Factors Influencing Adoption Intention for Electric Vehicles under a Subsidy Deduction: From Different City-Level Perspectives

Abstract

The popularization of electric vehicles (EVs) is beneficial to the sustainable development of energy and the environment. China’s promotion and development strategy for EVs will serve as a model for other countries. EV ownership has a significant difference between first/second-tier (FST) cities and third/fourth-tier (TFT) cities and there is a huge growth potential for the EV market in those TFT cities. This paper aims to explore the factors influencing the adoption intentions for EVs in FST and TFT cities under a subsidy deduction and to make a comparative analysis of their regional heterogeneity. Based on the extended theory of planned behavior (TPB) model, the structural equation model is used to compare the factors affecting the adoption intention for EVs of 858 respondents in China. The results show that attitude, subjective norms, novelty seeking, non-financial incentive, product cognition, and environmental concerns are positively related to intention in FST and TFT cities; however, infrastructure development only has a positive significant impact in the TFT cities. Additionally, the subsidy deduction has a more negative impact on the adoption intentions in FST cities. Our findings provide vital insights for formulating government regulations and marketing strategies depending on the diverse sizes and attributes of Chinese cities.

1. Introduction

The transportation sector is responsible for a large proportion of global energy consumption, and the ensuing tailpipe emissions are amongst the major sources of air pollutants in China [1]. According to survey statistics of the IEA, China’s CO₂ emissions reached 9876.5 million tons in 2019, of which the transportation sector accounted for 9.12%, ranking third among the industrial sectors (Figure 1) [2]. As a perfect alternative to conventional fuel vehicles, electric vehicles (EVs) are adopted to reduce air pollution and greenhouse gas (GHG) emissions [1,3]. The reasons for China’s commitment to promoting EVs are as follows. First, EVs are more energy efficient than conventional vehicles. The energy efficiency of EVs is close to 60%, while that of conventional fuel vehicles is less than 20% [4]. Moreover, the emissions of EVs are nearly zero [5]. Second, the promotion of EVs reduces energy consumption. China is an oil-poor country with an external oil dependency of 72.05% [6]. Additionally, the mismatch between the demand and supply of energy and environmental pollution has become an urgent problem for the Chinese government [7] and promoting electric vehicles (EVs) adoption is helpful for the transformation of the automotive industry, reducing the pressure caused by fossil energy consumption and alleviating environmental pressure.

Some critics point out that the electricity needed for EVs comes mainly from fossil fuels such as coal, which may lead to an increase in carbon emissions [8]. Instead, Valdez et al. [9] argued that EVs are environmentally friendly because the electricity they require can be provided by renewable energy sources such as wind, solar, hydro, and biomass. The use of renewable energy can reduce carbon emissions [10], which motivates China to encourage the development of the renewable energy industry. In 2021, China’s installed renewable energy capacity exceeded 1 billion kilowatts, of which installed wind and solar photovoltaic capacity have both exceeded 300 million kilowatts [11]. Predictably, the development of EVs is one of the crucial ways to preserve the environment and achieve energy sustainability.

The New Energy Vehicle Industry Development Plan (2021–2035) [12] declares that, the development of EVs is a necessary path for China to transform from a large to a strong automobile country; the sale for EVs will reach about 20% of the total vehicle sale in 2025, and pure EVs will be the mainstream of EVs sales by 2035. According to the data from the Chinese Ministry of Public Security in October 2021 [13], China has 296 million motor vehicles, including 6.78 million EVs, which account for only 2.29% of the total. This is far from the target (20%) for the New Energy Vehicle Industry Development Plan [12]. In this context, it is necessary to study the factors influencing the adoption intention for EVs. Furthermore, the subsidy deduction will be fully enforced in 2022 to change the market from being subsidy-oriented to market-oriented. This will affect the adoption intentions for EVs, given the subsidy deduction’s potential for causing a rise in EVs prices.

Factors influencing the adoption intention for EVs vary among the different city levels. According to Figure 2 [14], there was an enormous sales gap between the first/second-tier (FST) cities and third/fourth-tier (TFT) cities. The proportion of EVs was 69.11% in the FST in 2019, but only 28.48% in the TFT cities. As shown in Figure 3, public charging piles in China are mainly concentrated in Guangzhou, Jiangsu, Beijing and Shanghai, accounting for 46% of the total. The uneven distribution of charging piles is indicative of the uneven distribution of EVs. What causes this phenomenon? There is a huge sales market in TFT cities with large populations. At such a time, clarifying the differences in the factors influencing EVs adoption intentions between the FST and TFT cities contributes to the formulation of targeted guidance policies in cities at different levels to improve EVs adoption. In addition, it has a significant theoretical and practical significance to promote the popularity of EVs. Moreover, China has many cities, including cosmopolitan cities such as Beijing, Shanghai, Guangzhou, and Shenzhen, as well as some smaller cities that are relatively less cosmopolitan. As the largest market of EVs, the case of China cities can serve as a reference for cities in the world at different levels. Exploring the factors that influence consumer adoption intentions can provide lessons for multinational automotive companies to formulate strategies in the Chinese market; however, the main factors affecting EVs in TFT cities are not well understood so far. This article is dedicated to identifying the factors that influence the sales of EVs and the huge gap between the sales of EVs in FST cities and TFT cities.

Previous studies have explored the factors influencing the procurement of electric vehicles [3,15,16]. Zhang et al. [10] investigated residents’ perceptions and adoption intentions of EVs from an informational perspective in 10 pilot cities in China. He et al. [3] conducted an in-depth study and found that the effect of environmental concern on EVs adoption behavior was quantitatively determined. Previous research on the adoption intention for EVs was limited to consumers in metropolises, provincial capitals, or pilot cities, such as Beijing, Shanghai, Guangzhou, Shenzhen, and Tianjin [17,18,19]. Helveston et al. [20]) conducted a comparative analysis of adoption intentions in Beijing, Shanghai, Shenzhen, Chengdu, and American cities; however, the existing studies still have some deficiencies. First, the factors influencing the adoption intention of EVs in TFT cities have not yet been well explored. Second, it is necessary to conduct a comparative analysis of the differences in EV adoption intentions between different city levels, which is rarely considered in the existing literature. To address these research gaps, we added additional influencing factors to the theoretical model after reviewing previous literature to form an extended theory of planned behavior (TPB) model. This paper aims to explore the factors influencing the willingness to adopt EVs in FST and TFT cities under a subsidy deduction and to make a comparative analysis of their regional heterogeneity based on a questionnaire survey.

The following parts are organized as follows. Section 2 reviews the relative literature, Section 3 introduces the model hypotheses, Section 4 describes the methodology, Section 5 presents the main results and discussion, and Section 6 concludes this study and proposes the policy recommendations.

2. Literature Review

The relative literature currently focuses on the following aspects.

