Whether enterprises adopt low-carbon technologies is not merely a strategic choice, but a pivotal step within the systematic project of carbon-emission reduction that demands coordinated support from policy, capital, and technology [20]. The PEST model is a strategic analysis tool proposed by Johnson and Scholes to help companies examine the external macro environment and then make decisions [21]. This tool has been applied to research on green transformation [22], clean energy substitution green strategy formulation [23]. The diffusion of LCT in enterprises is mainly affected by political, economic, social and technological factors. The PEST model can be used to systematically analyze these four types of external environmental factors.

Political factors. The influencing factors can be divided into incentive policies and punitive policies. Incentive policies encompass subsidies aimed at encouraging enterprises to adopt LCT, such as new energy vehicles and photovoltaic power generation. These subsidies are designed to support advancements in sustainable practices [9]. Conversely, punitive measures are in place to effectively penalize polluting enterprises. China’s carbon tax policy requires enterprises that fail to adopt LCT to purchase carbon quotas or face government fines, thereby reinforcing environmental accountability [9]. Economic factors. Currently, there continues to be a price premium for green products compared to traditional ones. This enables enterprises to command higher sales prices for their green products, thereby enhancing both their profit margins and market competitiveness [24]. Social factors. The awakening of consumer green consciousness has fostered the emergence of low carbon preferences, making green products more popular in the market and further driving enterprises to adopt LCT. The increase in consumers’ low-carbon preferences also promote low-carbon decision-making within the supply chain [10]. Technical factors. The consideration of technological costs mainly covers expenses during the application of LCT. These include technology transfer, personnel training, and improved management practices needed for its adoption [11].

According to the resource dependence theory, while enterprise decision-making behaviors are complex, they are also traceable and explicable. Influenced by herd behavior and competitive advantage theory, when upstream and downstream enterprises in an enterprise’s supply chain, or competing firms, widely adopt LCT, decision-makers are influenced by herd mentality to accelerate LCT adoption [13]. On the other hand, influenced by green pressure from neighboring enterprises that have adopted LCT in the majority, businesses strive to maintain market share and a competitive edge, thereby hastening the adoption of LCT as well [17]. Inertia is a significant factor in real-world decision-making processes for individuals prone to procrastination [17]. Similarly, influenced by decision-making habits and cognitive biases, not all enterprises adopt LCT purely rationally; organizational inertia characterizes this resistance to decision-making, serving as a pivotal concept. The existing literature is insufficient to address factors such as herd behavior and decision inertia in the decision-making of companies using LCT. In addition, there is a lack of in-depth research on how these factors affect the enterprises’ decisions to adopt LCT. Most of them focus on a single factor, and research on combined factors is insufficient.

As scholars gain a deeper understanding of how the characteristics of social network topology and the heterogeneity of network actors influence differentiated enterprise decision-making, methods based on networked evolutionary game theory are employed to simulate the dynamic evolution of agents within complex social systems. This approach aims to explain the mechanisms underlying decision-making by modeling the dynamics and outcomes of agents in complex social systems. In the realm of energy and environment studies, enterprises exhibit intricate connections and interactions, characterized by complex network dynamics. Through the lens of complex network evolutionary game theory, the exploration of inter-enterprise technology diffusion offers a novel approach to understanding enterprise decision-making processes. Extant research has examined the dissemination dynamics of various corporate endeavors, including the substitution of clean energy [11], the production of green products [22] and technological advancements [25, 26]. Specifically, within the context of complex network evolutionary game theory, the literature exploring the diffusion of LCT among enterprises predominantly focuses on two pivotal aspects: first, the influence of diverse network structures on the propagation of these technologies; and second, the role of varying diffusion mechanisms in facilitating their widespread adoption.

