools. Among them, fast rides are cheaper, and the quality or brand of the car is standard, while for tailored rides, the quality or brand of the car is better, and the price is higher. For carpools, the quality or brand of the car is similar to that of fast ride, but the price level is cheaper than that of fast rides. The aim of this research is to investigate the travel mode choice behavior of ride-hailing services. Firstly, the factors that affect travelers’ choice behavior are summarized as socio-demographic factors, usage experience, socio-psychology variables, and scenario factors. Secondly, two models are constructed to examine the effect of related factors on the travel mode choice behavior regarding ride-hailing services. In the first model, there were no socio-psychology variables, while these were considered in the second model. To investigate the socio-psychological impact on the choice behavior of ride-hailing services, value (PV), subjective norm (SN), and behavioral intention (BI) were introduced. PV can be defined as the overall assessment or evaluation that a consumer makes regarding the benefits they receive from a product, service, or experience [24]. SN refers to an individual’s perception of the expectations of significant others, such as family, friends, etc. [25]. BI can be defined as an individual’s willingness to perform a specific behavior [25]. Thirdly, travelers’ choice behaviors regarding public transportation, taxis, and cars were further investigated and compared in different travel scenarios. In addition, travel scenarios were designed according to activity, period, and price level.

This study can provide some contributions to existing research. Firstly, in contrast to previous studies that have frequently concentrated on ride-hailing services as a collective entity, this research examines the choice behavior across a multitude of forms of ride-hailing services, with intricates distinctions in service quality, pricing, and user preferences within the ride-hailing services. Secondly, by considering travel scenarios in greater depth, the research extends beyond the conventional variables to offer a more comprehensive insight into decision making in the context of ride-hailing. Thirdly, this study employs two distinct models, without and with socio-psychological variables, to provide a comparative perspective on the impact of these variables, which enables a further understanding of the psychological effects regarding ride-hailing services. In addition, the investigation of the effects of scenarios and socio-psychological factors on travelers’ behavior in this study further sheds light on the consumer behavior related to electronic commerce.

The rest of this paper is organized as follows. In Section 2, the research design method and variables are illustrated; then, the multinomial probit models are constructed. The empirical results are presented and discussed in Section 3. Finally, in Section 4, conclusions and future research directions are described.

2. Literature Review

2.1. Factors Affecting the Use of Ride-Hailing Services

A considerable number of researchers have conducted extensive studies on the behavior of travelers utilizing ride-hailing services from various viewpoints. From extant research on the adoption of ride-hailing services, consumers who use ride-hailing services are normally younger, well-educated individuals with higher income [4,26,27,28,29]. In addition, females and males show different attitudes towards ride-hailing services. For example, [30] indicates that for females, security concerns may influence their behavior with ride-hailing services. Moreover, users who use ride-hailing tend to have a lower rate of car ownership [15,31].

The majority of current studies concentrate on behavioral intentions, as evidenced by [22,32,33,34,35]. Furthermore, some studies have concentrated on specific selection behaviors or discontinuation behaviors [15]. For instance, [36] investigated the factors influencing behavioral intentions to continue using ride-hailing services by integrating the Diffusion of Innovations Theory (DIT) and the Technology Acceptance Model (TAM). The study cited in [33] examined the key factors influencing travelers’ behavioral attitudes towards ride-hailing services in Egypt and further confirmed the relationship between behavioral attitudes and behavioral intentions. However, there has been a paucity of research investigating the choice behavior of ride-hailing services.

Generally, the main factors included in current studies are socio-demographic information and attitude. Furthermore, extant literature has demonstrated that factors regarding consumers’ perception are substantial factors for studying consumers’ behavior [24,25,37,38]. With the case of DiDi Chuxing, [15] studied the effect of risk perception on travelers’ intention to discontinue use behavior. This shows that perceived risk has a negative effect on travelers’ discontinuance behavior through variables of trust and attitude. In addition, [38] studied consumers’ attitudes toward ride-hailing services through the integration of the Theory of Planned Behavior (TPB) and TAM.

In addition to research on behavioral intention regarding the use of ride-hailing services, existing studies have also focused on travelers’ choice behavior. The utilization of ride-hailing services is typically associated with travelers’ activities. The study cited in [39] analyzed the travel behavior of ride-hailing users based on activities and classified the users through variables such as activity purpose, age, and work status. Concurrently, travelers may also exhibit certain preferences when confronted with specific travel modes in ride-hailing services. For instance, [40] investigated the factors influencing travelers’ selection of distinct types of ride-hailing services, demonstrating that the ownership of a private vehicle may influence travelers’ choice behavior. Furthermore, ride-hailing services exhibit a complementary relationship with other transport modes within urban areas. The study cited in [23] conducted interviews with 597 ride-hailing users in Chengdu to investigate the impact of ride-hailing services on travel frequency and travel mode choice. The study demonstrated that ride-hailing services exert a complementary effect on public transport travel.

Moreover, the usage of ride-hailing services has a significant impact on the choice of public transport [4,12,18]. Based on a multinomial linear regression analysis in the case of Las Vegas, [17] argued that ride-hailing services have a negative effect on the usage of taxicabs. Interestingly, [18] suggested that ride-hailing is not only in competition with, but also complementing, public transport. With the expansion of ride-hailing services, the prices of taxis have decreased and more affordable travel modes have come into existence [41,42]. Further, the choice behavior of travelers may change in different travel scenarios as according to travel period, weather, or price level. From various literature, it is indicated that consumer behavior depends on factors related to the scenario [15,28,43]. Overall, it is possible to study travel mode choice behavior for ride-hailing services based on a discrete choice model.

