Summer 2018 examination

EC441
Microeconomics for MRes students

Suitable for all candidates

Instructions to candidates

This paper contains six questions. Answer any TWO questions from section
A and any TWO questions from section B. All questions will be given equal
weight (25%).

Reading Time:15 minutes

Time allowed:
Writing Time: 3 hours
You are supplied with: no additional materials.

You may also use: no additional materials.

Calculators: Calculators are not allowed in this examination.

c LSE ST 2018/EC441

Page 1 of 6
 Section A

1. Consider an economy with two consumers, A and B, three goods, X, Y, and Z,

and a firm. The initial endowment of Consumer A consists of zero units of X, four
units of Y, and zero units of Z. The initial endowment of Consumer B consists of
two units of X, zero units of Y, and zero units of Z. The firm is owned by Consumer
A and has the production function
√
z1 = 2y1 + 8 x1
where z1 is the amount of Z produced, and x1 and y1 are the amounts of X and Y
used as inputs, respectively. Consumer A’s utility function is


uA xA , y A , z A = xA + y A ,
and Consumer B’s utility function is


uB xB , y B , z B = y B + z B ,
where xi is i’s consumption of X, y i is i’s consumption of Y, and z i is i’s consump-
tion of Z. Consumers and the firm are price-takers.

(a) [5 marks] Find the profit function of the firm.

(b) [10 marks] Find the Pareto efficient allocations.
(c) [10 marks] Find a Walrasian equilibrium price vector and allocation.

2. Consider a normal form, two-player game G = {Si , ui }i=1,2 , where Si is the set of
pure strategies of Player i and
ui : S1 × S2 × [0, 1] → R
is the payoff function of Player i, i = 1, 2. Note that the payoff of each player
depends on the profile of strategies played and a parameter in [0, 1]. The sets
Si are compact and convex subsets of R for i = 1, 2. The functions ui (s1 , s2 , γ),
i = 1, 2, are continuous and concave on S1 × S2 × [0, 1].

(a) [10 marks] Verify whether the conditions for existence of a Nash equilibrium
in pure strategies are satisfied for each value of the parameter γ in [0, 1].
(b) [15 marks] Define N E(γ) ⊆ S1 × S2 to be the set of Nash equilibria when
the value of the parameter is γ ∈ [0, 1]. Consider a sequence {γ n }∞ n=1 in [0, 1]
∗ n n ∞
converging to γ , and a sequence {(s1 , s2 )}n=1 in S1 × S2 that converges to
(s∗1 , s∗2 ) and is such that (sn1 , sn2 ) ∈ NE(γ n ) for n = 1, 2, .... Do the assump-
tions on Si and ui , i = 1, 2, imply that (s∗1 , s∗2 ) ∈ N E(γ ∗ )? Provide a proof
for your answer.

c LSE ST 2018/EC441

Page 2 of 6
 3. Consider the following bargaining game. Two risk neutral players, A and B, must
agree on how to divide a pie of size 1. The bargaining protocol is as follows.
The game lasts at most T rounds. In each of the rounds reached by the players,
one player is randomly selected to propose a division of the pie that the other
player can accept or reject. The probability with which a player is selected as the
proposer is constant across rounds: in each of the rounds reached by the players
Player A is selected with probability p > 0 and Player B with probability 1−p > 0,
independently of previous rounds. When Player i is selected as the proposer and
Player j accepts Player i’s proposal, i = j, i, j = A, B, the game terminates and
the proposed division is implemented. If Player i’s proposal is rejected by Player
j in round t < T , the game moves to the next round with probability 0 < δ < 1
and terminates with no agreement with probability 1 − δ. If Player i’s proposal is
rejected by Player j in round T , the game terminates with no agreement. Whenever
the game terminates with no agreement, the players get an outside option: Player
A gets zA
0 and Player B gets zB
0, where zA + zB < 1. If Player i, i = A, B,
gets y (her share of the pie or her outside option) in any round t, her payoff is y
(no discounting).

