Atsushi Kajii
Most problems are written by Atsushi Kajii, but some of them are taken form variaous textbooks or academic
papers without acknowledgement (in fact he does not remember where they are taken from).
1
2
(c) Let ) be a function on C. Suppose that (a, `) C R : ) (a) _ ` is a convex set. Verify
that ) is a convex function.
(d) Let ) and p be convex functions on C. Show that max ), p (that is, the function a
max ) (a) , p (a)) is a convex function. [Hint. look at the epigraph.]
4. Show that:
(a) p (a
1
, ..., a
1
) =
1
=1
p

(a

) is a concave function if each p

, / = 1, ..., 1 is a concave
function.
(b) if ) (a
1
, ..., a
1
) is quasi concave and homogeneous of degree one, ) is concave.
5. [Homogeneity and convexity] A function ) : R
1
+
R is homogeneous of degree 0 if ) (ta) =
t
0
) (a) for any a and t 0. A set C in R
1
is said to be convex if for any a, j C and t [0, 1] ,
ta + (1 t) j C. A function ) : C R is said to be convex if for any a, j C, and for any
t [0, 1], t) (a) + (1 t) ) (j) _ ) (ta + (1 t) j). A function ) : C R is said to be concave
if ) is convex. A function ) : C R is said to be quasiconvex if a : ) (a) _ c is a convex
set for any c. A function ) : C R is said to be quasiconcave if a : ) (a) _ c is a convex
set for any c.
(a) Show that ) (a
1
, ..., a
1
) = (a
1
)
o
1
(a
2
)
o
2
(a
1
)
o
K
with c

0, / = 1, ..., 1 and c
1
+
c
2
+ c
1
= 1 is homogeneous with degree 1.
(b) Show that if a function ) : R R is nondecreasing, it is quasiconvex and quasiconcave.
(c) Give a simple example of a quasiconcave function which is not concave.
(d) Show that p (a
1
, ..., a
1
) =
1
=1
p

(a
1
, ..., a
1
) is a concave function if each p

, / = 1, ..., 1
is a concave function.
(e) Show that if / : R R is increasing and ) : C R is concave, then /() (a)) is a quasi
concave function of a.
(f) Show that if ) (a
1
, ..., a
1
) is quasi concave and homogeneous of degree one, ) is concave.
(g) Show that ) (a
1
, ..., a
1
) = (a
1
)
o
1
(a
2
)
o
2
(a
1
)
o
K
is concave if c

0, / = 1, ..., 1 and
c
1
+c
2
+ c
1
_ 1.
6. A simpler version of the Gale Nikaido theorem. Let be the 11 dimensional simplex
in R
1
, i.e., :=
_
_
j
1
, ..., j
1
_
_ 0 :
1
l=1
j
l
= 1
_
, and let ) : R
1
be a continuous function
such that j ) (j) _ 0 for any j . (Note that ) is dened even for the case where some
prices are zero.) Let
) (j) := max
l
)
l
(j), and (j) := :
l
= 0 if )
l
(j) <
) (j). So,
(j) if and only if ) (j) =
) (j).
(a) Show that satises the conditions for Kakutanis xed point theorem.
(b) So by Kakutanis theorem, there is j
(j
). Show that ) (j
) _ 0 (i.e., )
l
(j
) _ 0 for
every ) must hold.
(c) Assume in addition that j ) (j) = 0 for any j , and if j
l
= 0 for some , there is 
0
(may be the same as ) such that j
l
0
= 0 and )
l
0
(j) 0. Show that a xed point j
found
above in fact must satisfy ) (j
) = 0.
7. A xed point theorem. Let be the 1 1 dimensional simplex in R
1
, and let ) : R
1
be a continuous function such that
1
l=1
)
l
(j) = 0 for any j . The FanBrouwer theorem
says that ) has a xed point if ) is inward pointing, that is, if j
l
= 0 implies )
l
0
(j) 0 for
some 
0
with j
l
0
= 0 (
0
may be same as ). This question asks you to establish this result from
the GaleNikaido theorem:
Microeconomic Theory problems by A. Kajii 3
(a) Show that if ) is inward pointing, then j
l
= 0 implies )
l
(j) _ 0. [Hint: consider a sequence
j
n
convergent to j, such that elements of j
n
are positive except for ]
Assume on the contrary that ) does not have a xed point from now on.
(b) Dene c(j) :=
_
:
l
= 0 if )
l
(j) < j
l
_
. Show that c is nonempty convex valued
and its graph is closed (i.e., upper hemicontinuous).
(c) So by Kakutanis theorem there is j such that j c( j). Show that this contradicts
the assumption that ) is inward  pointing.
2 Producer Theory
1. For a CobbDouglas production function ) (.
1
, .
2
) = (.
1
)
o
(.
2
)
o
where .
1
and .
2
are inputs,
c +a _ 1, c, a 0,
(a) Find the prot function, supply correspondence, cost function, factor demand function.
(b) Suppose .
2
is xed, but not sunk. That is, the producer can use either .
2
units of factor
2 or choose zero input (and then ) = 0 by construction). Find the average cost curve, the
marginal cost curve, and the supply curve. Draw graphs.
(c) As .
2
changes, how do they change?
(d) What happens if c 1? And if c +a 1? Interpret.
2. Show that:
(a) p (a
1
, ..., a
1
) =
1
=1
p

(a

) is a concave function if each p

, / = 1, ..., 1 is a concave
function.
(b) if ) (a
1
, ..., a
1
) is quasi concave and homogeneous of degree one, ) is concave.
3. For a CES production function with 1 factors of production ) (.) = )
_
.
1
, .
2
, ..., .
1
_
=
_
1
=1
a

_
.

_
o
_ 1
,
where a

0, / = 1, ..., 1, are constants with

a

= 1, and c _ 1, c ,= 0,
(a) Show that this technology exhibits constant returns to scale.
(b) Show that ) is a concave function.
(c) Find the cost function and the conditional factor demand function
4. Let c (n, j) be the cost function and . (n, j) be the conditional factor demand function. Prove:
(a) c (tn, j) = tc (n, j) for any t 0,
(b) 1
u
c (n, j) = . (n, j)
(c) if the corresponding technology exhibits constant returns to scale, c (n, tj) = tc (n, j) for
any t 0.
(d) c (n, j) is concave in n.
(e) c (n, j) is convex in j, if the corresponding production function ) is concave.
5. Consider a general class of linear activity production set 1 =
_
1
=1
t

a

: t

_ 0
_
, where a

,
/ = 1, ..., 1, is a xed technology vector in R
1
.
(a) For an example, let 1 = 2, 1 = 3, and a
1
= (3, 1) , a
2
= (2, 2) , a
3
= (1, 5). Draw
production set 1 . If price vector j is given by j = (1, 1), what is the supply (set)? What
if j = (5, 1) , or j = (2, 1)?
4
(b) Show that production set 1 exhibits CRS.
(c) Show that for a given price vector j: if j a

0 for some /, then the maximum prot is
+. If j a

_ 0 for all /, then the maximum prot is zero, and the supply correspondence
is given by y (j) =
_
1
=1
t

a

: t

_ 0, and t

= 0 if j a

< 0
_
.
6. Consider a competitive rm with a concave production function )
_
.
1
, ..., .
1
_
, where .

is the
amount of good / used as an input. The price of output is one, and the price of good / is n

, so
the total cost of inputs is
1
=1
n

.

