EC202 Lecture 3: Pareto efficiency and

Welfare Theorems in Pure Exchange

Economies.

In this lecture we define the efficiency criterion of Pareto Optimality and

discuss it’s links to Walrasian Equilibrium. In this Lecture we present our
central results of the first and second Theorems of Welfare Economics.
This lecture is split into the following sections:

1. Definition and discussion of Pareto efficiency.

2. Pareto efficiency in 2 ◊ 2 economies.

3. The First Fundamental Theorem of Welfare Economics.

4. The Second Fundamental Theorem of Welfare Economics.

For further reading see chapter 16 of Nechyba or chapters 10, 15 and 16 of

Mas-Colell.

1 Pareto efficiency
1.1 Definition
Pareto efficiency, also known as Pareto optimality is often regarded as a
minimal efficiency criterion. In words, it says that:

1
 You can’t make anyone strictly better off without making some-
body else strictly worse off.

Using our logic skills from Lecture 1, we can rephrase that as

There is no alternative in which everybody is weakly better off

and somebody is strictly better off.

To apply this to our current context, we need to define the set of possible
alternatives:

Definition 1.1. The feasible set of allocations is

I J
ÿ ÿ
x = (xi )iœI œ RI◊J
Ø0 | xi = ei
iœI iœI

In words, it is the set of allocations that satisfy market clearing: that for
every good the amount demanded equals the amount supplied. Elements of
this set are called feasible allocations.

Having defined the feasible set, we can now define Pareto efficiency in our
context:

Definition 1.2. An allocation x = (xi )iœI in the feasible set is Pareto effi-
cient if for every feasible allocation x̂ = (x̂i )iœI

÷i œ I such that x̂i º xi =∆ ÷j œ I such that x̂j ª xj

In words, for any alternative feasible allocation, if somebody is strictly

better off under the alternative, then somebody else must be strictly worse
off. (Because under the original feasible set there was Pareto Optimality)

It should hopefully be clear that this captures the idea of Pareto efficiency
expressed in the initial statement. But there are many ways of expressing
this, giving alternative but equivalent definitions of Pareto efficiency. Here

2
 is another definition which puts the second phrase defining Pareto efficiency
into mathematical notation:

Definition 1.3. An allocation x = (xi )iœI in the feasible set is Pareto effi-
cient if there does not exist another feasible allocation x̂ = (x̂i )iœI satisfying:
1. x̂i ≤ xi ’i œ I
2. ÷i œ I such that x̂i º xi

Since Definitions 1.2 and 1.3 are equivalent to each other and both define
Pareto efficiency you can think in terms of either. Here is another definition:
To drill down further into this, we can express the notion of one allocation
Pareto dominating another:

Pareto Domination :
Definition 1.4. The allocation x̂ = (x̂i )iœI Pareto dominates the allocation
→
There is another x = (xi )iœI if:
allocation in which
at least one person is
1. x̂i ≤ xi ’i œ I
better off and no one 2. ÷i œ I such that x̂i º xi
is worse off

Comparing Definitions 1.3 and 1.4, we see we can define Pareto efficiency
a third way:

Definition 1.5. An allocation x = (xi )iœI is Pareto efficient if it is feasible

and there is no other feasible allocation that Pareto dominates it.

Defining Pareto efficiency in this last way can be quite useful for arguing
whether something is Pareto efficient:
Remark 1.1. To argue that an allocation is not Pareto efficient, it suffices to
find an alternative allocation that Pareto dominates it.

Definition 1.6. We call the set of all Pareto efficient allocations the Pareto
Set.

When there are only two individuals each having preferences represented
by a utility function, by plotting the utilities associated with all feasible
allocations we can represent the Pareto set diagrammatically:

3
 𝑢2 𝑢2

PPF
(Normally every Set of
Set of point on the PPF Feasible
Feasible is Pareto Optimal) Utility
Utility combinations 𝑥
combinations

𝑢1 𝑢1

Pareto Set is the boundary as for every

point on the boundary, there are no Some points on boundary are
feasible points to the northeast not Pareto efficient as Pareto
dominated by x

Normally the set of feasible utility combinations will look like the dia-
gram on the left and so the Pareto Set is the allocations giving utilities on
the northeast boundary. But the diagram on the right shows this doesn’t
necessarily have to be the case.

