Summary: Using examples and nomenclature familiar to social workers, this article provides a brief, nontechnical introduction to hierarchically structured data and multilevel modeling.

Findings: Data structured hierarchically consist of units clustered within higher-level groups, such as individuals nested within families and families nested within counties. Longitudinal observations can also be thought of as having a hierarchical structure, with repeated measures clustered within a given individual. Because lower-level units within a group tend to be systematically more similar to one another than to units from another group, there is often a need to make statistical adjustments for within-group dependence. Hierarchical or multilevel modeling provides a method for researchers to account for possible within-group correlations while also explicitly modeling group-level attributes and membership, acknowledging the potential importance of contextual effects on lower-level outcomes.

Applications: Although hierarchical or multilevel modeling has been discussed (and used) extensively by researchers in other disciplines and fields, it has been more recently adopted as a modeling tool by social work researchers. Because hierarchical modeling is now a commonly employed technique, it is important that those in the field have a basic understanding of these models so that they remain informed consumers of research. This article offers a conceptual overview of hierarchical linear models written for social work practitioners, policy-makers, and other consumers of social welfare research.