Buses represent the main mode for intercity passenger transportation in China. In recent years, a multichannel ticketing strategy has been widely employed in the bus passenger transportation industry. However, the mechanisms and key drivers of the channels through which bus passengers purchase tickets are underexplored. Thus, the aim of this study is to empirically apply an integrated choice and latent variable (ICLV) approach to analyze ticketing channel choice behavior and the heterogeneous preferences of bus passengers. The variables incorporated in the model include the socioeconomic characteristics of passengers, trip attributes, and latent attitudes with 12 ordinal indicators. Based on the data of 1800 participants collected from the city of Beijing, China, this study develops a ticketing channel choice ICLV model merging a discrete choice model with a structural equation model. The key factors that affect the channel preference are further discussed through a comparison with a conventional multinomial logit (MNL) model. The results reveal that the three attitudinal variables have a significant influence on ticketing channel choice. Furthermore, this study indicates that perceptual differences exist due to various socioeconomic and trip characteristics. Personal privacy is a major obstacle that prevents passengers from choosing online channels, especially for older passengers and those with lower education.
Over the last decade, the multichannel strategy has become widely employed in various fields. It has also been targeted from the perspective of the transportation industry in recent years. Passenger transportation operation organizations are selling their tickets or services via multiple channels to reduce distribution costs and provide passengers with convenient service. Taking the example of bus passenger transportation in China, there was only a single channel (counter service at the station) from which passengers purchased tickets for a long period of time. In order to overcome the disadvantages resulting from the limitation of utilizing only a single ticketing channel, together with the integration of modern information technology, bus passenger transportation agencies have developed multiple distribution channels to reserve tickets, such as service counters and ticket vending machines (tvms) at stations, ticketing agencies, websites, mobile apps, etc. The use of multiple channels may result in tremendous savings in monetary and non-monetary costs from a passenger’s perspective. Consequently, the ticketing behavior of bus passengers has gradually changed and it is expected to be reconstructed, especially during the holidays. During 40 days of the Spring Festival in 2018, a total of 2.97 billion passengers were reported to have traveled, of which 2.48 billion traveled by bus transportation. A survey from a third party suggested that 63% of the bus passengers used online channels to purchase tickets. A multichannel strategy for ticketing services makes travelers feel better and more satisfied. It is believed that the utilization of the multichannel strategy for ticketing is conducive to the sustainable development of public transportation and to passengers’ travel behavior. As discussed by Birgelen et al. [
Since the multichannel strategy has become more widely adopted, channel performance has been paid greater attention by researchers [
The rest of the paper is organized as follows. Section
The previous research on channel choice has considered the consumer purchasing process as a sophisticated decision-making procedure. However, discrete choice analysis is supposed to be a proper approach for evaluating choice behavior and the decision process. Over the last few decades, numerous developments have been made within the field of discrete choice modeling in order to better unravel the underlying process leading up to observed choice outcomes. Behavior researchers have particularly attempted to improve models with the aim of enriching the collaboration of components of the decision-making process, such as attitudes, perception, incentives, etc. Reviewing the current literature indicates that decision-makers’ attitudes influence their parameter estimates [
Ticket purchasing can be regarded as a kind of consumption behavior in the transportation industry, which is also a complex decision-making process. Apparently, there exists an objective complexity which can be evaluated as a role both of individual properties of the passenger such as gender, age, revenue, employment status, etc., and of the trip attributes, i.e., travel distance, departure time, etc. Apart from the objective factors listed above, choice behavior is affected by a number of not directly observable factors such as attitude, habit, perception, etc. It means that there is a perceived complexity which also influences passengers’ behavioral responses and which is the subject of interest of our study.
