At scale implementation and the perils of fragmentation

Governments frequently rely on multiple actors to deliver technology-adoption policies at scale, often presuming that layering low-cost interventions, such as information, onto existing programs is at least weakly beneficial.We show that this need not hold.Using a cluster-randomized trial of private-sector sanitation microloans and an NGO-led information campaign in rural India, combined with quasi-random variation in government policy implementation, we find that information can reduce adoption when delivery is fragmented across actors.The mechanism operates through household expectations about complementarities that fail to materialize.Our findings offer a new explanation for why interventions may lose effectiveness at scale.