Objective. Using Poisson-based negative binomial regression, we estimate the effect of neighborhood factors on homicides in two cities (San Antonio, Texas and San Diego, California) that have large Mexican-origin populations.

Methods. Three independent data sources (official homicide police reports, medical examiner records, and the U.S. Census) are used to construct the dependent homicide, and independent neighborhood, variables. Census tracts represent the unit of analysis, which serve as a proxy for neighborhoods. Given the spatial nature of the data, spatial estimation procedures were also modeled.

Results. Spatial proximity to violence, neighborhood disadvantage, and affluence (in San Antonio) consistently buffered homicide across neighborhoods, even in heavily populated Latino neighborhoods.

Conclusions. Spatial embeddedness and neighborhood characteristics are important for improving our understanding about ethnic neighborhood variations in levels of violence. Comparative approaches across places, namely, Latino-dominated cities, can yield considerable insight into how the local context intersects race/ethnicity and violent crime.