The increased emphasis on using student growth measures in teacher evaluation has raised questions about how to treat test scores from students with disabilities. This study explores the consequences of three common approaches for treating scores from students with disabilities in statistical approaches to estimating teacher effectiveness: (1) including scores from students with disabilities with no additional disability-related covariates; (2) including scores from students with disability as well as disability-related covariates; and (3) removing these students entirely when estimating teachers' value-added scores. A single state's administrative data over the period of 2007-2009 provided information about all students in grades 3-5 who took the state general summative assessment; vertically-scaled student test scores were available in reading and mathematics. Researchers created both a single-cohort sample and a two-cohort sample for the sample analysis. Several value-added models (VAMs) and median student growth percentiles (SGPs) were used to estimate teacher scores. The following metrics were used to understand how and for whom model choice might matter: Pearson correlation coefficients; average percentile ranks; and percent movement across quintiles. Findings suggest that including test scores from students with disabilities allows more teachers to be evaluated and does not substantially affect teachers' scores. Moreover, including disability-related covariates can allow for fairer evaluations for teachers with many students with disabilities in their class. Tables and figures are appended.