Predicting Financial Stress in Emerging Countries

Abstract

Economic policies are designed to eliminate vulnerabilities while also increasing the welfare of the citizens. Yet, some undesirable factors could hinder these goals by distorting the functioning of the financial system, and even leading the economy into a crisis. Being prepared for this kind of negativities stands out as one of the most important responsibilities of policy makers. For this reason, constructing early warning indicators that can capture business cycles has been studied for more than 40 years. However, progress in this field has remained limited; a great number of studies have focused only on building financial stress indices to foresee a crisis. Furthermore, although dynamic non-linear models promise high potential when used with early warning indicators, there is not enough work on this subject. Finally, carrying out a study in particular for developing countries was not popular in the business cycle literature, and the vast majority of the studies addressed advanced countries. Therefore, the aim of this thesis is to bring novelty by predicting financial stress turning points through a non-linear Markov regime switching model with time-varying transition probabilities, which has not been applied for a group of emerging countries before, and by using the “spread” variable of Sönmez and Kandemir Kocaaslan (2022) as an explanatory variable in predicting turning points. To this end, firstly, financial stress indices were constructed for each country. Evidence suggests that industrial production growth promotes a high financial stress period in all countries in the next few quarters. In Chile, South Africa and Türkiye, the CBOE Volatility Index gives signals up to four quarters before exiting from high financial stress period. In Chile and Poland, private sector credit growth predicts regime switches from high to low financial stress period up to eight quarters in advance. And finally, “spread” between business and mortgage credit interests predicts exit from tense state in all countries one to six quarters in advance, except Türkiye.