If an effective school is defined as a school that causes student improvement on a number of important educational outcomes, the problem of identifying effective schools becomes one of establishing legitimate predictions of student performance and comparing those predictions to actual student or school outcomes. In attempting to identify effective schools, it is essential that methodologies be applied that are value-added and that account for student-entry characteristics as well as student growth on the outcomes of interest. This paper examines the similarities and differences between the student-based regression-based model used by the Dallas Independent School District (DISD), Texas, for the past 3 years and a number of models using two-level hierarchical linear modeling. Specifically, the study sought to determine the differences in results produced by a regression model using first- and second-order interactions and a two-level hierarchical linear model (HLM). The results suggest that student-based models that account for first- and second-order interactions among variables and two-level hierarchical linear models produce very similar school ranks. Ranking schools purely on the basis of absolute test scores is clearly biased against schools with large populations of poor students, minority students, and black students. It is recommended that two-level HLMs be used to rank schools in combination with a first regression and prediction stage that accounts for the variance associated with student-level background variables. It is important that both student-level and school-level contextual information be included in models for identifying effective schools. (Contains 21 references and 14 tables.) (LMI)