. The rm is under a nancial constraint so that there is
a maximum amount it can spend for purchase of factors of production. Let be this maximum
amount. So the total cost of rms inputs must be no larger than . Answer the following
questions.
(a) Write this rms prot maximization problem.
(b) Show that if is large enough, the total cost of prot maximizing combination of inputs
will be less than . (That is, the nancial constraint does not matter.)
(c) Show that the rms prot is non decreasing in ; that is, as the nancial constraint gets
less severe, the prot tends to increase.
3 Consumer Theory
1. For a CobbDouglas utility function n
_
a
1
, a
2
_
=
_
a
1
_
o
_
a
2
_
o
with c + a 0, c, a 0, nd
the demand function, the minimum expenditure function, Hicksian demand function, and the
indirect utility function.
2. Consider utility function n
_
a
1
, a
2
_
= min
_
c
1
a
1
, c
2
a
2
_
where c
1
and c
2
are positive constants.
Draw the indierence curve corresponding to n
_
a
1
, a
2
_
= 1. Find the demand function, the
minimum expenditure function, Hicksian demand function, and the indirect utility function.
3. Consider utility function n
_
a
1
, a
2
_
= c
1
a
1
+c
2
a
2
where c
1
and c
2
are positive constants. Draw
the indierence curve corresponding to n
_
a
1
, a
2
_
= 1. Find the demand function, the minimum
expenditure function, Hicksian demand function, and the indirect utility function.
4. For a quasilinear utility function n
_
a
1
, a
2
_
=
_
a
1
_
+ a
2
, where
0
0,
00
< 0, show that
the demand for good 1 does not depend on income. Can you say anything about the form of
the minimum expenditure function, Hicksian demand function, and the indirect utility function?
[Do not worry about the boundary: do as if negative consumption is allowed.]
5. Let a(j, :) =
_
a

(j, :)
_
1
=1
be a demand function.
(a) Show that
1
=1
j
 J
Jn
a

= 1.
(b)
J
J
l
a
l
(j, :)
_
i
l
l
_
is called the price elasticity of demand for good  (with respect to its
own price). Show that the consumer will spend less on good  as j
l
(marginally) goes up if
and only if the price elasticity is more than one.
(c) If
J
Jn
a
l
(j, :) _ 0, good  is said to be normal at (j, :). A good is called a normal good
if it is normal at any (j, :), otherwise it is called an inferior good. If
J
J
l
a
l
(j, :) 0
occurs, good  is called a Gien good. Show that a Gien good cannot be normal. Show
graphically that an inferior good is not necessarily a Gien good.
6. Consider an additively separable utility function n(a) = n
1
_
a
1
_
+ +n
1
_
a
1
_
.
Microeconomic Theory problems by A. Kajii 5
(a) Show that the preference relation represented by a CobbDouglas utility function in (1)
can be represented by an additively separable utility function.
(b) Show that if each n
l
is concave, so is n.
(c) Show that goods are normal.
7. Prove that the derivatives of the minimum expenditure function give the Hicksian demand.
8. Revealed Preference. Let there be two goods and consider a consumer with income : facing
prices j =
_
j
1
, j
2
_
. Suppose that this consumer exhibits the consumption pattern as in the
following table:
prices j

income :

chosen bundle
(a) (1, 2) 5 (3, 1)
(b) (2, 1) 5 (1, 3)
(c) (1, 1) 4
_
j
1
, j
2
_
(d)
_
1
2
,
1
3
_
3 (4, 3)
(e) (1, 3) 6 (1, 2)
For instance, this consumer chooses to consume (3, 1) when prices are (1, 2) and : = 5. The
chosen bundle j =
_
j
1
, j
2
_
for case (c) is missing in the data.
(a) By drawing a picture, show that choices (a) and (b) above are consistent with utility
maximization.
(b) By drawing a picture, show that choices (a) and (e) above are inconsistent with utility
maximization.
(c) By drawing a picture, nd the area for a that choices (a), (b) and (c) above are consistent
with utility maximization.
(d) Now assume that
_
j
1
, j
2
_
= (2, 2). Dene utility function n on R
2
by the rule n(a) =
min
2fo,b,c,Jg
`

(j

a) +c

, where / denote each case (i.e., j
o
= (1, 2)), each `

is
some positive number, and c

is a constant. Show that by appropriately choosing `

s and
c

s, n(a) is a concave function which justies cases (a) to (d). [Hint. Let `
o
= `
b
= 1 and
`
c
=
5
4
, and c
o
= c
b
= c
c
= 0. Draw the indierence curve for utility level 5, pretending
that / = d is never a minimizer, i.e., as if n(a) = min
2fo,b,cg
`

