Identification of Causal Relationship between Attitudinal Factors and Intention to Use Transportation Mode

Abstract

Based on the theory of planned behavior, this study identifies the causal relationship between attitudinal factors and intention to use transportation mode. A structural equation model was developed based on twelve hypotheses. The main findings and implications of this study are as follows. First, people who want to express themselves through cars have a high intention to use personal vehicles, and they purchase cars for this purpose. If the shared vehicle service provides a vehicle rental that reflects individual tastes, those who want to own the vehicle will use the shared vehicle. This could be a solution to the parking problem. Second, those who perceive travel as a disutility have a low intention to use public transportation. If fare discounts are applied when transferring public transportation and micro-mobility, it is expected that the use of public transportation will increase due to reduction of access time for public transportation. Third, people who like to drive have a high intention to use personal vehicles and micro-mobility. Providing space for driving cars as a leisure activity may be one of the ways to prevent traffic accidents that may occur in the future due to a mixed flow of autonomous vehicles and conventional vehicles.

1. Introduction

As technology advances, transportation modes are gradually diversifying. Micro-mobility such as e-bicycles and e-scooters has been introduced, and new transportation such as urban air mobility is being developed. The diversification of transportation modes provides people with more opportunities to choose from. As the opportunity to choose transportation modes increases, people can consider more factors when choosing one. For example, micro-mobility can act an alternative to walking when traveling a distance of 500 m.

In the transportation mode choice modeling literature, choice behavior is mainly explained by the utility theory calculated based on travel time and travel cost [1,2,3]. However, recent studies have shown that attitudinal factors such as preference, attitude, and perception explain the mode choice model better than socio-economic variables such as gender, age, and income [4]. Cao et al. [5] suggested that the built environment affects travel behavior, and that the built environment is determined according to the individual attitudinal factors. That is, attitudinal factors indirectly affect travel behavior. Likewise, in Ewing and Cervero [6], it was found that the built environment influenced travel behavior. Considering self-selection, the effect of the built environment on travel behavior increased, and it was interpreted that attitudinal factors have affected travel behavior. Choo and Mokhtarian [7] found that attitudinal factors and lifestyle affect vehicle type choice.

Many studies have attempted to identify the relationship between attitudinal factors and behavior, and they have concluded that attitudinal factors are not the only factors that determine behavior [8,9]. Based on this conclusion, the Theory of Reasoned Action (TRA) was established by Fishbein and Ajzen [10] and Trafimow [11]. Later, the theory was embodied in the Theory of Planned Behavior (TPB) by Ajzen [12]. According to the TPB, intentions intervene in the relationship between attitudinal factors and behavior. Therefore, attitudinal factors do not directly affect behavior. For example, an individual can think of the bicycle as a healthy, eco-friendly mode of transportation. This attitude does not increase bicycle use but creates positive intentions for bicycle use. The good intention for bicycles increases bicycle preferences, and eventually, bicycles are used as a means of travel.

In summary, as a new transportation mode is developed, people are likely to choose it according to their tastes of travel. Attitudinal factors have begun to be considered in the mode choice model, and their importance is increasing in the model while extending travel options. According to the TPB, behavior changes according to attitude, but it does not have a direct effect on behavior.

This study attempts to identify the relationship between attitudinal factors and the intention to use transportation mode. The causal structure of attitudinal factors and intentions was established in consideration of the concept of the TPB. The intention to use transportation mode was constructed using activity indicators such as transportation mode preference, travel frequency, and travel time. In addition, transportation was finally classified into three categories, personal vehicle, public transportation, and micro-mobility. Finally, by using the structural equation model (SEM), the causal relationship between the attitudinal factors and the intention to use transportation mode is explored.

