This study of graduation outcomes in Baltimore uses multivariate analysis of longitudinal student cohort data to examine the impact of factors identified in previous research as early warning indicators of a dropout outcome. Student cohort files were constructed from longitudinal administrative data (following all first-time 2004-2005 and 2005-2006 9th graders forward in time until their on-time graduation year and 1 year past). Sequentially estimated logistic regression hierarchical linear modeling models indicated the strongest predictors of graduation were 9th-grade attendance and course failure, although gender was still significant. Multinomial logistic regression models were used to analyze the relationship between the 4 categories of college enrollment outcomes (enrollment in a 4-year college, enrollment in a 2-year college, graduation with no college enrollment, and nongraduation) and student-level predictor variables, including grade point average (GPA) and 8th-grade test scores. Results suggest that equipping schools to implement interventions to address chronic absenteeism and course failure in 9th grade is a crucial strategy for increasing both high school graduation and college enrollment. (Contains 8 tables and 6 footnotes.)