Correlation regimes: when the stock–bond correlation flips

What you will learn

• What a correlation regime is and why the stock–bond correlation is not a constant
• How inflation and growth shocks decide whether bonds hedge equities
• How to read a rolling-correlation table without over-interpreting noise
• How much a regime flip changes the volatility and bad-month loss of a 60/40 portfolio

Explanation

A correlation regime is a period of several years in which the correlation between two assets stays in a recognisable range. How to compute a correlation coefficient from returns was covered in the course Forex & Currency Trading in Depth; here we apply it to allocation. The most important regime for a multi-asset investor is the stock–bond correlation, the correlation between returns of broad equities and government bonds.

The historical observation is this: from roughly the late 1990s to 2021 the correlation between US and European equities and their government bonds was mostly negative, so bonds rose when stocks fell and worked as a hedge. In earlier decades with high and volatile inflation, such as the 1970s and 1980s, the correlation was mostly positive. In 2022 broad global equities and government bonds both fell in the same year, and the correlation turned positive again. Past regimes do not tell us the next one, but they show that the sign can change.

The economic logic uses two types of shock. A growth shock (recession fear) lowers expected profits and lowers expected interest rates: stocks fall, bonds rise, correlation negative. An inflation shock raises expected interest rates, which lowers the present value of both bond coupons and company profits: both fall, correlation positive. Which shock dominates depends on the inflation regime: when inflation is low and stable, growth news dominates; when inflation is high or uncertain, inflation news dominates and bonds stop hedging.

A rolling correlation is calculated over a moving window, for example the last 36 monthly returns, and recalculated each month. Three reading rules: a 36-month window reacts slowly, so a flip appears about a year after it begins; 12-month windows are noisy and one crash month can push them from +0.3 to −0.5; judge the level over several years, not the last point.

Worked example

Hypothetical example: a 60/40 portfolio of global equities (assumed volatility 16%) and euro government bonds (assumed volatility 6%). We compare three correlation assumptions: −0.3 (hedging regime), 0.0 and +0.5 (inflation regime). Fees, spreads and taxes excluded.

Variance = 0.6² × 0.16² + 0.4² × 0.06² + 2 × 0.6 × 0.4 × ρ × 0.16 × 0.06. The first two terms are 0.009216 + 0.000576 = 0.009792; the cross term is 0.004608 × ρ.

The same capital weights produce about 20% more volatility in the positive-correlation regime than in the negative one. On a €250,000 portfolio a one-standard-deviation year is about €23,000 versus €27,500.

Bad-month check. Suppose equities fall 8% in a month. In the hedging regime bonds might gain 2%: portfolio return = 0.6 × (−8%) + 0.4 × (+2%) = −4.0%, or −€10,000 on €250,000. In the inflation regime bonds might fall 2%: 0.6 × (−8%) + 0.4 × (−2%) = −5.6%, or −€14,000. The bond block changed from a cushion into an additional loss of €2,000 instead of a gain of €2,000.

Reading a rolling table. Imagine 36-month rolling correlations of −0.35, −0.30, −0.25, −0.10, +0.15, +0.30 at six-month intervals. The steady climb across five readings is a regime signal; a single jump from −0.3 to +0.2 followed by a return to −0.3 would be noise from one extreme month entering the window.

Common mistake

Practice activity

A 50/50 portfolio holds equities with 18% volatility and long-duration government bonds with 10% volatility. Calculate the portfolio volatility for ρ = −0.4 and ρ = +0.6. Then state which single change (shorter bond duration, lower equity share, or adding a third asset with low correlation to both) reduces the dependence on the regime most directly, and why.

Key takeaways

• The stock–bond correlation depends on which shock dominates: growth shocks make it negative, inflation shocks make it positive.
• A flip from −0.3 to +0.5 raises the volatility of a 60/40 portfolio by roughly a fifth and removes the bond cushion in bad months.
• Read rolling correlations by trend across several readings; a 36-month window lags by about a year.
• Size equity so that the inflation-regime outcome is still within your tolerance.

Next step

Volatility is one half of an allocation decision; the other half is the expected return of each asset, and the next lesson shows how to estimate it honestly from valuations and yields.