The government is the direct subject of online food safety governance and plays an important role in the formulation of food safety system (26, 27). The government can alleviate the food safety problem by increasing the inspection probability of enterprises and improving the punishment of self-discipline (28). In relevant research, Ortega et al. found that Chinese consumers have the highest willingness to pay for government certification programs, followed by third-party certification, traceability systems, and product specific information labels (29). A strict monitoring system can not only improve consumers’ welfare, but also restore consumers’ trust and increase social welfare. However, in the online food trading market, there are many and scattered food suppliers, heavy food safety supervision tasks, backward food safety supervision technology, and other problems, which make the defects of the government’s single governance of the online food trading market increasingly prominent (30, 31). Among them, Deng believed that the single government supervision model is a system obstacle to produce food safety risks. It is an inevitable choice for the reform of the food safety supervision model to move from a single government supervision model to a social co governance model (32). Hu pointed out that at present, China’s food safety supervision is faced with such dilemmas as endogenous conflict of policy objectives, structural mismatch of resources and powers, and adverse incentive of regulatory behavior (33). In view of the problems existing in government governance, scholars have given relevant solutions. Among them, Liu and Ma believed that the government should strengthen food safety education to enable the public to have basic knowledge (34). Wei and Yao pointed out that the government should improve the security system of data governance, clarify the operator access mechanism, limit the monopoly of network giants, build a cooperative rights protection mechanism, and reform the regulatory governance system (12). Yang et al. believed that the government should increase punishment, reduce supervision costs, strengthen self-discipline of enterprises, guide the public to participate in governance, and build a multi-agent collaborative governance mechanism (35).

In order to effectively solve the failure of government regulation and improve the problem of insufficient government regulation, some scholars introduced media to participate in the governance of online food safety problems. Cao et al. studied the role of new media in government food safety supervision and found that efficient and accurate new media supervision can effectively restrain food enterprises’ adulteration behavior and urge the government to perform due diligence supervision (36). Chen et al. believed that the media would take the lead in exposing food safety problems, grasp the guidance of public opinion, and help government regulators to strengthen supervision (37). Zhang et al. found that strengthening third-party supervision is conducive to promoting government regulatory authorities to strengthen supervision and improve enterprise food safety governance (38). Xie et al. found that the sensitivity of producers to perceived reputation loss, the subjective value judgment of media participation in social governance, and the government’s normalization of regulatory penalties are three important constraints for media participation in social governance of food safety (39). Zhang et al. found that reducing the cost of media supervision will not only stimulate the enthusiasm of consumers to complain, but also improve the efficiency of food safety supervision, so that food safety risks are kept at a low level for a long time (40). However, the media have the characteristics of timeliness and low cost and high income, which makes it difficult for them to fully understand the whole process of food events, and it is easy for them to report in a partial way (21, 41).

In the online food trading market, the online food trading platform is the bridge connecting food sellers and food buyers, and the main carrier of online food sales. Compared with the government, food sellers, food buyers, media, and other subjects, online food trading platforms have more advantages in information acquisition, collation, and analysis. They are direct participants in online food safety governance and have the rights and obligations to manage online food trading market (42, 43). Cheng and Dong pointed out that the online food trading platform should give play to its own advantages in technology, information and resources, take the initiative to undertake food seller information review, food information disclosure and other work, so as to ease the regulatory pressure of the government (44). Zhang et al. found that the daily supervision and management of online food trading platform on food safety is crucial to improve the level of food safety supervision on the platform (24). Although the online food trading platform can improve the supervision efficiency of the market, the lack of direct judicial punishment power and the market behavior of pursuing profit maximization will also make the platform “fail” in the process of online food safety governance (45). Therefore, in the online food trading market, it is necessary to strengthen the government’s supervision on the platform and strengthen the platform’s supervision responsibility (21).

With the continuous occurrence of food safety incidents, consumers’ safety awareness has gradually increased and participated in the governance of online food safety problems. Wang and Miao found that consumers’ participation in supervision will affect the production decisions of food enterprises, and enterprises will eventually transform to producing high-quality products (46). Wang and Sha believed that consumers’ education level and objective cognitive ability had a positive impact on online food safety risk prevention and control (47). Niu and Wu believed that the public should be encouraged to supervise and report, and the public interest litigation system and punitive compensation system should be established and improved (48). Zhu and Rong found that the increase of consumers’ real evaluation and complaints about rights protection can effectively promote manufacturers to provide high-quality products (49).

