Contagion, spillover, and interdependence

This paper reviews the empirical literature on international spillovers and contagion. Theoretical models of spillover and contagion imply that the reduced-form observable variables suffer from two possible sources of bias: endogeneity and omitted variables. These econometric problems, in combination with the heteroskedasticity that plagues the data, produce time-varying biases. Several empirical methodologies are evaluated from this perspective: nonparametric techniques, such as correlations and principal components; and parametric methods, such as OLS, VAR, event studies, ARCH, and nonlinear regressions. The paper concludes that there is no single technique that can solve the full-fledged problem and discusses three methodologies that can partially address some of the questions in the literature.