We note that relevant research papers have adopted various kinds of channels, but most of them only focus on certain industries, especially within the retail industry [
As discussed above, the consumers’ internal process of preference formulation and the role of not directly observable factors such as attitudes, habits, etc. remain unexplained, forming the so-called black box in conventional discrete choice models. To address this challenge, this study provides a first attempt to employ an ICLV model which merges classic choice models with the structural equation model (SEM) for latent variables in order to deal with ticketing channel issues. The ICLV model consists of two main components: a choice model and a latent variable model, as shown in Figure
Framework of the integrated choice and latent variable model.
The choice model is a multinomial logit (MNL) model that is utilized to analyze passengers’ response to ticketing channel choice. The passenger
Assuming that
The latent variable model is a MIMIC (Multiple Indicator Multiple Cause) model that consists of two submodels. The first is a structural model that represents the interrelationship between the latent variables and the observed explanatory variables. The formulation is expressed as follows:
The second submodel is a measurement model which is a confirmatory factor analysis model linking the latent variables with their corresponding indicators. A linear factor model is specified to illustrate the mapping of the latent variables on the indicators and is expressed as follows:
The estimation of ICLV models is computationally complicated, and this complexity increases in proportion to the number of latent variables [
The data used in this study originate from a survey that took place in Beijing, the capital of China, in March 2018. The questionnaire is designed with the aim of capturing passengers’ ticketing channel choice behavior and the multidimensional nature of bus passenger transportation. There are four major sections in the questionnaire. The first section investigates ticketing channel choice through three questions. The second section contains 12 questions about attitude of passengers to the ticketing channel adopted. The third section includes six questions about trip characteristics of the respondents. The fourth section is composed of five questions focusing on the sociodemographic information of the passengers.
This study investigates demographic information, trip characteristics, and respondents’ current ticketing channel choices. In terms of ticketing channel usage, we provided five distribution channels for respondents to choose from, including service counters at a station, ticketing agencies, tvms, websites, and apps. Passengers could mark their current choices across the ticket purchasing process, and this enabled us to understand passengers’ current channel preference. We also investigated trip characteristics related to trip frequency, trip purpose, decision-making period prior to a given trip, and the timeframe prior to any given ticket purchase. We only required participants to report the current trip, including business trip or nonbusiness trip. The sociodemographic information collected in the last section of the questionnaire encompassed questions regarding gender, age, education background, occupation, monthly income, etc.
In this study, we consider three latent constructs: perceived risk, perceived usability, and perceived accessibility. The first latent variable, perceived risk, shows an individual’s concerns toward privacy and security issues. Miyazaki and Fernandez [
The indicators that are designed to measure the three attitudinal and perceptual latent variables are presented in Table
Measurement items of latent variables.
Latent Variable | Observed Indicator | Item Description |
---|---|---|
Perceived risk |
|
My personal information is protected in the process of purchasing a ticket |
|
I am protected from fraud in the process of purchasing a ticket (e.g., fake ticket) | |
|
My ticketing habits and transaction records are not tracked | |
|
I can prevent a mistake from being made (e.g., billing error) | |
|
||
Perceived usability |
|
I receive clear, complete, and timely information about my trip |
|
Using this channel, I can save time in the overall ticketing process | |
|
I can purchase my ticket efficiently | |
|
I can complete my transaction more conveniently | |
|
This channel is reliable and makes me feel reassured | |
|
||
Perceived accessibility |
|
It is easy access to this channel |
|
I can obtain clear operating instructions with this channel | |
|
I expended little effort in learning to use this channel |
The survey was administered both on-field and online method. The paper-based questionnaire survey was conducted simultaneously at 7 stations for bus passenger transportation in Beijing. These 7 stations contain almost all the intercity bus passengers. The surveys were collected on March 6, 2018, and March 11, 2018. The surveyors team consisted of postgraduate and undergraduate university students. In order to avoid mistakes, the surveyors were sufficiently trained. The research team had a good cooperation with the intercity bus stations where the survey took place. There were 8 experienced surveyors at each station, who were deployed at the waiting hall during operating hours for data collection. Initially, respondents were approached and asked about their willingness to participate in the survey. After successful completion of the survey, respondents were rewarded with an umbrella to acknowledge their effort. However, respondents were free to agree on whether or not they wanted to participate in the survey. The mean completion time was 15.6 minutes of each survey. The number of surveys collected at each station are 79,70,65,89,101,58,72, respectively. From the 534 surveys, 491 were valid.
