Indirect effects of an aid program: how do liquidity injections affect non-eligibles' consumption?

Abstract

Aid programs in developing countries are likely to affect both the treated and the non-treated households living in the targeted areas. Studies that focus on the treatment effecton the treated may fail to capture important spillover effects. We exploit the unique designof an aid program's experimental trial to identify its indirect effect on consumption for non-eligible households living in treated areas. We find that this effect is positive, and that itoccurs through changes in the insurance and credit markets: non-eligible households receivemore transfers, and borrow more when hit by a negative idiosyncratic shock, because of theprogram liquidity injection; thus they can reduce their precautionary savings. We also testfor general equilibrium effects in the local labor and goods markets; we find no significantchanges in labor income and prices, while there is a reduction in earnings from sales ofagricultural products, which are now consumed rather than sold. We show that this classof aid programs has important positive externalities; thus their overall effect is larger thanthe effect on the treated. Our results confirm that a key identifying assumption - that thetreatment has no effect on the non-treated - is likely to be violated in similar policy designs. Aid programs in developing countries are likely to affect both the treated and the non-treated households living in the targeted areas. Studies that focus on the treatment effecton the treated may fail to capture important spillover effects. We exploit the unique designof an aid program's experimental trial to identify its indirect effect on consumption for non-eligible households living in treated areas. We find that this effect is positive, and that itoccurs through changes in the insurance and credit markets: non-eligible households receivemore transfers, and borrow more when hit by a negative idiosyncratic shock, because of theprogram liquidity injection; thus they can reduce their precautionary savings. We also testfor general equilibrium effects in the local labor and goods markets; we find no significantchanges in labor income and prices, while there is a reduction in earnings from sales ofagricultural products, which are now consumed rather than sold. We show that this classof aid programs has important positive externalities; thus their overall effect is larger thanthe effect on the treated. Our results confirm that a key identifying assumption - that thetreatment has no effect on the non-treated - is likely to be violated in similar policy designs.