A robust model for optimal bank asset structure

Given a liability structure, the optimal bank asset structure problem consists in determining an asset allocation that maximizes profit, subject to restrictions on Basel III ratios and credit, liquidity and market risks. Most bank asset allocation models are very sensitive to inputs, making them difficult to use in practice, due to rigidities in the balance sheet. Our first contribution is to develop an optimization method that guarantees the stability of the allocations against the parameters, based on turnover constraints. On the other hand, bank allocation models have not been tested using historical data. We develop such tests, which document the superior performance of optimization strategies when compared to heuristic rules, resulting in an average annual out-of-sample outperformance of 9.4% in terms of Return on Equity using our data set. The tests also confirm that turnover constraints are important in order to achieve smooth allocations that can be implemented in practice.