 |
 |
Credit Commentary - Nov. '07
Tuning the LoanDynamics™ Model
by Anne Ching and Kyle Lundstedt
The LoanDynamics™ Model (LDM) is an economic model of prepayment, delinquency, default, and loss which is calibrated to historical data. As such, it provides critical information about the relative effects that different variables have on borrower behavior. However, one of the most valuable features of LDM is its flexibility, since it allows users to scale major parameters and inputs within the model.
AD&Co believes that users should have the ability to perform sensitivity analyses, and to express a view about how a particular collateral type or issuer should perform. Similarly, there are situations when the current market deviates significantly from historical performance, and thus behavioral models require adjustments to reflect market dislocations.
There are six different tuning categories within LDM, which are described in detail below.
1. SATO (Spread at Origination) Residual is a very important variable in LDM which proxies for unobserved loan characteristics not disclosed in servicing data, such as DTI ratio, prior default history, employment history, etc. By tuning SATO Residual, users can either reduce or amplify the SATO Residual effect.
2. Transition Probabilities Users have the ability to scale the transition probability equations for six transitions: Delinquency, Default, Cure, Delinquent Prepayment, Cure and Liquidation. The default tuning is 1.0. However, a user may want to increase the rate at which loans transition from one state to another. A value of 1.5 would increase a particular transition rate by 50%. The model will automatically adjust the remaining transition probabilities so that they sum to 100.
3. Loss Parameters There are two parts to calculating the expected loss of a loan: the probability of loss and the magnitude of the loss. Users have the ability to scale both for each of the three states from which losses can occur: Current, Delinquent and Seriously Delinquent.
