---
title: "Forecasting Real Housing Price Growth in the Eighth District States"
authors: "David E. Rapach; Jack K. Strauss"
source: "Regional Economic Development (St. Louis Fed)"
source_url: "https://cre.org/real-estate-issues/a-brief-review-of-house-price-forecasting-methods"
paper_id: "Regional Economic Development 31(2):33-42 (2007)"
harvested: 2026-08-22
year: 2007
keywords: [house prices, ARDL, forecast combination, state-level, leading indicator]
---

## Abstract

David E. Rapach and Jack K. Strauss forecast real housing price growth for states in the
Federal Reserve's Eighth District using a **forecast-combination** approach. The central
methodological finding — consistent with the broader forecasting literature — is that
**combining (pooling) multiple autoregressive-distributed-lag (ARDL) models outperforms any
single model**. They exploit cross-state heterogeneity in housing markets to show that no one
model or specification dominates; pooled / combined forecasts are more accurate and more robust
than picking a single best model.

**Relevance to StratMac RE-sector work:** Rapach–Strauss is the canonical "combine, don't pick"
result. Its direct tension with StratMac's finding (from `re-predictive-modeling.md`) is that
the parsimonious top-5 model **beat** the full model on out-of-sample HOUST/PERMIT prediction —
pooling toward a fuller set *lost* here. The paper is the right cite for the robustness section:
parsimonious-sparse vs pooled-ensemble is an empirical question, and StratMac has the walk-forward
evidence to argue the sparse side on these continuous targets.

Also cited in the *Counselors of Real Estate* house-price-forecasting review, which credits the
ARDL-combination approach for accurate state-level price forecasting.