An Analysis of Electric Vehicle Charging Intentions in Japan

Abstract

This study focuses on charging-related decisions for fast charging at highway service and parking areas, slow charging at home, fast charging at commercial facilities, and fast and slow charging at workplaces. This research contributes to the existing literature by estimating the charging behavior variables, as well as understanding the role of explanatory variables in influencing charging-related decisions. Responses from the stated preference (SP) survey in Japan in 2021 were analyzed with a mixed logit model (MXL). The results showed that, (1) when the battery level is 75% or higher, users of battery electric vehicles (BEVs) are not keen to charge their vehicles, but when the next trip is anticipated to be 50 or more kilometers, they choose to charge their vehicles; (2) individuals are not willing to tolerate any waiting time for their vehicles to be charged at each location; and (3) the recurrence of charging at the target location affects the charging decision of BEV users. We found significant relationships between socioeconomic characteristics and charging decisions. Furthermore, we examined the practical applications of the empirical findings in this study for policymaking and charging infrastructure planning.

1. Introduction

The ongoing increase in global temperatures has prompted governments and environmental scientists to exert pressure on car manufacturers to transition their product portfolio towards more sustainable vehicles [1]. The transport sector accounts for approximately 16% of global greenhouse gas emissions, with a significant portion attributed to road transport [2]. Electromobility emerges as a viable interim solution to address this problem by presenting an opportunity to significantly reduce emissions, as it does not depend on petroleum products [1], and EVs play a crucial role in efforts to achieve decarbonization in the transportation sector and fulfill sustainable policy objectives [3]. EVs contribute to reducing dependence on oil, lowering global and local emissions, and mitigating noise pollution, surpassing the environmental performance of conventional vehicles [4]. Additionally, EVs have several advantages such as higher energy efficiency, lower operational costs, and faster acceleration. Consequently, EVs are perceived as a good alternative to internal-combustion-engine (ICE) vehicles with great possibilities [5].

Generally, EVs are classified into five categories based on their fuel requirements: fuel cell electric vehicles (FCEVs), plug-in hybrid electric vehicles (PHEVs), hybrid electric vehicles (HEVs), range-extended electric vehicles (REEVs), and battery electric vehicles (BEVs). Typically, HEVs have an electric engine and an ICE. An electric engine helps an ICE increase its speed. In HEVs, the electric motor is internally charged by an ICE. A PHEV also has an electric motor and an ICE; however, unlike an HEV, the electric motor is charged externally. Similarly, a REEV has a generator (extender) and electric motor. The extender simply functions as a power generator instead of directly accelerating the wheels. FCEVs run on electricity; that is, fuel unit stacks generate energy. A BEV is exclusively driven by electrical energy and motors. In this study, we focused only on BEVs, which are referred to as EVs.

Increasing the adoption of EVs will be crucial for taking significant strides towards reducing transportation emissions [6]. EV sales in 2021 surpassed 6.6 million units globally; compared to the years between 2018 and 2020, the sale of EVs in the last two years has exponentially increased [7].

Several countries have made impressive strides in transitioning to EVs. The top countries leading in the EV market share are Norway, where all-electric vehicles accounted for 80% of passenger vehicle sales in 2022, followed by Iceland (41%), Sweden (32%), the Netherlands (24%), and China (22%). Norway, in particular, holds the global leadership position, achieving remarkable growth from less than 1% to 80% EV sales within 12 years. Meeting climate targets necessitates scaling up EV sales at a similar pace to Norway’s achievements. However, not all countries possess the same level of wealth as Norway or the market power and government structure of China. Nonetheless, electric vehicles can present economic and environmental advantages for a broad range of developing countries. Currently, 16 countries, including Canada, Japan, and the United Kingdom, have implemented policies mandating 100% EV sales by 2035 or even earlier [6].

The successful adoption of EVs relies heavily on the implementation of supportive policies and incentives [8]. Comprehensive planning and the development of charging infrastructure are essential to facilitating the widespread use of EVs [9]. The Japanese government has established a target of achieving full electrification for new passenger vehicle sales by 2035. Their objective is to create a society that boasts highly convenient and sustainable EV charging infrastructure, aligning with global standards [10].

BEVs have been hailed as one of the most promising means of achieving sustainable urban development because of their ability to minimize reliance on fossil fuels and tailpipe emissions [11]. The limited range and protracted recharging times pose considerable obstacles to the widespread adoption of EVs. Long journeys also necessitate caution when deciding in advance on charging stations to be utilized to prevent them from running out of power. For example, in 2019, inexpensive EVs typically covered a distance of approximately 250 km. Due to their restricted range, many recharge breaks are required for lengthy trips [12]. BEV users are concerned about the accessibility of charging infrastructure along travel routes, particularly for long-distance journeys [5]. A significant disparity has been observed between the demand for public charging stations and their actual availability, leading to low utilization rates despite high charging demand. This mismatch can be attributed to inaccurate estimations of charging demand and an ineffective deployment of public charging facilities, including issues related to their placement and charging capacity [13]. Therefore, it is important to gain a deeper understanding of the charging choice behavior of BEV users to more accurately assess charging needs. For this purpose, research studies have employed methods, such as revealed preference (RP) studies [9,10,11,12] and stated preference (SP) studies [11,13,14,15,16], to investigate the charging choice behavior of EV users. Many factors including charging duration [13,14,16,17], charging speed [16], the state of charge (SOC) [9,10,12,16], battery capacity [5], and the distance of the next trip [11,12] affect charging choice behavior. Moreover, there is considerable heterogeneity in the relationship between charging choice decisions and influencing factors among individuals.

Studies indicate that public charging stations and corridor charging stations are among the least utilized charging infrastructures [17], which can be attributed to the historically limited range of EVs and the resulting low likelihood of undertaking long-distance trips [18]. Driving behavior, charging behavior, and battery type are the primary factors influencing the charging load [19]. However, the existing literature on charging behavior lacks in-depth studies, resulting in models that assume predetermined charging behaviors when analyzing the integration of EV use and the power system rather than considering heterogeneity among drivers [20].