With the deepening of research, scholars found that the single subject regulation could not meet the needs of the rapid development of online food trading market. Therefore, it is very important to integrate the government, media, consumers, platforms, industry associations, and other subjects into the collaborative governance framework of online food safety through legalization, marketization, and other ways, and improve the collaborative governance capability through communication and cooperation (50, 51).

Game theory is a theory that uses rigorous mathematical models to study the optimal decision-making problem under the condition of conflict confrontation (52). In previous studies, Liu et al. built a signaling game model between online food trading platforms and sellers based on signaling game theory, and analyzed the formation conditions and results of different equilibria (53). Zhang et al. constructed a principal-agent model from the perspective of food sellers and food buyers, revealed the equilibrium evolution path of food quality and safety, and proved that the flooding of unsafe food in the market is the inevitable result of non-optimal equilibrium under the asymmetric information environment (23). Evolutionary game theory is a theory that combines game theory analysis with dynamic evolutionary process analysis. This method can help understand the dynamic process of group evolution, and explain why and how groups will reach this state. It has been widely used in management, economics, biology and many other fields (54). For example, in terms of environmental governance, Chu et al. built an evolutionary game model involving the central government, local governments and pollution enterprises, hoping to provide solutions for regional haze governance from the perspective of environmental regulation (55). In terms of online public opinion management, Wen constructed a game model for the evolution of online public opinion in colleges and universities involving the media, college students, universities and the government, and found that the main factors affecting the balance of the game system were the government’s supervision, the willingness of online media to report, the attention of colleges and universities to public opinion events, and college students’ self-awareness (56).

In recent years, scholars have also applied evolutionary game theory to online food safety governance. Liu constructed a static game payment matrix between the government and enterprises, and found that the combination of incentives and government supervision can effectively guide food enterprises to produce safe food (57). Xu et al. established an evolutionary game model involving suppliers and manufacturers, and found that the quality input strategy of food suppliers and manufacturers is closely related to the quality input-output ratio of both parties (58). Zhu and Sun built a tripartite evolutionary game model involving the government, food enterprises, and third-party testing institutions, and analyzed the interaction mechanism of strategy choices among different actors and the evolution trend of each subject’s strategy choices under different parameter changes (59). Wang et al. constructed an evolutionary game model of the behavior of the government and the seller involved in the platform supervision, and analyzed the strategies of the government and the seller under different supervision strengths of the platform (60). Cao et al. built a game model involving the government, the platform and the seller to discuss the collaborative supervision of the government and the platform on online food safety (61).

Therefore, this article constructs a tripartite evolutionary game model with the seller, buyer, and food trading platform, and analyzes the formation process and impact mechanism of the “lemon problem” in the online food trading market. The main contributions of this article are as follows: First, previous studies mainly analyzed online food safety from the theoretical level. However, this article focuses on the “lemon problem” in the online food trading market, and intuitively shows the formation process of the “lemon problem” in the online food trading market from the perspective of dynamic evolution, which can provide new evidence for the existence of the “lemon market.” Second, previous studies mainly focused on the relationship between the government, food sellers, and food trading platforms in the online food trading market, and mostly used the government and food trading platforms to regulate the behavior of food sellers. However, as direct participants in food transactions, food buyers’ behavior also has an important impact on the development and evolution of the market. Therefore, this article mainly analyzes the direct participants in the online food trading market, food sellers, food buyers, and food trading platforms, hoping to have a clearer understanding of the strategic choice of each subject. Third, previous studies mostly analyzed the conditions that affect the subject’s strategic choice behavior, and emphasized the role of supervision, but rarely conducted a comprehensive and in-depth analysis of the factors that affect the subject’s behavior. Therefore, based on the relevant conditions (stability conditions of evolutionary equilibrium) that affect the choice of the subject’s strategy, this article conducts an in-depth analysis of the factors that affect the choice of the subject’s strategy, and clarifies the extent and effect of different factors such as cost, punishment, subsidy, and income on each subject, which can provide a scientific basis for the online food trading platform and the government to formulate relevant institutional systems and policy measures.

Without considering the environment of online food trading market, it is assumed that seller, buyer, and online food trading platform can constitute a complete online food trading market based on the functions of each subject in the market. Assuming that each participant is a finite rational individual with information asymmetry among them, the following assumptions are made for the tripartite subjects based on evolutionary game theory.