(j

a) +c

. Play with
`
J
to accommodate (d).]
4 Risk and Uncertainty
1. Consider an agent with concave vNM utility function n. He can spend \ for a gamble (a
1
, a
2
)
 he receives a
1
with probability j and a
2
with probability 1 j. Show that he accepts more
gambles as the ratio r (\) =
u
00
(V)
u
0
(V)
[ArrowPratt measure of absolute risk aversion] gets
smaller. Do the same exercise for relative gambles (i.e., a
1
and a
2
represent the fractions of the
initial wealth and so he receives (a
1
\, a
2
\)) and
u
00
(V)V
u
0
(V)
[ArrowPratt measure of relative
risk aversion.]
2. Prove that the EU representation is unique up to positive ane transformation. [You may
assume that the set of outcomes is [a, a]. ]
3. Let n be a vNM utility function. The certainty equivalent for a random income A is a number
c that is dened by the rule 1 [n(A)] = n(c). Assume that A is a discrete random variable.
(a) Interpret c.
6
(b) Let A be the lottery which yields 4 with probability j and 1 with probability 1 j. Find
the certainty equivalent of A for the following utility functions:
i. n(a) =
_
a
ii. n(a) = aa where a 0 is a constant.
iii. n(a) = a
2
(c) Show that if is a positive ane transformation of n, the certainty equivalent of A for
is the same as that for n. That is, the certainty equivalent depends on the underlying risk
preferences, not a particular choice of vNM function.
(d) Show that c _ 1 [A] if n is concave (thus risk averse).What about the converse?
Suppose that there is another random variable o that may be correlated with A. Dene the
conditional certainty equivalent C (o) by the rule 1 [n(A) n(C (o)) [o] = 0, where 1 [[o]
denotes the conditional expectation given o. That is, C (o) is the certainty equivalent after
observing the realization of signal o. So o can be considered as a piece of information about
A that is revealed before the level of income gets known.
(e) Show that 1 [n(C (o))] = n(c).
(f) Show that 1 [C (o)] _ c.
(g) Is information o valuable?
4. By a lottery A we mean that a random variable whose cumulative distribution function is 1
.
We assume that the lotteries are bounded, so there is numbers c and a, c < a, such that every
outcome of each lottery we consider is included in [c, a]. A lottery A is said to (weakly) rst
order stochastically dominates (FOSD) a lottery 1 if 1
(.) 1
Y
(.) _ 0 for any . [c, a].
A lottery A is said to (weakly) second order stochastically dominates (SOSD) a lottery 1 if
_
:
o
[1
() 1
Y
()] d _ 0 for any . [c, a], and
_
o
o
[1
() 1
Y
()] d = 0.
(a) Let A be the lottery which yields 1 with probability one. Let 1 be a lottery which yields
2 with probability and 0 with probability 1 , where 0 < < 1.
i. Does A FOSD 1 ? Does 1 FOSD A?
ii. Does A SOSD 1 ? Does 1 SOSD A?
(b) Show that:
i. If A FOSD 1 , then 1 (A) _ 1 (1 ), i.e., A yields a higher outcome on average.
ii. If A SOSD 1 , then \ ar (1 ) _ \ ar (A) .
[Hint. You may assume that A and 1 are continuous random variable. Use integration
by parts.]
5. Consider the following situation. Your income today is zero, and that for tomorrow is given by
\, which may be a random variable. If you invest in a risky project, it will yield an additional
random income A. Let r be the (compensation) value of this investment for you, in the sense
that r is a constant that solves 1 [n(A +\)] = 1 [n(r +\)] .
(a) Show that if \ is not random and n is risk averse, we have r _ 1 [A].
(b) Show that if \ and A are independent, we have r _ 1 [A]. [Hint. Use the conditional
Jensens inequality]
(c) Give an example to show that if \ and A are not independent, r 1 [A] is possible.
Microeconomic Theory problems by A. Kajii 7
6. Simple investment problem. Suppose your income today is :
0
and tomorrow is :
1
. But
you can invest part of your income : today, and the gross return from a unit of investment is
given by a random variable 1. You are risk averse and your vNM utility function for income is
denoted by n, and you discount the future utility level by factor a. So, you would solve
max
i
n(:
0
a) +aEn(a1 +:
1
)
where E is the expectation with respect to 1.
(a) Write down the rst order condition for this problem.
(b) Suppose your income today :
0
increases. Does it imply that the demand for the risky
asset a increases? If not, give a sucient condition for this to be true.
(c) Suppose the return 1 gets more risky in the sense of the second order stochastic dominance.
Does it mean that the demand for the risky asset decrease? If not, give a sucient condition
for this to be true. (Hint: consider the property of the function an
0
(a)
7. Prudence and precautionary saving. Suppose your income is 1 tomorrow. 1 may be
random whereas you cannot insure against the income risk. But you can save part of your
income : today. Your vNM utility function for todays income is denoted by n and that for
tomorrow is denoted by . Assume that both are concave. So, you would solve
max
i
n(:a) +E (a +1 )
where E is the expectation with respect to 1 .
(a) Write down the rst order condition for this problem.
(b) Let 1 = j
0
for sure. Show that if your future income j
0
decreases, then your saving
increases.
(c) Suppose that 1
0
and 1
1
are random variables of the same mean and 1
1
is riskier than
1
0
. (i.e., 1
0
second order stochastically dominates 1
1
) Show that you save more under the
income risk 1
1
than you do under 1
0
if
000
0 (i.e.,
0
is convex). Interpret.
(d) The coecient of absolute prudence of is dened as the coecient of absolute risk aversion
for
0
. Assume that n = . Show that if the coecient of absolute prudence for increases,
you save more.
8. Variance and risk. Let A, 1 , 7 be random variables with 1 = A +7. We say 1 is as risky
as A if 1 [7[A] = 0.
(a) Show that if 1 is as risky as A, 1 [7] = 0 and Co [A, 7] = 0, and \ ar [1 ] _ \ ar [A].
(b) Show by an example that 1 [7] = 0 and Co [A, 7] = 0 do not imply that 1 is as risky as
A.
9. Simple insurance problem. Suppose your income is j
1
when the state of nature is H and
j
1
when the state of nature is 1. You believe that state H will occur with probability
1
and
the state 1 will occur with probability
1
(= 1
1
). There is an insurance which pays one
dollar per unit if state is ,  = H, 1. You may buy or sell these insurance contracts, and the
price of insurance that pays out in state  is given by j
~
. Writing a
~
for the amount of insurance
which pays out in state  (so negative a
~
means you sell), you are interested in maximizing
expected utility by choosing appropriate amounts of a
1
and a
1
, i.e.,
1
n(j
1
+a
1
) +
1
n(j
1
+a
1
)
subject to j
1
a
1
+j
1
a
1
= 0, where n is a concave vNM utility function.
8
(a) Setting c
~
= j
~
+ a
~
for  = H, 1, and write some indierence curves exhibiting your
preferences among (c
1
, c
1
) pairs.
(b) What is the marginal rate of substitution when c
1
= c
1
?
(c) Show that if
H
L
=
t
H
t
L
, you will choose a
1
and a
1
in such a way j
1
+a
1
= j
1
+a
1
.
(d) If
H
L
t
H
t
L
, in which state will you consume more?
10. Risk and demand. Suppose there are o equally likely states, and write a R
S
for a contingent
consumption plan: that is, a
~
,  = 1, ..., o, is consumption in state .
(a) Assuming o = 2, write the set of contingent consumption plans with average consumption
equal to c in a graph measuring a
1
horizontally and a
2
vertically.
(b) Assuming o = 2, and x a R
2
with a
1
a
2
0. Write the graph of cumulative
distribution function induced by a. (Hint: Pr(. _ c) =
1
2
if a
1
c a
2
.)
(c) Assuming o = 2, and x a R
2
with a
1
a
2
0. Write the set of contingent consumption
plans which rst order stochastically dominates a. Also write the set of contingent plans
which second order stochastically dominates a.
(d) Imagine an investor with an increasing and concave vNM utility n, n
0
0 and n
00
< 0, with
positive income n. There are o states, all equally likely. Denote the price of consumption
good in state  by j
~
0, and denote by j the vector of prices, i.e., j = ( , j
~
, ) .
Thus the investor wants to choose a contingent plan a =
_
a
1
, .., a
S
_
, where a
~
is the
amount of consumption in state , which maximizes the expected utility
S
~=1
~
n(a
~
)
given the budget
S
~=1
j
~
a
~
_ n, where
~
=
1
S
for  = 1, ..., o. Denote by a(j, n) :=
_
a
1
(j, n) , ..., a
S
(j, n)
_
the contingent plan which maximizes the expected utility.
i. Write the rst order condition of the maximization problem above.
ii. Show that when n increases, the investor will increase demand a
~
for all ; that is,
a
~
(j, n) is increasing in n for all .
iii. Show that if n
0
n, the random consumption resulting from a(j, n
0
) rst order
stochastically dominates that from a(j, n).
iv. Show that if j
~
= j
~
0
, then a
~
(j, n) = a
~
0
(j, n) . (i.e., if the price in state  and that
in state 
0
are the same, the investor consumes the same amount in these states.)
v. Show that the price and the consumption must be inversely related; that is, j
~
j
~
0
if and only if a
~
(j, n) < a
~
0
(j, n).
vi. [Harder] Show that if the price j and the consumption a are inversely related, then
there is a vNM utility function n such that a is utility maximizing.
11. The common ratio paradox. Denote by the lottery that yields $3000 for sure, by 1 the
lottery that yields $4000 with probability 0.8, and $0 otherwise. denote by C the lottery that
yields $3000 with probability 0.25, and $0 otherwise, and by 1 the lottery that yields $4000
with probability 0.2, and $0 otherwise. Many studies have shown a systematic tendency for
subjects to express a preference for over 1 and for 1 over C. Show that this choice pattern
violates the expected utility hypothesis.
12. The Ellsberg paradox. There are two urns identical balls. Urn 1 contains 49 white and 51
red balls. Urn 2 has 100 balls, but with unknown proportion of white and red balls. Consider
the following two bets: Bet 1: if red, $1000, otherwise 0; Bet 2: if white $1000, otherwise 0.
Write these problems in the Savage framework, and show that if an agent prefer urn 1 for both
bets, his preferences violate the surething principle.
Microeconomic Theory problems by A. Kajii 9
5 Classical Partial Equilibrium Analysis
1. Consider a quasilinear utility function n
_
a
1
, a
2
_
=
_
a
1
_
+ a
2
, where
0
0,
00
< 0. Suppose
the price of good one changes from j to j. Verify that the change in consumer surplus measures
the exact change of utility level.
2. Consider a quasilinear utility function n
_
a
1
, a
2
_
=
_
a
1
_
+ a
2
, where
0
0,
00
< 0. Suppose
the price of good one changes from j to j. Verify that the change in consumer surplus measures
the exact change of utility level. Also show that the equivalent variation and compensated
variation coincide, and relate these to the change in consumer surplus.
3. Aggregation of producers. Consider J producers, producing a homogeneous good from an
homogenous input. Price of output is j and that of input is 1. Producer using a technology
represented by a cost function c
(0) = 0 and c
0
0 and c
00
0,
and hence the supply is characterized by c
0
) = j. Write 
(j) =
_
c
0
_
1
(j), i.e., 
I
(j) is the
quantity supplied by the rm at price j. By the assumptions, each 
is increasing function,
and so is the aggregate supply o (j) :=
=1

=1
c
(
(1 ())).
Notice that this function has the following interpretation. For the target output level , calculate
the price level 1 (), and ask each rm to produce taking 1 () as given. Then C () is the
total cost incurred by the rms.
Show that the representative rm with cost function C as given above will supply o (j) when
price is j (thus, o constitutes the supply curve for this ctitious rm.)
4. Producer surplus. Consider a rm with cost function c () = ) () + (), where (1) (0) = 0
and
0
0 and
00
< 0; (2) ) () = 1 0 if 0, and ) (0) = 0. So the constant 1 represents
a xed (but not sunk) cost of production.
(a) Find the quantity this rm will supply when the price of output is j 0.
(b) We shall write (j) for this quantity supplied found above, and j () for its inverse. Write
the graph of j () taking horizontally.
(c) Suppose the rm produces a positive amount at price j 0. The area to the left of the
marginal cost curve is called producer surplus: That is, the producer surplus is j ( j)
_
j( )
0
0
() d. Is the producer surplus the same as prots?
5. Ineciency of taxation. Let
1
(j) be a downward sloping demand curve and
S
(j) be
upward sloping supply curve, and denote by j
), and let
(t) be the quantity traded. Assume these functions are dierentiable, and
0
(0) < 0.
(a) Show graphically that taxation leads to a loss in social welfare (consumer surplus + pro
ducer surplus + tax revenue).
(b) Write the amount of social welfare explicitly, and dierentiate it with respect to t. Show
that the derivative is zero at t = 0.
(c) Dierentiate the social welfare twice in t. Can you sign the derivative at t = 0?
(d) Interpret your results above.
6. Monopoly. Consider a monopoly rm with a convex cost function c (). Given decreasing
inverse demand function j (), write
.
(b) Show that the price elasticity of demand,
0
j
, must be more than one at
.
(c) Show that the revenue function j () is concave if j () = a /, where a 0, / 0 are
positive constants (that is, the concavity assumption in 6a is satised for linear demand
models). Show that when a increases, both quantity
2
. The large rm can produce costlessly (i.e., zero marginal
cost and no xed cost). The demand for the product is given by j = 1 Q, j is the price of the
product and Q is the total demand. So if Q
1
is the amount the large rm produces, and Q
S
is
the sum of quantities the small rms produce, then the market price will be j = 1(Q
1
+Q
S
).
(a) Find the total quantity produced by the small rms when the price is j.
(b) Write the problem that the large rm solves, and nd the quantity which the large rm
will produce.
(c) If the social welfare is to be maximized, how much should the large rm produce? What
about the small rms?
(d) As the number of small rms increases, does the equilibrium approach a socially desirable
state? Discuss.
8. Third degree price discrimination. Consider a monopoly rm with a cost function with
constant marginal cost c. There are two types of consumers, and the demand of type t, t = 1, 2,
consumers is given by a downward sloping demand curve d

