Contents

3 Allocation and Distribution 1

A Case where The Allocation Branch Needs to Know What the
Distribution Branch is Doing . . . . . . . . . . . . . . . . . . 2
When can the Allocation Branch Ignore Distribution? . . . . . . . 3
The Case of Identical Cobb-Douglas Utilities . . . . . . . . . 4
More General Results . . . . . . . . . . . . . . . . . . . . . . 5
Samuelson Conditions . . . . . . . . . . . . . . . . . . . . . . 5
A Non-calculus Treatment . . . . . . . . . . . . . . . . . . . . 6
Example 3.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Example 3.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Lecture 3

Allocation and Distribution

The undisputed standard graduate public finance textbook when I was in

graduate school in the 1960’s was Richard Musgrave’s The Theory of Pub-
lic Finance [2]. In this book, Musgrave proposes that the main economic
functions of government could be divided among three branches, the Alloca-
tion, the Distribution, and the Stabilization Branches of government. The
job of the Allocation Branch is to “secure adjustments in the allocation of
resources”. The job of the Distribution Branch is to “secure adjustments
in the distribution of income and wealth”, and the job of the Stabilization
Branch is to secure “economic stabilization”.
Musgrave suggests that we think of each branch as run by a “man-
ager” who is instructed to “plan his job on the assumption that the other
two branches will perform their functions properly.” Thus the Allocation
Branch proceeds on the “assumption of full employment of resources and
that the proper distribution of income has been secured.” The distribution
branch assumes that “a full-employment income is available for distribution
and that the satisfaction of public wants is taken care of.” Similarly for
the Stabilization Branch–(but I’m afraid my class never got as far as the
stabilization part of Musgrave’s book.)
Musgrave’s proposed division of labor was and is an attractive one. A
good part of the appeal of this separation is that it approximately coin-
cides with lines of specialization in the academic world. The Stabilization
Branch could be staffed by macroeconomists, the Allocation Branch by mi-
croeconomists and the Distribution Branch by welfare economists, ethical
philosophers, and perhaps a few stray theologians and political scientists.
The macroeconomists and microeconomists would never have to communi-
cate directly and the microeconomists would rarely have to communicate
2 Lecture 3. Allocation and Distribution

with the Distribution mélange.

In this lecture we consider the relation between the Allocation Branch
and the Distribution Branch. In case utility is quasi-linear, this relation is
especially simple. In fact, the Allocation Branch can get its job done while
paying almost no attention to the actions of the Distribution Branch. As you
recall from our discussion in the last chapter, if there is quasilinear utility,
then so long as the Allocation Branch knows that the Distribution Branch
is not going to be so cold-hearted as to leave some consumers with zero
private goods, there is a unique Pareto optimal amount of public goods. All
the Allocation Branch needs to do is to solve for the Pareto optimal quantity
of public goods and provide it.1
But in general, the Allocation Branch will not be able to determine the
right amount of public goods to supply unless it knows what the Distribution
Branch is doing. This makes life more complicated, but does not necessarily
mean that we must abandon Musgrave’s program of divisional separation.
Recall that Musgrave’s suggestion was not that each branch should ignore
the actions of the others, but rather that each branch should assume that
the other branches “will perform their functions properly”. Let’s see how
this goes in an explicit example.

A Case where The Allocation Branch Needs to Know

What the Distribution Branch is Doing
We return to Cecil and Dorothy, but now suppose that they both have
Cobb–Douglas utility functions. In particular:

UC (XC , Y ) = XC Y 2 (3.1)

UD (XD , Y ) = XD Y (3.2)
Suppose that px = py = 1. A little calculation shows the Samuelson
condition to be
XC XD
2 + =1 (3.3)
Y Y
1
At the time when Musgrave’s book was written, public finance economists paid little
attention to the problem of how the Allocation Branch was to find out the utility functions
of consumers who would be willing to tell the truth about their preferences only if it was
in their interest. Perhaps if Musgrave were writing this book today, he would add an
Investigative branch, or to make it sound a little less sinister, an Econometric Survey
Research branch.
WHEN CAN THE ALLOCATION BRANCH IGNORE
DISTRIBUTION? 3

or equivalently:
2XC + XD = Y. (3.4)
This equation together with the family budget equation,