The impact of network structure on LCT diffusion. Zhao et al. [27] conducted a comprehensive comparison of authoritative networks at the present stage, including the nearest neighbor coupled network, small-world network, scale-free network, and random network. They found that the diffusion rate is highest in the nearest neighbor coupled network, followed by the Watts–Strogatz (WS) small-world network and Barabási-Albert (BA) scale-free network. However, in Erdős–Rényi (ER) random networks, the strategy for new energy vehicles cannot achieve complete diffusion within the observation period, primarily because of the low clustering coefficient and long average path length characteristic of ER random networks. In their study, Wang and Zheng [10] examined how average degree, degree distribution, and consumer environmental awareness impact the diffusion of LCT. They discovered that under conditions of low consumer environmental awareness, small-world networks facilitate diffusion more effectively than scale-free networks. However, beyond a certain threshold of consumer environmental awareness, their findings suggest the opposite conclusion. Li et al. [11] utilized a BA network to characterize enterprise clean energy substitution strategies. Fan et al. [28], in modeling the impact of demand-side policies on LCT, applied the BA network to depict connections between enterprises. Similarly, Wang et al. [5] employed a BA scale-free network when studying innovation diffusion in supply chain networks. In practice, the scale-free nature of networks, marked by growth and preferential attachment, mirrors the real-world situation where a few enterprises wield considerable influence over connections. This theoretical framework has been widely applied in the aforementioned studies involving enterprises. Hence, this paper employs the BA scale-free network as the foundational model for network construction.

The impact of diffusion mechanism on LCT diffusion. Wu et al. [29] designed a multilayer network involving local governments, enterprises, and consumers to investigate the impact of subsidies, carbon taxes, and other policies on the diffusion of LCT. They validated that a mixed policy approach incorporating targeted fines can significantly enhance the diffusion of LCT. When assessing whether enterprises adopt LCT, updating node strategies through the Fermi rule involves directly comparing the benefits of the enterprise itself and those of selecting neighboring nodes [30]. Alternatively, considering the impact of decision inertia, profits are updated based on the difference in benefits and the ratio of benefits to the enterprise itself [5]. Existing studies have also highlighted the influence of decision inertia on enterprises. Chang et al. [17] categorize enterprises into those influenced by organizational inertia and those unaffected by it, and update their strategies according to refined Fermi rules and standard Fermi rules, respectively. The threshold model, unlike pure benefit-based updating mechanisms, emphasizes the collective characteristics of individuals. In promoting LCT, researchers have used the threshold model to examine the spread of green behaviors among individuals. Relevant studies have set homogeneous updating thresholds and analyzed the behavioral updating process influenced by the proportion of neighbors adopting green low-carbon behaviors [31]. The threshold model is also applicable in product diffusion research; each node’s threshold is closely related to the scale of product dissemination. This perspective offers insights into the dissemination of LCT: within specific social networks, when some enterprises demonstrate a positive attitude toward LCT and disseminate related information, other enterprises may adopt corresponding low-carbon behaviors once they reach a certain threshold. In summary, the threshold model provides an effective framework for studying the diffusion of LCT within networks.

The current update mechanisms have aimed to integrate corporate decision-making inertia and revenue mechanisms into the diffusion process of LCT, albeit mostly as adaptations of the Fermi criterion. In practice, during the adoption of LCT, enterprises may face limitations due to their own scale and capabilities, making a direct comparison of the benefits of adopting versus rejecting such technologies difficult. Under the influence of herd behavior, however, enterprises may opt to adopt LCT directly, thereby reducing decision costs. It is also possible that when a large enough enterprise in the supply chain adopts LCT, the resulting pressure forces other enterprises to adopt these technologies as well. Herd behavior can reduce decision costs for enterprises and expedite the diffusion of LCT among them. However, it may also lead some enterprises to abandon adoption, resulting in adverse effects. The influence of herd behavior on enterprise adoption of LCT has not been fully integrated into diffusion mechanisms. Whether herd mentality within networks positively or negatively affects the diffusion of LCT remains to be verified.

Generally, enterprises tend to make decisions based on the principle of maximizing effects with a rational attitude. However, research indicates that the development of enterprise networks can make corporate decision-making susceptible to external environmental influences [5]. At times, enterprises may also exhibit irrational attitudes, leading to cognitive biases, especially in situations of information uncertainty, when they are more prone to conformity behavior. Therefore, we consider that there are two different paths for enterprises to decide to adopt LCT (Figure 1). Accordingly, we employ the PEST framework to characterize enterprises’ payoff updating mechanisms under the rational decision-making path [21], while a threshold model is introduced to describe strategy updating based on neighboring enterprises when behavioral factors are taken into account [32]. With probability α , enterprises update their strategy based on gains. With probability 1 − α , enterprises update their strategy according to its neighbors. This decision is influenced by the dual impacts of supply chain competitive pressures and herd behavior. When a certain proportion of neighboring enterprises adopt LCT, firms experience pressure from these neighbors, prompting them to also adopt LCT. We assume a perfectly competitive market producing homogeneous and fully substitutable products. The strategy set of each enterprise consists of two options: adopting LCT or rejecting LCT. Enterprises that adopt LCT are referred to as adopting enterprises and produce green products, whereas enterprises that reject LCT produce traditional products. Green products feature lower carbon emissions but higher prices and costs, while traditional products exhibit the opposite pattern.