1.2 Discussion
Pareto efficiency can be seen as a necessary requirement for efficiency but
perhaps not a sufficient one. Consider a scenario with just two alternatives,
lets call them x and y, and two individuals we’ll call Andy and Bob. Suppose
that Andy slightly prefers x to y while Bob has a very strong preference for
y over x. By our definition, both x and y are in our Pareto Set. Put another
way, Pareto efficiency does not put any relevance upon strength of preference.
Although for our uses in this section this is no great problem. In order
to measure strength of preference, we would need cardinal utility functions

4
 that allow for interpersonal comparisons of utility. Whereas, as discussed in
Lecture 1, in representing preferences by utility functions, we assume only
ordinality and hence there are infinitely many utility functions representing
the same preferences.
If we had utility functions that did allow for interpersonal comparisons
of utility, it might make sense to try to maximise the sum of the utilities.
But this makes no sense in our context here: Recall that u (x1 , x2 ) = x1 x2
Ô
represents the same preferences as v (x1 , x2 ) = ≠ 11. But when trying
x1 x2
17
to maximise the sum of agents’ utilities, how much resources you allocate to
an individual with these preferences depends on whether you represent their
preferences by u or v.
A second thing which Pareto efficiency does not consider is the notion of
fairness. For example, in an economy of two individuals, Andy and Bob who
Because
both have strongly monotone preferences, it is Pareto efficient for Andy to get
you can't give the
person with nothing more everything and Bob nothing. Similarly it is Pareto efficient for Andy to get
nothing and Bob everything, and there will be many other Pareto efficient
without making the
person
with
everything worse off .

points besides. While one might look to argue that it would be better to
have an allocation that makes both reasonably happy instead of an extreme
allocation like one person getting everything and the other nothing, Pareto
efficiency has nothing to say on the matter.

2 Pareto efficiency in 2x2 economies

Just like in Walrasian Equilibrium, we can illustrate our Pareto efficient
allocations using an Edgeworth box. To begin with, I talk about the most
analysed case, where our preferences are convex to the origin and Pareto
optima are found in the interior of the Edgeworth box. Although as we later
show, even with convex preferences, this is not guaranteed to be the case.

5
 Solutions are inside the Edgeworth
Box (not in c. edges of box where
UA = 0 or
UB = 0

2.1 Convex preferences with interior Pareto optima

As a note on terminology, I shorten “Pareto efficient allocations” to Pareto
optima.
We start with a nice an illustration of what economists often call “well-
behaved” preferences1 :
𝑥𝐴2 𝑥𝐴2

0𝐵 0𝐵
𝑥𝐵1 𝑥𝐵1
𝑢𝐵 = 𝑐1

𝑢𝐴
𝑢𝐵 = 𝑐2

𝑢𝐵 = 𝑐3 𝑢𝐵
•

𝑢𝐴 = 𝑘 Pareto
Set
𝑥𝐴1 𝑥𝐴1
0𝐴 0𝐴
𝑥𝐵2 𝑥𝐵2

The green line depicts the set of optimal

Optimal bundle for B given A bundles. This is where agent’s indifference
gets utility at least k curves are tangential to each other.
Pareto set connects all of the optimal
bundles

When finding the Pareto Set, one method would be:

For every indifference curve of Andy with segments in the Edge-

worth box, find the highest indifference curve Bob can reach
given that Andy must get at least that utility.
1
When drawing indifference curves, it is impossible to draw indifference curves for every
level of utility. Instead we draw a sample with it being implicitly assumed that between
any two indifference curves there are infinitely many more, all of which are similar in shape.
Note that indifference curves can leave the Edgeworth box, as drawn here, but they can’t
leave the consumption set. That means Andy’s indifference curves should never can cross
the upper and right boundaries of the Edgeworth box but cannot cross the lower or left
boundaries since Andy’s preferences are not well-defined when he has negative amount of
a good. Similar holds for Bob’s.

