In this study, the relationships between land use, population size, employment by sector, economic output and motor vehicle accidents are explored. Using comprehensive data from the largest county in Honolulu, the relationships are modeled using a uniform 0.1 square mile grid structure and various linear regression models. This is an advantage over other approaches which have typically used unevenly sized and shaped traffic analysis zones (TAZs). Positive, statistically significant relationships between population size, job counts, economic output, and accidents are identified. After sorting through some of the general effects, a Negative Binomial model is used to look at the absolute and relative effects of these factors on the number of pedestrian, bicycle, vehicle-to-vehicle and total accidents. By using a multivariate model, the different effects can be compared and the specific nature of the relationships between types of activities can be identified. While there is, in general, a significant relationship between all of these values, the effects are more pronounced with vehicular crashes than those involving pedestrians or bicyclists. In addition to the general effects, the influences of employment, economic development, and activities on the level and type of accidents are investigated. Some of the challenges associated with modeling these relationships are described as are implications for traffic safety research. The paper adds to the growing volume of traffic safety research integrating Negative Binomial regression models and GIS (geographic information systems).
Abstract