This research contributes to the existing literature by investigating the relationship of charging decisions with charging location (normal and fast charging), remaining battery level, and expected travel distance (km). In our study, explanatory variables are limited to the socioeconomic characteristics of BEV users, multiple charging locations, including fast- and normal-charging stations, and acceptable waiting times at charging stations. For this purpose, a stated preference survey was distributed in the Kanto and Chubu regions of Japan in 2021, obtaining a sample of 441 respondents, with the resulting data analyzed with a mixed logit model to understand under which conditions EV drivers choose to charge or not. The findings are useful for policymakers as well as public and private charging station suppliers for its association of socioeconomic characteristics with charging intentions, and this research is a first step in developing recommendations regarding localization and the power of charging infrastructure.

The remainder of this paper is organized as follows: the following section reviews past research on charging infrastructure and consumer preferences for charging behavior in order to examine the need for this study in the Japanese market. Section 3 and Section 4 explain the survey design, dataset, and modeling approach. The descriptive statics, results, and policy implications are presented in Section 5, Section 6, and Section 7, respectively. The final section concludes the paper.

2. Literature Review

2.1. Charging Infrastructure

Generally, EVs are charged in two ways: normal charging (also known as slow charging) and fast charging (also known as quick charging) [16]. Normal charging, which is performed at 120 or 240 volts, is considered the most frequent and can be performed at home and other locations. By contrast, fast charging, which can be performed at 480 volts or at higher voltages, plays a significant role in extremely long outings, or when an unexpected crisis emerges, and is usually performed at the charging station. In Japan, a field trial was conducted for BEVs, and it was observed that cars rarely required fast charging daily; however, over some time, all owners wanted fast charging. Japan has made consistent investments in charging infrastructure to support pure EVs. The widely adopted CHAdeMO fast-charging standard, used in numerous fast-charging stations, has seen approximately 6000 installations in Japan. This translates to an impressive ratio of 5.9 fast chargers per 100 vehicles compatible with fast charging in the country [18].

A critical requirement for the widespread adoption of EVs is the presence of a robust and extensive charging infrastructure network. Over the last decade, there has been a significant increase in the total number of public charging stations. Between 2018 and 2020, the number of chargers more than doubled, growing from 0.5 million to 1.2 million. Subsequently, between 2020 and 2022, there was another substantial increase from 1.2 million to 2.6 million chargers. China accounts for nearly two-thirds of all installed chargers. The United States and South Korea each contribute 6%, followed by Norway with almost 5%, and France and Germany with approximately 3% each. However, in many countries, the availability of public charging stations remains limited [21]. The availability of publicly accessible fast chargers plays a crucial role in facilitating longer journeys for EVs. In Europe, the number of public fast chargers has risen by over 30% to nearly 50,000 units. The United States boasts around 22,000 fast chargers, with Tesla superchargers accounting for nearly 60% of them. Korea has seen a 50% increase in fast chargers compared to 2020, with a total of 15,000 units [7].

There also exists an interaction with EV adoption and infrastructure [22], suggesting that, for the EV market, it is essential to have a widespread infrastructure due to longer refueling times, but a significant portion of this requirement can be addressed by the convenience of recharging EVs at home. Governments play a vital role in this process by facilitating investments and reducing barriers that may hinder the widespread deployment of charging infrastructure.

2.2. Charging Behavior

There have been many empirical studies on EV charging behavior. Potogolu et al. [23] conducted a comprehensive review assisting policymakers in understanding the public charging preferences of potential and current EV users, which aims to facilitate the implementation of effective government policies that support the expansion of public charging infrastructure and encourage the widespread adoption of EVs. Battery capacity, midnight indicator, initial state of charging (SOC), and several past charging events are the parameters that determined the charging choice behavior and location, using the mixed logit model (MXL) among 500 BEV users in Japan [11]. The likelihood of normal charging at home or in a company with vehicle kilometers of travel duration increases in commercial BEVs compared with private BEVs. Similarly, both private and commercial BEV users exhibit heterogeneity in their charging preferences based on factors such as working and non-working days, as well as the decision to opt for normal charging or not to charge after their last trip [24]. In Australia, however, the adoption of EV charging is expected to generate a new peak in electricity demand during the early evening hours [25]. Occasionally, EV users choose charging behavior concerning the charging time, such as normal charging after the end of the day and nighttime charging. EV charging behavior is heterogeneous among users. In the same conclusion, heterogeneities were observed in driving and charging behaviors using a machine learning approach on 50 BEV drivers in China [26]. BEV drivers demonstrated conservative charging behavior as they looked for an appropriate opportunity, regardless of the remaining range, and preferred charging during the daytime, whereas practical factors such as charging fee, parking time, excess range, queues at the charging stations, age, income, and driving experience were the factors that affected charging behavior in China [27].

The most important locations for charging are home, work, and public venues [17], in line with the study by Hoen et al. [28], which found that these locations provide great flexibility for charging, using a Norwegian dataset of 465 EV users. The findings of Latinopoulos et al. [29] indicate that out-of-home charging is more prevalent among individuals who have access to charging facilities at their workplace or when charging is available for free. In another work, in Germany, neighborhood charging locations were more popular among EV users, including parking lots with charging facilities [30]. Likewise, because of incentivization, charging at home in the evening is preferred in Ireland [31]. However, uncertainty regarding the accessibility of charging stations near one’s residence, such as through daytime charging, diminishes the intention to purchase EVs [32]. EV users prefer semi-fast charging (22 kW AC) and normal charging (3.7 kW AC) for frequently used charging stations, while fast charging (50 kW DC) is desired for stations that are less frequently utilized [33]. Additionally, the ratio of public fast-charging stations to BEVs may be similar to that of other alternative fuels (approximately one fast-charging station for every 1000 vehicles with high power rates of 150 kW) [34]. In terms of costs of charging, curb-side charging infrastructure costs more and can be replaced by other charging locations, such as workplaces and shopping centers in Germany [35]. Another study by Corinaldesi et al. [36] demonstrated how the unmanaged charging of EVs led to high power peaks, resulting in elevated overall costs, and how the implementation of a managed charging strategy showed promising results, with the Austrian office site’s parking lot experiencing a reduction in costs of over 34%.

Wang et al. [37] determines the best locations for charging stations to meet the charging demand in Singapore without imposing financial restrictions; through a fast power supplement mode, users expected charging demand and charging time to be determined so that multiple EVs could achieve charging [38]. To increase the adoption of EVs, an optimum solar photovoltaic infrastructure was also introduced and applied at Qatar University [39].