2.1. The History of EVs

With the increase of car ownership since the 1980s, fuel vehicles have caused a certain negative influence on energy and the environment [21,22,23]. The attention of governments and consumers has been gradually targeted at non-fossil energy consuming EVs. The development of a new power battery makes EVs the core trend in the future development of the automobile industry. BYD, a representative of pure EVs in China, launched the BYD E6 in 2011, which has a top speed of 140 km/h and a maximum driving range of 290 km. In 2013, Tesla launched the Model S with a maximum speed of 210 km/ h and a continuous driving range of 480 km, pushing pure EVs to a new level of development. With the increase of people’s environmental awareness, and the improvement of EV technology infrastructure, the ideals of EVs have increased considerably in the last decade [24]. According to the corporate website of Tesla, this company is leading the world transition to sustainable energy [25]. Tesla and other new car-making forces promote traditional enterprises transforming into green ones by creating new user experiences and product models, which makes a large number of high-quality EVs available to the market in succession.

2.2. The Impact of Subsidies on the Adoption of EVs

In studies on the adoption of EVs, the subsidy policy has been widely examined. China once offered a very generous subsidy policy to stimulate EVs sales; however, the subsidy has recently been gradually declining. The Ministry of Finance of China released a total of CNY 35.058 billion in subsidies for energy preservation and emissions reduction in 2019. Among them, a total of 1.11 million new energy vehicles were promoted, and CNY 14.41 billion was pre-appropriated for promotion subsidies. From the perspective of mileage subsidy standards, the subsidy per vehicle for pure EVs in 2019 could have reached up to CNY 25,000, which was halved compared with the CNY 50,000 in 2018. In addition, since 2019, local governments have no longer provided subsidies for the purchase of EVs. If local governments continue to grant purchase subsidies, the central government will subtract relevant financial subsidies accordingly [26]. Riesz et al. [27] suggested that a subsidy is an effective policy and financial incentives can promote the adoption of EVs [28,29]. Research from Norway, New Zealand, and Canada found that financial incentives and low prices can effectively encourage the adoption of EVs [30,31,32,33]. Yuan et al. [34] suggested that the monetary subsidy plays a significant role in the diffusion of EVs; the Chinese EV market is deeply influenced by subsidies rather than technology, thus, the subsidy deduction makes EVs less attractive [34,35]. Higueras-Castillo et al. [36] believed that incentives are still a key factor in stimulating the sales of EVs in Spain, and the development of EVs in China is greatly influenced by policy incentives currently, therefore, EVs sales are less affected by the market [35]. Wang et al. [37] used a system dynamic model to predict the status by 2030 and found that a subsidy elimination would decrease the market share for EVs by 42% in China. Xiong et al. [38] classified the residents consumption ability of EVs into high, medium, and low levels, and found that the acquisition of a subsidy incentive is effective, and that the implementation effect is most obvious in cities with a low-consuming capacity. Huang and Qian [39] made surveys on EVs consumers in Chinese Jiangsu, and concluded that consumers were more sensitive to the subsidy policy when adopting EVs in third-tier cities; however, Li et al. [40] suggested that the adoption rate of EVs was unlikely to drop significantly without subsidies.

2.3. The Impact of City Level on the Adoption of EVs

According to the number of urban populations in each city, Chinese cities are classified into different levels. The number of urban populations over 10 million are the first-tier cities. The number of urban populations of second-tier cities is 5 to 10 million people. The number of urban populations of third-tier cities is 1 to 5 million and the number of urban populations between 0.5 to 1 million are considered fourth-tier cities [41]. There are differences in the urban development and population distributions in China. The population is 427 million people in FST cities, and 953 million people in cities below the second tier [42]. There are also significant differences in the infrastructure, economic development, and residents incomes across the different city levels [43]. Thus, the purchase capability differs in different city levels [44].

Most studies on the adoption intention for EVs focused on the FST cities. For example, She et al. [17] found that the consumers in Tianjin had low interests in EVs. Lin and Wu [18] found that, the network externality, price acceptability, government subsidy, vehicle performance, environmental concern, gender, age and marital status have significant effects on the adoption intention for EVs, based on the investigation in four first-tier cities (Beijing, Shanghai, Guangzhou, and Shenzhen). Huang and Ge [16] suggested that in Beijing the attitude, perceived behavior control, cognitive state, product concern, and monetary incentive policy had positive effects on the adoption intention for EVs, and subjective norms and non-monetary incentive policy had insignificant effects. Helveston et al. [20] used the data from Beijing, Shanghai, Shenzhen, and Chengdu, to compare the adoption intention for EVs in these cities with the United States. Xiong et al. [38] found that the acquisition subsidy was most effective in cities with a low consumption ability, whereas the acquisition tax relief was most effective in cities with a medium consumption capacity, but was not effective in cities with a high and low consumption capacity. Huang and Qian [39] took the consumers from two second-tier and three third-tier cities in Chinese Jiangsu as an example, and found that the acquisition subsidy, charging facility and cruising range greatly affected the adoption intention. Li et al. [40] investigated consumers of EVs in four counties of Luoyang, Chinese Henan, and the results revealed that the charging policy had a great influence on purchase intentions, followed by the “right of way” policy, publicity policy and vehicle acquisition policy. Shi et al. [45] suggested that consumers in Weihai, Chinese Shandong, had a low adoption intention for EVs, due to the main restrictive factors such as price, cruising range, supporting facility, technical maturity and so on. Zhang et al. [10] recently investigated the perceptions and adoption intentions of residents in 10 pilot cities in China from an informational perspective, while Liu et al. [19] studied the diffusion of EVs using only Beijing as a case study.

2.4. The Impact of Behavioral Elements on the Adoption of EVs

An intention is defined as the subjective probability to initiate a certain behavior in the psychological area [46]. Some studies have regarded intention as the main predictor of actual behavior [47,48,49]. The TPB was used to predict intention especially in the area of green consumption [50,51] and the TPB suggests that behavior can be straightforwardly impacted by an intention that depends on attitude, subjective norms and perceived behavior control [51]. The attitude is the evaluation of the performance of a particular behavior while subjective norms refer to the social pressure to implement a particular behavior [52] and perceived behavior control describes the ease and difficulty of implementing the behavior [53]. The adoption intention for EVs was considered a manifestation of pro-environmental behavior [54,55].

Moons and De Pelsmacker [56] conducted a survey on Belgian consumers and found that the influence of subjective norms and perceived behavior control on the adoption intention for EVs were less important than the impact of attitude, although the subjective norms and perceived behavior control have a positive effect. Based on the extended TPB model, Mohamed et al. [57] found that the attitude and perceived behavior control are strong factors influencing the adoption intention for EVs, through a survey on 15,392 households in Canada. Huang and Ge [16] argued that, in Beijing, attitude and perceived behavior control have a significantly positive effect on the adoption intention, and that the subjective norm was not significant. Li et al. [40] used an extended TPB model to explore the relationship between psychological factors, policy mix characteristics, and adoption intentions for EVs in 88 surveyed pilot cities and found that the consistency of policy mix has a negative moderating influence on the attitude, subjective norms and personal norms; they are positive moderators for the relationship between perceived behavior control and adoption intentions for EVs.

