作者：张吉喜
作者介绍：西南政法大学诉讼法与司法改革研究中心教授
出版日期：2023年8月9日
内容提要 在逮捕社会危险性条件的评估方法中，统计学方法优于经验方法，分类评估方法优于综合评估方法。犯罪嫌疑人逃跑是逮捕社会危险性条件的具体情形之一，运用统计学方法对其风险进行评估是体现不同评估方法优势的选择。以取保候审期间逃跑的2133名犯罪嫌疑人和遵守取保候审义务的2006名犯罪嫌疑人为样本，通过统计学分析，发现户籍地、受教育程度、就业状况、年龄、前科、盗窃罪、诈骗类犯罪和可能判处的刑罚等是对犯罪嫌疑人的逃跑风险有显著性统计学意义的影响因素。在此基础上构建的犯罪嫌疑人逃跑风险的评估模型具有较好的预测能力。将该模型转化为计算机运算程序，可以作为犯罪嫌疑人逃跑风险的评估工具。该评估工具不能否定检察官的裁量权，在使用该评估工具时亦需要保障辩方的知情权和提出意见权。
关键词 逮捕 取保候审 逃跑 社会危险性 风险评估
Abstract: Among the methods for assessing pretrial risk, statistical assessment is better than empirical assessment, and classified assessment is better than comprehensive assessment. The flight of suspects is one pretrial risk. Using statistical methods to assess it is the best choice among different methods of assessment. The 2133 suspects who escaped during the bail pending trial and the 2006 suspects who abided by the bail pending trial were selected as samples. Through statistical analysis, we found that registered residence, education level, Abstract: Among the methods for assessing pretrial risk, statistical assessment is better than empirical assessment, and classified assessment is better than comprehensive assessment. The flight of suspects is one pretrial risk. Using statistical methods to assess it is the best choice among different methods of assessment. The 2133 suspects who escaped during the bail pending trial and the 2006 suspects who abided by the bail pending trial were selected as samples. Through statistical analysis, we found that registered residence, education level, employment status, age, criminal record, theft, fraud, and punishment were significant factors affecting the flight risk of suspects. On this basis, the model for assessing the flight risk of suspects is constructed, which has good predictive ability. The model can be transformed into a computer operation program, which can be used as a tool for assessing the flight risk of suspects. The tool cannot deny the discretion of procurators. The defense’s right to know and put forward opinions need to be guaranteed when using the tool.