Microeconomics I

Michaelmas Term
Matthew Chesnes
The London School of Economics
December 14, 2001

Week 1

1.1

The Theory of the Firm

Objective: Maximize Profits by choosing output and input quantities to generate production plans subject to technology constraints.
Production Plan: z = (z1 , ..., zn ) of net inputs. If zi is positive, zi is an output of
production and if zi is negative, zi is an input into production.
Choose z from a production set, Y , where Y is the set of all technologically feasible
production plans.
Case of one output, the production plan might look like (y, x) where x is a vector of
inputs such that x = (x1 , ..., xn ) is a nonnegative vector.
Define: Input Requirement Set, V (y).
V (y) = {x Rn+ : (y, x) Y }.
In words, all combinations of inputs that allow the firm to produce y units of output,
perhaps more.
[G-1.1] Define: An Isoquant, Q(y).
Q(y) = {x Rn+ : x V (y) and x
/ V (y 0 ) when y 0 > y}.
In words, all combinations of inputs that product exactly y units of output, but nothing
more.
Define: The Production Function: The maximum output obtainable from the input
combination, x = (x1 , ..., xn ).
Cobb-Douglas Production Function: y = x1 x2 .
In general, y = f (x1 , x2 , ..., xn ).
Consider the two input case where y = f (x1 , x2 ). Totally differentiating,
dy =

f
f
dx1 +
dx2 .
x1
x2

Along an isoquant, Q(y), dy = 0, so

f
f
dx1 =
dx2 .
x1
x2
Thus,
f
dx2
x
= 1.
f
dx1
x2
Which is defined as the Marginal Rate of Technical Substitution (MRTS).
2

 Consider the case of linear technology, y = ax1 + bx2 . Thus the technical rate of
substitution is ab . So the two inputs are perfect substitutes.
Consider the case of fixed proportion technology or a Leontief production function,
y = min{ax1 , bx2 }. The technical rate of substitution is undefined and thus the two
inputs are perfect compliments.
Returns to Scale.
f (tx) = tf (x) Constant Returns to Scale (CRS).
f (tx) > tf (x) Increasing Returns to Scale (IRS).
f (tx) < tf (x) Decreasing Returns to Scale (DRS).
For a Cobb-Douglas production function, DRS if + < 1, CRS if + = 1,
and IRS if + > 1.
Homogenous Function: a function f (x) is homogenous of degree k if f (tx) = tk f (x) t >
0 and x the domain. The Cobb-Douglas production function is homogeneous of degree + .
Homothetic Function: A function f (x) is homothetic if it is a monotonically increasing
transformation of a homogenous function of degree 1. That is, there exists an increasing
function g, and a homogeneous of degree 1 function h(x), such that f (x) = g(h(x)).
Profit Maximization in a perfectly competitive market.
Let p be the price of output and w = (w1 , ..., wn ) be the vector of input prices.
The objective of the firm is to choose a production plan (y, x) to maximize
profits,
= py wx,
subject to a constraint,
(y, x) Y.
[Note: wx = w1 x1 + w2 x2 + ... + wn xn . The constraint simply specifies that (y, x)
is technologically feasible.]

Week 2
The firms problem is to choose (y, x) to maximize,
= py wx,
subject to,
y = f (x).
Substituting the firm is maximizing, = p(f (x)) wx.
First Order Conditions (FOCs):
f

p
wi = 0.
xi
xi
p

f
= wi .
xi

Thus the value of the marginal product of an input equals the marginal cost of that
input.
Second Order Conditions (SOCs): Let fij =
second order derivatives is defined as,

pf11 pf12
pf21 pf22
H=
...
...
pfn1 pfn2

2f
. Thus the Hessian matrix of
xi xj

... pf1n
... pf2n
.
... ...
... pfnn

(1)

Necessary SOC H matrix must be negative semidefinite.

Sufficient SOC H matrix must be negative definite.
Both of these conditions are LOCAL conditions.
Solving the FOCs yields xi (p, w) for i = 1...n. These are the unconditional factor
demands, or the optimal amount of input i given prices p and x. Thus f (x(p, w)) is
the amount of output a firm chooses to produce given prices. Thus f (x(p, w)) is the
firms supply function. Denote:
f (x(p, w)) = y(p, w) Firms Supply Function.
Profits of the firm:
(p, w) = p y(p, w) w x(p, w).
Which is the maximum profits a firm can achieve at given prices.
4

 Both the supply function and the input demand function are homogeneous of degree
zero. Thus, for a constant, t,
y(tp, tw) = y(p, w).
x(tp, tw) = x(p, w).
Property of the supply function, y is that it is increasing in the output price, p. The
input demand functions are decreasing in the input prices, w.
Consider a general production function, f (x1 , x2 ). The FOCs would be,
p

f
w1 = 0,
x1

f
w2 = 0.
x2
Totally differentiating the FOCs yields,
p

2f
2f
dx
+
p
dx2 dw1 = 0,
1
x21
x1 x2

2f
2f
dx1 + p 2 dx2 dw2 = 0.
x2 x1
x2

Or, in matrix form,



pf11 pf12
pf21 pf22



dx1
dx2

1
0


dw1

0
1


dw2 = 0.

(2)

(3)

Thus, via cramers rule,



 1 pf12 


 0 pf22 
x1
pf22
=
= 

 pf11 pf12
w1
 pf11 pf12 

 pf21 pf22 
 pf21 pf22

Since the first principal minor is negative and the second is positive, the matrix is
negative semidefinite and thus all diagonal entries are non-positive so pf22 0. Thus,
pf22

 pf11 pf12

 pf21 pf22

x1
=
0.

w
1



(4)

Or the unconditional input demand is downward sloping in prices.

Warning! Profit maximization only works as described above if there are DRS. If there
are CRS or IRS, profits will grow without bounds by buying up more and more inputs.
5

2.1

Properties of the Profit Function

(p, w) is non-decreasing in p and is non-increasing in wi for i = 1...n.

Proof: Let p0 > p. Thus,
(p, w) p0 y(p, w) wx(p, w) (p0 , w)
This first inequality is just algebra, but the second comes from the fact that is
maximized for (p0 , w) so using any other (p , w ) will yield lower profits.
Thus if p0 > p, Then (p0 w) (p, w), or (p, w) is non-decreasing in p.
(p, w) is homogeneous of degree 1 in prices. Proof: By definition,
py(p, w) wx(p, w) py 0 wx0 (y 0 , x0 ) Y,
because the first expression of profits is evaluated at the profit maximizing values of p
and w. Multiplying both sides by t,
tpy(p, w) twx(p, w) tpy 0 twx0 (y 0 , x0 ) Y
Hence, (y(p, w), x(p, w)) is the production plan that maximizes profits at (tp, tw).
Thus (tp, tw) = t(p, w), or is homogeneous of degree 1.
(p, w) is convex in prices.
Proof: Take prices (p, w) and (p0 , w0 ) and define,
(p00 , w00 ) = (p, w) + (1 )(p0 , w0 ) for [0, 1].
Now, by definition of profit maximization,
py(p00 , w00 ) wx(p00 , w00 ) (p, w).
Mulitplying both sides by , ()
py(p00 , w00 ) wx(p00 , w00 ) (p, w).
Also by definition of profit maximization,
p0 y(p00 , w00 ) w0 x(p00 , w00 ) (p0 , w0 ).
Mulitplying both sides by (1 ), ()
(1 )p0 y(p00 , w00 ) (1 )w0 x(p00 , w00 ) (1 )(p0 , w0 ).
Summing () and (), yields,
p00 y(p00 , w00 ) w00 x(p00 , w00 ) (p, w) + (1 )(p0 , w0 ).
Thus,
(p00 , w00 ) (p, w) + (1 )(p0 , w0 ).
So, is convex. QED.
(p, w) is continuous on (p, w).
Proof: QED.
6

