A comment on "Seeing racial avoidance on New York City streets"

Dietrich and Sands (2023) used New York City traffic camera footage to experimentally examine the effect of a pair of racialized confederate bystanders on the distance pedestrians maintained from those bystanders as they passed them on the sidewalk. Across their block-randomized experimental conditions, the authors found that pedestrians deviated by around 4 inches on average, maintaining larger distances from Black individuals as opposed to White individuals. Their point estimate was significant at the 5% level. In this conceptual replication we use data from a new context and new data collection technique, using 3D videos and computer vision models to estimate the effect of minoritized bystanders on pedestrian distance in Berlin, Germany. Despite these differences to the original study, we directionally reproduce Dietrich and Sands' main claim. Writ large, pedestrians maintain a larger distance from Muslim bystanders of Middle Eastern descent wearing jalabiyas (a religious Muslim garment), as opposed to White bystanders wearing jeans and t-shirts, in the German context. Using bootstrapped sub-samples from our data, our point estimates range from 0.39 inches (1 cm) to 9.84 inches (25 cm), with an average difference in distance of about 5.9 inches (15 cm). This roughly mirrors Dietrich and Sands' finding of a 4-inch difference. However, our replication finds greater heterogeneity across locations, where different types of areas and different streets show opposite patterns. Our pooled results are only statistically significant at the 5% level when using a wild block bootstrap but are not significant when using clustered standard errors.