Correlation matrix · daily macro_monitor archive (90-day rolling window) → standardise each dimension to z-score → pairwise Pearson r. Diagonal = 1 by construction.

Weight optimisation · OLS regression of SPX 3-month-forward log return on the 12 standardised dimension scores. Coefficients re-scaled to sum to 100 (positive only — negative coefficients reduced toward 0 to preserve composite-score interpretability). Out-of-sample R² estimated via 5-fold time-series cross-validation.

Lead / lag · for each dimension i, compute Pearson r between today’s value and SPX return over the next 63 trading days. Positive r = leading (today’s reading predicts forward equity move). Near-zero r = lagging or noise.

Caveats · Backtest window is short and includes the late-cycle stress regime currently dominating; weights may shift materially in expansionary regimes. Geopolitical r underestimates true tail risk — rare events under-represented in 90-day window. Recommend regime-conditional re-fits when sample size permits.