Microeconomics IV

Section 2
Aleksei Beliaev
March 23, 2023

1 Education (PS1Q3): part 3(c)

Formulate the dominance requirement for PBE refinement. Assume that the parameters
are such that the set Ω of pooling equilibria surviving the dominance requirement is non-empty.
Does Ω contain the pooling equilibrium which delivers the highest (pooling equilibrium) payoff
to θH ?

Solution. We will be looking at three different dominance requirements. Let A denote the
set of all possible sender’s signals, and let S denote the set of all possible receivers’ response
profiles in the subgame following a signal.

Dominance 1. We say that action a ∈ A is strictly dominated for type θ if there is an action
a′ ∈ A such that
min
′
u1 (a′ , s′ , θ) > max u1 (a, s, θ). ([Link].1)
s ∈S s∈S

Then we require that, if a is strictly dominated for θ, then the receivers should believe that
µ(θ | a) = 0 (provided that a is not strictly dominated for some other θ′ ).

In our model, S = R2+ if there are two competing firms, since they can in principle offer
any non-negative wage following any signal. Thus, no action is strictly dominated, since the
right-hand side of ([Link].1) is unbounded, and any PBE survives Dominance 1.

Dominance 2. For an action a ∈ A, define S ∗ (a) to be the set of all possible receivers’ response
profiles in equilibria of the subgame following the signal a for some beliefs. We say that action
a ∈ A is strictly dominated (in the stronger sense) for type θ if there is an action a′ ∈ A such
that
′
min
∗ ′
u1 (a′ , s′ , θ) > max
∗
u1 (a, s, θ). ([Link].2)
s ∈S (a ) s∈S (a)

Then we require that, if a is strictly dominated for θ, then the receivers should believe that
µ(θ | a) = 0 (provided that a is not strictly dominated for some other θ′ ).

In our model, S ∗ (e) = [θL + e, θH + e] for any education level e ≥ 0. To see that, note that
∗
S (e) is the set of equilibrium responses in the subgame following e for some beliefs, and the
wage offered by the competitive firms who believe µ(θH |e) = α is w = αθH + (1 − α)θL + e
where α ∈ [0, 1].
We claim that the pooling equilibrium that brings the highest payoff to the high type does
survive Dominance 2. See the illustration below:

1
 Here, e∗ denotes the education level chosen by the workers in the pooling equilibrium with
the highest payoff for the high type. Let e′p denote the level of education that brings the low type
the highest payoff on the lowest break-even line. Denote by ẽ the point of intersection between
the highest break-even line and the low type’s indifference curve passing through (e′p , θL + e′p ).
Note that any education level e > ẽ is strictly dominated for θL : as evident from the picture,
the maximum possible payoff for the low type at e is strictly less than the minimum payoff
at e′p . Moreover, if e is not too large, it is not strictly dominated for θH . Thus, we can place
µ(θH | e) = 1 and thus w(e) = θH + e for all e > ẽ. One can also easily see that there are no
further restrictions on beliefs by Dominance 2.
Draw the high type’s indifference curve passing through (ẽ, θH + ẽ) with the red dashed
line, and denote the point of its rightmost intersection with the middle break-even line by e′′p .
Note that the existence of this intersection is equivalent to saying that the set of surviving
pooling equilibria Ω is non-empty. Note also that, given the relative shapes of the two types’
indifference curves, e∗ ∈ (e′p , e′′p ), and thus it does survive Dominance 2.

Equilibrium Dominance. We say that action a ∈ A is equilibrium dominated for type θ in

a PBE (a∗ (θ), s∗ (θ), µ) if
u∗1 (θ) > max
∗
u1 (a, s, θ). ([Link].3)
s∈S (a)

Then we require that, if a is equilibrium dominated for θ, then the receivers should believe
that µ(θ | a) = 0 (provided that a is not equilibrium dominated for some other θ′ ).

In our model, the pooling equilbrium delivering the highest payoff to the high type does not
survive Equilibrium Dominance. See the following illustration:

2
 Here, e∗ depicts the education level in the pooling equilibrium delivering the highest payoff
to the high type. To see why it does not survive Equilibrium Dominance, consider the education
level e as shown in the picture. It is equilibrium dominated for θL , and not for θH , so the firms
should place µ(θH | e) = 1 and thus w(e) = θH + e. But then the high type wants to deviate
from e∗ to e, so this equilibrium does not survive Equilibrium Dominance.

2 Insurance (PS2Q3): dealing with non-existence

For the solution of the problem itself, please see the slides for Section 2. In particular, we
know that, if the model has an equilibrium, it is separating and unique. The condition for
existence is that the low risk type’s indifference curve passing through his contract must not go
lower than the pooling break-even line. But this condition fails for a wide range of parameters,
so very often the model does not give any prediction. Thus, we would like to come up with
an alternative solution concept that would allow the model to have a solution even though the
equilibrium in the usual sense does not exist.

