GeoDevelopment Observatory of Bolivia / Projects
Regional shocks and spillover effects
Synthetic control methods for regional shocks whose effects leak to neighbouring regions, with scspill, an open Python package that estimates both the direct effect and the spillovers.
When a policy or a shock hits one region, its effects rarely stop at the border: neighbouring regions trade, migrate and invest across it. The standard synthetic control method builds a counterfactual for the treated region from untreated “donor” regions, and is biased when those donors are themselves affected. This project develops scspill, a Python package for synthetic control with spillovers, which estimates the effect on the treated region together with a spillover effect on each donor. It implements the Bayesian spatial-autoregressive model of Sakaguchi and Tagawa, with a single spillover intensity estimated by Metropolis within Gibbs, and the method of Cao and Dowd for a known exposure structure, with hypothesis tests, plots and a Monte Carlo engine. Worked case studies include California’s Proposition 99 and the 2011 secession of South Sudan.
Mendez, C., Sakaguchi, S., & Tagawa, H. (2026). Synthetic Control Models with Spillovers in Python (version 0.3.0).