Understanding Commuters’ Willingness to Shift to Transfer-Type Buses Using a Latent Class Model ^†

Abstract

The Korean government proposes introducing a transfer-type bus system to reduce urban congestion. Transfer-type buses turn around at the Seoul border, requiring passengers to transfer to other modes to reach downtown. These buses have shorter routes, allowing reduced headways and increased bus supply. While this approach reduces congestion in the downtown area, it may cause transfer resistance, making it essential to analyze willingness to shift (WTS). This study uses a latent class model to categorize potential interregional bus users into three types: transfer avoidance, cost-sensitive, and time-sensitive. Over 50% of users in each group express WTS, showing a positive response to the transfer-type bus introduction. The choice model results indicate that the travel time and cost of direct type buses affect WTS, suggesting that policies should consider these factors for effective implementation.

1. Introduction

As urban areas grow, the increase in commuting populations has led to significant congestion in downtown areas, especially in Seoul, South Korea, due to urban expansion. Efforts to alleviate congestion have included the implementation of congestion charges, but challenges persist with the high demand for interregional buses during peak times, leading to increased congestion and decreased competitiveness of public transportation [1]. In 2019, the Korean Ministry of Land, Infrastructure and Transport proposed a reorganization of the bus system, introducing direct-type buses, which enter downtown directly, and transfer-type buses, which require passengers to shift to other transportation modes at transfer centers located at the city borders [2]. This measure aims to reduce downtown congestion, shorten bus routes, and improve operation frequency and efficiency. However, the success of this initiative depends on passengers’ willingness to shift (WTS) to this new system despite potential resistance due to the inconvenience of transferring. Therefore, it is important to analyze users’ WTS before introducing the transfer-type bus system. Previous studies have discussed the concept and importance of WTS [3,4,5].

2. Methodology

This study was developed to estimate WTS by considering the latent preferences of interregional bus users. A stated preference (SP) survey was conducted with 502 interregional bus users in the Seoul Metropolitan Area (SMA), consisting of a latent preference survey and a bus choice survey. Using the latent preference survey data, latent class analysis (LCA) was performed to classify users based on their latent behavioral preferences, and the WTS for each class was estimated. A choice model was then developed to compare direct-type buses, which enter downtown directly, and transfer-type buses, which require transfers at the city boundary. The model’s reliability was verified.

Previous studies on WTS mainly relied on the binary logit model, which assumes that all individuals share the same utility function, limiting its ability to capture user diversity [6]. To address this, this study adopted the latent class model (LCM), which accounts for taste heterogeneity (TH) by assuming the existence of different latent classes with distinct choice utility functions [7]. The LCM consists of two components. The membership model uses LCA to categorize users into distinct classes based on shared characteristics and determine their probability of belonging to a specific class. The class-specific choice model estimates the probability of individuals selecting a transportation alternative within each latent class.

3. Results and Discussion

The three latent classes are estimated using LCA. Results show the probability of users belonging to each class based on their responses to the latent class survey questions. Class 1 was defined as the transfer-avoidance type because it exhibited negative characteristics related to transfer items and included 9.16% of users. Class 2 was defined as the cost-sensitive type due to its positive characteristics concerning cost items, comprising 41.24% of users. Class 3 was defined as the time-sensitive type as it showed common characteristics regarding time items, with 49.60% of users belonging to this class. The study analyzed the factors influencing the choice of transfer-type buses for each class of users, which are presented in

Table 1. Estimation results of the model.

Coefficient			Parameter	Standard Error	p-Value
Class 1	Variables	${\mathrm{X}}_{\mathrm{r}\mathrm{e}\mathrm{s}}$	−2.002	0.561	0.000
		${\mathrm{X}}_{\mathrm{t}\mathrm{p}}$	1.120	0.348	0.001
		${\mathrm{X}}_{\mathrm{t}\mathrm{t}\mathrm{b}\_\mathrm{s}\mathrm{u}\mathrm{p}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{t}}$	1.135	0.332	0.001
		${\mathrm{X}}_{\mathrm{a}\mathrm{g}\mathrm{e}\_20}$	0.908	0.233	0.000
		${\mathrm{X}}_{\mathrm{t}\mathrm{t}\mathrm{c}\_\mathrm{D}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}}$	0.031	0.017	0.070
	Summary statistics	Cox and Snell ${\mathrm{R}}^2$	0.133
		Nagelkerke ${\mathrm{R}}^2$	0.177
		−2 Log-likelihood	456.632
		Observation	368
Class 2	Variables	${\mathrm{X}}_{\mathrm{r}\mathrm{e}\mathrm{s}}$	0.307	0.183	0.094
		${\mathrm{X}}_{\mathrm{i}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{e}\_\mathrm{l}\mathrm{o}\mathrm{w}}$	0.287	0.117	0.014
		${\mathrm{X}}_{\mathrm{c}\mathrm{b}\mathrm{p}\_\mathrm{f}\mathrm{a}\mathrm{r}\mathrm{e}}$	0.452	0.107	0.000
		${\mathrm{X}}_{\mathrm{j}\mathrm{o}\mathrm{b}\_\mathrm{s}\mathrm{e}\mathrm{l}\mathrm{f}}$	0.836	0.327	0.011
		${\mathrm{X}}_{\mathrm{t}\mathrm{t}\mathrm{c}\_\mathrm{D}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}}$	0.033	0.008	0.000
	Summary statistics	Cox and Snell ${\mathrm{R}}^2$	0.029
		Nagelkerke ${\mathrm{R}}^2$	0.040
		−2 Log-likelihood	2,109,358
		Observation	1656
Class 3	Variables	${\mathrm{X}}_{\mathrm{t}\mathrm{t}\mathrm{b}\_\mathrm{s}\mathrm{u}\mathrm{p}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{t}}$	0.588	0.130	0.000
		${\mathrm{X}}_{\mathrm{c}\mathrm{b}\mathrm{p}\_\mathrm{f}\mathrm{a}\mathrm{r}\mathrm{e}}$	0.186	0.098	0.058
		${\mathrm{X}}_{\mathrm{j}\mathrm{o}\mathrm{b}\_\mathrm{s}\mathrm{t}\mathrm{u}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{t}}$	0.720	0.172	0.000
		${\mathrm{X}}_{\mathrm{t}\mathrm{t}\mathrm{c}\_\mathrm{D}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}}$	0.033	0.015	0.024
	Summary statistics	Cox and Snell ${\mathrm{R}}^2$	0.028
		Nagelkerke ${\mathrm{R}}^2$	0.037
		−2 Log-likelihood	2,686,118
		Observation	1992

.

For Class 1 (transfer-avoidance type), individuals were more likely to choose transfer-type buses if they lived in Seoul, made non-commute trips, supported the introduction of transfer-type buses, perceived high travel costs for direct-type buses, or were in their 20s. Despite their tendency to avoid transfers, they are willing to shift if they believe it will help reduce downtown congestion and if the cost of transfer-type buses is lower than that of direct-type buses. For Class 2 (cost-sensitive type), individuals were more likely to shift if they lived in the outskirts of Seoul, had a lower income (below USD 28,780), supported the new bus system, perceived high fares as a key issue, were self-employed, or found direct-type buses expensive. This class is highly sensitive to cost, which significantly influences their transportation choices. For Class 3 (time-sensitive type), higher willingness to shift was observed among students, those who use interregional buses to save time, individuals who support the transfer-type bus system, and those who perceive direct-type buses as too slow. This class prioritizes travel time, leading to the identification of time-related variables.