(a) [5 marks] Does there exist a Nash equilibrium in which the expected payoff
of Player A is equal to zA ? Explain your answer carefully.
(b) [5 marks] Does there exist a Subgame Perfect equilibrium in which the ex-
pected payoff of Player A is equal to zA ? Explain your answer carefully.
(c) [15 marks] Find the Subgame Perfect equilibrium payoffs when zA = zB = 0.

c LSE ST 2018/EC441

Page 3 of 6
 Section B

4. Suppose that a single object is sold to one of two bidders. The rules of the selling
mechanism are described as follows. Each bidder submits a bid (∈ R+ ) and the
winner of the object is the bidder with the highest bid. The winner pays the sum
of the bids and the loser pays nothing. The payoff of the winner is v − p, where v
is his valuation for the object and p is his payment. The payoff of the loser is zero.
(In case of ties, the winner is determined by a coin flip.) Prior to bidding each
bidder can invest a constant c in order to change the distribution of his valuation.
If a bidder invests c his valuation is uniformly distributed on [0, 1], otherwise
his valuation is 12 . The valuations are independently distributed conditional on
the investment decisions. The investment decisions are not observable and the
valuation is the private information of the bidder. Each bidder maximizes his
expected payoff.

(a) [3 marks] State the Revenue Equivalence Theorem as precisely as you can.
(b) [9 marks] Suppose that both bidders invested c. What is the symmetric
increasing equilibrium bidding strategy in this auction?
(c) [8 marks] For what values of c is there an equilibrium in which both bidders
invest c?
(d) [5 marks]What is the expected revenue of the seller if both bidders invest?

c LSE ST 2018/EC441

Page 4 of 6
 5. Consider the following specification of the Mussa-Rosen Model. There is a seller
who can produce a perfectly divisible good. His cost of producing x units is x2 /2.
There is a single buyer with quasi-linear utility, whose utility from consuming x
units of the good at price p (x) is xθ − p (x) where θ is the buyer’s privately known
type. The buyer’s payoff is quasi-linear in transfer. The buyer’s type distribution,
F , is uniform on [0, 2] and it is common knowledge. The trade between the buyer
and the seller proceeds as follows. The seller posts a price schedule {p (x)}x∈R+ .
Then the buyer chooses x and the seller delivers x units of the good at price p (x).

(a) [5 marks] Suppose that p is convex and differentiable and that the buyer with
type θ finds it optimal to buy a strictly positive quantity. Characterize the
quantity purchased by the buyer with type θ in terms of p′ .
(b) [20 marks] Characterize the profit-maximizing price schedule {p∗ (x)}x∈R+ .

c LSE ST 2018/EC441

Page 5 of 6
 6. Consider the following specification of the Moral Hazard Model studied in class.
The utility of the agent as a function of the wage, w, is ln w. If the agent exerts the
low effort, eH , the profit, π, is uniformly distributed on [2, 3]. That is, f (π|eH ) = 1
if π ∈ [2, 3] and zero otherwise. The agent’s cost of the low effort is zero and that
of the high effort is one. The agent’s outside option is zero.

(a) [5 marks] Suppose that the agent’s effort is observable and contractible. What
is the principal’s cost of implementing the high effort?
(b) [5 marks] Suppose now that effort is observable only if the agent, prior to
interacting with the principal, invests a fixed cost c (> 0) in a monitoring
technology. If the agent invests, this technology perfectly reveals the exerted
effort. Furthermore, the outcome of the monitoring is contractible. For what
values of c does the agent invest?
(c) [15 mark] Suppose now that effort is not contractible and the principal wants
to implement the high effort. Show that if the wage is a linear function of
the profit then the density, f (π|eL ), is also linear in profit.

