Front. Energy Res. Frontiers in Energy Research Front. Energy Res. 2296-598X Frontiers Media S.A. 782992 10.3389/fenrg.2021.782992 Energy Research Original Research How Does the Consumers’ Attention Affect the Sale Volumes of New Energy Vehicles: Evidence From China’s Market Jiang et al. Consumer's Attention Affects NEVs Sales Jiang Zhe 1 2 Long Yin 3 Zhang Lingling 1 * School of Economics and Management, University of Chinese Academy of Sciences, Beijing, China School of Energy and Environment, City University of Hong Kong, Hong Kong, China Troop of People’s Liberty Army, Beijing, China

Edited by: Quande Qin, Shenzhen University, China

Reviewed by: Huaping Sun, Jiangsu University, China

Hongyan Zhang, China University of Petroleum (Huadong), China

 *Correspondence: Lingling Zhang, zhangll@ucas.ac.cn

This article was submitted to Sustainable Energy Systems and Policies, a section of the journal Frontiers in Energy Research

 08 12 2021 2021 9 782992 25 09 2021 27 10 2021 Copyright © 2021 Jiang, Long and Zhang. 2021 Jiang, Long and Zhang

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

The promotion of new energy vehicles is a grand plan across countries to achieve carbon neutrality and air purification. The sale volume of new energy vehicles is affected by many factors, yet it is the attitude of consumers themselves that has the final decisive role. We use four representative Baidu search indexes as the variables representing the attention of consumers and take variables of economic, population, and income as control variables for regression analysis from the national and sub-economic regional perspectives respectively. Results show that search indexes of “new energy vehiclek.” “new energy vehicles battery”, and ‘charging pile’ all have significant positive impacts on the sales of new energy vehicles to varying degrees while the index of ‘automobile spontaneous combustion’ has a significant negative impact on the sale volume. This study, therefore, verifies that the consumer attention represented by search indexes is an important yet uncovered factor affecting the sale volume of new energy vehicles. Besides, due to people’s prejudice against spontaneous combustion accidents of new energy vehicles, consumers have a cognitive bias about the spontaneous combustion rate of new energy vehicles especially in less developed areas of China.

 new energy vehicle consumer attention sale volume cognitive bias search trend National Natural Science Foundation of China10.13039/501100001809 1 Introduction

The proliferation of new energy vehicles is considered an important step towards the reduction of harmful gases and achieving carbon neutrality (Thomas, 2009; Hill et al., 2019; Spangher et al., 2019). Accordingly, governments around the world plan to adopt new energy vehicles to build a sustainable transport system and have set goals to promote new energy vehicles adoption (Mock and Yang, 2014; Nie et al., 2016). Yet, despite continuous financial supports and technological advances, the current market penetration rate of new energy vehicles is still quite low (Wang et al., 2019). Therefore, the academic has conducted extensive research on the influencing factors of the sales of new energy vehicles. The existing literature investigates a comprehensive list of elements that are related to the adoption of new energy vehicles, including but not limited to the four following categories: policy, economic, socio-demographic, and personal preference factors. Table 1 summarizes some typical literature on factors impacting the sale volume of new energy vehicles.

Literature on factors affecting new energy vehicles.

Category References Country Research interest Main results
Policy Ma et al. (2017) China Purchase subsidy Positive co-integration for the relationship between the new energy vehicles market share and the new energy vehicles purchase subsidy
Wu et al. (2021) China tightened dual-credits regime Remarkable policy pressure and inevitable execution challenges of the recently tightened dual-credits regime
Jenn et al. (2013) US Energy Policy Act of 2005 It increased the sales of hybrids from 3 to 20% depending on the vehicle model considered
Economic Palmer et al. (2018) UK, US, and Japan total Cost of Ownership Market share was found to be strongly linked to hybrid electric vehicle Total Cost
Sovacool et al. (2019) China cost considerations Chinese consumers have strong preferences about the related costs when they think about purchasing electric vehicles
Soltani-Sobh et al. (2017) US electricity price Electricity price is negatively associated with electric vehicle use
Socio-demographic Zarazua de Rubens (2019) Nordic countries clusters demographics Three key elements regarding the price, the range, and the environmental attributes of electric vehicle
Higueras-Castillo et al. (2020) Spain clusters demographics The driving range is the most important factor
Berkeley et al. (2018) UK clusters demographics High purchase price and the availability of public charging stations emerged as the most substantive barriers to electric vehicle adoption
Personal preference Ma et al. (2019) China preference for electric vehicles Prices, car classification, and powertrain types were the most important factors influencing consumer response
Liao et al. (2019) Netherlands preference for electric vehicles Vehicle leasing is the most popular option while battery leasing is less preferred than full price purchase
Wang et al. (2020) China preference for electric vehicles Safety awareness is closely related to the sale volume of electric vehicle

Although the above literature has studied the factors affecting the sales of new energy vehicles from various aspects, there exists little research focusing on the impact of consumer attention on sales of new energy vehicles. Additionally, the sale volume of commodities is ultimately determined by the behavior of consumers and the quantification of consumer attention can extract the important factors affecting the sale volume. In previous literature, the depiction of consumers’ behavior is mainly obtained through the questionnaire survey (Hao et al., 2016; Liao et al., 2019; Maheshwari et al., 2016). However, the questionnaire survey has a certain lag and cannot reflect the real-time attitude and attention of consumers. On the contrary, the emergence of the Internet has changed consumers’ consumption habits for the development of search engines has made it possible to provide related information immediately. Thanks to the emergence of Internet search engines, for example, Google and Baidu, the detailed information of almost all products can be easily found through the Internet and consumers are now more likely to get information from the Internet before making a shopping decision.

Because automobiles are relatively higher-value commodities, people tend to spend longer time and obtain more information in comparing potential goods to make the decision(Mitchell and Prince, 1993). Due to the convenience of Internet search engines illustrated in the previous part, customers typically use online searches to access commodity information. Searching for pre-purchase information is regarded as an integral element of the consumer’s buying behavior (Akalamkam and Mitra, 2018). Specifically, about 50% of the customers spend more than ten hours identifying the best matching vehicle for their requirements(Wachter et al., 2019). Some scholars have used the search indexes in the research on the sales of new energy vehicles (Wachter et al., 2019; Yang and Zhang, 2020; Zhang et al., 2017). However, almost all the literature is aimed at sale volume forecasting and as far as we know, there is no attempt to explore the impact of consumer attitudes on sale volume through the search indexes.

To explore this possible relationship between consumer attention and commodity sales, we assume that on the whole, consumers’ search volume for phrases related to new energy vehicles may have a certain impact on the sale volume of new energy vehicles. We use the Baidu Index 1 , a measure of search volume similar to Google Trends, as our dependent variable for we focus on the Chinese market and Baidu is the largest search engine in China with around 79.89% market share 2 . Baidu index is based on the search volume of Internet users, taking keywords as the statistical object, scientifically calculation of the weighted sum of search frequency of the specific keyword in search engine. According to the main factors that consumers need to consider in the purchase of new energy vehicles, we choose Baidu indexes of four keywords as our research objects: “new energy vehicle” “new energy vehicle battery” “charging pile”, and “automobile spontaneous combustion”. Results show that the first three search indexes all have significant positive impacts on the sales of new energy vehicles to varying degrees while the index of ‘automobile spontaneous combustion’ has a significant negative impact on the sale volume.