(j).
(a) Suppose the rm cannot price discriminate; that is, the rm must charge the same price j
for both types. Characterize the monopoly price.
(b) Suppose that the rm can price discriminate, thus the rm charges j

for type t consumers.
Write down the rms problem and the corresponding rst order condition.
(c) Show that the ability of price discrimination is advantageous to the rm (i.e., the rm will
get at least as much prots as in the case of nondiscrimination).
(d) Show that if the rm sets j
1
j
2
, the demand of type 1 at j
1
is more elastic than that of
type 2 at j
2
.
9. Ramsey taxation. Consider a central authority who needs to raise tax revenue of ` by
specic (per unit) taxes on 1 goods. There are 1 + 1 goods, and the representative consumers
utility function is linear in the 1+1 st good. (This 1+1st good is not taxed). More specically,
assume that it is given by
1
l=1
l
_
a
l
_
+ a
1+1
. Write the demand for good  as a
l
_
j
l
_
. Given
price j
l
, the demand for good  when tax t
l
is levied will be a
l
_
j
l
+t
l
_
, and so the tax revenue
from good  will be a
l
_
j
l
+t
l
_
t
l
. Currently, the prices of goods are given by j =
_
j
1
, ..., j
1
_
and the price of good 1 + 1 is one.
Microeconomic Theory problems by A. Kajii 11
(a) Write the problem to minimize the loss of consumer surplus, given that the tax revenue
must be at least as much as `.
(b) Show that price elastic goods tend to get taxed less. Interpret.
10. Policy mix. Consider a monopoly rm with a positive constant marginal cost c. A decreasing
inverse demand function j () is given. Assume that the revenue function j () is concave in
. Write (c) for the amount that the rm produces. Also denote by
(c) (c) .
(b) Compare the following two policies. (1) quantity control polity, which requires the rm
to produce an amount ; (2) tax/subsidy policy, which requires the rm to pay t to the
governement per unit produced. Show that both policies can induce the socially optimal
amount. What is the sign of t? Discuss.
From now on assume a linear inverse demand function j () = 1 c, where c 0.
(c) Suppose now that the government does not know c although the rm does: the government
knows c is either c
1
or c
1
, where 1 c
1
c
1
1, and these are equally likely.
i. Find the quantity control policy which maximize the social welfare. Explain why the
social optimal amount is not necessarily induced?
ii. Also nd the tax/subsidy policy which maximize the social welfare, and explain the
source of ineciency.
iii. Consider the following policy mix: there is a xed target production level ^ , and the
rm must produce at least this amount. But the rm may produce more, and in such
a case the rm pays t per unit of the amount for the excess production: i.e., the rm
pays t ( ^ ) to the government if ( ^ ) 0. Show that this policy mix is better
than the two policies above.
6 Competitive Markets: General Equilibrium
1. Characterization of Pareto eciency. Consider a pure exchange economy where 1 goods
are traded. The total endowments are c R
1
++
, and assume each n
I
is strictly increasing and
continuous.
(a) Give the denition of a Pareto ecient allocation of this economy.
Now consider the following maximization problems: 1) Given `
I
0, i = 1, ..., 1,
`aai:i.c
i
1
2R
L
++
,...,i
I
2R
L
++
I
`
I
n
I
(a
I
) subject to
I
a
I
= c; (1)
2) Given n
I
0, i = 2, ..., 1,
`aai:i.c
i
1
2R
L
++
,...,i
I
2R
L
++
n
1
(a
1
) subject to n
I
(a
I
) _ n
I
, i = 2, ..., 1 and
I
a
I
= c (2)
(b) Show that (a
I
)
J
I=1
_
R
1
++
_
J
is Pareto ecient if and only if (a
I
)
J
I=1
solves problem 2 above
with some n. Is this true if n
I
were not strictly increasing?
12
(c) Suppose every n
I
is concave, dierentiable, and 1n
I
(a
I
) 0, n
I
(a
I
) 0 for any a
I
0.
Show that (a
I
)
J
I=1
_
R
1
++
_
J
solves problem 1 for some ` if and only if (a
I
)
J
I=1
solves problem
2 with some n. (Hint. Both problems are then concave so compare the KuhnTucker
condition.)
2. Pateto Ecient share of random income. Consider an economy with 1 agents. Each
agents preferences are represented by a vNM utility function n
I
. The aggregate income of this
economy is denoted by j. For simplicity, suppose that there are only nitely many possible
income level, and denote by j (j) the probability that the aggregate income is j. Denote by
a
I
(j) the amount agent i receives when aggregate income is j, and we call (a
I
)
J
I=1
an income
sharing rule if it is feasible allocation of income, i.e.,
J
I=1
a
I
(j) = j for any j.
(a) Give the denition of Pareto ecient income sharing rule.
(b) Assuming each n
I
satises n
0
I
0 and n
00
I
< 0, write the Pareto maximization of weighted
sum of utility functions and the rst order condition for Pareto eciency.
(c) Fix a positive `
I
, for i = 1, ..., 1, and let p (t) =
J
I=1
(n
0
I
)
1
_

X
i
_
, where (n
0
I
)
1
is the
inverse of n
0
I
. Show that if we set a
I
(j) = (n
0
I
)
1
_
1
()
X
i
_
, then (a
I
)
J
I=1
is a Pareto ecient
sharing rule.
(d) Assume further n
I
(.) =
1
o
i
exp[c
I
.], where c
I
is a positive constant, for every agent i.
Find the (class of) Pareto ecient income sharing rule.
(e) Do the same exercise for n
I
(.) =
1
1
i
.
1
i
where where
I
is a positive constant, for every
agent i.
3. State and prove the rst fundamental theorem of welfare economics. Discuss why completeness
of markets is important in this result.
4. weak eciency to eciency. Say that the budget is tight at prices j and income n for n
I
if j a
0
< n and a
0
A
I
implies that a
0
is not a utility maximizing demand vector (i.e., to
maximize utility, all the given income must be spent).
(a) Show that the budget is tight at prices j and income n for n
I
if and only if any demand
vector is cost minimizing in the following sense: if a maximizes utility n
I
in the budget set,
then for any a
0
A
I
, n
I
(a
0
) _ n
I
(a) implies j a
0
_ j a.
(b) Argue graphically that the budget might not be tight when n
I
is not strictly increasing
(i.e., a thick indierence curve is possible)
(c) Let (a
, j
, j
=1
,
_
.
I
, (0
I
)
=1
_
J
I=1
). Show that (a
, j
I
(given
prices j
and income j
.
I
+
0
I
_
j
_
) for n
I
for all consumer i.
5. Second fundamental theorem. Let j R
1
be a price system, j ,= 0. We say that a
is a
quasiutility maximizer if n
I
(a
0
) n
I
(a
) implies j a
0
_ j a
)
but j a
0
= j a
. Consider (1 t) a
0
+t^ a with small t 0].
(d) Comment on the second fundamental theorem of welfare economics discussed in the class.
6. the quasilinear economy. There are two goods, a and :, and there are : traders. There
are 1 traders and each trader i has a quasilinear utility function n
I
(a
I
) +:
I
and endowed with
( a
I
, :
I
), where n
I
is strictly increasing and concave. The price of good a in units of good j is
denoted by j.
(a) Show that a trader may be interpreted as a producer with a convex cost function. [Hint:
think of n
I
as the negative of cost function, and check that utility maximization given
budget is equivalent to prot maximization.]
(b) Show that (a
I
, :
I
)
J
I=1
is Pareto ecient if and only if (a) (a
I
)
J
I=1
maximizes
J
I=1
n
I
(a
I
)
subject to
I
a
I
=
I
a
I
, and (b)
I
:
I
=
I
:
I
hold.
(c) Show that the (general) equilibrium price j
depends on a =
I
a
I
only (so in particular,
it is independent of the way the total resource is initially allocated).
(d) Show that the (general) equilibrium price j
, j
, j
=1
,
_
.
I
, (0
I
)
=1
_
J
I=1
).
Show that (a
, j
I
(given prices j
and income
j
.
I
+
0
I
_
j
_
) for n
I
for all consumer i.
18. Uniquness of notrade equilibrium. Consider an exchange economy with 1 consumers and
1 goods. Suppose that the initial endowments . = (.
1
, ..., .
J
) 0 constitute a competitive
equilibrium; that is, there is no trade in this equilibrium. Show that . is in fact a unique
equilibrium allocation if utility functions are strictly quasiconcave.
18
19. The Negishi Method. Consider an exchange economy with 1 consumers and 1 goods with dif
ferentiable utility functions. Assume that for each `
J1
:=
_
(`
1
, ..., `
J
) _ 0 :
J
I=1
`
I
= 1
_
,
there is a unique Pareto Ecient allocation a
(`) = ( , a
I
(`) , ) which is obtained by max
imizing the weighted utility sum
I
`
I
n
I
(a
I
) subject to
I
(a
I
.
I
) = 0, and the corresponding
supporting price vector j
I
(`)), j
(`)a
I
j
(`)a
I
(`)). Assume
further that a
(`) and j
(`) (a
I
(`) .
I
),
and c(`) := (c
I
(`))
J
I=1
.
(a) Show that
I
c
I
(`) = 0 for any `.
(b) Show that if c
_
`
_
= 0 (i.e., c
I
_
`
_
= 0 for all i),
_
a
`
_
, j
`
__
constitutes a competitive
equilibrium.
(c) Assume that 1 = 2. Argue using a graph, that under the standard assumptions, there will
be
` such that c
_
`
_
= 0.
20. The full insurance theorem. Consider an exchange economy where each household / receives
a random endowment vector c
~