Week 3

3.1

Duality

We have shown thus far that f (x) y(p, w), x(p, w) (p, w).
But can we go the other direction? yes. To do this, we first need the envelope theorem.
Envelope Theorem: Supose that we maximize a function g(s, a) with respect to s where
a is a parameter. We would obtain a solution s(a). Now define the maximized value
of the function as M (a) = g(s(a), a). By the chain rule,
dM (a)
g s g
=
+
.
da
s a a
But since we already defined s(a) to be the maximium solution of g,

g
= 0. Thus
s

dM (a)
g
=
.
da
a
Now consider the profit function,
(p, w) = py(p, w) wx(p, w).
Thus, by the envelope theorem,

= y(p, w).
p

= xi (p, w).
wi
Thus,
2
y(p, w)
=
Slope of the Supply Function.
2
p
p
The slope of the supply function is positive because is convex. Also,
2
xi (p, wi )
=
0.
2
wi
wi
Thus, the input demands are downward sloping. Finally,
2
xi (p, w)
xi (p, w)
2
=
=
=
.
wi wj
wi
wj
wj wi
Because the order that you take the partials does not matter (Youngs Theorem).
This process of taking the partials of the profit function with respect to price, p, to find
the supply function, and with respect to the input price, w, to find the input demand
is called the Hotelling Lemma.
7

3.2

Cost Minimization

The objective of the firm can also be to choose an input combination x = (x1 , ..., xn )
to minimize wx subject to y = f (x). Thus,
L = wx (f (x) y).
FOCs:

L
= y f (x) = 0.

f
L
= wi
= 0 for i = 1...n.
wi
xi

Thus,
wi =

f
i.
xi

Thus,
f
wi
xi
=
.
f
wj
xj
Which implies that the slope of the isocost = the slope of the isoquant at equilibrium.
[G-3.1]
To determine the SOCs, we must look at the Boarded Hessian. Consider the FOCs:
y f (x) = 0.
f
= 0.
xi
Thus take the derivative of all the FOCs with respect to and all the inputs, x. Thus,
wi

H=

0
f1
f1 f11
f2 f21
...
...
fn fn1

...
...
...
...
...

fn
f1n
f2n
...
fnn

(5)

If we solve the FOCs, we find a function xi (w, y) i. These are the conditional factor
demands because they depend on the quantity of output, y. If we sum up all the
conditional factor demands multiplied by their costs,
wx(w, y) =

n
X

wi xi (w, y) = C(w, y),

i=1

Where C(w, y) is the cost function.

8

 Properties of the Cost Function, C(w, y).

C(w, y) is nondecreasing in wi for i = 1...n.
C(w, y) is homogeneous of degree 1 in w.
C(w, y) is concave in w.
C(w, y) is continuous in w.
Duality of the Cost Function:
C(w, y) =

n
X

wi xi (w, y).

i=1

Thus, by the envelope theorem,

C(w, y)
= xi (w, y) i.
wi
So by the Shephards Lemma, taking the partial of the cost function with respect to
the input prices yields the respective input demand function. Also,
2 C(w, y)
xi (w, y)
=
Slope of the Conditional Factor Demand.
2
wi
wi
Since the Cost function is concave, C is negative semidefinite. Thus all diagonal entries
are negative. Thus the conditional input demand functions are downward sloping.
Finally, having obtained now the cost function, we can write it simply as C(y). Thus,
= py C(y).
To maximize profits, a firm chooses y to maximize py C(y), thus the FOC:

C
p
= 0.
y
y
Or the familiar result,
Price = Marginal Cost.

Week 4
Costs: C(y) = V C(y)+F , where V C are the variable costs that depend on the quantity
of output and F are the fixed costs. Fixed costs can be divided into Long and Short
run fixed costs.
Average Costs:
AC(y) =

V C(y) F
+ = AV C(y) + AF C(y).
y
y

If AC(y) is decreasing in y, then we have economies of scale.

If AC(y) is increasing in y, then we have diseconomies of scale.
IRS Economies of Scale, but NOT necessarily vice versa. Symmetrically, DRS
Diseconomies of Scale but NOT vice versa.
Marginal Cost:

dC(y)
= C 0 (y) MC.
dy

Since AC(y) =

C(y)
,
y

"
#
C(y)
yC 0 (y) C(y)
C 0 (y) C(y)
d
AC 0 (y) =
=
=
2 .
dy
y
y2
y
y
Therefore, when AC 0 (y) = 0,
C 0 (y)
C(y)
= 2 .
y
y
C(y)
.
y
Or the M C intersects the AC curve at the minimum of the AC. The same can be
shown for the AV C : M C = AV C at the minimum of the AV C.
C 0 (y) = M C(y) =

Short Run Firm Supply.

Assume fixed costs are sunk. Thus,
SR = y(P AV C(y)) F.
Firms only supply product if P > AV C(y), so graphically [G-4.1] the short
run supply curve is the M C curve above the minimum of the AV C curve and 0
elsewhere.
Short run aggregate supply assuming m firms:
Y (p) =

m
X
i=1

10

yi (p).

 Long Run Firm Supply.

A firms supply is positive if p > min(AC).
Assume no barriers to entry.
Thus if p > min(AC), > 0 and firms have an incentive to enter and drive the
price down.
Also if p < min(AC), < 0 and firms have an incentive to exit and bring up the
price.
Thus, in equilibrium, p = min(AC), or the long run supply curve is perfectly
elastic.
More on Duality. We have shown that via shepherds lemma, we can go from the
cost function to the conditional factor demands by differentiating C(y) with respect
to w. But can we go from the conditional factor demands to the technology? Yes, in
most cases. By looking at several input and price combinations and determining the
equilibrium we begin to trace out an isoquant by looking at the slope of the isocost
lines at each point. This only works if the technology is convex! [G-4.2]

4.1

Consumer Theory

Consumer chooses best (based on preferences) available (based on budget constraint)

choice.
Domain of choice: the consumption set, X Rn+ . A typical element of X is x =
(x1 , ..., xn ), which is a consumption bundle.
Preferences:
x p y x is weakly prefered to y.
x p y x is strictly prefered to y if x p y but NOT y p x.
x p y x is indifferent to y if x p y and y p x.
Properties of Preferences.
Completeness: For any x, y X, either x p y or y p x, or both. Thus all
bundles can be ranked.
Transitivity: For x, y, z X, if x p y and y p z, then x p z. Also holds
for strict preference and indifference.
Theorem: Suppose the consumption set, X, is finite. If preferences are complete and
transitive, then there exists a function u : X 7 R 3 x, y X, u(x) u(y) if and
only if x p y.
Proof: Suppose, for simplicity, all bundles are ordered by strict preference. Then there
exists a most preferred bundle. To show this, proceed by contradiction. Assume there
is not a most preferred bundle. Then for any bundle, say x0 , I can find another bundle,
11

x00 , such that x00 p x0 . Starting with x00 , there exists a x000 such that x000 p x00 . And so
on... If there is only a finite number of bundles some bundles must be repeated. But
this contradicts transitivity. So there must exist a most preferred bundle, call it x .
Set u(x ) = c1 , where c1 is a constant. Now find the second most prefered bundle, y
and set u(y ) = c2 where c2 < c1 . This can continue until u() is completely defined as
described above. QED.