We posit that as enterprises adopt LCT, there may be an influx of more green products into the market, influencing consumer preferences towards green products. Therefore, consumer green preferences evolve over time. Specifically, with the maturity of LCT and changes in consumer satisfaction, this can be represented by iterative Equation 1 for consumer green preferences at time t : η t = η 0 + τ ∗ 1 − η 0 ∗ φ c t − φ c 0 (1) Where η 0 represents the initial value of consumer green preferences, φ c 0 denotes the proportion of enterprises adopting LCT at the initial time, φ c t represents the proportion of enterprises adopting LCT at time t . It is calculated as φ c ( t ) = N ( t ) N , where N ( t ) denotes the number of enterprises adopting LCT in the network at time t , and N denotes the total number of enterprises. The dynamic evolution of φ c ( t ) originates from the strategy updating process of enterprises. After each round of the evolutionary game, enterprises decide whether to adjust their strategies based on payoff comparisons with their neighbors in the network or under the influence of herding effects. As a result, the number of enterprises adopting low-carbon technology changes over time, which in turn leads to the dynamic evolution of φ c ( t ) . And τ is the preference adjustment intensity factor, which lies within the interval 0,1 . A larger value of τ indicates that consumer green preferences change more significantly with variations in the proportion of enterprises adopting LCT.

The average market demand for enterprises adopting LCT and rejecting LCT is represented by Equation 2: q g = η t Q φ c t N q t = 1 − η t Q 1 − φ c t N (2) where Q denotes the constant total market demand. Consumers’ preference for green products is represented by η ( t ) , indicating the proportion of consumers at time t who choose green products. The variable φ c ( t ) denotes the proportion of enterprises adopting LCT at time t , and N represents the total number of enterprises in the market. Accordingly, q g and q t denote the per-enterprise demand for green and traditional products, respectively.

The government compensates enterprises adopting LCT by subsidizing based on production output: S = s q g , government intervention is captured by S , which denotes the direct subsidy provided to enterprises adopting LCT. Where s represents the government subsidy intensity per unit of green product.

Assuming all enterprises in the market are subject to environmental regulations and carbon tax policies, denoted by carbon price p c , enterprises rejecting green transformation emit e units of carbon per unit of production, while those adopting LCT achieve an emission reduction ratio of r . Consequently, taxes for enterprises rejecting and adopting LCT are denoted as T t and T g respectively. The specific expressions are given in Equation 3: T t = e p c q t , T g = e 1 − r p c q g (3)

In the process of adopting LCT, upgrading existing technology and equipment for producing green products, along with providing relevant training for managers, is essential. Assuming the technical cost per unit output is represented by σ , the benefits for enterprises adopt ( G A ) or reject ( G R ) LCT are as Equation 4: G A = p g − c g − σ q g + S − T g , G R = p t − c t q t − T t (4)

The parameters c g and c t represent the production costs of grleen and traditional products, while p g and p t denote their associated sales prices. The level of punishment F = ω ( p t − c t ) q t is contingent upon the company’s profitability, and the enterprise’s profit is G R ′ = ( p t − c t ) q t − F − T t . Therefore, the income matrix of the enterprise is shown in Table 1. This presents the payoff matrix of pairwise interactions between two connected enterprises in the network. When an enterprise adopting LCT selects one of its neighbors for interaction, the payoffs depend on the strategy combination of the two enterprises. Specifically, if both the focal enterprise and its selected neighbor adopt LCT, they obtain identical payoffs, denoted as ( G A , G A ) . Similarly, if both enterprises reject LCT, their payoffs are also identical and given by ( G R , G R ) . However, when the two enterprises adopt different strategies, the enterprise adopting LCT obtains a payoff of G A , while the enterprise rejecting LCT incurs an additional penalty F on the basis of G R . In this case, the payoff of the non-adopting enterprise is denoted as G R ′ = G R − F , and the corresponding payoff pairs are ( G A , G R ′ ) or ( G R ′ , G A ) .

