Forecasting Real Estate Backlog: A Regime-Switching Approach

Abstract

The current housing affordability crisis can partially be attributed to a significant backlog in the real estate market, leading to shortages in housing supply. There is a substantial amount of literature focused on forecasting real estate demand and prices, which leaves a gap for more work to be done on the supply side. In this paper, we develop a hybrid learning framework to identify and predict both national and regional housing backlog regimes. Firstly, we use Spectral Clustering to identify high and low growth regimes. We find that these regimes tend to coincide with National Bureau of Economic Research (NBER) business cycles. Then, we leverage L1-regularized Logistic Regression to determine the top macroeconomic indicators for predicting changes between these regimes. We determine that construction material costs, the 30-year fixed rate, and the unemployment rate are the most significant predictors of housing backlog regimes.