关键词: 有限监管能力, 异质风险偏好, 智能监管机制, 污染物减排, 复杂网络, 演化博弈

Abstract: In recent years, with the continuous advancement of industrialization, the environmental problems caused by pollutant emissions have attracted much attention from society. It is difficult for regulators to supervise all potential polluting enterprises in a region at the same time, which leads to environmental degradation. Risk preferences may differ between enterprises. Heterogeneous risk preferences can lead to differences in behavior among enterprises. Regulators can determine the type of risk preferences of enterprises by observing their historical decisions, and strengthen supervision of risk-seeking firms that often adopt defection strategies.
In order to provide supervisory strategy support to supervisory authorities, the optimization of pollutant abatement supervision mechanism is studied based on assumptions of heterogeneous risk preferences of enterprises and limited supervision ability of regulators. Aiming at the issue of regional enterprises' pollutant abatement and supervision, a game model on the pollutant abatement alliance network is established combined with theories of complex network and evolutionary game. In this game model, regulators set two pollutant abatement targets: A minimum abatement target and a higher abatement target. The minimum abatement target is the minimum requirement to safeguard the region from environmental degradation. If enterprises fail to meet the minimum target, the supervised enterprises adopting the defection strategy are excepted from the shutdown penalty. If enterprises meet the minimum target, the supervised enterprises adopting the defection strategy are excepted from the shutdown penalty. Fines are imposed by supervised enterprises adopting defection strategies if they accomplish higher targets. Based on bounded rationality theory, we consider that enterprises have risk preferences, including risk seeking and risk avoidance. The distribution of enterprises' risk preferences obeys a uniform distribution. Usually, with limited supervision ability, regulators can only randomly select some enterprises for supervision, and we call this approach a random supervision mechanism. An improved supervision mechanism is proposed considering the enterprises' heterogeneous risk preferences and the constraint of supervisors' limited supervision ability, namely the intelligent supervision mechanism. Under the smart regulation mechanism, the regulator determines the preferences of enterprises by setting up a whitelist and observing historical decisions. Enterprises that are long-term cooperators are whitelisted, and enterprises in the whitelist are not supervised. The regulator randomly selects the supervised enterprises from the ones outside the whitelist. The intelligent supervision mechanism contains two main variables. First, the whitelist length is the maximum number of enterprises that can be accommodated within the whitelist. Second, the observation period is the number of consecutive cooperation strategies adopted by an enterprise that is greater than the observation period requirement to enter the whitelist.
Using numerical simulation, compared with the random supervision mechanism, the effectiveness of the intelligent supervision mechanism is verified. The research results show that different punishment strategies affect the decision-making of enterprises. For example, if faced with the risk of being shut down, enterprises show a stronger willingness to cooperate. In the case of limited supervision ability, the supervision ability and the effect of pollutant abatement are positively correlated. Compared with the random supervision mechanism, the intelligent supervision mechanism can promote enterprises to adopt cooperation strategies and achieve a better effect. The increase in the length of the whitelist can improve the overall cooperation level. If the length of the whitelist is small, an appropriate extension of the inspection period can improve the cooperation level. In contrast, if the length of the whitelist is large, an appropriate shortening of the inspection period can improve the cooperation level.

Key words: limited supervision ability, heterogeneous risk preference, intelligent supervision mechanism, pollutant abatement, complex network, evolutionary game