Abstract
This paper demonstrates the use of an agent-based model (ABM) to study the mechanism of social influence in the diffusion of new energy vehicles. We introduce the “consumat” cognition model so that agents with different need satisfaction thresholds have different cognitive processes. In addition, supported by survey data, our study considers more characteristics of opinion leaders, such as innovative behavior, lower sensitivity to price influence, and a better ability to judge the product quality. Through the primary group and control group experiments, the simulations demonstrated that the opinion leaders play a significant role in the spread of information and the percentage of product adoption. The results indicate that targeting opinion leaders will be a valuable marketing strategy for new energy vehicles. It also provides some advice for assessing policies that promote sustainable behaviors.
1. Introduction
Wired Travel has asked many Tesla vehicle owners, “Why did you buy Tesla?” there is always a similar answer: “because Musk is a person who is building cars, still building rockets and going to Mars. Such an artificially manufactured car must also be cool.” According to Tencent Technology, Musk has posted 74 microblog searches since 3rd December, 2020. Musk said on his social media, Twitter, that “he is considering quitting his job and becoming an online celebrity full-time.” It attracted the attention of netizens worldwide for a while, and almost 20,000 responses were received within an hour of its release, showing its significant social influence.
In today’s booming new energy vehicle market, reexamining the role of stars in the technology diffusion process will significantly help the popularization of new energy vehicles in China. The Chinese vehicle brands need to learn not only from Tesla’s strong product competitiveness but also from its unique marketing ability. However, more research studies need to be conducted on the impact of key opinion leaders on new energy vehicles. Therefore, this paper aims to discuss the impact of key opinion leaders based on agent-based simulation modeling (ABM). Based on the theory of complex adaptive systems and innovation diffusion, we have established a simulation model of the diffusion of new energy vehicles. Through the simulation of the consumer decision-making process, interaction process, and the market, the real society can be restored to a maximum reasonable extent, and the mechanism of the diffusion process of new energy vehicles and key influencing factors can be explored. In addition, we distinguish opinion leaders from general consumers according to their characteristics, so that we can observe the experimental results more easily. At the same time, in order to make the simulation results closer to the real situation of Chinese consumers, we collected empirical data from the Chinese automobile market through a questionnaire survey. Finally, according to the experimental results, we put forward some practical marketing suggestions for the promotion of new energy vehicles in China.
In the following sections, we first discuss cognition and parameter settings; subsequently, we describe the simulation model and experimental results. Finally, we draw conclusions and have some discussions.
2. Literature Review
Agent-based models (ABMs) have often been used to study the diffusion of new products and technologies because of their flexible architecture and ability to operate in various environments. Among them are many studies on the simulation research of new energy vehicles [1–5]. However, most focus on the macrolevel, such as charging infrastructure, financial subsidies, market share, and other factors, and pay less attention to the microlevel of social interaction, such as opinion leaders. Therefore, modeling new energy vehicles is still a new area of research.
2.1. Social Influence in ABM
Scholars have long found the importance of social influence in innovation communication and applied it to the traditional equation model of innovation diffusion (Bass) [6]. However, the limitation of the Bass model is that it does not introduce social interaction factors. Because the agent modeling method can accurately simulate interpersonal interaction at the microlevel, it is more prevalent among scholars when analyzing social influencing factors. Janssen and Jager analyzed social influence by defining two kinds of utility functions (social utility and personal utility) of consumer agents. The utility value of consumers is divided into two parts which are affected by norms and information, respectively. Information influence refers to the satisfaction of the learned commodity-related information to the consumer’s utility brought about by the consumer’s physiological needs and personal preferences. The normative influence affects the satisfaction of consumer products to social needs such as consumers’ social status and sense of belonging through social norms and group atmosphere. Consumers can decide to accept goods based on the quality of the goods. However, the social influence caused by neighboring consumers will also bring social pressure on them, thereby strengthening or changing their initial purchase intention [7]. Kiesling et al. differentiate between microlevel, mesolevel, and macrolevel social influence. Microlevel social influence refers to the impact between two subjects through pairwise communication links, such as word of mouth (WOM). Mesolevel social influence refers to the influence of the surrounding environment of the subject, such as neighbor influence and group influence (herd behavior, conspicuous consumption, etc.). Macrolevel influence refers to the influence of the social level on the subject, such as fashion and trends [8].
