寬恕政策為政府打擊卡特爾不可或缺的重要工具，為了維持市場競爭公平性，各國相繼將其引入法條之中，該政策透過廠商主動揭露涉案行為，使得政府可有效掌握證據將其處置。本文建立兩種賽局模型並分別利用子賽局完全均衡及序列均衡的概念，嘗試討論一般情況下寬恕政策的效率及納入資訊不對稱情形下的政策有效性，並由兩模型推論出：實行寬恕政策且廠商主動申報聯合行為為社會最有效率的均衡、透過政府制定適當的罰鍰區間引導下，主動申報聯合行為的行為可視為一區隔廠商型態的訊號。

The leniency policy plays an indispensable role in thwarting cartel formation. To maintain the fairness of market competition, most countries successively bring this policy into their antitrust legislation. After the enforcement ofthepolicy, the involved firms may have incentive to self-report and provide evidences to the Antitrust Authority. Therefore, the authorities can get enough evidence to convict those firms of being cartel members. In this paper, we develop two kinds of game theoretical model and use the concept of subgame perfect equilibrium and sequential equilibrium to discuss the efficiency of leniency policy in general conditions, and the effectiveness of the policy under the condition of information asymmetry. We show that it is efficient to the society and the authorities when the cartel members self-report under the enforcement of leniency policy. Moreover, by setting up an appropriate fine payment, self-reporting can be a signal for the authorities to segment the type of the involved firms.