基于双变量有序Probit模型的出租车事故致因分析

Analyzing the causes of taxi accidents using a bivariate ordered Probit model

出租车是我国城市公共交通的重要组成部分,但其事故率一直较高且缺乏深度研究。为有效研究出租车事故的致因机理,2015—2017年于西安、西宁、吉林、汕头四城市开展了随机问卷调查工作,收集了出租车驾驶员的基本信息、工作条件、健康状况、不良驾驶行为频率及过去两年内的事故经历,共获取2391份有效问卷。对各变量进行描述性统计及皮尔森相关性检验后,考虑到事故经历包含财产损失事故数和受伤事故数两类,且均为有序变量,引入双变量有序Probit模型分析驾驶员基本信息、工作条件、健康状况、不良驾驶行为与两类事故发生的关系。结果表明:模型似然比检验结果显著(χ^(2)=4.38,p=0.036),拒绝γ=0的假设,同时,误差项的相关系数ρ=0.145,表明财产损失事故与受伤事故间存在正相关关系,适合使用双变量模型进行建模;Wald检验的结果显著(χ^(2)=60.37,p<0.001)表明各致因因素对两类事故发生的影响不完全相同。驾驶员基本信息中,只有年龄与受伤事故呈显著负相关;工作条件中的4个变量与财产损失事故显著相关,但与受伤事故之间没有显著关联;驾驶员的健康状况和不良驾驶行为频率与两类事故的发生均呈显著正相关。出租车行业管理部门与出租车公司应协同合作,从改善工作条件、提升健康水平及减少不良驾驶行为频率3个角度出发,共同治理出租车事故,促进城市公共交通安全运营。

Taxis play an important role in public transportation in China with a relatively high accident rate,but the causes of taxi accidents have not been deeply researched.Therefore,a survey was conducted in four cities including Xi’an,Xining,Jilin,and Shantou between 2015 and 2017.The survey participants were randomly selected and a total of 2391 valid questionnaires were collected.The questionnaire concerned the information about taxi drivers’demographic data,working conditions,occupational health,risky driving behaviors,as well as involvement in traffic accidents in the past two years.These factors were firstly analyzed by the basic statistic method and Pearson test,and then a bivariate ordered Probit model was used to explore the relationship between taxi drivers’demographic data,working conditions,health status,risky driving behaviors,and accident involvement,considering there were two ordered dependent variables.The model results show that the likelihood ratio test is significant(χ^(2)=4.38,p=0.036)and the correlation coefficient of error termρ=0.145,which indicates that the hypothesisγ=0 is rejected and a positive relation is found between property damage and injury accidents.Therefore,the bivariate ordered Probit model is suitable for this study.The results of the Wald test are significant(χ^(2)=60.37,p<0.001),indicating that the influence of the factors for the two types of accidents are not exactly the same.For drivers’demographic data,only age has a negative correlation with the occurrence of injury accidents.Four variables in working conditions are positively correlated with property damage accidents,but no significant relationship is found between these factors and injury accidents.Besides,drivers’health status and risky driving behavior are positively correlated with both types of the accident.The management departments and taxi companies should work together to manage taxi accidents and promote the safety operation of urban public transport from the perspectives of improving working conditions and health status,as well as reducing risky driving behaviors.