Scholars have improved the predictability of the initial TPB model through introducing the additional behavior conception, in order to better predict behavioral intentions [57,58,59]. For example, environmental concerns, vehicle performance, product concerns, government policy, and socio-demographic characteristics have an impact on intentions [16,17,18,20,60,61,62]. She et al. [17] found that inadequate infrastructure is a major obstacle for the adoption for EVs. The charging infrastructure affects convenience, thereby affecting the adoption for EVs [15]. Meanwhile, Dagsvik et al. [63] suggested that EVs will be competitive if the supporting infrastructure is abundant. In the previous literature, few scholars introduced the notion of “novelty seeking” into the analysis of adoption intentions, therefore it is necessary to consider “novelty seeking”, because this factor influences the adoption intentions [64]. Consumers are more likely to adopt a new product if the product is unique and innovative [65]. Jansson [66] found that consumers adopting fuel vehicles were more novelty-seeking, while Yang et al. [67] found that the adoption intention for EVs was influenced by product concerns. The concern and evaluation for the attributes of EVs is one of the main factors [16]. Environmental concern is an important factor [57], and this has changed current personal behaviors [68]. White and Sintov [69] suggested that consumers concerns for climate change can positively affect their adoption intentions for EVs. Thus, in this paper, based on the TPB model, we added the variables, such as infrastructure, novelty seeking, subsidy deduction, non-financial incentive, product cognition, and environmental concern into our conceptual framework.

In summary, some shortcomings exist in the previous studies. First, most literature focusing on the impact of subsidy deduction on the adoption intentions for EVs is limited to some specific local areas, but it lacks the comparative analysis of factors in different city levels. Second, the previous studies on the adoption intentions for EVs mainly focused on the FST cities, but seldom focused on the TFT cities.

In this regard, this study attempts to contribute to the relative research area from the following two aspects. First, the study sample is expanded to FST and TFT cities, and the structural equation modeling (SEM) and analysis of variance (ANOVA) are applied to conduct a comparative analysis from different city-level perspectives. Second, based on the previous studies, the TPB model is expanded with some new variables, such as the infrastructure, non-financial incentive, product cognition, environmental concern, novelty seeking, and subsidy deduction.

3. Model Hypotheses

3.1. The Conceptual Model

Combined with the previous empirical research results, the one-stage normative model was developed by SEM based on the TPB through adding the relevant factors. The model shown in Figure 4 includes independent and dependent variables: attitude, subjective norm, perceived behavior control, infrastructure, novelty seeking, subsidy deduction, non-financial incentive, product cognition, environmental concern, and adoption intention.

3.2. The Hypotheses

(1)

Basic behavioral elements

In the TPB model, the behavior is motivated by the rational evaluation of the consequences of action and perceived controllability of behavior, which is expressed through the attitude [51]. The behavioral intention refers to the judgment for the probability of a subjective behavior occurring. The behavior attitude is the assessment for the performance of a particular behavior and the attitude influences the behavioral intention [16,70]. The subjective norm refers to the social pressure to perform a particular behavior [52], which is influenced by the normative belief and motivation. Scholars have proposed that the subjective norm represents the social influence [71]. The perceived behavior control represents the degree to which an individual’s perception that it is easy or difficult to perform a particular behavior, which describes the individual’s perception of factors that facilitate or hinder the behavioral performance. Thus, the following hypotheses are proposed.

Hypothesis 1 (H1).

The attitude has a positive influence on adoption intention for EVs.

Hypothesis 2 (H2).

The subjective norm has a positive influence on adoption intention for EVs.

Hypothesis 3 (H3).

The perceived behavior control has a positive influence on adoption intention for EVs.

(2)

Infrastructure

The inadequate charging infrastructure is a barrier limiting the adoption of EVs [31,72]. The survey by Du et al. [1] shows that, more than 2/3 of respondents were willing to adopt EVs if the charging infrastructure was improved since the infrastructure construction is unsatisfactory at present [31,70,73]. Most charging infrastructure is concentrated in the developed coastal areas of China. Therefore, there are great differences in the infrastructure construction among the different city levels. In this regard, the following hypotheses are proposed.

Hypothesis 4 (H4).

The high level of well-constructed public infrastructure has a positive impact on adoption intention for EVs.

(3)

Novelty seeking

Novelty seeking refers to the pursuit of new products [74] and the electric vehicle (EV) is defined as an environmental-protection product. Choi and Johnson [75] suggest that the willingness to purchase green products is high if there is a high desire for novelty, while consumers perceive the purchase of green products as a novelty-seeking experience [65]. Jahanshahi and Jia [64] find that novelty seeking is a key factor affecting the behavior in purchasing green products, therefore, the following hypothesis is formulated.

Hypothesis 5 (H5).

The novelty seeking has a positive influence on adoption intention for EVs.

(4)

Incentive policy

The current policy to stimulate the adoption intention for EVs can be divided into a direct monetary subsidy and policy incentive [16,70,76]. The direct monetary subsidy results in the direct reduction of expenditure on EVs. The policy incentive results in increasing convenience in the use of EVs, such as an exemption from taxes, free parking, a separate allocation of EV license plates, removal from traffic restrictions, and access to bus lanes. Previous studies suggest that a monetary subsidy can increase the adoption intention [36,77,78]. The Chinese government has partly implemented a subsidy deduction which will be completely enforced by 2022, to shift the development of the EV market from policy-oriented to market-oriented. The subsidy-deduction policy has a certain impact on the adoption intention for EVs. The policy incentive positively influences adoption intentions [16,70], and especially, the restricting of fuel vehicle licenses and removal from traffic restrictions have played a promoting role in the adoption of EVs [52]. Thus, the following hypotheses are suggested.

Hypothesis 6 (H6).

The subsidy deduction has a negative influence on adoption intention for EVs.

Hypothesis 7 (H7).

The non-financial incentive has a positive impact on adoption intention for EVs.

(5)

Product cognition

Personal knowledge and experience can expand the acceptance of EVs [79], while a lack of knowledge and experience with EVs creates a barrier to adopting them [80]. Huang and Ge [16] argued that consumers’ cognitive status affects EVs adoption intentions, such as knowledge and an incentive policy about EVs. Consumers’ perceptions of low-carbon vehicle policies influences consumer behaviors and adoption intentions [81]. Caperello and Kurani [82] argue that detailed energy-saving and pro-environment labels can increase consumers’ willingness to purchase energy-saving and pro-environment vehicles. Thus, the product cognition was added in the TPB model, and the following hypothesis is proposed.

Hypothesis 8 (H8).

The product cognition has a positive influence on adoption intention for EVs.