XC + XD + Y = W (3.5)

gives us two equations in the three unknowns, XC , XD , and Y . There is

not enough information in equations 3.4 and 3.5 to solve uniquely for Y ,
without postulating something about the distribution (XC , XD ) of income
between Cecil and Dorothy. Indeed if we use equation 3.5 to eliminate XD
from equation 3.4, we find that the efficiency conditions are satisfied for
any choice of Y and XC such that Y = XC /2 + W/2. This means that the
optimal amount of Y depends on how the private goods are divided between
Cecil and Dorothy. The more generously the Distribution Branch chooses to
treat Cecil relative to Dorothy, the more public good the Allocation Branch
should supply.
If the Allocation Branch knows the rule according to which the Distri-
bution Branch is going to operate, then in typical cases it can solve uniquely
for the right amount of public good. For our example, the rule used by the
Distributive branch adds one more equation to the two equations with which
the Allocation Branch has to work. Suppose, for instance, that the Distri-
bution Branch decides that Cecil and Dorothy should always have equal
incomes. Then in addition to equations A and B, we have

XC = XD . (3.6)

Solving the system of equations 3.4, 3.5, and 3.6, we find that Y = 35 W and
XC = XD = W/5.

When can the Allocation Branch Ignore Distribu-

tion?
We showed that when Cecil and Dorothy have quasi-linear utility functions,
there is a unique amount of public goods that satisfies the Samuelson condi-
tion. This fact generalizes to the case of many consumers, all of whom have
quasilinear utility functions. While quasilinear utility functions are very
easy to work with, the assumption of quasilinear utility in public goods is
not very realistic. If preferences are quasi-linear, then a consumer’s marginal
rate of substitution between public and private goods must be independent
of wealth. If this assumption held in the real world, we would expect to find
4 Lecture 3. Allocation and Distribution

that rich communities would choose the same menu of public goods as poor
communities. We would also expect to see that within a given community,
if the rich are taxed at a higher rate than the poor, then the rich would
always favor less public goods than the poor. As we will see later, both of
these conclusions are strongly refuted by available empirical evidence.
It turns out, however, that there is an interesting class of preferences,
broader than the class of quasilinear preferences, for which the Pareto opti-
mal amount of public goods does not change when income is redistributed
among consumers. Before examining a more general class of utility func-
tions that have this property, we will look at one specific example where
preferences are not quasilinear, but where there is a unique Y that satisfies
the Samuelson conditions.

The Case of Identical Cobb-Douglas Utilities

Suppose that there are n consumers, each of whom has a utility function of
the form:
U (Xi , Y ) = Xiα Y β . (3.7)
where α > 0 and β > 0. Suppose also that this economy begins with a
total endowment of W units of private good and no public goods, but it is
possible to produce public goods at a constant cost of c units of private good
per unit of public good. Then an allocation (X1 , . . . , Xn , Y ) ≥ 0 is feasible
if and only if
X + cY = W (3.8)

where X = ni=1 Xi . The Samuelson necessary condition for a Pareto op-
timal allocation requires that the sum of the marginal rates of substitution
between public and private goods equals the marginal cost of public goods.
Consumer i’s marginal rate of substitution between public and private goods
is
β Xi
.
αY
The sum of the marginal rates of substitution over all consumers is:
n
β Xi βX
= . (3.9)
α i=1 Y αY

Therefore the Samuelson condition can be written:

βX
= c. (3.10)
αY
WHEN CAN THE ALLOCATION BRANCH IGNORE
DISTRIBUTION? 5

We see that in this case, the sum of marginal rates of substitution depends
only on the total amount X of private consumption. Thus any redistribution
of income that leaves total private consumption unchanged will have no effect
on the sum of marginal rates of substitution. Although individual marginal
rates of substitution depend on individual private consumption, we see that
the sum of these marginal rates of substitution is not changed if private
consumption is redistributed while X remains constant.
Solving the simultaneous equations 3.8 and 3.10, we find that the unique
value of Y s that satisfies the Samuelson necessary condition for an efficient
allocation is
 
β W
Ȳ = (3.11)
α+β c
Since the sum of the marginal rates of substitution depends only on the
total amount of private consumption and not on who gets it, any allocation
(X1 , . . . , Xn , Ȳ ) ≥ 0 such that
 α
Xi = W − cȲ = W (3.12)
α+β

will be feasible and satisfy the Samuelson condition.

