Engle-Granger Test
Published Last updated
Key Takeaway
Two-step statistical procedure testing whether two non-stationary cryptocurrency prices maintain equilibrium relationship through cointegration, validating mean-reversion pair trading viability.
Learn These First
What Is Engle-Granger Test?
Two-step statistical procedure testing whether two non-stationary cryptocurrency prices maintain equilibrium relationship through cointegration, validating mean-reversion pair trading viability.
How Engle-Granger Test Works
Frequently Asked Questions
How do I run Engle-Granger test on cryptocurrency pairs?
Step 1: Obtain price data for both cryptocurrencies (minimum 100 observations, preferably 250+ for stability). Step 2: Run Augmented Dickey-Fuller test on both price series—confirm both show unit roots (p > 0.05). Step 3: Regress price1 against price2, extracting residuals. Step 4: Run Augmented Dickey-Fuller test on residuals. If p < 0.05, cointegration is confirmed and hedge ratio from Step 3 regression provides optimal weighting. Most Python libraries (statsmodels) include Engle-Granger implementations calculating automatically. Alternative: use specialized trading platforms offering built-in cointegration screening tools testing hundreds of cryptocurrency pairs simultaneously.
What Engle-Granger test p-value should I require for cryptocurrency pairs?
Standard statistical threshold is p < 0.05 indicating 95% confidence in cointegration. For cryptocurrency trading, more conservative threshold of p < 0.01 reduces false positives when screening many candidate pairs. Some traders use stepped approach: initial screening at p < 0.05, then additional validation through multiple windows before deployment. The hedging ratio strength also matters: stronger cointegration relationships show lower p-values and more stable hedge ratios. Bitcoin-Ethereum cointegration typically shows p-values 0.001-0.02 (very strong); altcoin pairs show 0.01-0.05 (moderate). Pairs near 0.05 threshold represent marginal cointegration deserving caution.
Why would an Engle-Granger test show cointegration one quarter but not the next?
Cryptocurrency market fundamentals shift: regulatory changes, technology developments, or adoption patterns affecting assets differently can break cointegration. Bitcoin-Ethereum cointegration might deteriorate if Ethereum's staking developments diverge from Bitcoin's trajectory. Market regime changes alter correlation structures: bull markets show different relationships than bear markets. Walk-forward testing across multiple quarters reveals whether cointegration persists or degrades. Declining cointegration signals relationship weakening; deteriorating p-values (increasing toward 0.05) warn of approaching cointegration collapse. Professional traders interpret Engle-Granger results as time-dependent relationship health indicators requiring continuous monitoring, not permanent validations.
Common Misconceptions About Engle-Granger Test
If Engle-Granger test shows cointegration, my pairs trading strategy will automatically profit.
Engle-Granger confirmation establishes mathematical mean-reversion foundation but guarantees neither profitability nor specific trade timing. Cointegrated spreads revert predictably but timing is unpredictable—spreads can widen substantially before reverting. Trading costs, slippage, and execution delays reduce theoretical profits. Cointegration is necessary foundation for mean-reversion strategy viability, not profitability guarantee. Build strategies through rigorous backtesting on cointegrated pairs, then validate through out-of-sample testing. Cointegration mathematical validity doesn't substitute for comprehensive strategy validation.
Engle-Granger p-value of 0.04 is practically identical to p-value of 0.01 for pairs trading purposes.
P-value proximity hides important strength differences: p < 0.01 represents 99% confidence in cointegration; p = 0.04 represents 96% confidence. This seemingly small difference indicates substantially weaker statistical evidence. Pairs near 0.05 threshold represent marginal cointegration showing high deterioration risk; p < 0.01 pairs show robust persistent cointegration. Professional traders apply different trading strategies to p < 0.01 pairs (more aggressive positioning) versus 0.01-0.05 pairs (conservative sizing). The difference matters practically despite statistical proximity.
Engle-Granger test is too technical and unnecessary for practical cryptocurrency trading.
Engle-Granger testing is standard validation separating professional traders from amateurs. Many retail traders skip formal testing, deploying strategies on correlated-but-not-cointegrated pairs lacking mean-reversion characteristics, experiencing losses confirming cointegration's importance. Professional quantitative teams universally employ Engle-Granger methodology. Modern software makes testing straightforward; technical complexity is no longer a barrier. Skipping Engle-Granger testing amounts to building mean-reversion strategies on potentially non-reverting relationships—a recipe for capital losses.