Assessing technical inefficiency of bus transit with adjustment of accidents

In the light of occasional involvement of traffic accidents and of seldom costless to dispose such undesirable accidents, conventional data envelopment analysis (DEA) methods, which assume that traffic accidents are freely disposable in bus transit efficiency measurement, may not be appropriate. To rectify this shortcoming, this paper attempts to construct directional output distance function models to assess the technical inefficiency of bus transit with adjustment of four types of prevailing accidents: fatalities, heavy injuries, light injuries, and property losses. Three input factors--capital, labor, fuel, one desirable output--vehicle-kilometer, and one undesirable output--accident index converted from four types of accidents are employed. The empirical results from ten Taipei bus companies over 2001-2006 indicate that the inefficiency scores and the number of efficient DMUs adjusted by accidents have significantly differed from those measured without considering the accidents. It is also noted that the publicly-owned bus operator has increased its efficiency after privatization effective January 1, 2004. Based on the empirical results, it is interesting to find that some transit operators are scale inefficient--with decreasing returns to scale, implying that these companies may consider downsizing their fleet size to the range of 300 to 400 vehicles so as to achieve scale efficiency.