Hits and runs: International variation in the impact of the 2008 financial crisis

Recent empirical work on the 2008-09 financial crisis has found mixed results on the usefulness of indicators to explain the cross-country variation in the incidence of the crisis in non-originating countries.  Models of the crisis typically focus on the impact of various crisis causes or sources of contagion on economic growth performance during the crisis period.  These crisis causes and contagion include several categories of pre-crisis indicators such as financial market conditions and policies, macroeconomic imbalances, financial and trade linkages, institutions, and geography.  While some authors have found success with various indicators, Rose and Spiegel (2009a,b) find that almost no indicators are robust.  We revisit the broad cross country dataset compiled by Rose and Spiegel (2010a) with the objective of checking the robustness of their results to various alternative empirical assumptions and methods, contributing to the building of econometric models to help predict risk of exposure to international financial shocks.   First, we employ Bayesian Model Averaging to check the results of Rose and Spiegel (2009a,b) under model uncertainty, confirming their findings.  Further,  we employ Latent Class Modeling to check the data for non-linearities.   We find that when allowing for sample heterogeneity, many variables prove to be effective, including the growth in banking credit from 2000-06, inflation, the degree of credit market regulation, status as a major commodity exporter, and several financial and institutional variables.  The sample can generally be split into 2 classes, according to initial (2006) income.  Our results demonstrate that the lack of empirical success in determing sources of international contagion can be explained in part by the use of linear models.