Comparison with traditional data-gathering methods, online methods present well due to inherent advantages of low cost, swiftness, and ease of storing data in databases. For the online survey, after the final web survey design was completed, the questionnaire was uploaded on survey website and ticketing app for 15 consecutive days. We recruited participants using different mechanisms. We contacted intercity bus operators and forums and asked them to forward to their members. Further, we disseminated survey information to social medias, such as Weibo and WeChat. For an extra incentive to participate, a draw of a present was organized for the respondents. It required 15 minutes to answer the online survey, while each survey took approximately 10-12 minutes. From the 1396 filled questionnaires, 1309 were acceptable.
The sample used for the modeling process consists of 1800 participants, of which 491 are on-field surveys and 1309 are online surveys. Majority of people who take intercity bus are floating population rather than resident population in Beijing. For this reason, we set out to obtain a sample that provided good coverage and enough sample of various income, age, and gender rather than strictly being population-proportional. Moreover, in this study, only intercity bus passengers were surveyed and asked about their ticketing channel choice behavior. The sample in this case has got certain advantages. Firstly, a relatively larger sample could be obtained due to higher response rates among users who have used an intercity bus for their trip. Secondly, this dataset could contain more expert opinions than is generally obtained from a larger sample.
Table
Demographic, trip characteristics, and ticket usage of respondents.
Characteristic | Group | Total sample | Online sample | On-field sample |
| |||
---|---|---|---|---|---|---|---|---|
Frequency | Percentage (%) | Frequency | Percentage (%) | Frequency | Percentage (%) | |||
Age | <26 | 423 | 23.5 | 361 | 27.6 | 110 | 12.6 | 0.32 |
26-55 | 986 | 54.8 | 690 | 52.7 | 268 | 60.3 | ||
>55 | 391 | 21.7 | 258 | 19.7 | 113 | 27.1 | ||
|
||||||||
Gender | Male | 804 | 44.7 | 544 | 41.6 | 260 | 53.0 | 0.4 |
Female | 996 | 55.3 | 765 | 58.4 | 231 | 47.1 | ||
|
||||||||
Occupation | Clerk | 371 | 20.6 | 264 | 20.2 | 107 | 21.8 | 0.29 |
Servant | 63 | 3.5 | 46 | 3.5 | 17 | 3.5 | ||
Student | 149 | 8.3 | 118 | 9.0 | 31 | 6.3 | ||
Freelancer | 302 | 16.8 | 211 | 16.1 | 91 | 18.5 | ||
Migrant worker | 675 | 37.5 | 472 | 36.1 | 203 | 41.3 | ||
Unemployed or retired | 240 | 13.3 | 198 | 15.1 | 42 | 8.6 | ||
|
||||||||
Monthly income | ≤3000 CNY | 130 | 7.2 | 93 | 7.1 | 37 | 7.5 | 0.51 |
3001-6000 CNY | 675 | 37.5 | 452 | 34.5 | 223 | 45.4 | ||
6001-9000 CNY | 691 | 38.4 | 496 | 37.9 | 195 | 39.7 | ||
>9000 CNY | 304 | 16.9 | 268 | 20.5 | 36 | 7.3 | ||
|
||||||||
Education | Junior high | 122 | 6.8 | 88 | 6.7 | 34 | 6.9 | 0.39 |
High school | 454 | 25.2 | 326 | 24.9 | 128 | 26.1 | ||
Junior college | 503 | 27.9 | 368 | 28.1 | 135 | 27.5 | ||
College | 538 | 29.9 | 407 | 31.1 | 131 | 26.7 | ||
Postgraduate | 183 | 10.2 | 120 | 9.2 | 63 | 12.8 | ||
|
||||||||
Trip purpose | business | 810 | 45.0 | 640 | 48.9 | 170 | 34.6 | 0.71 |
nonbusiness | 990 | 55.0 | 669 | 51.1 | 321 | 65.4 | ||
|
||||||||
Ticket channel | Counters | 626 | 34.8 | 404 | 30.9 | 222 | 45.2 | 0.58 |
Agencies | 133 | 7.4 | 88 | 6.7 | 45 | 9.2 | ||
Websites | 497 | 27.6 | 376 | 28.7 | 121 | 24.6 | ||
Apps | 395 | 21.9 | 313 | 23.9 | 82 | 16.7 | ||
Tvms | 149 | 8.3 | 128 | 9.8 | 21 | 4.3 |
Table
Scores of the measurement items for perceptions.