4.2

More Accurate Definiteness Properties

Test for definiteness as it applies to symmetric matrices (such as the Hessian):

The matrix is positive definite iff all the leading principal minors are strictly positive.
The matrix is negative definite iff all the odd-order leading principal minors are strictly
negative and all the even-order leading principal minors are strictly positive.
The matrix is positive semidefinite iff every principal minor (leading and nonleading)
is nonnegative.
The matrix is negative semidefinite iff every principal minor (leading and nonleading)
of odd-order is not positive, and every principal minor of even-order is not negative.
Of course, if we find a matrix to be postive definite then we know it is positive semidefinite and if we find matrix to be negative definite then it is negative semidefinite and
we therefore do not need to look at the nonleading principal minors.
As an example, consider a 3x3 symmetric matrix where H1 , H2 , and H3 are the leading
principal minors:
If H1 > 0, H2 > 0, H3 > 0,
then the matrix is Postive Definite.
If H1 < 0, H2 > 0, H3 < 0,
then the matrix is Negative Definite.
If H1 > 0, H2 > 0, H3 = 0,
then the matrix is not definite.
In this case the matrix is either Postive Semidefinite or Indefinite. To determine if
it is positive semidefinite, we must test the rest of the principal minors (i.e. the
non-leading ones). If every principal minor is nonnegative, then the matrix is indeed
positive semidefinite.
If H1 < 0, H2 > 0, H3 = 0,
12

then the matrix is either Negative Semidefinite or Indefinite.

Again, we must test all the principal minors to determine if it is negative semidefinite.
If H1 = 0, H2 > 0, H3 = 0,
then the matrix is either Postitive Semidefinite,
Negative Semidefinite, or Indefinite.
Again, we must test all principal minors to determine which it is.
NOTE: When considering a Boardered Hessian, everything is the exact opposite. You
get negative semi-definitness when the odd principal leading minor determinants are
all non-negative and the even are non-positive. This excludes the first, which is alway
zero, and the second which is always negative, so you only have to worry about the
3rd and greater. For a negative definite hessian matrix therefore, all principal minor
determinants must be strictly positive!

13

Week 5
Utility is an ordinal concept. Only the relative ranks matter. If u(x) represents the
preferences of the consumer, so does u(x) + c because all utility levels just shift by c
but the relative position remains the same.
Any monotonic transformation that preserves the intial utility ranking leaves the utility
function unchanged. Thus [u(x)]k and ln(u(x)) are ok transformations.
Definition: Utility function.
u : x 7 R 3 x p y iff u(x) u(y) for any x, y X.
Utility functions are transitive and complete.
Definition: Indifference Curves: The set of consumption bundles, x, for which u(x) is
constant.
Consider the case of two commodities so that x = (x1 , x2 ). An indifference curve is
then u(x1 , x2 ) = c, where c is constant. Totally differentiating,
u
u
dx1 +
dx2 = 0.
x1
x2
Thus,
dx2
=
dx1

u
x1
u
x2

Or,
dx2
= The slope of the indifference curve.
dx1
Thus the slope of the indifference curve equals the Marginal Rate of Substitution
(MRS).
Example: Cobb-Douglas Utility function. u(x1 , x2 ) = x1 x2 . Thus,
M RS =

x1
x2
1
x1 x1
2

x2
.
x1

Example: Perfectly Substitutable Commodities: u(x1 , x2 ) = ax1 + bx2 . Thus,

a
M RS = .
b
Example: Perfectly Complementary Commodities: u(x1 , x2 ) = min{ax1 , bx2 }. Thus,
M RS = Undefined.
This is because the indifference curves are rectangular so we get a corner solution where
the slope is undefined.
14

5.1

The Consumers Problem

A consumer chooses a consumption bundle x = (x1 , x2 , ..., xn ) to maximize u(x1 , x2 , ..., xn )

subject to p1 x1 + p2 x2 + ... + pn xn m, where m is the consumers income.
Notice that the constraint is written as an inequality. It can be written as a strict
equality if we have the following assumption.
Local NonSatiation (LN S)
For any x X and  > 0, y X 3 d(x, y) <  and u(y) > u(x).
Where d(x, y) represents the distance between x and y. So in words, it just means that
for any bundle of goods, we can find some other bundle close to it where the utility
at the new bundle is strictly greater. A little more (possibly very little) can always be
better.
If utility functions are monotonic, you definitely get LN S, but not vice versa. Having
LN S does not necessarily mean that u(x) is monotonic. Monotonicity is stronger than
LN S.
If LN S holds, you cannot be in an equilibrium state where you are not spending all
available income. You can spend a little bit more and make yourself strictly better
off. Thus, repeating this process, you eventually reach your income so the consumers
problem changes.
Consumers Problem assuming LN S. Choose x X to maximize,
u(x1 , x2 , ..., xn ),
subject to,
p1 x1 + p2 x2 + ... + pn xn = m.
Or in vector notation, maximize:
u(x) subject to px = m,
where p is a price vector.
Solving the Consumers Problem, via the LaGrangian,
L = u(x) + (m px).
FOCs:

L
= m px = 0.

L
u
=
pi = 0 i.
xi
xi
15

 SOCs. Let the first F OC be L1 and the second set of F OCs be L2 , L3 , .

H=

L1

L2

L3

...
Ln

L1
x1
L2
x1
L3
x1
...
Ln
x1

...
...
...
...
...

L1
xn
L2
xn
L3
xn
...
Ln
xn

0 p1
p1 u11
p2 u21
...
...
pn un1

... pn
... u1n

... u2n
.
... ...
... unn

(6)

should be negative semidefinite because we are

This boardered Hessian matrix, H,
looking for a maximum. Thus the leading principal minor determinants starting with
the 3rd should alternate in sign and the 3rd should be positive. Note again that this
applies only to boardered hessians because we are adding on an extra row and column
so everything is backward compared a regular matrix. Note that you start with the
3rd because the 1st is always 0 and the 2nd is always negative.
Now consider the second set of F OCs:
L
u
=
pi = 0 i.
xi
xi
Thus,
u
= pi .
xi
Therefore comparing commodity i and j, we have
u
xi
u
xj

pi
.
pj

Which in words says that the Marginal Rate of Substitution (M RS) is equal to the
price ratio, which is the slope of the Budget Constraint (BC).
In order to have one unique solution to the consumers problem, there must exist only
one tangency between the consumers indifference curves and budget constraint. Thus,
Convex indifference curves or therefore Quasi-concave utility functions will avoid any
problems with multiple solutions. [G-5.1]
Example using Cobb-Douglas Utility Function: u(x1 , x2 ) = x1 x2 . Thus,
M RS =

x2
p1
= .
x1
p2

p2 x2 =

16

p1 x1 .

And from the budget constraint, substituting in,

p1 x1 + p2 x2 = m.

p1 x1 = m.

p1 x1 (1 + ) = m.

p1 x1 +

Thus,
x1 =

m
.
p1 +

x2 =

m
.
p2 +

And symmetrically,

These last two equations are called the Marshallian Demands.

Examples in the notes show that if one utility function is a monotonic transformation
of another, then the Marshallian demands that come out of the maximization problem
will be the same. Also diminishing marginal utility is not meaningful in this framework
(for some reason). With perfect substitutes, the marshallian demands depend on the
prices of the two goods. If one good is cheaper than another, the consumer will just
consume all of the cheaper good (and vice versa). If the prices of the goods are equal,
then any choice on the budget line is optimal.