Anticipatory equilibrium (Wilson, 1977). A set of contracts is an equilibrium if no firm

has a profitable deviation that remains profitable once existing contracts that lose money after
the deviation are withdrawn.
The following picture illustrates an anticipatory equilibrium:

This is a pooling anticipatory equilibrium. The profitable deviations are shown by the gray
area on the graph. Suppose initially both firms offer the pooling contract, but firm 1 decides to
deviate by offering another contract somewhere in the gray area. By doing so, firm 1 attracts
all low risk agents and receives a positive profit, since the new contract lies below the low
risk break-even line. However, firm 2 is now serving only the high risk agents with the old
pooling contract, and since it lies above the high risk break-even line, firm 2 has a negative
payoff. Therefore, firm 2 withdraws this contract, so that now all agents are served by firm 1.
However, firm 1’s contract lies above the pooling break-even line, so firm 1’s payoff becomes
negative following the withdrawal. Thus, the initial deviation is no longer profitable.

Reactive equilibrium (Riley, 1979). A set of contracts is an equilibrium if no firm has a

profitable deviation that remains profitable once new profitable contracts after the deviation
are added.
The following picture illustrates a reactive equilibrium:

3
 This is a separating reactive equilibrium. The profitable deviations are shown by the gray
area on the graph. Suppose initially firm 1 offers the contract for the low risk type, and firm
2 for the high risk type. Suppose firm 1 deviates by offering another contract somewhere in
the gray area. By doing so, firm 1 attracts all agents and receives a positive profit, since the
new contract lies below the pooling break-even line. However, firm 2 who now has no clients,
can react to the deviation by offering another profitable contract in the gray area (shown in
green on the picture) that is better for all agents, so it attracts them all. As a result, firm 1
is left with no clients and now receives a zero payoff. Thus, the initial deviation is no longer
profitable.

3 Productive Tasks (PS2Q4)

Extend the job market screening model to the case in which tasks are productive. Assume
that a type θ worker produces θ(1 + µt) units of output when her task level is t where µ > 0.
Identify the SPNE of this game.

Solution. Let us describe the SPNE of this game using a treatment similar to Chapter 13.D
in MWG. We prove the following five lemmas.

Lemma 1. In any equilibrium, whether pooling or separating, both firms must earn zero profits.

Lemma 2. No pooling equilibria exist.

Lemma 3. If (wL , tL ), (wH , tH ) are contracts signed by the low- and the high-ability workers
in a separating equilibrium, then both contracts yield zero profits; that is, wL = θL (1 + µtL ) and
wH = θH (1 + µtH ).

Proofs of Lemmas 1—3. See proofs of Lemmas 13.D.1—13.D.3 in MWG; the reasoning is
completely analogous.

Lemma 4. In any separating equilibrium, the low ability workers accept contract (ŵL , t̂L ) such
that:
t̂L = arg max θL (1 + µtL ) − c(tL , θL ), ŵL = θL (1 + µt̂L ).
tL ≥0

Proof. The expression for ŵL is due to Lemma 3. As for t̂L , consider two cases:

4
 1. Let t̂L = 0. The proof is completely analogous to that of Lemma 13.D.4 in MWG.
2. Let t̂L > 0. See the illustration below:

Suppose by contradiction that the low-ability workers sign the contract (wL , tL ) on the
low break-even line such that tL is either to the left or to the right of t̂L , as shown in the
picture. Then one of the firms has a profitable deviation in which it offers a contract (w̃, t̃)
somewhere in the grey area: the low-ability workers would strictly prefer this contract to
the old one, and the firm will earn positive profits, since the new contract lies below the
low break-even line.

To formulate the next lemma, define t̂L as in Lemma 4. Then let tH be defined by
θH (1 + µtH ) − c(tH , θL ) = θL (1 + µt̂L ) − c(t̂L , θL ).

Lemma 5. In any separating equilibrium, the high-ability workers accept contract (ŵH , t̂H )
such that:
t̂H = arg max θH (1 + µtH ) − c(tH , θH ), ŵH = θH (1 + µt̂H ).
tH ≥tH

Proof. The expression for ŵH is due to Lemma 3. As for t̂H , consider two cases:
1. Let t̂H = tH . The proof is completely analogous to that of Lemma 13.D.5 in MWG.
2. Let t̂H > tH . See the illustration below:

5
 Suppose by contradiction that the high-ability workers sign the contract (wH , tH ) on the
high break-even line such that tH > t̂H , as shown in the picture. Then one of the firms
has a profitable deviation in which it offers a contract (w̃, t̃) somewhere in the gray area:
the high-ability workers would strictly prefer this contract to the old one, and the firm
will earn positive profits, since the new contract lies below the high break-even line.

We can summarize the characterization of SPNE in this game by the following proposition.

Proposition. If SPNE exists, it is unique and described by Lemmas 4—5. Moreover, this
SPNE exists if and only if the share λ of the high-ability workers is sufficiently low.

Proof. The uniqueness part is immediate from Lemmas 1—5. For the existence part, see
the illustration below:

As we can see, there are two cases (left pictures and right pictures). If t̂H = tH (left
pictures), there exists a threshold value λ̄ such that, if the share of high-ability workers λ > λ̄,
then the high type’s indifference curve passing through their only possible equilibrium contract
necessarily intersects the pooling break-even line, in which case there is a profitable deviation
(as shown in the left pictures) that attracts all types and brings positive profits, as it lies below
the pooling break-even line. If instead t̂H > tH , then for any λ, the pooling break-even line is
below the high type’s indifference curve, so the equilibrium always exists.