Looking at the decentralized provision of public education in a middle income country, this paper estimates the impact of local autonomy on service quality, finding large heterogeneity in the effect across different levels of local development. Colombian municipalities were assigned to administer their public education service autonomously solely on the basis of whether they exceeded the 100 thousand inhabitants threshold. Exploiting this discontinuity, I estimate the impact that autonomy has had on student test scores across municipalities, using a regression discontinuity design and fixed-effects regression on a discontinuity sample. I find a test score gap arising between autonomous municipalities in the top quartile and those in the bottom quartile of the development range, in a trend that reinforces over time. From analysis of detailed municipal balance sheet data, I show that the autonomous high-developed municipalities invest in education more than the ad hoc transfers they receive, supplementing these with own financial resources.
Indicators of municipal administration quality also show significant differences between the two groups of cities, helping to explain the education outcome patterns.