7.

2 Application to economics: Leontief Model

Wassily Leontief won the Nobel prize in economics in 1973.
The Leontief model is a model for the economics of a whole country or region.
In the model there are n industries producing n different products such that the
input equals the output or, in other words, consumption equals production. One
distinguishes two models:

open model: some production consumed internally by industries, rest consumed

by external bodies.
Problem: Find production level if external demand is given.

closed model: entire production consumed by industries.

Problem: Find relative price of each product.

The open Leontief Model

Let the n industries denoted by S1, S2 , . . ., Sn . The exchange of products can be
described by an
input-output graph

Here, aij denotes the number of units produced by industry Si necessary to

produce one unit by industry Sj and bi is the number of externally demanded units
of industry Si .
Example: Primitive model of the economy of Kansas in the 19th century.

1
 The following equations are satisfied:
Production of Total output = Internal consumption + External Demand
farming industry (in tons): x = 0.05 x + 0.5 y + 8000
.
horse industry: y = 0.01 x + 2000
(in 1000km horse rides)
In general, let x1 , x2 , . . ., xn , be the total output of industry S1, S2 , . . ., Sn ,
respectively. Then


x1 = a11 x1 + a12 x1 + + a1n x1 + b1

x2 = a21 x1 + a22 x1 + + a2n x2 + b2

...


xn = an1 x1 + an2 x1 + + ann xn + bn ,

since aij xj is the number of units produced by industry Si and consumed by indus-
try Sj . The total consumption equals the total production for the product of each
industry Si .
Let
a11 . . . a1n b1 x1
A = ... .. , B = .. , X = ..

. . .
an1 . . . ann bn xn .
A is called the input-output matrix, B the external demand vector and X the
production level vector. The above system of linear equations is equivalent to the
matrix equation
X = AX + B.

0
..
In the open Leontief model, A and B 6= . are given and the problem is to
0
determine X from this matrix equation.
We can transform this equation as follows:

In X AX = B
(In A)X = B
X = (In A)1 B

if the inverse of the matrix In A exists. ((In A)1 is then called the Leontief
inverse.) For a given realistic economy, a solution obviously must exist.
For our example we have:
! ! !
0.05 0.5 8, 000 x
A= , B= , X=
0.1 0 2, 000 y
.

2
We obtain therefore the solution
X = (I2 A)1 B
! !!1 !
1 0 0.05 0.5 8, 000
=
0 1 0.1 0 2, 000
!1 !
0.95 0.5 8, 000
=
0.1 1 2, 000
! !
1 10 5 8, 000
=
9 1 9.5 2, 000
!
10, 000
= ,
3, 000
i.e., x = 10, 000 tons wheat and y = 3 Million km horse!ride.
7, 300
If the external demand changes, ex. B 0 = , we get
2, 500
!0 ! ! !
x 1 10 5 7, 300 9, 500
= (I2 A)1 B 0 = = ,
y 9 1 9.5 2, 500 3, 450
i.e., one doesnt need to recompute (I2 A)1 .
One difficulty with the model: How to determine the matrix A from a given econ-

x1 b1
omy? Typically, X = ... is known, B = ... is known and (aij xj )i,j=1, ... n

xn bn
is known. One takes therefore the matrix (aij xj )i,j=1, ... n and divides the j-th column
by xj for j = 1, . . ., n to get A.
Example: An economy has the two industries R and S. The current consumption
is given by the table
consumption
R S external
Industry R production 50 50 20
Industry S production 60 40 100

Assume the new external demand is 100 units of R and 100 units of S. Determine
the new production levels.
Solution: The ! total production
! is 120 units for R and
! 200 units for S. We ! obtain
50 50
120 20 120 200 100
X = , B = , A = 60 40 , and B 0 = . The
200 100 120 200 100
solution is
! ! !
0 1 0 1 96 30 100 307.3
X = (I2 A) B = = .
41 60 70 100 317.0
The new production levels are 307.3 and 317.0 for R and S, respectively.

3
The closed Leontief Model
The closed Leontief model can be described by the matrix equation

X = AX,

i.e., there is no external demand. The matrix In A is usually not invertible.

0
(Otherwise, the only solution would be X = ... .)

0
The input-output graph looks now as follows:

There is only internal consumption.

Example: Extended model of the economy of Kansas in the 19th century including
labor.

