We analyze the effect of school expenditure on children's test scores at age 16 by means of an education production model. By using unique register data of English pupils, we exploit the availability of test scores across time, subjects and siblings to control for various sources of input omission and measurement error bias. We overcome one of the main criticisms against the value-added model by proposing a novel method to control for the endogeneity of the lagged test. We find evidence of a positive but small effect of per pupil expenditure on test scores.
education production function test scores school quality