You are on page 1of 35

Type author name/s here

Dougherty

Introduction to Econometrics,
5th edition
Chapter heading
Chapter 9: Simultaneous
Equations Estimation

© Christopher Dougherty, 2016. All rights reserved.


THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z   5 P   6Q   7 R  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

We will discuss the order condition for identification in two stages. First, we will look at an
example. Then, we will generalize.

1
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z   5 P   6Q   7 R  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Suppose that you have a four-equation simultaneous equations model that features four
endogenous variables W, X, Y, and Z, and three exogenous variables P, Q, and R.

2
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z   5 P   6Q   7 R  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

We will focus on the equation for Y. In principle, Y might be directly influenced by every
other endogenous variable and also by all of the exogenous variables, as shown above.

3
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z   5 P   6Q   7 R  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Obviously, this equation would not be identified. It has three endogenous variables on the
right side. There are three exogenous variables in the model, but all of them appear in their
own right on the right side and so none is available.
4
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z   5 P   6Q   7 R  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

To make any progress with the identification of this equation, we have to be able to impose
exclusion restrictions on it. An exclusion restriction is an assumption that a variable does
not appear on the right side. In other words, its coefficient is assumed to be zero.
5
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 1: P , Q , R excluded

If we are in a position to assert that P, Q, and R do not have direct effects on Y, we can
impose exclusion restrictions on them and write the model as shown.

6
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 1: P , Q , R excluded

The equation would now be exactly identified, because we have three variables needing
instruments, and we have three instruments available, P, Q, and R.

7
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   5 P  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 2: Z , Q , R excluded

Next, consider Variation 2, shown above. Is the equation for Y identified?

8
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   5 P  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 2: Z , Q , R excluded

Yes, it is. We now have only two exogenous variables excluded, but we need only two,
because one of the endogenous variables has also been excluded. Note that we still have
three exclusions in total: one endogenous and two exogenous variables.
9
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   5 P   6Q  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 3: X , Z , R excluded

Finally, consider Variation 3. Here, two endogenous variables are excluded. Therefore we
need only one instrument, which means that we need exclude only one explanatory variable
from the right side. The total number of exclusions remains three.
10
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   5 P   6Q  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 3: X , Z , R excluded

In each of the three variations, three variables were absent from the right side. This caused
the equation to be exactly identified. It did not matter whether the missing variables were
endogenous or exogenous.
11
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   5 P   6Q  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 3: X , Z , R excluded

If more than three variables had been absent, the number of available exogenous variables
would have been greater than the number needed for identification, and the equation would
have been overidentified.
12
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   5 P   6Q  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 3: X , Z , R excluded

Again, it would not matter whether the absent variables were endogenous or exogenous.

13
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z   5 P   6Q   7 R  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

We will now generalize. In the example, we saw that identification of the equation required
three exclusion restrictions, and it did not matter which three. This finding generalizes to
what is known as the order condition for identification.
14
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z   5 P   6Q   7 R  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

In a fully specified model, there will be as many equations as there are endogenous
variables. Let us suppose that there are G of each.

15
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z   5 P   6Q   7 R  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

The maximum number of endogenous variables that can appear on the right side of an
equation is G ‒ 1 (the other is the dependent variable of that equation).

16
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 1: P , Q , R excluded

In such a case, identification requires that at least G ‒ 1 exogenous variables should be


excluded from the equation, as in Variation 1.

17
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 1: P , Q , R excluded

This is a worst-case scenario because all the other endogenous variables are on the right
side of the equation for Y.

18
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 1: P , Q , R excluded

Suppose, however, that j endogenous variables are absent from the equation. We would then
need only G ‒ 1 ‒ j instruments, so only G ‒ 1 ‒ j exogenous variables would have to be
excluded from the equation.
9
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   4 Z  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 1: P , Q , R excluded

The total number of variables excluded remains the same: j endogenous variables and
exogenous variables make a total of G ‒ 1.