2.2. Hypothesis

Based on the existing literature, several hypotheses can be proposed. This study investigates the choice behavior regarding multiple forms of ride-hailing services considering socio-demographic information, car ownership, ride-hailing usage frequency, travel scenarios, and socio-psychological factors. Regarding the impact of socio-demographic factors, existing research indicates that ride-hailing users tend to be younger and well-educated, with distinct behaviors between females and males [4,28,30,44], providing support for this study. Further, although the prices of ride-hailing services are lower than those of taxis, they are still much higher than the prices of public transportation. Hence, travelers with higher income levels tend to adopt such services [14,27]. Therefore, the following hypothesis is proposed.

H1a: 

The preferences for multiple forms of ride-hailing services differ between females and males.

H1b: 

Ride-hailing users tend to be younger.

H1c: 

Ride-hailing users tend to have higher income levels.

For car ownership, it is assumed that travelers who own a private car are less likely to utilize ride-hailing services for regular commuting purposes [15,31]. This suggests that car owners perceive ride-hailing as a supplementary mode of transport for specific scenarios, rather than a substitute for private vehicle usage. Thus, H2 can be proposed.

H2: 

Travelers who own a private car will have a decreased probability of choosing ride-hailing services and public transportation.

Frequent users of ride-hailing services may exhibit higher levels of dependency on such services [27]. This suggests that the formation of ride-hailing habits increase with usage frequency, which in turn leads to a reduced likelihood of switching to other modes. Furthermore, if travelers can access ride-hailing services more easily, then their choice behavior with such services increases. Hence, H3a and H3b can be proposed.

H3a: 

Travelers who use ride-hailing services more frequently will be more inclined to choose such services.

H3b: 

Travelers will choose ride-hailing services more frequently if they can easily access such services.

Travel scenarios may potentially affect choice behavior regarding ride-hailing services [45,46,47]. This research considers the factors of travel activity, travel period, and price level. It is hypothesized that ride-hailing is preferred over other transport modes for recreational activities rather than commuting. During peak travel periods, when transport is congested, demand for ride-hailing services is expected to decrease. Additionally, travelers are more likely to choose ride-hailing when discounted rates are available. Thus, H4a, H4b, and H4c are proposed as follows.

H4a: 

Travelers are more likely to choose ride-hailing services for recreational activities.

H4b: 

Travelers are more likely to choose ride-hailing services during off-peak periods.

H4c: 

Travelers are more likely to choose ride-hailing services with a price discount.

Existing studies have shown that travelers’ behavior related to ride-hailing services is influenced by socio-psychological factors [21,22,32]. The perceived value of ride-hailing, which encompasses convenience and personalized service, has a positive influence on the choice of ride-hailing over other modes of transport, particularly among individuals who place high value on time and comfort. Subjective norms also play a role. Individuals whose social circles frequently use ride-hailing are more likely to adopt these services themselves, driven by social conformity and shared experiences. Regarding behavioral intention (BI), the intention to engage in ride-hailing behavior is significantly predicted by past positive experiences. Thus, this study proposes H5a, H5b, and H5c.

H5a: 

Travelers’ choice behavior of ride-hailing services is influenced by behavioral intention.

H5b: 

Travelers’ choice behavior of ride-hailing services is influenced by perceived value.

H5c: 

Travelers’ choice behavior of ride-hailing services is influenced by social norms.

2.3. Summary

Previous studies have examined travelers’ behavior regarding the use of ride-hailing services. The willingness to use and choice behavior regarding ride-hailing services has generated a substantial body of research results. However, the majority of studies have only considered general ride-hailing services, with very little literature focusing on travelers’ choice behavior for specific types of ride-hailing services. Furthermore, existing studies are less likely to consider the effect of multiple travel modes when analyzing their choice behavior regarding ride-hailing services. Existing studies tend to focus solely on ride-hailing services themselves when exploring travelers’ choice behavior, thereby ignoring their combination with other travel modes. This limitation becomes particularly evident in the context of ride-hailing services, which emphasizes the integration of different modes of transport to provide users with a more convenient, efficient, and personalized travel experience. Consequently, integrating multiple modes of travel in the analysis of ride-hailing services represents a significant innovation and practical approach, enabling a more comprehensive consideration of the impacts of multiple modes of travel and revealing more accurate behavioral characteristics in the context of ride-hailing services.

3. Methods

This section outlines the research design methods and the process of data collection, then provides definitions of the variables and presents a construction method of a multinomial probit model.

3.1. The Data

The survey was conducted in Nanjing, China, using paid services from Sojump, a questionnaire platform. This platform has a large user base and has been shown to be an efficient method for data collection [15,32]. The target population for this research consisted of residents familiar with ride-hailing in Nanjing. Five transportation researchers were invited to revise the survey design. A pretest of the questionnaire was conducted with over 50 participants to ensure clarity and relevance, with adjustments made based on feedback. A bonus of CNY 10 was provided to attract more respondents, and the probability of obtaining this bonus was 20%, meaning that there was a chance mechanism involved where only a certain percentage of participants was randomly selected to receive the incentive. Finally, a total of 503 samples were collected. To ensure data quality, some samples were removed according to completeness, obvious errors, response times, etc. Among them, 414 samples were valid for further analysis. Regarding the combinations of scenarios based on activity type (commute or recreation), travel period (peak or off-peak), and price level (normal or discount), each participant was invited to make decision on 8 scenarios. Consequently, the number of travel mode choice records was 3312.

