Differences

National studies of students, and studies that compare institutions, have identified many predictors of students ' graduation rates, including socioeconomic status and admission selectivity. When there predictive variables are applied to data from an individual school, it may be found that they have less predictive power. This paper presents a theoretical explanation of why this pattern might be expected, and uses the Statistical Analysis System (SAS) to estimate a hierarchical linear model of the size of the interinstitutional and intrainstitutional effects of admission selectivity on graduation rates for a 7-year panel of data on several colleges and universities. This statistical phenomenon can be interpreted in a variety of ways: (1) measuring "within groups " effects versus "between groups " effects; (2) analyzing "time-series " data versus "cross-sectional" data; (3) regression analysis with unobserved or omitted variables that cause the error term to be correlated with the explanatory variables; (4) intrainstitutional versus interinstitutional or national data. In short