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

Semi-structural core Model

A core block of estimated macro equations - output, inflation, labor, policy - wired together without full DSGE micro-foundations.

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 Steady State

The semi-structural core consists of nyn_yny​ observed variables, nsn_sns​ unobserved states (trends and gaps), and nen_ene​ structural shocks. We write the system in state-space form with the behavioral equations as the transition law and an observation equation mapping states to observables.

In the steady state, all gaps equal zero, inflation equals the target πˉ\bar{\pi}πˉ, and the policy rate equals r∗+πˉr^* + \bar{\pi}r∗+πˉ. The steady state is not derived from optimization; it is imposed as a calibration target. This is the fundamental methodological choice separating semi-structural from DSGE.

The behavioral equations in deviation form (all variables expressed as deviations from steady state or as gaps) are:

y~t=βlagy~t−1+βleadEt[y~t+1]−βr(it−Et[πt+1]−rt∗)−βzz~t+εtIS\tilde{y}_t = \beta_{\text{lag}} \tilde{y}_{t-1} + \beta_{\text{lead}} E_t[\tilde{y}_{t+1}] - \beta_r (i_t - E_t[\pi_{t+1}] - r^*_t) - \beta_z \tilde{z}_t + \varepsilon_t^{\text{IS}}y~​t​=βlag​y~​t−1​+βlead​Et​[y~​t+1​]−βr​(it​−Et​[πt+1​]−rt∗​)−βz​z~t​+εtIS​
y~t\tilde{y}_ty~​t​βlag\beta_{\text{lag}}βlag​βr\beta_rβr​iti_tit​rt∗r^*_trt∗​z~t\tilde{z}_tz~t​

IS curve: output gap depends on its own lags and leads, the real interest rate gap, the real exchange rate gap, and a demand shock. Under the appreciation-positive convention for the exchange rate gap, a real appreciation (positive z~t\tilde{z}_tz~t​) lowers the output gap through weaker net exports - hence the minus sign on βz\beta_zβz​.

πt=αleadEt[πt+1]+(1−αlead)πt−1+αyy~t−αzΔz~t+εtπ\pi_t = \alpha_{\text{lead}} E_t[\pi_{t+1}] + (1 - \alpha_{\text{lead}}) \pi_{t-1} + \alpha_y \tilde{y}_t - \alpha_z \Delta \tilde{z}_t + \varepsilon_t^{\pi}πt​=αlead​Et​[πt+1​]+(1−αlead​)πt−1​+αy​y~​t​−αz​Δz~t​+εtπ​
πt\pi_tπt​αlead\alpha_{\text{lead}}αlead​αy\alpha_yαy​y~t\tilde{y}_ty~​t​z~t\tilde{z}_tz~t​

New Keynesian Phillips curve: inflation is a weighted average of forward and backward expectations, plus output gap and import-price pass-through. Under the appreciation-positive convention, a real appreciation (Δz~t>0\Delta \tilde{z}_t > 0Δz~t​>0) lowers import-price inflation - hence the minus sign on αz\alpha_zαz​.

it=γlagit−1+(1−γlag)[rt∗+πˉ+γπ(πt4−πˉ)+γyy~t]+εtMPi_t = \gamma_{\text{lag}} i_{t-1} + (1 - \gamma_{\text{lag}}) \left[ r^*_t + \bar{\pi} + \gamma_{\pi}(\pi^4_t - \bar{\pi}) + \gamma_y \tilde{y}_t \right] + \varepsilon_t^{\text{MP}}it​=γlag​it−1​+(1−γlag​)[rt∗​+πˉ+γπ​(πt4​−πˉ)+γy​y~​t​]+εtMP​
iti_tit​rt∗r^*_trt∗​γπ\gamma_{\pi}γπ​y~t\tilde{y}_ty~​t​

Taylor-type rule with interest rate smoothing. π4\pi^4π4 is year-over-year inflation. γπ>1\gamma_{\pi} > 1γπ​>1 is required for determinacy.

z~t=δEt[z~t+1]+14(r~t−r~tforeign)+εtUIP\tilde{z}_t = \delta E_t[\tilde{z}_{t+1}] + \frac{1}{4}(\tilde{r}_t - \tilde{r}_t^{\text{foreign}}) + \varepsilon_t^{\text{UIP}}z~t​=δEt​[z~t+1​]+41​(r~t​−r~tforeign​)+εtUIP​
z~t\tilde{z}_tz~t​r~t\tilde{r}_tr~t​

Uncovered interest parity in gap form. δ=1\delta = 1δ=1 (pure forward-looking UIP) delivers Dornbusch-style overshooting; δ<1\delta < 1δ<1 dampens the exchange-rate jump relative to Dornbusch.