6
The left hand diagram above illustrates an example of this when uA = k.
That is, we solve the problem:

max uB (xB ) subject to uA Ø k and xA + xB = eA + eB

xA ,xB œR2Ø0

We can see that the highest indifference curve Bob can reach is uB = c3 and
he reaches this by choosing a bundle such that at that allocation Andy’s and
Bob’s indifference curves are tangential to each other. It should be clear from
the diagram that this point is Pareto efficient as claimed, since any allocation
which moves Andy onto a higher indifference curve would move Bob onto a
lower one and vice-versa.
The right hand diagram demonstrates what happens when we solve this
for every level of utility or indifference curve of Andy. We find the Pareto
set, which is the set of points where the indifference curves of the two agents
are tangential to each other, plus the two corners.
1≠—
Example 2.1. Let uA = x–A1 x1≠–
A2 and uB = xB1 xB2 . Let their be 10 units of
—

each good in the economy. Here all Pareto optima except the corners must
be interior since on the lower or left edges of the Edgeworth box uA = 0;
while on the upper and right edge uB = 0. So the diagram above applies.
There are a couple of ways we could find the Pareto Set here.

Method 1

This is to directly use the method above and for each value of k representing
an indifference curve inside the Edgeworth box solve

max uB (xB ) subject to uA Ø k and xA + xB = (10, 10)

xA ,xB œR2Ø0

7
 EXAMPLE 4. I

UA =
XI RAF ,

U'a , =
XXAT Kaz
" ' ×
-

=
d) CAT KA2
KAI KAI

U'az =
It -
a) KAI Kai
'

=lt%n
HA2Y

MR5A =
××#HAD 1k¥
HAD¥94 -41×+1

= ××
( d) KA1
I -

As ↳ = x } , xp ;D
MR5B = B
4- B) KB1

III. BB) \ It -

KB1

As sea , t xp , =
10
& XAZTXBZ =
10

& BH0 -

KAL)
=

( I d) KAI
-

4- B) 110 -

KAI )

④ ANI -

B) 110 XAI )
-
= It a) KAI )
-

1110 -

Kaz)

XXaz =
It -
a)⑦At )( B) 110 -

)
xaz

f- B) 110 -

KAI)

&X =
PXAI
(I -4110 Kaz)-

(t B) (10
-
-

Kai)
 Substituting in the second constraint, this is equivalent to solving

max (10 ≠ xA1 )— (10 ≠ xA2 )1≠— subject to x–A1 x1≠–

A2 Ø k
xA œ[0,10]2

Since this is solved where the constraint is satisfied with equality and
slope of objective function equals slope of constraint, we can solve by using
Lagrangian:
1 2
L = (10 ≠ xA1 )— (10 ≠ xA2 )1≠— + ⁄ x–A1 x1≠–
A2 ≠ k

dL
= ≠— (10 ≠ xA1 )—≠1 (10 ≠ xA2 )1≠— + ⁄–x–≠1 1≠–
A1 xA2
dxA1

dL
= ≠ (1 ≠ —) (10 ≠ xA1 )— (10 ≠ xA2 )≠— + ⁄ (1 ≠ –) x–A1 x≠–
A2
dxA2

dL
= k ≠ x–A1 x1≠–
A2
d⁄
Our solution will be where all three of the above equations equal 0. Thus
-
— (10 ≠ xA1 )—≠1 (10 ≠ xA2 )1≠— = ⁄–x–≠1 1≠–
A1 xA2

(1 ≠ —) (10 ≠ xA1 )— (10 ≠ xA2 )≠— = ⁄ (1 ≠ –) x–A1 x≠–

A2

which implies

— (10 ≠ xA1 )—≠1 (10 ≠ xA2 )1≠— ⁄–x–≠1 1≠–

A1 xA2
=
(1 ≠ —) (10 ≠ xA1 )— (10 ≠ xA2 )≠— ⁄ (1 ≠ –) x–A1 x≠–
A2

which simplifies to

-
—xA1 –xA2
= (*)
(1 ≠ —) (10 ≠ xA1 ) (1 ≠ –) (10 ≠ xA2 )
.

8
 We could then substitute (*) into dL
d⁄
= 0 to find xA1 (k) and xA2 (k).
Although I prefer to leave this in terms of equation (*) from which, for each
xA1 œ (0, 10) we can find some xA2 œ (0, 10) and vice-versa. Two things to
note:

• With – = —, equation (*) gives us xA1 = xA2 , ie the set of allocations

joining 0A and 0B with a straight line.

• For xA1 œ {0, 10}, equation (*) is not well defined. Equation (*) only
defines the interior Pareto optima and we have to add to this the two
corners.

So in all the Pareto Set is:

Ó Ô
(xA , xB ) œ R2◊2
Ø0 | xA + xB = (10, 10) and (*) holds fi{((10, 10) , (0, 0)) , ((0, 0) , (10, 10))}

This defines a curve joining the two corners of the Edgeworth box, where
the curvature depends on – and —.

