Understanding mortgage performance remains a key challenge for lenders and policymakers as borrower transitions between loan states depend on both individual financial behaviour and broader economic conditions.
This study introduces a predictive model that integrates a first-order Markov process, auto-regressive components, and logistic regression, offering a more dynamic approach to mortgage risk assessment.
By leveraging an extensive Fannie Mae dataset spanning origination vintages from 2001 to 2021, the study quantifies the dynamics governing transitions between key loan states—namely performing, arrears, default, modification, and foreclosure.
Our findings indicate that both borrower-specific characteristics and macroeconomic indicators, such as state unemployment, and house price index (HPI) play significant roles in predicting these transitions. The findings also demonstrate that loan modifications significantly increase cure rates, even though they are infrequently offered before severe delinquency in the U.S.
Finally, our transition models reveal that although servicers face pressure to intervene during high unemployment, rising house prices and delinquency volumes dampen their willingness to modify, reflecting both diminished incentives when equity recovers and heightened redefault concerns.
The research delivers actionable insights that can be leveraged for:
- Optimizing risk management practices and refining loan pricing strategies,
- Informing regulatory policy decisions, and
- Establishing a foundation for proactive mitigation measures in mortgage portfolios.
