The analyses summarized in Chapter 7 use hierarchical linear models (HLM) developed by Raudenbush and Bryk. These models are well suited to the combination of settlement- and individual-level data we analyze from the ADS survey. HLM models can account for the nonindependence of observations within settlements and allow us to simultaneously estimate within- and between-settlement equations. Our particular theoretical interest is in the collective or settlement-level variance in such variables as racial epithets as our key indicator of collective racial motivation and intent, but our analytic approach also incorporates individual-level variation in this and other factors. Tables A.1 and A.2 summarize descriptive characteristics from the twenty-two settlements introduced in this book.

The HLM models that form the background for the analysis in Chapter 7 involve the estimation of within- and between-settlement equations.