BANK DISTRESS PREDICTION IN THE REPUBLIC OF CROATIA BASED ON OFFICIAL FINANCIAL STATEMENTS

Članak prikazuje rezultate istraživanja predviđanja poslovnih teškoća banka u Republici Hrvatskoj. Na osnovi uzorka 38 banaka i osam omjera izračunanih iz javno dostupnih financijskih izvješća napravljena je početna klasifikacija banaka na grupu s poslovnim teškoćama i grupu bez poslovnih teškoća, primjenom metode višedimenzionalnih skala. Nakon toga, određena je orijentacija svake od dobijene četiri dimenzije primjenom višestrukih regresija. Na osnovi tako klasificiranih banaka obavljena je diskriminacijska analiza i izračunani su parametri logit modela i napravljene posteriori klasifikacije uzorka banaka.The paper deals with models for the prediction of bank distress in the Republic of Croatia based on official financial data. Eight financial ratios were calculated (three liquidity ratios, four profitability ratios and one capital adequacy ratio) for 38 banks – 2/3 of all banks, and 90% of bank assets in Croatia. Information for banks refers to 1998 financial year because it was characterized by the crisis in Croatian banking industry. To classify distressed and no distressed banks, multidimensional scaling method (MDS) was used. Satisfactory results were obtained (based on low raw stress coefficient, coefficient of alienation and Shepard diagram) with four-dimensional map. MDS classifies 13 banks as distressed - three banks rehabilitated before 1998, four rehabilitated after 1998, two banks went bankrupt in 1999, and two went bankrupt in 2002, two banks solved serious distress thru turnaround management after 1998. This classification in four-dimensional space was then orientated by multiple regressions of each ratio as dependent variable, and four dimensions as independent variable. Liquidity ratios were firmly associated with first, and less with second and fourth dimension, profitability ratios were associated with second dimension, and capital adequacy with second and fourth dimension. After that, backward-stepwise discriminant analysis was performed. This analysis revealed that profitability ratios are better discriminators between distressed and non distressed banks than liquidity ratios. Final result had four variables of discrimination, and posterior classification was successful. Logit model (Hooke-Jeeves method was used) revealed similar results. Final conclusion is that we can have confidence in officially published financial data of banks in Croatia and use them for classifying and prediction of banks with respect to the level of distress