R
1
if state  occurs. Household /, / = 1, .., H, has utility
function
S
~=1
~
n

(a
~

) where n

is strictly concave, and a
~

is the consumption vector in state
. Note that probability weights
~
,  = 1, .., o, do not depend on /. Assume that there is
no aggregate risk; that is, there is a vector c such that c =
1
=1
c

() for all . Show that
consumption does not depend on states in any Pareto ecient allocation.
21. Insurance Markets. Consider an exchange economy where each household / receives a random
endowment vector c
~

R
1
if state  occurs. Household /, / = 1, .., H, has utility function
S
~=1
~
n

(a
~

) where n

is strictly concave, and a
~

is the consumption vector in state . Note
that probability weights
~
,  = 1, .., o, do not depend on /. Assume that there is no aggregate
risk; that is, there is a vector c such that c =
1
=1
c
~

for all .
(a) Show that consumption does not depend on states in any Pareto ecient allocation.
(b) From now on, assume that o = H, and consider the sequential trade model with H assets,
where the asset / pays $1 if state  = / occurs. (i.e., these are the Arrow securities).
Assume moreover that c
~

= c if  = /, c
~

= a otherwise, where a c 0. Notice
that asset / can be interpreted as insurance for household /s income risk. Write .
~

for
the amount of asset  household / holds at the beginning of period 1, and there is no
consumption in period 0 (so only assets are traded).
i. Show that household / must buy the insurance for / (i.e., .


0) in any equilibrium.
ii. Does any household other than / buy insurance / in some equilibrium?
(c) Let H = o = 2, n
1
(.) = n
2
(.) = ln., c
1
= (1, 0), c
2
= (0, 1). Write
_
1
,
2
_
= (, 1 ).
Consider the sequential trade model, and nd the equilibrium price of the Arrow security
which pays $1 in state , as a function of , with normalization that the sum of Arrow
security prices is one. Also nd the price of a discount bond which pays $1 in every state.
22. Sequential trade and asset structure. Consider the two period sequential trade model of
an exchange economy, with one good in each state  = 1, ..., o. Let be a linear subspace of R
S
.
Consider the following optimization problem:
(*) max
i
n
I
(a) subject to
S
~=0
j
~
(a
~
I
.
~
I
) = 0 and
_
a
1
I
.
1
I
, ..., a
S
I
.
S
I
_
.
We say that (j, a) constitute an equilibrium if every consumer i solves the problem above at
a
I
and
J
I=1
(a
~
I
.
~
I
) for  = 0, 1, ..., o. So an Arrow  Debreu equilibrium is an Aequilibrium
where = R
S
.
Microeconomic Theory problems by A. Kajii 19
(a) Consider J assets,
_
r
1
, ..., r
which span , in the economy with these assets there is a Radner equilibrium with
equilibrium consumption idential to the equilibrium consumption.
(c) Show that if two collections of assets
_
r
1
, ..., r
and
_
^ r
1
, ..., ^ r
^
_
generate the same lin
ear subspace, then the sets of rational expectations equilibrium consumption allocations
coincide.
23. Comonotonicity of ecient risk sharing. Consider an exchange economy with a single
good and o states. Each consumer i receives a c
~
I
units of the good if state  occurs. Consumer
i, i = 1, .., 1, has utility function
S
~=1
~
n
I
(a
~
I
) where n
I
is strictly concave, and a
~
I
is the
consumption in state . Note that probability weights
~
,  = 1, .., o, do not depend on i.
Denote by c
~
:=
J
I=1
c
~
I
the total resource available in state .
(a) Write the rst order condition for Pareto eciency.
(b) Show that if a is a Pareto ecient allocation, a
~
I
_ a
~
0
I
implies a
~
_ a
~
0
(a
.) n
(a
).
(b) Show that if a feasible consumption allocation (a
1
, ..., a
J
) is Pareto ecient, it allows no
gains from bilateral trade.
(c) Let (a
1
, ..., a
J
) be a feasible consumption allocation such that a
I
0 for each i. Show
that if (a
1
, ..., a
J
) allows no gains from bilateral trade, (a
1
, ..., a
J
) is Pareto ecient. [Hint:
express the rst order condition for no gains from bilateral trade by a constrained maxi
mization problem.]
25. An Equilibrium model for interest rate. Consider a two period economy with one per
ishable good in each period, 0 and 1. There is a representative consumer with a concave,
dierentiable vNM utility function n and his utility is additively separable with discount factor
20
c (0, 1). The consumer is endowed with c
0
units of the good in period 0, and his endowment
of the good in period 1 is random, and it is represented by a random variable 1 . There is a
market for a riskless discount bond, which is a security which promises to pay one unit of good
in the second period for sure. The net supply of the bond is zero. Denote by a the amount of
the discount bond the consumer chooses to own. The price of the bond is . To sum up, the
representative consumer solves
max
i
n
_
c
0
a
_
+cEn(a +1 )
where E is the expectation with respect to 1 .
(a) Write down the rst order condition for the consumers problem.
(b) Derive the equilibrium bond price and the interest rate of this economy.
(c) When the consumer becomes more patient, i.e., the discount factor increases, what happens
to the equilibrium bond price?
(d) When the second period endowment gets riskier, that is, the random variable 1 changes
to 1
0
and 1
0
is a riskier random variable than 1 , what happens to the equilibrium bond
price?
26. Equilibrium asset pricing model. Consider an economy with a single (representative) con
sumer with concave vNM utility index n and discount factor a. There is one good in each period,
period 0 and 1. The representative consumers endowment in period 0 is n
0
R, and the total
random endowment is \ in period 1. There are J assets with zero net supply, which may be
traded before the uncertainty is resolved. Denote by 1
the price of asset . The price of good is normalized to be one in each period, and also
the units of assets are normalized so that the expected dividend 1 [1
=1
j
_
_
_
_
subject to (3)
a +
= n
0
.
(a) Assume that the rst order condition is sucient for utility maximization. Write the FOC.
(b) Since there is a single consumer and assets are in zero net supply, j
= 0 must hold in a
competitive equilibrium. Find the equilibrium price j
of asset .
(c) Note that for any random variables A and A, we have 1 [A1 ]1 [A] 1 [1 ] = CO\ [A, 1 ],
where CO\ indicates the covariance. Using this relation, rewrite the equilibrium pricing
formula above.
(d) Now assume that utility function is quadratic, n(a) = aa
1
2
a
2
, where a is positive. Re
write the formula you obtained above. Among those assets and asset / have the same
expected dividend. at type of assets tend to have high market price?
27. Equilibrium asset pricing model. Consider an economy with a single (representative) con
sumer with concave vNM utility index n and discount factor a. There is one good in each period,
period 0 and 1. The representative consumers endowment in period 0 is n
0
R, and the total
random endowment is \ in period 1. There are J assets with zero net supply, which may be
traded before the uncertainty is resolved. Denote by 1
the price of asset . The price of good is normalized to be one in each period. Thus the
problem of the consumer is to solve, given j,
max
i,
n(a) +a1
_
_
n
_
_
\ +
=1
j
_
_
_
_
subject to (4)
a +
= n
0
.
(a) Find the equilibrium price of asset .
(b) From now on, assume 1
1
= 1 for sure; that is Asset 1 is riskless. Suppose the random
endowment \ improves in the sense the new random endowment \
0
rst order stochas
tically dominates the original \. What will happen to the price of asset 1? Can you say
anything about the prices of the other assets?
(c) Suppose the random endowment \ improves in the sense the new random endowment
\
0
second order stochastically dominates the original \, i.e., the endowment is less risky.
Assume in addition that n
000
0. What will happen to the price of asset 1? Can you say
anything about the prices of the other assets?
28. Consider an agent with concave vNM utility function n. His total wealth is \ and he can invest
(save) in a riskless or a risky asset. The riskless asset pays 1 per unit in the next period and costs
1
per unit in this period. The risky asset pays r with probability 1 j and r with probability
j, and it costs
2
. Denote by j
1
(resp. j
2
) his demand for the riskless (resp. risky) asset.
(a) Write the budget constraint, for the case where short sales is allowed (i.e., he can hold a
negative amount of asset), and for the case it is not.
(b) What is the relationship between
1
and interest rate?
(c) Assuming short sales are allowed, show that assets are normal goods.
(d) Suppose he owns one unit of each asset. So his initial wealth is
1
+
2
. Suppose further
that he is the only trader in the markets (i.e., he is a representative trader (consumer).
Find the equation that characterizes the equilibrium asset prices.
(e) Assume n(.) = .
1
.
2
. Find the capital asset pricing formula in the question above.
29. Risk sharing. Consider an economy with 1 agents. Each agents preferences are represented
by a vNM utility function n
I
. Assume that each n
I
satises n
0
I
0 and n
00
I
< 0. The aggregate
income of this economy is denoted by j 0, and it will be distributed among the agents. Denote
by a
I
(j) _ 0 the income the agent i receives when the aggregate income is j, so
I
a
I
(j) = j
must hold. For simplicity, suppose that there are only nitely many possible aggregate income
levels, j
1
, , j
1
and denote by j (j

