Introductory Mathematical Economics

Lecture 1: Linear Algebra Prelims

Sugata Bag

Delhi School of Economics

11th October 2023

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 1 / 16
 Preliminaries What’s Linear Algebra?

What’s Linear algebra? Basic Concepts

Linear algebra is the branch of mathematics concerning vector spaces

and linear mappings between such spaces.
It includes the study of lines, planes, and subspaces; also involves the
analysis of rotations in space, least square …tting, solution of coupled
di¤erential equations, determination of a circle passing through three
given points; but is also concerned with properties common to all
vector spaces.
Linear algebra is our ticket to multidimensional space.
Unavoidable if we are into economics, statistics, econometrics,
computer graphics, physics, chemistry or anything quantitative. In
today’s world, that’s almost everything!!

We will start with Vector Spaces, but before that....

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 2 / 16
 Preliminaries What’s Linear?

What is “Linear” in Linear Algebra?

“Algebra” means, roughly, “relationships”.
School algebra explores the relationship between unknown numbers.
Without knowing x & y , we can still work out that expression
(x + y )2 = x 2 + 2xy + y 2 . But this is a quadratic function. [Why?].
Finding the roots of a polynomial is a nonlinear problem.
“Linear Algebra” means, roughly, “line-like relationships”.
The word “linear” is to be understood as in a linear function in
calculus [i.e. a line], such as
y = g (x ) = m x, where m, y & x are numbers.
! Said to be a linear relation (through origin) between y & x.
Or, add a constant, y = f (x ) = m x + a,
!However, this one is “a¢ ne” function [from the Latin, a¢ nis, "connected with"],
i.e. still a straight line but not through origin, has an intercept term.
More generally, if ai be the coe¢ cients (…xed real numbers) to the xi
variables and c be the constant, then a linear equation be -
a1 x1 + a2 x2 + +an xn = c
[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 3 / 16
 Preliminaries What’s Non-Linear?

Non-linear case [1]: Illustration & Issues

Ex: Let < denote the real numbers, and suppose we want to know
something about the set of common solutions of the equations
z = x 2 + xy 5 and
z2 = x + y4
These equations represent two surfaces in real three space <3 (why?),
so we’d expect the set of common solutions to lie on a curve.
Nonlinear systems, viz. in the above example, are usually di¢ cult to
solve; their theory involves highly sophisticated mathematics
Here it’s impossible to express the solutions in a “closed form”, but
we can study them “locally” using linear methods.
For example, both surfaces meet at (0, 0, 0) [any other point??], and they
both have a tangent plane at (0, 0, 0). The tangent line to the curve
of intersection at (0, 0, 0) is the intersection of these two tangents.

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 4 / 16
 Preliminaries What’s Non-Linear?

Non-linear case [2]: Linear Approximations

The Taylor’s expansion of a function f in calculus [you know this!?] is -
f (x ) = f (x0 ) + f 0 (x0 ) (x x0 ) + 1/2 f 00 (x0 )(x x0 )2 + ...
If we use only the …rst two terms of the Taylor expansion of f , we
have a linear approximation to the above function f :
f (x ) f (x0 ) + f 0 (x0 ) (x x0 )
= f 0 (x0 ) x + [f (x0 ) f 0 (x0 ) x0 ]
If we set
m = f 0 (x0 ), and a = [f (x0 ) f 0 (x0 ) x0 ],
We can see that this is a linear function just like any a¢ ne fn f (x )
above. This approximation works reasonably well near x0 .
(e.g. think of the CO/SINE function at π/2 and π/3).
It is the simplest approximation to f beyond saying “near x0 , it’s
probably something like f (x0 )”. That’s why the linear approximation
is called "easy".
However, you can go ahead and get a quadratic approximation to f
by using the …rst three terms of the Taylor expansion of f .
[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 5 / 16
 Preliminaries What’s Non-Linear?

Non-linear case [3]: How linear algebra plays a role

Now what do we do if we have a function that depends on several

variables (or, which is the same, a variable x that is a "vector"
containing these variables as components)? This is in the domain of
multivariate calculus (because, multiple variables).
In that case, f 0 (x0 )[:= m] becomes a vector that has to be multiplied
[how?# ] with anther vector (x x0 )[:= a].
Likewise, f 00 (x0 ) becomes a matrix (actually, a bilinear tensor) that
has to be multiplied twice with the vector (x x0 ). [how? Will see later]
As we have seen, the linear and quadratic approximations are the
easiest ways to think about complicated functions f .
Linear algebra deals with vectors, matrices, and tensors, and how to
operate them with each other so that everything makes sense and is
useful too. Let’s delve into it!

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 6 / 16
 Linear Syatem What’s a Linear System?

System of Simultaneous Equations & it’s Solutions

Let there be a function f : <n ! <m .
Fix y 2 <m . Then the equation f (x ) = y represents a system of ‘m’
simultaneous equations in ‘n’variables. The set of equations is -
f1 (x1 , ..., xn ) = y1 a11 x1 + .... + a1n xn = y1
............... 7 ! say , ...............
fm (x1 , ..., xn ) = ym am1 x1 + ... + amn xn = ym

where y = (y1 , ..., ym )T , x = (x1 , ..., xn )T , f (x ) = (f1 (x ), ..., fm (x )),

8x 2 <n . The component function fi : <n ! < 8 i 2 f1, ..., mg.
A vector ‘x’[why?] that satis…es f (x ) = y is a solution to system.
The system in matrix notation is: Ax = y , and
. solution is x = A 1 y , using a matrix inverse.
Note that whether the system of equations has a solution is the
same question as whether f 1 (fyg) is a "nonempty" set.
Other techniques e.g. Gaussian elimination are numerically more robust.
[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 7 / 16
 Linear Syatem What’s a Linear System?

