In embargo until 1 Sep 2020
A version will be available for download on 1 Sep 2020

This paper investigates the effects of India's rural roads programme (PMGSY) on morbidity, using data on 279 households drawn from 30 villages in upland Orissa. The households were surveyed in 2010 and 2013, yielding an unbalanced panel of 1580 individuals, 1076 of whom were present in both years. Ten of the villages had received a direct all-weather road connection between 2004 and 2013. To estimate the programme's effects, the villages were treated as units within the whole network of roads and medical facilities. The provision of a connection, whether direct or in the neighbourhood, was associated with the following reductions in morbidity (random-effects estimates): for each km of unpaved track so replaced, an inhabitant's probability of suffering a bout of incapacitating illness fell by 4.5 percentage points and the expected duration of such illness by 0.53 days. The village fixed-effects estimates are qualitatively the same, but smaller and less precise. A simple indicator variable for the presence or absence of such a connection yields qualitatively similar estimates, but with very large standard errors, which confirms the importance of employing fine measures of the network regressors.