Aller au contenu principal

Cookies sur Macro by Mark.Nous utilisons des cookies essentiels pour la connexion, la langue et la sécurité. Avec votre accord, nous ajoutons des cookies optionnels de préférences, d’analyse et de mesure publicitaire avec le pixel Meta (Facebook). Lisez notre politique relative aux cookies.

Macro by MarkMacro by Mark
  • Accueil
Chargement…
Chargement…
← Modèles
MacroConceptsPolitiqueModèles

VECM Model

Vector error-correction model - for cointegrated macro series where long-run equilibrium relationships pull the system back.

Vue d’ensembleGraphiqueMathématiquesRéférences

Understand the mathematics

Follow the same equations and assumptions one section at a time or in a continuous reading.

Hover or focus a symbol defined in this model to find its exact matches. Select it to keep the highlight; Escape clears the selection.

The Granger Representation Theorem

Start from a VAR(p)\text{VAR}(p)VAR(p) in levels: yt=A1yt−1+…+Apyt−p+c+uty_t = A_1 y_{t-1} + \ldots + A_p y_{t-p} + c + u_tyt​=A1​yt−1​+…+Ap​yt−p​+c+ut​, where yty_tyt​ is n×1n \times 1n×1 and the variables are I(1)I(1)I(1). Subtract yt−1y_{t-1}yt−1​ from both sides and rearrange to obtain the VECM representation. The Granger representation theorem states that if the variables are cointegrated with rank rrr, then three equivalent conditions hold simultaneously: (1) there exists an error correction representation with a rank-rrr long-run matrix Π\PiΠ, (2) there exists a common trends representation with n−rn - rn−r stochastic trends, and (3) the long-run impact matrix C(1)C(1)C(1) of the Wold representation of Δyt\Delta y_tΔyt​ has reduced rank n−rn - rn−r, with β′C(1)=0\beta'C(1) = 0β′C(1)=0 - the cointegrating combinations annihilate the permanent component.

The theorem provides the structural justification for the VECM: cointegration among I(1) variables is equivalent to the existence of an error correction mechanism. Any system with cointegration must have error correction; any system with error correction must be cointegrated.

yt=A1yt−1+⋯+Apyt−p+c+ut\mathbf{y}_t = \mathbf{A}_1 \mathbf{y}_{t-1} + \cdots + \mathbf{A}_p \mathbf{y}_{t-p} + \mathbf{c} + \mathbf{u}_tyt​=A1​yt−1​+⋯+Ap​yt−p​+c+ut​
ut\mathbf{u}_tut​

Starting point. The companion matrix has some eigenvalues on or near the unit circle.

Δyt=Πyt−1+∑i=1p−1ΓiΔyt−i+c+ut\Delta \mathbf{y}_t = \boldsymbol{\Pi} \mathbf{y}_{t-1} + \sum_{i=1}^{p-1} \boldsymbol{\Gamma}_i \Delta \mathbf{y}_{t-i} + \mathbf{c} + \mathbf{u}_tΔyt​=Πyt−1​+i=1∑p−1​Γi​Δyt−i​+c+ut​
Π\boldsymbol{\Pi}ΠΓi\boldsymbol{\Gamma}_iΓi​ut\mathbf{u}_tut​

VECM representation. Subtracting yt−1y_{t-1}yt−1​ and collecting terms yields the long-run matrix Π\PiΠ.

Π=αβ′,rank(Π)=r<n\boldsymbol{\Pi} = \boldsymbol{\alpha} \boldsymbol{\beta}', \quad \text{rank}(\boldsymbol{\Pi}) = r < nΠ=αβ′,rank(Π)=r<n
Π\boldsymbol{\Pi}Πα\boldsymbol{\alpha}αβ\boldsymbol{\beta}βrrr

The reduced-rank restriction. Π\PiΠ has rank rrr because there are rrr cointegrating relationships and n−rn - rn−r common stochastic trends.

yt=C∑s=1tus+C∗(L)ut+P0\mathbf{y}_t = \mathbf{C} \sum_{s=1}^{t} \mathbf{u}_s + \mathbf{C}^*(L) \mathbf{u}_t + \mathbf{P}_0yt​=Cs=1∑t​us​+C∗(L)ut​+P0​
ut\mathbf{u}_tut​

Beveridge-Nelson-like decomposition. The common stochastic trends are α⊥′∑s=1tus\alpha_\perp' \sum_{s=1}^{t} u_sα⊥′​∑s=1t​us​, the cumulated innovations loaded by α⊥′\alpha_\perp'α⊥′​.

