Publications - Working Papers IER SAS

In this paper we examine the consistency in the timing of crisis events of the most prominent databases of banking and fiscal crises. In order to do so we calculate Cohen's kappa indicator measuring level of commonality across databases. Additionally, we employ panel logit models with random effects to investigate predictive properties of early warning indicators selected from the Macroeconomic Imbalances Procedure scoreboard. We also identify the most influential crisis observations unique to each database using Pregibon’s delta-beta influence statistics. Our results confirm that the degree of commonality across databases is indeed relatively high, especially if introducing a one-year lag due to possible beginning- and end-of-the-year discrepancies. However, there is still a significant role played by a few influential observations that determine several heterogeneous findings for statistically significant EWIs. This problem is more pronounced in the banking crisis literature. Based on the empirical findings, we propose several suggestions that should be discussed and potentially adopted by research community in order to address the existing concerns.