Given the importance of large-scale assessments to educational policy conversations, it is critical that subpopulation achievement is estimated reliably and with sufficient precision. Despite this importance, biased subpopulation estimates have been found to occur when variables in the conditioning model side of a latent regression model contain measurement error. As such, this article proposes a method to correct for misclassification in the conditioning model by way of the misclassification simulation extrapolation (MC-SIMEX) method. Although the proposed method is computationally intensive, results from a simulation study show that the MC-SIMEX method improves latent regression coefficients and associated subpopulation achievement estimates. The method is demonstrated with PIRLS 2006 data. The importance of collecting high-priority, policy-relevant contextual data from at least two sources is emphasized and practical applications are discussed.