The corresponding matrix equation is:

x 0.05 0.5 0.5 x

y = 0.1 0 0.1 y
1331
z 0.4 0.1 1800 z .

4
If X is a solution, also t X for every t > 0 is a solution. (Usually, one gets a one
parameter family of solutions.) If x 6= 0, we can assume x = 1, 000 by choosing the
appropriate parameter t. One obtains then the solution
2900 18000
x = 1, 000, y= 263.63, z= 1636.36.
11 11
For this computation, it is important to use rational numbers (i.e., fractions) as
matrix entries since otherwise the approximation to the matrix In A usually will
be invertible and only the trivial uninteresting solution x = 0, y = 0, and z = 0
will exist. This is also the reason, why the entry a33 has large numerator and
denominator.
In a closed economy, the absolute units of output are less interesting. More
important is the relative consumption of a product.
We can normalize therefore the matrix A such that the sum of every row is 1.

1 1
e such that . e .
This is a matrix A, .. = A .. . The recipe is: Divide the i-th
1 1

1
..
row of A by the i-th component of A . (that is the sum of the i-th row).
1
For our example, we have
21

1 20
1

A 1 =
5
,

1 2231
1800

leading to the matrix

1 10 10

21 21 21
1 1

Ae =

1
2 0 1
2
,

e
A 1 = 1 .
720 180 1331 1 1
2231 2231 2231

e = (e
The entries of the matrix A ai,j )i, j=1, ..., n have the following meaning: a
eij is
the relative consumption of the product of industry Si by industry Sj .

Market prices
The consumption of products is regulated by prices. All income of an industry
is used for buying other (or the own) products, i.e., income equals expenditure.
Let P = (p1, . . . , pn ) the price vector; pi is the relative price of the product of
industry Si . We can draw the flow of money into the input-output graph, the money
flows in exchange for the products:

5
 One has

p1 = a11 p1 + a21 p2 + + a
en1 pn


p2 = a12 p1 + e
e a22 p2 + + an2 p2

...


pn = a1n p1 + e
e an2 pn + + a
enn pn
,

aij pi is the amount paid by industry Sj for products produced by industry Si .

since e
The total income of industry Sj equals the total price Sj has to pay to all other
industries.
Again, one can write this as a matrix equation:

P A = P.

This equation can be transformed in the following way

P In = P Ae
e = (0, . . . , 0).
P (In A)

1 0
e e .. .. .
The matrix In A is (similar as In A) not invertible, since (In A) . = .
1 0

0
One can show that this implies that there is also a solution P 6= ... . Since with

0
P also t P for t > 0 is a solution, only the relative price between the different
products has a well-defined meaning.
Example (continued): Assume p1 = $1, 000. One gets p2 = $ 40000 63 $ 634.92
and p3 = $ 11155500
567 $ 1967.37. We can compare these relative prices with the
production levels measured by the original units and obtain the following relative
1000
prices per unit: p1/x = 1000 = 1 for one ton of wheat, p2/y 634.92263 2.4 for
1000km horse ride, and p3 /z 1967.37
1636.36 1.2 for one man-year.
Since the above matrix equation for P is not of the usual form which we have
studied so far, we make a final modification. We define
e
Ae = (e
ai,j )i, j=1, ..., n ,
e e
where a
ei,j = e
aj,i .

6
This gives us (just by switching the role of rows and columns) the price equation

e
Pe = AePe ,

where e
ai,j is now the relative consumption of industry Sj by industry Si , so that the
e

p1
sum of each column is 1, and Pe = ... is the price column vector.

pn
e
e is again denoted by A and our Pe is denoted by
In the textbook, our matrix A
X. The price equation is therefore X = A X. However, one has to keep in mind
that this matrix A is different from the input-output matrix A we used in the open
Leontief model!
Example: Let
1 1 1
2 3 4

A=

1
4
1
3
1
4
.

1 1 1
4 3 2

Compute all wages, given

that
the wages for the 3rd product is $ 30, 000.
x

Solution: Let X = y be the different wages with z = 30, 000. We have to
z
solve

X = AX

0

(I3 A)X = 0
0
1

2 23 34 x 0

34 2
34 y = 0 .
3
34 23 1
2 30, 000 0

This system of linear equations for x and y has the solution x = 30, 000 and y =
22, 500. The wages for the first and second product are therefore $, 30, 000 and
$ 22, 500, respectively.