2. Research Model Development

2.1. Literature Review

There have been many studies to deal with the influence of attitudinal factors on transportation [13,14,15,16,17]. Atasoy et al. [18] distinguished public transportation and personal vehicle users by using the latent class model. For the analysis, the 20 residents in the Swiss canton of Vaud, Switzland, were surveyed for a week with the GPS data. Three latent variables (PT children, flexibility, and family oriented) were derived through factor analysis (FA) and two clusters were constructed. Based on this, class-specific discrete choice models were used. It was found that travel time, cost, vehicle ownership, and absence of children had a great influence on the mode choice model. Heinen et al. [19] analyzed travel patterns according to the attitudinal factors of bicycle commuters. An online survey of 3500 people in Delft, Zwolle, Midden-Delfland, and Pijnacker-Nootdrop in Netherlands was conducted to derive three attitudinal factors (direct benefits, awareness, and safety). It was found that bicycle users perceived bicycles positively as commuting time increased. Direct benefits such as time saving, and comfort have a great impact on bicycle use. Chen and Li [20] used SEM to derive a public transportation choice model in which attitudinal factors were considered. From the survey of 570 citizens in Chengdu, China, five attitudinal factors (convenience, personal safety, modal comfort, service environment, and waiting feelings) were extracted. The SEM model with the attitudinal factors showed higher fit than other models, revealing that attitudinal factors had a significant influence on public transportation choice. He and Thogersen [21] compared the difference between car use and public transportation use by reflecting attitudinal factors. Also, the influence of attitudinal factors on car purchase was examined. Data surveyed in Guangzhou, China in 2013 were used for analysis, and five attitudinal factors (well-being, functionality, externality, benefits, and effects) were derived through FA. After that, the effect of attitudinal factors on personal vehicle purchase intention was analyzed thorough logistic regression analysis. The attitudinal factors were found to be significant in car use compared to public transportation use. For the purchase intention, a model including attitudinal factors was more appropriate than a model with only socio-demographic variables. Acheampong and Cugurullo [22] conducted a survey of 507 residents of Dublin, Republic of Ireland, to identify the determinants of autonomous vehicle (AV) choice. Six attitudinal factors (affective attitude towards AVs, perceived benefits of AVs, fears and anxiety about AV, image, subjective norm, and attitude towards technology) were utilized in the SEM. AV use is evaluated positively when new technologies are preferred, but concerns about technology have a negative effect on AV use. Asgari and Jin [23] analyzed the intention of additional payments following the introduction of AVs. Based on the data of 878 people in 10 metropolitan cities in the United States, the revealed preference (RP) survey and the stated preference (SP) survey were conducted in 2017. Four attitudinal factors (joy of driving, choice reasoning, trust, and tech-savviness) were derived, and SEM was used to analyze the effect of attitudinal factors on AV choice. It was found that those who liked to drive had a low preference of AVs, and those who were technology savvy preferred AVs. Ashkrof et al. [24] conducted an analysis of the difference in preference for AVs by travel purpose and distance. Based on the online SP survey of 663 Dutch, Netherlands, five factors (trust in AVs, public transport interest, driving interest, positive toward AV efficiency, and eco- friendly) were derived through FA. They developed a logistic regression model for AV use. The model results showed that the greatest influence on AV use was the belief in AV technology, and most factors have positive effects on AV use. But the interest in driving was negatively related to AV use. Goldbach et al. [25] explored the effect of user’s attitudinal factors on the intention to use public AV. A survey of 435 students at Rhine-Waal University in Kleve, Germany, who are relatively familiar with the public AV concept, was conducted to measure three attitudinal factors (trust in AVs, performance expectancy, and social influence). Trust in AVs and performance expectancy were found to be positive for the willingness to use public AV.

The previous studies on the intention to use by transportation were investigated with respect to attitudinal factors. They analyzed the effects of attitudinal factors on personal cars and public transportation [18,20,21]. With the development of autonomous driving technology, research on the effect of attitudinal factors on AVs was also performed [22,23,24,25]. In addition, there was a study that analyzed the effect of attitudinal factors on the use of non-powered mode such as bicycles [19]. Various attitudinal factors were derived as factors affecting transportation mode preference, and the effect of attitudinal factors on transportation mode preference was found to be different depending on transportation.

In this study, the following two additional points that were not considered in previous studies will be applied. The first point is the introduction of new transportation. As technology advances, new transportation may occur, and performance of existing transportation may be improved. In line with these technological advances, research on AVs is being conducted. Since AV is a mode which improves performance upon conventional cars (CVs), it is difficult to say that it is a new transportation mode with difference characteristics from CV. Meanwhile, micro-mobility, which has recently attracted attention as a transportation mode, is a new mode because its characteristics are different from those used in the past. To analyze the effect of attitudinal factors on new transportation modes, we intend to classify transportation mode into personal vehicles, public transportation, and micro-mobility.

The second point is reflection of intention. Intention is an abstract concept and a factor that causes behavior. In most previous studies, intention was investigated as a variable of ‘preference’ and reflected as an observed variable. However, intention is a latent variable formed by abstract factors such as propensity and preference. To reflect these characteristics, the latent variable of intention was derived by using preference, travel time, and number of trips.

