\[\begin{align*}
y'_t\Gamma + x'_tB &= \epsilon'_t \\
y'_t\Gamma\Gamma^{-1} + x'_tB\Gamma^{-1} &= \epsilon'_t\Gamma^{-1} \\
y'_t - x'_t\Pi &= \nu'_t \\
y'_t &= x'_t\Pi + \nu'_t
\end{align*}\]
\[\begin{align*}
Y\Gamma + XB &= E \\
Y\Gamma\Gamma^{-1} + XB\Gamma^{-1} &= E\Gamma^{-1} \\
Y - X\Pi &= V \\
Y &= X\Pi + V
\end{align*}\]
Ejemplo:
Forma estructural y reducida del modelo de oferta y demanda
\[\begin{equation*}
\left\{\begin{aligned}
q_t^d &= \alpha_0 + \alpha_1 p_t + \alpha_2 x_t + \epsilon_t^d \\
q_t^s &= \beta_0 + \beta_1 p_t + \epsilon_t^s \\
q_t^d &= q_t^s = q
\end{aligned} \right.
\end{equation*}\]
Su forma estructural es
\[\begin{equation*}
\notation{\MAT{q_t & p_t}}{$y'_t$}\notation{\MAT{1 & 1 \\ -\alpha_1 & -\beta_1} }{$\Gamma$} +
\notation{\MAT{1 & x_t}}{$x'_t$}\notation{\MAT{-\alpha_0 & -\beta_0 \\ -\alpha_2 & 0}}{$B$} =
\notation{\MAT{\epsilon_t^d & \epsilon_t^s}}{$\epsilon'_t$}
\end{equation*}\]
Note que cada columna de \(\Gamma\) y de \(B\) corresponden a una ecuación del modelo.
Así,
\[\begin{equation*}
\Gamma = \MAT{1 & 1 \\ -\alpha_1 & -\beta_1} \Rightarrow \Gamma^{-1} = \frac{1}{\beta_1-\alpha_1}\MAT{\beta_1 & 1 \\ -\alpha_1 & - 1}
\end{equation*}\]
Los parámetros reducidos son:
\[\begin{equation*}
\Pi = -B\Gamma^{-1} = \tfrac{1}{\beta_1-\alpha_1}\MAT{\alpha_0 & \beta_0 \\ \alpha_2 & 0}\MAT{\beta_1 & 1 \\ -\alpha_1 & - 1}
=\tfrac{1}{\beta_1-\alpha_1}\MAT{\alpha_0\beta_1-\alpha_1\beta_0 & \alpha_0 - \beta_0 \\ \alpha_2\beta_1 & \alpha_2}
\end{equation*}\]
y los shocks reducidos son:
\[\begin{equation*}
\nu'_t = \epsilon'_t\Gamma^{-1}= \tfrac{1}{\beta_1-\alpha_1}\MAT{\epsilon^d_t & \epsilon^s_t}\MAT{\beta_1 & 1 \\ -\alpha_1 & - 1}
=\tfrac{1}{\beta_1-\alpha_1}\MAT{\beta_1\epsilon^d_t-\alpha_1\epsilon^s_t & \epsilon^d_t - \epsilon^s_t}
\end{equation*}\]
por lo que la forma reducida es:
\[\begin{equation*}
\notation{\MAT{q_t & p_t}}{$y'_t$}
=
\notation{\MAT{1 & x_t}}{$x'_t$}
\notation{\MAT{\tfrac{\alpha_0\beta_1-\alpha_1\beta_0}{\beta_1-\alpha_1} & \tfrac{\alpha_0 - \beta_0}{\beta_1-\alpha_1} \\ \tfrac{\alpha_2\beta_1}{\beta_1-\alpha_1} & \tfrac{\alpha_2}{\beta_1-\alpha_1}}}{$\Pi$}
+
\notation{\MAT{\tfrac{\beta_1\epsilon^d_t-\alpha_1\epsilon^s_t}{\beta_1-\alpha_1} & \tfrac{\epsilon^d_t - \epsilon^s_t}{\beta_1-\alpha_1}}}{$\nu'_t$}
\end{equation*}\]
La forma reducida corresponde a la cantidad y precio de equilibrio:
\[\begin{equation*}
\left\{\begin{aligned}
q_t^* &= \notation{\tfrac{\alpha_0\beta_1-\alpha_1\beta_0}{\beta_1-\alpha_1}}{$\pi_{11}$} + \notation{\tfrac{\alpha_2\beta_1}{\beta_1-\alpha_1}}{$\pi_{21}$}x_t +
\notation{\tfrac{\beta_1\epsilon_t^d - \alpha_1\epsilon_t^s}{\beta_1-\alpha_1}}{$\nu_1$} \\
p_t^* &= \notation{\tfrac{\alpha_0-\beta_0}{\beta_1-\alpha_1}}{$\pi_{12}$} + \notation{\tfrac{\alpha_2}{\beta_1-\alpha_1}}{$\pi_{22}$}x_t +
\notation{\tfrac{\epsilon_t^d - \epsilon_t^s}{\beta_1-\alpha_1} }{$\nu_2$}
\end{aligned} \right.