Charging duration, including waiting time, is another important factor in the overall assessment of a charging station. Ashkrof et al. [40] utilized the MXL model on an SP survey of 505 BEV users in the Netherlands and found that waiting time in the queue, charging duration and travel cost, influence charging behavior; however, the installation of fast-charging stations turns into a high cost [41]. Some other solutions included introducing real-time charging station systems linked to the operational conditions of the charging stations [42], considering average waiting time behavior [43], booking systems at charging stations in Spain [44], and optimal locations for charging stations [7,39].

The second primary concern among EV drivers is range anxiety [45], which is the fear of running out of battery power before completing a trip or reaching a charging station. This stressful situation can be alleviated by maintaining a buffer range of battery level [46], even though it may result in more frequent charging and longer charging times than is strictly necessary [47].

The conventional approach for collecting data for the understanding of charging behavior is through stated preference (SP) surveys [20]. These surveys gather information about respondents’ intentions in hypothetical scenarios. Utilizing SP data collection in China allowed researchers to design hypothetical scenarios to gain deeper insights into how EV drivers make trade-offs between key variables [13]. To estimate the impact of these variables on charging behavior, researchers typically employ discrete choice models [10,25,34,43] based on random utility maximization, assuming that individuals choose the option that provides them with the highest utility, where utility is influenced by the attributes of the available alternatives [48]. In our case, the utility could be influenced by several factors, for example, charging time or travel purposes, which differ among individuals and alternatives.

Previous studies conducted in Japan have examined multiple factors that influence charging behavior, but there is still a need to explore additional factors to enhance our understanding of charging behavior. Conversely, the international literature offers a wealth of studies investigating various parameters related to charging behavior. Based on this literature review, a research gap has been identified, leading to the formulation of the aims of this study.

2.3. Research Gaps and Aims

We conducted a comprehensive review of the charging behavior of EV users and have identified the following research gaps in the existing literature:

• Most studies examining EV charging behavior need more detailed information about socioeconomic and demographic factors, such as income status, vehicle characteristics, and the daily activities of drivers [25,45].
• Several studies have characterized EV charging behavior without considering the users’ perception of charging, particularly in relation to the remaining battery level and the expected distance to be traveled for their next trip [28].

Based on these identified gaps, this paper will use the SP dataset collected from 441 BEV users in Japan in 2021 to provide insights into private BEV charging intentions, with a focus on the remaining battery level and expected distance kilometers to be traveled. Also, we will explore the possible association between EV users’ characteristics and their charging decisions. By filling these gaps, we aim to provide valuable insights that can inform policy decisions and infrastructure planning.

3. Data Survey

3.1. Sample Recruitment

A questionnaire survey targeting EV owners in Japan was designed to collect data on the user experience of using EVs, including locations, frequency, and the time required to charge their vehicles, and this survey was administered to collect data from 441 Japanese respondents who owned EVs. The survey was conducted in November 2021 using a research marketing company. In total, 60,000 respondents were recruited via email invitations that embedded screening questions (Some of the screening questions regarding vehicles were as follows: (1) Do you have your own car? If so, how many private vehicles do you drive? (2) What is the make and model of your car in your household? (3) How often do you use your car? (4) What is the purpose of using your car? (5) What is the annual mileage of your car? (6) Please enter the model number of your car.). Of these, 441 respondents who owned EVs were subsequently invited to participate in the full survey, which comprised two main parts. The first part sought revealed preference (RP) data on current experience with the waiting time for charging their EV at various charging locations, as well as respondent socio-demographics (such as age, household income, residential area, and employment status) and EV-related questions. The second part comprised an SP experiment that had a respondent face 10 scenarios, with each asking them to indicate their charging preferences.

3.2. Survey Design

The questionnaire consisted of several sections, and the focus of this paper is on the stated choice (SC) experiment of charging behavior and its corresponding sub-scenarios, which included information about battery-level conditions and the expected travel distance for their next trip. For each sub-scenario, respondents were presented with a choice set consisting of six selectable options for charging their electric vehicles (EVs). Respondents had to select the most applicable alternative concerning each sub-scenario. Respondents were presented with a hypothetical situation describing the sub-scenarios and choice tasks as shown in

Table 1. Scenarios.

Fast Charging at Highway SA/PAs	Normal Charging at Home	Fast Charging at a Commercial Facility	Fast/Normal Charging at the Workplace
Please choose the most applicable alternative for each of the following conditions regarding whether you would conduct quick charging at a highway SA/PA, after considering the presence of a charging facility at a highway SA/PA, the remaining EV battery level, and future driving plans.	Please choose the most applicable alternative for each of the following conditions when you return home regarding whether you would conduct normal charging at home, after considering the remaining EV battery level and future driving plans. a. The case of “Return home, …next time” assumes the scenario where you have plans to drive but are unsure when that plan will be. b. If you do not have a charging facility, then please imagine what you would do if you had a charging facility.	Please choose the most applicable alternative for each of the following conditions regarding whether you would conduct quick charging at a large commercial facility, after considering the presence of a charging facility at a large commercial facility, the remaining EV battery level, and future driving plans. a. If there is no charging facility, then please imagine what you would do if there was a charging facility.	Please choose the most applicable alternative for each of the following conditions regarding whether you would conduct fast/normal charging of your private vehicle at work, after considering the presence of a charging facility at work, the remaining EV battery level, and future driving plans. a. If there is no charging facility, then please imagine what you would do if there was a charging facility. b. If you do not work, then please imagine what you would do if you worked.
Sub scenarios
When the battery level is 90%
When the battery level is 75%
When the battery level is 50%
When the battery level is 25%
When the battery level is 75%, and you plan to drive a distance of 50 km by the end of the day
When the battery level is 50%, and you plan to drive a distance of 50 km by the end of the day
When the battery level is 25%, and you plan to drive a distance of 50 km by the end of the day
When the battery level is 75%, and you expect to drive 10 km later today
When the battery level is 50%, and you expect to drive 10 km later today
When the battery level is 25%, and you expect to drive 10 km later today

and

Table 2. Choice set.

Choice Set
I am sure to charge
In many cases, I will charge
It depends on the time and situation
In most cases, I will not charge
I will not charge at all
I do not know

, respectively.