20
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   3 X   5 P  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 2: Z , Q , R excluded

Z excluded  j  1

This is illustrated by Variations 2 and 3. G is equal to 4. In Variation 2, Z has been excluded


and j = 1. G ‒ 1 ‒ j = 2, so we need only two exclusion restrictions on the exogenous
variables in order to have enough instruments. Q and R are excluded, so the equation is
identified. 21
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   5 P   6Q  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 3: X , Z , R excluded

X and Z excluded  j  2

In Variation 3, X and Z have been excluded and j = 2. G ‒ 1 ‒ j = 1, so we need only one


exclusion restriction on the exogenous variables in order to have enough instruments. R is
excluded, so the equation is identified.
22
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   5 P   6Q  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 3: X , Z , R excluded

X and Z excluded  j  2

Thus we come to the general conclusion that an equation in a simultaneous equations model
is likely to be identified if G ‒ 1 or more variables are missing from it.

23
THE ORDER CONDITION FOR IDENTIFICATION

Y   1   2W   5 P   6Q  uY

Three other equations with W , X , and Z as dependent variables


Endogenous variables: W , X , Y , Z
Exogenous variables: P , Q , R

Variation 3: X , Z , R excluded

X and Z excluded  j  2

If exactly G ‒ 1 are missing, it is likely to be exactly identified, and if more than are missing,
it is likely to be overidentified, calling for the use of TSLS.

24
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

The order condition is only a necessary condition for identification. In practice, it is usually a
sufficient one, but exceptions are possible.

25
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

In general, when one has a simultaneous equations model, the reduced form equation for
each endogenous variable will have all of the exogenous variables in the model on its right
side.
26
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

The interconnectedness of the model ensures that every exogenous variable affects every
endogenous variable, either directly or indirectly, as has been the case in all of the examples
so far. Those exogenous variables with indirect effects are available as instruments.
27
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

However, it is possible for the structure of the model to be such that it is not fully
interconnected. It is then possible that one or more of the exogenous variables may be
absent from a reduced form equation. Then they are not available as instruments, after all.
28
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

We will consider an example. In the model shown, X, Y, and Z are endogenous variables,
and Q is exogenous.

29
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

At first sight, the equations for X and Y both appear to be identified. Each has an
endogenous variable on the right side, but each excludes Q, so in principle Q could be used
as an instrument.
30
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

However, looking at the structural equations for X and Y, neither of those variables depends
on Z, and as a consequence, the reduced form equations for X and Y do not include Q. So Q
is not available as an instrument.
31
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

The rank condition is a mathematical way of determining whether the structure of the model
is sufficiently interconnected that all of the exogenous variables are available as
instruments.
32
THE ORDER CONDITION FOR IDENTIFICATION

X   1   2Y  u
Y  1   2 X  v
Z   1   2Y   3 X   4Q  w

It involves the use of linear algebra and will not be pursued here. Fortunately, failure of the
rank condition is unusual in practice.

33
Copyright Christopher Dougherty 2016.

These slideshows may be downloaded by anyone, anywhere for personal use.


Subject to respect for copyright and, where appropriate, attribution, they may be
used as a resource for teaching an econometrics course. There is no need to
refer to the author.

The content of this slideshow comes from Section 9.2 of C. Dougherty,


Introduction to Econometrics, fifth edition 2016, Oxford University Press.
Additional (free) resources for both students and instructors may be
downloaded from the OUP Online Resource Centre
http://www.oxfordtextbooks.co.uk/orc/dougherty5e/.

Individuals studying econometrics on their own who feel that they might benefit
from participation in a formal course should consider the London School of
Economics summer school course
EC212 Introduction to Econometrics
http://www2.lse.ac.uk/study/summerSchools/summerSchool/Home.aspx
or the University of London International Programmes distance learning course
EC2020 Elements of Econometrics
www.londoninternational.ac.uk/lse.

2016.05.18

You might also like