2023-10-09 - 2023-10-13
Armen Hayrapetyan
UniCredit Bank
Uncertainties that surround us can give rise to potential financial losses – or even gains. Such an obscurity with negative impact is interrelated with financial and non-financial risks. Both types of risks should be assessed to ensure capital adequacy of the financial institute and to verify robustness and reliability of internal estimates used to calculate own funds requirements. To this end, risk factors are accessed by means of statistical methods with the sole purpose to establish various metrics enabling a comprehensive overview of risks. This is exactly the stage where the scientific approach of, among others, mathematicians and physicists is made use of for developing and evaluating complex models, which cover a wide range of disciplines featuring, especially, Data Science and Probability Theory.
In this talk, we will put the emphasis on a prominent type of financial risk, the well-known credit risk, by providing an insight on why it is important for the bank to thoroughly assess it. We will demonstrate how the credit risk is modeled in terms of loss- and default-related parameters, such as the probability of default, loss given default and exposure at default. We will show possible ways of quantification of these parameters using estimation techniques based on Probability Theory and Big Data Analytics. Developed models are then evaluated within validation activities by means of diverse statistical measures for quantifying the performance of models. We will thus discuss assessment methodologies for verifying how good estimations reflect the reality, embodied in the so-called rank ordering and calibration tests. We will conclude with an exhibition of test approaches using a toy model and with arguing the relevance of modelling and validation processes, performed regularly to keep the model compliant with the internal and external regulatory requirements as well as to ensure the up-do-dateness of the model within the macroeconomic environment.