In many of the reviewed articles, the term“social influence”is used in the sense of mesolevel social influence. The research of relevant scholars shows that in agent-based simulation modeling, a suitable network structure significantly impacts the path and speed of diffusion of innovation [9]. The two most common types in the agent model are small-world networks and scale-free networks [10, 11]. Barabasi and Bonabeau both believe that a scale-free network is the closest to a social, interpersonal network in the real world [12, 13]. The cognitive model of the subject is also an essential factor in determining the success of ABM. A large amount of literature builds the cognitive model based on the social psychology theory. Ajzen’s theory of planned behavior (TPB) is one of the most widely used theories. Subject decision-making is affected by attitude, subjective norms, and perceived behavior control [14]. Jager et al. put forward another theory of consumat [15–17]. Consumat (consumer agent) makes purchase decisions according to satisfaction and perceived uncertainty. These four decisions are deliberating, social comparison, repeating, and imitation.
2.2. Opinion Leaders in ABM
There is no strict definition of opinion leaders in the academic field. Engel et al. believed that opinion leaders should be innovators or first users who can influence other consumers through innovative behavior and specific product knowledge [18]. Feick and Price proposed that market experts who have no specific professional knowledge but are more familiar with the market are more likely to become opinion leaders [19]. Rogers believes that opinion leaders are not sensitive to old social norms or are more willing to accept new social norms, so they may be more willing to accept innovative or rebellious products [20]. Lyons and Henderson put forward a more comprehensive definition of opinion leaders. Compared to consumers who solicit their opinions, opinion leaders often have more experience or expertise, they have disclosed or obtained more information about the products, they show more exploratory and innovative behaviors, and thus show a higher product participation. This definition distinguishes opinion leaders and ordinary consumers from the perspective of individual heterogeneity and provides great convenience for the setting, identification, and use of opinion leaders in ABM [21]. Nielsen believes that opinion leaders are often experts in the industry from the perspective of marketing, and their professional evaluation is a unique marketing method which is transmitted to mass consumers through particular ways (such as WOM) [22].
Valente and Davis first studied the role of opinion leaders in the agent model. They defined homogeneous consumer subjects. If 15% of the neighbors adopt the innovative product, they will also make consumption decisions. Moreover, if the network model that randomly assigns a relationship to each agent is applied, then there will be no very closed social network structure similar to the real world [23]. The simulation results show that when opinion leaders initiate the diffusion, it will diffuse faster and that purposeful opinion leaders can speed up the diffusion process. Van Eck et al. defined heterogeneous consumer subjects, considering the pursuit of quality by opinion leaders and the impact of knowledge level on consumers [24]. Like Janssen and Jager, the subject’s decision to accept the product is based on a utility threshold function that includes personal preferences and social influence. However, they did not simulate social influence as a fixed value but as a constant value. For example, if more neighbors accept the commodity, the normative influence of the commodity will increase. The scale-free network is used for simulation. Finally, it is concluded that opinion leaders increase the transmission speed of information, the acceptance process, and the maximum number of recipients ratio.
3. The Simulation Experiment
3.1. Agent and Cognitive Model
We use NetLogo software to develop an agent-based model of the diffusion of new energy vehicles. To simplify the model, we divide consumer agents into two categories, ordinary consumers () and opinion leaders (). The heterogeneity of consumers is shown by using the different attribute values between consumer agents, which distinguishes opinion leaders from ordinary consumers. Specifically, agents identified as opinion leaders will have the ability to know the actual quality of goods through the mass media (better quality judgment), the higher weight of personal utility in the total utility (not easily affected by norms), and the tremendous enthusiasm to try out new things (innovative and price-insensitive).
