Applied Econometrics Applied Econometrics

Simultaneous Equation Models Learning Objectives

1. Introduction: basic definitions 1. Understand the problem of simultaneity and

2. Consequences of ignoring simultaneity its consequences
3. The identification problem 2. Understand the identification problem
4. Estimation of simultaneous equation through macroeconomic examples
models
3. Understand and use the two-stage least
5. Example: IS–LM model
squares method of estimation

Applied Econometrics Applied Econometrics

Introduction Introduction (2)
Consider the well-known demand function:
All econometric models covered have dealt with
a single dependent variable and estimations of
single equations.
Economic analysis suggests that price and quantity
However, in modern world economics, typically are determined simultaneously by the
interdependence is very common. market processes, and therefore a full market
Several dependent variables are determined consists of a set of three different equations:
simultaneously, therefore appearing both as (a) demand function,
dependent and explanatory variables in a set (b) supply function and
of different equations. (c) condition for equilibrium

Applied Econometrics Applied Econometrics

Introduction (3) Solving the Model
Take the following three equations: Using the equilibrium condition and solving for
Pt we get

These are called structural equations of the or

simultaneous equations model, and the
coefficients β and γ are called structural
parameters.

1
 Applied Econometrics Applied Econometrics
Solving the Model (2) Solving the Model (3)

Substituting the expression for Pt in the supply The two new equations specify each of the
function we get: endogenous variables only in terms of the
exogenous variables, the parameters of the
model and the stochastic error terms.

These two equations are known as reduced form

equations and the πs are known as reduced
form parameters.

Applied Econometrics Applied Econometrics

Consequences of Ignoring Simultaneity Consequences of Ignoring Simultaneity (2)
One of the assumptions of CLRM states that the Consider the model:
error term of an equation should be
uncorrelated with each of the explanatory (11.8)
variables in the equation. (11.9)
If such a correlation exists, the OLS regression
equation is biased.
And think about an increase in e1t assuming
It should be evident from the reduced form
everything else stays constant.
equations that, in cases of simultaneous
equation models, such a bias exists.

Applied Econometrics Applied Econometrics

Consequences of Ignoring Simultaneity (3) Consequences of Ignoring Simultaneity (4)
(a) if e1t increases, this causes Y1t to increase Increase in the error term of an equation
because of Equation (11.8); then causes increase in explanatory variable in
the same equation.
(b) if Y1t increases (assuming that β2 is positive)
Y2t will also increase because of the relationship Assumption of no correlation among the error
in Equation (11.9); but term and the explanatory variables is
violated, leading to biased estimates.
(c) if Y2t increases in Equation (11.9) it also
increases in Equation (11.8) where it is an
explanatory variable.

2
 Applied Econometrics Applied Econometrics
Estimation of Simultaneous Equations Estimation of Simultaneous Equations (2)
Estimation of exactly identified equation: Estimation of over-identified equation:
ILS method TSLS method
To be used only when equations in simultaneous Basic idea behind TSLS method is to replace
equation model are found to be exactly identified. stochastic endogenous regressor (which is
Step 1 Find reduced form equations correlated with error term and causes bias) with
Step 2 Estimate the reduced form parameters by one that is nonstochastic and, consequently,
applying simple OLS to the reduced form equations independent of the error term.
Step 3 Obtain unique estimates of the structural This involves the following two stages (hence two-
parameters from the estimates of the parameters of stage least squares):
the reduced form equation in step 2

Applied Econometrics Applied Econometrics

Estimation of Simultaneous Equations (3) VAR Models and Causality
Stage 1 Regress each endogenous variable that is 1. Vector autoregressive (VAR) models
also a regressor on all the endogenous and lagged
2. Causality tests
endogenous variables in the entire system by
using simple OLS (equivalent to estimating the
reduced form equations) and obtain fitted values of
the endogenous variables of these regressions
Stage 2 Use the fitted values from stage 1 as proxies
or instruments for the endogenous regressors in
the original (structural form) equations

Applied Econometrics Applied Econometrics

Learning Objectives VAR Models
1. Differentiate between univariate and multivariate • Quite common in economics to have models
time series models
where some variables are not only explanatory
2. Understand VAR models and discuss advantages
variables for a given dependent variable; but
3. Understand the concept of causality and its
also explained by variables that they are used
importance in economic applications
to determine.
4. Use Granger causality test procedure
5. Use Sims causality test procedure • In those cases we have models of simultaneous
6. Estimate VAR models and test for Granger and
equations, in which it is necessary to clearly
Sims causality through the use of econometric identify which are endogenous and which are
software exogenous or predetermined variables.