More General Results

Bergstrom and Cornes [1] found a more general class of utility functions
for which the Pareto efficient amount of public goods is independent of the
distribution of private goods. Suppose that there is a single private good
and k public goods. Let Xi denote the amount of private goods consumed
by individual i and let Y = (Y1 , . . . , Yk ) be the vector of public goods. The
Bergstrom-Cornes family of utility functions take the following form for each
consumer i:

Ui (Xi , Y ) = A(Y )Xi + Bi (Y ). (3.13)

Notice that each individual has the same function A(·), but that the func-
tions Bi (·) can differ from person to person.

Samuelson Conditions
If there is just one public good and utility functions take this form, then the
marginal rate of substitution of each consumer i between the public good
6 Lecture 3. Allocation and Distribution

and the private good is seen to be:

A (Y ) B  (Y )
M RSi (Xi , Y ) = Xi + i . (3.14)
A(Y ) A(Y )
Summing Equation 3.13 over all i and rearranging terms slightly, we find
that the Samuelson condition can be written as

n
A (Y ) Bi (Y )
X+ = c. (3.15)
A(Y ) i=1
A(Y )

where X = ni=1 Xi .
Thus we see that the sum of the marginal rates of substitution depends
only on the aggregate amount of consumption X and on the amount of public
goods Y and does not depend on the distribution of the private goods among
individuals. We can use the feasibility condition X + cY = W to eliminate
the variable X from Equation 3.15. Then we have:
n
A (Y ) Bi (Y )
(W − cY ) + = c. (3.16)
A(Y ) i=1
A(Y )

The only variable in Equation 3.16 is Y . From the previous lecture we

know that (if utility is continuously differentiable) the Samuelson condition
is necessary for an interior Pareto optimum and that if utility functions are
also quasiconcave, then the Samuelson condition together with the feasibility
equation is sufficient for an allocation to be Pareto efficient. Therefore we
can conclude that at any interior Pareto optima, the amount of public goods
must solve Equation 3.16. We also know that if Equation 3.16 is satisfied

for Y = Ȳ , then every allocation (X1 , . . . , Xn , Ȳ ) such that Xi = W − cY
and such that Xi > 0 for all i is Pareto optimal.

A Non-calculus Treatment
We haven’t yet answered the question of when there is a unique value of
Y that satisfies Equation 3.16. Nor have we worked out the story of what
happens at boundary solutions. We could approach the uniqueness ques-
tions with calculus arguments and the boundary solutions with Kuhn-Tucker
methods, but I think it is more instructive to take a different approach. We
can use simpler arguments based on addition, multiplication, some inequal-
ities and some simple geometry of convex sets.
Let us begin by extending our discussion to a more general set of feasible
allocation than we have considered previously. Specifically, let us assume
WHEN CAN THE ALLOCATION BRANCH IGNORE
DISTRIBUTION? 7

that there is some closed bounded subset F of the Euclidean plane such that
the set of feasible allocations consists of all allocations (X1 , . . . , Xn , Y ) such

that ( Xi , Y ) ∈ F.2
Preferences that can be represented by utility functions of the Bergstrom-
Cornes form as in Equation 3.13 all have the same linear coefficient for
private consumption. Therefore, the sum of individual utilities is determined
by the amount Y of public goods and the total amount X of private goods
consumed and does not depend on how the private goods are divided among
individuals. Specifically, we have
n
 n

