Smart Social Safety Nets: Reducing Incomplete Take-Up Through Digital Innovation in Latin America

I develop a welfare-based framework that links a data-driven incremental innovation cycle to a take-up model and use it to study how digital tools can raise participation and welfare in rental-voucher programs. The welfare effects of digital interventions hinge on targeting and on the distribution of frictions and fiscal externalities across applicant types. I apply the framework to Chiles rental-voucher program using a large randomized evaluation (N11,149) of a digital counseling platform to increase voucher utilization. The evaluation randomized (i) basic program information, (ii) motivational neighborhood messaging, and (iii) comprehensive neighborhood search tools. Using administrative and survey data with machine learning, I show that information frictions (assignment timing, administrative access), supply constraints (landlords, geography, tenure), and demand factors (savings, income, family composition) predict utilization; yet unassisted utilization is only moderately predictable, indicating important post-assignment frictions. Causal forests reveal substantial heterogeneity: basic information helps economically vulnerable households with lower supply barriers; motivational messages mainly aid already-engaged renters; comprehensive tools have the broadestand growingreach. Counseling increased utilization without raising rents and modestly improved neighborhood amenities, at the cost of longer search. Results support pairing universal basic information with widely available search tools and complementary supports.