2.2. Hypotheses Development

As shown in Figure 1, a conceptual model was constricted, based on the review of the previous studies. Twelve hypotheses developed in the model are as follows.

Hypothesis H1.

The car symbolic is related to the intention to use a personal vehicle.

Hypothesis H2.

The car symbolic is related to the intention to use public transportation.

The car symbolic is to think that a car represents a status symbol. One of the reasons people want to own a car is to express themselves through it [26,27]. These people are likely to own a car. Cars have higher utility than public transportation, except for higher purchase costs [28]. If a person owns a car, the person will want to use a car with high utility, whereas the intention to use a public transportation will decrease [18,21,29,30]. Since micro-mobility is mainly used for short-distance travel, it is irrelevant to the use of cars or public transportation. Therefore, it is considered that there is no correlation between the car symbolic and the intention to use a micro-mobility.

Hypothesis H3.

The negative perception of travel is related to the intention to use a personal vehicle.

Hypothesis H4.

The negative perception of travel is related to the intention to use public transportation.

Hypothesis H5.

The negative perception of travel is related to the intention to use micro-mobility.

The negative perception of travel is to think of travel as disutility. In general, travel is a derived demand [31,32,33]. Thus, travel time is regarded as disutility, and people tend to minimize their travel time [34,35]. Public transportation takes longer to travel than cars because it runs along a fixed route. This will make people who want to reduce travel time more willing to use cars than public transportation [19,36,37]. Those who want to shorten travel time will try to use micro-mobility because micro-mobility is faster than walking.

Hypothesis H6.

The fun-driving is related to the intention to use a personal vehicle.

Hypothesis H7.

The fun-driving is related to the intention to use public transportation.

Hypothesis H8.

The fun-driving is related to the intention to use micro-mobility.

The fun-driving can be explained as person who enjoy driving. Cars and micro-mobility require users to drive themselves, but public transportation does not. People who enjoy driving will prefer modes which the users have to drive themselves [23,24].

Hypothesis H9.

The positive for waiting is related to the intention to use public transportation.

Hypothesis H10.

The positive for waiting is related to the intention to use micro-mobility.

The positive for waiting refers to a person who has no resistance to waiting. Public transportation has a fixed schedule. If the user fails to arrive at the stop at the scheduled time, waiting time occurs. Because of the characteristics of public transportation where waiting time occurs [20,38], people who are positive about waiting will have a higher intention to use public transportation than those who do not. As micro-mobility has recently begun to attract attention due to the expansion of the shared e-scooter market, most e-scooter users are using shared e-scooter services. Shared e-scooter services require rental and return procedures, and if there is no device available, the user must move to another location. This behavior can be considered waiting time to use micro-mobility, so it is thought that people who are not reluctant to wait prefer to use micro-mobility. Since there is no waiting time for personal vehicle use, the attitudinal factors reflecting waiting time are considered irrelevant to the intention to use a personal vehicle.

Hypothesis H11.

The willingness to pay extra is related to the intention to use micro-mobility.

The willingness to pay extra means additional payment intention for better service. Providing better service in transportation is equivalent to providing higher utility transportation, and utilities are calculated by travel time and travel cost [1,2,3]. It can be said that reducing travel time improves the service, so a personal vehicle is a better service than public transportation and micro-mobility is a better service than walking. Users will pay extra to reduce travel time [39], so the willingness to pay extra will increase the intention to use transportation that provides better service. The personal vehicle user must purchase the car, and the cost is included in the travel cost. Since users are sensitive to travel cost [38], it is difficult to interpret the difference in service between personal vehicles and public transportation only as a difference in travel time, considering the high cost of purchasing a car. Therefore, the effect of willingness to pay extra is considered to affect the intention to use micro-mobility only.

Hypothesis H12.

The new technology is related to the intention to use micro-mobility.

The new technology refers to positive attitudes towards new technology. As shared e-scooter services spread, micro-mobility gradually began to be used as a mode of transportation. Since the shared e-scooter service was first introduced in Korea in 2018 [40], micro-mobility can be said to be a new technology. New transportation is attractive to people who favor new technology [22,23,25]. Therefore, micro-mobility will be preferred by those who prefer new technology.

3. Data

We designed a survey to investigate the attitudinal factors of transportation users. The online survey was conducted on 309 people in Seoul, Korea, for three weeks from 16 July to 6 August 2021. To eliminate the bias in the survey, gender, age, and ratio of main transportation were organized in the same way as the 2016 Household Travel Survey conducted by the Statistics Korea. The survey divided transportation mode into fourteen categories: personal car, shared car, town bus, city bus, wide-area bus, subway, taxi, walk, motorcycle, bicycle, shared bicycle, e-scooter, shared e-scooter, and others. A driver’s license is required for users to use cars and e-scooters. The survey was conducted on drivers over the age of 20 because respondents must be able to use all modes.

