Portfolio Beta Isn't a Constant: A 20-Year Test of the 60/40 Risk Narrative

A 60/40 stock-bond portfolio's beta to the S&P 500 isn't 0.60. Over the last 20 years, it ranged from 0.30 to 0.79 — a 2.6× spread — driven almost entirely by the stock-bond correlation regime, not by the allocation itself. Thirty-five percent of the time, the rolling beta sat more than 20% above or below its full-sample average of 0.51.

If you're using a single historical beta to estimate how much equity risk your diversified portfolio actually carries, you're using yesterday's correlation to forecast tomorrow's risk. The correlation isn't stable. The beta can't be either.

What Beta Is Supposed to Tell You

Beta measures how sensitive a portfolio's returns are to movements in a benchmark — usually the S&P 500. A beta of 1.0 means the portfolio moves in lockstep with the market. A beta of 0.5 means it moves half as much. It's the single number most risk models use to translate "how much equity exposure do I have?" into a clean, comparable figure.

The appeal is obvious. One number captures systematic risk. You can compare a 60/40 portfolio to a 70/30 portfolio, to a hedge fund, to a risk-parity strategy — all on the same scale. The problem is that beta is a realized statistic. It's estimated from historical returns. And the inputs that determine it — specifically, the correlation between the portfolio's components and the market — are not constants. They're regime-dependent variables.

The Mechanism: Why a Fixed Allocation Produces a Moving Beta

A 60/40 portfolio holds 60% stocks (SPY) and 40% long-term Treasuries (TLT). The equity portion contributes a beta of 0.6 — that part is stable, since SPY's beta to itself is always 1.0. The bond portion's contribution is where the instability lives.

Portfolio beta decomposes as:

β_portfolio = 0.6 × β_SPY + 0.4 × β_TLT

where β_TLT (TLT's beta to SPY) is itself a product of two things:

β_TLT = ρ_SPY,TLT × (σ_TLT / σ_SPY)

The volatility ratio (σ_TLT / σ_SPY) moves slowly. Over our 2006–2026 sample, it ranged from about 0.59 to 1.22 — a factor of about 2×, but it changes gradually over years. The correlation (ρ_SPY,TLT) is the volatile input. It swung from -0.77 to +0.32 — a 1.08-point range — over the same period. When correlation is deeply negative, TLT's beta is negative, and the bond sleeve reduces portfolio beta below 0.6. When correlation turns positive, TLT's beta goes positive, and the bond sleeve adds to portfolio beta above 0.6.

This is the mechanism: the allocation is fixed at 60/40, but the risk contribution of the 40% bond sleeve flips sign depending on whether bonds are diversifying or correlating with equities. A single beta number averages over both regimes and tells you neither.

The Evidence: 20 Years of Rolling Beta

We computed the rolling 252-day (one-year) beta of a static 60/40 SPY/TLT portfolio against SPY using daily adjusted-close data from Tiingo, covering January 2006 through August 2026. The full-period backtest of the 60/40 portfolio produced a CAGR of 8.66%, annualized volatility of 11.26%, Sharpe of 0.79, and max drawdown of -29.9%. For comparison, SPY buy-and-hold returned 11.13% CAGR with 19.23% volatility, a 0.65 Sharpe, and -55.2% max drawdown. The 60/40 portfolio's full-sample beta to SPY was 0.51.

But that 0.51 is an average over fundamentally different regimes.

The blue line is the rolling portfolio beta. The dashed orange line is the rolling SPY-TLT correlation. They move together — and they should, because the correlation is the driver. When correlation deepens to -0.60 (2008–2009, 2014–2019), portfolio beta drops toward 0.30–0.45. When correlation climbs toward zero or positive (2022–2024), portfolio beta rises toward 0.63–0.67.

Sub-Period Decomposition

Read the table column by column. The allocation never changed — 60% SPY, 40% TLT, every day. But the portfolio beta was 50% higher in 2022 than in 2010–2019 (0.63 vs. 0.42). The entire shift came from the correlation column moving from -0.48 to +0.09. TLT's beta to SPY went from -0.45 (a diversifier reducing equity risk) to +0.07 (a co-mover adding equity risk). The 40% bond sleeve's risk contribution flipped sign.

During the 2010–2019 negative-correlation regime, the 60/40 portfolio's beta sat at 0.42 — meaning an investor looking at risk through a beta lens would have concluded the portfolio carried less than half the market's systematic risk. During the 2022–2024 positive-correlation regime, that same portfolio's beta was 0.65 — more than half the market's risk. Same portfolio. Different beta. Different risk conclusion.

Why This Isn't Just a 2022 Story

The 2022 stock-bond selloff brought this issue into the spotlight because it was the first year on record when both the S&P 500 and long-term Treasuries produced double-digit losses simultaneously. The closest prior case was 1969, when the S&P 500 lost 8.5% and long-term Treasuries lost 5.1% — painful, but not double-digit for either. As research summarized by Alpha Architect notes, the monthly stock-bond correlation was -0.28 from January 2000 through December 2021, but +0.18 from 1926 through 1999. The negative correlation that modern investors took as structural was, in historical terms, the anomaly.

