There is increasing evidence for the adverse effects of child maltreatment on academic performance; however, most of these studies used selective samples and did not account for potential confounding or mediating factors. We examined the relationship between child physical abuse (PA; severe and non-severe) and sexual abuse (SA) and educational attainment (years of education, failure to graduate from high school) with a Canadian community sample. We used data from the Ontario Child Health Study (N = 1,893), a province-wide longitudinal survey. Potential confounding variables (family socio-demographic and parental capacity) and child-level characteristics were assessed in 1983, and child abuse was determined in 2000-2001 based on retrospective self-report. Results showed that PA and SA were associated with several factors indicative of social disadvantage in childhood. Multilevel regression analyses for years of education revealed a significant estimate for severe PA based on the unadjusted model (-0.60 years, 95% CI = [-0.45, -0.76]); estimates for non-severe PA (0.05 years, CI = [-0.15, 0.26]) and SA (-0.25 years, CI = [-0.09, -0.42]) were not significant. In the adjusted full model, the only association to reach significance was between severe PA and reduced years of education (-0.31 years, CI = [-0.18, -0.44]). Multilevel regression analyses for failure to graduate from high school showed significant unadjusted estimates for severe PA (OR = 1.77, 95% CI = [1.21, 2.58]) and non-severe PA (OR = 1.61, CI = [1.01, 2.57]); SA was not associated with this outcome (OR = 1.40, CI = [0.94, 2.07]). In the adjusted full models, there were no significant associations between child abuse variables and failure to graduate. The magnitude of effect of PA on both outcomes was reduced largely by child individual characteristics. These findings generally support earlier research, indicating the adverse effects of child maltreatment on educational attainment. Of particular note, severe PA was associated with reduced years of education after accounting for a comprehensive set of potential confounding variables and child characteristics.