xbeta
Simultaneous equations model

Model

Let y i be a M×1 vector of endogenous variables and x i be a K×1 vector of exogenous covariates.

The structural model can be written as

Γ 0M×My iM×1+B 0M×Kx iK×1=ε iM×1,\underset{M \times M}{\Gamma_0} \underset{M \times 1}{y_i} + \underset{M \times K}{B_0} \underset{K \times 1}{x_i} = \underset{M \times 1}{\varepsilon_i},

E[x iε im]=0 for all m. Γ 0 and B 0 are the structural parameters.

If we define ε i(ε i1,ε i2,,ε iM) , then we have

E[x iε i ]=0K×M\E[x_i \varepsilon_i^\top] = \underset{K \times M}{0}

If Γ 0 is nonsingular, the corresponding reduced form model is

y i =Γ 0 1B 0x i+Γ 0 1ε i =Π 0 x i+ν i \begin{aligned} y_i &= -\Gamma_0^{-1} B_0 x_i + \Gamma_0^{-1} \varepsilon_i \\ &= \Pi_0^\top x_i + \nu_i \\ \end{aligned}

The relationship between the structural and reduced form parameters is thus

Π 0 M×K=Γ 0M×M 1B 0M×K\underset{M \times K}{\Pi_0^\top} = -\underset{M \times M}{\Gamma_0}^{-1} \underset{M \times K}{B_0}

and the error terms are related by

ν i=Γ 0 1M×Mε iM×1.\nu_i = \underset{M \times M}{\Gamma_0^{-1}} \underset{M \times 1}{\varepsilon_i}.

Note that we still have E[x iν i ]=0.

Examples

Supply and Demand

q i =γ 11p i+B 11+B 12a i+ε i1 q i =γ 21p i+B 21+B 22w i+ε i2 \begin{gathered} q_i &= \gamma_{11} p_i + B_{11} + B_{12} a_i + \varepsilon_{i1} \\ q_i &= \gamma_{21} p_i + B_{21} + B_{22} w_i + \varepsilon_{i2} \\ \end{gathered}

The first equation is the supply equation and the second is the demand demand equation. The exogenous variables are the wage w i, a supply shifter, and a i, a demand shifter. Price is endogenous to both equations.

See also: Full information maximum likelihood