The contribution of this paper is mainly reflected in the following aspects. First, to the best of our knowledge, it is the first study to explore the impacts of search indexes on new energy vehicle sales. In addition, we not only use the straight-forward keyword of “New energy vehicle” but also commodity-related factors that may affect consumers’ purchase intention as research objects, which sheds light on a new way to investigate the influence of consumer attitude on product sales. Second, the study suggests that consumers in regions of various economic development levels may have different concerns about new energy vehicles when making purchasing decisions, which provides useful information for manufacturers and governments to further promote new energy vehicles. Last, this study has found a cognitive bias in the adoption of new energy vehicles. Specifically, it reveals that people’s attention to ‘spontaneous combustion’ of new energy vehicles has an excessively negative impact on sale volume, but in fact, the spontaneous combustion rate of new energy vehicles is much lower than that of fuel vehicles.

The rest of the paper is organized as follows. Indexes Choice and Research Hypotheses introduces the background and puts forward the research hypotheses. Related Methodologies and Data presents the data and variables used in the study and the models for the research. Empirical Results analyzes the empirical results and Discussion and Conclusion discusses the main findings, gives policy implications, and provides several future research topics.

 2 Indexes Choice and Research Hypotheses

Baidu doesn’t provide search indexes for all keywords, so in our study, we select phrases that both exist in Baidu’s search index library and can point precisely to specific information. The most direct search keyword for consumers when making purchase choices is the product itself, and this choice of keywords can be found in many pieces of literature on the relationship between consumer search index and product sales(Fritzsch et al., 2020; Geva et al., 2015; Ruohonen and Hyrynsalmi, 2017). Additionally, in the previous literature about forecasting new energy vehicles sales using search indexes, researchers all use the products themselves as the search index. Figure 1 also shows the trends of China’s new energy vehicle sales and Baidu search indexes of term ‘new energy vehicle’ from 2016 to June 2021 and it can be seen that the two curves have a certain synchronization. Accordingly, we assume the Baidu index of ‘New energy vehicle’ has an important impact on the sales of new energy vehicles.

New energy vehicle sales and Baidu search index in China from 2016 to 2021.

The search index of “new energy vehicle” has a significant positive impact on the sales of new energy vehicle.

Further, we explore the important factors that influence consumers’ decision to buy new energy vehicles. The main reason why it is difficult for electric vehicles to replace traditional fuel vehicles is the short range restricted by battery capacity and the lack of charging pile facilities (Sun et al., 2018). A specific term that arises from it is “range anxiety”, defined as the psychological anxiety consumers feel when dealing with the limited range of new energy vehicles and it has been labeled as one of the most pressing obstacles to the adoption of new energy vehicles (Noel et al., 2019). In previous studies, Wang et al. (2020) suggest that infrastructure, battery technology, and safety awareness are three main factors influencing the market proportion of new energy vehicles in China by analyzing the comment data in Autohome.com, the most visited professional car website in China. Funke et al. (2019) also highlight the deficiencies of charging infrastructure construction in the promotion of new energy vehicles. Based on this, we choose two keywords, “new energy vehicle battery” and “charging pile” as the research objects.

The search index of “new energy vehicles battery” has a significant positive impact on the sales of new energy vehicles.

The search index of “charging pile” has a significant positive impact on the sales of new energy vehicles.

Another major concern for new energy vehicles is the problem of safety and it is undoubtedly an enduring issue in the automobile market. Egbue and Long (2012) found that reliability and safety are among the biggest concerns for respondents in an American technological university. Wang et al. (2020) also stress the importance of safety in consumers’ purchasing intentions for new energy vehicles. Additionally, spontaneous combustion is the most common safety accident in electric vehicles. Therefore, we use the Baidu index of “automobile spontaneous combustion” as the variable of consumer attention and propose the fourth hypothesis.

The search index of “automobile spontaneous combustion” has a significant negative impact on the sales of new energy vehicles.

 3 Related Methodologies and Data 3.1 Variables and Data 3.1.1 Dependent Variable

Sales refers to the monthly sale volumes of new energy vehicles in the studied provinces during the studied period, i.e., 1st January 2016 to 30th June 2021. The data is provided by Datavision (http://www.dataisvision.com/). It should be noted that only passenger cars are included, while other types of vehicles such as new energy buses are not considered. This is because such vehicles are purchased in a predetermined way with a reasonable procurement process that can not be influenced by consumers. Additionally, the other statistics in our panel data for this study are also measured monthly.

 3.1.2 Independent Variable

BI refers to the Baidu search indexes which are downloaded from Baidu.com. Baidu provides real-time data on keyword search volumes for a specific period and territorial scope. To match the monthly sales of new energy vehicles, we use the monthly Baidu index of each province from 2016 to June 2021. Additionally, we use four keywords in the empirical study: “new energy vehicle” “new energy vehicle battery” “charging pile”, and “automobile spontaneous combustion” and norm them as Index1, Index2, Index3, and Index4 respectively. It should be pointed out that we use the sum of the Baidu index of “new energy vehicle” and “electric vehicle” as Index1 and the sum of “new energy vehicle battery” and “electric vehicle battery” as Index2 for new energy vehicles are mainly electric vehicles in China’s market and these two kinds of terms refer to almost the same information on the Internet. We choose these four keywords as they are all directly related to the information consumers are most concerned about when buying new energy vehicles.

 3.1.3 Control Variables

In this study, following Guo et al. (2020), we use provincial monthly urban population ( U P ), gross domestic product ( G D P ), and disposable income per capita ( D I ) as control variables.

The urban population refers to the volume of permanent urban residents in the studied provinces during the studied period. Population size determines the overall potential demand for the sales of new energy vehicles, so it is one of the determinants of energy vehicles adoption (Egnér and Trosvik, 2018). Considering that China still has a large rural population and the charging facilities in rural areas are difficult to meet the daily use needs of new energy vehicles, we use the urban population as a control variable in the empirical study.

Gross domestic product refers to the total monetary or market value of all the finished goods and services produced within a province’s borders during the studied period. Some studies have found that regional economic development has a positive effect on the sales of new energy vehicles(Javid and Nejat, 2017; Kester et al., 2018). One possible explanation is that regions with better economic development have better infrastructure, and the use of electric cars relies on a large number of public charging devices which demands large-scale investment in public transit facilities(Needell et al., 2016).

Disposable income per capita is calculated by taking income earned from all sources (wages, government transfers, etc.) minus taxes, savings, and some non-tax payments. Disposable income per capita determines an individual’s ability to purchase goods or services, therefore, it is considered to be a significant factor affecting the adoption of new energy vehicles (Langbroek et al., 2016; Mersky et al., 2016; Egnér and Trosvik, 2018; Priessner et al., 2018).

