Difference-in-Differences with Spillovers: Identification through Heteroskedasticity
Difference-in-Differences (DiD) designs assume no spillovers.
Yet spillovers are common in economics and bias treatment effect estimates. This paper introduces a simple method to obtain spillover-robust treatment effects and directly estimate spillovers using changes in (co)variance between treated and control group.
The approach leverages the heteroskedasticity that many treatments in economics plausibly induce on theoretical grounds. Applications to merger analysis and minimum wage studies - where both spillovers and heteroskedasticity are likely - show significant spillovers and substantial bias in conventional estimates. The method provides a practical solution for researchers facing spillover concerns in DiD frameworks.