����vlog����

Urban change involves transformations in the physical appearance and the social composition of neighborhoods. Yet, the relationship between the physical and social components of urban change is not well understood due to the lack of comprehensive measures of neighborhood appearance. Here, we introduce a computer vision method to quantify change in physical appearance of streetscapes and generate a dataset of physical change for five large American cities. We combine this dataset with socioeconomic indicators to explore whether demographic and economic changes precede, follow, or co-occur with changes in physical appearance. We find that the strongest predictors of improvement in a neighborhood’s physical appearance are population density and share of college-educated adults. Other socioeconomic characteristics, like median income, share of vacant homes, and monthly rent, do not predict improvement in physical appearance. We also find that neighborhood appearances converge to the initial appearances of bordering areas, supporting the Burgess “invasion” theory. In addition, physical appearance is more likely to improve in neighborhoods proximal to the central business district. Finally, we find modest support for “tipping” and “filtering” theories of urban change.