The survey consisted of four parts, i.e., socio-demographic information, daily travel information, behavioral intention, and travel mode choice. Among these, this paper mainly focuses on travel mode choice. In the travel mode choice section of this survey, the respondents were assumed to have both topographical and financial access to all the main travel modes in the urban area. The travel modes considered in this research included subway, bus, car, taxi, and ride-hailing services. Moreover, ride-hailing services were categorized into fast rides, tailored rides, and carpools. A travel distance of 10 km was set, which is close to the average daily travel distance of commuters in Nanjing; thus, cycling and walking were excluded. Eight scenarios were then constructed by combining activity, travel period, and price level. As for activity (ACT), commute and recreation were considered. For activity (ACT), commute and recreation were considered. For travel period (PER), a distinction was made between peak and off-peak travel periods. Additionally, normal and discounted price levels of ride-hailing services were considered for the price level (PRI). Although the actual prices of ride-hailing services vary, it is possible for ride-hailing platforms to offer the same discount level. Moreover, the average time (waiting and on-road time), average cost, and reliability were provided to support the respondents’ decision making. Finally, respondents chose a travel mode in each scenario.

In addition, the PV, SN, and BI regarding ride-hailing services were evaluated using a 5-point Likert scale, which has proven to be an effective method for collecting data on travelers’ perceptions [48,49].

Perceived Value: Overall, using ride-hailing delivers me good value [24].

Subjective Norm: People who are important to me think that I should use ride-hailing services [25,50].

Behavioral Intention: I intend to use ride-hailing more often in the future [25,50].

Additionally, information on socio-demographic characteristics was included, i.e., gender, age, income, and car ownership. As for the usage experience, the usage frequency (RHFRE) and accessibility (ACC) of ride-hailing services were considered.

3.2. Variable Specification

Table 1. Definition of variables.

Variables	Definition
GEN	Male = 1, female = 0
AGE1	18 years old and below (yes = 1, no = 0)
AGE2	19–24 years old (yes = 1, no = 0)
AGE3	25–30 years old (yes = 1, no = 0)
AGE4	31–35 years old (yes = 1, no = 0)
AGE5	36 years old and more (yes = 1, no = 0)
INC1	CNY 1500 and below (yes = 1, no = 0)
INC2	CNY 1501–3500 (yes = 1, no = 0)
INC3	CNY 3501–5000 (yes = 1, no = 0)
INC4	CNY 5001–8000 (yes = 1, no = 0)
INC5	8001 and more (yes = 1, no = 0)
CAR	Do not have a car = 1, do not have a car, but plan to buy one = 2, have only one car = 3, have more than one car = 4
RHFRE	I never do this = 1, I do this, but not in the past 30 days = 2, I did this 1–3 times in the past 30 days = 3, I do this 1 day per week = 4, I do this 2 or more days per week = 5
ACC	In my resident area, I can have access to ride-hailing service easily (fully disagree = 1, disagree = 2, not sure = 3, agree = 4, fully agree = 5)
ACT	Commute = 0, recreation = 1
PER	Peak period = 0, off-peak period = 1
PRI	Discount = 0, normal price = 1
BI	I intend to use ride-hailing more often in the future (fully disagree = 1, disagree = 2, not sure = 3, agree = 4, fully agree = 5)
PV	Overall, using ride-hailing delivers me good value (fully disagree = 1, disagree = 2, not sure = 3, agree = 4, fully agree = 5)
SN	People who are important to me think that I should use ride-hailing services (fully disagree = 1, disagree = 2, not sure = 3, agree = 4, fully agree = 5)
Travel mode	subway = 1, bus = 2, car = 3, taxi = 4, fast ride = 5, tailored ride = 6, carpool = 7

presents the summary of the variables studied in this paper. The mean and standard deviation are included.

In Table 1, the dependent variable is the choice of travel mode provided in this research. The independent variables include socio-demographic information (gender, age, income, and car ownership), usage experience (usage frequency and accessibility), scenario combinations (activity, period, and price level), and socio-psychological variables (behavioral intention, perceived value, and subjective norm). Age and income variables are divided into binary categories to examine the choice behavior of different groups.

3.3. Model Construction

According to previous research, the discrete choice model is used to analyze choice behavior among a set of alternatives [51]. Moreover, the discrete choice model has been comprehensively applied in transportation research [52,53]. In travel mode choice research, travelers’ subjective factors, such as attitudes and perceptions, can be introduced as socio-psychological variables [15,22,49]. Since there are seven travel mode alternatives in this research, a discrete choice model can be established to reveal the relationship patterns between socio-demographic factors, experience, scenarios, socio-psychological variables, and travel mode choices.

In reality, all travel modes are potential choices for travelers, and travelers’ perceptions of ride-hailing services may influence their choices of other travel modes. Therefore, to examine the interactions between ride-hailing and other transportation modes, the utility functions for subway, bus, car, and other modes incorporate specific attributes or characteristics typically associated with ride-hailing services, such as usage frequency and accessibility. Discrete choice models have effectively addressed research issues in various areas. The most commonly used discrete choice models include the logit model, nested logit model, and probit model. However, the logit model has limitations due to the independence of irrelevant alternatives (IIA). Previous research has shown that the probit model can yield results similar to those of the logit model and can overcome the limitations of IIA. Therefore, a multinomial probit model is considered in this study. The Multinomial Probit Model is a statistical model designed to analyze the relationship between a categorical dependent variable and one or more independent variables [54,55]. Given the normality assumption, the probability of choosing a category is computed by evaluating the cumulative distribution function of the standard normal distribution along with the difference between the utility of the category and the maximum utility of all other categories. The Multinomial Probit Model accommodates more than two response categories and is grounded in the principle of utility maximization, making it suitable for this research.

The construction of the discrete choice model is based on an assumption of utility maximization, i.e., travelers will choose the travel mode which can provide the maximum utility. In this process, the characteristics of alternatives will be evaluated. Moreover, the factors of socio-demographic and socio-psychology variables play important roles. Assume that the choice set of traveler n is $A_n$, and the utility of alternative $j$ for traveler $n$ is $U_{jn}$. In this paper, $A=\{subway,bus,car,taxi,fast ride,tailored ride,carpool\}$. Then, the condition in which traveler n willing to choose alternative $i$ over $A_n$ is $U_{in}>U_{jn},i\ne j,j\in A_n$.

