Pairs Trading: Cointegration vs Plain Correlation — When Spreads Are Mean-Reverting
Engle–Granger intuition, half-life of spread, why high correlation does not imply tradable stationarity of the spread.
Higher education in Financial Engineering and Money & Capital Markets. SPK (Turkey CMB) licence. 16 years across institutional markets, research, and quant-driven analytics.
Correlation is a weak gate
Two series can exhibit high Pearson correlation yet their spread \(S_t = P^{(1)}_t - \beta P^{(2)}_t\) may still be non-stationary. Correlation measures co-movement, not equilibrium error correction.Cointegration as an equilibrium story
Cointegrated price series share a linear combination that is mean-reverting — representing a statistical fair value tether. This is the foundational assumption behind many venue-neutral stat-arb desks.
Estimation hygiene
Two-step Engle–Granger versus Johansen rank tests address different dimensions (pairwise vs baskets). Residual diagnostics require heteroskedasticity-robust standard errors.
Practical trading layer
Even with cointegration, transaction costs may destroy edge. Model half-life vs average holding cost drag.
Finvestopia context
When our macro or cross-asset charts show tight linkages, distinguish narrative correlation from tradeable mean reversion windows — the library entry encodes that discipline.Related entries
Continuous-time OU mapping, discrete AR(1) estimation, horizons vs turnover and transaction-cost floors.
Unit-root null vs stationarity null, size/power trade-offs, structural breaks & rolling windows for live models.
Educational content authored by our team — informational only, not investment advice.