The Temporal Instability Of Factors Affecting Crash Injury Severity In Truck-car Crashes

With the development of society and economy,cargo transportation is of great significance to promote the circulation of goods and economic development.However,while providing efficient cargo transportation,trucks often cause more serious accident due to their high quality,long body,visual blind spots on the rear side,large inertia,and long braking distance.So far,domestic and foreign research scholars have used a lot of parameters models and nonparameters models to study the factors which affecting the severity of crashes involving truck.However,due to the high heterogeneity and time instability of accident data,ignoring the above problems can lead to biased parameter estimates and even wrong conclusions.Therefore,it is necessary to pay attention to the time instability and heterogeneity of accident data in order to provide more reasonable conclusions and inferences.The object of this paper is truck-car crashes.Firstly,a total of 19,312 accident data are extracted from the British STATS19 accident database over four-year period(January 1,2015 to December 31,2018),and a wide range of 50 parameters were considered including truck and car driver characteristics,truck and car vehicle characteristics,road characteristics,environmental characteristics and time and space characteristics.Secondly,two likelihood ratio tests were conducted to assess the transferability of model estimation results from year to year.And the accident data were divided into 2015-2016,2017,2018 by year with a 95% confidence interval.Based on the accident data of three years,an ordered Logit model,a random parameter Logit model and a random parameter ordered Logit model were respectively established considering three possible crash injury-severity outcomes(slight injury,severe injury,and fatal).The information criterion and Mc Fadden pseudo-R2 evaluation index show that the random parameter ordered Logit model has the best fitting results.Finally,marginal effects of explanatory variables were also calculated to investigate the stability of individual parameter estimates on injury-severity probabilities across time-period combinations.There are 34 factors significantly affecting severity of truck-involved crashes in the ordered Logit model and the random parameter ordered Logit model,including the gender of the truck driver,the age of the truck and so on.However,the 28 factor variables are significant variables in the random parameter Logit model.It should be noted that male drivers of trucks and cars and overtaking behavior of trucks and cars have opposite effects on the severity of the crashes.Meanwhile,the results show instability in the effect of factors from year to year.However,there were 10 variables that exhibited relatively stable effects on injury-severity probabilities including male truck driver,truck turning,lane change,etc,and other factor variables are only significant in the accident model of the special year.Finally,both the random parameter Logit model and the random parameter ordered Logit model can capture potential unobserved heterogeneity.The study found that 11 factor variables are random parameters including male truck drivers,elderly truck drivers,truck overtaking,etc.The findings of this study should be useful to learn the factors that affect the severity of truck-car crashes from the aspects of truck and car.On the other hand,this study could provide a theoretical basis for decision makers and trucking companies to better formulate relevant traffic safety policies.