Subtract yt−1y_{t-1}yt−1​ from both sides of the VAR

Reparameterize the levels VAR into a first-difference form with an error correction term
yt−yt−1=(A1−I)yt−1+A2yt−2+⋯+Apyt−p+c+ut\mathbf{y}_t - \mathbf{y}_{t-1} = (\mathbf{A}_1 - \mathbf{I})\mathbf{y}_{t-1} + \mathbf{A}_2 \mathbf{y}_{t-2} + \cdots + \mathbf{A}_p \mathbf{y}_{t-p} + \mathbf{c} + \mathbf{u}_tyt​−yt−1​=(A1​−I)yt−1​+A2​yt−2​+⋯+Ap​yt−p​+c+ut​
ut\mathbf{u}_tut​

Collect terms by adding and subtracting lagged differences

Isolate the long-run matrix Π\PiΠ from the short-run Γ\GammaΓ matrices
Δyt=−(I−A1−⋯−Ap)⏟Πyt−1+∑i=1p−1−(Ai+1+⋯+Ap)⏟ΓiΔyt−i+c+ut\Delta \mathbf{y}_t = \underbrace{-(\mathbf{I} - \mathbf{A}_1 - \cdots - \mathbf{A}_p)}_{\boldsymbol{\Pi}} \mathbf{y}_{t-1} + \sum_{i=1}^{p-1} \underbrace{-(\mathbf{A}_{i+1} + \cdots + \mathbf{A}_p)}_{\boldsymbol{\Gamma}_i} \Delta \mathbf{y}_{t-i} + \mathbf{c} + \mathbf{u}_tΔyt​=Π−(I−A1​−⋯−Ap​)​​yt−1​+i=1∑p−1​Γi​−(Ai+1​+⋯+Ap​)​​Δyt−i​+c+ut​
ut\mathbf{u}_tut​

Determine the rank of Π\PiΠ

If all variables are I(1)I(1)I(1) and some are cointegrated, Π\PiΠ must have reduced rank 0<r<n0 < r < n0<r<n
rank(Π)=r⇔there are r cointegrating relationships and n−r common stochastic trends\text{rank}(\boldsymbol{\Pi}) = r \Leftrightarrow \text{there are } r \text{ cointegrating relationships and } n-r \text{ common stochastic trends}rank(Π)=r⇔there are r cointegrating relationships and n−r common stochastic trends
Π\boldsymbol{\Pi}Πrrr

Factor Π\PiΠ into αβ′\alpha \beta'αβ′

The factorization separates adjustment dynamics (α)(\alpha)(α) from equilibrium relationships (β)(\beta)(β)
Π=αβ′⇒Δyt=α(β′yt−1)+∑i=1p−1ΓiΔyt−i+c+ut\boldsymbol{\Pi} = \boldsymbol{\alpha} \boldsymbol{\beta}' \Rightarrow \Delta \mathbf{y}_t = \boldsymbol{\alpha}(\boldsymbol{\beta}'\mathbf{y}_{t-1}) + \sum_{i=1}^{p-1} \boldsymbol{\Gamma}_i \Delta \mathbf{y}_{t-i} + \mathbf{c} + \mathbf{u}_tΠ=αβ′⇒Δyt​=α(β′yt−1​)+i=1∑p−1​Γi​Δyt−i​+c+ut​
Π\boldsymbol{\Pi}Πα\boldsymbol{\alpha}αβ\boldsymbol{\beta}βΓi\boldsymbol{\Gamma}_iΓi​ut\mathbf{u}_tut​