Since the original dataset of U P , G D P , and D I is presented annually, we use a differencing process to transform the data into a monthly form. At first, we acquire the monthly growth rate g using the following equation: g =   q i q i − 1 12 − 1 Where q i denotes the value of the parameter (i.e., urban population, gross domestic product, or disposable income per capita) by the end of year   i , q i − 1 denotes the value of the parameter by the end of year i − 1 . Then, the monthly value of the parameter m i can be calculated using the following equation: m i =   m i − 1 ( 1 + g )

3.1.4 Data Description

This study uses monthly data of 30 provincial-level administrative regions in China from January 2016 to June 2021 for the database we obtained only contains sales statistics of new energy vehicles since 2016. Our study doesn’t include the city of Beijing because Beijing has implemented the automobile purchase restriction policy and Beijing’s new energy vehicle sales are largely determined by the number of vehicle licenses issued by the government of Beijing, but not the true market demand. We also exclude Hong Kong, Macau, and Taiwan in this paper due to missing statistical information. Table 2 summarizes the definitions of all the variables used in the study, as well as their data sources and units of measurement.

Definitions of variables and descriptive statistics.

Category Variables Definition Mean Std.Dev Min Max Obs
Dependent Variables Sales (thousand) Monthly sale volumes of EVs in the studied provinces during the studied period, that is Jan 2016 to June 2021 59.550 50.980 0.062 328.2 1950
Independent Variables Index1 Baidu search index of ‘new energy vehicle’ and ‘electric automobile’ during the studied period 0.441 0.222 0.000 1.598 1950
Index2 Baidu search index of ‘charging pile’ during the studied period 2.198 1.466 0.206 16.470 1950
Index3 Baidu search index of ‘new energy vehicle battery’ and ‘electric vehicle battery’ during the studied period 0.597 0.302 0.000 1.784 1950
Index4 Baidu search index of ‘automobile spontaneous combustion’ during the studied period 0.106 0.093 0.000 0.432 1950
Control Variables UP (million) Urban population, in terms of the volume of permanent urban residents in the studied provinces during the studied period 27.230 18.11 0.912 99.580 1950
GDP (trillion) Province-level GDP value in the studied provinces during the studied period 0.248 0.205 0.009 0.967 1950
DI (thousand) Disposable income per capita in the studied provinces during the studied period, a measurement of purchasing power 3.103 0.761 2.079 6.603 1950

The sales of new energy vehicles are obtained from Dataisvision (http://www.dataisvision.com/). The Baidu indexes are obtained from https://index.baidu.com/v2/index.html#/and the numbers of U P , G D P , and D I are obtained from the Wind database (http://www.wind.com.cn).

 3.2 Methodologies 3.2.1 Panel Unit Root Test

In the case of long panel type data, as non-stationary variables may lead to spurious regression, meaning that there is a high correlation even though variables are unrelated, we have to determine whether the variables contain panel unit roots. The unit root test is used to check whether variables in the panel model are stationary and we utilized Levin–Lin–Chu (LLC) test, and Fisher-type tests in our empirical study.

Levin et al. (2002) suggest a more powerful panel unit root test than performing individual unit root tests for each cross-section. The null hypothesis is that each time series contains a unit root against the alternative that each time series is stationary. The model used can be shown as follows: Δ y i , t =   ρ i y i , t − 1 +   ∑ j = 1 p i δ i , j Δ y i , t − j + ε i , t

The Fisher-type test uses an average statistic as the Im-Pesaran-Shin method (Im et al., 2003) but performs the panel unit root test by using the p-value in a meta-analysis. This test is the most commonly used unit root test and is known to have the highest power (Kim & Jun 2010). Fisher-ADF uses the p-value of the statistic obtained from the ADF unit root test while Fisher-PP uses the p-value of the individual cross-sectional data as follows: P =   − 2 ∑ i − 1 N ln ( p i )

3.2.2 Panel Analysis

When estimating static energy demand models by panel analysis, it is common to explain unobserved heterogeneity by using fixed or random effects. The following linear regression model considers the heterogeneity of the panel data: y i t =   α + β x i t + u i + e i t

In the fixed-effect model, the error term u i of the above equation is regarded as the parameter to be estimated. On the contrary, assuming   u i as a random variable would make it the random effect model.

When choosing between fixed and random effect models, the primary criterion is the inference of the heterogeneity of the panel data. If the panel data are derived from a random sampling of the population, then the error term u i can be assumed to follow a probability distribution. In terms of econometric theory, the choice between the two models is determined based on whether the assumption c o v ( x i t , u i ) = 0   is established. This is called the Hausman test, which sets the null and alternative hypotheses as follows (Hausman, 1978): H 0 :   c o v ( x i t , u i ) = 0

If the null hypothesis is rejected, the random effect model is more efficient, otherwise, the fixed effect model is chosen. Therefore, in this study, we conduct a panel analysis on the unobserved heterogeneity by using the Hausman test.

We analyze the relationship between Baidu search indexes and new energy vehicles sales in China through a panel analysis. We construct regression equations using four Baidu indexes, as follows: S a l e s i t =   α + β I n d e x k , i t + γ ′ ( U P i t , G D P i t , D I i t ) ′ + u i + e i t Where k   =   1,2,3,4 , and I n d e x k denotes the four Baidu indexes. U P , G D P , and D I represent the urban population, province-level GDP value, and deposit income respectively. i denotes the respective province of China, and t   denotes the time period. μ and e are error terms.

Then, we put all the Baidu indexes into the model for regression, as follows: S a l e s i t =   α + β 1 I n d e x 1 i t + β 2 I n d e x 2 i t + β 3 I n d e x 3 i t + β 4 I n d e x 4 i t + γ ′ ( U P i t , G D P i t , D I i t ) ′ + u i + e i t

4 Empirical Results 4.1 Stationary Check and Hausmann Test

Table 3 shows the unit root test results for each variable. If the variable has a unit root, we use the first difference to make the variable stationary. It can be seen that U P , D I , and I n d e x 2 cannot pass the unit root test, so we perform the difference process to obtain D . U P , D . D I , and D . I n d e x 2 to be stationary.

Unit root test results for each variable.

Variable LLC Fisher-ADF Hausman test
Sales −23.178*** −16.664***
GDP −4.237*** −4.9170***
UP 1.108 0.335
D.UP −1.935** −3.028***
DI 1.405 −8.766***
D.DI −31.173*** −10.840***
Index1 −3.426*** −9.897*** 79.81***
Index2 4.317 −6.024***
D.Index2 −32.255*** −17.958*** 158.79***
Index3 −2.057** −9.222*** 115.25***
Index4 -20.183*** −20.999*** 190.87***
All indexes 103.86***

This table shows the unit root test for variables and the Hausmann test for models using different Baidu indexes. The first column uses the LLC, test and the second column uses the Fisher-ADF, test. If the variable has a unit root, we perform the different treatment to make the variable stationary. The third column is the Hausmann test result of Chi-Square Statistic. Asterisk ***,**, and *denote the rejection of the null hypothesis at the 1, 5, and 10% significance level, respectively.

