Asset classification under the IFRS 9 framework for the construction of a banking investment portfolio

Abstract

Under the IFRS 9 framework, we analyze the tradeoff of classifying a financial asset at amortized cost versus at fair value. Defining an impairment model and based on historical (2003-2019) data for the 10-year Portuguese Government bonds, we analyze the annual performance (income/comprehensive income) of different investment allocations. Setting as objectives the maximization of the income and the minimization of the semivariance of the comprehensive income, we suggest a bi-objective model in order to find efficient allocations. Given the non-smoothness of the semivariance function, we compute the solution of the suggested model by means of a multi-objective derivative-free algorithm. Assuming that the yields and funding rates follow a correlated mean-reverting process and that the bonds' rating dynamics are described by an ordinal response model, we show a possible approach to mitigate the estimation error ingrained in the proposed bi-objective stochastic model. Finally, we assess the out-of-sample performance of some of the suggested efficient allocations.