A Pattern Recognition Model for Predicting a Financial Crisis in Turkey

Updated: May 21

Authors: Ilkay Boduroglu, Zeynep Erenay

Using the pattern recognition paradigm, we have designed a scalar composite leading indicator that predicts a financial crisis in Turkey about six months in advance. Logistic regression was employed to distinguish between near-crisis months and safe months from January 1998 through October 2000, which is the last month before the most recent crisis in Turkey. Thus trained and once validated, the Turkish Economy Stability Index (TESI) was then successfully tested using independent data from the 1994 crisis. One of the two dimensionless attributes that we integrated into TESI, the capital adequacy ratio of banks, a precursor of banking crises, was selected from among seven precursors of the 1997 Asian Financial Crisis. The other one, the ratio of international reserves to short term outstanding external debt, is itself a precursor of currency crises. Our method for designing TESI may easily be applied to other emerging markets provided that they have had at least one financial crisis in recent history.

Key words: financial crises, pattern recognition, data mining, classification, logistic regression, Turkey, emerging markets, South Asia, economic stability, advance warning, leading indicators

Fig.1 Monthly maximum overnight interest rate (divided by 100) and TESI

Fig.2 The Istanbul Stock Exchange -100 (IMKB 100) converted to USD units and TESI

Recent Posts

See All

Subscribe Form

©2020 by RiskOptima FinTech R&D Corp.. Proudly created with Wix.com