The Future Aging Housing Model (FAHM): Forecasting Affordability Risks for Older Adults

• In Event: Spatial Methods and Geographic Dimensions of Policy Evidence

Abstract

The United States is undergoing a rapid demographic transition as the population age 50 and older grows and lives longer. Housing affordability is increasingly central to whether older adults can remain in their homes and communities, yet policymakers lack forward-looking tools to anticipate how affordability risks will evolve over time. Most existing approaches rely on static indicators such as current cost burden rates, limiting their usefulness for planning interventions that support aging well in place.This paper introduces the Future Aging Housing Model (FAHM), an AI-enabled forecasting and decision-support system designed to project housing affordability trajectories and net worth outcomes for older households through 2030 and 2040. The model addresses a key gap in housing policy analysis by producing uncertainty-aware projections of future housing cost burden rather than backward-looking descriptive statistics.

FAHM integrates multiple national datasets—including the American Community Survey (ACS), Health and Retirement Survey (HRS), and Survey of Income and Program Participation (SIPP)—into a pseudo-panel modeling framework that reconstructs lifecycle trajectories from repeated cross-sections. Missing wealth and mortgage variables are imputed using supervised learning methods trained on HRS and SIPP. We then implement gradient-boosted decision-tree models (LightGBM) separately for homeowners and renters to capture nonlinear interactions among demographic characteristics, housing tenure, income dynamics, and local market conditions. Scenario simulations further test how shocks such as early retirement affect projected affordability trajectories between 2030 and 2040.

Preliminary results highlight several policy-relevant findings. First, projections suggest that affordability risks are increasingly concentrated in the upper tail of the distribution, with a growing share of older households expected to face extreme housing cost burdens exceeding conventional thresholds. Second, substantial heterogeneity emerges across states, underscoring the importance of state-level planning tools for senior housing policy. Third, projected net worth trajectories reveal a structural divide between renters with limited financial buffers and homeowners whose resources are concentrated in illiquid home equity. Finally, scenario simulations demonstrate that labor-force exit timing and related income shocks meaningfully alter affordability risk pathways over the next two decades.

In addition to producing point forecasts, FAHM generates distribution-sensitive projections that allow policymakers to identify households exposed to extreme affordability stress rather than relying solely on median trends. This feature is particularly important for housing and health policy, where risks are concentrated among households facing interacting constraints related to fixed incomes, health shocks, and tenure transitions. Because the model operates within a modular architecture, the affordability module serves as a foundation for integrating additional domains—including housing accessibility, supply availability, and climate-related insurance exposure—into a unified forecasting framework designed to support longer-term planning for population aging. By translating complex demographic and financial uncertainty into interpretable forecasts, FAHM provides a scalable modeling infrastructure to support federal, state, and local decision-making. The affordability module serves as a proof-of-concept for a broader modular platform designed to anticipate future housing stability challenges among older adults.

Author

• Linna Zhu, Urban Institute
  Presenting Author