17

Week 6
Perfect Complements Utility Functions: u(x1 , x2 ) = min{ax1 , bx2 }. Let a = b = 1 and
we know that because of the rectangular utility functions [G-6.1], the optimal solution
will be such that x1 = x2 . Thus the budget constraint becomes,
p1 x1 + p2 x2 = p1 x1 + p2 x1 = m.
Thus, the marhshallian demands,
m
.
p1 + p2
m
.
x2 =
p1 + p2

Quasi Linear Utility Functions: u(x1 , x2 ) = 2 x1 + x2 , where the first term is clearly
nonlinear but the rest is linear. To solve this for the Marshallian demands, do the
usual and set MRS = Price Ratio. Thus,
x1 =

u
x1
u
x2

1/2

p1
x
=
1
p2
1

p1
.
p2

Thus,

p2 2
).
p1
And substituting into the budget constraint,
p2
m p2
.
p1 ( )2 + p2 x2 = m x2 =
p1
p2 p1
x1 = (

The only problem with this is x2 might be negative under the right choices for m, p1 ,
and p2 . Thus Marshallian demands are,
p2
m p2
m p2
x1 = ( )2 and x2 =

if

0.
p1
p2 p1
p2 p1
m
m p2
x1 =
and x2 = 0 if

< 0.
p1
p2 p1
Thus with quasi-linear utility functions, one must be careful because the tangency
condition might yield negative demands and thus the solution needs to be adjusted to
reflect this. [G-6.2]

6.1

Price and and Income Changes

Price Offer Curve: Take your typical utility maximizing setup with the utility curve
just tangent to the budget constraint. Then change the price of one of the commodities.
Thus the budget line shifts and you will get a new equilibrium at a new utility level.
Continuing this process and connecting the equilibrium points yields the Price Offer
Curve. If the curve slopes downward in xi , pi space then the good is called Ordinary:
The lower the price of good i, the more of it that is demanded.
18

 If the price offer curve slopes upward in xi , pi space then the good is called a Giffen
Good: The lower the price of good i, the less of it that is demanded. [G-6.3]
Now consider the same utility/budget constraint set up and vary the income level.
Record the different equilibriums as you adjust income. The curve that connects all
the equilibriums is called the Income Expansion Path. Drawn in xi , m space it is
refered to as an Engel Curve. If the Engel curve slopes upwards in xi , m space, then
the good is called a Normal Good: The higher a consumers income level, the more of
good i that is demanded.
If the Engel curve slopes downwards in xi , m space, then the good is called an Inferior
Good: The higher a consumers income level, the less of good i that is demanded.
Examples of inferior goods are cheap food products and even cigarettes. [G-6.4]

6.2

Indirect Utility Functions

From Utility maximization, we obtain a vector of marshallian demands x(p, m). Plugging the vector of marshallian demands into the utility function, we obtain the Indirect
Utility Function, v(p, m). That is,
v(p, m) = u(x(p, m)).
Example of a Cobb Douglas Utility Function: u(x1 , x2 ) = x1 x2 . MRS = Price ratio
implies,
u
p1
x2
p1
x1
=

= .
u
p2
x1
p2
x2
Thus,

p2
).
p1
p1
x2 = x1 ( ).
p2
x1 = x2 (

And substituting into the budget constraint,

m = p1 x1 + p2 [x1 (
So Marshallian Demands,

p1
)] = 2p1 x1 .
p2

m
.
2p1
m
x2 =
.
2p2
x1 =

Substituting the Marshallian demands back into the utility functions,

v(p, m) = u(x1 (p1 , m), x2 (p2 , m)) =

19

m m
m2
=
.
2p1 2p2
4p1 p2

6.3

Properties of the Indirect Utility Function

Non-increasing in p and non-decreasing in m. See notes for simple proof.

Homogeneous of degree 0 in (p, m). If all prices and income level go up by the same
amount, the budget line doesnt shift at all, thus the utility level is constant. tpx =
tm px = m.
Quasi-Convex in Prices.
Proof: Take two prices vectors, p and p0 and consider another price vector,
p00 = p + (1 )p0 with [0, 1].
Now consider a consumption bundle, x 3 p00 x m. Thus x is affordable at prices,
p00 . This implies that x must be affordable at either p, p0 , or both. Thus px m or
p0 x m or both. Hence,
v(p00 , m) max{v(p, m), v(p0 , m)}.
Which is the definition of quasi-convex in p.
v(p, m) is continuous if u(x) is continuous. No proof.

6.4

Expenditure Function

A consumer chooses the bundle x to minimize px subject to u(x) u. As a simply analogue to the cost function of the firm, out of this optimisation we obtain an
expenditure function with the same properties as the cost function.
Properties of the Expenditure function e(p, u): 1) Non-decreasing in p; 2) Homogeneous
of degree 1 in p; 3) Concave in p; 4) Continuous.
Just as we took the partial of the cost function with respect to input prices to get
the conditional factor demands (via Shephards lemma), we can take the partial of
the expenditure function with respect to prices to get the compensated or Hicksian
Demands.
e(p, u)
= hi (p, u).
pi
Next we have a series of identities which we will use in the coming sections,
Identity 1: e(p, v(p, m)) = m.
Identity 2: v(p, e(p, u)) = u.
Identity 3: hi (p, v(p, m)) = xi (p, m).
Identity 4: xi (p, e(p, u)) = hi (p, u).

20

 Taking the total differential of the 2nd identity,

v(p, e(p, u)) v(p, e(p, u)) e(p, u)
+
= 0.
pi
m
pi
v(p, e(p, u)) v(p, e(p, u))
+
hi (p, u) = 0.
pi
m
Evalutated at u = v(p, m),
v(p, m) v(p, m)
+
xi (p, m) = 0.
pi
m
Hence solving for the Marshallian demand,
v(p, m)
pi
xi (p, m) =
.
v(p, m)
m
This last Equality is called Roys Identity named after Roy. You can obtain Marshallian
demands from the indirect utility function.
Example of finding an expenditure functions: Consider a Cobb Douglas Utility function: u = x1 x2 . Thus,
1 m2
v(p, m) =
.
4 p1 p2
Solving for m,
p

m = 2 p1 p2 v(p, m) = 2 p1 p2 u = e(p, u).

We would now like to develop restrictions on the Marshallian demands. Consider,
Identity 4: xi (p, e(p, u)) = hi (p, u).
Taking to total differential,
xi (p, e(p, u)) xi (p, e(p, u)) e(p, u)
hi (p, u)
+
=
.
pj
m
pj
pj
xi (p, e(p, u)) xi (p, e(p, u))
hi (p, u)
+
hj (p, u) =
.
pj
m
pj
Evluated at u = v(p, m),
xi (p, m) xi (p, m)
hi (p, u)
+
xj (p, m) =
.
pj
m
pj
And this last equality is called the Slutsky Equation. Note that the left side only
depends on p and m which are usually observable quantities.
21

 Simplfying the notation of the Slutsky equation,

ij (p, m) =

hi (p, u)
.
pj

Thus in matrix form,

11 . . . 1n

..
.
..
.. =
.
.

n1 . . . nn

h1
h1
...
p1
pn
.. . .
.
. ..
.
hn
hn
...
p1
pn

(7)

Notice that the right hand matrix of partials of the hicksian demand equation is equivalent to the matrix of second partials of the expenditure function. Thus since e(p, u)
is concave, this matrix must be negative semidefinite. Thus the left hand matrix must
also be negative semidefinite. The ij (p, m) matrix is called the Substitution Matrix.
Finally, rearranging the terms of the Slutsky equation,
x1
=
p1

h1
p1
|{z}

SubstitutionEf f ect

x1
m
|{z}

x1 .

IncomeEf f ect

h1
x1
(the substitution effect) is always negative and
is positive when x1 is a
p1
m
x1
normal good,
< 0 or demand for good i is downward sloping as would be expected.
p1
Since

x1
< 0 then we have an inferior good. If this income effect term is large enough
m
x1
(negative enough), then
might actually become positive which would imply that
p1
the commodity is a giffen good. Thus ALL giffen goods are inferior but not vice versa.

If

22

Week 7

7.1

Comparative Statics

The Consumers Problem: Choose x X to maximize u(x) subject to px = m. Set up

the lagrangian: L = u(x)+(mpx) which yields FOCs (for the case of 2 commodities),
L
= m p1 x1 p2 x2 = 0.