The different sub-scenarios were designed to cover different situations where (1) the first four sub-scenarios represent standalone situations without any restrictions, providing a baseline understanding of charging behavior with the assumption that many EV drivers do not care so much about the state of charging and the next trip distance; (2) the subsequent sub-scenarios introduce specific conditions and constraints to further analyze charging behavior with battery level and the expected next trip distance kilometer. These conditional sub-scenarios allow for a more comprehensive analysis by considering factors such as battery level, travel distance, and future trip expectations. By incorporating both unconditional and conditional sub-scenarios, this survey design aims to gain insights into various charging behaviors and explore the influence of specific conditions and constraints on decision-making processes. Since this is an SC experiment, the price for charging was assumed to be constant at each location.

The inclusion of both the RP and SP data in our survey design is essential for formulating and developing effective charging infrastructure policies. It enables us to gain insights into charging preferences in terms of location and charging facility, which are still under consideration. Understanding these preferences is crucial for strategizing the requirements of charging infrastructure and further accelerating the growth of EVs. Therefore, the survey questions play a pivotal role in informing the development of charging infrastructure strategies.

3.3. Data Statistics

Table 3. General characteristics of the respondents (N = 441).

Socio-Economic Characteristics	Level	Total Sample Percentage (%)	Residential Area (%)
			Chubu Region	Kanto Region
Gender	Male	58.5	57.0	70.0
	Female	41.5	43.0	29.0
Age (mean = 50)	18–24	5.2	3.1	3.5
	25–39	20.2	22.3	14.9
	40–54	34.9	37.5	32.5
	55–60	39.7	37.1	49.1
Annual household income (JPY) (mean = 8.19 million)	Less than 4 million	17.2	24.7	12.3
	4–8 million	32.2	40.2	36.0
	8–12 million	22.2	25.9	28.9
	12–20 million	7.5	7.1	13.2
	20 million or above	3.6	1.9	9.6
Employment	Full-time	60.5	63.7	61.4
	Part-time	18.5	20.7	18.4
	Unemployed	20.8	15.5	20.2
The number of vehicles owned (mean = 1.4)	1	56.5	49.4	71.9
	2 or more	43.5	50.6	28.0

exhibits the socioeconomic characteristics of respondents, showing the total percentage of the sample and the percentage according to residential area. Data were collected from private households in the Kanto and Chubu regions, covering 16 prefectures in Central Japan. The Kanto region contains the Tokyo metropolitan area, which is the largest and central part of Japan, while Chubu contains the Nagoya metropolitan area, one of the three largest cities, and is leading the nation in the automotive industry. The respondents had BEVs, and the manufacturers of these BEVs included Toyota Motor Corporation, Nissan Motor Corporation, Honda Motor Company, Mazda Motor Corporation, Mitsubishi Motors Corporation, Tesla, and other foreign manufacturers.

The dataset has a comparable percentage of male and female respondents, and, in comparison, there were 51.6% of females in the screening survey and 48.4% of males. A total of 32% percent of the respondents had an annual household income of 4–8 million Japanese yen (about 29 to 58 thousand USD). In contrast with the screening survey, around 34 percent of people had an average annual income of 4–8 million, and only 15 percent of the total sample had 8–20 million in annual household income. The study [49] reported the average yearly income of Japanese households to be between 5.5 and 6 million JPY, based on data from the National Livelihood Survey conducted by the Ministry of Health, Labour, and Welfare in 2017.

In our dataset, there is a worthwhile association between BEV ownership, income, and age. Japanese BEV users were over 40 years old for both residential regions, and this also relates to the Norwegian BEV users aged 36–55 years [50]. However, in the screening survey, 52 percent of the respondents were aged 40–59 years. The age distribution among BEV drivers can be attributed to the fact that EVs tend to be more appealing to middle-aged individuals. Middle-aged individuals typically exhibit greater openness to adopting new technologies and are often attracted to the innovative features and advancements offered by EVs. The descriptive statistics of income and BEV ownership highlight a contrast with the conventional assumption that higher-income households would be more likely to adopt EVs. One possible explanation for this phenomenon could be that lower operating costs and potential incentives make BEVs more attractive to lower-income households. This suggests that affordability plays a significant role in influencing the preferences of lower-income households when it comes to vehicle choices. Nevertheless, it is important to note that the demographic characteristics of our survey respondents may not be representative of the broader population in Japan but provide insights into the demographics of the populations of the Chubu and Kanto regions, where cars remain in higher demand. It is interesting to note that we did not observe any obvious biases related to socioeconomic characteristics resulting from the web-based survey. The descriptive statistics analysis alone does not provide conclusive findings. The insights derived from the descriptive statistics should be considered in conjunction with the results of the model analysis, which will provide potential findings.

In our dataset, the sample includes individuals who are employed full-time or part-time or who are unemployed. The employment statistics of the total survey indicate that 55.3% of the respondents were engaged in the manufacturing industry, finance industry, civil services, medical and welfare industries, and construction industries. Of the 441 respondents, 60.5% were working as full-time workers in regular or regular offices, and 18.5% were working part-time. Through the screening questions, the respondents who owned non-EVs were screened out because this study aims to understand BEV owners’ charging intentions for their vehicles. Unfortunately, we were unable to utilize vehicle registration data, which could have provided more population data on EVs and potentially introduced bias. However, the impact of bias on the findings is presented in the conclusion section.

4. Analysis

In this study, consumer preference for BEV charging decisions was studied through parameter estimates obtained through discrete choice models. Many authors have conducted pioneering studies on discrete choice models. In discrete choice modeling, one of the most widely used models is MXL [51]. The MXL estimates a full covariance matrix among the utility coefficients and permits all utility coefficients to have a random distribution. Such a model provides the kind of correlation that scale heterogeneity would produce, as well as other behavioral factors that can influence the overall strength of the correlation between utility coefficients. These models are computationally feasible in most circumstances [52]. Scale heterogeneity and all other types of correlations are supported by MXL [53]. MXL could be used to approximate any choice model with any preference distribution to any degree of precision [51]. MXL is the generalization of the standard logit model, summarized below:

$$P_{ni}=\int [{\prod }_{f=1}^T\frac{e^{{\beta X}_{nit}}}{{\sum }_{j\in C}{e}^{{\beta X}_{njt}}}]f\left(\beta |b,W\right)\mathrm{d}\beta$$

(1)

where $P_{ni}$ is the probability of choosing alternative i at scenario t (t = 1 to T) for decision maker n; $X_{nit}$ refers to the matrix of the explanatory variables associated with the alternative chosen by individual n in scenario t; $C$ is the choice set; $j\in C$ are alternative j in the choice set C; T is the total number of scenarios; and $f\left(\beta |b,W\right)$ is the density of $\beta$, assumed to be normal for random parameters with mean b and covariance W.

