Improving Moran’s I to Identify Hot Spots in Traffic Safety

corresponding author This chapter aims at identifying accident hot spots by means of a local indicator of spatial association (LISA), more in particular Moran’s I. A straightforward use of this LISA is impossible, since it is not tailor-made for applications in traffic safety. First of all, road accidents occur on a network, so Moran’s I needs to be adapted to account for this. Moreover, its regular distributional properties are not valid under the circumstances of Poisson distributed count data, as is the case for accidents. Therefore, a Monte Carlo simulation procedure is set up to determine the correct distribution of the indicator under study, though this can be generalized to any kind of LISA. Moran’s I will be adapted in such a way, that it can over-come all the previously stated problems. Results are presented on highways in a province in Flanders and in a city environment. They indicate that an incorrect use of the underlying distribution would lead to false results. Next to this, the impact of the weight function is thoroughly investigated and compared in both settings. The obtained results may have a large impact for policy makers, as money could be allocated in a completely wrong way when an unadjusted LISA is used.