) 0, / = 1, ..., 1, the probability that the aggregate
income is j

(so
1
=1
j (j

) = 1).
(a) Assume 1 = 1 = 2. So the utility of agent i, i = 1, 2, can be written as n
I
(a
I
(j
1
)) j (j
1
) +
n
I
(a
I
(j
2
)) j (j
2
).
i. Let j
1
j
2
. Think of a
I
(j

) as consumption of good /, and draw an Edgeworth box
which represents feasible income distributions among the agents. Explain graphically
the conditions a Pareto ecient allocation must satisfy.
ii. Show that if j
1
= j
2
, a
I
(j
1
) = a
I
(j
2
) must hold for both i, for any Pareto ecient
allocation.
22
(b) Now assume 1 2, 1 2. So agent i receives a vector of income a
I
:= (a
I
(j

))
1
=1
, which
yields the expected utility l
I
(a
I
) :=
1
=1
n
I
(a
I
(j

)) j (j

)
i. Give the denition for a feasible allocation a :=
_
, (a
I
(j

))
1
=1
,
_
. Give the
denition of a Pareto ecient allocation, which does not rely on special properties of
utility functions such as dierentiability.
ii. Fix `
I
0, i = 1, ..., 1, and write the Pareto maximization problem of weighted sum of
expected utility functions,
J
I=1
`
I
l
I
, and derive the rst order condition for Pareto
eciency for interior points.
iii. Assume that n
I
(.) =
1
1
i
.
1
i
where where
I
is a positive constant, for every agent
i. Show that if a =
_
, (a
I
(j

))
1
=1
,
_
is Pareto ecient, then a
I
(j

) = a
I
j

for
/ = 1, ..., 1 where a
I
0 is a constant.
30. Consider an economy as follows. There are two consumers, i = 1, 2. There are two periods, and in
each period, consumers trade a single, perishable commodity (i.e., they cannot store the good).
There is no uncertainty. Consumer is preferences are represented by n
I
_
a
0
I
, a
1
I
_
= lna
0
I
+c
I
lna
1
I
where c
I
(0, 1) is a constant. Each consumer is endowed with one unit of the good in each
period. Thus consumers preferences are identical, except for the discount factor c
I
. The price
of good is normalized to be 1 in both periods. In period 0, consumers can save: denote by 
I
the
amount that consumer i saves in period 0. Let r be the interest rate. That is, if consumer i saves