MXL, with its error terms, is more general than its specifications. With an error-component structure, random utility is arbitrarily close to the MXL [54]. In the choice model, the alternatives presented to the respondents in Table 2 were customized into three alternatives for realistic and easier interpretation. Respondents were expected to choose the alternative with the highest utility. The variables relating to charging are as follows:

• “I am sure to charge” and “in many cases, I will charge” were merged as yes (charge) for simplicity.
• “It depends on the time and situation” was used directly from the design named “it depends” in the model. This alternative includes the influence of situational context on the decision-making of charging.
• “In most cases, I will not charge” and “I will not charge at all” were merged as no charge for simplicity.
• The alternative “I do not know” was deleted from our model.

Table 4. Explanatory variables.

Variables	Definition	Percentage (%)
Male	1 if male; 0 otherwise	58
Youth	1 if the age of the respondent is 18 to 24; 0 otherwise	5.2
Young	1 if the age of the respondent is 25 to 39; 0 otherwise	20.2
Senior	1 if the age of the respondent is 40 to 54; 0 otherwise	34.9
Elderly (base)	1 if the age of the respondent is 55 to 80; 0 otherwise	39.7
Low-income household	1 if household annual income is lower than 4 million JPY; 0 otherwise	20.8
Lower-middle income household (base)	1 if household annual income is 4–8 million JPY; 0 otherwise	38.9
Middle-income household	1 if household annual income is 8–12 million JPY; 0 otherwise	26.8
Upper-middle income household	1 if household annual income is 12–20 million JPY; 0 otherwise	9.0
High-income household	1 if household annual income is 20 million JPY or more; 0 otherwise	4.4
Full-time worker	1 if civil servant, management office workers, technical worker, office worker; 0 otherwise	60.5
Part-time worker	1 if self-employed, free-lance, part-time job; 0 otherwise	18.5
Unemployed (base)	1 if full-time housewife, student, others, unemployed; 0 otherwise	20.8
Residential area	1 if Chubu region; 0 otherwise	69.2
No allowable waiting time (base)	1 if no waiting time; 0 otherwise	33.2
Short allowable waiting time	1 if the tolerance time for charging is 5 to less than 15 min; 0 otherwise	40.7
Medium allowable waiting time	1 if the tolerance time for charging is 15 to less than 60 min; 0 otherwise	21.5
Long allowable waiting time	1 if the tolerance time for charging is 60 to 90 min; 0 otherwise	4.5
One vehicle owned (base)	1 if a respondent has 1 vehicle; 0 otherwise	56.5
Two vehicles owned	1 if a respondent has 2 vehicles; 0 otherwise	43.5
Frequency of charging at the target location	1 if the respondent charges less frequently at the target location; 0 otherwise	21.1
Remaining battery level	1 if the battery level is more than 75%; 0 otherwise	-
Kilometer	1 if expected travel is more than 50 km; 0 otherwise	-
Fast charging	1 if fast charging; 0 otherwise	-

presents the variables considered in the deterministic part of our model utility estimation. Among these variables, sex, age, household income, employment, and residential area were the socioeconomic characteristics. The allowable waiting time presents the waiting time respondents have to wait until a spot becomes available. The variable frequency of charging considers the frequency of usage of the charging facility at target locations on two bases: if they will charge and if they will not charge, which means the frequency of charging will be zero. The remaining battery level, fast charging, and distance (km) were obtained from the sub-scenarios in the SP survey.

5. Results and Discussions

Descriptive Analysis of Charging Behavior

Table 5. Consumer responses toward charging.

Charging Choices	I Am Sure to Charge	In Many Cases, I Will Charge	It Depends on the Time and Situation	In Most Cases, I Will Not Charge	I Will Not Charge at All	I Do Not Know
Scenario: Fast charging at highway SA/PAs
Count	928	804	1071	438	656	513
Percentage (%)	21.0	18.2	24.2	9.9	15.0	12.0
Scenario: Normal charging at home
Count	1036	670	851	364	777	712
Percentage (%)	23.5	15.19	19.3	8.25	17.6	16.1
Scenario: Fast charging at large commercial facilities
Count	737	739	1175	462	766	531
Percentage (%)	16.7	17.0	27.0	10.4	17.3	12.0
Scenario: Fast charging at workplace
Count	852	714	998	432	811	603
Percentage (%)	19.3	16.1	22.6	9.7	18.4	13.7
Scenario: Normal charging at workplace
Count	908	648	953	444	831	626
Percentage (%)	21.0	15.0	22.0	10.0	19.0	14.0
Total	4461	3575	5048	2140	3841	2985

presents a descriptive analysis of consumer behavior toward charging in the five scenarios. Overall, the proportion of respondents choosing the alternative “it depends on the time and situation” is high when compared to other choices. The second and third largest proportions of choices were “I am sure to charge” and “in many cases, I will charge”. In the case of normal charging at home, the choice “I am sure to charge” had the highest number of responses. Thus, the charging behavior depended on time, and the situation was more important for respondents. The response for the “I do not know” case was low; therefore, it was not considered in the model. Workplaces, including fast and normal charging, had a higher proportion of people choosing to charge compared to large commercial facilities. Consumer behavior toward charging was heterogeneous and may be related to personal attributes. These results are consistent with those of past studies [55], which illustrated that PHEV drivers prefer charging at home and in the workplace, and 16% of PHEV drivers demonstrate charging behavior similar to that of EVs. However, a mixed logit model was applied to estimate the role of explanatory variables in the charging decisions. Soon, the number of EVs might increase more than the number of charging stations; therefore, inadequate charging facilities may be a bottleneck in EV market growth. It is important to have adequate charging infrastructure to increase EV adoption.

6. Estimation Results

Table 6. Estimation results.