Based on the above analysis assumptions, a benefit matrix of the tripartite game model is constructed with the seller, the buyer and the online food trading platform as the tripartite subjects. The benefits of all subjects under different scenarios are shown in Table 2. Among them, when the strategy of the seller, buyer and platform is {positive performance, positive participation, positive regulation}, the seller’s benefit is R₁ + R₅ + H − M − C₁, the buyer’s benefit is R₆ + R₈ + I − C₂, and the platform’s benefit is R₉ + R₁₀ − H − I − C₃.

When the seller, buyer and platform choose different strategies, they will get different benefits, as shown below.

By combining Formulae 1–3, the replicator dynamic system of the online food trading market can be obtained.

When U(x) = 0 ⋂ U(y) = 0 ⋂ U(z) = 0 in Formula 4, the equilibrium points of the replicator dynamic system can be obtained: E₁(0,0,0), E₂(0,0,1), E₃(0,1,0), E₄(0,1,1), E₅(1,0,0), E₆(1,0,1), E₇(1,1,0), E₈(1,1,1), and E₉(x*, y*, z*). In Su et al. (62), Xiao et al. (63), and other studies, scholars believe that the Evolutionary Stable Strategy (ESS) of the multi group evolutionary game must also be a pure strategy Nash equilibrium, that is, in an asymmetric game, the mixed strategy equilibrium must not be an evolutionary stability equilibrium. Therefore, this article will only analyze eight pure strategy equilibrium points, that is, do not discuss E₉(x*, y*, z*). The stability of each equilibrium point in the tripartite evolutionary game can be determined according to the Lyapunov stability theory, that is, when the replicator dynamic system is evolutionary stable, the eigenvalue of its Jacobian matrix is negative (64).

By calculating the partial derivatives of U(x), U(y) and U(z) for x, y, and z, respectively, the Jacobian matrix of replicator dynamic system of online food trading market can be obtained. The eigenvalues can be calculated by substituting the values of x, y, and z at each equilibrium point into Formula 5. For example, the eigenvalues of E₁ (0,0,0) is λ₁ = R₃ − R₄ − M, λ₂ = N − C₂, λ₃ = J + K − C₃. The eigenvalues of each equilibrium point are shown in Table 3.

For each equilibrium point, if its eigenvalues λ₁, λ₂, λ₃ are all negative, then it is the evolutionary stable point of the system. According to the actual operation of the online food trading market, C₂ > 0 and H + I + C₃ > 0 can be known by analyzing the equilibrium points. That is to say, the two equilibrium points E₇ (1,1,0) and E₈ (1,1,1) are unstable. If it is to achieve the evolution stability, the asymptotical stability of equilibrium points E₁ (0,0,0), E₂ (0,0,1), E₃ (0,1,0), E₄ (0,1,1), E₅ (1,0,0), and E₆ (1,0,1) need to be analyzed. The asymptotical stability conditions for each equilibrium point are shown in Table 4.

According to Table 4, we can know that many factors, such as the seller’s benefit, the seller’s cost, the buyer’s cost, the platform’s cost, the platform’s subsidy and punishment to the seller, the platform’s subsidy and punishment to the buyer, and the buyer’s punishment to the seller, will affect the eigenvalues of each equilibrium point. The stability of each equilibrium point in the asymptotic stability conditions ➀–➅ will be analyzed below (Table 5).

According to Table 5, under each asymptotic stability condition, there may be multiple equilibrium points in the replicator dynamic system. Specific conditions are as follows:

In the case of condition ➀, except that E₁ is the stable point of the Jacobian matrix of the replicator dynamic system, other equilibrium points are all unstable points. At this time, {negative performance, negative participation, negative regulation} is the evolutionary stability point of the system.

In the case of condition ➁, there is only E₂ as a stable point in the system. At this time, {negative performance, negative participation, positive regulation} is the evolutionary stability point of the system.

In the case of condition ➂, E₃ is the stable point, that is, {negative performance, active participation, negative regulation} is the evolutionary stable point of the system. In addition, the stability of E₅ and E₆ is uncertain. The specific conditions are as follows: When conditions ➂ and ➄ are met simultaneously, E₃ and E₅ are stable points and E₆ is unstable point. When conditions ➂ and ➅ are met simultaneously, E₃ and E₆ are stable points and E₅ is unstable point. In general, under condition ➂, there are at most two stable points in the system.