We then perform the Hausman test to choose the fixed effect or random effect model to perform the panel regressions. As shown in Table 3, the null hypothesis of the Hausman test is rejected at the 1% level. That is, the Hausman test shows that the fixed effect estimation is more suitable than the random effect estimation.

 4.2 Consumer Attention to the Sales of New Energy Vehicles

The estimation results are shown in Table 4. The models in the left four columns are regressions using four different search indexes respectively. The fifth model indicates the regression result using all four search indexes and in the sixth model, we omit the control variable of D . U P as it is persistently not significant in the previous regression. At last, we use the cluster method, as presented in the last column. It can be seen that the search amount of ‘new energy vehicle’, ‘new energy vehicle battery’, and ‘charging pile’ are significantly positive with the new energy vehicle sales indicating that the search behavior of these three kinds of terms may have an impact on the sales of new energy vehicles. Additionally, the search amount of ‘automobile spontaneous combustion’ is negative but not significant.

Model estimation results using Baidu search index in the same month.

Variables (1) Index1 (2) Index2 (3) Index3 (4) Index4 (5) All Indexes (6) Omit D.UP (7) VCE
Index1 10.842*** 8.828*** 8.843*** 8.843***
(0.66) (0.74) (0.74) (0.78)
D.Index2 27.523*** 9.829 9.842 9.842**
(7.73) (7.33) (7.33) (3.73)
Index3 45.502*** 26.972*** 26.947*** 26.947***
(4.10) (4.70) (4.70) (5.57)
Index4 −15.558 −45.185*** −45.308*** −45.308***
(11.09) (11.04) (11.03) (12.04)
GDP 27.853** −30.882** −34.489*** −30.480** 7.517 8.830 8.830
(13.23) (13.61) (13.23) (13.69) (13.48) (13.30) (29.75)
D.UP 1.968 3.751 3.796 3.696 1.968
(3.33) (3.55) (3.45) (3.56) (3.30)
D.DI −73.44 65.7 −137.161*** 92.069* −149.729*** −149.132*** −149.132***
(45.46) (47.53) (50.08) (48.03) (47.94) (47.92) (30.64)
Constant 29.850*** 65.893*** 42.846*** 67.079*** 29.179*** 28.959*** 28.959**
(3.99) (3.58) (4.01) (3.84) (4.25) (4.23) (11.29)
Observations 1950 1950 1950 1950 1950 1950 1950
R-squared 0.127 0.011 0.065 0.006 0.145 0.145 0.145
Number of provinces 30 30 30 30 30 30 30
province FE YES YES YES YES YES YES YES
Year FE YES YES YES YES YES YES YES
VCE cluster(province) NO NO NO NO NO NO YES

This table presents the estimation results for impacts of four Baidu search indexes on the sale of new energy vehicles sales of China’s market in the same month. For more details of prediction models, refer to Eqs 7, 8. The four columns on the left indicate results of regressions using four Baidu indexes, respectively. The fifth column is the result of regression using all indexes. The sixth column presents the result of regression in which the control variable D.UP, that remains insignificant in previous regressions is removed. In column 7, the VCE, cluster method is used to improve the robustness based on column 6. Asterisk ***,**, and *denote the rejection of the null hypothesis at the1, 5, and 10% significance level, respectively. Standard errors are presented in parentheses.

Consumers may take a longer time to compare and select commodities with higher values. It is more appropriate to use the previous month’s search index if the consumers use more than 1 month to gather information and make decisions to buy cars. Moreover, using the previous month’s search index can effectively avoid the possible endogeneity issue. We, therefore, use the lag terms of the four search indexes for regressions, and the results are shown in Table 5. The results are very similar to the previous regressions, indicating that search amount of ‘new energy vehicle’, ‘new energy vehicle battery’, and ‘charging pile’ are significantly positive with the new energy vehicle sales of the next month while the term of ‘automobile spontaneous combustion’ is significantly negative, which is in line with our proposed hypotheses.

Model estimation results using Baidu search index of 1 month ahead.

Variables (1) Index1 (2) Index2 (3) Index3 (4) Index4 (5) All Indexes (6) All Indexes omit D.UP (7) VCE
L.Index1 6.950*** 4.917*** 4.931*** 4.931***
(0.68) (0.77) (0.77) (0.90)
L.D.Index2 35.262*** 20.204*** 20.314*** 20.314**
(7.69) (7.68) (7.68) (8.32)
L.Index3 35.456*** 22.849*** 22.789*** 22.789***
(4.01) (4.77) (4.77) (5.95)
L.Index4 −0.685 −25.956** −26.118** −26.118
(11.25) (11.60) (11.60) (16.65)
GDP 5.707 −31.816** −36.460*** −28.527** −13.169 −11.011 −11.011
(13.69) (13.58) (13.39) (13.66) (13.99) (13.81) (28.76)
D.UP 3.266 3.59 4.047 3.809 3.312
(3.47) (3.54) (3.49) (3.56) (3.44)
D.DI 14.842 89.686* −41.873 80.974* −15.438 −14.137 −14.137
(46.69) (47.25) (48.59) (48.72) (48.63) (48.61) (18.85)
Constant 42.372*** 65.742*** 47.784*** 65.040*** 41.063*** 40.733*** 40.733***
(4.13) (3.56) (4.02) (3.79) (4.37) (4.35) (11.74)
Observations 1950 1950 1950 1950 1950 1950 1950
R-squared 0.057 0.016 0.044 0.005 0.073 0.072 0.072
Number of area 30 30 30 30 30 30 30
province FE YES YES YES YES YES YES YES
Year FE YES YES YES YES YES YES YES
VCE cluster (province) NO NO NO NO NO NO YES

This table presents the estimation results for impacts of four Baidu search indexes on the sale of new energy vehicles sales of China’s market in the next month. For more details of prediction models, refer to Eqs 7, 8. The four columns on the left indicate results of regressions using four Baidu indexes, respectively. The fifth column is the result of regression using all indexes. The sixth column presents the result of regression in which the control variable D.UP, that remains insignificant in previous regressions is removed. In column 7, the VCE, cluster method is used to improve the robustness based on column 6. Asterisk ***,**, and *denote the rejection of the null hypothesis at the 1, 5, and 10% significance level, respectively. Standard errors are presented in parentheses.

 4.3 Analysis of the Different Districts of China

Sales of new energy vehicles are closely related to economic, demographic, and other factors according to the verified results. China has the world’s largest population and the largest market for new energy vehicles. Meanwhile, China has a relatively complicated national condition because the regional development in China is unbalanced in economic development and the situation of the environment varies greatly in different areas. Given this, to explore the impacts of the online search indexes of consumers on new energy vehicles from a more precise perspective, we divide China into four economic regions, i.e. east region, central region, west region, and northeast region, according to the geographical classification standard of the National Bureau of Statistics of China 3 . The sale volumes of new energy vehicles of four economic regions are presented in Figure 2. It can be seen that eastern China accounts for the largest share of new energy vehicle sales, followed by central and western China, and the least by northeast China.