Ui (Xi , Y ) = A(Y )X + Bi (Y ) (3.17)
i=1 i=1

where X = ni=1 Xi .
Now consider the combination (X̄, Ȳ ) of public goods and aggregate
private good output that maximizes the sum of utilities subject to the feasi-
bility constraint, (X, Y ) ∈ F.3 Why should we be interested in the feasible
outcome that maximizes the sum of utilities? Because any allocation that
maximizes the sum of utilities must be Pareto optimal. (I leave this as an
exercise for you to prove.)
The sum of utilities will be the same at all allocations in which the
amount of public goods is Ȳ and the total amount of the private good is
X̄ although, of course, different allocations of the same total amount of
private goods will lead to different distributions of utility. But since (X̄, Ȳ )
maximizes the sum of utilities over all feasible allocations, it must be that
every allocation in which the amount of public goods is Ȳ and the total
amount of a private goods is X̄ is Pareto optimal. We can state this result
more formally.

Proposition 1 Suppose that preferences of all consumers can be repre-

sented by utility functions of the form Ui (Xi , Y ) = A(Y )Xi + Bi (Y ) and

suppose that (X̄, Ȳ ) maximizes A(Y )X + i Bi (Y ) over the set of all fea-

sible combinations of X and Y , where X = i Xi . Then every allocation
 
(X1 , . . . , Xn , Ȳ ) such that Xi = X̄ is Pareto optimal.
2
In the special case where there is a fixed initial endowment of private goods and
public goods are produced from private goods at constant cost of c per unit, the set F is
{(X, Y )|X + cY = W }.
3
A standard mathematical result (known as the Weierstrass theorem) tells us that if
the feasible set is a non-empty closed bounded set in a finite-dimensional space and if the
function to be maximized is continuous then there is at least one point in the feasible set
that maximizes the function over the feasible set. Therefore so there always is at least one
(X̄, Ȳ ) that solves this constrained maximization problem.
8 Lecture 3. Allocation and Distribution

Proposition 1 tells us that we can find a whole lot of Pareto optima by


choosing X̄ and Ȳ to maximize A(Y )X + Bi (Y ) subject to (X, Y ) ∈ F
and then distributing the total amount X̄ of private goods in any way that
adds up. This theorem does not, however, tell us whether Ȳ is the only
possible amount of public goods in a Pareto optimum, or even if Ȳ is the
only possible amount of public goods at an interior Pareto optimum.

Figure 3.1: Maximizing the Sum of Utilities

Y
 

Ȳ F

X̄ b X

Figure 3, which will look familiar to you from consumer theory will give
you a good idea of how to answer these questions. The crosshatched region
in Figure 3 shows the set of feasible allocations F. We have drawn two level

curves (indifference curves)  and  for the function A(Y )X + Bi (Y ).
Notice that we have drawn the set F as a convex set and we have also drawn
the level curves to be convex toward the origin, in such a way that the set of
points above each level curve is a convex set with no flat edges. A standard
result in consumer theory is that this is appropriate if and only if the function

A(Y )X + Bi (Y ) is a strictly quasi-concave function. The picture shows
the indifference curve  to be the highest indifference curve that touches the
feasible set F. The point of tangency is the point (X̄, Ȳ ). The line shown as

ab is tangent both to the set F and to the set {(X, Y )|A(Y )X + Bi (Y ) ≥

A(Ȳ )X̄ + Bi (Ȳ } at the one and only point (X̄, Ȳ ) belonging to both sets.
Looking at Figure 3, we see that when the feasible set F is closed,

bounded and convex and the function A(Y )X + Bi (Y ) is a strictly quasi-
concave function, that there is exactly one quantity Ȳ of public goods that
corresponds to an outcome which maximizes the sum of utilities. Bergstrom
WHEN CAN THE ALLOCATION BRANCH IGNORE
DISTRIBUTION? 9

and Cornes are able to show that when this is true, in any Pareto optimal
allocation that gives a positive amount of private goods to each consumer,
the amount of public goods must be the unique quantity Ȳ that maximizes
the sum of utilities.