In the case of condition ➃, E₄ is the stable point, and the stability of E₅ and E₆ is uncertain. The specific conditions are as follows: When conditions ➃ and ➄ are met simultaneously, E₄ and E₅ are stable points and E₆ is unstable point. When conditions ➃ and ➅ are met simultaneously, E₄ and E₆ are stable points and E₅ is unstable point. In general, under condition ➃, there are at most two stable points in the system.

In the case of condition ➄, E₅ is the stable point, and the stability of E₃ and E₄ is uncertain. The specific conditions are as follows: When conditions ➄ and ➂ are met simultaneously, E₅ and E₃ are stable points and E₄ is unstable point. When conditions ➄ and ➃ are met simultaneously, E₅ and E₄ are stable points and E₃ is unstable point. In general, under condition ➄, there are at most two stable points in the system.

In the case of condition ➅, E₆ is the stable point, and the stability of E₃ and E₄ is uncertain. The specific conditions are as follows: When conditions ➅ and ➂ are met simultaneously, E₆ and E₃ are stable points and E₄ is unstable point. When conditions ➅ and ➃ are met simultaneously, E₆ and E₄ are stable points and E₃ is unstable point. In general, under condition ➅, there are at most two stable points in the system.

By analyzing the stability of the equilibrium points in the asymptotical stability conditions ➀–➅, it can be found that there may be multiple stable points in the replicator dynamic system under various asymptotic stability conditions. Specifically, when condition ➀ or condition ➁ is met, there will be only one stable point in the system. When condition ➂ and ➄, ➂ and ➅, ➃ and ➄, ➃ and ➅ are met respectively, there will be two stable points in the system.

For different sellers, buyers and online food trading platforms, the initial strategic choice behavior may be affected in many ways and show some differences. In order to avoid this impact, this article refers to the method of setting the initial value of the system in the previous evolutionary game analysis (65), and sets the initial willingness of each subject to low, medium, and high levels, that is _x0,y0,z0 ∈ Ω(0.2, 0.5, 0.8). By combining the initial willingness of the three subjects, 27 different scenarios can be obtained. For the convenience of comparison, the following only explores the cases where the initial willingness of each subject are consistent.

In the process of assigning values to each variable, first, we set the variable parameters according to the stability of condition ➀, that is, each variable needs to meet R₃ − R₄ − M < 0, N − C₂ < 0, J + K − C₃ < 0 at the same time. After constant debugging, we finally determined the parameters of condition ➀. Then, for the convenience of exploring the changing situation of each subject in the online food trading market, this article takes the condition ➀ as the base, with reference to the actual operating structure and interest relationship of the online food trading market, and assign values for variables ➀, ➁, ➂, ➃, ➄, ➅, ➂ and ➄, ➂ and ➅, ➃ and ➄, and ➃ and ➅for different situations. For example, for condition ➁, it is difficult to satisfy R₅ + H + J + R₃ − R₄ − M < 0, R₈ + I + K + N − C₂ < 0, − (J + K − C₃) < 0 by substituting the variable parameters in condition ➀. To this end, it is necessary to compare the difference between condition ➀ and condition ➁, and adjust on the basis of the relevant parameters of condition ➀. It is found by comparison that: condition ➁ can be met when M, C₂, and C₃ are adjusted, condition ➂ can be met when M and N are adjusted, condition ➃ can be met when M and J are adjusted, condition ➄ can be met when R₃ is adjusted, condition ➅ can be met when K and C₂ are adjusted, condition ➂ and ➄ can be met when M, N, C₂, and R₃ are adjusted, condition ➂ and ➅ can be met when M, N, K, C₂, and R₃ are adjusted, condition ➃and ➅ can be met when M, J, and R₃ are adjusted, condition ➂ and ➅ can be met when M, N, J, K, C₂, and R₃ are adjusted. The specific assignment of each variable under different conditions is shown in Table 6.

In the online food trading market, many conditions will occur if the system is to reach evolutionary stable equilibrium. The following simulates and analyzes the strategic choice behaviors of each subject under different conditions.

Under different conditions, the strategy of seller, buyer, and platform in the online food trading market will eventually stabilize in six situations.

