METHODOLOGICAL STRATEGIES FOR PANEL DATA: THE CASE OF TYPICAL BANKS IN ARGENTINA

In this work we analyze the determinants of bank profitability in the case of Argentinian Typical banks through the period 2005-2018. Profitability is measured as return on assets (ROA) and the bank-specific determinants considered are Capital (Equity/Assets), Credit risk (Loan loss provisions/loans), Productivity (Gross total revenue/personnel), Operating expenses (Operating expenses/assets) and Size (logarithm of assets). Additionally, derived from the Herfindhal-Hirschman concentration index, the share of each bank of the system-total was considered as an industry-specific variable. As a first step a linear model was estimated by Ordinary Least Squares (OLS) allowing the early detection of outliers. In a second step, methods that take account of the hierarchical structure of the information were considered, such as Random Effects Model (RE) and Fixed Effects Model (FE), both with robust standard errors estimation and standard errors with the Driscoll and Kraay correction. In order to select the appropriate model, Hausman test of specific effects and tests to inquire into cross-sectional correlation were performed. Results indicate that a fixed Effects model with the Driscoll and Kraay correction of variances is the most appropriate model. The ratios that significantly and positively affect bank profitability are Capital, Productivity, Size and Herfindahl concentration; and negatively affecting Credit risk.