Home / Ideas / Cointegration ETF Pairs Trading (Stability-Filtered)
Cointegration ETF Pairs Trading (Stability-Filtered)
trade the spread between cointegrated ETF pairs on z-score extremes
A 2025 Journal of Asset Management study tests cointegration-based pairs on ETFs over 2000–2024 across 30 pairs and multiple z-score thresholds, finding low max drawdown (~-1.02%) and positive Sharpe when cointegration is stable — but long recovery periods when it breaks. Classic Gatev/Goetzmann/Rouwenhorst stat-arb roots. Edge survives only with a rolling cointegration-stability filter.
Why it's relevant now
Record-low cross-asset correlation raises the payoff to genuinely cointegrated (not merely correlated) pairs, since spurious co-movement is scarce and true relative-value dislocations widen further before reverting.
Universe
Liquid sector/style/commodity ETF universe (e.g. XLE/XOP, XLF/KRE, GLD/GDX, EWA/EWC, IWM/IWD) — screen all pairs for Engle-Granger / Johansen cointegration on a rolling window.
How it works
On each rolling window, keep pairs passing a cointegration p-value threshold; compute the hedge-ratio spread z-score; enter long-short at |z|>2, exit at |z|<0.5, hard stop if cointegration test fails mid-trade.
Expected performance
Research-derived Sharpe estimate: 0.6–1.2. paper-implied, before the realism discount; wide band due to sensitivity to the stability filter).
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 Cointegration pairs trade on liquid sector ETFs: rolling Engle-Granger screen, enter spread at z-score above 2, exit near zero, drop pairs that lose cointegration.
Get SignalChain — $49 →
Research & sources
- link.springer.com — link.springer.com
- arxiv.org — arxiv.org
- quantoisseur.com — quantoisseur.com
More mean-reversion and statistical arbitrage 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.