From the Maryland Population Research Center
"Research on neighborhood effects relies heavily on measures from the American Community Survey (ACS), typically treating tract estimates as fixed despite their published sampling uncertainty. We examine what that practice means for substantive conclusions, when it matters, and whether its consequences can be anticipated. Across more than 20,000 census tracts and 160 city-years in two independently assembled databases roughly a decade apart, the results show that the consequences are selective rather than pervasive. Broad socioeconomic composites change little, providing reassurance for many conventional applications. Immigrant concentration and measures of racial and ethnic composition, however, are more affected, particularly when the populations underlying those measures are small. Accounting for ACS uncertainty widens intervals but can also move coefficients. When coefficients move appreciably, they more often move away from zero, suggesting that treating ACS measures as fixed can obscure associations rather than simply make them appear more precise. Moreover, the additional uncertainty is systematically greater where the relevant population is less prevalent, making the measures and places most vulnerable to ACS uncertainty identifiable before an outcome model is fit. The findings clarify when conventional treatment of ACS measures is a reasonable approximation and when greater caution is warranted, with implications extending beyond research on crime to neighborhood and population research that relies on ACS estimates for relatively small population groups." Eric Baumer is Professor of Criminology and Criminal Justice at the University of Maryland.
Read more and register on the MPRC website
Copyright (c) District of Columbia Sociological Society. Contact us: dcsociologicalsociety@gmail.com