Comparative Statics: Introduction

In many economic problems, the solution, or equilibrium, depends on one or more parameters, and
we want to know how the solution changes when a parameter changes

This exercise is called Comparative Statics

• In the ISLM model we find a solution for output y and the interest rate r , which depends on
government spending G and the money supply M :

y = y(G, M ), r = r(G, M )

∂y ∂r
Then and tell us how y and r change with G
∂G ∂G

M ATHS C RASH C OURSE — C OMPARATIVE S TATICS 1

 Implicit Functions and Implicit Differentiation

Consider the equation F (x, y) = 0

For each value of x you could try to solve the equation, to obtain a corresponding value of y

When possible (if there’s a unique y for each x) the equation defines y as an implicit function of x

• F (x, y) = 3x2 − 2y
• F (x, y) = y + ln(y + 1) − x
• F (x, y) = x2 + y 2 − 4

Derivative of implicit function y(x) defined by F (x, y) = 0 is found by implicit differentiation. . .

. . . differentiate the whole equation wrt x, allowing for the fact that y is a function of x

 
dy 1 dy dy y+2 dy y+1
• y + ln(y + 1) = x ⇒ + =1 ⇒ =1 ⇒ =
dx y + 1 dx dx y+1 dx y+2

M ATHS C RASH C OURSE — C OMPARATIVE S TATICS 2

 An Alternative Interpretation

Note that if x and y satisfy F (x, y) = 0, and there is an infinitesimal change in x and y so that the
equation is still satisfied, then the total differential dF = 0 is

dy F1(x, y)
dF = F1dx + F2dy = 0 ⇒ =−
dx F2(x, y)

x+y x+y x+y dy (ex+y + 1)

• e + x + 2y = 0 ⇒ (e + 2)dy + (e + 1)dx = 0 ⇒ = − x+y
dx (e + 2)
• If u(x, y) is a (well-behaved) utility function, we can find the slope of an indifference curve

∂u ∂u dy ∂u . ∂u
u(x, y) = c ⇒ dx + dy = 0 ⇒ =−
∂x ∂y dx ∂x ∂y

M ATHS C RASH C OURSE — C OMPARATIVE S TATICS 3

 Homothetic Implicit Functions

If u(x, y) is homothetic then u(x, y) = g(f (x, y)) where f is h.o.d. r and g is monotonic

Writing ux and uy for the partial derivatives, note that ux = g 0(f )fx and uy = g 0(f )fy (so the
indifference curves are the same for any g )

The slope of an indifference curve at the point (tx, ty) is

dy ux(tx, ty) fx(tx, ty) fx(x, y)

(tx, ty) = − =− =−
dx uy (tx, ty) fy (tx, ty) fy (x, y)

so the slope is the same everywhere on a ray through the origin

M ATHS C RASH C OURSE — C OMPARATIVE S TATICS 4

 The Implicit Function Theorem

Suppose x ∈ Rm, and (x∗, y ∗) is a solution to F (x, y) = 0. Suppose that F is C 1 in an open ball
around (x∗, y ∗), with ∂F ∗ ∗ 1
∂y (x , y ) 6= 0. Then there is a C function y = y(x) defined on an open
ball around x∗ such that

1. y(x∗) = y ∗

2. F (x, y(x)) = 0
∂y ∂F . ∂F
3. (x, y) = −
∂xi ∂xi ∂y

Implication: If F (x, y) = 0 has a solution y ∗ for a particular x∗, then it is legitimate to think of y as
a function of x around that point, and you can determine how y changes from y ∗ when x changes
a little from x∗ by implicitly differentiating

• x2 + y 2 = 10 has a solution x = 3, y = −1. What is dy/dx at this point?

Now 2xdx + 2ydy = 0 ⇒ dy/dx = −x/y = 3

M ATHS C RASH C OURSE — C OMPARATIVE S TATICS 5

 The Implicit Function Theorem with Several Functions of Several Variables

Consider the system of n equations in n + m variables:

F1(x1, x2, . . . , xm; y1, y2, . . . , yn) = 0 y1 = y1(x1, x2, . . . , xm)

F2(x1, x2, . . . , xm; y1, y2, . . . , yn) = 0 y2 = y2(x1, x2, . . . , xm)
... suggesting: ...
Fn(x1, x2, . . . , xm; y1, y2, . . . , yn) = 0 yn = yn(x1, x2, . . . , xm)

We can write this as

m+n n
F(x, y) = 0 where F:R →R

Suppose (x∗, y∗) is a solution to F(x, y) = 0 and that each Fi is C 1 around this point, then a
generalisation of the Implicit Function Theorem above allows you to determine how yk changes
from yk∗ when xj changes a little from x∗j . Implicitly differentiating the ith equation wrt xj

∂Fi ∂Fi ∂y1 ∂Fi ∂y2 ∂Fi ∂yn

+ + + ··· + =0
∂xj ∂y1 ∂xj ∂y2 ∂xj ∂yn ∂xj

M ATHS C RASH C OURSE — C OMPARATIVE S TATICS 6

 Several Functions Continued

Writing this in matrix form:

   ∂y   ∂F 
∂F1 ∂F1 ∂F1 1 1
...
∂y1 ∂y2 ∂yn   ∂xj  ∂xj
  ∂y2
  
∂F2 ∂F2 ∂F2  ∂F2
...
  
  ∂xj
⇒ ∂y1 ∂y2 ∂yn  = −  ∂x. j
   

 ... ... ...  .
  ..   ..


    
∂Fn ∂Fn ∂Fn ∂yn ∂Fn
∂y1 ∂y2 ... ∂yn ∂xj
∂xj

and provided the Jacobian matrix is non-singular you can solve or use Cramer’s rule

More generally, you can allow for all the xj s to change by taking the total differential:
   X ∂F 
∂F1 ∂F1 ∂F1 1
dxj
...
 
∂y1 ∂y2 ∂yn dy1  Xj ∂xj
∂F2
  
∂F2 ∂F2 ∂F2
...   dy2 dxj
    
⇒ ∂y1 ∂y2 ∂yn  = − j ∂xj
   
... ... ...   .. ...
 .
 
   
   X 
∂Fn ∂Fn ∂Fn dyn ∂F n
∂y1 ∂y2 ... ∂yn ∂x dxj
j j

M ATHS C RASH C OURSE — C OMPARATIVE S TATICS 7

 Worked Example

Consider a model of the market for umbrellas. . .

Supply: QS = S(P − t), where t is a per unit tax; S is an increasing function

Demand: QD = D(P, r), where r is rainfall; D1 < 0, D2 > 0

How do changes in r and t affect equilibrium price and quantity?

 −1
∂P ∂S ∂QS ∂S ∂P
= D2 − D1 >0 ⇒ e.g. = >0
∂r ∂P ∂r ∂P ∂r
 −1
∂P 0 ∂S ∂QD ∂P
= S − D1 >0 ⇒ e.g. = D1 <0
∂t ∂P ∂t ∂t

M ATHS C RASH C OURSE — C OMPARATIVE S TATICS 8