Considering the characteristics of transportation, transportation was classified into three categories: personal vehicle, public transportation, and micro-mobility. Personal vehicles include only personal cars. Public transportation includes town buses, city buses, wide-area buses, and the subway. Micro-mobility includes bicycles, shared bicycles, e-scooters, and shared e-scooters. Among the samples, cases that never used all three modes were excluded from the analysis. As a result, six cases were removed, and 303 data were finally used for the analysis. The composition ratio of the sample is shown in

Table 1. Descriptive Statistics for the Sample.

Variables		Sample Size	Ratio (%)	2016 Survey Ratio (%)
Gender	Male	190	62.7%	62.7%
	Female	113	37.3%	37.3%
Age	20~29	45	14.9%	14.4%
	30~39	106	35.0%	34.7%
	40~49	85	28.1%	28.0%
	50~59	63	20.8%	22.9%
	60+	4	1.3%	1.3%
Main travel mode	Car	84	27.7%	28.7%
	Bus	99	32.7%	33.2%
	Subway	107	35.3%	36.4%
	Others	13	4.3%	1.7%

. As in the 2016 Household Travel Survey, 62.7% of the respondents were males, and 35.3% used the subway as a major mode of transportation. Those under the age of 19 were excluded from the survey, and those in their 30s accounted for the largest portion with 35.0%.

3.1. Attitudes and Personal Preferences

To analyze individual attitudes, 27 questions were investigated using the 5-point Likert scale method. The questions can be classified into seven categories (car symbolic, negative perception of travel, fun-driving, positive for waiting, willingness to pay extra, new technology, and preference by transportation), and the response results of each question are shown in

Table 2. Respondents’ Attitudes Toward Various Characteristics.

Items	1	2	3	4	5
Car symbolic
CS1: A car represents a driver’s inclination	0.7%	6.6%	28.4%	49.8%	14.5%
CS2: I want to have my car	2.3%	7.6%	31.4%	36.6%	22.1%
CS3: To me, cars aren’t just a vehicle for transportation	4.6%	32.3%	31.4%	28.7%	3.0%
CS4: I am self-conscious about what people think of my car	5.6%	28.7%	35.0%	28.4%	2.3%
Negative perception of travel
NP1: Travel for work is an unproductive activity	2.6%	22.8%	38.6%	28.7%	7.3%
NP2: Travel time itself is a waste of time	5.3%	27.1%	41.9%	22.1%	3.6%
NP3: The trip itself is boring	3.0%	23.8%	41.3%	28.7%	3.3%
NP4: The purpose of the trip is just to arrive at the destination	3.3%	26.7%	24.4%	38.3%	7.3%
Fun-driving
FD1: It is fun to drive	4.3%	17.2%	44.6%	28.4%	5.6%
FD2: I think it is safer for me to drive myself than to use other transportation	5.9%	23.4%	45.5%	21.8%	3.3%
FD3: I prefer driving myself to being in a vehicle driven by someone else	6.9%	21.1%	38.3%	26.7%	6.9%
Positive for waiting
PW1: Waiting time is a short break in a busy day	5.3%	20.5%	46.2%	24.4%	3.6%
PW2: Waiting time is not boring	0.7%	12.2%	30.0%	46.5%	10.6%
PW3: I don’t try to reduce the waiting time	14.5%	57.4%	24.1%	4.0%	0.0%
PW4: Waiting is a good opportunity to get something	2.6%	22.8%	46.9%	24.8%	3.0%
Willingness to pay extra
WP1: I am willing to pay extra for better service	1.0%	7.3%	32.0%	51.5%	8.3%
WP2: I am willing to pay extra to reduce travel time	3.0%	12.2%	39.6%	40.6%	4.6%
WP3: I am willing to pay extra to meet the scheduled Time	1.3%	4.6%	28.4%	53.8%	11.9%
New technology
NT1: When a new product or service is released, I use it before anyone else	6.3%	23.4%	42.2%	24.8%	3.3%
NT2: I am willing to use a service that I have not used in the past	0.0%	8.9%	35.6%	45.5%	9.9%
NT3: I have no problem using the new technology	0.0%	7.3%	34.0%	47.5%	11.2%
NT4: I buy a new product before anyone else	10.9%	42.9%	26.7%	10.2%	9.2%
Preference by transportation
I prefer driving a car to other means of transportation	3.3%	21.1%	42.9%	25.1%	7.6%
I don’t have any reluctance to transfer	4.3%	18.2%	31.4%	33.0%	13.2%
It is not inconvenient to move with strangers	3.6%	18.8%	35.3%	30.7%	11.6%
I prefer to use public transportation	2.6%	9.2%	41.9%	38.0%	8.3%
I prefer micro-mobility	3.3%	20.1%	33.0%	34.3%	9.2%

