A student’s peers are often thought to influence his or her educational outcomes. If so, an unequal distribution of advantaged and disadvantaged students across schools (‘sorting’) in a community will amplify existing inequalities. This paper explores the relationship between the degree of sorting across schools within a community and educational inequality as measured by the variance of standardized high school exam scores within the community. Cross-sectional OLS estimates suggest that the variance of test scores is related to sorting by ethnicity, but not to sorting by income or parental education. We then implement two strategies for addressing endogeneity in the degree of sorting: a standard unobserved effects (first-difference) approach, and a first-difference/instrumental variables approach in which the structure of school choice (number and relative size of schools) is used to construct instruments for the degree of sorting. The results from both approaches indicate that the variance of test scores is related to sorting by home language and parental education, but not to sorting by income. Our results also suggest that reducing sorting would have little effect on inequality of outcomes in the typical Alberta community, but would have substantial effects in the larger cities.