Modeling Mortgage Prepayments with Multinomial Logit
Summary
The note discusses loan-level modeling of mortgage prepayment rates when observations vary across loans and over time. It presents multinomial logistic regression as an established academic approach for modeling prepayment outcomes. Time-related variation can be represented with changing explanatory variables, including measures of credit availability and loan vintage. This offers a starting point for analyzing how loan characteristics and economic conditions relate to prepayment behavior, beyond applying a simple regression to pooled longitudinal and cross-sectional data.
The method has an important limitation: multinomial logit relies on the independence of irrelevant alternatives assumption, which may not hold for prepayment data. Extensions can address some of these concerns, but add estimation complexity. The discussion does not specify a canonical implementation or settle whether fixed effects are appropriate for a particular dataset. It points readers toward examining the modeling choices in relevant studies, since the right specification depends on the outcome structure and research question.
Key ideas
- Multinomial logistic regression is a commonly used framework for loan-level prepayment modeling.
- Time-varying predictors can represent changing credit conditions and loan vintage.
- The independence of irrelevant alternatives assumption may be unsuitable for prepayment outcomes.
- Extensions can relax aspects of the baseline model but make estimation more complex.
Tags
Full text
# Loan level model to understand drivers of mortgage prepayments # Loan level model to understand drivers of mortgage prepayments I am following up from my question here. As described there, I'm trying to assess the drivers of CPRs for a type of MBS. However, I want to understand, how a loan-level model of such a relationship would look? Specifically, how does one model the drivers of prepayment at the loan level, as a typical regression will not work given that the data is both longitudinal and cross-sectional in nature. So, in this case, is there a tried and tested regression or other procedure that is used? I am leaning towards a Fixed Effect model. ## Answer by Sharad (score 1) https://quant.stackexchange.com/a/58603 Probably the most established (thus far...) academic approach to this problem is to use a multinomial logistic regression (the search terms "prepayment model multinomial logit" should turn up dozens of papers). Longitudinal effects are captured in some of these models by including time-dependent explanatory variables such as an index that captures the availability of mortgage credit, the vintage of a loan etc. On the other hand, it is not clear that prepayment data satisfies a crucial assumption made by the model, namely the "independence of irrelevant alternatives." Various extensions have been proposed to the baseline MNL framework to handle this but these introduce additional layers of complexity to the estimation process. Unfortunately, there is no canonical reference (that I know of) that provides a guided tour through these nuances in the context of prepayment modeling. Going through the modeling details of some of the papers found by searching and working through their references and also asking specific questions appears to be the only way right now.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.