Abstract
We derive four facts from experiments on the spillover effects of cash and in-kind transfer programs onto non-recipients: (1) average spillovers on consumption and asset value are negligible; (2) negative psychological spillovers emerge in the longer term among households randomly excluded from transfers by within-village lotteries; (3) these spillovers are concentrated in consumption-poor households; and (4) they do not appear when the selection of transfer recipients is eligibility-based. These findings are consistent with household preferences against within-village randomization of transfers, and point to additional benefits of cluster-level randomized designs over household-level ones.