Apparently sophisticated school performance measures have been used to claim that giving schools autonomy from local government control improves pupil exam performance. This paper explores the extent to which inferring causality between autonomy and pupil achievement is reasonable given that pupils are not randomly assigned to schools and schools do not randomly acquire autonomous status. Rich administrative data and the Longitudinal Survey of Young People in England are used to evaluate whether school performance measures are confounded by pupil characteristics that explain both the chances of attending an autonomous school and academic achievement. The assignment of grant-maintained (and thus now foundation) status through a vote of parents is used to compare school that just did, and just did not, gain autonomy over a decade ago. These alternative estimation strategies suggest there is little evidence that foundation status casually yields superior school performance.