Step 1

The state vector contains both gap variables (first four) and trend variables (last four: potential output level, potential growth, equilibrium real rate, trend inflation). Trends are unobserved.
Define state vector: st=[y~t,πt,it,z~t,yˉt,gt,rt∗,πˉt]′\text{Define state vector: } s_t = [\tilde{y}_t, \pi_t, i_t, \tilde{z}_t, \bar{y}_t, g_t, r^*_t, \bar{\pi}_t]'Define state vector: st​=[y~​t​,πt​,it​,z~t​,yˉ​t​,gt​,rt∗​,πˉt​]′
y~t\tilde{y}_ty~​t​πt\pi_tπt​iti_tit​z~t\tilde{z}_tz~t​yˉt\bar{y}_tyˉ​t​rt∗r^*_trt∗​

Step 2

Potential output follows an integrated random walk with time-varying drift gtg_tgt​ (potential growth). The equilibrium real rate follows a driftless random walk. These are deliberately flexible to accommodate structural breaks in trend growth.
Trend laws of motion: yˉt=yˉt−1+gt−1+εtyˉ,gt=gt−1+εtg,rt∗=rt−1∗+εtr∗\text{Trend laws of motion: } \bar{y}_t = \bar{y}_{t-1} + g_{t-1} + \varepsilon_t^{\bar{y}}, \quad g_t = g_{t-1} + \varepsilon_t^g, \quad r^*_t = r^*_{t-1} + \varepsilon_t^{r^*}Trend laws of motion: yˉ​t​=yˉ​t−1​+gt−1​+εtyˉ​​,gt​=gt−1​+εtg​,rt∗​=rt−1∗​+εtr∗​
yˉt\bar{y}_tyˉ​t​rt∗r^*_trt∗​

Step 3

Collect all equations into matrix form. A0A_0A0​ may not be the identity matrix because some equations have contemporaneous terms. The BBB matrix captures forward-looking expectations. When BBB is nonzero, the system requires a rational expectations solution method.
Stack into state-space form: A0st=A1st−1+BEt[st+1]+Cεt\text{Stack into state-space form: } A_0 s_t = A_1 s_{t-1} + B E_t[s_{t+1}] + C \varepsilon_tStack into state-space form: A0​st​=A1​st−1​+BEt​[st+1​]+Cεt​

Step 4

XtX_tXt​ contains observed variables (GDP level, CPI, policy rate, nominal exchange rate). HHH selects or combines states to match observables. utu_tut​ is measurement error. GDP level observation: XtGDP=yˉt+y~tX_t^{\text{GDP}} = \bar{y}_t + \tilde{y}_tXtGDP​=yˉ​t​+y~​t​. This links the unobserved decomposition to the observed data. The observation equation is linear even when the transition contains expectations.
Observation equation: Xt=Hst+ut\text{Observation equation: } X_t = H s_t + u_tObservation equation: Xt​=Hst​+ut​

Step 5

Setting all shocks to zero and all expectations to their realized values, the gap equations produce zero gaps and inflation at target. This is by construction, not by derivation from optimality conditions. The calibrator chooses πˉ\bar{\pi}πˉ and r∗r^*r∗.
Steady state: y~ss=0,πss=πˉ,iss=r∗+πˉ,z~ss=0\text{Steady state: } \tilde{y}_{ss} = 0, \quad \pi_{ss} = \bar{\pi}, \quad i_{ss} = r^* + \bar{\pi}, \quad \tilde{z}_{ss} = 0Steady state: y~​ss​=0,πss​=πˉ,iss​=r∗+πˉ,z~ss​=0

Step 6

Each behavioral equation is verified at the steady state. The IS curve: the real rate gap is iss−πss−r∗=(r∗+πˉ)−πˉ−r∗=0i_{ss} - \pi_{ss} - r^* = (r^* + \bar{\pi}) - \bar{\pi} - r^* = 0iss​−πss​−r∗=(r∗+πˉ)−πˉ−r∗=0. All gaps are zero. The system is internally consistent.
Check: IS at steady state: 0=βlag⋅0+βlead⋅0−βr(r∗+πˉ−πˉ−r∗)−βz⋅0  ⟹  0=0  ✓\text{Check: IS at steady state: } 0 = \beta_{\text{lag}} \cdot 0 + \beta_{\text{lead}} \cdot 0 - \beta_r (r^* + \bar{\pi} - \bar{\pi} - r^*) - \beta_z \cdot 0 \implies 0 = 0 \; \checkmarkCheck: IS at steady state: 0=βlag​⋅0+βlead​⋅0−βr​(r∗+πˉ−πˉ−r∗)−βz​⋅0⟹0=0✓
βlag\beta_{\text{lag}}βlag​βr\beta_rβr​

Why not DSGE?

In a DSGE, the IS curve parameters (βlag,βr)(\beta_{\text{lag}}, \beta_r)(βlag​,βr​) would be derived from the household's Euler equation: βr=1/σ\beta_r = 1/\sigmaβr​=1/σ where σ\sigmaσ is the coefficient of relative risk aversion. The semi-structural approach treats βr\beta_rβr​ as a free parameter to be estimated or calibrated. This gains flexibility (βr\beta_rβr​ can differ from micro estimates) at the cost of Lucas critique vulnerability (βr\beta_rβr​ may shift when policy regimes change).

Notations

x~t\tilde{x}_tx~t​

Gap

xˉt\bar{x}_txˉt​

Trend value

Et[⋅]E_t[\cdot]Et​[⋅]

Rational expectation

εtj\varepsilon_t^jεtj​

Structural shock

About these symbols

Gap: Gap: deviation of variable x from its trend or equilibrium value

Trend value: Trend or equilibrium value of variable x (unobserved, estimated)

Rational expectation: Model-consistent rational expectation conditional on time-ttt information

Structural shock: Structural shock to equation jjj, i.i.d. N(0,σj2)N(0, \sigma_j^2)N(0,σj2​)

    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