Cointegration refers to a statistical property indicating a stable, long-run relationship between two or more time series variables, despite short-term deviations.
Cointegration is a statistical concept in econometrics that indicates a stable, long-run relationship between two or more time series variables, despite being individually non-stationary. When variables are cointegrated, their individual trends are aligned so that their long-term movements are connected in such a manner that any deviation from this equilibrium relationship is temporary.
Cointegration can be mathematically defined for two time series \(X_t\) and \(Y_t\) as follows:
Non-Stationarity:
Existence of a Linear Combination:
Formally, if \(X_t \sim I(1)\) and \(Y_t \sim I(1)\), then \(X_t\) and \(Y_t\) are cointegrated if there exists \(\beta\) such that:
When considering two time series, such as \(X_t\) and \(Y_t\), pairwise cointegration occurs if they share a single common stochastic trend.
This involves more than two non-stationary series (e.g., three or more variables) that may have multiple cointegrating vectors, indicating several long-run equilibrium relationships.
Prior to testing for cointegration, it’s necessary to establish that the individual time series are non-stationary through unit root tests such as the Augmented Dickey-Fuller (ADF) test or the Phillips-Perron test.
This consists of two stages:
The Johansen cointegration test allows for the identification of multiple cointegrating vectors in a system of equations, making it useful for analyzing more complex relationships.
Cointegration is extensively used in financial economics for pairs trading strategy, where securities with a stable, long-term relationship are traded to profit from temporary deviations from their long-run equilibrium.
In real estate, cointegration can help in understanding the long-term relationships between housing prices and macroeconomic indicators like interest rates or GDP.
Analyzing the cointegration between commodity prices (e.g., oil and gold) helps in developing hedging strategies and understanding market dynamics.
While correlation measures the strength and direction of a linear relationship between two variables, cointegration assesses the existence of a stable long-term equilibrium relationship despite short-run volatility.
Stationary processes have a constant mean and variance over time. In contrast, cointegrated series, though individually non-stationary, maintain a stationary linear combination.
Prioritize evidence that links Cointegration to source data, forecast assumptions, normalization adjustments, sensitivity cases, and valuation impact. The strongest evidence shows how the term changes cash flow, earnings quality, invested capital, discount rate, risk premium, or the multiple applied.
Use Cointegration when an analytical conclusion depends on a model input, adjustment, scenario, ratio, valuation method, or sensitivity. The practical issue is whether the term changes cash flow, invested capital, discount rate, terminal value, earnings quality, or risk premium.
Analysts should tie it to three model locations: the source data, the adjustment or assumption, and the output that changes. If it affects enterprise value, equity value, return on capital, leverage, margins, or comparability, show the impact explicitly. If it is qualitative, use it to frame the scenario or diligence question instead of hiding it inside a single point estimate.
For Cointegration, the decision impact is whether the analyst changes normalized earnings, cash flow, discount rate, multiple, terminal value, invested capital, or scenario weight. If the model output is unchanged, Cointegration is explanatory support rather than a valuation driver.
Verify Cointegration against the model tab, source data, normalization adjustment, peer set, discount-rate support, scenario case, and sensitivity output. Cointegration matters when value, return, leverage, margin, or comparability changes.
Trace Cointegration from source assumption to model cell, valuation bridge, sensitivity, and investment conclusion. Cointegration matters when it changes cash flow, discount rate, multiple, scenario weight, comparability adjustment, margin of safety, or explanation of why value differs from price.
The use boundary for Cointegration is reached when cash flow, discount rate, multiple, scenario weight, comparability adjustment, sensitivity, and margin of safety are unchanged. In that case, document the term as context but do not let it move valuation.
The evidence link for Cointegration is the source assumption, model cell, comparable set, sensitivity table, valuation bridge, or investment memo. Without that link, Cointegration should not move cash flow, discount rate, multiple, scenario weight, or margin of safety.
The risk check for Cointegration is whether a valuation conclusion depends on an untested assumption. Test cash-flow sensitivity, discount rate, multiple selection, peer comparability, scenario weights, terminal value, and whether the result survives a reasonable downside case.
Decision evidence for Cointegration should show the model cell, source assumption, comparable evidence, sensitivity, and valuation bridge affected. Cointegration can change valuation only when it alters cash flow, discount rate, multiple, scenario weight, or margin of safety.
Review evidence for Cointegration should make the valuation evidence traceable, not just definitional. For Cointegration, tie the evidence to the model workbook, forecast source, market data, comparable set, and management or analyst assumption file and explain why that evidence is reliable enough for the finance decision.
Before relying on Cointegration, document the decision context: the valuation date, forecast period, reporting date, and market multiple observation window. Keep the Cointegration evidence trail visible: sensitivity case, input tie-out, reviewer challenge, and support for discount rate, terminal value, or normalized earnings. In Valuation work, Cointegration matters when it changes intrinsic value, relative value, impairment analysis, deal pricing, or investment recommendation.
The practical risk for Cointegration is that valuation terms can create false precision unless assumptions, source data, and sensitivity ranges are explicit. If those facts are unavailable, keep Cointegration in the explanatory layer instead of treating it as decision-grade evidence.
Use Cointegration as a decision workflow, not a static glossary label: define the finance meaning, verify the evidence, and identify which conclusion changes. Start by linking Cointegration to forecast input, market data, comparable set, discount rate, sensitivity case, and recommendation effect. Only after those checks should Cointegration influence a valuation decision.
For Cointegration, confirm the source record, the date or jurisdiction that could change the answer, and the finance decision affected if the evidence were wrong. If those checks are incomplete, keep Cointegration as explanatory context rather than a decisive input.