Forecasting Mortgage Prepayments in Changing Interest Rate Regimes

Forecasting Mortgage Prepayments in Changing Interest Rate Regimes: A Hybrid Economic-Engineering Model with EMPC

Krabbenborg, Britt (TU Delft Mechanical, Maritime and Materials Engineering; TU Delft Delft Center for Systems and Control)

Mendel, M.B. (mentor)
Mazo, M. (graduation committee)
Hutters, C. (graduation committee)
van Zwol, Pieter (graduation committee)

Delft University of Technology

Mechanical Engineering | Systems and Control

2022-12-15

Banks such as Rabobank depend on multi-year mortgage prepayment forecasts in order to make provisions for the associated prepayment risks. The econometric models they use are fitted to historical data, and as a consequence their models are fitted to a decreasing interest rate regime. Given the current economic climate of increasing interest rates, Rabobank has ascertained that their models are underperforming. They expressed the need for an alternative modeling approach that performs better in changing interest rate regimes.

This thesis takes a systems and control approach motivated by this need. We split the development into two parts; a dynamical system for modeling mortgage payments and prepayments, and a controller for simulating mortgagor behavior.

We model the dynamical system by following the principles of economic engineering. Economic engineering is based on the method of analogs, and we develop specific analogies applicable to the mortgage market. We first derive a continuous model describing the mortgage payment and partial prepayment dynamics. This model is then extended towards a hybrid model to include the dynamics of full prepayment. The parameters of this economic engineering model can be identified with historical data and are relatively constant. The resulting model is not affected by the variation of interest rates and performs well in any interest rate regime.

We design an Economic Model Predictive Controller (EMPC) to simulate mortgagor behavior that minimizes an objective function of its costs. This controller minimizes an economic objective which is needed to simulate the behavior of mortgagors in changing interest rate regimes. For different interest rate scenarios, we forecast prepayments with the model by simulating this minimizing behavior. We perform simulations for different kinds of mortgagors by varying the model parameters and the objective function. Based on these simulations, we describe for each mortgagor both the exact cause and dynamics behind the mortgage prepayments supplied.

Interest rates
Hybrid modeling
Economic model predictive control
Forecasting
Mortgage prepayments

http://resolver.tudelft.nl/uuid:108f1838-20bc-4af4-90d0-6d473120633a

Student theses

master thesis

© 2022 Britt Krabbenborg