By including district variables at a third level, this study extended previous two-level hierarchical linear modeling (HLM) of multivariate relationships among Texas school-level characteristics and within-school mean Scholastic Assessment Test (SAT) scores and score differentiating factors (L. L. Hargrove, M. X. Mao, and G. Barkanic, 1996; Hargrove and L. T. Mellor, 1994; Hargrove, Mellor, and Mao, 1995). Most consistently, school percentage of students passing all state high school exit examinations, Advanced Placement (AP) examinees with scores of three to five, and students taking AP examinations related positively to school average scores and helped explain substantial total variance in separate three-level HLM models for verbal, mathematics, and total scores, while the district percentage passing all state exit assessments was positively linked to district average score in math and total score models. Similar and other school and district factors described smaller amounts of school and district variance in the average score differentiating effects of student grade point average, rank, advanced coursework, and college English placement/credit plans, further clarifying some construct patterns previously noted. Most importantly, it is noted that variables in these models were predominantly educational rather than demographic in concept, and thus among the more readily alterable variables of the educational context, using the (1989) terminology of J. Oakes. A number of findings were consistent with the previous modeling of Texas scores and have implications for policies regarding the use and reporting of SAT summary score results. (Contains 6 tables and 42 references.) (Author/SLD)