		Fast Charge at Highway SA/PAs	Normal Charge at Home	Fast Charge at Commercial Facility	Fast/Normal Charge at Work
Log-likelihood		−3701.40	−5509.50	−3703.20	−5906.80
McFadden R2		0.25	0.42	0.28	0.34
AIC		7554.83	11,170.97	7558.47	11,973.70
Sample size		3000	2920	3110	6090
Variable	Alternatives	Coefficients	Coefficients	Coefficients	Coefficients
Intercept (mean)	Charge	0.93 *** 1	0.56 ***	- 2	0.42 **
	No charge	-	−1.62 ***	−1.56 ***	−1.65 ***
Intercept (sd)	Charge	1.28 ***	0.61 ***	1.42 ***	0.24 ***
	No charge	-	1.03 ***	1.57 ***	1 ***
Male (mean)	Charge	-	0.27 *	0.41 **	−0.40 ***
	No charge	0.57 ***	2.11 ***	1.39 ***	1.08 ***
Male (sd)	Charge	0.77 ***	0.43 ***	0.38 ***	0.93 ***
	No charge	0.90 ***	1 ***	1.05 ***	0.98 ***
Youth (mean)	Charge	-	1.58 ***	1.42 ***	1.03 ***
	No charge	-	−1.48 ***	0.94 **	1.03 ***
Youth (sd)	Charge	1.31 ***	1.56 ***	-	1.86 ***
	No charge	1.14 ***	1.38 ***	3.21 ***	1.47 ***
Young (mean)	Charge	−0.3	−0.63 ***	0.32	−0.69 ***
	No charge	−0.30	−1.94 ***	−0.62 **	-
Young (sd)	Charge	1.31 ***	1.05 ***	0.82 ***	1.22 ***
	No charge	1.14 ***	0.94 ***	1.74 ***	2.14 ***
Senior (mean)	Charge	-	−0.67 ***	−0.36 **	−0.3 **
	No charge	-	−1.68 ***	-	0.62 ***
Senior (sd)	Charge	-	0.29 ***	0.72 ***	-
	No charge	1.50 ***	1.87 ***	-	0.75 ***
Low-income household (mean)	Charge	0.29 *	0.61 ***	0.56 **	-
	No charge	-	0.90 ***	0.59 ***	0.96 ***
Low-income household (sd)	Charge	0.86 ***	0.41 ***	1.63 ***	1.83 ***
	No charge	2.27 ***	0.36 ***	1.65 ***	1.11 ***
Middle-income household (mean)	Charge	−0.40 **	0.66 ***	0.33 **	0.36 ***
	No charge	−0.72 ***	−0.94 ***	−0.76 ***	−0.71 ***
Middle-income household (sd)	Charge	0.38 ***	0.28 ***	0.51 ***	0.14.
	No charge	0.60 ***	0.20 *	1.60 ***	1.46 ***
Upper-middle income household (mean)	Charge	−0.65 **	−0.89 ***	-	0.59 ***
	No charge	0.46 *	1.26 ***	-	1.09 ***
Upper-middle income household (sd)	Charge	-	1.22 ***	2.61 ***	1.15 ***
	No charge	0.55 *	1.08 ***	-	0.63 ***
High-income household (mean)	Charge	-	1.83 ***	−0.60 *	-
	No charge	-	-	0.74 **	0.95 ***
High- income household (sd)	Charge	-	0.90 ***	1.82 ***	0.79 ***
	No charge	3.43 ***	1.38 ***	2.30 ***	0.40.
Full-time worker (mean)	Charge	-	0.40 **	−0.34.	0.30 *
	No charge	−0.62 ***	-	−0.34 *	-
Full-time worker (sd)	Charge	0.40 ***	1.63 ***	0.27 ***	1.44 ***
	No charge	0.90 ***	1.35 ***	0.36 ***	0.77 ***
Part- time worker (mean)	Charge	−1.20 ***	-	-	-
	No charge	−1.02 ***	1.20 ***	−0.43 *	1.70 ***
Part-time worker (sd)	Charge	3.57 ***	1.74 ***	0.26 *	0.95 ***
	No charge	2.55 ***	1.40 ***	2.66 ***	1.06 ***
Chubu region (mean)	Charge	0.21.	-	-	0.97 ***
	No charge	-	0.82 ***	0.23.	1.47 ***
Chubu region (sd)	Charge	-	-	0.72 ***	0.97 ***
	No charge	1.04 ***	1.31 ***	0.18 *	1.47 ***
Remaining battery level (mean)	Charge	−0.41 ***	−0.97 ***	−0.33 ***	−1 ***
	No charge	0.64 ***	1.11 ***	0.50 ***	1.11 ***
Remaining battery level (sd)	Charge	0.84 ***	1.51 ***	0.33 ***	1.23 ***
	No charge	-	1.63 ***	0.67 ***	1.5 ***
Kilometer (mean)	Charge	−0.23 *	1.18 ***	−0.55 ***	0.74 ***
	No charge	−0.35 **	−1.27 ***	−0.5 ***	−0.58 ***
Kilometer (sd)	Charge	2.32 ***	2.67 ***	1.9 ***	0.65 ***
	No charge	1.29 ***	2.03 ***	1.71 ***	0.75 ***
Fast charging (mean)	Charge				-
	No charge				−0.2 *
Fast charging (sd)	Charge				0.61 ***
	No charge				0.32 ***
Short allowable waiting time (mean)	Charge	−0.32 *	-	-	−0.27 **
	No charge	-	-	-	−0.76 ***
Short allowable waiting time (sd)	Charge	-	-	-	0.19 **
	No charge	-	0.16 *	-	0.42 ***
Medium allowable waiting time (mean)	Charge	-	-	-	−0.19.
	No charge	-	-	-	−0.79 ***
Medium allowable waiting time (sd)	Charge	-	0.15 *	-	-
	No charge	-	-	-	0.33 ***
Long allowable waiting time (mean)	Charge	-	-	-	-
	No charge	-	-	0.70 *	−0.61 ***
Long allowable waiting time (sd)	Charge	-	0.26.	1.34 **	0.70 ***
	No charge	-	-	-	-
Two vehicles owned (mean)	Charge	0.93 ***	−0.55 ***	-	0.33 ***
	No charge	0.47 ***	−0.18.	-	-
Two vehicles owned (sd)	Charge	0.51 ***	1.90 ***	1.85 ***	1.72 ***
	No charge	-	2.10 ***	0.47 ***	0.16 *
Frequency (mean)	Charge	-	-	0.31 **	0.26 *
	No charge	-	-	0.37 **	0.30 **
Frequency (sd)	Charge	-	-	-	0.29 **
	No charge	0.36 **	0.24 ***	0.30 *	-

¹ *, **, and *** show statistical significance at 10%, 5%, and 1% levels, respectively. ² - parameters are not statistically significant.

presents the estimated parameters from the MXL for the 441 BEV owners (We also estimated multinomial logit model (MNL), ordered logit, and generalized ordered logit models for comparison purposes; the log likelihood and AIC values of these models were compared, indicating that MXL provides the best results. However, the detailed results for the generalized ordered logit model are presented in Appendix A. The results also suggest that the MXL model statistically fits our data better than the generalized ordered logit model.). A full set of variables was tested, and the statistically significant variables for less than 0.05%, 0.01%, and 0.001% are presented in the table. A wide range of sociodemographic variables, such as employment and daily EV usage patterns, including allowable waiting time and frequency of charging at the target locations, were found to be significant for charging decisions.