I
units of the good, he receives (1 +r) 
I
units of good at the beginning of period 1 before the
trade of the consumption good takes place. Hence consumer i chooses
_
a
0
I
, a
1
I
, 
I
_
R
+
R
+
R.
Notice that negative saving, i.e., borrowing is allowed.
(a) Write the consumer i s budget constraint.
(b) For each i, derive the saving function 
I
(r), that is, 
I
(r) is the amount of good consumer
i saves in period 0 when the interest rate is r.
(c) Show that when the saving market clears, i.e., 
1
(r) + 
2
(r) = 0, the good market clears
in both periods 0 and 1.
31. Consider a pure exchange economy with 1 consumers and 1 goods. Each consumer i is char
acterized by utility function n
I
and endowments c
I
=
_
c
1
I
, ..., c
1
I
_
. Write a
I
=
_
a
1
I
, ..., a
1
I
_
_ 0
for consumer is consumption, and write a = (a
1
, ..., a
I
, ...., a
J
). Denote by j =
_
j
1
, ..., j
1
_
the prices of goods. Assume that utility functions are concave, dierentiable, and 1n
I
0.
Answer the following questions.
(a) Write the denition of a competitive equilibrium of this economy.
We say that a feasible consumption allocation (a
1
, ..., a
J
) R
1
+
R
1
+
allows no
gains from bilateral trade, if for any pair of consumers i, , there is no . R
1
such that
n
I
(a
I
+.) n
I
(a
I
) and n
(a
.) n
(a
).
(b) Show that if a feasible consumption allocation (a
1
, ..., a
J
) is Pareto ecient, it allows no
gains from bilateral trade.
(c) Let (a
1
, ..., a
J
) be a feasible consumption allocation such that a
I
0 for each i. Show
that if (a
1
, ..., a
J
) allows no gains from bilateral trade, (a
1
, ..., a
J
) is Pareto ecient. [Hint:
express the rst order condition for no gains from bilateral trade by a constrained maxi
mization problem.]
Microeconomic Theory problems by A. Kajii 23
32. [core (non)convergence] Consider an economy with two goods, two consumers. The con
sumers 1 and 2 have an identical utility function n
I
_
a
1
I
, a
2
I
_
= min
_
a
1
I
, a
2
I
_
, and their endow
ments are (2, 1) and (1, 2), respectively.
(a) Find the core of this economy (you may answer graphically).
(b) Suppose that this economy is replicated. What will happen to the core?
33. [Replica economy] Consider an economy with two goods, two types of consumers. The type 1
and 2 consumers have an identical utility function n
I
_
a
1
I
, a
2
I
_
= lna
1
I
+ lna
2
I
, and their endow
ments are (8, 2) and (2, 8), respectively.
(a) Verify that the allocation a
1
= (4, 4), a
2
= (6, 6) belongs to the core of the two consumers
exchange economy where there is one consumer of each type.
(b) Suppose that this economy is replicated once so that there are two consumers of each
type. Show that the allocation where type one consume a
1
= (4, 4), and type 2 consume
a
2
= (6, 6) is not in the core of this economy.
34. [fair allocation] Consider an exchange economy with 1 consumers and 1 goods, where the total
endowments of goods is . 0. A feasible allocation a = ( , a
I
, ) is called envyfree if
for any i and , n
I
(a
I
) _ n
I
(a
to produce.
(a) Formulate this problem as a strategic form game.
(b) Find a Nash equilibrium. Is it in dominant strategies?
(c) Find quantity
1
per rm that maximizes the sum of prots of the two rms. The formulate
a strategic form game where each rm can choose either
1
or the quantity you found in
( 1b) above. Find a Nash equilibrium of this game. Is it in dominant strategies? Discuss.
2. Cournot Competition and iterative deletion of dominated strategies. There are two
identical rms with constant marginal cost of production 0. The total demand for the product
is given by = 1 j. Each rm , = 1, 2, freely determines the quantity
to produce.
(a) Formulate this problem as a strategic form game.
(b) Find a Nash equilibrium. Is it in dominant strategies?
(c) Find dominated strategies. (Hint. Will the rm ever produce less than a monopoly rm
produces?)
(d) Assuming that the other rm will never choose a dominated strategy, nd dominated
strategy. (Hint. If a rm is sure that the other rm never produces more than the monopoly
amount, it should produce some positive amount.)
(e) Repeat this process of iteratively eliminating dominated strategies What do you get in the
limit?
3. Bertrand Competition. There are two identical rms with constant marginal cost of produc
tion c. The total demand for the product is given by = a /j, where a, / 0 Each rm freely
sets its price, but if they set dierent prices, every consumer chooses to buy the product from
the rm with the cheaper price. The demand will be split evenly if the prices are the same.
(a) Formulate this problem as a strategic form game.
(b) Find a Nash equilibrium.
4. Price competition with capacity constraint. In the setting in (3), we shall assume that
each rm , = 1, 2, can sell only up to a preset capacity limit /
0. When rm sets a
higher price j
than that of the other rm, rm does not necessarily lose all the demand if the
26
other rm is selling at its capacity /
I
. In such a case, demand for rm is the residual demand
1 /
I
j
. To simplify computation, set a = / = 1 for the demand function, and c = 0 for the
marginal cost.
(a) Formulate this problem as a strategic form game.
(b) Show that if /
1
_ 1 and /
2
_ 1, j
1
= j
2
= 0 constitute a Nash equilibrium.
(c) Show that j
1
= j
2
= 0 is not a Nash equilibrium if /
1
< 1 or /
2
< 1.
(d) Assume /
1
= /
2
=
1
3
. Find a Nash equilibrium.
5. Monopolistic Competition. There are two rms, 1 and 2, with constant marginal cost of
production c
, 0 < c
= 1 j
, where
. Discuss.
6. Free Rider Problem. Consider a community with 2 individual. Each individual own 1 unit of
consumption good. Individual is utility depends on private consumption of the good as well as
the amount of public good available in the community. Specically, individual is utility is a
I
+j,
where a
I
is the amount of good privately consumed, and j is the amount of the public good. The
public good can only be produced from consumption good: from . units of consumption good,
) (.) units of public good can be produced. So when each individual i decides to consume a
I
units of consumption good privately, ) (2 (a
1
+a
2
)) units of public good is produced. Assume
that ) (0) = 0, )
0
0, and )
00
< 0. Each individual i chooses a
I
strategically, and negative
consumption is not allowed.
(a) Dene a feasible allocation of consumption good and public good; that is, describe (a
1
, a
2
, j)
which can be achieved in this community. Then write the denition of a Pareto ecient
allocation of this community.
(b) Assume that )
0
(2)
1
2
. Find all Pareto ecient allocations.
(c) Will a Pareto ecient allocation be realized when )
0
(2)
1
2
? Explain.
(d) Will a Pareto ecient allocation be realized when )
0
(0) <
1
2
? Explain.(25 points)
7. Consider a community with 2 individuals. Each individual own 1 unit of consumption good.
Individual is utility depends on private consumption of the good as well as the amount of public
good available in the community. Specically, individual is utility is a
I
+ j, where a
I
is the
amount of good privately consumed, and j is the amount of the public good. The public good
can only be produced from consumption good: from . units of consumption good, ) (.) units
of public good can be produced. So when each individual i decides to consume a
I
units of
consumption good privately, ) (2 (a
1
+a
2
)) units of public good is produced. Assume that
) (0) = 0, )
0
0, and )
00
< 0, and )
0
(0)
1
2
)
0
(2) . Negative consumption is not allowed.
(a) Dene a feasible allocation of consumption good and public good, and write the denition
of a Pareto ecient allocation of this community.
(b) Find the rst order condition which characterizes Pareto ecient allocations.
Microeconomic Theory problems by A. Kajii 27
(c) Consider a game where both individuals simultaneously choose a
1
and a
2
. Give the den
ition of a Nash equilibrium of this game.
(d) Is a Nash equilibrium Pareto ecient? Why?
(e) Consider a game where the game above is repeated twice as follows: in each of the periods,
each individual is endowed with one unit of consumption good, and simultaneously choose
private consumption level. The utility is a discounted sum of utilities with discount factor
a (0, 1). Is there a subgame perfect Nash equilibrium where the rst period allocation
or the second period allocation are ecient? Why?
8. An Exhaustible Resource Commons Problem. (Dutta) Suppose two players, 1 and 2,
share a xed supply of j sh. Each player lives for exactly two periods. In the rst period, each
player i can consume a nonnegative amount of sh, c
I
, provided that c
1
+c
2
_ j. In the second
period, any remaining sh, j c
1
c
2
, are divided equally between the players. Player is utility
is given by
n
I
(c
I
, c
I
) = lnc
I
+ ln
_
j c
1
c
2
2
_
,
where lna is the natural logarithm of a. By convention, utility is if consumption of sh is
zero in the second period.
(a) Consider this as a simultaneous move game where each player chooses a strategy 
I
such
that 0 _ 
I
< j (that is, choosing 
I
= j is not possible). If 
1
+ 
2
_ j then c
I
= 
I
, but
if 
1
+
2
j then c
1
= c
2
= j2. Find each players best response rule and nd the Nash
equilibrium of the game.
(b) Suppose that a central planner can set c
1
and c
2
(subject to c
1
+ c
2
_ j). Assume the
planner aims to maximize social welfare given by n
1
+n
2
. Find her choice of c
1
and c
2
.
Now consider an analogous situation with players. Player is utility is now given by
n
I
(c
I
, c
I
) = lnc
I
+ ln
_
j c
1
6=I
c
_
where
6=I
c
as a strategic variable.
The marginal cost of production is zero.
(a) Suppose three rms choose quantities simultaneously. Find a Nash equilibrium.
(b) Suppose that the decisions are made sequentially by rms 1, 2, and 3 in this order. That is,
after rm 1 chooses
1
, rm 2 chooses
2
observing
1
, and so on. Find a subgame perfect
Nash equilibrium.
(c) Suppose that after rm 1 chooses its quantity, rms 2 and 3 choose their quantities simul
taneously. Thus rms 2 and 3 knows
1
but not each others choice of quantity. Find a
subgame perfect Nash equilibrium.
(d) Suppose that rm 1 chooses its quantity after rms 2 and 3 choose their quantities simul
taneously. Thus rm 1 knows
2
and
3
, but rms 2 and 3 must decide without knowing
any other rms decision. Find a subgame perfect Nash equilibrium.
2. Entry deterrence. Consider two rms, Incumbent and Entrant. Both rms can produce a
consumption good with a constant marginal cost c, 0 _ c < 1. The inverse demand for the
good is given by 1 Q, where Q is the total production of the good. Entrant however incurs a
xed entry cost 1 _ 0 if it chooses to produce. Denote by
J
and
J
the level of production of
Entrant and Incumbent, respectively.
(a) Suppose that once Entrant pays the entry cost, both rms will compete in the Cournot
fashion.
i. Find a subgame perfect equilibrium in which Entrant does enter, when 1 is small
enough.
ii. Find a subgame perfect equilibrium where Entrant chooses not to enter if 1 is large.
iii. Is subsidizing entry cost 1 (thus Entrants eective cost for entry is zero) a good policy
from the point of view of social welfare?
(b) Suppose that Incumbent can commit to its quantity produced
J
before Entrant makes its
entry decision, and its level of production.
i. Let 1 = 0. Find a subgame perfect equilibrium.
ii. Let 1 0. Will Incumbent produce more than the amount for the case of 1 = 0? If
so, why? [in this example this is going to be a degenerate case]
iii. Is subsidizing entry cost 1 (thus Entrants eective cost for entry is zero) a good policy
from the point of view of social welfare?
(c) Suppose that Incumbent can invest to reduce its marginal cost of production to 0, before
Entrant makes its entry decision. The xed cost of the investment is c 0. Once entry
takes place, both rms compete a la Cournot.
i. Show that if c is small and 1 is large, there is a subgame perfect equilibrium where
Incumbent invests, and Entrant does not enter.
ii. Is the cost reducing investment above socially desirable?
3. Imagine a developer building houses near a lake. There are two periods, t = 1 and 2. In each
period, the developer can construct a house per unit cost c 0. Denote by a

the total unit
Microeconomic Theory problems by A. Kajii 29
constructed in period t. Also denote by j