(6)

Environmental concern

The environmental concern reflects a personal concern for the environment [83]. Ellen et al. [84] suggests that environmental concern is a key motivator of environmentally conscious behavior [75], while Hanson [85] and Maichum et al. [86] suggested that environmental concern affects green-product-purchasing behavior. As we know, the EVs are environment-friendly green products, and the adoption for EVs rather than fuel vehicles is helpful for environmental protection [87]. Thus, it can be assumed that environmental concern positively influences adoption intention for EVs.

Hypothesis 9 (H9).

Environmental concern has a positive influence on adoption intention for EVs.

4. Methodology

4.1. Questionnaire Design

The questionnaire method was engaged to collect data in this paper, which contains two parts. The first part was a survey on the attitude, subjective norms, perceived behavior control and adoption intention, the scale of which was modified based on Huang and Ge [16], Ajzen [51] and Bamberg et al. [88]. The scale of “infrastructure” was developed based on Lin and Wu [18] and Du et al. [1]. The scale of “novelty seeking” was developed based on Choi and Johnson [75]. The scale of “subsidy deduction” was self-developed. The scale of “product cognition” was modified based on Huang and Ge [16] and Lin and Wu [18]. The scale of “environmental concern” was developed based on Choi and Johnson [75] and Wang et al. [70]. The second part was a survey on demographic variables, including gender, age, education level, annual income after tax, marital status, number of children in a family, and number of vehicles in a family. Except for the demographic variables, the questions in the questionnaire were assigned on a Likert 5-point scale from 1–5 in terms of importance (1 for strongly disagree and 5 for strongly agree) (

Table A1. Questionnaire items.

Constructs	Item
Attitude (AT)	I think the development of EVs is good for the environment (AT1)
	I support the country to introduce more policies to encourage individuals to buy EVs (AT2)
	I think buying an EV is a good choice (AT3)
Subject Norm (SN)	The opinion of my family members is an important factor in my decision to buy an EV (SN1)
	If someone around me buys an EV, his behavior will motivate me to buy an EV (SN2)
	Media positive coverage of EVs will motivate me to buy an EV (SN3)
Perceived Behavioral Control (PBC)	It is up to me to buy an EV (PBC1)
	I can afford to buy an EV (PBC2)
	It is easy to buy EVs in my city (PBC3)
Infrastructure (IC)	I live in a city with a very good EVs infrastructure (IC1)
	I live or work in or near a place where I can charge an EV (IC2)
	I would choose to buy an EV if charging facilities were better (IC3)
	I live in an area where EVs charging facilities are available (IC4)
Novelty Seeking (NS)	I am always looking for information about new products and brands (NS1)
	I am always looking for new product experiences (NS2)
	I consider EVs to be a fashionable and cutting-edge technology (NS3)
Subsidy Deduction (SD)	I would not buy an EVs if the EVs purchase subsidy was reduced (SD1)
	I care about EVs subsidies (SD2)
	A reduction in subsidies would increase the price of EVs, which would affect my adoption intention to purchase them (SD3)
Non-Financial Incentive (NFI)	Removal from traffic restrictions for EVs will encourage me to buy one (NFI1)
	The policy of unlimited license plates for EVs would encourage me to buy one (NFI2)
	The ability to use bus lane will encourage me to buy (NFI3)
	Exemption from vehicle purchase tax will motivate me to buy an EVs (NFI4)
Product cognition (PC)	I understand that EVs are environmentally friendly products (PC1)
	I know that EVs can save energy and protect the environment through fuel substitution (PC2)
	I am aware of the preferential policies of EVs (PC3)
Environmental Concern (EC)	I care about energy saving and environmental protection (EC1)
	I have a sense of mission to protect the environment and save energy (EC2)
	I think vehicle exhaust is causing a lot of pollution to the environment (EC3)
Adoption Intention (AI)	I am willing to buy an EV in the future (AI1)
	I will recommend my friends and relatives to buy EVs in the future (AI2)
	I will choose an EV when I buy my second car (AI3)

). Due to COVID-19, the questionnaire was distributed to respondents in Chinese FST and TFT cities, except for Hong Kong, Macau and Taiwan, through the online “Wenjuanxing” platform in February 2021. After four months of intermittent distribution, the questionnaires were collected through social platforms and questionnaire collection and payment services. Although an online questionnaire is a very convenient and fast method to collect information, it still has some limitations. For example, when interviewees have questions about the questionnaire, the interviewers cannot answer them in time. In addition, it is impossible to supervise whether those respondents answer the questions seriously. The longer respondents take to answer the questionnaire, the higher the quality of the information obtained [89], for example, a questionnaire completed in five minutes or more was considered valid. In addition, the sample size of subjects should ideally be above 200 if stable SEM analysis results are sought [90]. In total, 955 questionnaires were collected. The number of valid questionnaires was 858, of which 434 were from FST cities and 424 were from TFT cities. The city classification is based on the official criteria of the Chinese State Council. Due to the questionnaire setup, the respondents were not able to submit the questionnaire if there were unfilled items, therefore, there was no missing data.

4.2. Demographic Variables

The distribution of respondents is shown in

Table 1. The questionnaire sample distribution.

Respondents’ Characteristic	FST City (N = 434)		TFT City (N = 424)
	Frequency	Percentage (%)	Frequency	Percentage (%)
Gender
Female	181	41.70%	150	35.40%
Male	253	58.30%	274	64.60%
Age
Under 18	3	0.70%	2	0.50%
18–25	126	29.00%	180	42.50%
26–30	143	32.90%	86	20.30%
31–40	99	22.80%	68	16.00%
41–50	44	10.10%	49	11.60%
51–60	19	4.40%	38	9.00%
60 above	0	0.00%	1	0.20%
Education
Junior high school and below	2	0.50%	8	1.90%
High school	20	4.60%	41	9.70%
Associate degree	62	14.30%	88	20.80%
Bachelor degree	204	47.00%	215	50.70%
Master degree	125	28.80%	59	13.90%
Doctor degree	21	4.80%	13	3.10%
Annual income after tax (RMB)
50,000 below	85	19.60%	159	37.50%
50,000–100,000	126	29.00%	135	31.80%
110,000–150,000	75	17.30%	73	17.20%
160,000–200,000	64	14.70%	33	7.80%
210,000–250,000	25	5.80%	8	1.90%
250,000 above	59	13.60%	16	3.80%
Marital status
Single	231	53.20%	242	57.10%
Married	198	45.60%	175	41.30%
Others (Widowed or Divorced)	5	1.20%	7	1.70%
Number of vehicles in family
0	89	20.50%	92	21.70%
1	226	52.10%	212	50.00%
2	86	19.80%	101	23.80%
3	22	5.10%	15	3.50%
4	6	1.40%	2	0.50%
5	0	0.00%	1	0.20%
6 or more	5	1.20%	1	0.20%

. The proportion of men in the questionnaire was higher than that of women. Additionally, there are more men (51.24%) than women (48.76%) in the national statistics [91]. There was a great difference in annual income after-tax between the FST and TFT cities. According to the new energy vehicle industry report released by CTR in May 2020, our survey samples were mostly concentrated in the age range of 18–25, 26–30 and 31–40, which is in line with the demographic characteristics of EV consumers, therefore the sample was reasonable and representative.

