ORIGINAL ARTICLE Systemic early warning systems for EU14 based on the 2008 crisis: proposed estimation and model assessment for classification forecasting Savas Papadopoulos 1 • Pantelis Stavroulias 2 • Thomas Sager 3 Ó Springer Nature Limited 2018 Abstract Reliable forecasts of an economic crisis well in advance of its onset could permit effective preventative measures to mitigate its consequences and become a valuable tool for banking regulation and macroprudential policy. Using the EU14 crisis of 2007–2008 as a template, we develop methodology that can accurately predict a banking crisis several quarters in advance in each country. The data for our predictions are standard, publicly avail- able macroeconomic and market variables that are pre- processed by moving averages and filtering. The prediction models then utilize the filtered data to distinguish pre-crisis from normal quarters through standard statistical classifi- cation methodology plus one proposed method, enhanced by an innovative goodness-of-fit measure used in the esti- mation and in the threshold selection. Empirical results are quite satisfactory and can be used by policy makers, investors and researchers who are interested in estimating the probability of a crisis as much as one and a half years in advance in order to deploy prudential policies. Implications to bank regulatory policy are discussed. Keywords Banking crisis  Macroprudential policy  Classification methods  Decision trees and C5.0  Goodness-of-fit measures Introduction ‘‘Achieving financial stability is perhaps the most urgent task facing the world economy at the present time. If the international financial system cannot be made to operate in a more stable way, the prospects for an open and liberal approach to trade and capital flows are poor … the fun- damental goals of development and poverty alleviation will be set back’’ Andrew Crockett [1]. Although research ori- ented toward developing early warning system (EWS) models of financial crisis was underway in the mid-1990s, a strong stimulus was given in recent years following the global financial crisis, which started in the USA in 2007 and spread through the rest of the world in the following years. The burgeoning of EWS was to be expected since the direct and indirect cost of the global financial crisis was vast and had a significant impact on the health and soundness of the entire global financial system. Subse- quently, policy makers started to reconsider the existing early warning models, the predictors of financial instability and their policy tools in order to build new ones with the objective of preventing or at least reducing the intensity of future financial crises, using all the knowledge so far. Since the global crisis of 2007–2008, the literature on EWS has proliferated. For a historical review of EWS and for a classification of crisis generations, see Kaminsky [2]. For a comparison of EWS methods before 2008, see Davis and Karim [3]. The following comments briefly summarize some of the literature that is most relevant for this article EU14 stands for the EU15 countries in the European Union prior to the accession of ten candidate countries on 1 May 2004, minus Luxemburg. Luxemburg was excluded due to missing values of variables important for the study. & Savas Papadopoulos sapapa@bankofgreece.gr 1 Department of Financial Stability, Bank of Greece, Amerikis 3, 102 50 Athens, Greece 2 Economics Department, Democritus University of Thrace, Komotini, Greece 3 Department of IROM, B6500, University of Texas at Austin, 1 University Station, Austin, TX 78712, USA J Bank Regul https://doi.org/10.1057/s41261-018-0085-0