A detailed interpretation was carried out to capture the role of explanatory variables as well as deterministic and random heterogeneity in preferences. This included interacting with the remaining battery level, expected distance kilometer, allowable waiting time, gender, age, income, employment status, area, and vehicle ownership with the charging decision between the charging and not charging alternatives. In addition to socio-demographics, attitudes and perceptions are important behavioral determinants in many areas of human behavior.

Alternative-specific constants were included for charging vs. not charging. The means for intercepts for charging were positive and significant for all locations except for fast charging at commercial facilities, showing that, all else being equal, respondents prefer the charging choice. For normal charging at home and fast charging at commercial facilities, males exhibited a higher likelihood of charging or not charging than the reference category, depending on the situation. However, they demonstrated a stronger inclination towards not charging at the highway SA/PA, while displaying lower intentions for charging and higher intentions for not charging at the workplace. It is quite possible that individual preferences, access to charging infrastructure, and distance from charging facilities could also play a role in shaping these patterns. In general, there was heterogeneity in choosing charging alternatives, but most men preferred charging, and some respondents tended to behave the opposite. Youth were more likely to charge via normal charging at each location, while there was no significance for fast charging at highway SA/PAs. This can be attributed to the fact that charging on highways is not a common practice due to the infrequent nature of highway travel for most individuals. Youth show a lower likelihood of both charging and no charging at highway SA/PAs and normal charging at home, while they demonstrate higher intention and lower intention for charging at commercial facilities and the workplace, respectively. This can be the influence of factors such as the availability of convenient charging facilities or charging incentives at commercial facilities. Seniors show a lower intention of charging at home, commercial facilities, and workplaces but show a strong intention of not charging at workplaces. It seems that seniors rely more on the situation for charging decisions. This is consistent with [55], indicating that personal attributes such as age affect the charging choice of BEV drivers.

Household income is a significant variable for charging decisions. Low-income households show a lower likelihood of charging and not charging at home and commercial facilities, while they show a higher intention of charging and not charging at highway SA/PAs and the workplace, respectively. These patterns in charging behavior among low-income households reflect the complex interplay between financial considerations, access to charging infrastructure, and the perceived convenience of different charging options. Middle-income respondents show a strong intention of charging and a low intention of not charging at home, commercial facilities, and the workplace, while showing a lower intention of charging and not charging at highway SA/PAs. It is quite possible that charging at home, the workplace, and commercial facilities are more convenient and align with their daily routines, and they may have a lower intention to use highway charging facilities. Upper-middle-income households show a lower likelihood of charging and a higher likelihood of not charging at highway SA/PAs and at home, while they show a higher preference for charging and not charging at the workplace. This reflects the fact that charging at the workplace may be perceived as convenient and cost-effective. High-income households show a higher likelihood of charging at home and a higher intention of not charging at commercial facilities and the workplace. According to previous studies conducted in California, people (BEV owners) with high incomes, old age, and owners of detached houses prefer to charge at home [56]. Well-educated and wealthy people with EVs rarely charge at public stations or workplaces [57].

Employment status hits differently for the charging decision. Full-time workers show strong intentions for normal charging at home and the workplace, while they show lower intention for not charging on highways and at commercial facilities. However, part-time workers demonstrate a lower intention of charging and not charging on highways, which seems to depend more on the situation, while showing a strong intention of not charging at home and the workplace.

In our survey, the majority of the respondents were from the Chubu region. They showed a strong intention of charging on highways while depending on the situation for charging at the workplace. However, they showed a strong intention of not charging at home or in commercial facilities. Charging infrastructure availability, convenience, and individual preferences influenced the charging intentions of respondents from the Chubu region.

The next two variables are acceptable waiting time and the frequency of charging. The acceptable waiting time, which is the time until the charging spot becomes available at the charging station, affects the charging decision. The allowable waiting time preferences, including short, medium, and long allowable waiting times, indicate that respondents were less likely to prefer charging or not charging at the workplace while showing a strong desire for not charging at commercial facilities if they have to wait for a longer time. The variable frequency indicates the frequency of charging at target locations, such as for fast charging at commercial facilities and the workplace. Respondents show a higher likelihood of charging or not charging. Additionally, respondents with multiple vehicles show a higher likelihood of charging or not charging compared to the reference category and tend to have the opposite likelihood of normal charging at home. They show a strong desire to charge at the workplace. It should be noted that variable fast charging is only considered at the workplace, which is significant because there is less likelihood of not charging at the workplace.

Aside from the above, the remaining battery level and expected distance kilometers to be traveled also impact charging intention. The remaining battery level indicates that, when the remaining battery level is more than 75% and respondents are planning for the next trip, they show lower intention to charge at each location as the battery level decreases; respondents are more willing to charge at all target locations. Additionally, when respondents are planning their next trip and the expected travel distance is more than 50 km, they will depend on the situation to charge at highway SA/PAs and commercial facilities. Respondents showed a higher intention to charge at home and the workplace when the expected travel distance was more than 50 km. This is probably because charging at home and the workplace is convenient. This study is consistent with past findings that the probability of normal charging at home increases when the travel duration on the previous or next travel day increases in the case of commercial BEVs [11].

The empirical findings could provide some insights for policymakers to understand charging preferences such as charging intention linked with the socioeconomic characteristics of EV drivers and charging location choice with type of charging facility. The key findings indicate that, with a low remaining battery level, EV drivers prefer to charge at locations, i.e., home, workplace, commercial facilities, and highways. EV drivers follow the same manner with expected travel distance, such as if they are planning a trip with more than 50 km, they will prefer to charge at home. In addition, EV drivers are reluctant to endure longer waiting times. This information is crucial for determining the capacity of charging stations.