Observed Indicator | Total sample | Online sample | On-field sample |
| |||
---|---|---|---|---|---|---|---|
Mean | S.D. | Mean | S.D. | Mean | S.D. | (online vs. on-field) | |
y1 | 3.95 | 0.78 | 4.29 | 0.65 | 3.61 | 0.81 | 0.43 |
y2 | 4.32 | 0.80 | 4.38 | 0.92 | 4.26 | 0.76 | 0.19 |
y3 | 3.59 | 1.16 | 2.98 | 1.08 | 4.20 | 1.21 | 0.81 |
y4 | 4.05 | 0.79 | 4.09 | 0.67 | 4.01 | 0.82 | 0.72 |
y5 | 4.24 | 0.88 | 4.71 | 0.93 | 3.77 | 0.81 | 0.38 |
y6 | 2.55 | 1.36 | 3.12 | 1.56 | 1.98 | 1.21 | 0.16 |
y7 | 3.74 | 1.07 | 3.91 | 1.02 | 3.57 | 1.18 | 0.90 |
y8 | 3.99 | 1.02 | 4.01 | 0.87 | 3.97 | 1.23 | 0.53 |
y9 | 4.20 | 0.67 | 3.01 | 0.89 | 5.39 | 0.61 | 0.34 |
y10 | 2.43 | 1.19 | 2.59 | 1.28 | 2.27 | 0.97 | 0.39 |
y11 | 4.03 | 0.80 | 4.01 | 0.95 | 4.05 | 0.76 | 0.26 |
y12 | 4.31 | 0.75 | 3.98 | 0.61 | 4.64 | 0.92 | 0.87 |
In Tables
In order to ensure that all the samples are consistent and stable, we used Cronbach’s alpha to test the reliability of the latent variables. At the same time, we adopted Average Variance Extracted (AVE) and Kaiser–Meyer–Olkin (KMO) tests to guarantee a qualified validity. Reliability and validity tests of the sample are shown in Figure
Reliability and validity tests of the sample.
Figure
The indicators scores of the three latent variables are presented in Figure
Indicators scores of the three latent variables.
This section presents and discusses the estimation results of the model. In the MIMIC model that relates the latent variables to observed explanatory variables, we considered age, gender, occupation, income, and education levels as potential explanatory variables. In the ticketing channel choice model, we considered additional trip-specific variables (such as purpose, frequency, etc.), as well as the latent variables themselves. Annotations and discretization results of the observed variables are represented in Table
Annotations of the observed variables.
Variables | Annotation | Formulation |
---|---|---|
Age |
|
|
Gender |
|
|
Occupation |
|
|
|
||
Income |
|
|
|
||
Education (edu) |
|
|
|
||
|
||
Trip purpose |
|
|
Preset time of decision-making (dm) |
|
|
Preset time of buying tickets (bt) |
|
|
Frequency of the coach (fr) |
|
|
Taking into account the discrepancy of the effect and choice mechanism between channels, we develop counters, ticketing agencies, websites, apps, and tvms MIMIC submodels, respectively, in this section. Several goodness of fit indices of the models are presented in Table
Goodness of fit indices of Multiple Indicator Multiple Cause (MIMIC) submodels.
