Predicting Financial Crises in Emerging Markets Using a Composite Non-Parametric Model

Posted: 30 Oct 2012 Last revised: 31 Oct 2012

Date Written: December 1, 2005

Abstract

The large number of financial crises in emerging markets over the past ten years has left many observers, both from academia and financial institutions, puzzled by an apparent lack of homogenous causal relations between endogenous economic variables and the bursting of large financial shocks. The paper aims at showing that the key difficulty is not the identification of proper endogenous variables, but the ability to combine them in a way that is able to capture the combinatorial aspect of such causal relations. The paper is based on a newly developed non-parametric methodology for country risk signaling: the RiskMonitor CDM-Model. Using a combination of macroeconomic indicators and a composite model of 5 modern non-parametric classification methods, we constructed 9 early warning signals to predict financial crises in emerging markets. These signals are constructed for 3 types of crises (cyclical crises, exchange rate crises and transfer crises) and over 3 horizons (less than 1 year, 1 to 3 years, 3 to 5 years). This complex use of quantitative models is able to provide excellent early warning information, with impressive back-testing results on 50 developing countries over the period 1980 to 2002.

Keywords: financial crises, leading indicators, early warning systems, classification methods, emerging countries

JEL Classification: C10, C44, C45, F31, F34, F37, F40, O1

Suggested Citation

Barthelemy, Sylvain and Apoteker, Thierry, Predicting Financial Crises in Emerging Markets Using a Composite Non-Parametric Model (December 1, 2005). Emerging Markets Review, Vol. 6, No. 4, 2005, Available at SSRN: https://ssrn.com/abstract=2168803

Sylvain Barthelemy (Contact Author)

TAC ( email )

La Saigeais
35140 Saint Hilaire des Landes
France

Thierry Apoteker

TAC ( email )

La Saigeais
35140 Saint Hilaire des Landes
France

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
342
PlumX Metrics