Exploring the Spatial Dependence and Selection Bias of Double Parking Citations Data

Abstract

Parking violation citations, often used to identify factors contributing to parking violation behavior, offer one of the most valuable datasets for traffic operation research. However, little has been done to examine their spatial dependence caused by location-specific differences in features such as traffic, land use, etc., and potential selection biases resulting from different effects of traffic enforcement. This study leveraged extensive data on double parking citations in Manhattan, New York City, in 2015, along with other relevant datasets including land use, transportation, and sociodemographic features. Moran’s I statistics confirmed that double parking tickets were spatially correlated so that spatial lag and spatial error models were proposed to account for the spatial dependence of parking tickets to avoid biased estimates. To investigate whether selection bias exists in issuing tickets, we estimated the effects of parking ticket density and police precinct distance, when controlling for variables such as commercial area, truck activity, taxi demand, population, hotels, and restaurants. Parking ticket density and police precinct distance were used as indicators of the enforcement levels and coverage and were found to be statistically significant. This indicated the existence of selection bias due to heterogeneity in enforcement levels or coverage across different regions. Moreover, patrol patterns of traffic enforcement officers revealed that the majority had less than three daily patterns. These findings can assist with proper usage of the citation data by recommending that researchers and agencies consider spatial dependence as well as selection bias, and provide insights for parking violation management strategies.