Let G = ( V , E ) denote the network, where V = { V 1 , V 2 , … , V N } represents the set of all nodes in the network, and E denotes the set of relationships among nodes, which can be expressed as: E = e 11 e 12 ⋯ e 1 N e 21 e 22 ⋯ e 2 N ⋮ ⋮ ⋱ ⋮ e N 1 e N 2 ⋯ e N N .

Here, e i j indicates the relationship between nodes v i and v j . The value of e i j reflects whether a cooperative relationship exists between the two nodes. Specifically, when e i j = 1 , nodes v i and v j are connected by a cooperative relationship; when e i j = 0 , no cooperative relationship exists between them. Since this study considers only first-order neighbors, when e i j = 1 , node j is regarded as a neighbor of node i .

The strategy set consists of adopting LCT or refusing it. The probability of selecting a strategy correlate positively with the expected cumulative payoff, allowing for stochastic decision-making. Enterprises engage exclusively with their neighbors and consider only first-order neighbors. Strategies are updated over time steps according to the same gaming rules. The focus of the study is on whether enterprises adopt LCT. Enterprises are categorized into two groups: those adopting LCT ( A ) and those refusing it ( R ) . The mechanism by which enterprises decide whether to adopt LCT is outlined as follows.

Figure 3 depicts the variation in the proportion of enterprises adopting LCT in the network over time t, under different preferences for benefits and decision biases. The panel (a) shows a network with a total of N = 300 nodes, while the panel (b) shows a network with N = 500 nodes. We consider different values of α as 0, 1, 0.2, 0.5, and 0.8. Here, the first two values represent single-strategy decisions: a value of 0 indicates strategy updates based solely on the influence of LCT adoption proportions among neighbors, while a value of 1 indicates updates based solely on benefit. The results show that under the two network scales, the proportion of enterprises initially adopting LCT in the network is 10 % . This proportion is obviously relatively low, which is also the state of the market faced by most enterprises in the initial stage of green transformation. At this stage, benefit-driven decisions would favor the adoption of LCT by more enterprises in the network. However, increasing the influence of herd behavior on decision-making mechanisms ultimately results in a lower proportion of collaborators in the game network. It is worth noting that the network size has little effect on the proportion and evolution trend of the final collaborators, and the conclusion is consistent with the existing research [11]. Therefore, in the follow-up study, we set the number of enterprises in the network to 300.

The conclusions drawn from Figure 3 are obtained under a fixed external environment, in which key parameters remain constant across the network. In reality, however, enterprises operate in a dynamic external environment characterized by continuously evolving PEST factors. To better reflect this reality, the subsequent analysis investigates how different external environments influence enterprises’ adoption of LCT. Given that profit motives play a fundamental role in firms’ strategic decisions, the profit preference parameter α is set to 0.2 in the following analysis.

Hereafter, we examine how variations in the decision-making inertia parameter λ influence the equilibrium proportion of enterprises adopting LCT within the network. Considering that some enterprises may exhibit herd behavior by adjusting decisions based on the adoption of their neighbors, we analyze how different levels of organizational inertia—none, weak, moderate, strong, and very strong—affect final adoption outcomes. As shown in Figure 8, organizational inertia does not follow a strictly linear relationship with the final adoption proportion, indicating that greater inertia does not necessarily promote LCT adoption. Given an initial adoption level of 0.1, enterprises tend to follow majority-neighbor behavior, which reduces LCT adoption. This suggests that enterprises should remain strategically cautious and avoid blind conformity, especially in the early stages of industry-wide LCT adoption. The results reveal a non-monotonic relationship between organizational inertia and the diffusion of LCT, indicating that stronger inertia does not necessarily lead to higher adoption levels. In the early stage of diffusion, when the initial adoption proportion is low, herd behavior dominates decision-making, causing firms to follow the majority of neighboring enterprises and thus discouraging early adoption. Moderate levels of organizational inertia, however, can help stabilize firms’ strategies once adoption has gained sufficient momentum, preventing excessive switching and supporting sustained diffusion. These findings underscore the importance of stage-dependent decision strategies and caution against blind conformity in the early phases of industry-wide low-carbon transitions.

