This study aims at investigating the association between pedestrian crashes and various roadway, socio-economic, and land-use features.

Bayesian Conditional Autoregressive models considering spatial correlation among Traffic Analysis Zones (TAZs) were established.

Seven spatial weight features were developed and compared to characterize the spatial correlations among TAZs.

Geometric centroid-distance-order, which was introduced in macro-level safety analysis for the first time, performed best.

More pedestrian crashes were associated with larger population, higher road density, and medium land use intensity.