Note: 1 = Strongly disagree, 2 = Disagree, 3 = Indifferent, 4 = Agree, 5 = Strongly agree.

.

Among the respondents, 64.4% said that cars show driver tendencies, and 58.7% said they want to own a car. Compared to the percentage of respondents who enjoy driving (34.0%), the percentage of people who want to own a car was high. This implies that it is recognized as a positive transportation even though it is inconvenient for users to drive on their own [28]. The percentage of respondents who said that cars are not just a vehicle for transportation (31.7%) is similar to the percentage who said they self-conscious about what people think of their car (30.7%). This means that one of the reasons for using a car is to show off oneself through a car [26,27]. People’s perception of travel was similar, with 25.4~32.3% positive and 25.7~45.5% negative. About half of the respondents (57.1%) perceive waiting positively, which is the influence of portable digital technology and mobile communication development. These developments have increased the positive effect on waiting time by enabling various activities such as watching movies during waiting time [41]. Nevertheless, most people (71.9%) want to reduce their waiting time. As expected, most respondents are positive about paying extra for better service (59.8%) or reducing travel time (45.2%) because travel time is disutility [32,34,35]. About half of the respondents (55.4~58.7%) said they were positive about the use of new technologies. Most of the respondents are not elderly (over 60 years old), so they are used to using new technologies. Finally, six questions were designed for preference analysis by transportation. As a result of the preference response by modes, personal vehicles were 32.7%, public transportation was 43.5%, and micro-mobility was 43.5%, which was lower in preference for personal vehicles than other modes.

3.2. Characteristics of Transportation

To understand the characteristics of respondents’ use by transportation mode, the number of trips and travel time for three modes were investigated.

Table 3. Descriptive Statistics by Transportation.

Variables	Sample Size	The Number of Trip (trip/week)			Total Travel Time (min/week)
		Avg	Min	Max	Avg	Min	Max
Total Mode	303	10.78	1.00	70.00	420.45	10.00	2373.00
Personal vehicle	215	4.00	1.00	50.00	183.91	10.00	1000.00
Public transportation	267	8.11	1.00	56.00	302.79	2.00	2100.00
Micro-mobility	72	3.36	1.00	20.00	97.35	1.00	540.00

represents the result of aggregating the number of trips and travel times for a week and classifying them by transportation. The number of users by mode per week was 215 personal vehicles, 267 public transportation, and 72 micro-mobility. Since transfer can occur in public transportation, the average number of trips per week by modes is the highest at 8.11 in public transportation. The average weekly travel time by modes was the longest for public transportation and the shortest for micro-mobility. Public transportation travels along a fixed route, and personal vehicles travel along the shortest route. In addition, public transportation stops at designated stops even if it is not the user’s destination. Due to these characteristics, public transportation has a longer travel time than personal vehicles. Micro-mobility is responsible for short-distance travel, such as first/last mile travel [42,43]. As a result, it was found that the travel time was shorter because the travel distance was shorter than other transportation.

4. Methodology

The purpose of this study is to identify the causal relationship between attitudinal factors and the intention to use each transportation mode. A total of twelve hypotheses were constructed based on the results derived from the previous study. Before proving the hypotheses, we derived six attitudinal factors using exploratory factor analysis (EFA). A SEM was used to verify twelve hypotheses based on the derived attitudinal factors.

Structural Equation Model

SEM is one of the representative statistical techniques used to verify complex causal relationships and correlations between various variables [44,45]. Not only the relationship between several exogenous variables and endogenous variables, but also causal relationship between endogenous variables can be estimated at the same time. In addition, since correlation between error terms of exogenous variables or correlation between error terms of endogenous variables can be assumed, individually separated regression models can be analyzed at once.