Sales of new energy vehicles in China’s four major economic regions from 2016 to the first half of 2021.

We use existing data to perform the regression analysis for each of the four economic regions to evaluate the influences of Internet search behavior on new energy vehicles sales separately. The results are shown in Tables 6–9. In the analysis of each area, we first perform the regression on all of the Baidu indexes as well as the control variables and then omit the non-significant control variables in the results of regressions using the single Baidu index. Using the same method in Consumer Attention to the Sales of New Energy Vehicles, we also use the Baidu search indexes with 1 month lag in the regressions.

Model estimation results using Baidu search index regarding the east district of China.

Variables (1) All indexes (2) All indexes omit D.DI (3) All indexes with VCE (4) All lag indexes (5) All lag indexes omit D.DI (6) All lag indexes with VCE
Index1 9.810*** 9.715*** 9.715***
(1.56) (1.56) (1.16)
D.Index2 12.880 12.895 12.895**
(14.25) (14.27) (4.56)
Index3 38.254*** 34.316*** 34.316**
(11.70) (11.37) (11.53)
Index4 −16.940 −18.486 −18.486
(24.67) (24.67) (14.16)
L.Index1 6.368*** 6.375*** 6.375***
(1.63) (1.62) (1.03)
L.D.Index2 4.499 4.382 4.382
(14.95) (14.87) (8.29)
L.Index3 30.366** 30.477** 30.477**
(11.96) (11.88) (12.17)
L.Index4 6.431 6.751 6.751
(25.87) (25.59) (22.80)
GDP 31.862 35.095 35.095 22.862 22.768 22.768
(31.89) (31.84) (56.86) (33.42) (33.38) (51.44)
D.UP 9.880 11.112 11.112 5.621 5.490 5.490
(11.96) (11.94) (16.07) (12.66) (12.56) (13.88)
D.DI -158.311 9.857
(112.44) (113.73)
Constant 23.169* 22.357* 22.357 37.656*** 37.736*** 37.736
(13.57) (13.57) (28.60) (13.92) (13.87) (27.97)
Observations 585 585 585 585 585 585
R-squared 0.154 0.151 0.151 0.085 0.085 0.085
Number of provinces 9 9 9 9 9 9
province FE YES YES YES YES YES YES
Year FE YES YES YES YES YES YES
VCE cluster (province) NO NO YES NO NO YES

This table presents the estimation results for impacts of four Baidu search indexes on the sale of new energy vehicles sales of eastern China’s market. For more details of prediction models, refer to Eq. 8. The left three columns indicate results of the impacts of consumer attention on the sale volume in the same month and the right three columns indicate the impacts on the sale volume of the next month. We first perform regression using all Baidu indexes and then omit the persistently insignificant control variables if any and finally use the VCE, cluster method. Asterisk ***,**, and *denote the rejection of the null hypothesis at the 1, 5, and 10% significance level, respectively. Standard errors are presented in parentheses.

Model estimation results using Baidu search index regarding the central district of China.

Variables (1) All indexes (2) All indexes VCE (3) All lag indexes (4) All lag indexes omit D.DI (5) All lag indexes with VCE
Index1 8.325*** 8.325***
(1.31) (0.81)
D.Index2 −14.165 −14.165
(18.73) (12.86)
Index3 36.593*** 36.593***
(9.91) (5.79)
Index4 −85.775*** −85.775***
(20.37) (11.91)
L.Index1 2.398* 2.373* 2.373***
(1.38) (1.37) (0.33)
L.D.Index2 63.568*** 64.669*** 64.669***
(19.87) (19.76) (10.70)
L.Index3 33.319*** 32.656*** 32.656***
(10.25) (10.17) (4.70)
L.Index4 −75.323*** −77.972*** −77.972**
(21.99) (21.48) (28.65)
GDP −106.897*** −106.897*** −115.136*** −113.229*** −113.229***
(31.43) (19.31) (33.34) (33.14) (12.49)
D.UP −8.273 −8.273** 0.441 0.071 0.071
(5.43) (2.56) (5.74) (5.70) (1.41)
D.DI −185.627** −185.627* −53.420
(89.17) (87.91) (93.30)
Constant 65.173*** 65.173*** 81.203*** 80.786*** 80.786***
(9.62) (4.77) (9.98) (9.94) (3.77)
Observations 390 390 390 390 390
R-squared 0.230 0.230 0.139 0.139 0.139
Number of provinces 6 6 6 6 6
province FE YES YES YES YES YES
Year FE YES YES YES YES YES
VCE cluster (province) NO YES NO NO YES

This table presents the estimation results for impacts of four Baidu search indexes on the sale of new energy vehicles sales of central China’s market. For more details of prediction models, refer to Eq. 8. The left two columns indicate results of the impacts of consumer attention on the sale volume in the same month and the right three columns indicate the impacts on the sale volume of the next month. We first perform regression using all Baidu indexes and then omit the persistently insignificant control variables if any and finally use the VCE, cluster method. Asterisk ***,**, and *denote the rejection of the null hypothesis at the 1, 5, and 10% significance level, respectively. Standard errors are presented in parentheses.

Model estimation results using Baidu search index regarding the west district of China.

Variables (1) All indexes (2) All indexes with VCE (3) All lag indexes (4) All lag indexes omit D.DI (5) All lag indexes with VCE
Index1 5.669*** 5.669***
(1.15) (0.48)
D.Index2 −3.072 −3.072
(8.73) (4.64)
Index3 28.089*** 28.089***
(5.30) (8.00)
Index4 −86.145*** −86.145***
(13.51) (15.14)
L.Index1 3.426*** 3.480*** 3.480***
(1.19) (1.19) (0.39)
L.D.Index2 22.763** 23.445*** 23.445***
(9.05) (9.02) (6.48)
L.Index3 16.183*** 15.028*** 15.028***
(5.40) (5.28) (4.43)
L.Index4 −58.742*** −59.658*** −59.658***
(14.00) (13.96) (11.41)
GDP −38.367* −38.367*** −50.445** −49.006** −49.006***
(20.51) (11.12) (20.98) (20.93) (8.20)
D.UP -1.294 -1.294 -9.190 -7.461 -7.461
(12.53) (14.77) (13.12) (13.00) (12.01)
D.DI −159.402*** −159.402** −55.241
(55.15) (63.75) (55.46)
Constant 27.680*** 27.680*** 34.582*** 33.905*** 33.905***
(3.81) (2.97) (3.87) (3.81) (1.78)
Observations 780 780 780 780 780
R-squared 0.141 0.141 0.080 0.079 0.079
Number of provinces 12 12 12 12 12
province FE YES YES YES YES YES
Year FE YES YES YES YES YES
VCE cluster (province) NO YES NO NO YES

This table presents the estimation results for impacts of four Baidu search indexes on the sale of new energy vehicles sales of western China’s market. For more details of prediction models, refer to Eq. 8. The left two columns indicate results of the impacts of consumer attention on the sale volume in the same month and the right three columns indicate the impacts on the sale volume of the next month. We first perform regression using all Baidu indexes and then omit the persistently insignificant control variables if any and finally use the VCE, cluster method. Asterisk ***,**, and *denote the rejection of the null hypothesis at the 1, 5, and 10% significance level, respectively. Standard errors are presented in parentheses.