Proposition 2 Suppose that preferences of all consumers can be repre-

sented by strictly quasiconcave utility functions of the form Ui (Xi , Y ) =
A(Y )Xi + Bi (Y ) and that the set F of feasible combinations of aggregate
output and public good supply is closed, convex and bounded. Then there
is a unique quantity of public goods Ȳ such that in every Pareto optimal
allocation in which each consumer has a positive amount of private goods,
the amount of public goods must be Ȳ .

Although any allocation that maximizes the sum of utilities must be

Pareto optimal, (You will be asked to prove this in an exercise), it is not
in general true that every Pareto optimum maximizes the sum of utilities.
Example 3.1, shows why this is the case. Example 3.2 shows why we need
the convexity assumption for Proposition 2.

Example 3.1
Suppose that there are two persons, √ 1 and 2, and that each consumer i has
utility function Ui (X, Y ) = Xi + Y . Public goods can be produced from
private goods at a cost of 1 unit of private goods per unit of public goods
and there are initially 3 units of private goods which can either be used to
produce public goods or can be distributed between persons 1 and 2. Thus
the set of feasible allocations is {(X1 , X2 , Y ) ≥ 0|X1 + X √2 + Y ≤ 3}. The
sum of utilities
√ is U1 (X1 , Y ) + U2 (X2 , Y ) = X1 + X2 + 2 Y which is equal
to X + 2 Y where X = X1 + X√2 . Therefore we would maximize the sum
of utilities by maximizing X + 2 Y subject to X + Y ≤ 3. The solution to
this constrained maximization problem is Y = 1 and X = 2. Any allocation
(X1 , X2 , 1) ≥ 0 such that X1 + X2 = 2 is a Pareto optimum.
In Figure 3, we draw the utility possibility set. We start by finding the
utility distributions that maximize the sum of utilities and in which Y = 1
and X1 + X2 = 2. At the allocation (2, 0, 1), where Person 1 gets all of the
private goods, we have U1 = 2 + 1 = 3 and U2 = 0 + 1 = 1. This is the
point A. If Person 2 gets all of the private goods, then U1 = 0 + 1 = 1 and
U2 = 2 + 1 = 3. This is the point B. Any point on the line AB can be
achieved by supplying 1 unit of public goods and dividing 2 units of private
goods between Persons 1 and 2 in some proportions.
10 Lecture 3. Allocation and Distribution

Figure 3.2: A Utility Possibility Set

U2

D
B
F

A
C

E U1

Now let’s find the Pareto optimal points that do not maximize the sum
of utilities. Consider, for example, the point that maximizes Person 1’s
utility subject to the feasibility constraint X1 + X2 + Y = 3. Since Person
1 has no interest in Person 2’s √ consumption, we will find this point by
maximizing U1 (X1 , Y ) = X1 + Y subject to X1 +Y = 3. This is a standard
consumer theory problem. If you set the marginal rate of substitution equal
to the relative prices, you will find that the solution is Y = 1/4 and X1 =
2 34 . With this allocation, U1 = 3 14 and U2 = 1/2. This is the point C
on Figure 3. Notice that when Person 1 controls all of the resources and
maximizes his own utility, he still leaves some crumbs for Person 2, by
providing public goods though he provides them from purely selfish motives.
The curved line segment CA comprises the utility distributions that result
from allocations (W − Y, 0, Y ) where Y is varied over the interval [1/4, 1].
An exactly symmetric argument will find the segment DB of the utility
possibility frontier that corresponds to allocations in which Person 2 gets no
private goods.
The utility possibility frontier, which is the northeast boundary of the
utility possibility set, is the curve CABD. There are also some boundary
points of the utility possibility set that are not Pareto optimal. The curve
segment CE consists of the distributions of utility corresponding to alloca-
tions (W − Y, 0, Y ) where Y is varied over the interval [0, 1/4]. At these
allocations, Person 1 has all of the private goods and the amount of pub-
lic goods is less than the amount that Person 1 would prefer to supply for
WHEN CAN THE ALLOCATION BRANCH IGNORE
DISTRIBUTION? 11

himself. Symmetrically, there is the curve segment DF in which Person 2

has all of the private goods and the amount of public goods is less than
1/2. Finally, every utility distribution in the interior of the region could be
achieved by means of an allocation in which X1 + X2 + Y < 3.
In this example, we see that at every Pareto optimal allocation in which
each consumer gets a positive amount of private goods the amount of public
goods must be Y = 1, which is the amount that maximizes the sum of
utilities.

