An economics and business concept defining a measure, method, or organizational practice used for analysis and decision-making. It specifies how information is generated or used to guide allocation of resources and evaluation of outcomes. It does not ensure correctness without clear assumptions, reliable inputs, and appropriate review of results. It materially affects planning, performance, and risk by shaping decisions and incentives within organizations and markets. The concept is generally stable, though methods and tools evolve over time.

Principle

If nonstationary series share the same stochastic trend(s), their common stochastic components can cancel in a linear combination, producing stationary residuals; cointegration justifies modeling both long-run relations and short-run adjustments via error-correction representations.

Demonstration

Two I(1) series x_t and y_t may be cointegrated if y_t − β x_t is I(0); the Engle–Granger approach estimates β by OLS and tests stationarity of residuals, while Johansen’s system method tests for the dimension of the cointegration space.

Misapplication

Testing for cointegration on poorly specified deterministic components, ignoring structural breaks, or using series of different integration orders (e.g., I(1) and I(0)) can produce spurious detection or miss true long-run relations.

Consequence

Evidence of cointegration implies the existence of long-run equilibrium constraints and validates the use of error-correction models that separate transient dynamics from stable long-term relationships, improving inference and forecast stability.

Reversal

Absence of cointegration implies no stable linear long-run equilibrium; regressions in levels among non-cointegrated I(1) variables are prone to spurious relationships and require differencing or alternative modeling.

Boundary

Cointegration typically requires that the involved series be integrated of the same order (commonly I(1)); small-sample issues, deterministic trends, regime changes, and fractional integration complicate testing and interpretation.

Semantic Tension

Cointegration is distinct from high contemporaneous correlation: series can be highly correlated but not cointegrated if their stochastic trends differ; cointegration also differs from causality, which addresses directional predictive content rather than equilibrium constraints.

Synthesis

Cointegration identifies stationary linear combinations among nonstationary series and thereby formalizes long-run equilibrium relationships; recognizing cointegration determines whether to model levels with error-correction dynamics or to difference series for short-run analysis.