Model estimation results using Baidu search index regarding the northeast area of China.

Variables (1) All indexes (2) All indexes omit GDP and D.UP (3) All indexes with VCE (4) All lag indexes (5) All lag indexes omit GDP and D.UP (6) All lag indexes with VCE
Index1 6.266*** 6.352*** 6.352**
(1.63) (1.62) (1.03)
D.Index2 48.716*** 48.043*** 48.043*
(14.99) (14.85) (11.91)
Index3 −15.113** −14.135* −14.135
(7.60) (7.49) (6.63)
Index4 33.945* 35.228* 35.228
(18.97) (18.69) (19.19)
L.Index1 1.863 1.853 1.853*
(1.73) (1.72) (0.46)
L.D.Index2 25.075 24.451 24.451
(16.02) (15.95) (10.97)
L.Index3 −10.163 −8.699 −8.699***
(7.88) (7.76) (0.84)
L.Index4 15.139 17.528 17.528
(19.98) (19.62) (14.29)
GDP 28.746 48.340
(56.86) (60.75)
D.UP 7.594 8.798
(13.34) (14.25)
D.DI 121.423 123.143* 123.143** 186.762** 188.090** 188.090**
(73.72) (72.73) (22.29) (75.43) (74.37) (23.74)
Constant 24.408*** 27.698*** 27.698*** 26.665*** 32.553*** 32.553***
(8.59) (3.17) (2.33) (9.32) (3.38) (1.61)
Observations 195 195 195 195 195 195
R-squared 0.165 0.161 0.161 0.063 0.055 0.055
Number of provinces 3 3 3 3 3 3
province FE YES YES YES YES YES YES
Year FE YES YES YES YES YES YES
VCE cluster (province) NO NO YES NO NO YES

This table presents the estimation results for impacts of four Baidu search indexes on the sale of new energy vehicles sales of the northeast region of China. For more details of prediction models, refer to Eq. 8. The left three columns indicate results of the impacts of consumer attention on the sale volume in the same month and the right three columns indicate the impacts on the sale volume of the next month. We first perform regression using all Baidu indexes and then omit the persistently insignificant control variables if any and finally use the VCE, cluster method. Asterisk ***,**, and *denote the rejection of the null hypothesis at the 1, 5, and 10% significance level, respectively. Standard errors are presented in parentheses.

With the development of modern industry in China, eastern China has become the most economically developed region and the infrastructure construction in this area is also of the highest level in China. The three largest urban agglomerations in China, the Beijing-Tianjin-Hebei Region, the Yangtze River Delta, and the Pearl River Delta, are all located in eastern China. According to China’s seventh national census, the population in eastern China accounted for 39.93 percent. As illustrated in Table 6, the search trend of ‘new energy vehicle’, and ‘new energy vehicle battery’ both have significant positive impacts on the sales of new energy vehicles while ‘charging pile’, and ‘automobile spontaneous combustion’ are not significant.

The central region of China is consists of six provinces. Because it is located inland, the development level of the central region is not balanced, and the level of economic openness is not as good as that of the eastern region. At present, the economic development level of the central region is lower than that of the eastern economic zone while higher than that of the western economic zone, but the growth rate of economic development is lower than that of west China. As shown in Table 7, the search trends of ‘charging pile’, ‘new energy vehicle battery’, and ‘automobile spontaneous combustion’ all have significant impacts on the sales of new energy vehicles of the next month. It should be noted that the attention of ‘automobile spontaneous combustion’ is negatively correlated with the sales indicating consumers in the central region of China is more sensitive to the new energy vehicles accidents.

Western China generally lags behind central and eastern China. It accounts for 70 percent of the country’s land area and about 29 percent of the total population. Due to the terrain and climate conditions in western China, although the area is vast, the population density is relatively sparse, which magnifies the mileage problem of new energy vehicles. At the same time, the energy resources in western China are abundant and the electricity cost is low, which provides favorable conditions for the application of new energy vehicles. As shown in Table 8, the search trend of ‘new energy automobile’, ‘new energy vehicle battery’, and ‘automobile spontaneous combustion’ all have significant impacts on the sales of new energy vehicles of the next month. The attention of ‘automobile spontaneous combustion’ is also negatively correlated with the sales, which is similar to the facts of the central region of China.

The Northeast region of China is an old industrial area in China, but its economy develops slowly in recent years and is in a period of economic transition. According to the result presented in Table 9, we found that most of the variables were not significant. We believed that this is because the overall sample was too small and overall sales volume was only in the hundreds. The climate in Northeast China was not suitable for the use of new energy vehicles could be one of the possible reasons.

 5 Discussion and Conclusion

Exploring the impacts of consumer attention to different aspects of new energy vehicles on the sale volume can provide effective information for the promotion of new energy vehicles. Accordingly, we use four representative Baidu search indexes, ‘new energy vehicle’, ‘new energy vehicles battery’, ‘charging pile’, and ‘automobile spontaneous combustion’, as variables representing the attention of consumers and adopt variables of economic, population, and residents income as control variables for regression analysis. We first analyzed the data of the whole China’s new energy vehicle market and found that search amounts of ‘new energy automobile’, ‘new energy vehicle battery’, and ‘charging pile’ have significant positive impacts on the new energy vehicle sales while the term of ‘automobile spontaneous combustion’ has a significant negative impact. To address the problem of endogeneity and consider the likelihood that consumers may make longer purchase decisions, we also used the lagging term of Baidu indexes for regressions and the results are in line with the conclusions from models using same month indexes.

Given the unbalanced regional development in China, we perform the regressions using the data of four major economic regions with different development levels respectively. In our subregional study, we found some interesting results. First, the search index of ‘charging pile’ has no significant impact on the sales of new energy vehicles in the economically developed region, specifically eastern China while in less developed areas this relationship exists, that is, more search for ‘charging pile’ mean more sales of new energy vehicles. One possible explanation is that better infrastructure and the availability of new energy charging devices in economically developed regions have made consumers less anxious about charging piles while in the less developed area, the consumers still have such concerns. Second, the search trend of ‘new energy vehicle battery’ has significant positive impacts on the sale volumes in all regions of China indicating that the efficiency of new energy vehicle batteries is still an important concern of consumers.

