Abstract
We provide a critical review of the literature on early warning indicators of economics crises and propose methods to overcome several pitfalls of the previous contributions. We use a quarterly panel of 40 EU and OECD countries for the period 1970-2010.
As the response variable, we construct a continuous index of crisis incidence capturing the real costs for the economy. As the potential warning indicators, we evaluate a wide range of variables, selected according to the previous literature and our own considerations.
For each potential indicator we determine the optimal lead employing panel vector autoregression, then we select useful indicators employing Bayesian model averaging. We re-estimate the resulting specification by system GMM to account for potential endogeneity of some indicators.
Subsequently, to allow for country heterogeneity, we evaluate the random coefficients estimator and illustrate the stability among endogenous clusters. Our results suggest that global variables rank among the most useful early warning indicators.
In addition, housing prices emerge consistently as an important domestic source of risk.