BE

510 Business Economics 1 - Autumn 2021

Problem Set 11

1. Solve the following game using the concept of Perfect Bayesian Equilibrium.

2, 1 3, 0
Up
Up

Left Right

Player 2 Player 1(a) Player 2

Do
1, 2 Do 0.5 0, 5

w
n
Nature

1, 5 0.5 3, 9
Up
Up

Player 2 Player 1(b) Player 2

Do
Do

w
0, 2 2, 5

2. Consider the following situation with two firms and a potential employee. The firms want to
hire the employee and they compete for her by making salary offers. A difficulty is that the
firms do not know how productive the employee is. What they do know is that her
productivity is either 𝐿 = 5 or 𝐻 = 10 and that the chance of either is 50-50. The firm that
succeeds in hiring the employee will thus earn either 𝜋! = 5 − Salary or 𝜋" = 10 − Salary.
Not hiring her implies a payoff of 0. The employee’s payoff is 𝑢 = Salary and does not depend
on her type. If she rejects both offers her payoff is (normalized as) 0.
(a) The sequence of moves is as follows. First, the firms simultaneously and independently
make a salary offer each (𝑤# and 𝑤$ ). Second, the employee observes these offers and
then decides which, if any, offer to accept. Find the Perfect Bayesian Equilibrium.
(b) Consider now the introduction of a Master in Management program. How does this affect
the equilibrium analysis? The sequence of moves is as follows. In stage 1, the firms
simultaneously and independently make two salary offers each: One for employees
without a Master in Management Degree (𝑤% ) and one for employees with such a degree
(𝑣% ≥ 𝑤% ). In stage 2, the potential employee observes these offers and then decides
whether or not to obtain the degree. Getting such a degree is more costly for the low-
productivity employee (the cost is 6) than for the high-productivity employee (the cost is
2). Assume that the employee knows her own productivity. In stage 3, the employee
decides which, if any, offer to accept.
Hints: Organize your search for solutions by looking at pooling cases and separating cases in terms of
getting the degree. Also, focus on equilibria that are symmetric regarding firms’ behavior, i.e. consider
only solutions in which the firms end up offering the same salaries (not excluding 𝑤! ≠ 𝑣! ).
 BE 510 Business Economics 1 - Autumn 2021

Problem Set 11 - Solutions

No guarantees for correctness. If you find errors in the proposed solutions, please let us know.

1. Solve the following game using the concept of Perfect Bayesian Equilibrium.

2, 1 3, 0
Up
Up

Left Right

Player 2 Player 1(a) Player 2

Do
1, 2 Do 0.5 0, 5

w
n
Nature

1, 5 0.5 3, 9
Up
Up

Player 2 Player 1(b) Player 2

Do
Do

w
0, 2 2, 5

n

In such signaling games where an informed player moves first we often refer to “separating”
and “pooling” equilibria. In a separating equilibrium the informed player conditions her
choice on Nature’s move (doing different things depending on the state of the world). In a
pooling equilibrium, on the other hand, she chooses the same action either way.
Consider first the possibility that player 1(a) chooses Left whereas player 1(b) chooses Right.
Player 2’s beliefs are then determined by this given strategy since both of his information sets
are reached with positive probability (there is no “off-the-equilibrium path”). Hence, under
this strategy player 2 concludes that the Left action reliably signals that player 1 is of type (a)
and the Right action reliably signals that player 1 is of type (b). His best response to this is to
choose Down in response to Left and Up in response to Right. Now we must ask conversely:
Is Left-Right a best response to Down-Up? Well, no: Given the Down-Up strategy player 1(a)
obtains a payoff of 1 from Left but 3 from Right.
Second possibility: Player 1(a) chooses Right and player 1(b) chooses Left. That is a non-
starter: For player 1(b) Left is actually strictly dominated.
Thus, this game has no separating equilibria!
In a pooling equilibrium player 1 would either choose Left independently of her type or Right
independently of her type. Again, we can immediately rule out Left-Left because this cannot
be a best response for player 1(b). Thus, we are left with Right-Right. Given this strategy what
is player 2’s best response? On the equilibrium path he will observe that Right has been
Problem Set 11 - Solutions BE 510 Business Economics 1 - Autumn 2021

