Program Evaluation Using Spatial Instruments: Evidence from Indonesia by Mark Pitt and Nidhiya Menon :: NEUDC 2007 Papers :: Northeast Universities Development Consortium Conference :: Center for International Development at Harvard University (CID)

This paper estimates the effects of non-randomly distributed government programs in Indonesia using a new instrumental variables technique. Based on the maintained assumptions that governments are responsive to regional attributes and not to individual characteristics, and that conditional on the presence of programs, individual outcomes are uncorrelated with mean characteristics, regional level averages of individual attributes may be used as instruments to correct for the endogeneity of government programs. Three categories of spatial instruments are constructed; averages of exogenous variables over districts, averages over neighboring districts, and averages over provinces. An application of these instruments to cross-sectional data from Indonesia demonstrates that public programs such as grade schools, junior schools, senior schools, maternal/child health clinics, health clinics associated with the Puskesmas program, and family planning clinics, have significant effects. Moreover, a comparison of these instrumental variables estimates with “naive estimates” that treat programs exogenously clearly highlights the extent of bias.