Method 2

An alternate and slightly quicker way of deriving equation (*) is to calculate

the Marginal rates of substitution and set them equal:

–xA2 —xB2
MRSA
12 = MRS12 ≈∆
B
=
(1 ≠ –) xA1 (1 ≠ —) xB1
Then substituting in xA + xB = (10, 10) to the above gives us equation (*).

2.2 Finding Pareto optima more generally

In general though things are more complicated. The above method only
worked because we could see from drawing the indifference curves on the
Edgeworth box that the Pareto optima had to be at interior points where

9
the indifference curves are tangential to each other. However this need not
be the case:
Remark 2.1. If preferences of one or both agents are not convex then an
allocation at which marginal rates of substitution are equal to one another
is not necessarily Pareto optimal

Exercise 2.1. Demonstrate an Example of Remark 2.1 by drawing some

indifference curves on an Edgeworth box.

Remark 2.2. Even if preferences are convex, we might not be able to use the
methods of Section 2.1 because the Marginal rates of substitution are never
equal to one another.

Example 2.2. Consider both agents with perfect substitutes: uA = –xA1 +

xA2 , uB = —xB1 + xB2 , where –, — > 0. By drawing indifference curves we
derive the Pareto Set. The logic below works regardless of amount of goods
in the economy.
𝑥𝐴2
0𝐵
𝑥𝐵1

𝑥𝐴1
0𝐴
𝑥𝐵2

10
 MRSA
12 = – and MRS12 = — at any allocation. If – = — then marginal
B

rates of substitution coincide everywhere and all allocations are Pareto op-
timal. But if – ”= — then the marginal rates of substitution will never be
equal. However we can still plot indifference curves onto our Edgeworth box
and ask, for each indifference curve of Andy, what is the highest indifference
curve of Bob that we can reach? We find the answer depends on which is
higher between – and —.
If – > — then the Pareto Set is the lower and right boundaries of the Edge-
worth box. In words, starting from the point 0A where Andy gets nothing
and Bob everything, if we want to trace out the Pareto optimal allocations
as Andy’s utility increases: we first give Andy units of good 1; then once he
has all units of good 1, we start giving him units of good 2 until we reach 0B
where Andy has everything.
If – < — then this is the opposite case to above: if we want to gradually
increase Andy’s utility while maintaining efficiency, we first give Andy units
of good 2 before good 1. Hence the Pareto Set is the left and upper boundaries
of the Edgeworth box.

In Example 2.2 preferences were convex but not strictly convex. This
distinction though doesn’t matter, as it is also possible to replicate this result
whit strictly convex preferences. For example imagine adding a very slight
curvature to the indifference curves above. We can also demonstrate this
point algebraically:

Exercise 2.2. Y
Let there be one unit of each good in
Y the economy and prefer-
]ln xA1 + xA2 xA1 > 0
_ ]xB1 + ln xB2 xB2 > 0
_
ences be uA = and uB = .
_ _
[≠Œ xA1 = 0 [≠Œ xB2 = 0
1. Show that everywhere in Edgeworth box MRSA
12 Ø MRS12 with equal-
B

ity only at (xA , xB ) = ((1, 0) , (0, 1)).

2. Use point 1. to conclude that the Pareto Set is the lower and right
boundaries of the Edgeworth box.

11
EXERCISE 2- 2

1) UA =
Inka ,
t
KA2 UB =
XB , t In KB2

Ub
Uh =L
= 1
,
,
"
At U 'D, =L
KB2
✓A2 = I

MR5B KB2
MRSA =L
=

XAI MR5B =
I -

XAZ

I
7 I
% KA2
-

I 2 KAI -
XAIXAZ
07 XA -
XAI KA2 -
/
,

For all OEXA, c 1 and 0 <

Xaz It MR5A >
MR5B
For XA , =/ and xnz=0 MR5A =
MR5B =D
,
As a > p the Pareto lower & Edgeworth Box
optimal solution is in the
right boundaries of the
 To sum up, we can say:

Lemma 2.1. If preferences are convex and differentiable then for any in-
terior allocation of the Edgeworth box, it is Pareto optimal if and only if it
satisfies MRSA
12 = MRS12 .
B

However this only tells us about interior points. There can still be other
Pareto optimal points on the boundary of the Edgeworth box.
In all the examples so far, the two corners have been Pareto optimal.
Although even this is not guaranteed to be the case:

Example 2.3. Consider uA = xA1 + xA2 , uB = min {xB1 , xB2 } and let there
be one unit of good 1 and two units of good 2 in the economy. Then the
allocation (xA , xB ) = ((0, 0) , (1, 2)) where Bob gets everything is not Pareto
optimal since we can redistribute one unit of good 2 to Andy making Andy
better off without making Bob worse off.