c LSE ST 2018/EC441

Page 6 of 6
 Solutions
1.
16p2Z
(a) If 2pZ > pY , profits are unbounded. If 2pZ ≤ pY , the demand for x1 is
p2X
16p2Z
and profits are
pX
(b) First note that any Pareto efficient allocation y B = 0 because of the marginal
products of Y in production. Obviously, xB = 0 and z A = 0. Also assigning
z B units of Z to consumer B, we have from the production function that

 
z B = 2 4 − y A + 8 2 − xA
and thus
 zB
yA = 4 2 − xA + 4 − .
2
where 0 ≤ y A ≤ 4. Substituting, the utility of consumer A is equal to
 zB
4 2 − xA + 4 − + xA
2
which is decreasing in xA for xA ≤ 2. This is intuitive: when 0 ≤ x1 ≤ 2, the
marginal product of X is larger than the marginal product of Y. In particular,
we can find dx1 + dy1 < 0, where dy1 < 0, that keep the production of Z
constant (and increase the utility of consumer A as dxA + dyA > 0). Hence,
in a Pareto
√ efficient allocation, y1 > 0 only when x1 = 2, that is, when
z B ≥ 8 2. Thus, the Pareto efficient allocations are

√ √ 
(xA , y A , z A , xB , y B , z B , x1 , y1 , z1 ) = 2 − α, 4, 0, 0, 0, 8 α, α, 0, 8 α
for 0 ≤ α ≤ 2 and
 √ √ 
A A A B B B
(x , y , z , x , y , z , x1 , y1 , z1 ) = 0, 4 − α, 0, 0, 0, 2α + 8 2, 2, 0, 2α + 8 2

for 0 ≤ α ≤ 4.
(c) Set pX = 1. In any competitive equilibrium allocation, z A = xB = y B = 0
because of the First Welfare Theorem. Profits are 16p2Z from (a). Let’s try
the case in which consumer A purchases  both X and Y. Then pY = 1. By
16p2Z
(b), the production of Z is equal to 8 = 32pZ . Thus, equilibrium in
p2X
the Z market is given by
2
32pZ =
pZ
1
which yields pZ = . The equilibrium allocation is
4
(xA , y A , z A , xB , y B , z B , x1 , y1 , z1 ) = (1, 4, 0, 0, 0, 8, 1, 0, 8)

c LSE ST 2018/EC441

Page 7 of 6
 2.

(a) Standard verification.

(b) Suppose not. Without loss of generality, suppose there exists s1 ∈ S1 , and
ε > 0 such that
u1 (s1 , s∗2 , γ ∗ ) > u1 (s∗1 , s∗2 , γ ∗ ) + ε
By continuity, there exists N such that, for n > N

u1 (s1 , sn2 , γ n ) > u1 (s∗1 , s∗2 , γ ∗ ) + ε

By continuity, there exists N ′ such that, for n > N ′ ,

u1 (s∗1 , s∗2 , γ ∗ ) + ε > u1 (sn1 , sn2 , γ n )

Hence, for n > max{N, N ′ },

u1 (s1 , sn2 , γ n ) > u1 (sn1 , sn2 , γ n )

which contradicts the hypothesis (sn1 , sn2 ) ∈ NE(γ n) for n = 1, ..., ∞.

3.

(a) Yes. The strategy in which both players demand the entire pie when selected
as proposers and reject any proposals is an NE. The outcome for both players
is the outside option.
(b) No. In period T , in any subgame perfect equilibrium Player i demands 1 − zj
and this demand is accepted by player j. To see this, first note that since
any demand by Player i smaller than 1 − zj will be accepted by Player j,
Player i’s payoff as a proposer must be at least as large as 1 − zj . The claim
then follows from observing that the payoff of Player j as a responder must
be at least as large as zj . Thus, upon reaching period T , Player A’s payoff
is p (1 − zB ) + (1 − p)zA > zA as 1 − zB > zA . Then Player A, against any
SPE strategy of Player B, by demanding the entire pie and rejecting any
proposal before T , reaches T with positive probability and gets at least zA in
any previous round.
(c) Consider any round t and suppose that in round t + 1 there is a unique SPE
expected payoff for players A and B that is the same as in round T , that is,
p and 1 − p. Then, by the same reasoning in (b) for round T , in any subgame
perfect equilibrium, Player A, when selected as the proposer, demands

