Forecasting Rent vs. Buy Housing Costs Using Metro-Level Zillow Data
Abstract
Housing affordability is a growing economic concern in the United States, and individuals increasingly face uncertainty when deciding whether to rent or purchase a home under changing market and macroeconomic conditions. This project aims to develop a data-driven forecasting framework to compare rental and mortgage costs across U.S. metropolitan regions and support informed rent-versus-buy decisions. The central research question asks whether time-series forecasting models incorporating housing and macroeconomic indicators can reliably estimate future rent and mortgage costs at the metro level. We hypothesize that multivariate time-series models will outperform traditional regression approaches due to the smooth, trend-driven nature of housing markets. Using Zillow metro-level housing data combined with macroeconomic variables such as inflation, unemployment, and mortgage rates, extensive feature engineering was conducted, including lagged variables, rolling averages, seasonal encodings, and market structure ratios. Two modeling approaches were evaluated: a baseline linear regression model and an advanced Prophet time-series forecasting model with external regressors. Models were trained using historical monthly data with a 12-month holdout period and evaluated using mean absolute error (MAE) and root mean squared error (RMSE). Results show that Prophet consistently outperformed linear regression for both rent and mortgage value forecasting, achieving lower error rates and improved trend capture across metropolitan areas. Based on these forecasts, a Rent vs. Buy Index was constructed as the ratio of predicted rent to predicted mortgage costs, enabling dynamic affordability comparisons across regions. This study demonstrates that metro-level time-series forecasting can provide meaningful insights into housing affordability and offers a practical foundation for decision-support tools in housing economics and policy analysis.
Start Time
15-4-2026 9:00 AM
End Time
15-4-2026 12:00 PM
Room Number
Culp Ballroom 316
Poster Number
4
Presentation Type
Poster
Presentation Subtype
Posters - Competitive
Presentation Category
Science, Technology, and Engineering
Student Type
Graduate and Professional Degree Students, Residents, Fellows
Faculty Mentor
Ghaith Husari
Forecasting Rent vs. Buy Housing Costs Using Metro-Level Zillow Data
Culp Ballroom 316
Housing affordability is a growing economic concern in the United States, and individuals increasingly face uncertainty when deciding whether to rent or purchase a home under changing market and macroeconomic conditions. This project aims to develop a data-driven forecasting framework to compare rental and mortgage costs across U.S. metropolitan regions and support informed rent-versus-buy decisions. The central research question asks whether time-series forecasting models incorporating housing and macroeconomic indicators can reliably estimate future rent and mortgage costs at the metro level. We hypothesize that multivariate time-series models will outperform traditional regression approaches due to the smooth, trend-driven nature of housing markets. Using Zillow metro-level housing data combined with macroeconomic variables such as inflation, unemployment, and mortgage rates, extensive feature engineering was conducted, including lagged variables, rolling averages, seasonal encodings, and market structure ratios. Two modeling approaches were evaluated: a baseline linear regression model and an advanced Prophet time-series forecasting model with external regressors. Models were trained using historical monthly data with a 12-month holdout period and evaluated using mean absolute error (MAE) and root mean squared error (RMSE). Results show that Prophet consistently outperformed linear regression for both rent and mortgage value forecasting, achieving lower error rates and improved trend capture across metropolitan areas. Based on these forecasts, a Rent vs. Buy Index was constructed as the ratio of predicted rent to predicted mortgage costs, enabling dynamic affordability comparisons across regions. This study demonstrates that metro-level time-series forecasting can provide meaningful insights into housing affordability and offers a practical foundation for decision-support tools in housing economics and policy analysis.
