A Heterogeneous Agent Model of Mortgage Servicing: An Income-based Relief Analysis

Deepeka Garg, Benjamin Patrick Evans, Leo Ardon, Annapoorani Lakshmi Narayanan, Jared Vann, Udari Madhushani, Makada Henry-Nickie, Sumitra GanesharXiv 2024 · DOI · OpenReview

TL;DR

This paper introduces a novel agent-based model (ABM) to simulate the complex interactions between mortgage borrowers and servicers, particularly during financial distress. The model reveals that lower-income households are disproportionately affected by income shocks and benefit most from mortgage relief options, highlighting the need for tailored solutions.

What Problem Does This Paper Address?

Mortgages represent a massive portion of household debt, and financial hardship can lead to significant distress for borrowers. While mortgage servicers play a crucial role in offering assistance, there's been a lack of detailed modeling of the intricate relationship between borrowers and servicers, especially concerning the effectiveness of various relief measures. Existing agent-based models of housing markets often overlook the servicer's perspective and their impact on borrower behavior. This paper aims to bridge this gap by developing an ABM that captures heterogeneous borrower behavior and analyzes the efficacy of different mortgage servicing strategies during financial distress.

Methods

The authors developed an adaptive multi-agent model using the "Phantom" framework. Each time step represents one month, aligning with typical mortgage payment schedules. The model includes several key agents:

  • Economy Agent: Simulates the broader economic environment, providing income to households, applying exogenous shocks (like income reductions), and updating house prices.
  • Borrower Agents: Represent households with realistic financial attributes (income, expenses, savings). These agents are designed as adaptive learners, using a Proximal Policy Optimization (PPO) algorithm to make decisions that optimize their utility, which is a trade-off between liquidity and equity.
  • Servicer Agent: Manages loan activities, acting as an intermediary. It follows a regulatory hierarchy (Regulation X) and is responsible for offering relief to borrowers and advancing missed payments to mortgage owners.
  • Mortgage Owner Agent: Represents the entity that owns the mortgage (e.g., a financial institution or GSE).

The model allows for analyzing the impact of various external shocks and different mortgage relief strategies on specific borrower subgroups.

Datasets

The paper mentions using data from the following sources to inform the borrower's financial characteristics:

  • S1901 US Census (for income)
  • S2506 US Census (for housing expenses)
  • Consumer Expenditure Survey (for non-housing expenses)
  • Survey of Income and Program Participation (for cash savings)

The paper also references data sources for servicing fees and costs, such as Kaul et al. (2019) and Getter (2022).

Key Insights and Ideas

  • Heterogeneity Matters: The model clearly demonstrates that borrower characteristics, particularly income level, significantly influence their vulnerability to income shocks and their response to relief measures.
  • Lower-Income Vulnerability: Lower-income borrowers are disproportionately affected by income shocks. This is because mortgage payments consume a larger portion of their income, leaving less liquidity, and they often have fewer liquid cash savings to weather the storm.
  • Servicer Profitability and Risk: Servicers' profits are heavily influenced by higher-income quintiles due to the servicing fee structure and the greater stability of these borrowers. However, servicers face temporary liquidity pressure when advancing missed payments during widespread shocks.
  • Effectiveness of Relief: The model can evaluate the impact of different relief strategies, such as mortgage reserve accounts (MRAs). MRAs show particular promise for lower-income households, significantly reducing foreclosure rates and providing additional months of payment flexibility.
  • Model Validation: The model's ability to replicate real-world mortgage studies and trends lends credence to its findings and its potential as a tool for "what-if" scenario analysis.

Key Results

  • Affected Rate and Time-to-Affect: Figure 2a shows that lower-income quintiles experience a higher "affected rate" (proportion of borrowers missing payments) even with the same relative income shock. Figure 2b illustrates that these same borrowers also have a shorter "time-to-affect" after an income shock.
  • Servicer Net Profit: Figure 3 indicates that while servicers generally maintain a positive net profit, larger income shocks can lead to temporary negative profit due to the need to advance missed payments.
  • Mortgage Reserve Accounts (MRAs): Figure 4a and 4b show that MRAs significantly reduce foreclosure rates and provide additional months of payment support, with lower-income households benefiting the most. For instance, a $5000 MRA can save 12 percentage points of foreclosures for lower-income borrowers. Figure 4c highlights that lower-income borrowers are more likely to uptake MRAs.

Limitations and Future Work

  • Model Granularity: While the model captures heterogeneity, the authors suggest that further granularity in borrower characteristics and servicer strategies could enhance its predictive power.
  • External Shocks: The current model applies exogenous income shocks. Future work could explore more complex and interconnected economic shocks.
  • Servicer Behavior: The servicer agent's behavior is modeled based on Regulation X. Exploring variations in servicer strategies and their impact on different mortgage owners could be a valuable extension.
  • New Products: The paper demonstrates the model's utility in analyzing new products like MRAs. Future work could involve simulating a broader range of innovative mortgage relief products and policies to foster financial inclusion and homeownership.

Summary updated

Abstract

Mortgages account for the largest portion of household debt in the United States, totaling around \$12 trillion nationwide. In times of financial hardship, alleviating mortgage burdens is essential for supporting affected households. The mortgage servicing industry plays a vital role in offering this assistance, yet there has been limited research modelling the complex relationship between households and servicers. To bridge this gap, we developed an agent-based model that explores household behavior and the effectiveness of relief measures during financial distress. Our model represents households as adaptive learning agents with realistic financial attributes. These households experience exogenous income shocks, which may influence their ability to make mortgage payments. Mortgage servicers provide relief options to these households, who then choose the most suitable relief based on their unique financial circumstances and individual preferences. We analyze the impact of various external shocks and the success of different mortgage relief strategies on specific borrower subgroups. Through this analysis, we show that our model can not only replicate real-world mortgage studies but also act as a tool for conducting a broad range of what-if scenario analyses. Our approach offers fine-grained insights that can inform the development of more effective and inclusive mortgage relief solutions.
BibTeX
@misc{garg2024heterogeneous,
  author = {Deepeka Garg and Benjamin Patrick Evans and Leo Ardon and Annapoorani Lakshmi Narayanan and Jared Vann and Udari Madhushani and Makada Henry-Nickie and Sumitra Ganesh},
  title = {{A Heterogeneous Agent Model of Mortgage Servicing: An Income-based Relief Analysis}},
  year = {2024},
  eprint = {2402.17932},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2402.17932},
}
Figure