1

SIMULTANEOUS EQUATION MODELS

In this section, we look at unique estimation problems arising in economic models where
the values of several variables are determined simultaneously. Some examples include the usual
Keynesian and monetary macroeconomic models as well as supply and demand equations of
microeconomic markets. So far we assumed that Cov(X,u) = 0, that is the stochastic process
which generates the regressor X operates independently of the error term ut. This assumption
allows the determination of X to be treated independently of Y.
But there are many situations where such a one-way or unidirectional cause-and-effect
relationship is not meaningful. This occurs if Y is not only determined by the X’s, but some of
the X’s are, in turn, determined by Y. In short, there is a two-way, or simultaneous, relationship
between Y and (some of) the X’s. When the dependent variable in one equation is also an
explanatory variable in some other equations, we have simultaneous equation system or model.

Simultaneous Equation Problem

Economic data are often generated by a set of processes that are interdependent. Modeling such
interdependent processes leads to the consideration of systems of simultaneous equations. These
equations have the characteristics that each may contain several dependent variables that also
occur in other equations.

Examples Of Simultaneous Equation Models

1. Gas consumption
Gas consumption = α + β(thermostat setting) + θ(sq.ft) + ε
Thermostat setting = η + φ(income) + μ
Problem: Because μ and ε are correlated, and μ affects thermostat setting, thermostat
setting is correlated with ε.

2. Keynesian Model Of Income Determination:

Consider the simple Keynesian model of income determination.
Consumption Function Ct = 0 + 1Yt + Ut 0 < 1 < 1 …… (1)
Income identity Yt = Ct + It (= St) (2)
where C = Consumption expenditure, Y = Income
I = Investment (assumed exogenous), S = Savings
t = Time, U = Stochastic disturbance term, 0 and 1 = Parameters
The parameter 1 is known as the Marginal Propensity to Consume (MPC) (the amount of extra
consumption expenditure resulting from an extra Pula of income). From economic theory 1 is
expected to lie between 0 and 1.

Definitions
Endogenous: Those (whose values are) determined within the model. The endogenous variables
are regarded as stochastic. In the above model C (consumption) and Y (income) are the
endogenous variables.
Predetermined: Those (whose values are) determined outside the model. They are treated as
non-stochastic. The predetermined variables are divided into two categories: EXOGENOUS
current as well as lagged and lagged endogenous. In the above model I (investment expenditure)
is treated as an exogenous variable.
 2

The equations appearing in the model are known as the Structural, or behavioural equations
because they may portray the structure of an economy or the behaviour of an economic agent.
The parameters 0 and 1 are known as the structural parameters or coefficients. There is one
behavioural/structural equation for each endogenous variable in the system

Reduced Form Equation (RFE) is one which expresses an endogenous variable solely in terms
of the predetermined variables and the stochastic disturbances.

Consider the structural model:

M = α + βP + θN + ε
P = λ + φM + ηR + μ

The RFE is
               
M       R    N    ……….. (3)
 1     1     1     1   
i.e. M = a + bR + cN + e
               
P       R    N    (4)
 1     1     1     1   
i.e. P = d + fR + gN + u
 Reduced form equations are equations (3) and (4) with reduced form parameters a, b, c, d,f,g.

The Identification Problem

By the identification problem we mean whether numerical estimates of the parameters of a
structural equation can be obtained from the estimated reduced – form coefficients. If this can be
done, we say that the particular equation is identified. If this cannot be done, then we say that the
equation under consideration is unidentified, or under-identified.
An identified equation may be either exactly (or fully or just) identified or over-identified.
It is said to be exactly identified if unique numerical values of the structural parameters can be
obtained. It is said to be over-identified if more than one numerical value can be obtained for
some of the parameters of the structural equations.

Identification from the Structure Form

Rules For Identification

There are two conditions which must be fulfilled for an equation to be identified.
1. The Order Condition For Identification
A necessary (but not sufficient) condition of identification is known as the order
condition. If G is the total number of endogenous variables in the system and K is the
total number of variables missing (excluded) from the equation, then we can say:
(i) If K = G – 1 the equation is just identified
(ii) If K > G – 1 the equation is over identified
(iii) If K < G – 1 the equation is un(der) identified
2. The Rank Condition For Identification
This is otherwise called the necessary and sufficient condition for identification. The rank
condition states that: in a system of G equations any particular equation is identified if and only if
 3

it is possible to construct at least one-non-zero determinant of order (G – 1) from the coefficients

of the variables excluded from that particular equation but contained in the other equations of the
model.

To apply the rank condition one may proceed as follows:

Write the parameters of all the equations of the model in a separate table noting that the parameter
of a variable excluded from an equation is equal to zero.
Strike out the row of coefficients of the equation which is being examined for identification.
Strike out the columns in which a non-zero coefficient of the equation being examined appears.
Form the determinant(s) of order (G – 1) and examine their value. If all the det. of order (G – 1)
are zero the equation is under-identified.

