Evidence on the effectiveness of school inputs remains inconclusive, partly due to the challenge of identification as families sort themselves into school districts and resources are potentially allocated to compensate (or reinforce) differences in pupil abilities. Using variation in school resources induced by the location of waterfalls in Norway, we examine the effect of school expenditures on pupil performance at age 16. Higher school expenditures, triggered by higher revenues from local taxes on hydropower plants, have a significantly positive effect on pupil performance. This positive IV estimate contrasts a zero effect based on least squares. A downward biased estimate using a standard cross section estimator is expected in a context of compensating resource allocation across educational units. (Contains 8 tables and 3 figures.)