L
u
=
p1 = 0.
x1
x1
L
u
=
p2 = 0.
x2
x2
Totally differentiating the first FOC for utility maximization yields,
dm x1 dp1 p1 dx1 x2 dp2 p2 dx2 = 0.
Or,
p1 dx1 p2 dx2 = x1 dp1 + x2 dp2 dm.
The second FOC:
u11 dx1 + u12 dx2 dp1 p1 d = 0.
Or,
p1 d + u11 dx1 + u12 dx2 = dp1 .
Similarly for the third FOC:
u22 dx2 + u21 dx1 dp2 p2 d = 0.
Or,
p2 d + u21 dx1 + u22 dx2 = dp2 .
Written in matrix form then,

0 p1 p2
d
x1
x2
1
p1 u11 u12 dx1 = dp1 + 0 dp2 + 0 dm.
0

0
p2 u12 u22
dx2

(8)

Via cramers rule on the above matrix, one can find all the effects of prices and income
and consumption under utility maximization. It can also show the decomposition of
dx1
into income and substitution effects (Slutsky Equations).
dp1

23

7.2

Labor Supply

Definitions: c consumption, l leisure, p price of consumption, and w wage

rate.
An individuals utility is derived from consumption and leisure, u(c, l). T is the total
time available to the consumer.
The consumer chooses consumption and leisure to maximize u(c, l) subject to
pc = (T l)w.
Or,
pc + wl = |{z}
wT .
| {z }

Expenditure

Income

Solving, we obtain the demand for consumption: c(p, w, wT ) and the demand for
leisure: l(p, w, wT ). Differentiating the leisure equation with respect to wage, one
obtains,
dl
l
l d(wT )
l
l
l
l
=
+
=
+
T =
+
T.
dw
w wT dw
w wT
w m
Because wT = m, or the total possible income available if the individual worked as
much as possible.
Now the change in leisure (or T l = work hours) from a change in the wage rate can
be decomposed into an income and substitution effect. Thus,
dl
=
dw

l
l
l
+
T.
m m
|{z}
Substitution
s
|{z}

Income

Rearranging,
+

dl
l
= s +(T l)
.
dw
m

l
dl
If s < (T l)
, then
> 0 so as the wage rises, people work less. Hence if the
m
dw
substitution effect is outweighed by a strong income effect, we see the possibility of
backward bending labor supply curves.

7.3

Welfare

Consider money as the standard for welfare in an economy. Consider the following
definitions: p0 is the initial price vector, p1 is the final price vector, u0 is the utility of
the consumer at p0 and u1 is the utility of the consumer at p1 .

24

 Consider an initial set up of 2 commodities with price vector and corresponding utility
level, p0 and u0 . Suppose the price of good 1 falls pivoting the budget constraint out.
New utility level at p1 , u1 . Instead of the price changing to reach u1 , we could have
also obtained it via an increase in income. By definition, the income needed to reach
u1 at the old prices is the total expenditure: e(p0 , u1 ). Thus the change in income
necessary for this increase in utility is:
e(p0 , u1 ) e(p0 , u0 ) = e(p0 , u1 ) m Equivalent Variation EV.
Note: Old Prices, New Utility. [G-7.1]
Next consider an initial set up of 2 commodities with price vector and corresponding
utility level, p0 and u0 . Suppose the price of good 1 falls pivoting the budget constraint
out. New utility level at p1 , u1 . After the price change, we could take away some of the
consumers income to bring him back to u0 . By definition, the income needed to reach
u0 at the new prices is the total expenditure: e(p1 , u0 ). Thus the change in income
necessary to maintain the original level of utility is:
e(p1 , u1 ) e(p1 , u0 ) = m e(p1 , u0 ) Compensating Variation CV.
Note: New Prices, Old Utility.
Thus this can be summarized by:
V (p1 , m CV ) = u0 .
V (p0 , m + EV ) = u1 .
In general, CV 6= EV .
Suppose in the new price vector, only the price of good 1 falls, while the rest of the
commodity prices remain constant. Thus p11 < p01 , where the superscript represents the
price vector and the subscript is the commodity index. We have defined,
EV = e(p0 , u1 ) m = e(p0 , u1 ) e(p1 , u1 ).
NOTE: f (x0 ) f (x1 ) =

R x0 df
dy. Thus,
x1
dy
Z
EV =

p01

p11

de
dp1 .
dp

The derivative of the expenditure function is the hicksian demand, so:

Z p01
EV =
h1 (p1 , p02 , p03 , ..., p0n , u1 )dp1 .
p11

Similarly,
CV = m e(p1 , u0 ) = e(p0 , u0 ) e(p1 , u0 ).
Z p01
CV =
h1 (p1 , p02 , p03 , ..., p0n , u0 )dp1 .
p11

25

 Note that the only difference in the CV and EV is the utility levels. If p1 falls, u1 > u0 ,
but comparing EV and CV will depend on if x1 is normal or inferior.
Consider the marshallian demand which by an identity,
x1 (p, m) = h1 (p, v(p, m)).
Since v(p, m) is increasing in m, if good 1 is a normal good,
0 or

dx1
dh1
> 0. Thus
>
dm
dv(p, m)

dh1
> 0. Thus we have EV CV iff the good is normal.
du

Since we have defined EV and CV in terms of integrals, they can be viewed graphically
as areas under curves. Consider good 1 as a normal good. Thus,


0 0
0
0
0 0
0
0 0
0
h1 (p1 , p2 , p3 , ..., pn , u ) h1 p1 , p2 , p3 , ..., pn , V (p1 , p2 , p3 , ..., pn , m) h1 (p1 , p02 , p03 , ..., p0n , u1 ).
|
{z
}
M arshallian Demand f or x1

for p11 p1 p01 .

A graphical display of this can be seen in notes. [G-7.2] In general, CV Consumer
Surplus (CS) EV .
If Marshallian does not depend on income, h1 (p) = x1 (p) or the above inequality
becomes equality and CV = CS = EV .
Example of Welfare comparison. Suppose there are K consumers. Let xji (p, mj ) be the
demand for good i by consumer j having income mj . Thus in aggregate,
Di =

K
X

xji (p, mj ) Aggregate Demand.

j=1

More specifically, let

xji = j (p) + (p)mj .
Thus, in aggregate,
Di =

K
X

(p) + (p)

j=1

K
X

mj .

j=1

So demand
good i can be throught of as consumer js demand with aggregate
P for
j
income j m . However, aggregate demands are distribution independent iff every
consumer j has an indirect utility function,
v j (p, mj ) = aj (p) + b(p)mj .
(Or any monotonic tranformation there of.)
26

 This type of indirect utility function is called a Gorman Form. It is satisfied if

consumers have identical homothetic preferences or quasilinear preferences (relatively
strict conditions).
Now suppose that consumers do have indirect utilities of the Gorman Form. Thus the
aggregate indirect utility is,
V (p, m) = a(p) + b(p) + m,
P
PK
j
j
where a(p) = K
j=1 a (p) and m =
j=1 m .
Suppose V (p1 , m) V (p0 , m) = c > 0. At the price level p1 , the aggregate indirect
utility function of this society is higher. So is society better off? Yes, but only if there
is a redistribution of income. Consider new income levels:
(m
1, m
2 , ..., m
K)
such that,

c
aj (p0 ) + b(p0 )m
j = aj (p0 ) + b(p0 )mj + .
k
The last term is the redistribution term of the utility surplus spread out equally over
all consumers.
Does this work? To see that it does, sum both sides across consumers,
X
X
V (p0 ,
m
j ) = V (p0 ,
mj ) + c.
j

Hence,
K
X

m
j = m.

j=1

Moreover, at income levels, (m

1, m
2 , ..., m
K ), every consumer is better off. Fascinating.