3
 Applied Econometrics Applied Econometrics
VAR Models (2) VAR Models (3)
Sims (1980) suggests: For example, time series yt that is affected by
• if there is simultaneity among a number of current and past values of xt and,
variables, then all these variables should be simultaneously, the time series xt to be a series
treated in the same way. that is affected by current and past values of
• So, all variables are treated as endogenous. the yt series.
• This means that in its general reduced form In this case, simple bivariate model is given by:
each equation has the same set of
regressors. yt = β10 − β12xt + γ11yt−1 + γ12xt−1 + uyt
xt = β20 − β21yt + γ21yt−1 + γ22xt−1 + uxt

Applied Econometrics Applied Econometrics

VAR Models (4) VAR Models (5)
This is a first-order VAR model, because the Or
longest lag length is unity. Bzt= Γ0 + Γ1zt−1 + ut
Where
Rewriting the system with matrix algebra:
1 12  y   
1 12   yt   10   11  12   yt 1  u yt  B  , zt   t , 0   10 
     21 1   xt    20 

 21 1   xt    20   21  21   xt 1  u xt 
  12  u 
1   11  and ut   yt 
 21  21  u xt 

Applied Econometrics Applied Econometrics

VAR Models (6) VAR Models (7)
Advantages of VAR models: Disadvantages of VAR models:
(a) Very simple (no worry about which variables (a) A-theoretic as they are not based on any
are endogenous or exogenous) economic theory ‘everything causes everything’
(b) Estimation is very simple (usual OLS) (resolved by statistical inference and causality
(c) Forecasts from VAR models are better than tests)
those obtained from far more complex (b) Loss of degrees of freedom
simultaneous equation models (c) Obtained coefficients of VAR models are
difficult to interpret as they lack theoretical
background (overcome by impulse response
functions)

4
 Applied Econometrics Applied Econometrics
Causality Tests Causality Tests (2)
Suppose two variables, say yt and xt , affect each other Granger (1969) developed a relatively simple
with distributed lags.
test that defined causality as follows:
The relationship between those variables can be
captured by a VAR model. A variable yt is said to Granger-cause xt, if xt
In this case it is possible to have that: can be predicted with greater accuracy by
(a) yt causes xt using past values of the yt variable rather
(b) xt causes yt than not using such past values, all other
(c) there is bi-directional feedback (causality among terms remaining unchanged.
variables)
(d) the two variables are independent

Applied Econometrics Applied Econometrics

Causality Tests (3) Causality Tests (4)
• Case 1 Lagged x terms in (1) may be statistically different from
First step is to estimate following VAR model:
zero as a group, and lagged y terms in (2) not statistically
n m different from zero. In this case we have that xt causes yt.
yt  a1    i xt i    j yt  j  e1t • Case 2 Lagged y terms in (1) may be statistically different from
i 1 j 1 zero as a group, and lagged x terms in (1) not statistically
different from zero. In this case we have that yt causes xt.
n m
xt  a 2    i xt  i    j y t  j  e2 t
• Case 3 Both sets of x and y terms are statistically different from
zero in (1) and (2), so that have bi-directional causality.
i 1 j 1 • Case 4 Both sets of x and y terms are not statistically different
from zero in (1) and (2), so that xt is independent of yt.

Applied Econometrics Applied Econometrics

Causality Tests (5) Causality Tests (6)
Step 1 Regress yt on lagged y terms and obtain RSS Step 4 Calculate F statistic for normal Wald test on
of this regression (which is the restricted one) and coefficient restrictions
label it as RSSR
Step 5 If computed F value exceeds
Step 2 Regress yt on lagged y terms plus lagged x
F-critical value, reject the null hypothesis and
and obtain RSS of this regression (which now is
conclude that xt causes yt
the unrestricted one) and label it as RSSU
Step 3 Set the null and the alternative hypotheses:
Ho: coefficients of the lagged terms of x are equal to zero, or x−/→y
Ha: coefficients of the lagged terms of x are not equal to zero, or x→y