Fitness Index | Counters | Ticketing Agencies | Websites | Apps | Tvms |
---|---|---|---|---|---|
|
2.596 | 2.763 | 2.875 | 3.529 | 2.578 |
CFI (Comparative Fit Index) | 0.906 | 0.981 | 0.929 | 0.932 | 0.891 |
TLI (Tucker–Lewis Index) | 0.949 | 0.905 | 0.951 | 0.916 | 0.915 |
SRMR (Standardized Root Mean Squared Residual) | 0.040 | 0.039 | 0.048 | 0.049 | 0.039 |
RMSEA (Root Mean Squared Error of Approximation) | 0.076 | 0.040 | 0.068 | 0.762 | 0.790 |
To achieve good model fit, it is recommended that CFI and TLI should be greater than 0.9 [
Effects on latent variables dependent upon demographic characteristics.
Channels | Latent Variables | Age | Gender | Occupation | Income | Education |
---|---|---|---|---|---|---|
Counters |
|
0.119 |
0.022 | −0.035 | −0.098 | −0.008 |
(1.78) | (−0.45) | (−0.50) | (−1.19) | (−0.78) | ||
|
0.055 | −0.040 | −0.039 | 0.012 | −0.009 | |
(0.87) | (−0.91) | (0.86) | (1.32) | (−1.11) | ||
|
0.018 | 0.037 | 0.010 | 0.056 |
0.002 | |
(1.18) | (−0.64) | (0.75) | (1.75) | (−0.08) | ||
|
||||||
Ticketing agencies |
|
0.005 | 0.112 | 0.014 | 0.024 | 0.022 |
(−0.65) | (−1.33) | (1.44) | (0.80) | (−0.99) | ||
|
0.020 | −0.067 | 0.068 |
0.085 | −0.131 | |
(−1.1) | (0.58) | (2.06) | (1.11) | (−1.21) | ||
|
0.018 | −0.098 | −0.049 | 0.005 | 0.074 | |
(1.04) | (0.61) | (−1.27) | (0.59) | (1.30) | ||
|
||||||
Websites |
|
0.014 |
−0.152 | 0.017 | −0.121 | 0.038 |
(1.85) | (−0.38) | (1.40) | (1.22) | (0.56) | ||
|
0.017 |
−0.045 | −0.068 |
−0.024 | 0.044 | |
(−2.02) | (−1.29) | (1.88) | (−0.66) | (1.35) | ||
|
0.045 | 0.004 | 0.049 | 0.083 | 0.029 | |
(1.49) | (0.99) | (1.41) | (−0.92) | (1.92) | ||
|
||||||
Apps |
|
0.101 | 0.005 | 0.014 | 0.005 | 0.007 |
(0.81) | (−0.97) | (1.05) | (−1.11) | (1.20) | ||
|
−0.222 |
0.051 | −0.028 | 0.068 |
0.145 | |
(−2.88) | (0.86) | (−0.58) | (2.51) | (4.90) | ||
|
−0.018 |
−0.001 | 0.025 | 0.092 | 0.127 | |
(1.69) | (1.40) | (0.36) | (−1.43) | (−3.33) | ||
|
||||||
Tvms |
|
−0.083 |
0.016 | 0.009 | −0.004 | 0.014 |
(−1.81) | (−0.39) | (−0.99) | (0.81) | (1.30) | ||
|
−0.132 |
−0.031 | −0.014 | 0.108 |
0.069 | |
(−3.85) | (−0.49) | (0.44) | (3.14) | (−1.17) | ||
|
0.024 | 0.004 | 0.105 | 0.046 | 0.087 | |
(1.59) | (1.09) | (−0.76) | (−1.43) | (5.59) |
Notes:
Table
Though widely investigated in the field of retail consumption, the influence of perceived risk on ticketing choice has not been studied thoroughly before. This paper indicates that the perceived risk seems not to be affected by demographic characteristics, except the age. In the choice of websites, elder passengers have a high perceived risk than the younger, while it is quite the contrary in tvms.
