← Back to archive

1. Background

This article updates our ongoing tracking of cross-asset correlations, last reported in our April 2025 analysis of correlation breakdown during stress events. The August 2025 volatility episode provides a fresh data point for assessing whether the post-2022 positive stock-bond correlation regime persists.

2. The August 2025 Event

In early August 2025, a combination of weaker-than-expected payroll data, escalating geopolitical tensions, and a sharp unwind in crowded momentum positions triggered a multi-day volatility spike. The VIX reached 38.6, its highest level since the April tariff shock. What made this event analytically interesting was the behaviour of cross-asset correlations: the stock-bond correlation, which had been oscillating around +0.15 for most of 2025, spiked to +0.45 during the acute phase — indicating that both stocks and bonds sold off simultaneously.

3. Updated Correlation Matrix

Pair2024 Avg2025 CalmAug 2025 StressΔ
SPX – UST+0.08+0.15+0.45+0.30
SPX – Gold−0.02+0.04−0.21−0.25
SPX – Crude+0.31+0.28+0.61+0.33
SPX – EM Eq+0.72+0.69+0.91+0.22
UST – Gold+0.18+0.22+0.38+0.16
Crude – EUR+0.24+0.19+0.47+0.28

Table 1: 20-day rolling correlations before and during the August 2025 stress event.

4. The Stock-Bond Regime

The positive stock-bond correlation first appeared during the 2022 rate-hiking cycle and has not fully reversed. Our rolling 60-day correlation estimate has remained positive for 80% of trading days since January 2023. The August 2025 event reinforced this regime: bonds failed to provide their traditional hedge during the equity sell-off. This has profound implications for the 60/40 portfolio and for any systematic strategy that relies on the stock-bond hedge for risk management.

import numpy as np

def rolling_correlation(x, y, window=60):
    """Compute rolling Pearson correlation."""
    T = len(x)
    corr = np.full(T, np.nan)
    for t in range(window, T):
        corr[t] = np.corrcoef(x[t-window:t], y[t-window:t])[0,1]
    return corr

def correlation_regime_probability(corr_series, threshold=0):
    """Fraction of time correlation exceeds threshold."""
    valid = corr_series[~np.isnan(corr_series)]
    return np.mean(valid > threshold)

5. Gold as the Replacement Hedge

Gold was the only major asset to show a negative correlation shift with equities during the August event (−0.21 vs. +0.04 in calm periods). This is consistent with our earlier finding that gold has replaced Treasuries as the primary flight-to-safety asset in the post-2022 positive-stock-bond-correlation regime. However, the magnitude of the hedge is smaller than the historical Treasury hedge: the equity-gold correlation during stress was −0.21 compared to the pre-2022 equity-Treasury correlation during stress of approximately −0.35.

6. DCC-GARCH Model Update

We re-estimate our Dynamic Conditional Correlation (Engle, 2002) model through September 2025. The DCC model captures the time-varying nature of correlations and provides one-step-ahead forecasts. The model predicted an equity-bond correlation of +0.28 for August 2025; the realised value was +0.45. The underestimate of 0.17 is within the model’s historical error distribution (95th percentile of forecast error is 0.22) but represents a meaningful miss for risk management purposes.

7. Portfolio Risk Implications

For a standard 60/40 equity-bond portfolio, the positive stock-bond correlation increases portfolio volatility by approximately 15% relative to the pre-2022 regime. For systematic multi-asset strategies that use correlation-based position sizing, the key implication is that calm-period correlation estimates underestimate stress-period risk — a point we have emphasised repeatedly. We recommend either using stress-period correlations as the baseline for sizing, or applying a dynamic adjustment based on the VIX level.

8. Conclusion

The August 2025 volatility event confirms that the positive stock-bond correlation regime remains intact. Gold has partially filled the hedging role previously played by Treasuries, but with smaller magnitude. Systematic traders should assume positive stock-bond correlations in their risk models until there is clear evidence of a regime reversal — which would likely require sustained disinflation and a return to the zero-interest-rate environment.

References

  1. Engle, R. (2002). "Dynamic Conditional Correlation." J. Business & Economic Statistics, 20(3), 339–350.
  2. Longin, F. and Solnik, B. (2001). "Extreme Correlation of International Equity Markets." Journal of Finance, 56(2), 649–676.
  3. Page, S. and Panariello, R.A. (2018). "When Diversification Fails." Financial Analysts Journal, 74(3), 19–32.
  4. Campbell, J.Y., Sunderam, A. and Viceira, L.M. (2017). "Inflation Bets or Deflation Hedges?" Journal of Finance, 72(4), 1529–1563.