SEM consists of two components: a structural model and a measurement model. If there is no latent variable in the model, the SEM can only be expressed as a structural model. In this study, both the structural model and the measurement model are included as attitudinal factors are reflected as latent variables. A SEM containing latent variables can be expressed by Equation (1).

$${\eta }_i={\alpha }_{\eta }+\mathsf{{\rm B}}{\eta }_i+\mathsf{\Gamma }{\xi }_i+{\zeta }_i$$

(1)

In this equation, the subscript i means the ith case, ${\eta }_i$ means a vector of latent endogenous variables, ${\alpha }_{\eta }$ means the vector of intercepts, $\mathsf{{\rm B}}$ means a coefficient matrix that gives the expected effect of ${\eta }_i$ for ${\eta }_i$ with a main diagonal matrix of zero, ${\xi }_i$ means a vector of latent exogenous variables, $\mathsf{\Gamma }$ means a coefficient matrix that gives the expected effect of ${\xi }_i$ on ${\eta }_i$, and ${\zeta }_i$ means a vector of equation [46].

The endogenous variables and the exogenous variables can be formulated as shown in Equations (2) and (3), respectively.

$$y_i={\alpha }_y+{\mathsf{\Lambda }}_y{\eta }_i+{\epsilon }_i$$

(2)

$$x_i={\alpha }_x+{\mathsf{\Lambda }}_x{\xi }_i+{\delta }_i$$

(3)

In these equations, $y_i$ means a vector of indicator ${\eta }_i$, ${\mathsf{\Lambda }}_y$ means a factor loading matrix that gives the expected effect ${\eta }_i$ on $y_i$, ${\epsilon }_i$ means a vector of disturbances that is not explained by ${\eta }_i$, $x_i$ means a vector of indicator ${\xi }_i$, and ${\delta }_i$ means a vector of disturbances that consists of all the other influences on $x_i$ except for the effects of ${\xi }_i$ [46]. The SEM has a structure in which Equations (2) and (3) are combined with Equation (1).

The measurement model has the following assumptions: The means of disturbances (E[${\epsilon }_i$], E[${\delta }_i$]) are zero, each disturbance is uncorrelated with each other, and each disturbance is uncorrelated with the latent exogenous variables [46]. The mathematical detailed formulas of the SEM are not specified in this study, and the basic concepts and mathematical formulas can be found in Mueller [47] and Byrne [48].

5. Result

5.1. Exploratory Factor Analysis

EFA was performed through the statistical package SPSS to derive attitudinal factors for verifying hypotheses. Varimax rotation, commonly used for analysis [49], is used, and it is a good choice if there is no correlation between factors [50]. The analysis results are shown in

Table 4. Result of the Exploratory Factor Analysis.

Variables		1	2	3	4	5	6
Car symbolic	CS1	0.767
	CS2	0.618
	CS3	0.675
	CS4	0.658
Negative perception of travel	NP1		0.828
	NP2		0.829
	NP3		0.621
	NP4		0.603
Fun-driving	FD1			0.751
	FD2			0.733
	FD3			0.768
Positive for waiting	PW1				0.825
	PW2				0.731
	PW3				0.646
	PW4				0.816
Willingness to pay extra	WP1					0.748
	WP3					0.767
	WP4					0.750
New technology	NT1						0.778
	NT2						0.762
	NT3						0.643
	NT4						0.816
Cronbach $\alpha$		0.737	0.758	0.706	0.761	0.0.761	0.767
Kaiser-Meyer-Olkin		0.715
Bartlett’s test of Sphericity		1616.796 (p < 0.001)

. To verify the model, we performed a Bartlett’s test of sphericity (BTS) and Kaiser-Meyer-Olkin (KMO) analysis. If the null hypothesis is rejected in BTS, the model can be interpreted as suitable [51,52], and in the case of KMO, it can be interpreted as appropriate if the KMO value is 0.5 or more [53]. As a result of EFA, BTS was 1442.208 (p-value < 0.001) and KMO was 0.715, so the EFA results were statistically appropriate.

In Table 4, the display format was set to suppress when the factor loading value was less than 0.5. The results identified six factors and have at least three eigenvalues. Six factors can explain 61% of all variables, and all factors are reliable because the Cronbach alpha for each factor is greater than 0.7 [54].

5.2. Goodness-of-Fit of SEM

SEM was conducted based on six attitudinal factors derived by EFA. For the analysis, STATA 16.0, a statistical package program, was used. Maximum likelihood estimation is widely used in SEM due to its advantages of easy interpretation [55,56,57], and our study also used maximum likelihood estimation.