We employed Monte Carlo simulations to assess the integrated effects of different factor combinations on the diffusion of LCT across enterprises. Specifically, we considered three dimensions—government, consumer, and public—encompassing policy subsidies, consumer green preferences, and public complaint penalties. To minimize stochastic errors in simulations, each scenario was run 50 times in independent experiments. Figure 9a shows that when consumer green preferences and government subsidies vary simultaneously, the diffusion rate of low-carbon technologies (LCT) in the network reaches 0.8–0.9. However, the relationship between these factors and the diffusion rate is nonlinear. When both preferences and subsidies are low, their effects are similar; as consumer preferences approach 0.9–1, increasing subsidies beyond 0.06 has little additional effect on diffusion, reflecting the diminishing marginal utility of subsidies and the nonlinear dynamics of adoption. Figure 9b indicates that enhancing subsidies and strengthening public complaint penalties can maintain the diffusion rate at 0.8–0.9, and compared to Figure 9a, the blue region is much larger, suggesting that these policies significantly promote LCT adoption even at early stages when consumer preferences remain constant. Figure 9c illustrates that the combination of consumer preferences and penalty intensity can further increase LCT diffusion, potentially exceeding 0.9, although as consumer preferences strengthen, the effect of adjusting penalties diminishes. The results reveal pronounced interaction and substitution effects among different policy and societal factors in promoting LCT diffusion. When consumer green preferences are sufficiently strong, demand-side pressure alone can sustain high adoption levels, reducing the marginal effectiveness of additional subsidies or penalties. In contrast, under weaker consumer preferences, combinations of government subsidies and public complaint penalties exhibit strong complementarity, significantly accelerating early-stage diffusion. These findings highlight the necessity of coordinated policy design that accounts for nonlinear interactions among multiple instruments, rather than relying on isolated measures to promote low-carbon technology adoption.

This study constructs a comprehensive analytical framework to simulate the factors influencing enterprises’ adoption of low-carbon technology (LCT), emphasizing both external and internal determinants. By introducing a threshold model into a complex network framework, the study captures the interactive consistency among enterprises. Building upon the PEST model, we further incorporate behavioral factors, namely herd behavior and organizational inertia, thereby addressing an important gap in existing research that tends to overlook enterprises’ internal decision-making dynamics [9, 29, 33].

The findings both confirm and extend previous studies. Consistent with existing literature, carbon taxes, subsidies, and penalties are found to promote the diffusion of LCT, which aligns with prior empirical and theoretical results [29]. However, our results further indicate that carbon taxes may have limited or even adverse effects in the short term, particularly in the new energy vehicle (NEV) sector, echoing the findings reported in [33]. In addition, consumer green preferences and public complaint penalties significantly enhance adoption rates, highlighting the importance of social pressure and demand-side forces in the diffusion process. Notably, green premiums emerge as a critical yet underexplored economic factor: moderate green premiums stimulate LCT adoption, whereas excessively high premiums make adoption outcomes highly dependent on consumers’ willingness and ability to pay.

This research advances the literature by quantitatively examining the roles of herd behavior and organizational inertia in LCT diffusion. The results suggest that excessive herding can inhibit technology adoption when initial participation is low, while a moderate degree of organizational inertia helps sustain long-term strategic commitment, thereby facilitating stable diffusion across the network. Compared with existing studies such as [5, 11, 29], which primarily examine the effects of single policy instruments on low-carbon technology diffusion, this study further explores the combined impacts of multiple policy and demand-side factors within a unified analytical framework. By jointly considering government subsidies, consumer green preferences, and public complaint penalties, our results demonstrate that the diffusion process exhibits pronounced nonlinear and synergistic effects under combined policy scenarios. In particular, the effectiveness of subsidies and penalties depends critically on the level of consumer green preferences, suggesting that coordinated policy mixes are more effective than isolated instruments in promoting sustained LCT diffusion.

From a policy perspective, this study provides empirical evidence supporting integrated policy design. When LCT adoption is initially low, combining subsidies and carbon taxes can improve enterprise profitability and counteract herd-driven abandonment. Policymakers should also compare the efficiency of different incentive tools to optimize resource allocation. For the NEV industry, subsidies remain more effective than carbon taxes in encouraging LCT adoption. Meanwhile, demand-side drivers, such as fostering green consumer preferences and enforcing public scrutiny mechanisms, can amplify policy effects. Governments should promote green consumption culture through education, subsidies, and awareness campaigns to create sustained demand for eco-friendly products. For enterprise managers, this study highlights the importance of maintaining strategic persistence during early adoption stages to build competitive advantages. Given the rising institutional pressures for environmental protection, firms should integrate green transformation strategies aligned with national sustainability goals.