The 2023–2026 data confirms this isn't a one-off. The rolling correlation has remained near zero to slightly positive through August 2026. Portfolio beta has remained elevated at 0.64–0.65. If you built a risk model in 2015 using the trailing 10-year beta of 0.42, you would have underestimated your equity exposure by roughly 50% for the next decade.

Research by Molenaar, Sénéchal, Swinkels, and Wang published in the Financial Analysts Journal (2024) confirms the mechanism: inflation, real rates, and government creditworthiness are the primary drivers of the stock-bond correlation. Their U.S. data going back to 1875 shows the correlation averaged +0.35 from 1970–1999 and -0.31 from 2000–2022. A 2025 CFA Institute Research Foundation brief by Baumann, Nazemi, and Fabozzi reached similar conclusions using machine learning on macroeconomic data: the stock-bond correlation is regime-dependent, and the regime is driven by inflation and monetary policy conditions — not by a permanent structural relationship. The basic intuition — that beta measures an asset's sensitivity to market movements — is sound; the problem is that the sensitivity itself is non-stationary, as the standard definition of beta as cov(R_i, R_m) / var(R_m) makes clear when you remember that the covariance term is the unstable part.

What Breaks Beta as a Risk Measure

Beta's failure mode isn't poor math — it's a stability assumption applied to an unstable input. Three specific problems:

• Window sensitivity. The 252-day beta we computed ranged from 0.30 to 0.79. A 63-day window ranged even wider: 0.21 to 0.89. Your "beta" depends on how far back you look, and no window length is correct — they all average over different regime mixes.
• Non-stationarity. The correlation regime isn't a mean-reverting noise process. It's driven by structural macroeconomic conditions (inflation regime, monetary policy framework). A 20-year sample that happens to span one regime will produce a beta that's wrong for the next regime.
• Aggregation masking. A full-sample beta of 0.51 sounds precise. It's actually the average of a 0.30–0.79 range. Reporting the average without the range is like reporting average temperature without mentioning summer and winter exist.

The honest use of beta is as a current-state diagnostic: "what is my portfolio's equity sensitivity right now, given the current correlation regime?" Not as a fixed parameter for forward-looking risk estimation.

Where This Meets Our Work

This is why RiskHarvest's methodology doesn't rely on a static beta target. The premia-decomposition test we apply to every holding examines what risk premium each asset actually contributes — and whether a "diversifier" is genuinely diversifying or just diluting the same equity premium under a different label. The stock-bond correlation shift of 2022 is the textbook case: TLT was a genuine diversifier when correlation was -0.48 (it contributed a negative beta, reducing portfolio equity risk) and a redundant co-mover when correlation flipped positive (it added beta without adding a different premium). A static beta model can't distinguish those states; a premia decomposition can.

Our portfolio-level rebalance trigger is the second place this matters. Rather than rebalancing on a fixed calendar (which implicitly assumes risk contributions are stable), the trigger responds to whether the portfolio's risk structure has drifted beyond its target — including drift driven by correlation regime change. When the bond sleeve stops diversifying, the portfolio's effective equity exposure rises even if the allocation hasn't moved. A trigger that monitors risk structure, not just weights, catches that.

The Practical Takeaway

The one thing to walk away with: a portfolio's beta is a snapshot of one correlation regime, not a permanent property of the allocation.

If you use beta to estimate risk, check it under multiple regimes. Compute rolling beta over 1-year and 3-year windows and look at the range, not just the average. If the range spans 50%+ of the average — as ours did — your risk estimate is only as good as your assumption about which correlation regime you're in. And that assumption is a macro call, not a portfolio construction input.

For diversified portfolios with a bond sleeve, the question isn't "what's my beta?" It's "what's my beta if the stock-bond correlation stays negative, and what's my beta if it doesn't?" If you can't live with the second answer, the allocation is riskier than the first beta suggested — and the first beta is the one your risk model is using.

Educational only. Not financial advice. Past performance does not guarantee future results. Data: Tiingo daily adjusted closes, SPY and TLT, January 2006 – August 2026. Backtest assumes 10 bps round-trip transaction costs.

Sources

• Molenaar, R., Sénéchal, E., Swinkels, L., & Wang, Z. (2024). "Empirical Evidence on the Stock–Bond Correlation." Financial Analysts Journal, 80(3), 17–36.
• Swedroe, L. (2023). "Implications of Regime-Shifting Stock-Bond Correlation." Alpha Architect.
• Baumann, F., Nazemi, A., & Fabozzi, F. (2025). "Macroeconomic Drivers of Stocks and Bonds." CFA Institute Research Foundation.
• Investopedia: "What Beta Means for Investors."
• Wikipedia: "Beta (finance)."
• Price data: Tiingo via research-service, SPY and TLT daily adjusted closes, 2006-01-03 to 2026-08-21.