the price of house in period t which the developer
determines. The developer is interested in the discounted sum of prots with discount factor a,
0 < a _ 1. That is, the developer maximizes the sum of period 1 prots and a times period 2
prots. Because of nancial reasons, the developer must sell all units of houses in the
period they are built. In each period, after houses are constructed, many potential buyers
come to see the houses. Each buyer buys at most one unit. If a potential buyer buys a house,
his payo (in terms of money) is a (a
1
+a
2
), where a is a positive constant with a c. If he
does not buy a house, his payo is 0. Note that in period 1, the second period houses are yet
to be constructed, thus the buyers decision will depend on the expectation of a
2
. Assume that
if a buyer is indierent between buying a house and not buying, he will buy. Assume that a

s
are real numbers to simplify the question.
(a) Note that the payo from a house is decreasing in a
1
+a
2
. Interpret.
(b) Suppose that a
1
units have already been sold in period 1. So, if a
2
units of houses are
constructed, each buyers payo from a house is a ( a
1
+a
2
).
i. If the developer wishes to sell a
2
units in period 2, what j
2
will the developer choose?
ii. Find the number of houses a
2
the developer constructs.
(c) Suppose that at the beginning of period 1, the developer can somehow convince the buyers
that the number of houses sold in the next period will be a
2
, and assume that the developer
keeps his promise and construct a
2
units of houses in the second period.
i. If the developer wishes to sell a
1
units of houses in period 1, what will be the prices of
houses in period 1 and period 2?
ii. What will be the number of houses constructed in period 1,
iii. Find a
2
which maximizes the total prot, assuming that the developer constructs a
2
units of houses in the second period.
iv. In the solution above, will the developer has incentive to construct a
2
units in period
2, after the rst period houses are all sold?
(d) Suppose that the developer cannot commit to the number of houses to be constructed in the
second period, thus the buyers will take into account what the developer will do in period
2. How many houses will be constructed in period 1, and what will the total number of
houses?
4. Trade war. Consider two countries, 1 and 2. There is one domestic rm in each country and
they produce an identical good at zero marginal cost. Demand for the product is given by an
inverse demand function j = 1 in each county (so the demand is identical). Denote by
1
and
2
the quantity produced towards the domestic market in each country and denote by c
1
and c
2
the quantity exported form country 1 and 2 to the other country. So for instance the
price of good in country 1 will be 1 (
1
+c
2
).
(a) Suppose the rms behave as price takers, and there is no international trade. Find the
equilibrium production level for each country.
(b) Suppose the rms behave strategically a la Cournot. That is, rm 1 maximizes the sum
of prots from country 1 and 2 by changing
1
and c
1
, given
2
and c
2
, for instance. Find
equilibrium quantities
I
, c
I
, i = 1, 2.
(c) Consider the following game: rst, each country i x a per unit tari t
I
on import, and
then the rm compete a la Cournot fashion. Solve this game by rst nding the second
30
stage (i.e., the game where (t
1
, t
2
) is already selected) equilibrium, and then nding a Nash
equilibrium in tari (t
1
, t
2
).
5. Ination target. Consider two players, the public (P) and the monetary authority (M). The
payo of P is given by (
t
)
2
, where
t
is the expected rate of ination held by P and is
the actual rate of ination. The payo of M is given by
2
(j j)
2
, where j is the realized
GDP of the economy and j is the natural (full employment) level of GDP. That is, ination as
well as over/under production are costly to M. Suppose that there is a tradeo between the rate
of ination and GDP (i.e., a Phillips curve) given by (c j j) + (
t
) = 0 where 0 < c _ 1
is a constant, and this relation is known to the both players. So if c < 1, the full employment
GDP can be realized only with a surprise ination, i.e.,
t
0.
(a) Suppose P selects
t
and M controls simultaneously. Find a Nash equilibrium of this
game.
(b) Suppose P moves rst. That is, M can select the rate of ination after observing the
expected rate of ination. What happens?
(c) Suppose M moves rst. That is, M can commit to its target rate of ination before P
forms its expectation. What happens?
6. Alternative for prot maximization. Consider two rms, 1 and 2, producing an identical
good at marginal cost c (suciently low, say c < 1). The demand for the product is given by
j = 1. The quantity produced
1
and
2
by the rms 1 and 2 are simultaneously determined.
But for each rm, the production decision is done by a manager who is independent of the
owner of the rm; that is, the managers hired by the rms decide
1
and
2
. The rm needs
to pay for the manager, thus the total cost of production from the view point of the rm is c
(revenue of rm
)+(1 c
= 0, the manager of rm wants to maximize the prot of the rm, but otherwise
he may not be interested in prot maximization. If c
!
!
!
!
!
!
!
!
!
a
a
a
a
a
a
a
a
@
@
@
@
@
@
T
T
T
T
T
T
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
L
L
L
L
L
L
L
L
L
u
u u
u
u
u u
(8,8)
Nature
Player 1
Player 2
Player 2
(8,8) (10,0) (10,0)
(10,0) (0,10) (0,10) (0,10)
Figure 1: Game
1. (a) Does Player 2 observes Player 1s choice of action? Does the players observe Natures move?
(b) Show that both players choosing left at every information set constitutes a perfect
Bayesian equilibrium.
(c) Suppose instead Natures move is observable to both players. Write this environment in an
extensive form game. Find the unique subgame perfect equilibrium.
(d) Compare ex ante utility levels of players in (1b) and (1c). Comment.
2. A risk neutral principal and a risk averse agent. The agent chooses eort level c in [0, 1], which
is not observable. The agents concave vNM utility function is n(a) c, where a is the amount
of wage received. The output may be j
1
or j
1
, with j
1
j
1
. The conditional probability of
achieving j
1
given c is denoted by (c), where
0
0,
00
< 0. The realized output is observable
and veriable. Wage schedule is written as (n
1
, n
1
), i.e., when the observed output is high, the
agent receives n
1
, otherwise n
1
. The reservation utility level for the agent is normalized to be
0. Assume enough degree of dierentiability in the following.
34
(a) Suppose c were observable. Write down the rst order condition that characterize the
optimal eort level for the principle. Find the wage level that implements the optimal.
From now on, assume that c is not observable.
(b) Write down the agents utility maximization problem, given (n
1
, n
1
).
(c) Write down the principles problem. Show that at the optimum:
i. The rst best eort level you found in 2a cannot be achieved.
ii. The individual rationality constraint must be binding;
iii. n
1
n
1
.
3. Consider the Spence job market signaling model, with the following parameters:
output of high type 4
output of low type 2
cost of education for low type 1
wage for educated worker n
wage for noneducated worker 1
Assume high type and low type are equally likely. The wage for noneducated worker is xed at
1 by law, and the reservation payo for workers is 0.
(a) Suppose n 1 has been xed. Formulate the problem as in a simple game as in the
previous question. Find an equilibrium which yields the best outcome for the employer.
(b) Suppose the employer rst pick n, then the game considered above takes place. Find an
equilibrium that yields the best outcome for the employer.
(c) Suppose that n is chosen after education is nished, but before the productivity is revealed.
Find an equilibrium that yields the best outcome for the employer.
4. Consider a two player game as follows. Player 1 (Sender) observes his type rst, and then send
a message to Player 2. Player 2 (Receiver) chooses an action after observing the message. The
Player 1s types are, t
1
, or t
2
, and possible messages are 
1
or 
2
, and Player 2s actions are a
1
or a
2
. The types are equally likely. The payos are given in the following table, where c is a
constant. Notice that the payos are independent of the choice of messages. In words, there is
no cost for sending a message, i.e., player 1 is "just talking".
Player 1s payos
a
1
a
2
t
1
1 0
t
2
c 1
Player 2s payos
a
1
a
2
t
1
2 1
t
2
1 2
(a) Write this game in extensive form.
(b) Consider the following strategy prole: Player 2 chooses a
1
or a
2
with probability
1
2
whichever signal she receives (i.e., she ignores the message), and Player 1 sends 
1
or

2
with equal chance whichever the type is. Show that this prole constitutes a perfect
Bayesian equilibrium with appropriate beliefs.
(c) Consider the following strategy prole: Player 2 chooses a
1
or a
2
with probability
1
2
whichever signal she receives, and Player 1 sends 
1
if his type is t
1
and 
2
if his type
is t
2
. Does this constitute a perfect Bayesian equilibrium with appropriate beliefs?
Microeconomic Theory problems by A. Kajii 35
(d) Consider the following strategy prole: Player 1 sends 
1
if his type is t
1
, and 
2
if t
2
.
Player 2 chooses a
1
if the message is 
1
, chooses a
2
if the message is 
2
. Show that this
constitutes a perfect Bayesian equilibrium with appropriate beliefs, if c < 1. (So, in this
case, the message conveys some meaning.)
(e) Is there a perfect Bayesian equilibrium where the message conveys some meaning when
c 1?