1 − δ (1 − p)

and her demand is accepted by Player B. Symmetrically, Player B demands

1 − δp

c LSE ST 2018/EC441

Page 8 of 6
 and her demand is accepted by Player A. Then, the equilibrium payoff of
Player A in round t is then

p (1 − δ (1 − p)) + δp (1 − p) = p

which is the same as the expected payoff in round t + 1 (and T ). Similarly,

the equilibrium payoff of Player B in round t is

δ (1 − p) p + (1 − p) (1 − δp) = 1 − p

which is the same as the expected payoff in period t + 1 (and T ). A simple

induction argument shows that these payoffs are the unique Subgame Perfect
equilibrium payoffs.

c LSE ST 2018/EC441

Page 9 of 6
 4. Suppose that a single object is sold to one of two bidders. The rules of the selling
mechanism are described as follows. Each bidder submits a bid (∈ R+ ) and the winner
of the object is the bidder with the highest bid. The winner pays the sum of the bids
and the looser pays nothing. The payoff of the winner is v − p, where v is his valuation
for the object and p is his payment. The payoff of the looser is zero. (In case of ties, the
winner is determined by a coin flips.) Prior to bidding each bidder can invest a constant
c in order to change the distribution of his valuation. If a bidder invests c his valuation
is uniformly distributed on [0, 1], otherwise his valuation is one half. The valuations
are independently distributed conditional on the investment decisions. The investment
decisions are not observable and the valuation is the private information of the bidder.
Each bidder maximizes his expected payoff.

a. [3 marks] State the Revenue Equivalence Theorem as precisely as you can.

solution:
Consider a single-object, n—bidder, IPV auction environment. Consider two mecha-
nisms which satisfy the following two properties: 1. the allocation of the object is the
same conditional on any realization of the value profile, and 2. the payoff of bidder i (for
all i) with the lowest possible valuation is the same in both mechanisms. Then 1. the
expected payoff of each bidder with a given valuation is the same in both mechanisms,
and 2. the expected revenue to the seller is the same in both mechanisms.

b. [9 marks] Suppose that both bidders invested c. What is the symmetric increasing
equilibrium bidding strategy in this auction?

solution:
The payoff of a bidder with v in the second-price auction would be
 v
2 v2 v2
v − xdx = v2 − = .
0 2 2

Let b (v) denote the equilibrium strategy in our mechanism. Then the payoff of a bidder
with v is  v
2
v − vb (v) − b (x) dx
0
because the probability of winning the object is v. By the Revenue Equivalence Theorem,
 v
v2 2
= v − vb (v) − b (x) dx,
2 0

and hence,  v
v2
= vb (v) + b (x) dx.
2 0
This is an identity, so after differentiating both sides by v

v = vb′ (v) + 2b (v) .

c LSE ST 2018/EC441

Page 10 of 6
This differential equation is solved by b (v) = v/3.

c. [8 marks] For what values of c is there an equilibrium in which both bidders invest
c?

solution
If both bidders invest, then by the Revenue Equivalence Theorem and the argument
in part b., the expected payoff of each bidder is
 1 2 1
v v3  1
dv − c =  − c = − c.
0 2 6 0 6

If a bidder deviates at the investment stage and does not invest, his payoff from the
auction would be the same as his payoff after investing the auction with v = 1/2. By
the argument in part b., this payoff is 1/8. Hence, there is no incentive to deviate if and
only if,
1 1
−c≥ ,
6 8
that is, c ≤ 1/24.

d. [5 marks]What is the expected revenue of the seller if both bidders invest?

solution
The expected payment of a bidder in the second-price auction is
 1 2
v 1
dv = .
0 2 6

The expected revenue is the sum of the expected payments, that is, 1/3. By the Revenue
Equivalence Theorem, the revenue is the same in our auction too.

