认罪认罚量刑从宽实效研究——基于故意伤害罪轻罪的数据解读

A Study on Impact of Leniency on Sentencing for Suspects Admitting Guilt&Accepting Punishments—Based on Data of Misdemeanor Crime of Intentional Injury

我国认罪认罚从宽制度设计具有明显的公正优先价值导向。通过构建非线性量刑模型分析177161份故意伤害罪致人轻伤的初审判决书发现,司法实践中表现出明显的“从宽”克制倾向:认罪认罚对轻罪量刑的作用倾向于在“底线”游走,接近量刑情节“坦白”;传统量刑情节仍在习惯性适用,认罪认罚发挥实效的空间不足。在法定主义背景下推动更高层次上公正与效率的统一,可以考虑,一方面建立认罪认罚分级分类制度,适当加大对轻罪的从宽幅度;另一方面扩大认罪认罚从宽幅度的法定空间,在立法上赋予认罪认罚“减轻”情节地位,破解量刑折扣零和博弈的困局。

Does suspects admitting guilt&accepting punishments contribute to lenient punishment?This is the key point of a new mechanism—Leniency for Suspects Admitting Guilt&Accepting Punishments.Through analyzing the first-instance judgments of 177,161 intentional assaults causing minor injuries,we find that the effect of the mechanism on the sentence of misdemeanors is close to the“confession”.Further inspections manifest that the implementation of the mechanism did not change the effects of traditional sentencing circumstances for lighter punishment significantly.Research conclusions indicate that the implementation of the mechanism shows a tendency of justice priority.At the same time,compared with the traditional sentencing circumstances for lighter punishment,the application of the new mechanism is conservative for misdemeanors,and sentences are often determined in the lower part of the sentencing interval,which implies that the leniency range or the application of probation in misdemeanors can be appropriately increased,so as to achieve the unity of justice and efficiency.In terms of analysis methods,a non-linear sentence model that highly simulates judicial practice is constructed.Considering the spatial consistency and correlation effect over time of criminal judgments,we use the Recursive Least Squares algorithm to estimate the actual impact and possible changes of the mechanism,and use Error Probability Inequality to test the reliability of estimates under limited data.Our methods have two characteristics:First,the independent and identical distributed assumption of noises is not required,which is more suitable for the nature of social data such as judicial judgments.Secondly,it is possible to observe the trends of main characteristics over time.