According to random utility theory, utility is a random variable which consists of fixed utility and error term, as shown in $U_{in}=V_{in}+{\epsilon }_{in}$. In this equation, $V_{in}$ is the fixed utility and ${\epsilon }_{in}$ is the error term. Further, in probit model, ${\epsilon }_{in}$ follows a multivariate normal distribution. The mean of this error term is 0, and the variance is $\Sigma$, i.e., $\epsilon ~MVN[0,\Sigma ]$. Then, the probability of traveler n choosing travel mode i can be obtained as follows:

$$P_{in}=Prob\left(U_{in}>U_{jn};i\ne j,j\in A_n\right)=Prob(V_{in}+{\epsilon }_{in}>V_{jn}+{\epsilon }_{jn};i\ne j,j\in A_n)$$

In this equation, $0\le P_{in}\le 1,\sum _i{P}_{in}=1$. When there are more than 2 alternatives, the probit model will be much more difficult to solve. In this paper, Stata 14.1 is used to estimate the parameters in the multinomial probit model.

4. Results and Discussion

4.1. Descriptive Analysis

Using the method of random sampling, 414 valid samples were collected. The distribution of respondents according to various socio-demographic variables can be further described. Among all respondents, 50.5% were male and 49.5% were female. Regarding age, the dominant age group in this sample was between 25 and 30 years old (53.4%). Younger generations appeared to predominate in the sample data, which may be because ride-hailing users tend to be younger and the specific representativeness is difficult to evaluate [23]. In addition, although the majority of respondents were younger, the results are also reliable, as indicated in previous literature [13,23,56,57]. However, income distribution was more evenly spread across five groups: CNY 1500 and below (21.0%), CNY 1501–3500 (13.8%), CNY 3501–5000 (15.2%), CNY 5001–8000 (25.1%), and CNY 8001 and above (24.9%). Regarding car ownership, the proportion of respondents who did not own a car was slightly higher than that of those who did. Furthermore, most respondents had easy access to ride-hailing services. Notably, more than half of the samples had experience using ride-hailing services, with 20.1% of them using it at least once per week. Moreover, the top three concerns for travelers regarding ride-hailing services were average price (75.1%), average waiting time (69.8%), and safety (69.6%), as shown in Figure 1.

As described above, there were 3312 observations of choice behavior. The descriptive results of travel mode choices in various scenarios are summarized in

Table 2. Descriptive analysis (%).

		Subway	Bus	Car	Taxi	Fast Ride	Tailored Ride	Carpool
Total		47.9	9.7	13.3	3.9	15.8	2.9	6.4
Activity	Commute	53.3	10.6	13.6	3.1	12.7	1.6	5.0
	Recreation	42.5	8.8	13.0	4.8	18.9	4.2	7.9
Period	Peak	50.1	8.7	13.2	4.0	15.6	2.7	5.7
	Off-peak	43.8	10.4	13.9	4.0	16.8	3.4	7.7
Price	Normal	53.0	11.2	13.6	4.3	12.2	1.2	4.4
	Discount	42.8	8.3	13.0	3.5	19.4	4.6	8.5

. The subway was the most popular travel mode across all scenarios. Ride-hailing services were chosen more often in scenarios involving recreation, off-peak travel periods, and discounted prices. Conversely, in scenarios involving commute, peak travel periods, and normal price levels, public transport was chosen more frequently. However, travelers showed little difference in their choice behavior between cars and taxis in most scenarios.

4.2. Empirical Results and Discussion

The empirical results of travel mode choice behavior are presented in

Table 3. Hypothesis test.

Hypothesis		Fast Ride	Tailored Ride	Carpool
H1a	The preferences for multiple forms of ride-hailing services differ between females and males.	Yes	Yes	Yes
H1b	Ride-hailing users tend to be younger.	Yes	No	Yes
H1c	Ride-hailing users tend to have higher income levels.	Yes	Yes	No
H2	Travelers who own a private car have a decreased possibility of choosing ride-hailing services and public transportation.	Yes	Yes	Yes
H3a	Travelers who use ride-hailing services more frequently will be more inclined to choose such services.	Yes	Yes	Yes
H3b	Travelers will choose ride-hailing services more frequently if they can easily access such services.	No	No	No
H4a	Travelers are more likely to choose ride-hailing services for recreational activities.	Yes	Yes	Yes
H4b	Travelers are more likely to choose ride-hailing services during off-peak periods.	No	No	Yes
H4c	Travelers are more likely to choose ride-hailing services with a price discount.	Yes	Yes	Yes
H5a	Travelers’ choice behavior of ride-hailing services is influenced by behavioral intention.	Yes	Yes	No
H5b	Travelers’ choice behavior of ride-hailing services is influenced by perceived value.	Yes	No	Yes
H5c	Travelers’ choice behavior of ride-hailing services is influenced by social norms.	Yes	Yes	Yes

and

Table 4. Results of multinomial probit regression.