The cognitive model proposed by Jager et al. is a good description of social influence [15], but subsequent studies have been applied to analyzing small-world networks. We further introduce the“consumat”into the scale-free network and modify the original practical function of the consumer agent. Each consumer agent will receive a commodity satisfaction threshold and an uncertainty threshold . The commodity satisfaction threshold is the lowest utility value the utility brought about when choosing a commodity that can satisfy consumers. When the utility value obtained by the consumers from the selected commodity is lower than the satisfaction threshold, then the consumer agent is in a state of dissatisfaction or low satisfaction; otherwise, it indicates that consumers are satisfied with the selected commodity. When the uncertainty is higher than the uncertainty threshold, it indicates that the consumers are skeptical about the goods they currently buy and will refer to the use of others to decide what goods they will buy next. Therefore, the model in this paper sets the cognitive process of consumers as follows (see Figure 1):(1)Repeat: when and , consumat will continue to purchase the products purchased in the last time step.(2)Deliberation: when and , consumat will estimate the utility value of each product and will finally choose the product with the highest score. With the same utility value, the choice will be at random.(3)Imitation: when and , consumat will evaluate the product that is consumed most in its social network. That which is consumed the most in the previous time step will be chosen for current consumption. With the same holding rate, the choice will be made at random.(4)Social comparison: when and , the consumer agent calculates the model with the highest purchases in its social network. The candidate product with the highest market share will be chosen for consumption.
3.2. Parameters and Assumptions
We collected many consumer data from China’s auto market through questionnaires and interviews. Finally, we selected price, fuel consumption, and subsidy attributes to build the initial agent assignment. Two indicators measured the promotion of new energy vehicles: the diffusion of information related to new energy vehicles and the adoption of new energy vehicles. That is, the diffusion of new energy vehicles is observed by the awareness rate of consumers about new energy vehicles and the purchase rate of new energy vehicles. This model divides the process into three stages for the diffusion of new energy vehicles in the network: mass communication, word of mouth, and strategic decision. These three stages constitute the diffusion mechanism of new energy vehicles in this model, which is continually repeated during the operation of the model. The microlevel decision-making and consumption behavior of consumer agents continue to evolve through this, and finally, the actual trend of the automobile market emerges at the macrolevel.
Combined with the abovementioned collected data, the initial data of the three attributes of the two types of automotive agents are shown in Table 1.
After referring to the previous literature and by comprehensively considering the effect of NetLogo operation and efficiency, this paper sets the initial number of agents in the model to people. This value can be adjusted by using the slider, which can be set according to the needs of the subsequent experiments. Among them, are fuel vehicle consumers, while the remaining are potential vehicle consumers who intend to buy a vehicle but have not yet bought it. According to the theory described previously, we define consumers with the highest score on the questionnaire as opinion leaders. Among these people, about of the agents will be set as opinion leaders, and their characteristic values are only partially consistent with those of ordinary consumers. The settings of the various attributes of the consumer subjects are shown in Table 2.
In this model, the opinion leaders’ innovation will be reflected in the following aspects: opinion leaders have higher fuel consumption and price thresholds for buying cars and can afford higher prices than ordinary consumers. The fuel consumption and price thresholds of each opinion leader are evenly distributed between and , while the fuel consumption and price thresholds of ordinary consumers are evenly distributed between and . The gaps can not only reflect the differences between opinion leaders and ordinary consumers but also control the differences to avoid the situation that opinion leaders are too innovative. If an opinion leader is too innovative, he will likely accept those innovative products that prove unsuccessful and then lose his position in the field [16].
For ordinary consumers, we have three assumptions. First, each consumer has different characteristic values. Second, the threshold of ordinary consumers for fuel consumption and the price is lower than that of opinion leaders. Third, ordinary consumers cannot accurately judge the quality of goods based on information.
3.3. Experimental Manipulation
To accurately judge the differences in the results of the simulation under different initial conditions, first, we complete the experiment of the experimental group, that is, simulate the diffusion of new energy vehicles under the initial conditions given above, to provide a basis for the subsequent treatment of the results of the control group experiment. Next, we compare different simulation results with the data obtained from control group experiments to judge the role of opinion leaders in the diffusion of new energy vehicles and what factors will affect the role of opinion leaders in the group. The setting of the opinion leaders and general consumers attributes are shown in Table 3.
4. Results and Analysis
4.1. Number of Opinion Leaders
We conduct two control group experiments to test whether opinion leaders play a role in the diffusion of new energy vehicles. In , the number of opinion leaders in the network is only half of that in normal circumstances. At the same time, in , there is no distinction between opinion leaders and ordinary consumers, and all consumer subjects have the same attributes and characteristics.