Table 7. Comparison of MXL with other logit models.

		Fast Charge at Highway	Normal Charge at Home	Fast Charge at Large Commercial Facilities	Fast/Normal Charge at Workplace
Ordered logit model	Loglikelihood	−4883	−9019.5	−5103.4	−8709.55
	AIC	9806	18,079.08	10,246.8	17,461.1
Generalized ordered logit model	Loglikelihood	−4832.88	−8881.37	−5040.25	−8455.87
	AIC	9741.77	17,838.75	10,156.5	16,991.74
Multinomial logit model	Loglikelihood	−4831.2	−8884.9	−5036.5	−8461
	AIC	9738.42	17,845.8	10,148.9	17,001.9
Mixed logit model	Loglikelihood	−3701.4	−5509.5	−3703.2	−5906.8
	AIC	7554.83	11,170.97	7558.47	11,973.7

presents a comparison of MXL with other models, including the ordered logit model, the generalized ordered logit model [58], and MNL. Based on the AIC, the MXL fits best, as it has a lower AIC for each scenario, which could be due to differences in the flexibility and assumptions of the models.

7. Policy Implementation

Many studies have already been conducted on EV adoption; however, the intention for the adoption of EV charging is not well prepared. The study majorly contributes within the given condition of (fast charging at highway SA/PAs, normal charging at home, fast charging at large commercial facilities, and fast/normal charging at the workplace), followed by the users’ behavioral intention of BEV charging. Hence, empirical findings from this study can be useful for policymakers and charging infrastructure planners.

Intuitively, it was noted that higher income groups are less susceptible to outdoor fast charging, preferring in-home charging. It is expected that, in the future, the demand for fast charging in the home may gain increment, particularly for high-income households. A broad policy viewpoint could be that the government may ask the car manufacturers to introduce an affordable fast EV charging infrastructure in order to benefit lower- and middle-income households. In addition, the government may subside the electricity benefits for the users of charging infrastructure.

The results show that charging intention is significantly related to the remaining battery level and expected distance per kilometer. Furthermore, BEV drivers do not tend to bear longer waiting times at charging stations. This information could help the government to cap the incentives or subsidies of electricity at charging stations; for example, for those who have BEV with lower-range anxiety, the government can enact a policy to prioritize them at charging stations, as they do not need to wait for longer times. For example, China and Norway have increased non-financial incentives for EV drivers. Time-based discounts can be offered to drivers or reduced charging fees or free charging for a certain duration can be offered to allow them to charge their vehicles.

The results also indicate that BEV drivers show a strong intention to charge at home and the workplace. This can be an important consideration for policymakers in the development of home-location charging, particularly for residential buildings with individual units and for those without private parking. Workplace charging could be a feasible alternative for non-home charging locations. In the near future, workplace charging might also gain more increment. Probably, policymakers should introduce a reduced charging cost to employees. These are few examples of policy planning, but it is crucial, however, to understand that actual charging infrastructure planning is more complex, and other factors need to be considered.

8. Conclusions and Future Work

In this study, we used results from 441 respondents to capture charging intentions at multiple locations in Japan. Based on the role of explanatory variables, we could analyze the different patterns of charging behavior for BEV owners. Using a mixed logit model, we analyzed potential factors that may characterize different charging intentions. Socio-demographic characteristics like gender, household income, the age of the driver, residential area, job, and vehicle ownership as well as charging behavior variables like remaining battery level, expected distance kilometer to be traveled, fast charging at the workplace, acceptable waiting time, and frequency of charging were found to be key factors influencing the charging decisions.

The remaining battery level and expected distance kilometer to be traveled are crucial factors for making a charging decision, indicating that those with a remaining battery level of more than 75% are more inclined to not charge. Kilometers to be traveled show a mixed usage of charging infrastructure for charging intentions, indicating respondents are more inclined to charge at home and the workplace when expected travel is more than 50 km, depending on the situation for charging at highway SA/PAs and commercial facilities. Other than that, BEV drivers are not willing to bear longer acceptable waiting times. The result here shows the importance of sociodemographic and charging behavior variables for charging intentions. This highlights the importance of having an integrated infrastructure investment plan that will account for different locational charging intentions, considering BEV drivers’ preferences. Additionally, the statistically significant coefficients of the estimated MXL model of gender, age, income, and employment status suggests the potential relationship with environmental awareness and affordability, which are associated with basic characteristics of EVs—using EVs can have environmental benefits.

The empirical findings above could be attributed to the studies of charging behavior from theoretical and practical perspectives. On the theoretical side, the charging decisions of BEVs, consumer preferences for charging location and facilities, and behavioral factors that influence charging decisions have received ample attention in previous studies. The empirical findings in this paper could provide a better understanding of the charging behavior of BEVs which could be further used for explicitly modelling charging behavior at the individual level. From a practical standpoint, the empirical findings could help stakeholders and policymakers to ensure adequate charging infrastructure availability, leading to improved accessibility and convenience for EV users, as well as supportive policies for the new and existing charging infrastructure at prioritized charging locations of EVs.

However, several factors may be important for charging choices that could have been but were not included in the survey and defined as limitations. First, the survey focused on the general charging choice behavior and passed over the potential influence of some important factors, such as trip purpose, range anxiety (which could vary across different brands of BEVs), and charging station characteristics. To address this, future questionnaires could be expanded to include more detailed questions about these factors. Second, the MXL model was used to provide insights into the relationships between individual attributes and charging behavior. More complex models such as nested logit models could be used to further explore the interrelationships among different individual choices. Third, the potential bias of the survey compared to a general sample lies in terms of its exclusive focus in Japan, as we used the Japanese EV owner dataset. In other cases, charging behavior might be different because of the existing charging infrastructure and road network. Further analyses are necessary to investigate the potential influence of these biases on empirical findings. Moreover, winter conditions are also interesting impacts not covered in this study.

The results can have implications for charging infrastructure planning, as the obvious rise in EV adoption depends on the charging infrastructure, which is consistent with past findings [59]. Homeowners and owners of detached houses with private charging focus more on adopting EVs [54,57]. Additionally, this research can be expanded in the future to comprehend the effects of charging behavior on potential EV users. Future investigations should focus on the variables that affect charging decisions, such as drivers’ travel routines for traveling, traveling purposes, attitudes, and perceptions of the charging infrastructure.