We started with a baseline of municipalities (defined as “places” in most states, but also including “county subdivisions” in others). Next, we obtained lists of municipalities that have proposed restrictive ordinances and regulations from various sources, including the American Civil Liberties Union, Latino Justice PRLDEF, the Fair Immigration Reform Movement, the National Immigration Law Center, and the Migration Policy Institute. We then validated these lists by making phone calls to jurisdictions noted as considering or passing ordinances, as well as by monitoring news stories on local ordinances through December 2011. Still, the data on municipal ordinances became less reliable after 2007, due to a sharp decline in newspaper reports of new municipal ordinances and no further tracking of municipal legislation by national advocacy groups. We merged information on the proposal and passage of ordinances with demographic data from the 2000 Census and the 2005–2009 American Community Survey. Our municipal analysis is presented in Table A.1.

Multivariate Regression Analyses

Model I is a standard logistic regression, which is appropriate for regressions where outcomes take on binary values of 1 and 0. Logistic regressions also have the benefit of allowing coefficients to be converted easily to odds ratios. Model II is a rare-events logistic regression, using the Relogit statistical software package from http://gking.harvard.edu/relogit. The model weights the sample such that the weighted ratio of 0s and 1s in the sample matches that of the population, a process designed to correct for selection on the dependent variable.