Hypothetically, a compact city form decreases Vehicle Miles Travels (VMT) and the reduced VMT indicates lower exposure of pedestrian crash. The evidences of empirical studies, however, have shown a mixed findings. We speculate that the mismatches between theory and empirical evidences may be caused by the spatial scale of study areas, focusing on spatial distribution feature of job and housing density. To explore the relationship between compactness of the built environment and pedestrian crash, we conducted regression analysis at both city and neighborhood scale. The results showed that the association between compactness and pedestrian crashes varies by the spatial scale of study areas. At the city scale, the compactness does not have a significant association with pedestrian safety, whereas, at neighborhood scale, density of employment was associated with greater risk of pedestrian crash while population density was negatively associated with pedestrian crashes. The finding implies that the spatial distribution of housing and jobs do an important role in pedestrian safety. In addition, we found larger than moderate level of confounding effect of demographic features in explaining pedestrian crash risk.