\end{equation*}\]
A partir de la forma reducida, es fácil calcular el efecto de shocks o de cambios en variables exógenas sobre las endógenas. Por ejemplo:
\[\begin{equation*}
\marginal{q^*_t}{x_t} = \pi_{21} = \tfrac{\alpha_2\beta_1}{\beta_1-\alpha_1} \qquad\qquad
\marginal{p^*_t}{\epsilon^s_t} = \tfrac{-1}{\beta_1-\alpha_1}
\end{equation*}\]
Nótese que a partir de la forma reducida, el efecto de una variable exógena es observable (es un parámetro reducido), no así el efecto de un shock estructural (está combinado con los demás shocks estructurales en el shock reducido).
Ejemplo:
Forma estructural y reducida del modelo de demanda agregada
\[\begin{equation*}
\left\{\begin{aligned}
C_t &= \alpha_0 + \alpha_1 Y_t + \epsilon_{1t} \\
I_t &= \beta_0 + \beta_1\left(Y_t - Y_{t-1}\right) + \epsilon_{2t} \\
Y_t &= C_t + I_t + G_t
\end{aligned} \right.
\end{equation*}\]
Su forma estructural es
\[\begin{equation*}
\notation{\MAT{1 & 0 & -\alpha_1 \\ 0 & 1 & -\beta_1 \\ -1 & -1 & 1} }{$\Gamma'$}\notation{\MAT{C_t \\ I_t \\ Y_t}}{$y_t$} +
\notation{\MAT{-\alpha_0 & 0 & 0 \\ -\beta_0 & \beta_1 & 0 \\ 0 & 0 &-1}}{$B'$}\notation{\MAT{1 \\ Y_{t-1} \\ G_t}}{$x_t$} =
\notation{\MAT{\epsilon_{1t} \\ \epsilon_{2t} \\ 0}}{$\epsilon_t$}
\end{equation*}\]
Como tomamos la transpuesta, note que cada fila de \(\Gamma'\) y de \(B'\) corresponden a una ecuación del modelo.
