Charter school competition can only work as a policy to improve public education if schools that do not contribute to this goal are allowed to fail. We estimate survival regressions to assess the effects of various factors on the probability of school failure. We find that students' test scores are the most important determinant of survival: a one standard deviation increase reduces the probability of failure by 76%. Higher expenditures per student and a longer wait list result in smaller, but significant, reductions. Enrollment, average performance in the host district, and student demographics do not significantly affect school survival. (JEL I21, H75)