In the cases of {negative performance, negative participation, negative regulation}, {negative performance, negative participation, positive regulation}, {negative performance, positive participation, negative regulation}, {negative performance, positive participation, positive regulation}, the seller will eventually choose negative strategy. That is to say, with the development of the online food trading market, more and more sellers on the market will choose to provide low-quality products, and the seller who choose to provide high-quality products will gradually be crowded out of the market, eventually leading to the “lemon problem” in the online food trading market.

In the case of {positive performance, negative participation, negative regulation}, the seller will eventually choose positive strategy, the buyer and the platform will choose negative strategy. Although the seller will choose positive strategy in this case, satisfying this scenario requires that the benefits the seller adopts positive strategy obtains are far greater than those of the seller adopting negative strategy. In actual situations, when the buyer and the online food trading platform choose negative strategy, the seller may obtain higher benefits from providing low-quality products.

In the case of {positive performance, negative participation, positive regulation}, the seller and the platform will choose positive strategy, while the buyer will choose negative strategy. However, in order to meet this scenario, it will be more costly for the buyer to positively participate, and the platform will also have to impose severer punishment on negative buyer. In reality, the cost of for buyer with positive participation is relatively low, and the platform is more to encourage buyer to choose negative performance than to impose higher punishment on the buyer. Because high punishment will cause a large number of buyer withdraw from the market, and make the entire online food trading market disappear.

In general, during the development and evolution of the online food trading market, the “lemon problem” will occur and will not disappear. To solve the “lemon problem” in the online food trading market, it is necessary to analyze the factors affecting the development and evolution of the online food trading market and formulate the corresponding measures to suppress and reduce the occurrence of the “lemon problem.”

In order to explore the influence of C₂ and C₃ on the choice of each subject’s strategy, we will analyze the evolution law of each subject’s strategy under scenarios {C₂ = 1, C₂ = 2, C₂ = 3}, {C₃ = 2, C₃ = 3, C₃ = 4}. Compare and analyze the evolution law of each subject’s strategy when C₂ and C₃ change (Figure 12). We can know the following.

When C₂ changes, the strategies of the seller and buyer will change significantly, and the platform will change little. Specifically, with the increase of C₂, the probability and speed of the seller and buyer to choose positive strategy will gradually decrease, and the probability and speed of the platform to choose positive strategy will gradually increase. In other words, C₂ will have a negative impact on the positive strategy choice of the seller and the buyer, and will have a positive impact on the positive strategy choice of the online food trading platform. The reason is that if a large amount of investment cannot be exchanged for the same or more benefits, the buyer’s enthusiasm for participating in market governance and rights protection will also be greatly reduced. The reduction of the buyer’s willingness to defend their rights will reduce the supervision of the seller’s behavior and make them tend to choose negative strategy. At this time, in order to maintain market order and retain more users, the food trading platform will choose positive strategy.

When C₃ changes, the strategies of the seller, buyer, and platform will change to some extent. Specifically, with the increase of C₃, the probability and speed of the seller, buyer, and platform to choose positive strategy will gradually decrease. In other words, C₃ will have a negative impact on the positive strategy choice of the seller, buyer, and platform. The reason is that when the platform increases the regulation cost, it will restrict the behavior of the seller and buyer from many aspects, and will transfer the cost to the seller and buyer, which will reduce the enthusiasm of the seller and buyer and drive them to withdraw from the online food trading market.

In order to explore the influence of N, J, and K on the choice of each subject’s strategy, we will analyze the evolution law each subject’s strategy under scenarios {N = 0.1, N = 1, N = 2}, {J = 0.1, J = 1, J = 2}, {K = 0.1, K = 1, K = 2}. Compare and analyze the evolution law of each subject’s strategy when N, J, and K change (Figure 13). We can know the following.

When N changes, the strategies of the seller and buyer will change significantly, and platform will change slightly. Specifically, with the increase of N, the probability and speed of the seller and buyer to choose positive strategy will gradually increase, and the probability and speed of the platform to choose positive strategy will gradually decrease. In other words, N will have a positive impact on the positive strategy choice of the seller and buyer, and will have a negative impact on the positive strategy choice of the online food trading platform.

When J changes, the seller will change significantly, the online food trading platform will change slightly, and the buyer will not change. Specifically, with the increase of J, the probability and speed of the seller and platform to choose positive strategy will gradually increase. In other words, J will have a positive impact on the positive strategy choice of the seller and platform.

