Discriminant analysis has been employed in this paper in order to identify and explain key features of
bank profitability levels. Bank profitability is set up in the form of two categorical variables: profit or
loss recorded and above or below average return on equity. Predictor variables are selected from
various groups of financial indicators usually included in the empirical work on microeconomic
determinants of bank profitability. The data from the Croatian banking sector is analyzed using the
Enter method. General recommendations for a more profitable business of banking found in the bank
management literature and existing empirical framework such as rationalization of overhead costs,
asset growth, increase of non-interest income by expanding scale and scope of financial products
proved to be important for classification of banks in different profitability levels. A higher market
share may bring additional advantages. Classification results, canonical correlation and Wilks’
Lambda test confirm statistical significance of research results. Altogether, discriminant analysis turns
out to be a suitable statistical method for solving presented research problem and moving forward
from the bankruptcy, credit rating or default issues in finance