27

Week 8

8.1

Cost of Living Indices

Suppose p0 is the initial price vector and corresponds to a utility function for the
consumer of u0 . Also, p1 is the final price vector corresponding to the utility for the
consumer of u1 .
A true measure of the cost of living would be the ratio of the expenditure functions at
the different prices levels that would result in the same utility level:
e(p1 , u0 )
.
e(p0 , u0 )
But since utility levels are unobservable, we estimate it with the Laspeyres index, L,
such that,
P 1X 0
L = 0 0,
P X
0
where X is the initial bundle of goods.
e(p1 , u0 )
because the L index compares the same bundle of goods.
e(p0 , u0 )
When the price level changes, consumers will reoptimize their consumption and most
likely they will consume different quantities of goods in the bundle depending on the
new relative prices. Thus, when there is a price decrease in the economy for example,
the L estimate of the cost living over estimates the true measure.

In all likelihood, L

8.2

Markets

We have developed a measure for the concept of welfare for the consumer: EV , CV ,
and consumer surplus. But now we would like to find a similar (monetary) measure
for the producer. The obvious candidate is profit.
Suppose the output price increases from p0 to p1 where these prices are single prices
and not a vector as in previous sections.
The change in profits for the producer is:
= (p1 ) (p0 ).
Via the hotelling Lemma,
(p, w)
= y(p, w).
p
Thus, we can write as an integral as follows:
Z p1
Z p1
(p, w)
=
dp =
y(p)dp.
p
p0
p0
28

 Therefore, as seen in the notes [G-8.1], the difference in profits, , or otherwise called
Producer Surplus, can be shown as the different between the two prices under the
supply curve.
The same graph can be drawn for aggregate supply with the area representing aggregate
producer surplus.

8.3

Price Formation

Equilibrium market price is defined in elementary economics where Demand, D(p),

equals Supply, S(p).
To measure the responsiveness of these curves to prices, one could look at the slopes of
the curves, but unfortunately because of a difference in units, two equilivalent demand
or supply functions might have different slopes. Thus, consider elasticities:
dD(p) p
.
dp D(p)

Elasticity of Demand: D =
Elasticity of Supply: S =

8.4

dS(p) p
.
dp S(p)

Taxes

Two types of taxes, quantity taxes and value taxes.

Let pD be the price paid by the consumer and let pS be the price received by producers.
A Quantity tax at rate t per unit would be of form:
pD = pS + t.
A Value tax at a rate t would be of the form:
pD = pS (1 + t).
Graphically it can be shown (See notes for graph [G-8.2]) that the tax revenue from
the tax is less than the loss of consumer and produce surpluses together. Thus, taxes
are distortionary and inefficient and result in a dead weight loss:
Dead Weight Loss = Consumer Surplus + Producer Surplus - Tax Revenue > 0.

29

8.5

General Equilibrium - Beyond Mock Exam

Consider a pure exchange economy with NO production.

Suppose there are n commodities and K consumers who wish to trade their endowments
of goods.
Each consumer J is described as hairy and ugly (Ed) with (uJ , wJ ) where uJ is Js
utility function and wJ is Js endowment such that,
wJ = (w1J , w2J , . . . , wnJ ).
Consider the following example which will be developed throughout the analysis.
Suppose K = 2 and n = 2. Let w1 = (3, 1) and w2 = (1, 3). The utility
functions of the two consumers are cobb-douglas such that,
u1 = x11 x12 ,
and,
u2 = x21 x22 .
Note that the superscripts are consumer indices and the subscripts are
commodity indices.
An allocation, x = (x1 , x2 , . . . , xK ) assigns a consumption bundle xJ = (xJ1 , xJ2 , . . . , xJK )
to each consumer J = 1...K.
An allocation, x = (x1 , x2 , . . . , xK ) is feasible if:
K
X

J=1

K
X

wj .

J=1

This just means that an allocation is feasible if the quantities consumed in the bundle
do not exceed the total endowments of the goods.
In our example, x1 = x2 = (2, 2) is a feasible allocation. Note that
allocations that add up to less than the total endowment are obviously
feasible as well even though it would result in having to destroy some
goods to reach that allocation.

8.6

Walrasian Equilibrium

Walrasian equilibrium, or otherwise called competitive or normal equilibrium, is a pair

(
p, x) where,
p = (
p1 , . . . , pn )
is a price vector and,
x = (
x1 , . . . , xK )
is an allocation such that,
30

 (1) x is feasible.
(2) xJ maximizes uJ (xJ ) such that pxJ pwJ .
So the allocation is feasible in that the total allocation quantities do not exceed the
total endowments. Also the allocation maximizes the consumer utility such that at the
current price level, each consumers bundle is affordable for him.
Back to our example.
The Marshallian demands coming from the cobb-douglas utility functions
are:
3p1 + p2
3p1 + p2
x11 =
and x12 =
.
2p1
2p2
p1 + 3p2
p1 + 3p2
x21 =
and x22 =
.
2p1
2p2
Looking at the individual market conditions (for each good):
Good 1:

3p1 + p2 p1 + 3p2
+
= 4.
2p1
2p1

3p1 + p2 p1 + 3p2
+
= 4.
2p1
2p1
Solving the good 1 equation yields,
Good 2:

p1 = p2 .
Substituting this into the market for good 2 just tells us that the equation
is valid! We cannot actually solve for either prices but we do find
that the prices must be equal. Thus an equilibrium supports any set
of prices such that the two prices are equal.
NOTE: Once we found the equilibrium in market 1, this automatically implied
equilibrium in market 2.
Back in the general setting, consider the vector of demands for consumer J:
xJ (p, pwJ ).
xJ (p, pwJ ) is homogeneous of degree 0 in prices because if you increase prices by t,
you increase both income (via the endowments) and the price of goods the consumer
purchases. The net effect is 0. Thus, if p is an equilibrium price vector, so is t
p. (Thus
we have a multiplicity of equilibrium price vectors.)
Define a vector of excess demands, Z J , as:
Z J (p, pwJ ) = xJ (p, pwJ ) wJ .
Z J is how much more of each good in the bundle that consumer J wants beyond
his endowment. Clearly ZiJ could be negative if a consumer would rather sell his
endowment than have more of good i.
31

 Assuming local non-satiation, the value of excess demand must be 0 because consumers
choose bundles on their budget constraint. Thus,
pZ J (p, pwJ ) = 0.
In the aggregate, define:
Z(p) =

K
X

Z J (p, pwJ ),

J=1

as the aggregate excess demand over all consumers.

Since pZ J (p, pwJ ) = 0 for all J = 1...K consumers, then,
pZ(p) = 0.
Thus Walras Law:
The value of aggregate excess demand must be zero.
Which in other words just means that aggregate expenditure equals aggregate income.
Consider the following implication of Walras Law:
Walras law says:
pZ(P ) = p1 Z1 (p) + p2 Z2 (p) + + pn Zn (p) = 0.
Suppose we have equilibrium in all but the nth market. Thus
pi Zi (p) = 0 i < n.
Thus, the last market must also be in equilibrium. So pn Zn (p) = 0.
NOTE: When we solve for competitive equilibrium prices, we set one price equal to
one (ie p1 = 1) or define a numeraire. All other prices are measured in terms of the
numeraire.

8.7

Pareto Efficiency

Consider K consumers. An allocation, x = (

x1 , . . . , xK ) is pareto efficient if the following is true:
(1) x is feasible.
(2) @ a different feasible allocation, x = (
x1 , . . . , xK ) 3
uJ (
xJ ) uJ (
xJ ) J = 1...K,
and,
uJ (
xJ ) > uJ (
xJ ) for some J.
Thus all consumers cannot be at least as good off and at least one consumer cannot
be strictly better off.
32

Week 9

9.1

The First Theorem of Welfare

Assume L.N.S. Then any competitive equilibrium allocation is pareto efficient.