To guarantee the corners are Pareto optimal, we would need to impose

a condition that meant consuming all goods in the economy is an agent’s
unique optimal bundle subject to feasibility.

Lemma 2.2. If both agents’ preferences satisfy strong monotonicity then

both corner allocations are Pareto optimal.

Remark 2.3. When finding the Pareto Set before, I said we can proceed
by finding the optimal bundle for agent i œ {A, B}, keeping utility of j œ
{A, B} fixed and repeat for all indifference curves of j. However, note that in
Example 2.3, for k = 0, (xA , xB ) = ((0, 0) , (1, 2)) is a solution to the problem

max uB (xB ) subject to uA Ø k and xA + xB = eA + eB

xA ,xB œR2Ø0

This example demonstrates that solving this problem is not a sufficient

condition for Pareto optimality. However, being a solution for some value of

12
k is a necessary condition since if (xA , xB ) didn’t solve this problem, then by
definition there is another feasible allocation that makes B better off while
not making A worse off. In other words, if we found the set of allocations
that solved this maximisation problem for all feasible values of k then the
Pareto Set is a subset of that. A sufficient condition for Pareto optimality is
that it also solve the reverse maximisation problem: That is, it’s a solution
to maximising Andy’s utility subject to uB Ø k1 for some k1 œ R; and also
maximises Bob’s utility subject to uA Ø k2 for some k2 œ R.
Hopefully I have demonstrated that when finding the Pareto set, you need
to take each example on it’s own merits: Draw an Edgeworth box to get a
sense of what the indifference curves look like and then use sound logic to
reason the Pareto set.

3 The First Fundamental Theorem of Wel-

fare Economics.
In this Section we pull together our Lecture 2 analysis with what we have
just done to discuss the Pareto efficiency of Walrasian Equilibria.
I start again with an example of “well-behaved preferences”. As can be
seen in the diagram below, if we start at an initial endowment (like e in
the diagram below) not on the Pareto Set then there is a lens of points,
formed by agents’ indifference curves through e, that both prefer to that
initial endowment. If we move to another point inside that lens like eÕ , which
is still not Pareto optimal then we can find a new lens of points that Pareto
dominate eÕ . Although at eÕ it’s a smaller lens. As we carry on this process,
moving closer to the Pareto Set, the lens becomes smaller and on the Pareto
Set that lens becomes the empty set.

13
 𝑥𝐴2

0𝐵
𝑥𝐵1

𝒆′

Pareto 𝒆
Set
𝑥𝐴1
0𝐴
𝑥𝐵2

We now define the contract curve2 .

Definition 3.1. The contract curve relative to an initial endowment is the

set of feasible allocations such that:
1. All players are at least as well off as under their initial endowment.
2. No further mutually beneficial trades are possible. II.e. Pareto Optimal)

In other words, it is the set of Pareto efficient points which Pareto domi-
nate the initial allocation. It is called the contract curve because one would
2
Note of caution: Nechyba and some online resources I’ve seen instead define the
contract curve as being the same as the Pareto Set. With opinion split, I have chosen
to follow Mas-Colell which is the most revered Microeconomics textbook. I think it also
makes sense to have Pareto Set and Contract Curve mean different things - and it is
useful to have a word for what I will henceforth call the contract curve and as I will argue
“contract curve” seems an appropriate name for it.

14
expect two agents bargaining with each other to reach an agreement (or con-
tract) on that curve. If they are not at a Pareto efficient allocation, they
can continue negotiating and trading until they reach one. While individual
rationality necessitates that each agent should do at least as well as their
initial allocation.
Remark 3.1. If the initial endowment is a Pareto optimal allocation, then
the contract curve is a singleton, consisting of only that allocation.
As we will see, Walrasian Equilibrium will generally lead us onto the
contract curve.

