The effects of road geometrics and traffic regulations on driver-preferred speeds in northern Italy. An exploratory analysis

• Free-flow speed data collection on urban arterials and collector roads.
• Panel data models on hierarchical data structure.
• Calibration of random effect model for the estimation of speed distribution.
• Selection of significant variables according to the Bayesian Information Criterion.
• Speed differential among lanes on urban multilane roads.

Speeds are affected by several variables such as driver characteristics, vehicle performance, road geometrics, environmental conditions and driving regulations. It is therefore important to study the relationships between speed and such variables in order to facilitate conscious speed management on existing and planned roads, and to induce drivers to select a speed consistent with the posted limit. This relationship is of great interest to those who wish to achieve roadway functionality and improve overall road safety.A small number of studies have focused on this objective; however, few of them concern urban roads and they are limited to specific road types and recently built-up areas. These studies often refer to the 85th percentile of the speed distribution and are relevant to locations which are homogeneous in terms of geometry, environment, driving regulations and vehicle type.This paper presents results obtained from a study carried out on urban arterials and collectors characterized by dissimilar geometric features which facilitated the inclusion of a fully representative range of variables. A general model able to predict operating speed for a generic percentile was calibrated using three different strategies: (a) a simple multiple regression analysis in which the variables were selected using the Bayesian Information Criterion (BIC); (b) the analysis of covariance method including random effects on the same set of variables as in (a); and, finally, (c) the analysis of covariance method with random effects and a new selection of variables (again using BIC). The analysis shows a dramatic variation in results depending on the method selected. In particular, when random effects are considered, almost all the variables are found to be statistically significant.