Findings from a study designed to evaluate the Houston Independent School District's (HISD) program for identifying students at risk of dropping out are presented in this report. In the 1988-89 school year HISD employed a scheme of four state-mandated criteria for identifying at-risk students, and this scheme's performance in predicting actual dropouts is assessed. Also examined are four alternative methods for stratifying the identified at-risk populations into groups that can be ranked according to within-group dropout rate. These methods are explored as possible ways to improve dropout prediction accuracy and consequently to identify a priority group for intervention purposes. Results show that in terms of annual dropouts, the state-mandated criteria overidentify students as potential dropouts and fail to identify many potential dropouts. The highest degree of predictive reliability was obtained by individual examination of the 15 combinations of the 4 criteria. The recommendation is made to explore further methodologies for making the best use of the criteria and to expand the list of at-risk criteria. Eight tables are included. (LMI)