The experiment is the primary experiment; on an average, opinion leaders account for of the total leaders. The results are reflected in the diamond-shaped dot line. Experiment has an average of % of opinion leaders; the experimental data are shown in a straight line of the figure. removes opinion leaders from the consumption circle; there are only ordinary consumers in the model. The data in this model are represented in the figure by triangular segments. It can be seen that the number of opinion leaders has a significant impact on the diffusion of information and products for new energy vehicles. When opinion leaders do not reach the average proportion in the crowd, the diffusion of vehicle-related information is significantly constrained, which can only reach coverage. Its diffusion speed has also been significantly decreased compared to the time steps in the preliminary experiment. Although the diffusion of information in has also increased significantly in the initial stage, it still takes an average of time steps to reach the maximum cognitive rate of information from new energy vehicles, and the buyers of new energy vehicles only reach about of that under normal circumstances (see Figure 2).
(a)
(b)
Among consumer groups without opinion leaders, the diffusion process of new energy vehicle information is slow and extremely limited. After 25 time steps, only about 110 consumers can learn about the new energy vehicles. Considering that the probability of diffusion of new energy vehicles through the mass media is about 1%, it can be considered that in the consumer market without opinion leaders, the ability to communicate is almost zero, and the weak ability of information diffusion also leads to the result of a purchase of in the .
The t-test analysis of the three groups of experimental data can yield a more actual and intuitive conclusion. At the significance level (when α is ), there are significant differences in the cognitive rate and the purchase rate of consumers in these three groups of experiments. It can be seen that different numbers of opinion leaders make different diffusion of new energy vehicles. We can conclude that opinion leaders play an essential role in the diffusion of new energy vehicles, and the speed of information and product diffusion depends on the number of opinion leaders.
4.2. Degrees of Innovation
To reflect the innovation of opinion leaders, we raised the thresholds of opinion leaders for fuel consumption and selling price; that is, they need to pay more attention to the price of the products. Figure 3 shows the diffusion of new energy vehicles in the network with different innovation capabilities of opinion leaders.
The solid line represents the data of the preliminary experiment. In contrast, the dotted line represents the diffusion of new energy vehicles in the environment of opinion leaders after the decline of innovation (that is, opinion leaders have the same fuel consumption and price threshold as general consumers). It can be seen from the two figures that the decline in the innovation of opinion leaders has helped the diffusion of new energy vehicles. In the experiment , new energy vehicles can reach the peak of information dissemination in an average of 2.5 time steps, covering all the 500 consumer subjects. The number of people who finally buy new energy vehicles is nearly two more than that in experiment . In both experiments, opinion leaders account for about half of the consumers who buy new energy vehicles. By comparing these data, opinion leaders’degree of innovation plays a minor role in the diffusion of new energy vehicles in this model (see Figure 3).
(a)
(b)
This result can be explained by the fact that the creative expression of opinion leaders in the model needs to be more accurate, and the difference in threshold cannot bring advantages to the final decision-making of choice. The second is the error caused by randomness in the model. Therefore, it needs to be further discussed in subsequent experiments whether opinion leaders’ innovation plays a positive role in diffusion.
4.3. Weights of Normative Influence
We conducted experiments and , respectively, which simulated the role of opinion leaders with different weights of regulatory influence in the diffusion process of new energy vehicles. The data analysis is shown in Figure 4.
(a)
(b)
From the abovementioned data analysis, whether opinion leaders value the impact of norms has nothing to do with the diffusion of new energy vehicle information. All three experiments can let consumers know about the relevant information of new energy vehicles in a relatively short time (the average time of is 2.71, is 2.82 time steps, and only needs 2.84 time steps). However, the weight of the normative influence of opinion leaders has produced unexpected results in purchasing new energy vehicles. Through comparison and analysis, the following reasons may cause it: Although car ownership in China has increased significantly in recent years, it has yet to be entirely popularized. For most Chinese consumers, car consumption may still be in the initial stage. Consumers are too closely connected with their identity, status, and strength when buying a car. They also hope to gain honor and reputation through car consumption (see Figure 4).
The weight of social utility (compared to personal utility) for Chinese consumers is too large compared to western consumers. Does this need to be adjusted accordingly in the initialization of the model? It may need further discussion in future experiments.