Así,
\[\begin{equation*}
\Gamma' = \MAT{1 & 0 &-\alpha_1 \\ 0 & 1 & - \beta_1 \\ -1 & -1 & 1} \Rightarrow {\Gamma'}^{-1} = \frac{1}{1-\alpha_1-\beta_1}\MAT{1-\beta_1 & \alpha_1 & \alpha_1 \\ \beta_1 & 1-\alpha_1 & \beta_1 \\ 1 & 1 & 1}
\end{equation*}\]
Los parámetros reducidos son:
\[\begin{align*}
\Pi' = -{\Gamma'}^{-1}B' &= \tfrac{1}{1-\alpha_1-\beta_1}\MAT{1-\beta_1 & \alpha_1 & \alpha_1 \\ \beta_1 & 1-\alpha_1 & \beta_1 \\ 1 & 1 & 1}
\MAT{\alpha_0 & 0 & 0 \\ \beta_0 & -\beta_1 & 0 \\ 0 & 0 & 1} \\
&= \tfrac{1}{1-\alpha_1-\beta_1}\MAT{\alpha_0 - \alpha_0\beta_1 + \alpha_1\beta_0 & -\alpha_1\beta_1 & \alpha_1 \\ \alpha_0\beta_1 + \beta_0 - \alpha1\beta_0 & \alpha_1\beta_1 - \beta_1 & \beta_1 \\ \alpha_0+\beta_0 & -\beta_1 & 1}
\end{align*}\]
y los shocks reducidos son:
\[\begin{align*}
\nu_t = {\Gamma'}^{-1}\epsilon_t &= \tfrac{1}{1-\alpha_1-\beta_1}\MAT{1-\beta_1 & \alpha_1 & \alpha_1 \\ \beta_1 & 1-\alpha_1 & \beta_1 \\ 1 & 1 & 1}
\MAT{\epsilon_{1t} \\ \epsilon_{2t} \\ 0} \\
&=\tfrac{1}{1-\alpha_1-\beta_1}\MAT{(1-\beta_1)\epsilon_{1t} + \alpha_1\epsilon_{2t} \\ \beta_1\epsilon_{1t} + (1-\alpha_1)\epsilon_{2t} \\ \epsilon_{1t} + \epsilon_{2t}}
\end{align*}\]
por lo que la forma reducida es:
\[\begin{equation*}
\notation{\MAT{C_t \\ I_t \\ Y_t}}{$y_t$}
=
\notation{\tfrac{1}{1-\alpha_1-\beta_1}\MAT{\alpha_0 - \alpha_0\beta_1 + \alpha_1\beta_0 & -\alpha_1\beta_1 & \alpha_1 \\ \alpha_0\beta_1 + \beta_0 - \alpha1\beta_0 & \alpha_1\beta_1 - \beta_1 & \beta_1 \\ \alpha_0+\beta_0 & -\beta_1 & 1}}{$\Pi'$}
\notation{\MAT{1 \\ Y_{t-1} \\ G_t}}{$x_t$}
+
\notation{\tfrac{1}{1-\alpha_1-\beta_1}\MAT{(1-\beta_1)\epsilon_{1t} + \alpha_1\epsilon_{2t} \\ \beta_1\epsilon_{1t} + (1-\alpha_1)\epsilon_{2t} \\ \epsilon_{1t} + \epsilon_{2t}}}{$\nu_t$}
\end{equation*}\]
La forma reducida corresponde al consumo, inversión, e ingreso de equilibrio:
\[\begin{equation*}
\left\{\begin{aligned}
C_t^* &= \tfrac{\alpha_0 - \alpha_0\beta_1 + \alpha_1\beta_0}{1-\alpha_1-\beta_1} - \tfrac{\alpha_1\beta_1}{1-\alpha_1-\beta_1}Y_{t-1} + \tfrac{ \alpha_1 }{1-\alpha_1-\beta_1}G_t + \tfrac{(1-\beta_1)\epsilon_{1t} + \alpha_1\epsilon_{2t}}{1-\alpha_1-\beta_1} \\
I_t^* &= \tfrac{ \alpha_0\beta_1 + \beta_0 - \alpha1\beta_0}{1-\alpha_1-\beta_1} + \tfrac{\alpha_1\beta_1 - \beta_1}{1-\alpha_1-\beta_1}Y_{t-1} + \tfrac{\beta_1}{1-\alpha_1-\beta_1}G_t + \tfrac{\beta_1\epsilon_{1t} + (1-\alpha_1)\epsilon_{2t} }{1-\alpha_1-\beta_1} \\
Y_t^* &= \tfrac{ \alpha_0+\beta_0 }{1-\alpha_1-\beta_1} - \tfrac{\beta_1 }{1-\alpha_1-\beta_1}Y_{t-1} + \tfrac{1}{1-\alpha_1-\beta_1}G_t + \tfrac{\epsilon_{1t} + \epsilon_{2t}}{1-\alpha_1-\beta_1}
\end{aligned} \right.
\end{equation*}\]
Así, el multiplicador del gasto público es:
\[\begin{equation*}
\marginal{Y^*_t}{G_t} = \pi_{33} = \tfrac{1}{1-\alpha_1-\beta_1}
\end{equation*}\]
De nuevo, observe que los shocks reducidos son combinaciones lineales de los shocks estructurales.