Home / Ideas / Stock-bond correlation regime switch
Stock-bond correlation regime switch
flip between a diversified 60/40 and a defensive posture based on realized SPY-TLT correlation
The stock-bond correlation is unstable (ranged roughly -0.6 to +0.5) and its *sign* determines whether bonds hedge equities. It flipped positive in 2022 as monetary/inflation shocks dominated, and remained elevated into 2026 (E-piphany, June 2026). Conditioning allocation on the realized rolling correlation — rather than assuming a static negative one — is a second-moment regime signal absent from the published set.
Universe
SPY (equity) and TLT/IEF (duration) as the correlation pair and core holdings; GLD and SHY/cash as the diversifier when the equity-bond hedge breaks down.
How it works
Monthly, estimate the ~63-day rolling return correlation between SPY and TLT; when correlation is negative run a balanced risk-weighted SPY/TLT book, when correlation is positive cut bond weight and shift the hedge sleeve to GLD/SHY, rebalance monthly.
Expected performance
Research-derived Sharpe estimate: 0.4-0.8. regime-conditioning improves risk-adjusted return over static 60/40, especially in drawdown control; modest because the switch is infrequent and lagged).
Backtest this idea with SignalChain
This is a research lead — not a finished backtest. SignalChain takes an idea like this and runs the whole pipeline inside Claude Code: it researches the concept against academic and practitioner sources, sets benchmarks, writes and lints a VectorBT backtest, runs it, and grades the result PASS/FAIL. One command:
/signalchain Monthly regime switch: when 63-day SPY-TLT return correlation is negative hold risk-weighted SPY/TLT, else cut TLT and add GLD and SHY
Get SignalChain — $49 →
Research & sources
- inflationguy.blog — inflationguy.blog
- digitalcommons.ursinus.edu — digitalcommons.ursinus.edu
- arxiv.org — arxiv.org
More Macro, rates & cross-asset (ETF-based) ideas
Not financial advice. This page describes a research idea, not a recommendation. Any performance figures are hypothetical, research-derived estimates and are not indicative of future results. SignalChain is a research and educational tool; you are solely responsible for any decisions you make.