Skip to main content

Cookies on Macro by Mark.We use essential cookies for sign-in, locale, and security. With your permission, we also use optional cookies for preferences, for analytics, and for advertising measurement with the Meta (Facebook) Pixel. Read our cookie policy.

Macro by MarkMacro by Mark
  • Home
Loading…
Loading…
← Models
MacroConceptsPolicyModels

Threshold VAR Model

VAR with regime-dependent dynamics, where impulse responses change after a threshold variable crosses a specified boundary.

OverviewGraphMathReferences

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.

Model Specification and Regime Structure

The Threshold VAR partitions a standard reduced-form VAR into MMM regimes governed by an observable threshold variable. We work with the two-regime case; extension to M>2M > 2M>2 is conceptually identical but compounds the grid search.

Define the nnn-dimensional endogenous vector yty_tyt​, the threshold variable qtq_tqt​ (which may be an element of yty_tyt​ or an exogenous series), the delay parameter ddd, and the threshold γ\gammaγ. The model is a piecewise-linear system.

yt=c(1)+∑i=1pAi(1)yt−i+ut(1),ut(1)∼N(0,Σ(1))if qt−d≤γy_t = c^{(1)} + \sum_{i=1}^{p} A_i^{(1)} y_{t-i} + u_t^{(1)}, \quad u_t^{(1)} \sim \mathcal{N}(0, \Sigma^{(1)}) \quad \text{if } q_{t-d} \leq \gammayt​=c(1)+i=1∑p​Ai(1)​yt−i​+ut(1)​,ut(1)​∼N(0,Σ(1))if qt−d​≤γ
qt−dq_{t-d}qt−d​γ\gammaγ

When the threshold variable is at or below γ\gammaγ, Regime 1 parameters govern the data-generating process.

yt=c(2)+∑i=1pAi(2)yt−i+ut(2),ut(2)∼N(0,Σ(2))if qt−d>γy_t = c^{(2)} + \sum_{i=1}^{p} A_i^{(2)} y_{t-i} + u_t^{(2)}, \quad u_t^{(2)} \sim \mathcal{N}(0, \Sigma^{(2)}) \quad \text{if } q_{t-d} > \gammayt​=c(2)+i=1∑p​Ai(2)​yt−i​+ut(2)​,ut(2)​∼N(0,Σ(2))if qt−d​>γ
qt−dq_{t-d}qt−d​γ\gammaγ

Above the threshold, a different set of VAR parameters applies.

yt=[c(1)+A(1)(L)yt]⋅1(qt−d≤γ)+[c(2)+A(2)(L)yt]⋅1(qt−d>γ)+uty_t = \bigl[c^{(1)} + A^{(1)}(L) y_t\bigr] \cdot \mathbb{1}(q_{t-d} \leq \gamma) + \bigl[c^{(2)} + A^{(2)}(L) y_t\bigr] \cdot \mathbb{1}(q_{t-d} > \gamma) + u_tyt​=[c(1)+A(1)(L)yt​]⋅1(qt−d​≤γ)+[c(2)+A(2)(L)yt​]⋅1(qt−d​>γ)+ut​
A(1)A^{(1)}A(1)1(qt−d≤γ)\mathbb{1}(q_{t-d} \leq \gamma)1(qt−d​≤γ)A(2)A^{(2)}A(2)qt−dq_{t-d}qt−d​γ\gammaγ

The compact representation stacks both regimes using indicator functions. This is the form used for estimation.

yt=Xt(γ)′β+ut,Xt(γ)=(xt⋅1(qt−d≤γ)xt⋅1(qt−d>γ)),β=(β(1)β(2))y_t = X_t(\gamma)^\prime \beta + u_t, \quad X_t(\gamma) = \begin{pmatrix} x_t \cdot \mathbb{1}(q_{t-d} \leq \gamma) \\ x_t \cdot \mathbb{1}(q_{t-d} > \gamma) \end{pmatrix}, \quad \beta = \begin{pmatrix} \beta^{(1)} \\ \beta^{(2)} \end{pmatrix}yt​=Xt​(γ)′β+ut​,Xt​(γ)=(xt​⋅1(qt−d​≤γ)xt​⋅1(qt−d​>γ)​),β=(β(1)β(2)​)
γ\gammaγ1(qt−d≤γ)\mathbb{1}(q_{t-d} \leq \gamma)1(qt−d​≤γ)qt−dq_{t-d}qt−d​

Vectorizing the model this way shows that, conditional on γ\gammaγ, the Threshold VAR is just a linear regression on an expanded regressor set.

Why delay matters

Using qt−dq_{t-d}qt−d​ rather than qtq_tqt​ prevents the current shock from simultaneously determining the regime and being determined by it. With d≥1d \geq 1d≥1, regime assignment is predetermined at time ttt.

Transition checks

Impact jump
x0=Bε0,xh=AhBε0(h=1,2,…)x_0=B\varepsilon_0,\qquad x_h=A^hB\varepsilon_0\quad(h=1,2,\ldots)x0​=Bε0​,xh​=AhBε0​(h=1,2,…)

In a linear state-space representation, B maps the innovation into the contemporaneous state. A delayed variable has a zero row in B and moves only after A propagates the shock. A jump is therefore a timing restriction, not a charting artifact.

A nonzero point at h=0 means the shock changes the plotted variable on impact. It is not the end of a smooth pre-shock segment.

Do not smooth through h=0 when the model says the shock enters contemporaneously.

Decay or hump
rh=aρhorrh=a he−λhρh,∣ρ∣<1r_h=a\rho^h\quad\text{or}\quad r_h=a\,h e^{-\lambda h}\rho^h,\qquad |\rho|<1rh​=aρhorrh​=ahe−λhρh,∣ρ∣<1

Stable eigenvalues shrink repeated applications of A. If a variable is reached through another state first, the response rises before the stable decay dominates, producing a hump-shaped path.

Post-impact movement follows the model recursion. A hump means delayed propagation, not interpolation by the chart.

Do not treat a hump as evidence of a linear ramp unless the model includes that ramp.

Variable timing

Innovation

ε0≠0,εh=0 for h>0\varepsilon_0\ne 0,\qquad \varepsilon_h=0\ \text{for}\ h>0ε0​=0,εh​=0 for h>0

The shock is dated at horizon zero. A response at h=0 is an impact effect, not the end of a hidden ramp.

State vector

xh=AhBε0x_h=A^hB\varepsilon_0xh​=AhBε0​

A state can move on impact only through the relevant row of B; otherwise it moves after A propagates the shock.

Reported response

rh=CAhBε0+Dεhr_h=CA^hB\varepsilon_0+D\varepsilon_hrh​=CAhBε0​+Dεh​

The plotted variable can jump directly through D or indirectly through the state vector.

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
Help
Privacy PolicyTerms of UseEthics and Compliance

© 2026 Mark Jayson Nation LLC