Charter schools have become a very popular instrument for reforming public schools, because they expand choices, facilitate local innovation, and provide incentives for the regular public schools while remaining under public control. Despite their conceptual appeal, evaluating their performance has been hindered by the selective nature of their student populations. This paper investigates the quality of charter schools in Texas in terms of mathematics and reading achievement and finds that, after an initial start-up period, average school quality in the charter sector is not significantly different from that in regular public schools. Perhaps most important, the parental decision to exit a charter school is much more sensitive to education quality than the decision to exit a regular public school, consistent with the notion that the introduction of charter schools substantially reduces the transactions costs of switching schools. Low income charter school families are, however, less sensitive to school quality than higher income families.