4.3. Data Analysis

4.3.1. Measurement Model

The reliability of the survey data was tested. For example, if the value of α is less than 0.5, the reliability is insufficient [92]. The results showed that the Cronbach’s α coefficients for all variables were greater than 0.5, except for perceived behavior control (denoted by “PBC”). The values of α coefficient of the PBC in the FST cities rose to 0.618, and the values in the TFT cities rose to 0.546 after the first question for perceived behavior control (PBC1) was deleted. The values of Cronbach’s α coefficient in the questionnaire for the FST cities and TFT cities were 0.925 and 0.915, respectively. This reflects the high reliability of the questionnaire. The KMO value and Bartlett’s sphericity were tested. The results showed that the KMO values for the FST and TFT cities were 0.899 and 0.887, respectively, which exceeded the critical KMO value of 0.7 [93,94]. This indicated that the questionnaire passed the validity test. The result of the Bartlett’s sphericity test was significant (p < 0.000) [95]. A confirmatory factor analysis was performed on each variable using AMOS24.0 software. The results are shown in

Table 2. The results for confirmatory factor analysis.

Constructs	Indicators	S.td Factor Loading		SMC		Convergent Validity				Ave Square Root		Cronbach’s Alpha (α)		p Value
						CR		AVE
		FST Cities	TFT Cities	FST Cities	TFT Cities	FST Cities	TFT Cities	FST Cities	TFT Cities	FST Cities	TFT Cities	FST Cities	TFT Cities	FST Cities	TFT Cities
AT	AT1	0.8663	0.8634	0.7505	0.7455	0.8595	0.8444	0.6721	0.6496	0.8198	0.8060	0.848	0.815	***	***
	AT2	0.8508	0.9053	0.7239	0.8196									***
	AT3	0.7361	0.6195	0.5418	0.3838										***
SN	SN2	0.8428	0.7630	0.7103	0.5822	0.8246	0.7460	0.7015	0.5950	0.8376	0.7714	0.818	0.733	***	***
	SN3	0.8323	0.7796	0.6927	0.6078
NFN	NFN1	0.9122	0.8855	0.8321	0.7841	0.9101	0.8682	0.7717	0.6892	0.8785	0.8302	0.899	0.859
	NFN2	0.9012	0.8809	0.8122	0.7760									***	***
	NFN4	0.8191	0.7124	0.6709	0.5075									***	***
SD	SD1	0.7493	0.6621	0.5615	0.4384	0.8702	0.8168	0.6919	0.6002	0.8318	0.7747	0.869	0.814	***	***
	SD2	0.8740	0.8056	0.7639	0.6490
	SD3	0.8663	0.8446	0.7505	0.7133									***	***
NS	NS1	0.9040	0.8065	0.8172	0.6504	0.8566	0.7992	0.7496	0.6656	0.8658	0.8158	0.781	0.740
	NS2	0.8258	0.8251	0.6819	0.6808									***	***
IC	IC1	0.8953	0.8229	0.8016	0.6772	0.8322	0.8870	0.7135	0.7978	0.8447	0.8932	0.709	0.745
	IC2	0.7908	0.9584	0.6254	0.9185									***	***
EC	EC1	0.9048	0.9199	0.8187	0.8462	0.8872	0.8945	0.7973	0.8092	0.8929	0.8996	0.768	0.768
	EC2	0.8809	0.8787	0.7760	0.7721									***	***
PC	PC1	0.8751	0.8279	0.7658	0.6854	0.8375	0.7862	0.7207	0.6479	0.8489	0.8049	0.710	0.639
	PC2	0.8219	0.7813	0.6755	0.6104									***	***
AI	AI1	0.8376	0.8050	0.7016	0.6480	0.8602	0.8512	0.6730	0.6564	0.8204	0.8102	0.855	0.850
	AI2	0.8703	0.8490	0.7574	0.7208									***	***
	AI3	0.7484	0.7748	0.5601	0.6003									***	***

Note: Attitude is denoted by AT, Subjective Norm is denoted by SN, Perceived Behavior Control is denoted by PBC, Infrastructure is denoted by IC, Novelty Seeking is denoted by NS, Subsidy Deduction is denoted by SD, Non-Financial Incentive is denoted by NFI, Product Cognition is denoted by PC, Environmental Concern is denoted by EC, and Adoption Intention is denoted by AI. The number on the far right for indicators represents the question order for these variables in the questionnaire. Square Multiple Correlations is denoted by SMC. Note: *** p < 0.01.

. The standardized factor loading and composite reliability (CR) were greater than 0.7. This indicates that there was sufficient convergence or internal consistency between the observed indicators [96]. The average variance extraction amount was greater than or equal to 0.5, therefore the data had sufficient convergent validity. Thus, unreasonable items such as subjective norm (SN1), perceived behavioral control (PBC1, PBC2, PBC3), infrastructure (IC3, IC4), product cognition (PC3), non-financial incentive (NFN3), and environmental concern (EC3) should be removed from the analysis. Table 2 shows the results of modified confirmatory factor analysis. Each variable met the evaluation index, indicating that the questionnaire had decent convergence validity.

4.3.2. Common Method Variance

Many studies have suggested that there is an artificial covariance between predictor variables and criterion variables created by the context of questionnaire items and the characteristics of the items themselves, which is called a common method variance (CMV) [97]. Most scholars have believed that the self-reported, single-source, cross-sectional survey commonly used by researchers is prone to common method variance [98,99]. The survey data of this study were obtained from self-reported questionnaires. Therefore, it is necessary to analyze whether the result was affected by CMV. This study used two methods to verify this.

First, Harman’s one-way test was constructed by extracting eigenvalues equal to 1 from all survey items in exploratory factor analysis [100]. The largest explained variance of factors was 32.98% and 36.10% for the FST cities and TFT cities, respectively, which indicates that there was no significant CMV. Secondly, some studies have showed that the Harman’s one-way test has certain limitations [97]. Therefore, all indicators were used as new indicators of the single-factor model for confirmatory factor analysis [101]. Compared with the fitting results of the original CFA model (χ²/df = 2.538, CFI = 0.9562, TLI = 0.941, RMSEA = 0.059), the fitting results of the FST cities (χ²/df = 14.344, CFI = 0.541, TLI = 0.493, RMSEA = 0.176) were very poor, compared with the fitting results of the original CFA model (χ²/df = 15.228, CFI = 0.508, TLI = 0.459, RMSEA = 0.181), and the fitting results of the TFT cities (χ²/df = 2.370, CFI = 0.953, TLI = 0.938, RMSEA = 0.057) were very poor, which shows that there was no significant CMV.