Another interesting finding from our study is people’s cognitive bias of excessive concerns about spontaneous combustion of new energy vehicles. Cognitive bias is an error in thinking that occurs when people are processing and interpreting information in the world around them and affects the decisions and judgments that they make (Bedi and Toshniwal, 2018). The empirical study shows that the search index of ‘auto spontaneous combustion’ has a significant negative effect on new energy vehicles sales in less developed regions, i.e. central and western China compared to eastern China indicating that the information on automobile spontaneous combustion will significantly increase people’s negative safety evaluation of new energy vehicles in certain areas. This concern about the spontaneous combustion of new energy vehicles may be exaggerated by the related news because the news of the spontaneous combustion of traditional fuel vehicles is not as attractive as that of new energy vehicles, although the spontaneous combustion rate of traditional fuel vehicles is higher than that of new energy vehicles. In fact, according to the “Travel Big Data Report of Small Pure Electric Passenger Vehicles” released by “National Big Data Alliance for New Energy Vehicles” in2020 4 , the spontaneous combustion probability of domestic new energy vehicles in 2019 was 0.0049% and in 2020, the probability dropped to 0.0026%. On the other hand, according to data released by the Departments of the Public Security of China, the annual spontaneous combustion rate of traditional fuel vehicles in 2020 was about 0.01–0.02%, significantly greater than that of new energy vehicles. The different responses to spontaneous combustion of new energy vehicles between central as well as western regions, and eastern regions also reflect that this cognitive bias may be related to the level of economic development. People in economically developed regions are less likely to produce such cognitive bias.

This study provides some useful implications for the promotion of new energy vehicles. First, battery performance and charging devices are very important factors when consumers decide whether to buy new energy vehicles. Accordingly, how to improve the range and charging convenience should be the first thing that automobile manufacturers need to solve. Second, the safety of new energy vehicles should be more publicized, so that the cognitive bias that new energy vehicles are more prone to spontaneous combustion can be appropriately alleviated.

This study has several limitations that bode well for future research opportunities. First, as Baidu only provides search index information for specific keywords, some search terms related to new energy vehicles that do not appear in the search database may also provide interesting findings. Therefore, we can further study the impact of consumer attention on new energy vehicles through other proxies. Second, since relevant Baidu search terms can have an impact on the sales of new energy vehicles, we can try to predict the sales by integrating these indexes to provide effective information for new energy vehicle manufacturers and the government. Last, the potential relationship between economic development level and cognitive bias needs to be further demonstrated through social empirical research.

 Data Availability Statement

The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.

 Author Contributions

ZJ carried out methodology, coded, and wrote the original manuscript YL collected the data did the review and edition LZ acquired the fund and did the thesis modification.

 Funding

Research reported in this publication was supported by funding from the National Natural Science Foundation of China (No. 72071194).

 Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

 Publisher’s Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

https://index.baidu.com/

https://gs.statcounter.com/search-engine-market-share/all/china

The information is found in http://www.stats.gov.cn/ztjc/zthd/sjtjr/dejtjkfr/tjkp/201106/t20110613_71947.htm.

The official website of “National Big Data Alliance for New Energy Vehicles” is http://www.ndanev.com/.

