Spillover Effects with Nonrandom Sample Selection

Pablo Estrada^1
1Emory University

Paper | Five-minute Summary

Summary

This paper presents a method to estimate spillover effects of a random treatment using a nonrandom sample of individuals with observable outcomes. The approach uses an exposure monotonicity assumption to bound spillover effects, accounting for network dependence. It also incorporates high-dimensional covariates through machine learning to tighten the bounds.

Replication

Data_Laptops.ipynb contains the code to generate the data used in the empirical application.

Spillover_Bounds.ipynb contains the code to estimate the spillover bounds.

Simulation.ipynb contains the code to replicate the simulation study.

Citation

@misc{Estrada_2024,
  title={Spillover Effects with Nonrandom Sample Selection},
  author={Estrada, Pablo},
  year={2024}
}