When K changes, the seller, the buyer, and the online food trading platform will change to some extent. Specifically, with the increase of K, the probability and speed of the seller, buyer, and platform to choose positive strategy will gradually increase. In other words, K will have a positive impact on the positive strategy choice of the seller, buyer and platform.

In general, a certain degree of punishment will prompt relevant subjects to choose positive strategy. Because the benefits obtained in the process of punishment will urge the subject who implements punishment to choose positive strategy, while the subject who accepts punishment will also choose positive strategy in order to avoid losses.

In order to explore the influence of H, I, R₅, and R₈ on the choice of each subject’s strategy, we will analyze the evolution law of each subject’s strategy under scenarios {H = 0.1, H = 1, H = 2}, {I = 0.1, I = 1, I = 2}, {R₅ = 0.1, R₅ = 1, R₅ = 2}, {R₈ = 0.1, R₈ = 1, R₈ = 2}. Compare and analyze the evolution law of each subject’s strategy when H, I, R₅ and R₈ change (Figure 14). We can know the following.

When H, I, R₅, and R₈ change, the strategies of the seller, buyer and platform will change slightly. This is mainly because the decision on how much subsidy and additional benefit the subject can obtain is made by other subjects rather than itself. From the perspective of interests, other entities will not provide them with more subsidies, which will not make major changes due to changes in H, I, R₅, and R₈.

With the increase of H, the probability and speed of the seller to choose a positive strategy will gradually increase, while the probability and speed of the buyer and platform to choose a positive strategy will gradually decrease. That is to say, H will have a positive impact on the positive strategy choice of the seller, and will have a negative impact on the positive strategy choice of the buyer and the online food trading platform.

With the increase of I, the probability and speed of the seller and platform to choose a positive strategy will gradually decrease, while the probability and speed of the buyers to choose a positive strategy will gradually increase. That is to say, I will have a negative impact on the positive strategy choice of the seller and platform, and will have a positive impact on the positive strategy choice of the buyer.

With the increase of R₅, the probability and speed of the seller to choose a positive strategy will gradually increase, while the probability and speed of the buyer and platform to choose a positive strategy will gradually decrease. That is to say, R₅ will have a positive impact on the positive strategy choice of the seller, and will have a negative impact on the positive strategy choice of the buyer and platform.

With the increase of R₈, the probability and speed of the seller and buyer to choose a positive strategy will gradually increase, while the probability and speed of the platform to choose a positive strategy will gradually decrease. That is to say, R₈ will have a positive impact on the positive strategy choice of the seller and buyer, and will have a negative impact on the positive strategy choice of the platform.

The analysis shows that R₁, R₂, R₃, and R₄ do not directly affect the evolution game of the system, but affect the system through the benefit difference between the positive and negative performance of the seller, that is, affect the evolution law of the system through R₁ − R₂ and R₃ − R₄. Figure 15 shows the tripartite strategy evolution trends when R₁ and R₂ change and R₁ − R₂ unchanged, and Figure 16 shows the tripartite strategy evolution trends when R₃ and R₄ change and R₃ − R₄ unchanged.

In order to explore the influence of M, R₁ − R₂, and R₃ − R₄ on the choice of each subject’s strategy, we will analyze the evolution law of each subject’s strategy under scenarios {M = 1, M = 2, M = 3}, {R₁ − R₂ = 0.1, R₁ − R₂ = 1, R₁ − R₂ = 2}, {R₃ − R₄ = 0.1, R₃ − R₄ = 1, R₃ − R₄ = 2}. Compare and analyze the evolution law of each subject’s strategy when M, R₁ − R₂, and R₃ − R₄ change (Figure 17). We can know the following.

When M, R₁ − R₂, and R₃ − R₄ change, the seller’s strategy changes most significantly, followed by the buyer, and the platform’s strategy changes less. The reason is that when the cost difference and benefit difference between the positive and negative performance of the seller are large, the seller will quickly react and choose strategies that are beneficial to itself, and the seller’s different strategies will also have an impact on the buyer and cause them to change indirectly.

With the increase of M, the probability and speed of the seller to choose a positive strategy will gradually decrease, and the probability and speed of the buyer and platform to choose a positive strategy will gradually increase. That is to say, M will have a negative impact on the positive strategy choice of the seller, and will have a positive impact on the positive strategy choice of the buyer and platform.