	Model 1												Model 2
Mode	Subway		Bus		Taxi		Fast Ride		Tailored Ride		Carpool		Subway		Bus		Taxi		Fast Ride		Tailored Ride		Carpool
Value	Coef.	P	Coef.	p	Coef.	p	Coef.	P	Coef.	p	Coef.	p	Coef.	p	Coef.	p	Coef.	p	Coef.	p	Coef.	p	Coef.	p
Gen	0.087	0.304	0.032	0.754	−0.046	0.700	−0.053	0.571	0.291	**	0.002	0.985	0.117	0.169	0.040	0.697	−0.025	0.839	−0.043	0.649	0.301	**	0.023	0.833
Age1	0.416	0.525	2.383	***	1.607	**	1.362	**	1.181	0.108	1.662	**	0.358	0.577	2.267	***	1.385	*	1.224	*	0.974	0.179	1.629	**
Age2	0.079	0.691	0.019	0.937	−0.393	0.168	0.401	*	−0.828	***	−0.306	0.225	0.106	0.595	0.050	0.839	−0.357	0.214	0.492	**	−0.772	**	−0.224	0.382
Age3	−0.167	0.252	−0.127	0.512	−0.257	0.227	−0.196	0.242	−0.954	***	−0.436	**	−0.175	0.230	−0.121	0.533	−0.284	0.186	−0.196	0.244	−0.963	***	−0.411	**
Age4	−0.518	***	−0.060	0.778	−0.391	0.113	−0.260	0.154	−0.407	*	−0.459	**	−0.536	***	−0.069	0.745	−0.430	*	−0.289	0.117	−0.415	*	−0.451	**
Inc1	0.005	0.975	0.111	0.582	−0.264	0.291	0.005	0.979	0.057	0.841	−0.136	0.538	0.034	0.844	0.143	0.483	−0.192	0.447	0.012	0.950	0.110	0.704	−0.171	0.451
Inc2	0.101	0.496	0.395	**	0.550	***	0.090	0.594	0.268	0.264	0.214	0.253	0.151	0.314	0.426	**	0.562	***	0.052	0.765	0.266	0.272	0.184	0.334
Inc3	−0.224	*	0.352	**	0.105	0.589	0.088	0.545	0.188	0.382	−0.086	0.620	−0.232	*	0.338	**	0.136	0.487	0.092	0.533	0.190	0.380	−0.119	0.497
Inc4	0.091	0.410	0.372	***	0.234	0.168	0.221	*	0.431	**	0.230	0.111	0.109	0.333	0.402	***	0.242	0.161	0.197	0.121	0.432	**	0.198	0.176
Car	−0.793	***	−0.863	***	−0.821	***	−0.561	***	−0.612	***	−0.708	***	−0.801	***	−0.863	***	−0.817	***	−0.556	***	−0.602	***	−0.717	***
RHFre	0.070	*	0.213	***	0.063	0.297	0.481	***	0.221	***	0.417	***	0.133	***	0.227	***	0.065	0.317	0.445	***	0.193	***	0.399	***
Acc	0.103	**	−0.088	*	−0.124	**	0.008	0.869	0.113	0.112	−0.094	*	0.140	***	−0.064	0.205	−0.115	*	−0.024	0.616	0.106	0.154	−0.132	**
Act	−0.212	**	−0.120	0.229	0.248	**	0.303	***	0.547	***	0.288	***	−0.217	***	−0.120	0.231	0.259	**	0.308	***	0.559	***	0.290	***
Per	−0.239	***	0.017	0.867	−0.044	0.709	0.020	0.824	0.158	0.229	0.183	*	−0.239	***	0.018	0.857	−0.045	0.708	0.025	0.792	0.171	0.198	0.185	*
Pri	0.192	**	0.173	*	0.091	0.445	−0.358	***	−0.751	***	−0.426	***	0.196	**	0.174	*	0.098	0.413	−0.367	***	−0.759	***	−0.434	***
cons	2.349	***	1.200	***	1.409	***	−0.009	0.979	−0.318	0.500	0.584	0.116	2.688	***	1.290	***	1.513	***	−0.536	0.147	−0.594	0.257	0.085	0.841
BI													0.060	0.302	0.103	0.140	0.328	***	0.268	***	0.299	***	0.069	0.363
PV													−0.039	0.551	−0.049	0.527	−0.132	0.141	0.165	**	−0.002	0.985	0.242	***
SN													−0.223	***	−0.120	*	−0.228	***	−0.220	***	−0.186	**	−0.126	*
Wald chi2	1044.143												1112.670
Prob > chi2	0.000												0.000
Count R2	0.514												0.518
AIC	2.778												2.756

Notes: (a)* p < 0.1 ** p < 0.05 *** p < 0.01. (b) The baseline is car.

and Figure 2 and Figure 3. The results consist of two main parts. Firstly, the analysis results without considering the socio-psychological variables of travelers’ travel mode choice behavior are shown as Model 1. Secondly, several socio-psychology variables are introduced to analyze travelers’ travel mode choice behavior, and the results are shown as Model 2. The socio-psychology variables include behavioral intention (BI), perceived value (PV), and subjective norm (SN). Additionally, the model fitness criteria are summarized in Table 4. Wald chi-square (chi2), Prob > chi2, Count R-squared (R2), and AIC (Akaike Information Criterion) were included to assess model fitness. A large chi-square value, accompanied by a small p-value, implies that certain factors contribute to explaining the variation in the dependent variable. A higher Count R-squared indicates a better fit to the data. AIC allows for model comparison. When comparing multiple models, the one with the lowest AIC is considered to strike the best balance between complexity and fit.

Based on the values below, both models demonstrated good fit, with the model including socio-psychology variables performing better than the model without them. The empirical analysis was conducted using multinomial probit regression in Stata; therefore, a common baseline was necessary. In this study, the baseline was set as “CAR”. In this context, the coefficients presented in Table 4 indicate the change in the outcome associated with each variable level relative to the “CAR” level. Thus, the effect of each specific variable on different travel modes can be compared with the common baseline. The empirical results of Models 1 and 2 are compared and summarized across four groups of variables: socio-demographic variables, experience variables, scenario variables, and socio-psychology variables. Overall, there is minimal difference between the outcomes of Models 1 and 2, but it is important to illustrate the results by comparing these two models.

