Macro Financial Modeling Diagnostics
Quarterly macro series, regression-trap diagnostics, and PCA rate-factor linkages across 105 observations.
The normalized line chart shows why macro-financial panels can be deceptive in levels: real GDP and CPI rise steadily over the full sample, while rates and unemployment cycle sharply around recessions and policy pivots. Common trend does not necessarily imply a stable predictive relationship.
The bar diagnostics quantify two separate failure modes. First, a very high in-sample fit in levels weakens substantially after differencing, consistent with a spurious regression problem. Second, allowing lookahead information produces a much stronger R² than a lagged-data specification, illustrating how timing misalignment can exaggerate model performance.
- Factor 1 is tightly tied to Fed Funds and also strongly linked to the 10-year yield, indicating a broad level/rate-regime component.
- Factor 2 has a more moderate relationship with the 10-year yield and a weak inverse relationship with Fed Funds, suggesting a different slope or curve-shape dimension.
- The combined view separates trend, timing, and factor-interpretation risks in one panelized diagnostic report.
| Metric | Value |
|---|---|
| Spurious regression R² (levels) | 98.1% |
| Spurious regression adj. R² (levels) | 98.0% |
| Corrected regression R² (differences) | 19.0% |
| Corrected regression adj. R² (differences) | 16.6% |
| Naive lookahead R² | 80.0% |
| Naive lookahead adj. R² | 79.4% |
| Realistic lagged-data R² | 18.6% |
| Realistic lagged-data adj. R² | 16.2% |
| R² degradation after removing lookahead | +61.3pp |
| Corr: Factor 1 vs Fed Funds | +0.985 |
| Corr: Factor 1 vs 10Y yield | +0.867 |
| Corr: Factor 2 vs 10Y yield | +0.499 |
| Corr: Factor 2 vs Fed Funds | -0.175 |