This paper proposed a bootstrap fixed-effect panel regression model for predicting return on assets (ROA) using variables in the capital adequacy, asset quality, management efficiency and liquidity (CAML) framework as determinants. Bootstrap is a machine learning technique that allows for resampling of observed data to obtain robust and efficient estimates when the strict normal probability distribution assumption required for modeling is not met.
The study adopted a longitudinal research design. Original samples consisted of data on 19 Ghanaian commercial banks for the period 2011 to 2022. To eradicate the influence of time-invariant characteristics and the data being heavily tailed, both classical fixed-effect and bootstrap fixed-effect regression analyses were conducted.
Controlling for bank size, it emerged that capital adequacy and management efficiency ratios have significant positive influence on ROA, while the liquidity ratio recorded significant negative influence. Also, a rather surprising finding suggests that the relationship between asset quality and ROA is positive and significant, contrary to a previous assertion that asset quality and return on bank assets are negatively related.
Studies on bank performance mostly utilized the CAMEL framework. CAML variables were used as predictors of ROA in a bootstrap regression for panel data that is not normally distributed. CAML variables play critical policy decision roles in times of bank crises management, where premium is given to key bank lifesaving indicators.