c LSE ST 2018/EC441

Page 11 of 6
 5. Consider the following specification of the Mussa-Rosen Model. There is a seller
who can produce a perfectly divisible good. His cost of producing x units is x2 /2. There
is a single buyer with quasi-linear utility, whose utility from consuming x units of the
good at price p (x) is xθ − p (x) where θ is the buyer’s privately known type. The buyer’s
payoff is quasi-linear in transfer. The buyer’s type distribution, F , is uniform on [0, 2]
and it is common knowledge. The trade between the buyer and the seller proceeds as
follows. The seller posts a price schedule {p (x)}x∈R+ . Then the buyer chooses x and
the seller delivers x units of the good at price p (x).

a. [5 marks] Suppose that p is convex and differentiable and that the buyer with
type θ finds it optimal to buy a strictly positive quantity. Characterize the quantity
purchased by the buyer with type θ in terms of p′ .

solution:
The buyer’s (with type θ) problem is

max θx − p (x) ,
x

the first order condition is θ = p′ (x). That is, x = p′−1 (θ).

b. [20 marks] Characterize the profit-maximizing price schedule {p∗ (x)}x∈R+ .

solution:
We know from Mussa-Rosen that the quantity purchased by type θ, x (θ), is charac-
terized by
1 − F (θ)
θ− = c′ (x (θ)) ,
f (θ)
if the left-hand side is positive and x (θ) is zero otherwise. In this example, this equation
is
1− θ
θ − 1 2 = 2θ − 2 = x (θ) ,
2
if θ ≥ 1 and zero otherwise. Hence,

x (θ) + 2
θ= .
2
By part a., this implies that
x+2 x
p′ (x) = = + 1.
2 2
The solution is p (x) = x2 /4 + x + c. Then the constant c = 0 is pinned down by the
requirement buying nothing costs nothing.

c LSE ST 2018/EC441

Page 12 of 6
 6. Consider the following specification of the Moral Hazard Model studied in class.
The utility of the agent as a function of the wage, w, is ln w. If the agent exerts the
low effort, eH , the profit, π, is uniformly distributed on [2, 3]. That is, f (π|eH ) = 1 if
π ∈ [2, 3] and zero otherwise. The agent’s cost of the low effort is zero and that of the
high effort is one. The agent’s outside option is zero.

a. [5 marks] Suppose that the agent’s effort is observable and contractible. What is
the principal’s cost of implementing the high effort?

solution:

If effort is contractible then the principal pays a constant wage at which the partici-
pation binds, that is,
v (w) − c (eH ) = ln w − 1 = 0,
so w = e.

b. [5 marks] Suppose now that effort is observable only if the agent, prior to interact-
ing with the principal, invests a fix cost c (> 0) in a monitoring technology. If the agent
invests, this technology perfectly reveals the exerted effort. Furthermore, the outcome
of the monitoring is contractible. For what values of c does the agent invests?

solution
Irrespective of which effort is implemented and whether or not the effort is con-
tractible, the agent’s participation constraint binds. As a consequence, the agent never
invests.

c. [15 mark] Suppose now that effort is not contractible and the principal wants to
implement the high effort. Show that the wage is a linear function of the profit then the
density, f (π|eL ), is also linear in profit.

solution:
In general,

1 f (π|eL )
=λ+µ 1− ,
v′
(w (π)) f (π|eH )
where λ and µ are the Lagrange multipliers corresponding the participation and incentive
constraints. In this case, by v (w) = ln w and f (π|eH ) = 1, the previous equation
becomes
w (π) = λ + µ [1 − f (π|eL )] .
If w (π) = aπ + b (because w is linear in π) then
aπ + b = λ + µ [1 − f (π|eL )] .
This means that
b − λ aπ
f (π|eL ) = 1 − − ,
µ µ
which means that f (π|eL ) is indeed linear in π.

c LSE ST 2018/EC441

Page 13 of 6