chosen. However, given that both types choose Right this observation contains no information
about which type he is facing. Hence, player 2’s beliefs are determined by the probabilities
with which Nature made its initial move. Thus, in equilibrium player 2 attaches a 50-50
chance to being located in the top-right or in the bottom-right decision node. As a result,
player 2 calculates his expected utility from Up as 0.5 × 0 + 0.5 × 9 = 4.5 and his expected
utility from Down as 0.5 × 5 + 0.5 × 5 = 5. Down is better than Up. However, this means that
player 1(a) shouldn’t choose Right (which would yield 0 for her) but Left because that
guarantees a better payoff (either 1 or 2). Thus, again there is no equilibrium.
We have now ruled out all possible separating and pooling equilibria. What remains are
hybrid equilibria involving mixed strategies. However, the only type of player 1 who might
randomize is (a) because (b) still has got her strictly dominant option (Right). Thus, assume
that player 1(a) chooses Left with probability 𝑝 and Right with probability 1 − 𝑝.
What are the consequences of this for player 2’s beliefs? Well, now we are again in a situation
where both of player 2’s information sets are reached with positive probability. What can he
learn from observing “Left” and from observing “Right”? One thing is clear: Only type (a)
chooses Left with any positive probability; type (b) chooses Right with probability 1.
Therefore, when player 2 observes a Left this must mean that he is facing type (a) and his best
response is Down.
What should player 2 make of an observation of “Right”? This can either come from the
randomizing type (a) or from the strictly incentivized type (b). We can calculate the relative
chances of these two possibilities using Bayes’ rule. We get

1
Pr(Right|a) Pr(a) (1 − 𝑝) 2 1−𝑝
Pr(a|Right) = = =
1
Pr(Right|a) Pr(a) + Pr(Right|b) Pr(b) (1 − 𝑝) + 1 × 1 2−𝑝
2 2
and
1
Pr(b|Right) = 1 − Pr(a|Right) =
2−𝑝
Which value of 𝑝 is compatible with an equilibrium? We started by assuming that type (a)
randomizes. But (a) may only randomize if Left and Right provide the same expected utility
for her. Since player 2’s best response to Left is Down, as discussed above, player 1(a) realizes
that she will get a payoff of 1 by choosing Left. Choosing Right, on the other hand, will give her
either 3 or 0. Hence, to become indifferent between Left and Right, player 2 must randomize
in just the right way to achieve this. But this requires that player 2 is also indifferent between
his pure strategies on the right hand side of the tree. Therefore, his expected utility from
choosing Up and from choosing Down must be the same. Given his beliefs (expressed by 𝑝),
player 2’s expected utility from choosing Up is

1−𝑝 1 9
𝜋Up = ×0+ ×9=
2−𝑝 2−𝑝 2−𝑝

2

Problem Set 11 - Solutions BE 510 Business Economics 1 - Autumn 2021

and his expected utility from choosing Down is 5. They are the same when 𝑝 = 1⁄5.
Conversely, suppose that in response to Right player 2 chooses Up with probability 𝑞 and
Down with probability 1 − 𝑞. Then player 1(a)’s expected utility from Right is 3𝑞 and her
utility from Left is 1, as discussed. These are the same when 𝑞 = 1⁄3.
In summary, we get a unique mixed-strategy PBE in which
§ player 1(a) chooses Left with probability 1⁄5 and Right with probability 4⁄5;
§ player 1(b) chooses Right with probability 1;
§ after observing Left player 2 believes that player 1 is of type (a) with probability 1;
§ after observing Right player 2 believes that player 1 is of type (a) with probability 4⁄9
and is of type (b) with probability 5⁄9;
§ in response to Left player 2 chooses Down;
§ in response to Right player 2 chooses Up with prob. 1⁄3 and Down with prob. 2⁄3.

