The Asset Liability Management (ALM) process in a bank is multidimensional in nature. The best possible trade off solution for profitability will have to strike an appropriate balance among the key drivers viz., advances, investments, deposits and other income (non-interest income), while simultaneously taking care of the regulatory and other constraints. The objective of this paper is to estimate, in a robust manner, the relative importance of advances, investments, deposits and other income in predicting profits. A comparative assessment is made of the two methods: Ordinary Least Square (OLS) and Robust Regression based on Least Absolute Deviation (LAD) in order to select the one that is appropriate in this situation. The results show that the robust regression outperforms OLS in terms of predictive accuracy, particularly in the context characterized by outliers and non-normal distribution with longer tails. Elasticity coefficients have been computed using the estimated slopes of the robust regression as inputs for arriving at the percentage relative importance of each driver of profitability. For this study, data filtering for inconsistencies warranted exclusion of some banks. Secondly, the focus is mainly on predictive accuracy and not hypothesis testing where OLS may still prove to be more useful. These are the two limitations of the study.

Profit of a bank is predominantly driven by advances, investments, deposits and other income which are interconnected in terms of high pair-wise correlation. Increase in advances to get higher interest income may impact the deposits in terms of higher interest cost particularly term deposits. Increase in deposits may be achieved at the expense of advances and investments because risk-weighted assets will increase leading to a lower capital adequacy. The assets and liabilities composition in the bank balance sheet will have to be carefully planned because their interactions can impact the profitability either positively or negatively. This exercise in terms of best possible solution should take care of the multidimensional nature of the assets and liabilities management.

Profits of a bank are, therefore, a function of advances, investments, deposits and other income. It is essential to accurately estimate the response of profit to changes in advances, investments, deposits and other income. There are many possible ways to quantify the response of the dependent variable to changes in independent variables.

The least squares regression has dominated the statistical literature for a long time and it is still considered to be very important. This popularity can be attributed to the fact that the theory is simple, convenient, well-articulated and documented. It provides a platform for testing the hypothesis of the parameters and the goodness of fit when the errors are independent and follow a normal distribution with mean zero and a common unknown variance. The outliers occurring with extreme values of the independent variables can be very disruptive and may spoil the estimates in a significant manner. In the robust regression involving Least Absolute Deviation (LAD), we minimize the sum of absolute errors by varying the intercept and the regression coefficients. Because of its resistance to outliers, it provides better estimates than the least squares regression when there are outliers in the data set and residuals have non-normal distributions.

Bank Management Journal, Indian Banks, Asset Liability Management, Data Filtering, Least Absolute Deviation, Decision-Making Group, Commercial Banks, Ordinary Least Square, Banking Industry, Kenyan Banks, Least Squares Regression, Mutual Fund Industry, Linear Programming, Financial Markets, Capital Required Adequacy Ratio, Public Sector Banks.