Regarding gender, the positive and significant coefficient for tailored rides indicates that males prefer tailored rides over cars more than females do, which supports H1a. Travel mode choices vary across different age groups. Younger generations (AGE1 and AGE2) show a stronger preference for ride-hailing services compared to older generations (AGE3 and AGE4), which aligns with previous research [4,26]. Additionally, the negative and significant coefficients of AGE3 and AGE4 indicate that older generations prefer cars over all other travel modes. Overall, ride-hailing services are more popular than taxis across all age groups. In addition, younger generations prefer to use public transportation compared to older generations. Among the choices of ride-hailing services, fast rides and carpools are more popular with younger generations, while older generations are more inclined to choose tailored rides. Thus, H1b is supported with fast rides and carpools, but not supported with tailored rides. This may be caused by the income levels of different age groups. In many studies, income has been one of the most important factors reflecting socio-demographic information [10,13,21]. Table 3 suggests that lower-income groups more frequently choose public transportation modes (subway and bus) compared to higher-income groups. Conversely, the conclusions regarding the effect of income on the choice of ride-hailing services are markedly different. There appears to be a trend where travelers with higher incomes are more likely to accept particularly tailored rides and fast rides. Therefore, H1c is confirmed with fast rides and tailored rides, but not confirmed with carpools.

Interestingly, all the coefficients of CAR are negative and highly significant. This suggests that car ownership significantly influences travelers’ travel mode choice behavior. If travelers own a car, their frequency of car usage will be higher than other travel modes. However, the choice of all travel modes involving ride-hailing services remains superior to subway, bus, and taxi travel. Hence, H2 is supported with all forms of ride-hailing services. Moreover, fast rides and tailored rides are preferred over carpools. Travelers who own a car are likely accustomed to and prefer traveling by car and similar modes of transport.

Experience may contribute to the choice behavior of ride-hailing services, and the pattern of travelers’ travel mode choice may change due to the accumulated experience of ride-hailing usage. As for ride-hailing services’ usage frequency (RHFRE), it verifies the effect of ride-hailing experience on travel mode choice behavior according to positive and highly significant coefficients [10]. Moreover, the coefficients of fast ride, tailored ride, and carpool in Model 1 and Model 2 are significant and relatively higher than those for public transportation and taxi. This indicates that RHFRE plays an important role in the choice behavior of ride-hailing services, which further suggests that travelers with higher ride-hailing usage frequency tend to choose ride-hailing services more frequently. Thus, H3a is confirmed with all forms of ride-hailing services. Interestingly, the impact of access to ride-hailing services (ACC) is not substantial. Hence, H3b is not supported. This is likely because most respondents in this survey were urban residents who have easy access to ride-hailing services.

Three scenario factors were introduced, creating a total of eight scenarios for the research on travel mode choice. The highly significant coefficients in Table 4 indicate that travelers’ choice behavior is significantly influenced by scenarios, as is consistent with previous literature [45,46,47]. The evidence implies that travelers’ preference for public transportation is higher than car transportation when commuting. Furthermore, when engaging in recreational activity, travelers prefer to choose ride-hailing services and taxis as their travel modes. In addition, all forms of ride-hailing services are preferred with recreation. Thus, H4a is supported by the results on fast rides, tailored rides, and carpools. Interestingly, during recreational activities, travelers tend to prefer fast rides and tailored rides over carpools. Concerning the results related to travel periods, carpools are more preferred during off-peak travel periods. However, the impact of the travel period on the preference for fast rides and tailored rides is not significant. Hence, H4b is only supported with carpools. Further, there is a strong possibility that travelers prefer to choose ride-hailing services in the situation of a discounted price for ride-hailing. Therefore, when the price level of ride-hailing services turns to normal, it is more likely that travelers will choose the subway and bus as their travel modes. During discounted periods offered by ride-hailing platforms, travelers tend to choose tailored rides more frequently than other ride-hailing service modes. Therefore, H4c is supported for all the three forms of ride-hailing services.

Clearly, the inclusion of socio-psychology variables has improved the fit of Model 1. Several meaningful and interesting results highlight the significant role of socio-psychology variables [15,21,22,38]. Additionally, H5 can be confirmed as well. In Table 4, it is suggested that the behavioral intention (BI) has a positive effect on the choice of ride-hailing services. In addition, the choice of taxi is positively and significantly related to travelers’ behavioral intentions. This may be caused by travel demand, since travelers with higher behavioral intentions to use ride-hailing services normally tend to have larger travel demand for similar travel modes. Further, the perceived value (PV) of ride-hailing services also affects travelers’ choice behavior regarding ride-hailing services. Furthermore, the highly significant coefficients of subjective norm (SN) suggest that the subjective norm contributes to travelers’ travel mode choice behavior. Although the coefficients are negative, they can imply that travelers prefer to use ride-hailing services due to the increase in subjective norm. Regarding ride-hailing services, the behavioral intention is more influential on the choice of fast rides and tailored rides. Therefore, H5a is confirmed with the results of fast rides and tailored rides, but is not confirmed with carpools. The reason may be that fast rides and tailored rides can better represent the characteristics of ride-hailing services and provide more customized travel services. Further, from the coefficients of perceived value, carpools and fast rides are more popular than tailored rides. Hence, H5b is supported with fast rides and carpools, but not supported with tailored rides. Maybe the high prices of tailored rides have decreased the perceived value of such travel modes. From the viewpoint of subjective norms, the effect on choice behavior for all forms of ride-hailing services is significant, and travelers tend to choose carpool as their travel mode among ride-hailing services. Thus, H5c is supported with all three forms of ride-hailing services. It is possible that travelers who are highly affected by others may be more likely to carpool with their friends or family members.