Example 3.2
Suppose that as in Example 3.1, there are two persons,
√ 1 and 2, and each
person i has utility function Ui (X, Y ) = Xi + Y . As in the previous
example, public goods can be produced from private goods at a cost of 1
unit of private goods per unit of public goods and there are initially 3 units
of private goods. But in this example, the amount of public goods supplied
must be either Y = 0 or Y = 1/4. In this case the set of possible allocations
is not a convex set.

Figure 3.3: A Non-convex Utility Possibility Set

U2

F
D
B

A E
C

U1

The utility possibility set includes the two lines AB and CD. Points on
the line AB show all of the utility distributions that are possible when Y = 1
and 2 units of private goods are divided between the persons 1 and 2. Points
on the line CD show the utility distributions that are possible when Y = 1/4
and 2 34 units of private goods are divided between the persons 1 and 2. The
12 Lecture 3. Allocation and Distribution

cross-hatched area shows the entire utility possibility set. (Points that are
not on AB or CD are obtained by wasting some of the private goods.) The
utility possibility frontier consists of the three line segments AB, CE, and
DF . The Pareto optimal allocations on the line segments CE and DF are
reached with Y = 1/4 rather than Y = 1, even though Y = 1 maximizes
the sum of utilities. Moreover, except for the endpoints C and D, points
on these lines correspond to allocations in which both persons get a positive
amount of public goods.
EXERCISES 13

Exercises

3.1 Suppose in the example where Cecil and Dorothy have utility functions
XC Y 2 and XD Y respectively, the Distribution Branch has the rule that
XC = 2XD . Solve for the Pareto optimal choice of Y by the Allocation
Branch.

3.2 Where α > 0 and β > 0, show that if all consumers have identical
Cobb-Douglas utility functions Xiα Y β then these same preferences can also
be represented by a utility function of the form A(Y )Xi + Bi (Y ). What are
the functions A(Y ) and Bi (Y )?

Hint: What monotonic transformation of the Cobb-Douglas func-

tions will give a utility function of the Bergstrom-Cornes form?

3.3 Consider an economy with two individuals. Person i has utility function
Y (Xi + ki ) where ki > 0. Public goods can be produced from private goods
at a cost of one unit of private goods per unit of public goods, and there is
an initial allocation of W units of private goods.

a). Find the unique amount of public goods that satisfies the Samuelson
condition.

b). Show that there are some Pareto optima that do not satisfy the Samuel-
son condition and that have a different amount of public goods.

c). Describe the utility possibility set and the utility possibility frontier.
Sketch the way it would look, qualitatively.

d). Suppose that one or both of the ki ’s are negative. Compare the quan-
tity of public goods at Pareto optimal outcomes that do not satisfy
the Samuelson conditions with those at Pareto optimal outcomes that
do. Interpret your result.

3.4 Consider an economy with n individuals where individual i has utility


function Ui (Xi , Y ) = Y α (Xi + βi Y + γi ), where 0 < α < 1, i βi = 0, and
γi > 0 for all i. Assume that public goods can be produced from private
goods at a cost of one unit of private goods per unit of public goods, and
that there is an initial allocation of W units of private goods. Find the
unique quantity of Y that satisfies the Samuelson conditions.
14 Lecture 3. Allocation and Distribution

3.5 Prove the following results which are claimed in the text of the lecture:
a). An allocation that maximizes the sum of individual utilities over all
feasible allocations must be Pareto optimal.

b). Where ai > 0 for all i = 1, . . . , n, any allocation that maximizes the
sum n
ai Ui (Xi , Y )
i=1
of individual utilities over all feasible allocations must be Pareto opti-
mal.

