Accident Frequency Model Using Zero Probability Process

Statistics show that the South Korean railroad system ran about 114 million vehicle miles in 2004 and operated through almost 3,600 highway–rail grade crossings nationwide. Highway–rail grade crossing accidents result in severe crashes with trains and other vehicles every year and account for approximately 95% of all grade crossing fatalities. Despite the number of highway–rail grade crossing accidents, quantification of the effect of possible countermeasures has been surprisingly limited because of the absence of data needed to analyze rigorously those factors that affect the frequency of highway–rail grade crossing accidents. This study provides some initial insight into this important problem, including detailed data from a database of roadway geometric and grade crossing elements that were used to analyze highway–rail grade crossing accidents on 100 individual grade crossings in South Korea. Zero-inflated models of accident frequency were developed, and the findings isolated a wide range of factors that significantly influence highway–rail grade crossing accident frequency. The marginal effects of these factors were computed to provide an indication of the effectiveness of potential countermeasures. The findings show promise for the methodological approach undertaken and provide important directions for future railroad research.