4.3.3. Model Fit Test

In order to solve the problem of the multivariate non-normality of data, the maximum likelihood estimation was used to estimate the model, and the Bollen–Stine bootstrap was used to adjust the model. The evaluation indicators were all within acceptable ranges. The results of the FST cities were χ² = 434.4391, 316 df, p < 0.001, χ² /df = 1.37, CFI = 0.98, IFI = 0.98, TLI = 0.98, RMSEA = 0.03, Bollen–Stine p = 0.000. The results of the TFT cities were χ² = 423.30, 312 df, p < 0.001, χ²/df = 1.36, CFI = 0.98, IFI = 0.98, TLI = 0.97, RMSEA = 0.03, Bollen–Stine p = 0.000. The criteria were as following: CFI was higher than 0.95; RMSEA was less than 0.05; IFI was higher than 0.95, and TLI was higher than 0.95 [90]. These results indicate that the theoretical framework hypothesized in this analysis were consistent with the actual survey data.

5. Results and Discussion

5.1. Differences in the Affecting Factors on Adoption Intention

The results for the hypothesis test were obtained through the SEM, which is shown in

Table 3. Results for the hypothesis test.

Hypothesis	Path	Unstandardized Coefficient		S.E.		p Value		Standardized Coefficient		Test Result
		FST Cities	TFT Cities	FST Cities	TFT Cities	FST Cities	TFT Cities	FST Cities	TFT Cities	FST Cities	TFT Cities
H1	AI←AT	0.1975	0.1565	0.0433	0.0604	0.0943 *	0.0096 ***	0.2659	0.1691	Supported	Supported
H2	AI←SN	0.1889	0.2361	0.0399	0.055	***	***	0.2857	0.2844	Supported	Supported
H3	AI←PBC	-	-	-	-	0	0	-	-	NONE	NONE
H4	AI←IC	0.0373	0.0573	0.0311	0.0294	0.2306	0.0514 *	0.0578	0.0986	Unsupported	Supported
H5	AI←NS	0.1362	0.1886	0.0281	0.0406	***	***	0.246	0.272	Supported	Supported
H6	AI←SD	−0.1128	−0.0998	0.0352	0.0412	0.0013 **	0.0154 *	−0.165	−0.1463	Supported	Supported
H7	AI←NFN	0.2767	0.2291	0.0501	0.0429	***	***	0.3442	0.341	Supported	Supported
H8	AI←PC	0.3249	0.2941	0.0514	0.0623	***	***	0.4773	0.3598	Supported	Supported
H9	AI←EC	0.0724	0.172	0.0433	0.051	0.0943 *	***	0.0816	0.1881	Supported	Supported

Note: * p < 0.1, ** p < 0.05, *** p < 0.01.

. Hypothesis 3 failed the validity test, therefore it was removed. Hypothesis 4 was valid for the TFT cities but was not valid for the FST cities. The remaining hypotheses were all valid for the FST and TFT cities.

As for Hypothesis 1, the empirical result shows that attitude had a greater positive influence on the adoption intention for EVs in the FST cities ($\beta =0.2659$) (Figure 5) than the positive influence in the TFT cities ($\beta =0.1691)$(Figure 6). In previous studies, attitude is regarded as an important factor to predict an adoption intention [56]. Consumers with a positive attitude have a strong adoption intention for green products [102,103] and this attitude is more influenced by knowledge and experience [104]. EVs are more accessible to consumers in FST cities, because the number of EVs owned in FST cities are higher than the number in TFT cities. Moreover, it is convenient for consumers in FST cities to obtain the knowledge and experience of EVs, which is helpful for the positive influence of attitude on their adoption intentions for EVs.

The empirical result for Hypothesis 2 shows that the subjective norm has a positive impact on the adoption intention, with a small variability between the FST and TFT cities. The impact of a subjective norm is stronger than attitude. The results from Rogers [105] and Lane and Potter [80] indicate that media and social networks often affect consumer choices. According to the Annual Report of the Massive Data of New Energy Vehicles in China [106], EV users mainly obtain information through the recommendations of their friends, media evaluation, and the Internet. Because there are nearly no obvious differences in the different ways of obtaining information, the impact of subjective norms on the adoption intentions for EVs differed slightly between the FST and TFT cities. Some scholars have proposed that attitude and subjective norms have a positive significant influence on adoption intentions for EVs [4,40,107]. Du et al. [1] suggest that the subjective norm is a powerful factor affecting the adoption intention, compared with other factors. These studies support our results.

As expected, the empirical results confirm that the TPB model was applicable to the survey respondents. This study was an extension of the TPB model. Adoption intention is significantly positively correlated with attitude and subjective norms. Previous studies have found that attitude is the most important influence on an adoption intention [4,86]. Empirical results show that the findings of this study were different from those of previous studies as the impact of subjective norms on adoption intentions was stronger than attitude. This shows that customers’ adoption intentions towards EVs can be influenced by external factors rather than internal factors. As for Hypothesis 3, the perceived behavior control was removed from this study because it failed the validity test. Some scholars have suggested that perceived behavior control has an impact on behavioral intention [4,16]; however, in this study, this result could not be found. This can be explained through the differences in the sample characteristics.

The result for Hypothesis 4 was not significant in the FST cities. The TFT cities have a disadvantage in their infrastructure, especially in the number of public charging piles. The data from the China Electric Vehicle Charging Infrastructure Promotion Alliance [108] shows that the number of EVs and public charging piles in China is 3.855 million and 496 million, respectively, most of which are concentrated in the developed coastal areas. Thus, infrastructure more greatly affects the adoption intentions in TFT cities. This result is supported by Lin and Wu [18] and Huang and Qian [39]. In recent years, a relevant policy has been implemented by the Chinese government to promote the development of the EV industry and the construction of charging piles. There is a regional difference in the infrastructure construction for EVs, because of the difference in government financial investment between the FST and TFT cities. The infrastructure construction is now concentrated in the FST cities where EV users are active. This results in an uneven distribution of charging infrastructure in the different regions, thereby resulting in differences in the adoption intentions for EVs between the FST and TFT cities.

The result for Hypothesis 5 shows that novelty seeking had a positive influence on the adoption intention for EVs in the different city levels. The adoption intention in the TFT cities ($\beta =0.2720)$ was more greatly influenced than the intention in the FST cities ($\beta =0.2460)$. This indicates that customers in the TFT cities are more willing to accept new products. This novelty seeking factor has rarely been considered as an influencing factor on the adoption intentions for EVs in previous studies. The previous studies have showed that consumers are more willing to accept green products when they have a high desire for the newness in the shopping process and adoption decision-making process [105]. Therefore, the desire for newness has a positive impact on early adoption [74,75,109,110]. The adoption intention in the TFT cities ($\beta =0.2720)$ was more greatly influenced than in the FST cities ($\beta =0.2460$). The EV industry normally has earlier access to the markets in FST cities, but it is also easy for customers in TFT cities to learn of new products and services in the information era. Conversely, it is inconvenient for customers in TFT cities to adopt novel products and services in the beginning because of barriers, such as imperfect infrastructure and sales channels, which attract customers’ innovation desires. This can explain the reason why novelty seeking had a higher influence on the adoption intention of EVs in the TFT cities than the FST cities.