With the increase of R₁ − R₂ and R₃ − R₄, the probability and speed of the seller to choose a positive strategy will gradually increase, and the probability and speed of the buyer and platform to choose a positive strategy will gradually decrease. That is to say, R₁ − R₂ and R₃ − R₄ will have a positive impact on the positive strategy choice of the seller, and will have a negative impact on the positive strategy choice of the buyer and platform.

Through systematic analysis of the structure and operation mode of the online food trading market, this article takes the seller, buyer, and online food trading platform as tripartite subjects and constructs a tripartite evolution game model. With the help of MATLAB software, it is possible not only to simulate the formation process of the “lemon problem” in the online food trading market dynamically, but also to conduct an in-depth analysis of the strategic choice behavior of each subject. In general, the following conclusions can be drawn.

Through the analysis of the stable points of evolution of the seller, buyer, and online food trading platform, it is found that the “lemon problem” occurs with the development and evolution of the online food trading market and will not dissipate. In the online food trading market, the strategy of the tripartite subjects will stabilize in six situations. And the “lemon problem” occurs in all six situations. In the research of Zhang et al., it was also pointed out that under the network trading environment of asymmetric information, the characteristics of trusted products of food safety and the self-interest motives of various stakeholders in the food supply chain make unsafe food flooding the market an inevitable result (23). In order to control the “lemon problem” in the online food trading market, it is necessary to deeply analyze the factors that affect the development and evolution of the market, and use advanced technical means and effective management mechanisms to alleviate the problem of information asymmetry in the market (66, 67).

In the online food trading market, different factors have different influence directions on the subject strategy. For the seller, N, H, J, K, R₁ − R₂, R₃ − R₄, R₅, and R₈ will have a positive impact on the seller’s positive strategy choice, while M, I, C₂, and C₃ will have a negative impact on the seller’s positive strategy choice. For the buyer, M, N, I, K, and R₈ will have a positive impact on the buyer’s positive strategy choice, while H, C₂, C₃, R₁ − R₂, R₃ − R₄, and R₅ will have a negative impact on the buyer’s positive strategy choice. For the online food trading platform, M, J, K, and C₂ will have a positive impact on the platform’s positive strategy choice, while N, H, I, C₃, R₁ − R₂, R₃ − R₄, R₅, and R₈ will have a negative impact on the platform’s positive strategy choice. Clarifying the impact of different factors on each subject can not only help understand the strategic choice behavior of each subject in the online food trading market, but also help and guide the subject’s behavior from various aspects. For example, we can change M, J, K, and C₂ to promote the platform to choose a positive strategy.

In the online food trading market, different factors have different influence degrees on the subject strategy. Among them, M, N, H, I, J, K, C₂, C₃, R₁ − R₂, R₃ − R₄, and R₅ will have a significant impact on the seller’s strategic choice behavior, M, N, K, C₂, C₃, and R₈ will have a significant impact on the buyer’s strategic choice behavior, H and C₃ will have a significant impact on the platform’s strategic choice behavior. Analyzing the degree of influence of different factors on the subject’s strategy can help us clearly identify the key factors that affect the subject’s strategy choice and point out the direction for the subsequent development of solutions. However, the degree of influence mentioned in this article is more the result of the comparison among multiple subjects, rather than quantitative comparison. In order to more clearly identify the differences between the impact degrees, we will collect realistic data from various aspects in the future to verify and deepen relevant conclusions.

In the online food trading market, cost, punishment, subsidy and benefit have different effects on the subject’s strategy. Among them, cost and cost difference have the most significant impact on the subject’s strategy, followed by punishment and benefit difference, and subsidy and additional benefit have less impact on the subject’s strategy. Therefore, if we want to build a good and sustainable online food trading market, we need to focus on cost reduction and reveal food quality and safety information at the lowest management cost (68). At present, establishing a good institution (such as signal detection mechanism, reputation mechanism, and reward and punishment mechanism) has become an inevitable requirement for the healthy development of online food trading market (53). However, most of the supervision in the market is ex-post supervision, and although the incentive forms are diverse, there is no clear and scientific basis, which requires comprehensive management of the online food trading market from various aspects.

In order to effectively suppress the “lemon problem” in the online food trading market, provide consumers with rich and diverse, healthy and safe food, and bring more benefits to food suppliers and food trading platforms, this paper presents the following suggestions.