| Indicator | Coefficient |
|---|---|
| Industrial Production Index | +0.725 |
| Retail & Food Services Sales | +0.624 |
| Initial Unemployment Claims | -0.012 |
| Total Nonfarm Payrolls | +1.117 |
| Housing Starts | +0.071 |
Out-of-sample, this model beats the naive baseline by 70% (RMSE 2.40 vs. 7.90 for naive) across 97 expanding-window backtested quarters — no lookahead, each quarter predicted using only prior data.
Caveat on that number: the 2020 COVID quarters (a -28% crash then a +34.9% rebound) make the naive baseline look catastrophically bad almost by definition — "assume no change" is guaranteed to fail across a V-shaped shock, regardless of model quality. Excluding those four quarters, the improvement is a more honest 29% (RMSE 1.81 vs. 2.54 over 93 quarters) — still a real improvement, just not the inflated headline number.
2026Q2 nowcast: +3.72% (±1.87 at 1 residual std dev, from a fit with R²=0.83). The official figure for this quarter has already been released; shown for reference against the model.