The result for Hypothesis 6 shows that the subsidy deduction had a negative effect on the adoption intention for EVs. The adoption intentions in the FST cities ($\beta =-0.1650)$ were more vulnerable to the subsidy deduction, compared with the TFT cities ($\beta =-0.1463)$. Wang [37] found a subsidy deduction of 50% in 2019 compared to 2018. Therefore, the sales numbers of EVs were 1.206 million in 2019, which had declined by 4% compared with 2018. This was the first decrease since 2009. The study by Li et al. [40] shows that the subsidy deduction had little impact on the adoption intention for EVs in Luoyang, China. The reason for this is that the consumers in small- and medium-sized cities are more concerned about the price rather than the subsidy [39]. In our study, we found that the adoption intentions for EVs in FST cities were more vulnerable to the subsidy deduction. The reason for this may be that the customers in TFT cities focus more on product quality, infrastructure, and so on, rather than the subsidy deduction. Relatively, the customers in FST cities are more sensitive to subsidies because of good infrastructure and fuel vehicle restriction.

The result for Hypothesis 7 shows that the non-financial incentive had a greater influence on the adoption intention for EVs in the FST cities ($\beta =0.3442)$ compared with the TFT cities ($\beta =0.3410)$, but the difference was relatively small. In the FST cities, fuel vehicles are restricted by policies, such as traffic control and use restrictions. The non-financial incentive factor includes no traffic control, no purchase restrictions, and the use of bus lanes. The main reason for customers in FST cities to adopt EVs is the restrictions on driving and license plates on fuel vehicles. EVs provide convenience for customers to travel compared to fuel vehicles while the restrictions on the use and adoption of fuel vehicles in TFT cities are not as strict as those in FST cities. Therefore, the customers in TFT cities are relatively slightly affected by the non-financial incentive compared with FST cities.

The result for Hypothesis 8 shows that the product cognition was the key factor influencing the adoption intention for EVs. The product cognition had a greater impact on the adoption intention in the FST cities ($\beta =0.4773)$ than its impact in the TFT cities ($\beta =0.3598)$. Some scholars have suggested that the adoption intention is influenced by product performance [61,62,111,112] and the adoption intention is affected by the product cognition, because consumers sometimes have prejudices against the product for due to the unknown real performance of EVs [4,18]. Thus, the adoption intention for EVs in the FST cities was more likely to be influenced by product cognition. The reason for this is the restriction in most FST cities, for example, that only one EV can be used as the main household vehicle. Some consumers would be more likely to adopt EVs than traditional fuel vehicles because the cost of access to EVs is lower than for traditional fuel vehicles, according to EVs policies in Shanghai. Thus, customers in FST cities will pay more attention to the knowledge of product polices for EVs. Conversely, EVs can be used as an alternative household vehicle in the TFT cities but they have more alternatives to choose from and do not necessarily choose EVs.

The result for Hypothesis 9 shows that the environmental concern had a positive effect on the adoption intention, which is consistent with the findings in the study of Zhang et al. [61]. Environmental concern had a greater impact on the adoption intention in the TFT cities ($\beta =0.1881)$ than the FST cities ($\beta =0.0816)$. EVs are more environmentally-friendly compared to traditional fuel vehicles and customers with higher environmental concern are more likely to adopt the use of EVs. In recent years, the TFT cities have been more eager to create a national ecologically-civilized city, and local governments have paid more attention to environmental protection. As a result, the higher environmental concern of residents in TFT cities has more greatly stimulated the adoption intention for EVs.

5.2. Differences in the Effects of Demographic Characteristics on Adoption Intention

The demographic variables were tested using the ANOVA to explore the differences in adoption intention for EVs between the FST and TFT cities. The result is shown in

Table 4. Effects of demographic characteristics in FST and TFT cities.

Variable	FST Cities				TFT Cities
	Mean	Std. Deviation	F	p Value	Mean	Std. Deviation	F	p Value
Gender
Female	3.8508	0.6642	2.745 *	0.098	3.8844	0.7204	7.758 ***	0.006
Male	3.7154	0.9451			3.6472	0.8966
Age
Under 18	3.2222	0.1925	0.406	0.845	2.6667	0.9428	3.732 ***	0.001
18–25	3.791	0.8597			3.7889	0.7280
26–30	3.7366	0.8096			3.8023	0.6905
31–40	3.7879	0.9264			3.5392	1.0042
41–50	3.7803	0.8163			3.4218	1.1522
51–60	3.8947	0.6091			4.0877	0.7215
60 above	0				4
Education
Junior high school and below	4	0	0.268	0.931	3.1667	0.6424	2.022 *	0.075
High school	3.7	0.6389			3.6748	0.9986
Associate degree	3.8226	0.9286			3.8409	0.8562
Bachelor degree	3.799	0.8567			3.7349	0.8175
Master degree	3.7253	0.7977			3.7797	0.7444
Doctor degree	3.6825	0.9278			3.2308	1.0575
Annual income after tax (RMB)
50,000 below	3.9059	0.8398	1.423	0.214	3.6813	0.8116	1.863 *	0.1
50,000–100,000	3.7857	0.7260			3.7531	0.8395
110,000–150,000	3.8133	0.7637			3.7945	0.8474
160,000–200,000	3.7396	0.9804			3.596	1.0367
210,000–250,000	3.76	0.6125			3.3333	0.7346
250,000 above	3.5367	1.0468			4.2292	0.6962
Marital status
Single	3.7734	0.8393	0.192	0.826	3.7617	0.7686	0.44	0.644
Married	3.7643	0.8493			3.6857	0.9433
Others (Widowed or Divorced)	4	0.7071			3.8095	0.8576
Number of vehicles in family
0	3.7566	0.7650	3.459 ***	0.004	3.7138	0.8639	0.288	0.943
1	3.8083	0.8181			3.7516	0.7950
2	3.6938	0.8559			3.7063	0.9525
3	4.0303	0.7267			3.6444	0.8015
4	3.9444	0.8798			3.6667	0
5	0				4.6667
6 or more	2.4	1.9494			4

Note: * p < 0.1, *** p < 0.01.

.