2. Consider the following situation with two firms and a potential employee. The firms want to
hire the employee and they compete for her by making salary offers. A difficulty is that the
firms do not know how productive the employee is. What they do know is that her
productivity is either 𝐿 = 5 or 𝐻 = 10 and that the chance of either is 50-50. The firm that
succeeds in hiring the employee will thus earn either 𝜋! = 5 − Salary or 𝜋" = 10 − Salary.
Not hiring her implies a payoff of 0. The employee’s payoff is 𝑢 = Salary and does not depend
on her type. If she rejects both offers her payoff is (normalized as) 0.
(a) The sequence of moves is as follows. First, the firms simultaneously and independently
make a salary offer each (𝑤# and 𝑤$ ). Second, the employee observes these offers and
then decides which, if any, offer to accept. Find the Perfect Bayesian Equilibrium.
In this situation the employee will accept any salary offer and will choose the best one
independent of her type. One could debate whether she would accept a zero-salary where
she becomes indifferent between accepting and rejecting. I will assume that she would
accept even that. Although the firms anticipate this behavior, it is not true that they can
offer salaries near zero. The reason is the competition between them. If one firm offers a
salary of, say, 2 the other firm should offer a slightly better salary (2 + 𝜀) to outbid the
first firm and attract the employee. The successful firm can expect to earn 𝜋% =
(5 + 10)⁄2 − 𝑤% . Thus, no firm would offer more than 7.5 and in equilibrium both firms
will indeed offer precisely 7.5 because any higher offer would lead to a loss if accepted,
and any lower offer would invite the rival firm to bid more. Given the two identical offers,
the employee may pick either firm (let’s say with probability 0.5).
(b) Consider now the introduction of a Master in Management program. How does this affect
the equilibrium analysis? The sequence of moves is as follows. In stage 1, the firms
simultaneously and independently make two salary offers each: One for employees
without a Master in Management Degree (𝑤% ) and one for employees with such a degree

3

Problem Set 11 - Solutions BE 510 Business Economics 1 - Autumn 2021

(𝑣% ≥ 𝑤% ). In stage 2, the potential employee observes these offers and then decides
whether or not to obtain the degree. Getting such a degree is more costly for the low-
productivity employee (the cost is 6) than for the high-productivity employee (the cost is
2). Assume that the employee knows her own productivity. In stage 3, the employee
decides which, if any, offer to accept.
Hints: Organize your search for solutions by looking at pooling cases and separating cases in terms of
getting the degree. Also, focus on equilibria that are symmetric regarding firms’ behavior, i.e. consider
only solutions in which the firms end up offering the same salaries (not excluding 𝑤! ≠ 𝑣! ).

If the employee obtains the degree she will have the choice between 𝑣# and 𝑣$ and will
accept the larger one. I will refer to the larger salary as 𝑣̅ = max{𝑣# , 𝑣$ }. If the employee
does not obtain the degree she will have the choice between 𝑤# and 𝑤$ and will again
accept the larger one (𝑤 G).
Should the employee get the degree? This depends on how much more money she can
earn by obtaining the degree. The high type will choose the degree as long as 𝑣̅ − 𝑤
G ≥ 2,
and the low type will choose the degree as long as 𝑣̅ − 𝑤
G ≥ 6.
Pooling case 1: Both types obtain the degree.
The question is whether there is such an equilibrium. This case would require that the
salary difference is very substantial: 𝑣̅ − 𝑤 G ≥ 6. Let’s begin by assuming that there is such
an equilibrium and let’s refer to the equilibrium salaries as 𝑣#∗ = 𝑣$∗ and 𝑤#∗ = 𝑤$∗ . Thus,
firm 𝑗 offers salaries 𝑣'∗ and 𝑤'∗ ≤ 𝑣'∗ − 6. Under what circumstances is it a best response
for firm 𝑖 to choose the same salaries? To find the answer, note that what matters in
particular is the level of 𝑣'∗ because that is the salary that will actually be paid. If both
types of employee acquire the degree, the expected profit of the firm that ends up hiring
the employee is 𝜋% = 7.5 − 𝑣% . Therefore, firm 𝑖 would have an incentive to outbid firm 𝑗
for any 𝑣'∗ < 7.5 and would have an incentive to undercut firm 𝑗 for any 𝑣'∗ > 7.5. Thus,
we must have 𝑣#∗ = 𝑣$∗ = 7.5 (and both firms earn zero). This, in turn, implies 𝑤'∗ ≤ 1.5.
Given that 𝑤'∗ ≤ 1.5 could firm 𝑖 increase its profit by offering more to employees without
a degree? It can indeed. Suppose firm 𝑖 offers some 𝑤% > 1.5 such that the low-
productivity employee would no longer want to acquire the degree. Instead, the low-
productivity employee will now choose firm 𝑖 for sure and firm 𝑖 will earn 5 − 𝑤% when
the employee turns out to be the low type and earn 10 − 7.5 with a 50% chance when the
employee turns out to be the high type. This is clearly better than the zero profit from the
original plan. Thus, there cannot be an equilibrium of this kind.
Pooling case 2: Neither type obtains the degree.
This case would require that the salary difference is very small: 𝑣̅ − 𝑤
G < 2. Assume again
that there is such an equilibrium. Firm 𝑗 offers salaries 𝑤' and 𝑣' ∈ [𝑤'∗ , 𝑤'∗ + 2). Just as
∗ ∗

