Working Papers

Water, Spillovers, and Free Riding: Provision of Local Public Goods in a Spatial Network (pdfSSRN)
In many low-income countries, local public goods are provided by non-governmental organizations and local communities, rather than by government. In rural Tanzania, more than 500 organizations install different technologies of water pumps which local communities are then responsible for maintaining. One third of pumps are non-functional, despite low access to clean water. I show evidence that communities free ride on their neighbors' pump maintenance investments, but also benefit from spillovers that reduce the cost of maintaining their own pump. When pump maintenance spillovers are large, maintenance decisions are strategic complements, but when spillovers are small, free riding effects dominate and maintenance decisions are strategic substitutes. I develop and estimate a spatial network model of communities' pump maintenance decisions to quantify the importance of free riding and maintenance spillovers on pump functionality and child outcomes, and to estimate the effects of two policies that have been proposed in Tanzania. The model estimates that: (i) water collection fees mitigate free riding -- if adopted universally, pump functionality rates would increase by 11 percentage points; (ii) pump technology standardization across communities reduces maintenance costs -- full coordination by installing organizations would increase pump functionality by 6 percentage points; (iii) increased pump functionality improves child outcomes -- a 10 percentage point increase in functionality increases the child survival rate by 0.9 percentage points and school attendance by 1.5 percentage points.

Identifying Causal Effects in Experiments with Social Interactions and Non-compliance, with Francis DiTraglia, Camilo Garcia-Jimeno and Alejandro Sanchez-Becerra (pdf)
This paper shows how to use a randomized saturation experimental design to identify and estimate causal effects in the presence of social interactions - one person’s treatment may affect another’s outcome - and one-sided non-compliance - subjects can only be offered treatment, not compelled to take it up. Two distinct causal effects are of interest in this setting: direct effects quantify how a person’s own treatment changes her outcome, while indirect effects quantify how her peers’ treatments change her outcome. We consider the case in which social interactions occur only within known groups, and take-up decisions do not depend on peers’ offers. In this setting we point identify local average treatment effects, both direct and indirect, in a flexible random coefficients model that allows for both heterogenous treatment effects and endogeneous selection into treatment. We go on to propose a feasible, kernel-based estimator.

Work in Progress

Market-Mediated Effects of Cash Transfer Programmes, with Natalie Quinn

Distributional Effects of Cash Transfers, with Stefan Dercon, Robert Garlick, Kate Orkin and Natalie Quinn

The Spillover Effects of Financial Interventions: Evidence from a Savings Program in Ugandawith Chris Heitzig