Hint: Consider an allocation that is feasible and Pareto superior

to the allocation that solves your maximization problem. What is
true of the sum or weighted sum of the utilities in this allocation?
Can this allocation be feasible? Why not?

3.6 There are two consumers and one public good. Person 1 always prefers
more of the public good to less. Person 2’s preferences are more subtle.
Their utility functions are given by

U1 (X1 , Y ) = (1 + X1 )Y
1
U2 (X2 , Y ) = X2 Y − Y 2 .
2
The feasible allocations are those such that X1 + X2 + Y = W .

a). Are these utility functions of the Bergstrom-Cornes form?

b). Draw some sample indifference curves for Person 2 between private
and public goods. How would you describe Person 2’s attitude toward
public goods?

c). Find the allocations that maximize the sum of utilities. Take care to
distinguish the case where W is large enough for there to be an interior
solution from the case where it is not.

d). In the case where W = 4, find all of the Pareto optimal allocations and
draw the utility possibility set and show the utility possibility frontier.

e). In the case where W = 1/2, find all of the Pareto optimal allocations
and draw the utility possibility frontier. What distributions of private
consumption are consistent with a Pareto optimal allocation?
EXERCISES 15

f). For what values, if any, of W are there Pareto optimal allocations in
which both consumers consume some private goods and where the sum
of utilities over the set of feasible allocations is not maximized.
3.7 Bergstrom and Cornes prove that under fairly weak assumptions rep-
resentability of preferences in the functional form A(Y )Xi + Bi (y) is both
necessary and sufficient for it to be true that regardless of the level of aggre-
gate income starting from a Pareto optimal allocations in which both con-
sumers have some private goods, if one leaves the amount of public goods
unchanged and redistributes private goods to reach another allocation in
which all consumers have some private goods, the resulting outcome will
also be Pareto optimal. This exercise shows that the “necessity” part of this
proposition depends critically on the qualification regardless of the level of
aggregate income.
Consider an economy with one public good, one private good, and two
consumers. Consumer 1 likes the public good and Consumer 2 hates it. The
amount of public good provided must be either 0 or 1. The public good is
costless to produce. There is one unit of private good which can be divided
between Consumers 1 and 2 in any way such that private goods consumption
adds to 1. Thus the set of feasible allocations is {(x1 , x2 , y)|x1 ≥ 0, x2 ≥
0, x1 + x2 = 1, y ∈ {0, 1}, }. The utility functions of Consumers 1 and 2
respectively are:
U1 (x1 , y) = (x1 + 1)(1 + y) = (1 + y)x1 + y + 1
U2 (x2 , y) = (x2 + 1)(2 − y) = (2 − y)x2 − y + 2.
a). Show that no monotonic transformations of these utility functions will
make it possible to write them both in the Bergstrom-Cornes form
A(Y )Xi + Bi (y).
Hint: If this were possible, the sum of marginal rates of sub-
stitution would not change after a redistribution of private
goods.
b). Draw the utility possibility frontier for this economy.
c). Assuming that lotteries are not possible, show that every possible al-
location is Pareto optimal.
d). Suppose that the functions U1 and U2 are the von Neuman Morgen-
stern representations. Starting from allocations in which both con-
sumers have some private goods, would it be possible to find a lottery
that would make both consumers better off?
16 Lecture 3. Allocation and Distribution

e). Suppose that instead of 1 unit of private goods, there were two units
of private goods to be allocated, so that the set of feasible allocations
is {(x1 , x2 , y)|x1 ≥ 0, x2 ≥ 0, x1 + x2 = 2, y ∈ {0, 1}, }. Draw and
label two separate lines; one showing the utility distributions possible
if y = 0 and one showing the utility distributions possible if y = 1. Do
these lines cross? Identify the utility possibility frontier and show that
the Pareto optimal amount of public goods depends on the distribution
of income.
Bibliography

[1] Theodore Bergstrom and Richard Cornes. Independence of allocative

efficiency from distribution in the theory of public goods. Econometrica,
51(6):1753–1765, November 1983.

[2] Richard Musgrave. The Theory of Public Finance. McGraw Hill, New
York, 1959.