Further, according to the results and discussions above, the research gap is closed. Firstly, as shown in Table 3, there exists a significant distinction between travelers’ choice behavior towards fast rides, tailored rides, and carpools. Currently, ride-hailing platforms tend to perform various services to cover the demand of different groups of consumers. Additionally, the results of this study further strengthen the importance of investigating specific forms of ride-hailing services. Hence, this can contribute to existing studies on travelers’ behavior related to ride-hailing services. Secondly, this study demonstrates that travelers’ choices of ride-hailing services rely on activities, travel periods, and price level, which verifies the impact of scenarios on consumer behavior. By including travel scenarios, it adds a layer of complexity that reflects real-world decision-making processes. This comprehensive approach not only enhances the practical relevance of the research, but also contributes to a deeper theoretical understanding of the factors influencing ride-hailing choices. Thirdly, with consideration of socio-psychological variables, it allows for a comparative analysis that highlights the significance of psychological factors in consumer decisions. This result regarding socio-psychological factors offers a clearer picture of how socio-psychological aspects can shape the demand for ride-hailing services, which contributes to existing studies.

4.3. Marginal Effects Analysis

Based on the results of Model 1 and Model 2, an analysis of marginal effects can further be carried out. The results of marginal effects are presented in Table 4. Marginal effects analysis mainly focuses on the effects of independent variables on the probabilities of seven travel modes considered in this study.

Table 5. Results of marginal effects.

Mode	Subway		Bus		Car		Taxi		Fast Ride		Tailored Ride		Carpool
Value	dy/dx	p	dy/dx	p	dy/dx	p	dy/dx	p	dy/dx	p	dy/dx	p	dy/dx	p
GEN	0.030	*	−0.003	0.787	−0.010	0.375	−0.006	0.385	−0.022	*	0.013	**	−0.003	0.705
AGE1	−0.254	***	0.207	***	−0.139	0.110	0.033	0.153	0.074	0.218	0.004	0.848	0.076	**
AGE2	0.018	0.632	−0.001	0.953	−0.010	0.711	−0.028	*	0.094	***	−0.043	***	−0.031	0.112
AGE3	0.012	0.707	0.013	0.544	0.035	*	−0.005	0.668	0.004	0.856	−0.038	***	−0.021	0.177
AGE4	−0.093	***	0.044	*	0.058	**	−0.006	0.694	0.013	0.624	−0.004	0.634	−0.012	0.507
INC1	0.010	0.752	0.020	0.301	−0.002	0.932	−0.014	0.291	0.001	0.981	0.005	0.659	−0.020	0.253
INC2	−0.014	0.632	0.036	**	−0.028	0.146	0.026	**	−0.026	0.246	0.005	0.629	0.001	0.954
INC3	−0.099	***	0.056	***	0.006	0.732	0.011	0.315	0.027	0.183	0.011	0.261	−0.011	0.430
INC4	−0.031	0.187	0.034	**	−0.028	**	0.004	0.659	0.005	0.765	0.013	*	0.002	0.870
CAR	−0.075	***	−0.030	***	0.105	***	−0.011	***	0.018	***	0.001	0.859	−0.008	*
RHFRE	−0.032	***	0.004	0.435	−0.033	***	−0.009	***	0.051	***	−0.001	0.754	0.020	***
ACC	0.053	***	−0.014	***	−0.005	0.307	−0.010	***	−0.011	*	0.005	0.167	−0.017	***
ACT	−0.109	***	−0.019	*	−0.004	0.738	0.017	***	0.061	***	0.026	***	0.026	***
PER	−0.082	***	0.014	0.161	0.011	0.297	0.002	0.805	0.018	0.127	0.011	*	0.026	***
PRI	0.103	***	0.028	***	0.006	0.550	0.008	0.207	−0.070	***	−0.036	***	−0.040	***
BI	−0.028	**	−0.002	0.720	−0.020	***	0.015	***	0.032	***	0.009	**	−0.006	0.263
PV	−0.027	**	−0.010	0.192	−0.004	0.668	−0.010	**	0.030	***	−0.002	0.682	0.023	***
SN	−0.026	**	0.008	0.225	0.028	***	−0.004	0.386	−0.011	0.182	−0.001	0.768	0.005	0.392

Notes: * p < 0.1 ** p < 0.05 *** p < 0.01

verifies most of the conclusions drawn from the results above. However, some interesting results can also be observed from the perspective of marginal effects.

Among the travel modes offered by ride-hailing platforms, females prefer fast rides, while males prefer tailored rides. Travelers in the AGE3 and AGE4 groups show a higher preference for cars compared to other travel modes. This preference may stem from the fact that older generations can afford to own cars and tend to prefer more comfortable travel modes. The marginal effect of income reinforces the conclusions drawn from the multinomial probit model. The highly significant and positive marginal effect of CAR on car choice suggests that travelers who own cars are more likely to choose cars as their primary mode of travel. Travelers with higher ride-hailing usage frequency show a higher probability of choosing ride-hailing services compared to other travel modes.

As for the factors of scenario, more significant relationships emerge, as shown in Table 4, which jointly suggests that travelers’ travel mode choice behavior is profoundly influenced by scenarios. And the results are similar to the multinomial probit model in that travelers prefer ride-hailing services in scenarios of recreation, off-peak travel periods, and discounted prices. Moreover, in commuting, peak travel period, and normal-price-level scenarios, travelers prefer to choose public transportation. In addition, when engaged in recreational activity, taxis are also considered as one of the most popular travel modes. The marginal effects of socio-psychology variables demonstrate that travelers’ decision-making behavior normally relies on perceived value, subjective norm, and behavioral intention. Similarly, the variables of behavioral intention and perceived value have positive effects on the choice of ride-hailing services. However, the marginal effect of subjective norm on the choice of ride-hailing services is insignificant. And the negative and significant effects of such socio-psychology variables on subway, bus, car, and taxi suggest that higher perceived value, behavioral intention, and subjective norm toward ride-hailing services lead to a decrease in the probability of travelers choosing these travel modes. Finally, according to the marginal effects shown in Table 5, the preference for travel modes among ride-hailing services is similar to the results demonstrated in Table 2.