The goodness-of-fit of the model was explored. The model fit of the SEM can be examined with absolute fit, incremental fit, and parsimonious fit. The absolute fit is an indicator of how well the model predicts the sample covariance matrix based on the covariance matrix theory. Mainly used indicators include Chi-square (${\chi }^2$), Goodness-of-Fit Index (GFI), Root Mean Square Residual (RMR), and Root Mean Square Error of Approximation (RMSEA). The incremental fit is a variable that indicates how good the study model is to measure compared to the null model. Incremental Fit Index (IFI), Normal Fit Index (NFI), Relative Fit Index (RFI), Tucker-Lexis Index (TLI), and Comparative Fit Index (CFI) are used as incremental fit indicators. The parsimonious fit considers the complexity of the model and uses it to compare the fit between different models. It is represented by Parsimony Ratio (PRATIO), Parsimonious GFI (PGFI), Parsimonious NFI (PNFI), and Parsimonious CFI (PCFI) [44,45,48,58]. In this study, parsimonious fit was not reviewed because no comparisons were made between models, and the fit of the model was investigated using absolute and incremental fit.

We used RMSEA as an absolute fit measure. The basic indicator of the absolute fit is Chi-square, but it is greatly affected by the size of the sample. It is easy to reject the null hypothesis that “the model is appropriate” when there are more than 200 samples [45,48]. In this study, the Chi-square is not appropriate because there are more than 200 samples analyzed. Therefore, we adopted RMSEA, which has the advantage of being less affected by the size of sample. RMSEA is considered a good model when it is less than 0.05 [44,58]. The incremental fit represents the improvement of the proposed model over the null model between 0 and 1, and the closer to 1, the better. The basic indicator of the incremental fit is NFI, but as recent studies [43,58] recommend TLI and CFI, TLI and CFI were used instead of NFI. When the TLI and CFI of the model are over 0.9, the model is considered acceptable [43,58]. The SEM model is statistically significant as a result of analyzing the goodness-of-fit of the model designed in this study, with the RMSEA value of 0.035, the TLI value of 0.908, and the CFI value of 0.922.

5.3. Hypotheses Test

According to the TPB, attitudinal factors do not directly affect behavior, but attitudinal factors change intention, and intention causes behavior [12]. Based on this theory, the causal relationship between attitudinal factors and intention to use transportation mode was established. The intention to use each transportation mode was derived through mode preference, number of trips, and travel time by referring to previous studies [59,60]. SEM was performed through twelve hypotheses to analyze the causal relationship between attitudinal factors and the intention to use transportation mode, and the results of the standardized coefficient are shown in Figure 2.

People with Car symbolic are shown to have a high intention to use personal vehicles ($\beta$ = 0.23, p < 0.10) and low intention to use public transportation ($\beta$ = −0.42, p < 0.10), supporting H1 and H2. These people want to own a car to express themselves through the car [26,27]. Moreover, owning a car increase car use and reduces public transportation use [13,21,29,30]. Therefore, the Car symbolic increases the intention to use a personal vehicle (PV) and lowers the intention to use public transportation (PT) by generating car purchases.

Negative perception of travel has a positive effect on PV ($\beta$ = 0.30, p < 0.01) and the intention to use micro-mobility (MM; $\beta$ = 0.25, p < 0.01), while it has a negative effect on PT ($\beta$ = −0.56, p < 0.01), supporting H3, H4, and H5. Unlike other transportations, public transportation runs along a fixed route. It also stops at all stops, even if it is not user’s destination. This makes public transportation spend more time than personal vehicles when traveling to the same destination. A person who feels negative about travel recognize it as disutility and tries to reduce travel time [36,37]. As transportation with short travel time is preferred, PV increases and PT decreases. Micro-mobility is used as an alternative to walking because it is used for relatively short distances [42,43,61]. It is also effective in reducing short distance travel time because it is faster than walking [19]. According to this characteristic, if travel is considered negatively, MM increases.

As expected, fun-driving increases PV ($\beta$ = 0.78, p < 0.01) and MM ($\beta$ = 0.46, p < 0.01), but reduces PT ($\beta$ = −0.44, p < 0.01) which support H6, H7, and H8. Public transportation cannot be driven by users, but personal vehicles and micro-mobility can be driven by users. In a study on the preference of AVs and CVs, it was found that people who like driving prefer CVs to AVs [23,24]. Likewise, people who enjoy driving prefer the mode in which they can drive.

