Forecast of forecast: An analytical approach to stressed impairment forecasting

Abstract

The new impairment reporting standards require banks to move from incurred loss models to sophisticated macroeconomic based expected credit loss models for current impairment estimation. While the impairment estimation is mainly focused on business as usual macroeconomic projections, there is a demand and expectation from regulators that the new impairment models should also be a foundation for the next generation stress tests. The expected credit loss models are the base for loss provisioning and hence are subject to stress testing. The use of impairment models in stressed situations of course have profound implications on banks model, calibration and validation approach. This paper proposes a forecast of forecast approach to stressed impairment estimation. The paper captures consistently the initial stressed macroeconomic development (which can, for example, be a regulatory mandated stress scenario), the bank specific assumptions about new business generated in the initial stress development and an estimation of the forward impairment forecast. After an initial motivating example of the forecast of forecast approach using a delinquency state transition model we show how an analytical and computationally tractable forecast of forecast approach can be simply implemented using only minor changes to the current impairment models calculation, especially for state transition models with no or limited state path history tracking, which covers the majority of models used by banks in practice.