With regard to the effects on the perceived usability, the age is insignificant. Specifically, young passengers who have a higher income tend to feel pleased with the functions provided by apps and tvms. In contrast, elder passengers feel a higher perceived usability of the website channel. Besides, education was also shown to have a significant positive effect on this latent variable in the channel choice of apps. It is implied that operators should improve the design quality of apps for passengers with a lower education background, which make up the majority of intercity bus passengers in Beijing according to our investigation. It should be noted that the occupation does have influence on the latent variable, but the perceived usability of the websites and the ticketing agencies only.
Perceived accessibility is significantly affected by the education towards the choice of online channels. The positive coefficient indicates that better educated passengers will have a higher value of the perceived accessibility. What is more, age still plays an important role of the perceived accessibility in the choice of apps, partly because of the lower popularity of the smart phone among the old people.
MIMIC submodel for service counters.
MIMIC submodel for ticketing agencies.
MIMIC submodel for websites.
MIMIC submodel for apps.
MIMIC submodel for tvms.
It can be seen from Figures
Taking service counters as an example, this has long been the predominant channel for passengers because counters are free of complicated operations, as embodied in the positive high weight of
In the following, both results from a conventional choice model and from an integrated choice and latent variable model are presented, in order to gain insight into the process of passengers’ decision-making and to assess to what extent the inclusion of the latent variable model provides additional explanatory power compared with a classic choice model, as discussed by Heike [
The estimation results of the ICVL model and MNL model without latent variables are presented in Table
Comparison of the estimation results.
Variables | Integrated choice and latent variable model | Multinomial logit (MNL) model | ||||||
---|---|---|---|---|---|---|---|---|
Counters | Ticketing Agencies | Apps | Tvms | Counters | Ticketing Agencies | Apps | Tvms | |
Constant | −1.076 |
−0.132 | 1.531 |
−0.105 | −1.002 | −0.15 | 1.011 | 0.183 |
|
0.819 | 11.55 |
−0.586 | 0.128 |
1.392 | 15.79 |
−0.839 |
0.212 |
|
1.855 |
−0.219 | −1.206 |
0.532 | 0.557 |
−0.598 | 0.851 | 0.022 |
|
1.329 |
2.257 |
−0.281 | −12.769 |
1.032 | 0.409 |
−0.667 |
−0.175 |
|
0.972 | 0.311 | 1.056 |
0.657 | 0.751 |
0.219 |
0.971 | 0.356 |
|
3.197 | 2.988 | 1.336 | 0.715 | 2.187 | 3.15 | 1.159 | 1.76 |
|
−0.196 |
−0.803 |
0.033 | −0.901 |
−0.083 | −3.5 | 1.902 |
0.026 |
|
−0.597 | −0.169 | 0.891 | −0.359 |
−0.081 |
−0.453 |
0.661 |
0.229 |
|
1.794 |
1.208 |
−0.926 |
3.971 | 1.097 | 1.946 | 0.061 | 2.667 |
|
−0.898 | −10.238 |
0.121 |
−0.303 | −1.186 | 9.927 | 0.198 |
−0.664 |
|
0.279 |
0.13 | 1.07 | 0.398 | - | - | - | - |
|
−0.39 | −0.35 |
1.362 |
0.18 |
- | - | - | - |
|
0.17 | 0.271 | 0.117 | 0.138 | - | - | - | - |
|
||||||||
Samples | 1800 | 1800 | ||||||
likelihood | 2981.77 | 1761.29 | ||||||
log-likelihood | −100.87 | −589.26 | ||||||
R2 | 0.435 | 0.379 |
Notes:
Overall, the estimated values of the parameters are in agreement with prior expectations. From the perspective of constant terms, websites seem more attractive than counters and ticketing agencies, but less so compared to apps. Interestingly, in this case, the constants do seem to reflect the preference of bus passengers in Beijing. Indeed, more and more people are using smartphones and the Internet in China.
