Sign up
ContentsHideShow

Quant glossary · Time series

Cointegration residual

A cointegration residual is the estimated deviation from a long-run level relation between nonstationary series.

Definition

A cointegration residual is the estimated deviation from a long-run level relation between nonstationary series. In an Engle-Granger setup, it is the residual from a levels regression, including any specified intercept or trend. Cointegration requires this residual to be stationary under an appropriate residual-based test.

How it is calculated

e^t=yt−a^−β^xt\hat e_t=y_t-\hat a-\hat\beta x_t

Estimated cointegration residual e-hat at t = y_t minus estimated intercept a-hat minus estimated slope beta-hat times x_t.

Symbol Meaning Unit
e cointegration residual y units
y dependent level series level units
x explanatory level series level units
a intercept y units
beta long-run slope y units per x unit
  1. Fit the specified levels relation.

  2. Subtract its fitted intercept and slope contribution to obtain the estimated residual.

  3. Test that residual with the appropriate residual-based unit-root procedure.

Interpretation

The residual is the current distance from the fitted long-run line. It has a time-invariant marginal distribution under stationarity even though each level can wander. Estimating the line and testing the residual are linked steps, not separate evidence from an arbitrary spread.

Assumptions and conventions

  • The levels regression's deterministic terms are specified before testing.

  • The series are integrated in a compatible order, commonly I(1).

  • Residual stationarity is assessed with the appropriate Engle-Granger critical values and lag treatment.

The residual and intercept use y units; beta has y units per x unit.

Limitations and common misreadings

  • A visually mean-reverting spread is not enough to establish cointegration.

  • The residual's stationarity test has nonstandard critical values after estimating the relation.

References

  1. Robert F. Engle, C. W. J. Granger. (1987). Co-Integration and Error Correction: Representation, Estimation, and Testing. Econometrica. Theorem 1 and Section 2. doi:10.2307/1913236. Source

Cointegration residual FAQ

What is Cointegration residual?

A cointegration residual is the estimated deviation from a long-run level relation between nonstationary series.

How is Cointegration residual calculated?

Estimated cointegration residual e-hat at t = y\_t minus estimated intercept a-hat minus estimated slope beta-hat times x\_t.

What does Cointegration residual not tell you or assume?

A visually mean-reverting spread is not enough to establish cointegration.

Browse all glossary termsTime series termsQuant research library

Definitions are educational and descriptive, not investment advice. Worked examples use labelled sample data.