Proof: Consider a competitive equilibrium, (
p, x). Proceed by contradiction. Suppose
x is not pareto efficient. There there exists a different feasible allocation x, such that,
(1) : uJ (
xJ ) uJ (
xJ ), J = 1...K.
(2) : uJ (
xJ ) > uJ (
xJ ), for at least one J.
By LN S, (1) implies, pxJ pwJ . If not, there would exist another bundle, xJ , that
would make the consumer strictly better off: uJ (
xJ ) > uJ (
xJ ) uJ (
xJ ).
Therefore, pxJ pwJ for J = 1...K. Hence (2) implies that at the equilibrium prices,
the consumer cannot afford xJ (p
xJ > pwJ ). Then summing across consumers,
p

K
X

xJ > p

J=1

K
X

wJ .

J=1

Which contradicts feasibility. QED.

Note that pareto efficiency does not imply desirability.

9.2

Is free trade a good thing?

Suppose, utilities of consumers 1 and 2 with superscripts representing consumer indices

and the subscripts representing commodity indices:
u1 = x11 x12
and
u2 = x21 x22
with endowments:
w1 = (3, 1)
and
w2 = (1, 3).
The equilibrium price would therefore be p1 = p2 = 1 and the equilibrium allocation
would be (2, 2) and (2, 2).
Suppose now the utility of consumer 1 is:
u1 = x11 x12 2x21 .
And for consumer 2:
u2 = x21 x22 2x12 .
33

 Thus both consumers experience negative externalities from the others consumption.
Plugging in the equilibrium values (2, 2),(2, 2) gives us u1 = u2 = 0. However, if trade
is not allowed and we consider the endowment utility level (3, 1), (1, 3), the utilities
are u1 = u2 = 1. Thus the consumers are better off without trade! Thus a reason for
regulation.

9.3

Edgeworth Box Analysis

Refer to graphs in notes [G-9.1]. Under this setting the social planners problem is:
Choose an allocation (x1 , x2 ) to maximize,
u1 (x1 )
subject to:
u2 (x2 ) u.
The first order conditions are therefore tangency of each consumers indifference curves
to each other (and of course to their budget constraints.) Thus,
M RS 1 = M RS 2 .
And,
x1 + x2 = w 1 + w 2 .
The second contraint is just the feasibility constraint.
Note that this condition only holds for interior solutions.
The set of Pareto efficient allocations is called the Contract Curve. [G-9.2]
The Core is the area on the contract curve that makes both consumers better off (hence
it is in DEE LENZ.)
Since there are many pareto efficient equilibrium, which one is optimal? Thus there
is room for regulation, but not in setting consumption levels but rather in setting endowments and letting the market determine where people consume (Hence the Second
Theorem of Welfare).
The FOCs mentioned above for determining pareto efficient equilibria relies on the
fact that preferences are convex.

9.4

Competitive Equilibrium with Production

Consider 2 production processes that produce 2 commodities, y1 and y2 using the same
input, x.
Production process 1 is owned by agent A and is defined as:

y1 = x1 .
34

 Production process 2 is owned by agent B and is defined as:

y2 = x2 .
The two agents, A and B, also have utility functions,
uA = y1A y2A .
uB = y1B y2B .
Consider intial endowments,
wA = (y1 , y2 , x) = (0, 0, 5).
wB = (y1 , y2 , x) = (0, 0, 3).
So the total amount of X available in the economy is 8 and note that the two agents
get no utility from possessing x so they would rather use it in production.
Define P1 as the price of y1 and P2 as the price of y2 . Let r be the price of x. Set a
numeraire,
P2 = 1.
Thus the price ratio is defined as,
P =

P1
P1
=
= P1 .
P2
1

As producers, both A and B maximize profits. Define profits:

A = P1 y1 rx1 = P1 x1 rx1 .

B = P2 y2 rx2 = P2 x2 rx2 .
Thus first order conditions,
A
1
1 1/2
1/2
P1 x 1
r = P x1
r = 0.
x1
2
2
B
1
1 1/2
1/2
P2 x 2
r = x2
r = 0.
x2
2
2
Solving the FOCs for x1 :

1 1/2
Px
r = 0.
2 1
1/2
P x1
= 2r.
2r
1/2
x1
= .
P
P
1/2
x1 = .
2r
P2
x1 = 2 .
4r
35

 Solving the FOCs for x2 :

1 1/2
x
r = 0.
2 2
1/2

x2

= 2r.

1
.
2r
1
x2 = 2 .
4r
1/2

x2

Substituting the input demand into the production function yields the supply of y1
and y2 ,
y1 =
y2 =

P2
P
= .
2
4r
2r

1
1
= .
2
4r
2r

x1 =
x2 =

Therefore substituting into the profit functions,

P
P2
2P 2 P 2
P2
r 2 =

=
.
2r
4r
4r
4r
4r
1
2
1
1
1
r 2 =

= .
B = y2 rx2 =
2r
4r
4r 4r
4r

A = P y1 rx1 = P

At the same time, both consumers maximize their utility. The income of each consumer
is made up of their profits from making yi , and also what they get from selling there
endowment of x. Thus,
P2
+ 5r.
mA =
4r
And,
1
mB =
+ 3r.
4r
Because the utility functions are cobb-douglas, we can write the quantity demands of
yi as each consumers income level divided by two times the prices level. The demand
for commodity y1 for example:
P2
+ 5r
y1A = 4r
.
2P
1
+ 3r
y1B = 4r
.
2P
Similiarly,
36

P2
+ 5r
y2A = 4r
.
2
1
+ 3r
y2B = 4r
.
2
Setting supply of y1 equal to the demand for y1 ,
1
P2
+ 5r
+ 3r
P
4r
+ 4r
= .
2P
2P
2r
"
#
1
P
1 P2
+ 5r +
+ 3r = .
2P 4r
4r
2r
P2
1
+ 10r2 + + 6r2 = 2P 2 .
2
2
2
P
1
+ 16r2 + = 2P 2 .
2
2
2
2
P + 32r + 1 = 4P 2 .
32r2 + 1 = 3P 2 .
We also have the condition that the supply of x must add up to the total endowment
of x. Thus,
P2
1
+ 2 = 8.
2
4r
4r
Thus,
P 2 + 1 = 32r2 .
Substituting this into the equation above,
(P 2 + 1) + 1 = 3P 2 .
P 2 + 2 = 3P 2 .
2 = 2P 2 .
P = 1.
Substituting this into the above equation to solve for r,
32r2 + 1 = 3(1).
32r2 = 2.
1
1
r2 =
r= .
16
4
37

Substituting in again to find the quantity of y1A and y2B ,

P2
1 5
+ 5r
+4
9
y1A = 4r
= 1
= .
2P
2
8
1
1 3
+ 3r
+4
7
y1B = 4r
= 1
= .
2P
2
8
Setting supply of y2 equal to the demand for y2 ,
P2
1
+ 3r
+ 5r
1
4r
+ 4r
= .
2
2
2r
But we dont need to solve this system because we used the numeraire so we can just
plug in P and r to find the quantity of y2A and y2B :
P2
1 5
+ 5r
+
9
y2A = 4r
= 1 4 = .
2
2
8
1
1 3
+ 3r
+
7
y2B = 4r
== 1 4 = .
2
2
8

9.5

Two Production Processes - Marginal Rate of Transformation

Consider two production processes: y1 = f1 (x1 ) and y2 = f2 (x2 ) where x1 + x2 = x.

