Spatially Disaggregate Panel Models of Crash and Injury Counts: The Effect of Speed Limit and Design

This work statistically examines the impacts of the 1996 speed limit changes based on over 6,000 Washington State highway segments. Fixed-effects and random-effects Poisson and negative binomial regression specifications are employed to estimate six crash count measures during the 1993-1996 period. The modeled crash measures are the numbers of fatalities, injuries, crashes, fatal crashes, injury crashes, and property-damage-only (PDO) crashes. The average segment length is just 0.131 mile, permitting tight control of geometric characteristics, such as curvature and grade, as well as vehicle miles traveled (VMT). A 10 mph speed limit increase, typical of U.S. state policies pursued in the mid-1980s (on rural interstates) and in the mid-1990s, is estimated to increase fatalities and injuries by 78 and 24 percent, respectively, assuming other factors remain constant. Speed limit effects on total crashes (and property-damage-only crashes), however, are predicted to be slightly negative, suggesting that crashes become more severe, but not more common. As expected, tighter horizontal curves, fewer lanes, and lower traffic volumes (per lane) increase fatal and injurious crash rates