The more positive people feel about waiting, the greater the PT ($\beta$ = 0.35, p < 0.01) and MM ($\beta$ = 0.12, p < 0.10), which supports H9 and H10. Due to the nature of operating in a fixed schedule, waiting time is inevitable for public transportation. Accordingly, the more positive people feel about waiting time, the higher the preference for public transportation [20,38]. Since the behavior of increasing public transportation preference is the result of an increase in the intention to use public transportation, it can be said that those who are positive about waiting time have a greater intention to use public transportation than those who do not. The most representative micro-mobility is the e-scooter. As the shared e-scooter market has recently expanded around the world [62,63,64], many people are using shared e-scooters, and most micro-mobility users are shared e-scooter users. When using a shared e-scooter, rental and return must be performed, which can be seen as waiting time. Walking, a mode that competes with micro-mobility, has no waiting time. Considering this, it can be concluded that the more positive people feel about waiting time, the higher the PT.

If there is a willingness to pay extra, MM appears to increase ($\beta$ = 0.19, p < 0.10), supporting H11. Those who want better service even if they pay extra choose roads that offer fast and comfortable driving environment even if the toll fee is high [39]. Likewise, micro-mobility costs money to use, but can move faster and more comfortably than walking.

People who like to use new technology were found to have high MM ($\beta$ = 0.13, p < 0.10), supporting H12. The invention of micro-mobility is old, but it is drawing attention with the recent launch of the shared e-scooter service. In the United States, the shared e-scooter service was first introduced in 2017, and in Korea, it was first introduced in 2018 [40]. Only recently have users started using e-scooters, recognizing them as new technologies. This perception makes people who prefer new technology prefer micro-mobility. This is similar in research on AVs. In a study that analyzed the preference for using AVs and CVs, it was found that the more new technologies were preferred, the more AVs were preferred than CVs [22,23,25].

6. Conclusions

6.1. Discussion and Implications

In this study, we use a SEM to understand how various attitudinal factors relate to the intention to use transportation mode. Data of 303 people in Seoul, Korea, were used for analysis in 2021, and transportation was divided in to three categories (personal vehicle, public transportation, and micro-mobility). The findings of this study provide the following insights.

First, people who desire to own a car have high intention to use personal vehicles. They try to use personal vehicles to express themselves, and buy a car for this reason. There are many problems due to the lack of parking spaces in the city center [65,66]. If a person who wants to own a car purchases a car and uses it, the parking problem will inevitably become serious as the total number of cars increases. To address this problem, there is a way to promote the use of shared vehicles [67,68]. Shared car services are gradually developing. Initially, only vehicle types could be selected, but with service development, it becomes possible to select driving assistance technologies such as navigation and cruise driving systems. However, these services still provide only technical options and cannot be selected in areas where individual characteristics can be expressed, such as vehicle color and interior lighting. If options can be reflected in areas where individual characteristics can be expressed, users who want to purchase cars as status symbols are expected to become customers of the shared vehicle market.

Second, people who think of travel as a disutility have a high intention to use personal vehicles and micro-mobility, but low intention to use public transportation. They want to use personal vehicles rather than public transportation and micro-mobility rather than walking because they want to reduce travel time. Generally, transportation aims to reduce the use of personal vehicles and promote the use of public transportation to ease traffic congestion. One way to promote public transportation is to reduce public transportation travel time. The introduction of a discount on transfer fees between public transportation and micro-mobility is expected to shorten public transportation access time and promote public transportation use.

Third, people who like to drive have a high intention to use personal vehicles and micro-mobility. Technological advances will lead to the introduction of AVs, and in the future, most vehicles will eventually turn into AVs. However, there will still be people who want to drive their cars themselves, and these people will want to use CVs rather than AVs [23,24]. This leads to a mixed flow of AVs and CVs in the future. Since the mixed flow of AVs and CVs negatively affects safety and traffic flow [68,69], the mixed flow should be prevented. Providing a space to enjoy driving is expected to prevent a mix of AVs and CVs. And this can promote the use of micro-mobility, considering that most of the purpose of using e-scooters is leisure [61,62,63].

6.2. Limitations and Future Research

This study has some limitations due to data availability. The data in the study consisted of 303 samples, sufficient for data analysis, but small to reflect various variables. For example, less than 10 respondents used the shared car, so it was insufficient to set shared car as a mode when classifying transportation. If more samples are investigated, it is expected that more detailed mode classification such as classification of transportation according to sharing and ownership will be possible. In addition, socio-demographic factors such as age, gender, and income are likely to affect the intention to use transportation mode as well as the attitudinal factors. Thus, a better model can be developed if socio-demographic variables are considered together in the future study.