Not surprisingly, the explanatory variables have a significant impact on ticketing channel choice. Compared with the websites channel, passengers with a low level of education are more likely to choose offline channels, which offer them a more accessible way to obtain necessary assistance. Meanwhile, passengers with a higher income demonstrate a tendency to choose websites or apps; it is also discussed by Cheng [
Table
The results also show that the R2 improves from 0.379 for MNL to 0.435 for the ICLV model. The differences might be due to the absence of latent variables in the MNL model that led to the underestimation of the significance of most of the sociodemographic and trip-specific variables in explaining the preference heterogeneity. Therefore, it can be said that the perception has a significant effect on the ticketing channels choice.
In the last decade, discrete choice models have evolved to include an explicit recognition of psychological factors, enabling us to gain a deeper understanding of the decision processes of individuals. One such model formulation is the ICLV model, which is increasingly being considered in many fields. The aim of this study is to develop a ticketing channel choice ICLV model to analyze intercity bus passengers’ ticketing channel preferences and choices. To the authors’ knowledge, this paper presents the first ICLV model in field of ticketing channel choice behavior for bus passenger transportation.
The data used for the model estimation originate from a survey that took place in Beijing in March 2018, which was administered through both online and paper-based questionnaires. In our research, we include three latent psychological factors to explain ticketing channel choice: perceived risk, perceived usability, and perceived accessibility. The indicators for these constructs are collected in the survey on a five-point Likert ordinal scale. Our study shows that the three latent variables play significant roles in ticketing channel choice. In terms of perceived risk, passengers are more concerned about privacy and anti-fraud, especially for online channels. In terms of perceived usability, time saving and efficiency are primary incentives, while convenience also should be taken into consideration.
The ICVL and MNL model, as a base model, are estimated in this study. The result shows that the
Furthermore, the results indicate that passengers with different sociodemographic and trip characteristics have differing preferences regarding ticketing channels. Generally speaking, young or highly paid people tend to buy tickets through online channels such as apps or websites, while old or poorly educated people tend to choose counters or ticketing agencies.
The findings are very useful for researchers and authorities dealing with bus passenger transportation issues. In particular, a bus passenger transportation operation agency may better understand the real needs of passengers and reallocate relevant resources to the appropriate channels. Although the mobile channel is developing rapidly, the use of offline channels remains the favorite among passengers in this study. The intercity bus operators should offer incentives for passengers to use online channel. At the same time, they must demonstrate protective mechanisms for privacy and financial information to eliminate passengers’ concerns.
Passengers with various sociodemographics have different needs for ticketing channel preference. This implies a need to adopt different marketing strategies so as to precisely and effectively influence ticketing choice behavior. For the younger and female passengers who prefer to purchase ticket online, operators can optimize the function of the ticketing website and provide precision and timely travel information. The elderly passengers and the passengers with low income are more likely to purchase tickets at the service counter and the ticketing agent. Operators should improve the quality and service level of ticket sale to better meet the needs of such passenger groups.
Although this study provides an appropriate modeling approach for dealing with ticketing channel choice behavior, several research directions remain to be explored in future studies. The empirical case was conducted in Beijing, and this research did not emphasize the potential perception differences caused by the various nature of the transportation industry. Future studies are suggested to collect passengers’ related information based on the research area. Moreover, the data used in this study were collected by a questionnaire survey. Some of the variables, such as cost of the fare, knowledge of the system, and current payment options, have not been included in the model. It is recommended that passive data collection technology be used to enhance the model, such as databases collected by ticketing software.
The data used to support this study incorporate personal information of participants. Considering the protection of personal information, the data are partly available from the corresponding author upon request.
The authors declare no conflicts of interest.
The authors appreciate the support from the Beijing Transportation Information Center and the Beijing Liuliqiao road passenger station in research data collection. This research was supported by the National Natural Science Foundation of China (No. 71390332).