Purpose: We assess whether school start times predict the engagement and achievement outcomes of middle grades students. Our focus on middle grades is important because biological changes in sleep often begin when adolescents are in middle school and because middle school is a time when more students struggle academically. Research Methods/Approach: We use 6 years (2011-12 through 2016-17) of statewide administrative data from North Carolina to assess how school start times predict the school attendance, disciplinary records, and test scores of middle grades (6-8) students. We estimate a range of models--school fixed effect, student fixed effect, propensity score--and include a rich set of covariates to isolate the impact of start times. Findings: Our school engagement results are somewhat inconsistent but suggest that later start times predict a reduction in absences and suspensions. Later start times consistently predict higher test scores in mathematics and reading. Subgroup analyses return mixed results regarding which students benefit more from later middle school start times. Implications: Our results emphasize the broader connections between health and academic outcomes and indicate that policy makers should delay start times for middle grades students. States can instigate start time changes by incentivizing districts to delay or requiring that districts delay start times. Districts can independently delay their start times. In doing so, it is important that district officials take time to build support for the policy change and think comprehensively about the start times of all--elementary, middle, and high-district schools.