Models must contend with unprecedented challenges
One challenge for modelers is that some of the measures and outcomes are so extreme – millions signing on for unemployment each week, potentially negative nominal interest rates, widespread mortgage non-payments for example – that they fall well outside the normal assumptions used to model GDP growth, default rates, and interest rate risk, as just a few examples. Some models will be able to handle these extremes and produce valuable results; others may not.
Another challenge is the absence of reliable data to use in models. For example? Currently, in the UK, banks have introduced payment holidays on mortgages, personal loans, and credit card debt. While welcome news for hard-pressed borrowers, the absence of default data means trying to assess the impact on default rates during the pandemic is almost impossible to estimate. It will only become clear once the situation returns to somewhere near normal.
To get around these issues, and provide insight and options for senior decision-makers, model teams are having to develop new models – at speed – that reflect the new reality and provide fresh answers. Where their existing models are not proving adequate, they are developing End-User Computing (EUC) applications using platforms like Python, R, MATLAB, and even highly complex spreadsheets.
These can be set up quickly and can accommodate new assumptions and new types of inputs to create meaningful insights into areas like portfolio performance, credit risk, market risk, as well as wider decision-making options.