 References Akalamkam K. Mitra J. K. (2018). Consumer Pre-purchase Search in Online Shopping: Role of Offline and Online Information Sources. Business Perspect. Res. 6 (1), 42–60. 10.1177/2278533717730448 Bedi J. Toshniwal D. (2018). Empirical Mode Decomposition Based Deep Learning for Electricity Demand Forecasting. IEEE Access 6, 49144–49156. 10.1109/access.2018.2867681 Berkeley N. Jarvis D. Jones A. (2018). Analysing the Take up of Battery Electric Vehicles: An Investigation of Barriers Amongst Drivers in the UK. Transportation Res. D: Transport Environ. 63, 466–481. 10.1016/j.trd.2018.06.016 Egbue O. Long S. (2012). Barriers to Widespread Adoption of Electric Vehicles: An Analysis of Consumer Attitudes and Perceptions. Energy Policy 48, 717–729. 10.1016/j.enpol.2012.06.009 Egnér F. Trosvik L. (2018). Electric Vehicle Adoption in Sweden and the Impact of Local Policy Instruments. Energy Policy 121, 584–596. 10.1016/j.enpol.2018.06.040 Fritzsch B. Wenger K. Sibbertsen P. Ullmann G. (2020). Can Google Trends Improve Sales Forecasts on a Product Level. Appl. Econ. Lett. 27 (17), 1409–1414. 10.1080/13504851.2019.1686110 Funke S. Á. Sprei F. Gnann T. Plötz P. (2019). How Much Charging Infrastructure Do Electric Vehicles Need? A Review of the Evidence and International Comparison. Transportation Res. Part D: Transport Environ. 77, 224–242. 10.1016/j.trd.2019.10.024 Geva T. Oestreicher-Singer G. Efron N. Shimshoni Y. (2015). Using Forum and Search Data for Sales Prediction of High-Involvement Products. Tomer Geva, Gal Oestreicher-Singer, Niv Efron, Yair Shimshoni. MIS Q. 41, 65–82. 10.25300/MISQ/2017/41.1.04 Guo J. Zhang X. Gu F. Zhang H. Fan Y. (2020). Does Air Pollution Stimulate Electric Vehicle Sales? Empirical Evidence from Twenty Major Cities in China. J. Clean. Prod. 249, 119372. 10.1016/j.jclepro.2019.119372 Hao Y. Dong X. Y. Deng Y. X. Li L. X. Ma Y. (2016). What Influences Personal Purchases of New Energy Vehicles in China? an Empirical Study Based on a Survey of Chinese Citizens. J. Renew. Sust. Energ. 8 (6), 065904. 10.1063/1.4966908 Hausman J. A. (1978). Specification Tests in Econometrics. Econometrica 46, 1251–1271. 10.2307/1913827 Higueras-Castillo E. Molinillo S. Coca-Stefaniak J. A. Liebana-Cabanillas F. (2020). Potential Early Adopters of Hybrid and Electric Vehicles in Spain-Towards a Customer Profile. Sustainability 12 (11), 4345. 10.3390/su12114345 Hill G. Heidrich O. Creutzig F. Blythe P. (2019). The Role of Electric Vehicles in Near-Term Mitigation Pathways and Achieving the UK's Carbon Budget. Appl. Energ. 251, 113111. 10.1016/j.apenergy.2019.04.107 Im K. S. Pesaran M. H. Shin Y. (2003). Testing for Unit Roots in Heterogeneous Panels. J. Econom. 115 (1), 53–74. 10.1016/s0304-4076(03)00092-7 Javid R. J. Nejat A. (2017). A Comprehensive Model of Regional Electric Vehicle Adoption and Penetration. Transport Policy 54, 30–42. 10.1016/j.tranpol.2016.11.003 Jenn A. Azevedo I. L. Ferreira P. (2013). The Impact of Federal Incentives on the Adoption of Hybrid Electric Vehicles in the United States. Energ. Econ. 40, 936–942. 10.1016/j.eneco.2013.07.025 Kester J. Noel L. Zarazua de Rubens G. Sovacool B. K. (2018). Policy Mechanisms to Accelerate Electric Vehicle Adoption: a Qualitative Review from the Nordic Region. Renew. Sust. Energ. Rev. 94, 719–731. 10.1016/j.rser.2018.05.067 Kim Y. Jun C. H. (2010). A New Approach to Unit Root Testing of Panel Data. Seoul, South Korea: Korea Inst. Ind. Eng., 1462–1469. Langbroek J. H. M. Franklin J. P. Susilo Y. O. (2016). The Effect of Policy Incentives on Electric Vehicle Adoption. Energy Policy 94, 94–103. 10.1016/j.enpol.2016.03.050 Levin A. Lin C.-F. James Chu C.-S. (2002). Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties. J. Econom. 108 (1), 1–24. 10.1016/s0304-4076(01)00098-7 Liao F. Molin E. Timmermans H. van Wee B. (2019). Consumer Preferences for Business Models in Electric Vehicle Adoption. Transport Policy 73, 12–24. 10.1016/j.tranpol.2018.10.006 Ma S.-C. Fan Y. Feng L. (2017). An Evaluation of Government Incentives for New Energy Vehicles in China Focusing on Vehicle Purchasing Restrictions. Energy Policy 110, 609–618. 10.1016/j.enpol.2017.07.057 Ma S.-C. Fan Y. Guo J.-F. Xu J.-H. Zhu J. (2019). Analysing Online Behaviour to Determine Chinese Consumers' Preferences for Electric Vehicles. J. Clean. Prod. 229, 244–255. 10.1016/j.jclepro.2019.04.374 Maheshwari P. Seth N. Gupta A. K. (2016). An Empirical Approach to Consumer Buying Behavior in Indian Automobile Sector. Ind. Commercial Train. 48 (3), 156–162. 10.1108/ict-09-2015-0061 Mersky A. C. Sprei F. Samaras C. Qian Z. (2016). Effectiveness of Incentives on Electric Vehicle Adoption in Norway. Transportation Res. Part D: Transport Environ. 46, 56–68. 10.1016/j.trd.2016.03.011 Mitchell V. W. Prince G. (1993). Retailing to Experienced and Inexperienced Consumers: A Perceived Risk Approach. West Yorkshire, UK: International Journal of Retail & Distribution Management. Mock P. Yang Z. (2014). Driving Electrification: A Global Comparison of Fiscal Incentive Policy for Electric vehiclesICCT, the Internationa council on Clean Transportation. Needell Z. A. McNerney J. Chang M. T. Trancik J. E. (2016). Potential for Widespread Electrification of Personal Vehicle Travel in the United States. Nat. Energ. 1 (9), 16112. 10.1038/nenergy.2016.112 Nie Y. Ghamami M. Zockaie A. Xiao F. (2016). Optimization of Incentive Polices for Plug-In Electric Vehicles. Transportation Res. B: Methodological 84, 103–123. 10.1016/j.trb.2015.12.011 Noel L. Zarazua de Rubens G. Sovacool B. K. Kester J. (2019). Fear and Loathing of Electric Vehicles: The Reactionary Rhetoric of Range Anxiety. Energ. Res. Soc. Sci. 48, 96–107. 10.1016/j.erss.2018.10.001 Palmer K. Tate J. E. Wadud Z. Nellthorp J. (2018). Total Cost of Ownership and Market Share for Hybrid and Electric Vehicles in the UK, US and Japan. Appl. Energ. 209, 108–119. 10.1016/j.apenergy.2017.10.089 Priessner A. Sposato R. Hampl N. (2018). Predictors of Electric Vehicle Adoption: An Analysis of Potential Electric Vehicle Drivers in Austria. Energy Policy 122, 701–714. 10.1016/j.enpol.2018.07.058 Ruohonen J. Hyrynsalmi S. (2017). Evaluating the Use of Internet Search Volumes for Time Series Modeling of Sales in the Video Game Industry. Electron. Markets 27 (4), 351–370. 10.1007/s12525-016-0244-z Soltani-Sobh A. Heaslip K. Stevanovic A. Bosworth R. Radivojevic D. (2017). Analysis of the Electric Vehicles Adoption over the United States. Transportation Res. Proced. 22, 203–212. 10.1016/j.trpro.2017.03.027 Sovacool B. K. Abrahamse W. Zhang L. Ren J. (2019). Pleasure or Profit? Surveying the Purchasing Intentions of Potential Electric Vehicle Adopters in China. Transportation Res. A: Pol. Pract. 124, 69–81. 10.1016/j.tra.2019.03.002 Spangher L. Gorman W. Bauer G. Xu Y. Atkinson C. (2019). Quantifying the Impact of U.S. Electric Vehicle Sales on Light-Duty Vehicle Fleet CO2 Emissions Using a Novel Agent-Based Simulation. Transportation Res. Part D: Transport Environ. 72, 358–377. 10.1016/j.trd.2019.05.004 Sun H. Geng Y. Hu L. Shi L. Xu T. (2018). Measuring China's New Energy Vehicle Patents: A Social Network Analysis Approach. Energy 153, 685–693. 10.1016/j.energy.2018.04.077 Thomas C. E. (2009). Fuel Cell and Battery Electric Vehicles Compared. Int. J. Hydrogen Energ. 34 (15), 6005–6020. 10.1016/j.ijhydene.2009.06.003 Wachter P. Widmer T. Klein A. (2019). “Predicting Automotive Sales Using Pre-purchase Online Search Data,” in 2019 Federated Conference on Computer Science and Information Systems (FedCSIS), Leipzig, Germany, 1-4 Sept. 2019 (IEEE). 10.15439/2019F239 Wang L. Fu Z. L. Guo W. Liang R. Y. Shao H. Y. (2020). What Influences Sales Market of New Energy Vehicles in China? Empirical Study Based on Survey of Consumers' purchase Reasons. Energy Policy 142, 111484. 10.1016/j.enpol.2020.111484 Wang N. Tang L. Pan H. (2019). A Global Comparison and Assessment of Incentive Policy on Electric Vehicle Promotion. Sust. Cities Soc. 44, 597–603. 10.1016/j.scs.2018.10.024 Wu Y. A. Ng A. W. Yu Z. Huang J. Meng K. Dong Z. Y. (2021). A Review of Evolutionary Policy Incentives for Sustainable Development of Electric Vehicles in China: Strategic Implications. Energy Policy 148, 111983. 10.1016/j.enpol.2020.111983 Yang Z. Zhang C. (2020). “Automobile Sales Forecast Based on Web Search and Social Network Data,” in 2020 The 11th International Conference on E-business, Management and Economics, Beijing, China, July 15–17, 2020 (Beijing, China: Association for Computing Machinery), 37–41. 10.1145/3414752.3414759 Zarazua de Rubens G. (2019). Who Will Buy Electric Vehicles after Early Adopters? Using Machine Learning to Identify the Electric Vehicle Mainstream Market. Energy 172, 243–254. 10.1016/j.energy.2019.01.114 Zhang Y. Zhong M. Geng N. Jiang Y. (2017). Forecasting Electric Vehicles Sales with Univariate and Multivariate Time Series Models: The Case of China. PloS one 12 (5), e0176729. 10.1371/journal.pone.0176729