Example 3.1. Let uA = xA1 xA2 and uB = xB1 xB2 and the initial allocation
be eA = (0.9, 0.1), eB = (0.1, 0.9).
𝑥𝐴2
0𝐵
𝑥𝐵1
Pareto Label the following:
Set 1. Walrasian
Equilibrium allocation.
2. Lens of points
preferred by both
agents.
3. Contract curve.

Initial
endowment
Price line defining
budget constraints
𝑥𝐴1
0𝐴
𝑥𝐵2

The Walrasian Equilibrium is

33 4 3 44
1 1 1 1
(xA , xB ) = , , , and p = (1, 1)
2 2 2 2

15
 The First Welfare Theorem says this is no fluke:

-
Theorem 3.1. The First Fundamental Theorem of Welfare Economics:
If (markets are complete and) everyone’s preferences are locally non-
satiated then any Walrasian Equilibrium is Pareto optimal.

The above Theorem holds regardless of how many Walrasian Equilibria

there are. If there are none then it holds trivially. If there are multiple
Walrasian Equilibrium outcomes, then all of them must be Pareto optimal.
It works as long as preferences are locally non-satiated. We don’t even need
to impose any convexity requirements. You don’t need to worry about the
bracketed term about markets being complete at the moment. This phrase
will make more sense after Lecture 6 on externalities. The proof is given
below. I encourage you to read and try to follow it but this will not be
examined.

Proof. We suppose that we have a Walrasian Equilibrium (p, x) which is not

Pareto optimal and aim to generate a contradiction. That is suppose there
is an alternative feasible allocation x̂ such that

x̂i ≤ xi ’i œ I and ÷j œ I with xˆj º xj

Since agent j is solving his UMP, it must be the case that bundle xˆj
wasn’t affordable to him at prices p. That is

Price Pareto Price Walrasian

p.xˆj > p.ej of of
=

Optimal Bundle Equilibrium bundle

Meanwhile every other agent is solving their UMP at x and x̂i ≤ xi ’i œ I.

Since they all have locally non-satiated preferences, every agent i must spend
their entire budget at bundle xi and any bundle at least as good as xi cannot
be cheaper. So
p.x̂i Ø p.xi = p.ei ’i œ I

16
 Putting this together we get
ÿ ÿ ÿ ÿ
p.x̂i > p.ei ≈∆ p. x̂i > p. ei
iœI iœI iœI iœI

This contradicts feasibility since

ÿ ÿ ÿ ÿ
x̂i = ei =∆ p. x̂i = p. ei
iœI iœI iœI iœI

Remark 3.2. Notice the role local non-satiation plays in this proof. Without
it, we can’t conclude that any bundle at least as good as xi must also exhaust
i’s budget.

Exercise 3.1. Let uA = xA1 xA2 and Bob have preferences uB = ÂxB1 xB2 Ê
where ÂcÊ means the largest integer Æ c. Let there be 10 units of each good in
the economy and show that the First Welfare Theorem fails. (Hint: to make
things easier start an initial allocation where eA1 = eA2 but eB1 eB2 œ
/ Z. Show
there is a Walrasian Equilibrium with allocation equal to initial endowment
but that this is not Pareto optimal. Draw an Edgeworth box to explain this
diagrammatically. Note that Bob has “thick” indifference curves.

Despite, the exercise above, local non-satiation is generally considered a

very weak assumption. As discussed in Lecture 1, it is weaker than mono-
tonicity and strong monotonicity and hence the Theorem would also hold
with either of these two assumptions or any other assumption that implied
local non-satiation. It adds to the strength of this Theorem that it holds
under such mild assumptions3 . As such this result is often quoted by free-
market economists who say we should leave the market to get on with things
without government interference.
3
Although recall that in Lecture 2 we discussed the assumptions behind the whole
Walrasian Equilibrium concept.

17
 Corollary 3.1. (Assuming markets are complete) Since trade can’t make
agents worse off, we can state an even stronger result: that the Walrasian
Equilibrium must lie on the contract curve. In Lecture 10, we define a
stronger concept still, the core, and find that Walrasian Equilibria must lie
within the core.

Revision tip

Set yourself lots of examples with different preferences (mix and match be-
tween those preferences in Problem set 1) and different endowments. Cal-
culate Walrasian Equilibria and the Pareto Set. Verify that the Walrasian
Equilibrium lies on the contract curve.

