Starting at the Top
Setting Up the Problem
Considering Multiple Vintages
Volatility Analysis and Economic Capital
Credit Scores and Account Management
Analysis of the US Mortgage Crisis
An Example Using [email protected] Data
Examples of Modelling Vintages
Epilogue: it’s about time
“Push-button” forecasting does not exist in retail lending. No credit risk model can predict future account management policies or new bookings plans, and they should not be expected to predict the future economic environment. Forecasting the economy requires a distinctly different set of models. These items should be included as scenarios driving a portfolio forecast. Consequently, the only forecasting in retail lending is scenario-based forecasting.
Saying that, many models do not accept input for these scenario elements. If the scenario is not explicit in the model, it must then be implicitly assumed not to change or be forecasted by the model. For example, the typical roll-rate model, described later in the chapter, is a trend follower. If originations have been trending upwards or the economy is trending downwards, then those trends are assumed to continue.
In this chapter we discuss true scenario-based forecasting, where a clear distinction is drawn between what can accurately be predicted (lifecycle, seasonality and existing vintage quality) and what can only be guessed (future economic environment and future originations).
The goal of a retail lending po