In randomized control trials (RCTs) of educational interventions, there is a growing literature on impact estimation methods to adjust for missing student outcome data using such methods as multiple imputation, the construction of nonresponse weights, casewise deletion, and maximum likelihood methods (see, for example, Allison, 2002; Graham, 2009; Peugh & Enders, 2004; Puma, Olsen, Bell & Price, 2009; Schafer & Graham, 2002). Much less attention, however, has been devoted in education RCTs to developing statistical methods to adjust for the systematic misreporting of student outcome data for those with nonmissing data. Without appropriate adjustments, misreporting could lead to biased impact estimates, which could be exacerbated if the intervention leads to treatment-control differences in misreporting rates and the composition of students with misreported data. Misreporting could also affect the variance of the estimated impacts, and hence, significance levels from statistical hypothesis tests of intervention effects. This article develops a parametric statistical framework to test and adjust for the misreporting of binary outcomes in the estimation of average treatment effects (ATEs) for school-based RCTs. The author considers a realistic scenario where it is assumed that binary outcomes on sensitive topics can be misreported for students with truly undesirable outcomes, but not for those with truly desirable outcomes. A latent index framework is employed where misreporting and binary outcome decision processes are modeled using available baseline data and normality assumptions about model error terms. This approach yields a "double hurdle" random effects probit model that can be estimated separately for treatments and controls. The article discusses quasi-Newton ML methods for obtaining consistent estimates of the unobserved misreporting rates, the ATEs on the considered binary outcomes, and standard errors of the estimates that are not contaminated by misreporting. The article also discusses how the approach can be applied to continuous outcomes and to nonclustered, student-level RCT designs