Derive the common trends representation

The n−rn - rn−r stochastic trends driving the system are spanned by α⊥′∑s=1tus\alpha_\perp' \sum_{s=1}^{t} u_sα⊥′​∑s=1t​us​, the cumulated innovations
yt=β⊥(α⊥′Γβ⊥)−1α⊥′∑s=1tus+I(0) terms\mathbf{y}_t = \boldsymbol{\beta}_{\perp}(\boldsymbol{\alpha}_{\perp}' \boldsymbol{\Gamma} \boldsymbol{\beta}_{\perp})^{-1} \boldsymbol{\alpha}_{\perp}' \sum_{s=1}^{t} \mathbf{u}_s + I(0) \text{ terms}yt​=β⊥​(α⊥​′Γβ⊥​)−1α⊥​′s=1∑t​us​+I(0) terms
β⊥\boldsymbol{\beta}_{\perp}β⊥​α⊥\boldsymbol{\alpha}_{\perp}α⊥​

Verify that β′yt\beta' y_tβ′yt​ is stationary

The cointegrating combinations must be I(0) for the error correction mechanism to function
β′yt=β′C∑s=1tus+I(0)=0⋅∑s=1tus+I(0)(since β′C=0)\boldsymbol{\beta}'\mathbf{y}_t = \boldsymbol{\beta}'\mathbf{C}\sum_{s=1}^t \mathbf{u}_s + I(0) = \mathbf{0} \cdot \sum_{s=1}^t \mathbf{u}_s + I(0) \quad (\text{since } \boldsymbol{\beta}'\mathbf{C} = \mathbf{0})β′yt​=β′Cs=1∑t​us​+I(0)=0⋅s=1∑t​us​+I(0)(since β′C=0)
β\boldsymbol{\beta}β

The role of the orthogonal complements

The orthogonal complements α⊥\alpha_\perpα⊥​ and β⊥\beta_\perpβ⊥​ identify the common trends. α⊥′ut\alpha_\perp' u_tα⊥′​ut​ are the permanent shocks (they have a long-run impact); α′ut\alpha' u_tα′ut​ are the transitory shocks (their effects die out). β⊥\beta_\perpβ⊥​ spans the directions in which the system is nonstationary.

Notations

Π\boldsymbol{\Pi}Π

Long-run impact matrix

α⊥\boldsymbol{\alpha}_{\perp}α⊥​

Loading complement

β⊥\boldsymbol{\beta}_{\perp}β⊥​

Cointegration complement

About these symbols

Long-run impact matrix: Long-run impact matrix: Π=αβ′\Pi = \alpha \beta'Π=αβ′

Loading complement: n×(n−r)n \times (n-r)n×(n−r) orthogonal complement of α\alphaα, spanning the common trend loadings

Cointegration complement: n×(n−r)n \times (n-r)n×(n−r) orthogonal complement of β\betaβ, spanning the nonstationary directions

    Transition checks

    Sampled dynamics
    zt=G(zt−1;θ)+ηt,yt=Hzt+εtz_t=G(z_{t-1};\theta)+\eta_t,\qquad y_t=H z_t+\varepsilon_tzt​=G(zt−1​;θ)+ηt​,yt​=Hzt​+εt​
    HHH

    The state is updated at observed dates. A line between dates is a display convention unless the model specifies behavior inside the interval.

    Read adjacent points as adjacent sample calculations. Smoothness is a property of the estimator, not of the chart line.

    Do not infer within-period behavior from a sampled estimator.

    Variable timing

    Latent state

    zt=G(zt−1;θ)+ηtz_t=G(z_{t-1};\theta)+\eta_tzt​=G(zt−1​;θ)+ηt​

    The state updates at sample dates. The visible line is only as smooth as the estimated state sequence.

    Observation equation

    yt=Hzt+εty_t=Hz_t+\varepsilon_tyt​=Hzt​+εt​
    HHH

    Observed series can move differently from latent states because measurement noise and loadings matter.

    Keep going

    Read next

    Continue with a related concept, model, or example.

    Theory model

    IS-LM Model

    a well-connected theoretical model two steps from here

    New angle: theoretical model →
    Macro by Mark
    Aide
    Politique de confidentialitéConditions d’utilisationÉthique et conformité

    © 2026 Mark Jayson Nation LLC