Forecasting the Quality of Service of Bogota’s Sidewalks from Pedestrian Perceptions: An Ordered Probit MIMIC Approach

Abstract

A variety of different pedestrian performance or service indicators (PPSI), such as the pedestrian level of service (PLOS), or quality of service (QoS), have been developed to evaluate pedestrian infrastructure. Although modeling approaches vary (e.g., ordinal least squares, scoring system), the great majority use on-site measurable attributes, such as sidewalk width or adjacent traffic, to estimate the PPSI. However, most of these models have been developed without jointly considering objective and subjective variables and their interactions. This study had two objectives: (i) to develop a model that simultaneously uses objective and subjective variables to estimate the pedestrian perception of sidewalks’ QoS in Bogota, Colombia, and (ii) to identify the interactions between objective variables and pedestrian perceptions of sidewalk attributes. To do so, data was gathered from 1056 users of 30 sidewalks in the city and an Ordered Probit Multiple Indicator and Multiple Cause model was estimated and validated using match score, error distribution, and chi squared test. Using the model, it was possible to correctly forecast the perceived QoS in 26 of the 30 sidewalks, considering the interaction between users’ characteristics and on-site sidewalk measured attributes with four latent variables (sidewalk characteristics, surrounding, discomfort, and externalities) based on pedestrian perceptions. We also proposed guidelines that provide decision makers with the tools to identify which sidewalk attributes actually influence pedestrian perception of QoS.