Cointegration
Published Last updated
Key Takeaway
Statistical relationship where two non-stationary cryptocurrency prices move together predictably despite individual price trends, enabling profitable mean-reversion pairs trading strategies.
Learn These First
What Is Cointegration?
Statistical relationship where two non-stationary cryptocurrency prices move together predictably despite individual price trends, enabling profitable mean-reversion pairs trading strategies.
How Cointegration Works
Frequently Asked Questions
How do I test whether two cryptocurrencies are cointegrated?
Use Engle-Granger two-step procedure: (1) Run Augmented Dickey-Fuller test on each individual price series—both must show unit roots (p-value > 0.05) confirming non-stationarity; (2) Regress one price against the other, extracting residuals; (3) Run Augmented Dickey-Fuller test on residuals—if p-value < 0.05, residuals are stationary, confirming cointegration. The hedge ratio from step 2 regression determines optimal weighting for mean-reversion trading. Cointegration software calculates automatically; you specify two price series and receive confirmation of cointegration plus optimal hedge ratio. Bitcoin-Ethereum cointegration is common; smaller altcoin pairs rarely achieve confirmed cointegration.
Why is cointegration more important than simple correlation for cryptocurrency pairs trading?
Correlation measures whether prices move together; cointegration measures whether spreads maintain stable equilibrium. Two assets can be perfectly correlated (moving identically) yet both trending upward indefinitely without equilibrium—unsuitable for mean reversion. Cointegrated pairs guarantee equilibrium: spreads deviate but systematically revert toward equilibrium, creating exploitable trading opportunities. Correlation-only pairs lack predictable reversion; mean-reversion traders deploying on correlated-but-not-cointegrated pairs experience losses. Cointegration is mathematical guarantee of mean-reverting relationship; correlation provides no such guarantee. Professional traders demand cointegration confirmation before deploying capital, ignoring correlation-only relationships.
What should I do if my cointegrated cryptocurrency pair shows deteriorating cointegration?
Deteriorating cointegration signals relationship breakdown: fundamental drivers diverging, market structure changing, or regulatory environment shifting. Monitor cointegration continuously through walk-forward testing (quarterly minimum); if p-values trend higher (approaching 0.05 or exceeding), cointegration is degrading. Upon detection: (1) reduce position size immediately, (2) increase monitoring frequency, (3) investigate underlying cause (correlations breaking down, volatility changing), (4) consider strategy pause until relationship restabilizes, (5) if cointegration disappears entirely (p > 0.05), retire the pair. Continuing to trade pairs with deteriorated cointegration produces losses because fundamental mean-reversion dynamics have collapsed.
Common Misconceptions About Cointegration
Two highly correlated cryptocurrencies are automatically cointegrated and suitable for mean-reversion trading.
Correlation and cointegration are distinct properties. Bitcoin and Ethereum might show 0.95 correlation (moving together) yet fail cointegration tests if both individually trend without equilibrium. Cointegration requires non-stationary individual prices combining into stationary spread—correlation alone doesn't guarantee this. Many traders confuse these concepts, building mean-reversion systems on correlated-but-not-cointegrated pairs that lack mean-reverting characteristics. Always test cointegration formally through Engle-Granger procedure before deploying capital. High correlation is necessary but insufficient for mean-reversion suitability.
If two cryptocurrencies were cointegrated last year, they'll remain cointegrated indefinitely.
Cointegration is time-dependent; relationships stable for years can break suddenly. Bitcoin-Ethereum cointegration might deteriorate if regulatory treatment diverges, technology development differs, or market structure shifts. Professional traders monitor cointegration continuously through rolling-window testing, not relying on historical validation. Pairs once cointegrated can transition to non-cointegration as fundamental drivers change. Walk-forward analysis reveals this degradation early, enabling proactive strategy adjustment before losses accumulate. Treat cointegration as provisional ongoing relationship requiring continuous validation.
Cointegration guarantees my mean-reversion strategy will profit.
Cointegration confirms mathematical mean-reversion property but doesn't guarantee profitability. Cointegrated spreads revert predictably, but reversion timing is unpredictable—spreads can widen dramatically before reverting. Additionally, trading costs, slippage, and execution delays may eliminate theoretical profits. Cointegration is necessary foundation for mean-reversion, not profitability guarantee. Build cointegrated strategies through rigorous backtesting with realistic cost assumptions, then validate through out-of-sample testing. Cointegration mathematical validity doesn't substitute for practical strategy validation.