To see whether the equation is exactly identified or over-identified we use the order condition.
If K = G – 1 the equation is exactly identified
K > G – 1 the equation is over-identified.
1. EXAMPLE
Assume the following simple version of the keynesian model of income determination.
Consumption function Ct = a0 + a1Yt - a2Tt + U
Investment function It = b0 + b1Yt – 1 + V
Taxation function Tt = c0 + c1Yt + W
Definition Y t = C t + It + G t
This model is mathematically complete in the sense that it contains as many equations as
endogenous variables. There are four endogenous variables, C, I, T, Y and two predetermined
variable lagged income (Yt – 1) and government expenditure (G)
A. The First Equation
1. Order Condition:
K = 3, G = 4,
K = 3 and G - 1 = 3
Thus K = G - 1, which shows that the order condition for identification is satisfied.
2. Rank Condition: The table of structural coefficients is as follows: (Write all equations
equal to 0).
Equations Variables
C Y T I Yt – 1 G
1st equation -1 a1 - a2 0 0 0
nd
2 equation 0 0 0 -1 b1 0
rd
3 equation 0 C1 -1 0 0 0
th
4 equation 1 -1 0 1 0 1
We strike out the first row and the three first columns of the table and thus obtain the table of
coefficients of excluded variables:

Table Of Coefficients Of Excluded Variables

-1 b1 0
0 0 0
1 0 1
We evaluate the determinant of this table. Clearly the value of this determinant is zero since the
second row contains only zero. Consequently we cannot form any non zero determinant of order 3
 4

(= G - 1). The rank condition is violated. Hence we concludes that the consumption function is
not identified; despite the satisfaction of the order criterion.

B. The Investment Function Is Over-identified

1. Order Condition.
The investment function includes two variables. Hence
K= 4
Clearly K > (G - 1) given that G - 1 = 3. The order condition is fulfilled.
2. Rank Condition
Deleting the second row and the fourth and fifth columns of the structural-coefficients
table we obtain.
Complete Table Of Structural Parameters
-1 a1 a2 0 0 0
0 0 0 1 b1 0
0 C1 -1 0 0 0
1 -1 0 1 0 1

Table Of Coefficients Of Excluded Variables

-1 a1 a2 0
0 C1 -1 0
1 -1 0 1

Models Of Simultaneous Relationship

The value of the first 3 x 3 determinant of excluded variables is
1 = - 1 c1 - 1 - a1 0 -1 + a2 0 c1 = 1 + a1 – a2c1  0
-1 0 1 0 1 1
(provided a1 - a2c1  -1)
The rank condition is satisfied since we can construct at least one non-zero determinant of order 3
= (G - 1)
Applying the counting rule K > (G - 1) we see that the inequality sign holds 4 > 3; hence the
investment function is over-identified.

The detection of the identifiability of the tax equation is left to the student as an exercise.
Exercise
The following is a system of structural equation.
y1t = - 3y2t + 5x1t + 6x2t + x3t - 9x4t + e1t
y2t = 2y1t + y3t - x1t + 5x3t - 8x4t + e2t
y3t = y1t - 4x2t + 6x3t + x4t + e3t
The y’s are endogenous variables while the x’s are exogenous. Assess the identifiability of the
model using both order and rank conditions.
 5

Estimation of parameters of a Simultaneous Equation Model

We adopt two approaches to estimate the structural equation coefficients, namely, single-equation
methods, also known as limited information methods, and system methods, also known as full
information methods. In the former, we estimate each equation in the system individually while
in the latter we estimate all the equations at once.

Indirect Least Squares

One method for obtaining consistent estimates for just or exactly identified structural coefficients
from the OLS estimates of the RF coefficients is known as the method of indirect least squares
(ILS), and the estimates thus obtained are known as the ILS estimates. ILS involves the
following three steps:

Step 1: Obtain the RFE, where all the RHS variables are exogenous. For instance,
               
M       R    N    ……….. (3)
 1     1     1     1   
i.e. M = a + bR + cN + e ……. (3a)
               
P       R    N    (4)
 1     1     1     1   
i.e. P = d + fR + gN + u ……… (4a)

Step 2: Solve these equations (3a and 4a) individually by OLS method and get the RFE
parameters.

Step 3: We obtain the original structural coefficients from the estimated RF coefficients.

Two Stage Least Squares (2SLS)

This is used when the equation in a simultaneous systems are either just identified or over
identified. For just identified equation, ILS and 2SLS estimates are the same. To perform this
method, we following the two steps below:
Step 1: Regress each endogenous variable against all exogenous variables, get the estimated
endogenous variable in each case.
Step 2: Regress original equations, replacing the endogenous explanatory variables with their
predicted values.

Three Stage Least Square

This is a system estimation method. That is, all the equations are considered at once, unlike
individual equations as seen in ILS,RFE and 2SLS. To perform 3SLS estimation:

1. Perform the two steps of 2SLS

2. Save residuals from the step 2 of 2SLS, labeling each to correspond to its equation.
3. Re-estimate structural equations with the error terms (residuals) included as explanatory
variables.
This method is better than the rest considered so far, because it uses more information in getting
its estimates.