Both physical and population-compositional characteristics were used to develop structural models of property crime in a sample of over 900 U.S. cities. The primary data source was the 1977 County and City Data File, which includes a wide range of census data on 910 U.S. cities with a 1975 population of at least 25,000. Population characteristics examined were ethnicity, age, income, percent below poverty level, employment diversity, education, citizenship, and sex. The physical characteristics of cities encompassed density, households with more than one person per room, city size, region, division of labor, commercial differentiation, and housing factors. Bivariate analysis was used for data cleaning and reduction, and multiple regression analysis was used for model construction and validation. The explanatory importance of environmental variables and population-characteristics variables differed by type of property crime, place size, and, for robbery, region of the country. The study confirms the necessity of using both personal and environmental variables to explain crime rates. 5 tables and 36 footnotes.