We can define a production possibilities frontier (see graph in notes [G-9.3]) such that,
y2 = T (y1 ).
Thus, define the Marginal Rate of Transformation (MRT) as the slope of this curve:
dT (y1 )
f0
y2 /x2
= 20 =
.
dy1
f1
y1 /x1
The MRT is the trade off between y1 and y2 given that the input, x, is shared between
the two production processes.

38

10
10.1

Week 10
More on Pareto Efficiency

With reference to the model from last week, we had two agents acting as both producers
and consumers. We get pareto efficiency in the following way. Choose (y1A , y1B , y2A , y2B )
to maximize,
uA (y1A , y2A ),
subject to,
uB (y1B , y2B ) = u,
and,
y2A + y2B = T (y1A + y1B ).
So we maximize the one agents utility subject to the other attaining some fixed level
of utility. The second constraint just provides that we can get the full amount of y2 by
applying technology to y1 .
By varying u, we can find all possible pareto efficient points.
Setting up the Lagrangian of the maximation above, yeilds FOCs. Solving the FOCs
yields,
M RS A = M RS B = M RT.

10.2

Decision Under Uncertainty

Let C be the set of N possible outcomes.

A simply lottery, L, is a probability distribution, (P1 , . . . , PN ) over C, where Pi is the
probability of outcome i.
Example C = (0, 100) and L = ( 21 , 12 ) is like a simple coin toss lottery with equal
probability of winning 0 or 100.
A Compound Lottery is (1 , . . . , K , L1 , . . . , LK ) where J is the probability of lottery
LJ . From this compound lottery, we get a simple lottery, (P1 , . . . , PN ) where,
Pi =

K
X

J PiJ ,

J=1

where PiJ is the probability of outcome i in lottery J.

Example of a Compound Lottery. Let C = (0, 100), L1 = ( 12 , 12 ), and L2 = (0, 1).
Thus the compound lottery is ( 21 , 12 , L1 , L2 ). The outcomes are still 0 and 100. But
now consider the probabilities of each of these outcomes. They are simply equal to

39

the probability of being in each lottery times the probability of attaining the outcome.
Thus,
1
1 1 1
P rob(0) = 1 P11 + 2 P12 = + 0 = .
2 2 2
4
3
1 1 1
P rob(100) = 1 P21 + 2 P22 = + 1 = .
2 2 2
4
1 3
Thus the simple lottery becomes ( 4 , 4 ).

10.3

Preference Properties Involving Lotteries

Let L be the set of simple lotteries and let P represent preferences of L. We have
several assumptions about P .
Assumption 1: P is complete and transitive.
Assumption 2: Continuity. For any 3 simple lotteries, L, L0 , and L00 L, the sets,
)
(
L [0, 1] 3 L + (1 )L0 P L00
and,
(

L [0, 1] 3 L00 P L + (1 )L0

are closed sets. This is just a technical assumption. It says we have no jumps in
preferences.
Assumption 3: Independence Axiom. For an 3 lotteries, L, L0 , and L00 , and [0, 1],
then,
L P L0 if and only if L + (1 )L00 P L0 + (1 )L00 .
Note that L00 appears on both sides, so we really shouldnt take it into account when
determining our preferences between L and L0 .
Note that the independence axiom assumes the independence of the outcomes, but
clearly in this case, you wouldnt want to read about Paris knowing that you werent
ever going to get to go. People have regrets. But I still think this was a poor example.
Now assuming that the assumptions and axioms above hold, then there exists a vector
of utilities (u1 , . . . , uN ) 3 for any 2 lotteries, L = (P1 , . . . , PN ) and L0 = (P10 , . . . , PN0 ),
L P L0 if and only if EU (L) EU (L0 ).
P

L  L if and only if

N
X

Pi u i

i=1

Where EU () is the expected utility of the lottery.

40

N
X
i=1

Pi0 ui .

 Note that if (
u1 , . . . , uN ) represents the same preferences as (u1 , . . . , uN ), then it must
be the case that,
ui = ui + ,
with > 0 and a constant. Note that just having a monotonic transformation of a
utility function is not enough (not strong enough). The transformation must be linear
and increasing to maintain the preference relationships.
Consider the importance of the Independence axiom with the following example. Let
C = (C1 , C2 , C3 ), L = (1, 0, 0), L0 = (0, 1, 0), and L00 = (0, 0, 1). Thus,
L + (1 )L00 has utility: u1 + (1 )u3 .
L0 + (1 )L00 has utility: u2 + (1 )u3 .
Now suppose,
L P L0 .
Therefore by the independence axiom,
L + (1 )L00 P L0 + (1 )L00 .
Thus from above,
u1 + (1 )u3 u2 + (1 )u3 .
Simplifying,
u1 u2 .
Fascinating.

10.4

Risk Preferences

Now consider outcomes as wealth or consumption. A lottery in this case is a

cummulative distribution function, F (x) where F (x) = Prob(wealth x).
The utility of F (x) is:
Z
u(x)dF (x).
If f (x) is the density of F , the utility is then,
Z
u(x)f (x)dx.
Which is the definition of expected utility as each level of utility, u(x), is weighted
by its probability, f (x). In this case u(x) is called a Bernoulli or Von Neumann Morganstern (VNM) utility function.
41

 3 cases.
For any lottery, F (x), then
!

Z
u

xdF (x)

u(x)dF (x)

implies Risk Aversion and Concave Utility Functions.

For any lottery, F (x), then
!

Z
u

xdF (x)

Z
=

u(x)dF (x)

implies Risk Neutrality and Linear Utility Functions.

For any lottery, F (x), then
!

Z
u

xdF (x)

u(x)dF (x)

implies Risk Lovingness and Convex Utility Functions.

10.5

Example of the Demand for Insurance

Let M = initial wealth. L = Possible Loss. P = Probability of Loss. S = Amount of

Insurance Coverage. r = unit premium or price of insurance.
Suppose the consumer buys S units of insurance. Then the lottery is as follows: With
probability (1 P ), the consumer does not experience a loss and has payoff M rS.
With probablity, P , the consumer has the loss and has payoff M L rS + S =
M L + (1 r)S.
The consumer chooses S to maximize,
"
#

"

(1 p)u M rS + pu M L + (1 r)S .

Yields FOC(S):
(1 p)ru0 (M rS) + p(1 r)u0 (M L + (1 r)S) = 0.

42

 Now consider the case of Fair Insurance, or p = r. The first order condition becomes,
(1 p)ru0 (M rS) = p(1 r)u0 (M L + (1 r)S).
u0 (M rS) = u0 (M L + (1 r)S).
M rS = M L + (1 r)S.
rS = L + (1 r)S.
L = S.
Thus Fair Insurance implies Full Insurance.
NOTE!!!! that the SOC for maximization is only satisfied if the consumer if risk averse.
Otherwise it IS NOT! Important!!
Finally, since we have shown that two utility functions representing the same preferences can have different second order derivatives, (as one is a linear combination of
the other), we cannot rely on u00 (x) alone to determine the concavity or convexity of
preferences and therefore the risk level of the consumer.
Define: The Coefficient of Absolute Risk Aversion (CARA) as,
r(x) =

u00 (x)
.
u0 (x)

Example: u(x) = eax . Thus u0 (x) = aeax and u00 (x) = a2 eax . Thus,
r(x) =

a2 eax
u00 (x)
=

= a.
u0 (x)
aeax

Suppose that u1 (x) has a higher CARA than u2 (x). Then u1 (x) = u2 (x) where is
concave. Thus u1 (x) is more concave and represents a higher degree of risk aversion.
So the higher the value of CARA that a consumers utility function has, the more
concave it is and thus the more risk averse the consumer is.
Note that subjectivity plays a role in determining probablities of all these types of
problems.

43