Search Results

Many social phenomena are observed repeatedly for an individual or unit nested within a contextual level. The problem of modeling unobserved unit heterogeneity, which is nested in unobserved context heterogeneity, is referred to as a nested heterogeneity problem. To address this problem, this article develops a systematic method, where the problem is formally specified and issues such as correct model specifications, appropriate estimators, tests and recovery of each level of heterogeneity, and assessment of the relative importance of each level of hetergeneity are discussed. We proceed from a simple case where there exists no context-specific and unit-specific observed variables to a more inclusive case with both context-specific and unit-specific variables. We then extend the nested heterogeneity model to allow the context-specific variables to vary across time, where we develop a new estimator using the difference-in-differences method. We illustrate our methods using an empirical example of child and family heterogeneity in a study of children's behavior problems.