4 The Second Fundamental Theorem of Wel-

fare Economics.
This gives a partial converse to the First Welfare Theorem: roughly speaking
it says that for any Pareto optimal allocation we might want to achieve, we
can redistribute resources in such a way as to achieve it. This however
requires slightly stronger assumptions than our First Welfare Theorem. I
state the Theorem but omit the proof as it’s beyond the scope of this course.

Theorem 4.1. The Second Fundamental Theorem of Welfare Economics:

(Assuming markets are complete) Let x be a Pareto efficient allocation. If
Given local non -

all agents have continuous, convex and locally non-satiated preferences then
satiation & additional
there exists a reallocation of resources such that for some price schedule p,
assumptions :

Pareto Eff Walrus ian E9

(p, x) is a Walrasian Equilibrium.
.
.

with price vector

where slope ICA =
The easiest way to achieve this is by reallocating resources so that the
slope KB
new allocation is x and argue that we can find some prices p such that every
agent solves their UMP at p.

18
 Example 4.1. Continuation of Example 3.1.
Let uA = xA1 xA2 and uB = xB1 xB2 and the initial allocation be eA =
(0.9, 0.1), eB = (0.1, 0.9). The Walrasian Equilibrium is
33 4 3 44
1 1 1 1
(xA , xB ) = , , , and p = (1, 1)
2 2 2 2

But what if, instead we wanted to implement a different Pareto efficient

allocation? As an Example take
33 4 3 44
2 2 3 3
(xA , xB ) = , , ,
5 5 5 5

𝑥𝐴2 3 1
5 2 0𝐵
𝑥𝐵1
Pareto
Set
Walrasian
Equilibrium
without
Price line
reallocation
after
reallocation
1 A 1
2 2
When we move from
A to B 2 3
we are just
Pareto 5 5 Initial
choosing another B

optimal point we can see

endowment
Walrasian
that there exists 1 CA8 Equilibrium Price line defining
1CB such that ICA = ICB after budget constraints
reallocation without
@ point B implying that
reallocation
Pareto Eff walrasi an Eq 𝑥𝐴1
0𝐴 2 1
.

5 2 𝑥𝐵2

As this diagram shows (you could do algebra to confirm), this is a Wal-

rasian Equilibrium with price vector p = (1, 1) if the initial allocation is
33 4 3 44
2 2 3 3
(eA , eB ) = , , ,
5 5 5 5

19
Starting from this allocation, no trade would take place as both agents are
already solving their UMPs. We could have also redistributed resources to
any other point on the line labelled “Price line after reallocation” and trade
would take us to our desired allocation.
The redistribution could also have taken the form of “money”. Comparing
the two budget lines before and after reallocation, note that, evaluated at
prices p = (1, 1), Andy has 0.2 less income and Bob has 0.2 more income.
So an authority with the power to tax people money could achieve the same
result by taxing Andy 0.2 and giving the money to Bob.

Necessity of convexity

The diagram below illustrates a case where Andy has Leontieff preferences
like uA = min {xA1 , xA2 } while Bob has a quasi-convex utility function (yield-
ing non-convex preferences) such as uB = xB1 + x2B2
𝑥𝐴2
𝑀 0𝐵
𝑥𝐵1
Pareto
Set

𝑢𝐵
𝒙

𝑢𝐴

𝒑′ 𝒑

𝑥𝐴1
0𝐴
𝑁
𝑥𝐵2

Allocation x is Pareto optimal but not achievable as Walrasian Equilib-

20
rium. To illustrate this point suppose we had price vector p which is same
slope as Bob’s indifference curve, then subject to budget constraint, Bob
would be best consuming at point N where he takes only good 2. Alterna-
tively point M where he takes only good 1 is also an improvement on x. If we
change prices, we could fix one of these issues but not both simultaneously.
For example at price pÕ Bob no longer has an incentive to consume only good
1 over x but his optimal demand would be to only consume good 2 and at a
point way outside the Edgeworth box.

Lecture Recap
Here are the skills and knowledge you should have acquired:

• Knowledge of the definition of Pareto optimality. You should be able

to state a formal definition and intuitively understand the concept.

• Be able to depict the Pareto Set on an Edgeworth box.

• Contract curve: definition and drawing on Edgeworth box.

• Knowledge of the two Welfare Theorems, particularly the first one:

– Be able to quote them and understand what they mean.

– Know what conditions are necessary for the conclusion to hold.
– Have some intuitive understnading of why they hold.

21