4.4. Judgment of Product Quality
Because opinion leaders have a more in-depth understanding and involvement in a particular field, they have more professional knowledge than ordinary consumers. They can more accurately judge the quality of relevant goods. In the experiment , the ability of opinion leaders has been stripped away; that is, opinion leaders can only make a general judgment on the quality of new energy vehicles. The data analysis is shown in Figure 5. The judgment ability of opinion leaders on product quality is not the main reason they can effectively promote the information and product diffusion of new energy vehicles (see Figure 5).
(a)
(b)
4.5. Weights of Price
Previous empirical studies have shown that, unlike ordinary consumers, opinion leaders do not care too much about the price of goods when buying. To test whether this characteristic is the factor that promotes the diffusion of new energy vehicles, we carry out the following data analysis for experiments and . It is not difficult to see that the price weight does not play a decisive role in the new energy vehicle market. The result may also be explained by the fact that the gap between ordinary consumers and opinion leaders is not very prominent, coupled with the randomness of the simulation system itself (see Figure 6).
(a)
(b)
4.6. Degrees of Trust
Ordinary consumers only believe in commodity information from opinion leaders and those who have used experience, reflecting ordinary consumers’ trust and dependence on opinion leaders in a specific field. To test whether the change in ordinary consumers’ trust in opinion leaders will affect the diffusion of new energy vehicles, the data analysis of the basic experiment and control group experiment is shown in Figure 7.
(a)
(b)
The graph obtained from the data analysis is similar to the graph of the number of different opinion leaders in Figure 3. It can be seen that the degrees of trust play an essential role in the diffusion of new energy vehicles. It is due to the great trust that they will recognize the information recommended by the opinion leaders, which makes the information related to new energy vehicles spread smoothly and quickly (see Figure 7).
5. Conclusions and Future Research
In this study, we explore the role of opinion leaders in the diffusion of new energy vehicles using an agent-based simulation model based on the NetLogo platform. This study comes to the following conclusions. First, opinion leaders play a significant role in the diffusion of new energy vehicles. It increases the diffusion speed of new energy vehicle information and promotes the sales of new energy vehicles to a certain extent. Second, the study also found that the critical factor causing this impact is the trust of ordinary consumers in opinion leaders. In particular, the role of normative influence on the diffusion of new energy vehicles is more significant than that attained by the conclusions of the relevant literature.
Through the simulation experiments, we understand the mechanisms that may influence the adoption rate of new energy vehicles. This model aims to provide practical information to automobile manufacturers and governmental policymakers on how consumers may be motivated to change their preferences more favorably toward the new energy vehicles. From the perspective of automobile manufacturers, they should fully use the role of opinion leaders in the group. Especially at the initial stage, they can target opinion leaders in the automotive consumer market, who may be auto enthusiasts, professionals in automotive industries, or other fields. From the perspective of government policymakers, we should deliberately cultivate opinion leaders to become propagandists and leaders of new social norms. Living in a society and feeling pressure from all aspects of society all the time, the consumers will comply with the opinions of mainstream ideas to a certain extent to maintain their status and interpersonal relationships in the group.
Our findings have some limitations that can be addressed in future research studies. First, the research object of the model designed in this paper is limited only to two types of agents. The promotion and diffusion of new energy vehicles involve a more comprehensive range of objects, such as automobile manufacturers and government agencies. In addition, to simplify the model, the attribute settings of the two entity types are relatively small, which cannot fully reflect all the characteristics of their actual market. Second, we expressed the heterogeneity between agents by using the characteristics of ABM during initialization; the characteristics of the whole social relationship group have yet to be indeed reproduced, such as the strength of the relationship between various subjects.
Future research may introduce new subjects, such as the government and automobile manufacturers, into the authenticity and complexity so as to bring them closer to the actual situation. At the same time, the model could include more complex network structures, such as dynamic networks, and be closer to the Chinese market.
Data Availability
The data used to support the findings of the study can be obtained from the corresponding author upon request.
Additional Points
Note: 1. WeChat Official Accounts: https://news.pedaily.cn/202112/483292.shtml. 2. Tencent Technology: https://new.qq.com/ch/tech/.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
This work was supported by the Fundamental Research Funds for the Central Universities (Grant No. ND2020001) and the Humanities and Social Sciences Project of the Ministry of Education (Grant No. 21YJA630088).