before, this can only work if the salary that is actually paid is 7.5 and both firms earn zero.
Thus, 𝑤'∗ = 7.5 and 𝑣'∗ < 9.5. But given that, firm 𝑖 could do better by offering 𝑣%∗ = 9.5

4

Problem Set 11 - Solutions BE 510 Business Economics 1 - Autumn 2021

and incentivize the high-productivity type to obtain the degree and join firm 𝑖. Then, firm
𝑖’s payoff will be 10 − 9.5 = 0.5 if Nature selects the high-productivity type. Firm 𝑖 would
no longer want to match firm 𝑗’s offer of 7.5 for an employee without a degree because
that would now definitely be the low type. For example, firm 𝑖 might choose 𝑤% = 5
(anything below 7.5 will do). Overall, firm 𝑖 would make a positive profit in expectation.
Therefore, there is no pooling equilibrium such that neither type obtains the degree.
Separating case: Only the high-productivity type acquires the degree.
This case would require that the salary difference is moderate: 2 ≤ 𝑣̅ − 𝑤 G < 6. Again, the
competitive pressure between the firms would drive up the salaries. Anticipating that the
high type gets the degree and the low type doesn’t, the salaries would have to reflect the
respective productivities, i.e. 𝑤'∗ = 5 and 𝑣'∗ = 10. Offering a 𝑤% < 5 does not bring any
improvement to firm 𝑖. This would have no effect on the high-productivity type and would
only mean that the low-productivity type will no longer want to join firm 𝑖. Offering a
𝑤% > 5 would make things worse because now the low-productivity type would definitely
want to joint firm 𝑖 but this would cause a loss for 𝑖. Offering a 𝑤% > 8 would change
things because now the high-productivity employee would no longer want to obtain the
degree and both types would join firm 𝑖. Again, however, this implies a loss for firm 𝑖.
Offering a 𝑣% > 10 would obviously be crazy and, finally, offering a 𝑣% < 10 would not
bring about an improvement either because it just means that the high-productivity,
degree-holding type will prefer firm 𝑗.
The last thing to double-check is whether the two types of employee really want to get
the degree/not get the degree under the incentives they face. The high type obtains net
earnings of 10 − 2 = 8 when she gets the degree; she has no incentive to mimic the low
type who earns only 5. The low type, on the other hand, would earn even less, 10 − 6 =
4, if she obtained the degree. Hence, this a Perfect Bayesian Equilibrium.
Brief discussion: In this very basic model the purpose of the degree is to identify the high-
productivity type. However, this is actually not a signaling game but a screening game.
The key difference is that the uninformed party moves first here. Another, more general
point to make is that because the degree does not change the employee’s productivity (by
assumption, just like in the example discussed in the lecture) there is—from a societal
perspective—actually no real economic benefit from having it. On the contrary, in
equilibrium it creates a cost of 2 when Nature decides to endow the employee with high
productivity. Notice something interesting here: While it seems obvious that this cost is
paid by the high-productivity type, there is a deeper truth hiding underneath this
observation. Compare the results from parts (a) and (b). Who actually benefits from the
introduction of the program? Who loses out?