Simultaneous Equations - ‘Contour surfaces’

Example:
Consider a function f : <2 ! <2 , where f (x1 , x2 ) = (x1 x2 ; x1 + x2 ).
For what values of y in the codomain does the equation f (x ) = y have a
solution?
[Let’s draw an analogy with consumer’s equilibrium problem]
One way to look at a solution: the intersection of (hyper) surfaces.
For the example above, given y = (y1 , y2 ), there are two equations:
x1 x2 = y1 and x1 + x2 = y2 . Both of these are 1-dimensional surfaces /
curves in <2 , and their intersection is the set of solutions.
For the more general n-variable case, fi (x1 , ..., xn ) = yi describes an
(n 1)-dimensional surface in <n , and the solution set is the
intersection of the ‘m’such surfaces.

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 8 / 16
 Linear Syatem Linear System Example

System of Linear Equations & it’s Solutions ... [1]

Basic Idea -
Example-1:
x = 1
2y (1)
3x + 2y = 11 (2)
In matrix form -

1 2 x 1
= ,
3 2 y 11
[what are the vectors here?]

! ! ! 1! 3
) [A] X = b ) X = [A] b =
1
The ‘ordered pair’ [3, 1]T is the solution, and not [1, 3]T
What about row reduction method? TRY IT!!
[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 9 / 16
 Linear Syatem Linear System Example

System of Linear Equations & it’s Solution ... [2]

(Ex-1) Row (Line) Picture:
Let’s draw the conventional picture: one equation is one line.

Solution is the intersection between the two. Simple for 2-eqn case!!
What about if a system has: 3 or more eqns? more than 2 variables?
The matrix inversion method is often cumbersome, and beyond 3-D
graphical method is next to impossible!!
[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 10 / 16
 Linear Syatem Linear System Example

System of Linear Equations & it’s Solution ... [3]

(Ex-1) Column (Vector) Picture: Try to recognize linear system as
a vector eqn. Separating original system into its columns instead of its
rows, we get “linear combination” of ‘column vectors’of matrix A -
1 2 1
x +y = =b ) Sol :: 3. col 1 + 1. col 2 = b
3 2 11

Applying, vector addition method [can you recollect?], we get the solution.
[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 11 / 16
 Linear Syatem Linear System Example

System of Linear Equations & it’s Solutions ... [4]

Example-2: with 3 unknowns with 3 equations ...

x + 2y + 3z = 6 (1)
2x + 5y + 2z = 4 (2)
6x 3y + z = 2 (3)
In matrix
2 form - 3 2 3 2 3
1 2 3 x 6
42 5 25 4y 5 = 445,
6 3 1 z 2
2 3
0
! ! ! !
Solution to [A] X = b ) X = [A] 1 b = 405
2
Can you see some thing interesting here?? (…rst 2 entries)
What about row reduction method? TRY IT!!

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 12 / 16
 Linear Syatem Linear System Example

System of Linear Equations & it’s Solution ... [5]

(Ex-2) Row (Line) Picture: Each equation is a plane in 3-D space.

First equation crosses X, Y and Z axes at points (6, 0, 0), (0, 3, 0)
and (0, 0, 2) respectively, and that de…ne the plane. Similarly others.

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 13 / 16
 Linear Syatem Linear System Example

System of Linear Equations & it’s Solution ... [6]

(Ex-2)
2 3 Column
2 3Picture:
2 3Let’2
s start
3 with the vector form of equation
1 2 3 6
x 425 + y 4 5 5 + z 425 = 445 )Correct combination is (0, 0, 2)
6 3 1 2

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 14 / 16
 Linear Syatem Linear System Example

Simultaneous Equations in Economics.

{First order conditions in optimization problems; General equilibrium
& Nash equilibrium in some problems}
All these can be cast as solutions to a system of equations z(x ) = 0,
!
where z : <n ! <n and 0 2 <n .
General equlm is often described as a …xed point of a fn. (i.e., say
f : <n ! <n , x is a …xed point if it satis…es f (x ) = x). But x is a
…xed point of f i¤ it is a zero of z(x ) f (x ) x, so the the
question is really of …nding the zeros of F i.e. solving z(x ) = 0.
Continuity of z is an important player in the existence of a solution.
(Just as in 1-D case: if f : < ! <, is continuous, and
f (x1 ) > 0 > f (x2 ), then the intermediate value theorem assures a
solution x 2 (x1 , x2 ).) For the more general higher dimensional case,
in computational economics methods like Gauss-Jacobi make use of
1-dimensional solutions to the n di¤erent equations in an iterative
way to converge to a solution. (see, Judd 1998 - Numerical Methods
in Economics).
[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 15 / 16
 Linear Syatem References

Reference:

Strang 3rd Edition:

1 Chapter 2 (for solution to linear systems)
2 Chapter 1 (for intro to vector analysis)

[SB] (Delhi School of Economics) 002: Introductory Math Econ 11th October 2023 16 / 16