5. Conclusions

5.1. Main Results

This study examines the travel mode choice behavior towards multiple forms of ride-hailing services using a multinomial probit model. By referring to previous research on travel mode choice behavior, multiple socio-psychology variables are introduced. The results indicate that the fitness of the model is improved by the consideration of perceived value, behavioral intention, and subjective norm. The study designed the choice behavior and created eight scenarios by combining three scenario factors—activity (commute and recreation), travel period (peak and off-peak), and ride-hailing price levels (discounted and normal)—to analyze travelers’ mode choices. Furthermore, the study examined how travelers’ preferences for travel modes change across different situations. Additionally, the study investigated the influence of ride-hailing usage on traditional travel modes. Finally, the conclusions drawn from the multinomial probit model results were further tested using marginal effects.

The main conclusions of this study are summarized as follows. As is consistent with previous research, consumers of ride-hailing services tend to be younger, have higher incomes, and do not own cars. Among ride-hailing services, younger generations prefer fast rides and carpools, while older travelers with higher incomes tend to choose tailored rides. For public transportation, younger travelers with lower incomes are more inclined to choose the subway and bus. Additionally, while taxis provide a similar transportation experience to ride-hailing services, there is a general preference for ride-hailing services over taxis. Regarding the effect of gender on choice of ride-hailing services, males tend to prefer tailored rides, while females tend to prefer fast rides.

In many studies, car ownership is considered as one of the most important socio-demographic factors. It strongly indicates that travelers prefer to travel by car more than other travel modes if they own one. The effect of ride-hailing usage on the choice behavior towards subway, bus, car, and taxi transportation is further examined through the coefficient and marginal effect of ride-hailing usage frequency (RHFRE). This suggests that, if travelers have more experience with ride-hailing usage, they will choose traditional travel modes less frequently than ride-hailing services. In addition, the probability of choosing fast rides, tailored rides, or carpools increases with higher ride-hailing usage experience. This result further demonstrates that ride-hailing usage has a significant effect on traditional travel modes.

Significantly, the scenario factors have a substantial impact on travelers’ travel mode choice behavior. Generally, travelers are more inclined to choose ride-hailing services in the recreational scenarios, during normal travel periods, and when there are discounted prices for ride-hailing. By contrast, in scenarios characterized by commuting, periods of heavy travel, and normal price levels of ride-hailing, travelers tend to choose public transportation as their travel mode. The choice of taxis shows a similar trend to ride-hailing services. Behavioral intention (BI), perceived value (PV) and subjective norm (SN) are factors considered as socio-psychology variables in this study. The results of socio-psychology variables indicate that travelers’ perceived value and behavioral intentions have a positive effect on the choice behavior regarding ride-hailing services. Moreover, an increase in subjective norm leads travelers to have a higher probability of choosing ride-hailing services over other travel modes.

5.2. Implication

Several theoretical implications can be deduced from this study. Firstly, this study makes a significant contribution to the literature on urban mobility patterns and electronic commerce by providing a detailed examination of travelers’ choice behavior across different ride-hailing modes and comparing them with traditional transportation options. This study deepens our understanding of how socio-demographic characteristics, usage experience, socio-psychological factors, and specific travel scenarios interact to shape preferences for specific forms of ride-hailing services. Secondly, by designing and analyzing a range of travel scenarios, the research enhances the understanding of the way in which contextual factors interact with individual characteristics to influence travel choices. This contributes to the development of behavioral theory, underscoring the pivotal role of situational dynamics in decision-making processes, which is also a compensation for studies related to electronic commerce. Thirdly, the introduction of two modeling approaches, one with socio-psychological variables and one without, represents a methodological advancement in the study of travel behavior. This comparative analysis serves to reinforce the robustness of the findings and to highlight the importance of taking psychological aspects into account when attempting to predict consumer choices.

In addition, several practical implications can be provided according to the results of this study. Firstly, an understanding of the subtleties of travelers’ preferences for different forms of ride-hailing services can inform policymakers in the design of fair and effective regulatory frameworks. This encompasses considerations pertaining to pricing policies, vehicle standards, and incentives for carpooling, with the objective of mitigating congestion and environmental impacts. Secondly, the interdependence between ride-hailing and other transport modes under different scenarios highlights the necessity of integrated transport planning. Policymakers can use these insights to develop strategies that promote connectivity and a shift in modal use towards more sustainable options by considering different scenarios. Thirdly, the insights on socio-psychological factors can inform the development of user-centered policies and services that cater to travelers’ diverse needs and preferences, thereby fostering a more inclusive and satisfying travel experience.

5.3. Limitations and Future Directions

There are still some limitations of this research, and more work can be continued in the future. Firstly, more factors that reflect the characteristics of ride-hailing services should be considered. As a service or product, the consumer behavior of ride-hailing services is complicated. In future, more research will be carried out using a combination of behavioral decision theory and issues of ride-hailing services. Further, more advanced modeling methods should be considered in the next stage. Furthermore, there is still a limitation regarding the scale of the data sample in this study. In a future study, a survey with a larger scale needs to be put forward. For example, a larger dataset considering a more detailed age structure is required; this would be helpful in understanding the general patterns concerning ride-hailing services more comprehensively. In addition, a comparison of samples from different cities or cultural background would be an interesting research direction.