Females in FST and TFT cities are more likely to adopt EVs than males while residents in FST cities are more likely to adopt EVs when they have three vehicles in their family. Hidrue et al. [113] found that multicar households preferred to adopt EVs. Non-adopting households were mainly affected by the cruising range of EVs, whereas the multi-vehicle households had a backup vehicle to extend cruising range, therefore, they were not limited by the short cruising range of EVs [114,115]. Marital status had no effect on the adoption intention in the FST or TFT cities. In the FST and TFT cities, 48.6% and 69.3% of respondents, respectively, had an annual after-tax income below CNY 100,000. Income above CNY 250,000 Yuan was 13.6% in the FST cities and only 3.8% in the TFT cities (Table 1). Thus, there is a serious income gap between the FST and TFT cities. The income level in the TFT cities affected the adoption intention, while the income level in the FST cities did not. This is consistent with the study by Xiong et al. [38]. The adoption intention was stronger in the TFT cities with an upper-middle education level. Zeng et al. [116] showed that the consumer’s education presented a positive relationship with green purchasing participation. The adoption intention for EVs in the FST cities was not impacted by the education factor. Age had a significant impact on the adoption intention in the TFT cities, which is consistent with the findings of previous studies [18,117]. The respondents aged 51–60 and 26–30 were more willing to adopt EVs. The reason for this may be that older people aged 51–60 have a stable career and relatively strong economic power to adopt more expensive EVs whereas EVs are emerging products that are more attractive to younger people aged 26–30. In addition, our survey revealed that there were significantly more factors affecting the adoption intentions for EVs in TFT cities in terms of demographic variables, such as age, education and income, compared to FST cities. The reason for this is that, based on the policy, customers in FST cities have limited options to adopt vehicles because of vehicle use and purchase restrictions, which reduce the influence of demographic variables on the adoption intentions for EVs. Conversely, customers in TFT cities can consider more factors in their adoption of EVs, because the traffic and purchase restriction policies are not obvious. Therefore, there are regional differences in the adoption intentions for EVs between FST and TFT cities.

According to the empirical results, there were differences between the FST cities and TFT cities in the factors affecting consumers’ adoption intentions for EVs. For the TFT cities with poor infrastructure, consumers were more concerned about the degree of infrastructure improvement, while in the FST cities they were not. Product cognition was the key influencing factor in this study, which is different from previous studies. A study on Finland consumers found that the adoption intention of EVs was relatively low when consumers had insufficient knowledge about EVs [118]. Similarly, the lack of understanding of EVs leads to a weak intention to adopt them among UK consumers [119,120]. Previous studies, meanwhile, have found attitude to be the strongest factor [4,86], but subjective norms had a stronger impact than attitudes in our study. Due to the positive external effect of media campaigns and social networks, Chinese consumers are more willing to purchase EVs than Brazilian and Russian consumers [121]. Moreover, compared with Americans, the Chinese are more willing to purchase EVs due to environmental concerns [20]. Furthermore, subsidy deduction was the only influencing factor that negatively and significantly affected adoption intentions, indicating that consumers prefer to purchase under subsidies, which is consistent with previous studies [16,87].

6. Conclusions and Policy Implications

Considering the lack of analysis of regional differences in the adoption intentions for EVs from different city levels in the existing studies, this paper originally investigated the differences in influencing factors on the adoption intention for EVs between Chinese FST and TFT cities. This study provides a new insight into related research areas and expands the TPB model using survey data from respondents in Chinese FST and TFT cities to explore the different effects of demographic characteristic variables and factors, such as attitude, subjective norms, infrastructure, novelty seeking, subsidy deduction, non-financial incentives, product cognition, and environmental concerns, on the adoption intention for EVs between those Chinese FST and TFT cities.

The following conclusions can be drawn. All factors have a significant positive impact on the adoption intention for EVs between FST and TFT cities, except the perceived behavior control and subsidy deduction. The infrastructure has no significant impact on the adoption intention for EVs in FST cities while the subsidy deduction has a negative impact on the adoption intentions for EVs in both FST cities and TFT cities. Product cognition has the most positive influence on the adoption intention for EVs in different tier cities. Moreover, there are differences in the influencing factors on adoption intention for EVs between FST cities and TFT cities. The attitude, subjective norm, subsidy deduction, non-financial incentive and product cognition has a greater impact on the adoption intentions for EVs in FST cities than the impact in TFT cities. The infrastructure, novelty seeking and environmental concerns have a greater influence on customers’ adoption intentions for EVs in TFT cities than their influence in FST cities. As for the demographic variables, there are more factors affecting TFT cities, namely, females are more likely to adopt EVs, only gender and the number of vehicles in a family have an impact on adoption intentions in FST cities, and gender, age, income and education level have an influence on the adoption intentions in TFT cities.

The findings of this study can be widely applied to companies and provide implications for government policy development. The following constructive policy suggestions will be indicated. First, there are regional differences in influencing factors on the adoption intentions for EVs in different tier cities. In this regard, the government and enterprises should consider the regional differences when formulating policies and promoting sale strategies based on the local conditions. Second, the subsidy deduction has a more negative impact on the adoption intention for EVs in FST cities than TFT cities. In the period of subsidies deductions, the government should replace the subsidy of the whole vehicle with a charging subsidy, power exchange subsidy, insurance subsidy, maintenance subsidy, and right-of-way concessions, to reduce the operating costs of EVs. Moreover, it is essential to improve the production technology and reduce production costs to lower the selling price of EVs, making them more cost-effective compared with traditional fuel vehicles. Third, the infrastructure has an impact on adoption intentions only in TFT cities. In addition, the TFT cities have an enormous population base, low vehicle ownership and huge market potential. Thus, the government should introduce more policies and increase funding to support the development of the EV market and its infrastructure for TFT cities. Fourth, enterprises should pay more attention to the promotion of advertisement campaigns for EVs, develop and promote EVs and consider the differences of demographic characteristics in different regions. Fifth, the findings showed that product cognition is the strongest influencing factor for the adoption intention for EVs. Therefore, informing consumers about the real performance and environmentally-friendly features of EVs can promote consumers’ ecological values and social responsibility. The government should publicize the incentive policy and environmental protection effect of EVs. Companies can cultivate and establish short video accounts and WeChat official accounts to popularize environmental protection knowledge and the use of EVs. Sixth, novelty seeking has a greater impact on the adoption intentions in TFT cities than FST cities, therefore more attention should be focused on propaganda of the latest EV technological innovations that is advertised to consumers in TFT cities. People who are willing to pursue new things are more likely to accept EVs. Thus, more attention should be paid to the novelty of EVs to attract customers. Other countries can learn from the success cases of Chinese cities to develop their own electric vehicle promotion according to their own local situations.

There are some limitations in our research. In our future research, we plan to conduct a separate analysis for EV owners vs. non-EV owners to obtain more meaningful results. We could further divide cities according to their characteristics, such as urban economic level, transportation, tourism, education, and vehicle ownership. In addition, mediating or moderating effects could be added to future models, such as adding demographic variables as a moderating variable to our model. Moreover, factors affecting different tier cities could be examined in one framework and since other countries were not fully considered in this study and only China was examined, it is necessary to carry out empirical research and comparative analysis on the purchasing intention of EVs in other countries in the future.