Cheap Talk and Signaling

Sukanta Bhattacharya

Sep-Dec 2019

Sukanta Game Theory 9 November, 2019 1 / 28

Cheap Talk or Direct Communication
Works well when the players’ interests are well aligned
Assurance Game

Sally
Local Starbucks
Harry Local 2, 2 0, 0
Starbucks 0, 0 1, 1
Both Harry & Sally want to meet and prefer the Local, so one saying
“Let’s meet at the Local” will work.
Stage 1: H: Saying where to meet
Stage 2: Both go to Local if “Local” said, both to Starbucks if
“Starbucks” said.
Doesn’t work at all if players’ interes ts are diametrically opposed, as
in a Zero-Sum game, such as Penalty Shootout
Only a babbling equilibrium.
Sukanta Game Theory 9 November, 2019 2 / 28
What if there is some commonality of interest?

Will a message lead to a clear N.E.?

Will there be a cheap-talk equilibrium at one of the two N.E.?
Or will the message be ignored, with a babbling equilibrium at either
N.E.?
Battle of Sexes:
Bar Cafe
Bar 3, 1 0, 0
Cafe 0, 0 1, 3
Husband says Bar and Wife meets him there in the backward
induction equilibrium.

Sukanta Game Theory 9 November, 2019 3 / 28

The Broker’s Game

The Investor relies on the Broker to know whether a stock is Good or

Bad.
The Broker knows the stock quality. The Investor does not.
The Investor wants to buy when the stock is Good and sell when Bad.

Sukanta Game Theory 9 November, 2019 4 / 28

Broker’s incentive to tell the truth

x is a secret commission the Broker gets from inducing the Investor

to buy a Bad stock .
Suppose the Investor believes the Broker.
Broker’s payoffs:

Stock quality Broker’s advice Investor decision Broker’s payoff

G G Buy 1
G B Sell −1
B G Buy −1 + x
B B Sell 1

Sukanta Game Theory 9 November, 2019 5 / 28

Broker’s incentive to tell the truth

If the stock is Good, the Broker has no incentive to lie.

If the stock is Bad,
I if x < 2, there will be a Cheap-Talk equilibrium, with no incentive to
lie.
I if x > 2, the Broker has an incentive to lie.
If the Investor knows that x > 2, he may distrust the Broker’s advice,
and ignore it.
Babbling equilibrium.

Sukanta Game Theory 9 November, 2019 6 / 28

Signaling Examples

Entry deterrence
I Incumbent tries to signal its resolve to fight to deter entrants
Credence Goods
I Used car warranties
How can strong informed players distinguish themselves?
Can weak players signal-jam?

Sukanta Game Theory 9 November, 2019 7 / 28

Credibility?

If you have an incentive to exaggerate your own worth, how can you
credibly convey your private information?
An independent third party?
I If the information is truly private, this option may be unavailable.
A reputation for honesty?
I A valuable asset, and recognized as such. But without the visibility or
longevity in the market to develop such a reputation, unavailable.
How else?
What if there is some action — a signal — that is costly to take and
visible to the other party.
The signal must be more costly if lying than if telling the truth. Then
the other party might see the action and infer truth telling.

Sukanta Game Theory 9 November, 2019 8 / 28

Signals as communication device

Signals are a form of credible communication: one’s actions speak

louder than words. Talk is cheap ...
The job seeker.
I Private information: you could do the job well if hired.
I Sufficient to tell the employer?
I Credible signal: harder to get for an incompetent worker than for a
competent worker.
The would-be borrower.
I Private information: whether you intend to repay the loan.
I Sufficient to tell the bank?
I A good credit record as a signal.

Sukanta Game Theory 9 November, 2019 9 / 28

Signals as communication device

When signaling cannot occur — when information cannot be credibly

communicated — markets don’t function well, and inefficiencies
occur.

An inefficient outcome: although possible with complete information,

mutually beneficial trade does not occur.

Sukanta Game Theory 9 November, 2019 10 / 28

The beer & quiche model

A monopolist can be either a tough incumbent or a wimp (not tough).

Incumbent earns 4 if the entrant stays out.
Incumbent earns 2 if the entrant enters.
Entrant earns 2 if it enters against wimp.
Entrant earns -1 if it enters against tough.
Entrant gets 0 if it stays out.

Sukanta Game Theory 9 November, 2019 11 / 28

Beer & Quiche

Prior to the entrant’s decision to enter or stay out, the incumbent

gets to choose its “breakfast.”
The incumbent can have beer or quiche for breakfast.
Breakfasts are consumed in public.
Eating quiche “costs” 0.
Drinking beer costs differently according to type:
I a beer breakfast costs a tough incumbent 1. . .
I but costs a wimp incumbent 3.

Sukanta Game Theory 9 November, 2019 12 / 28

What’s Beer?

Toughness Beer
Excess capacity High output
Low costs Low price
Deep pockets Extravagant advertising and limit pricing

Sukanta Game Theory 9 November, 2019 13 / 28

Beer and Quiche

Sukanta Game Theory 9 November, 2019 14 / 28

Signaling Equilibrium

Can the Incumbent credibly use Beer to signal Toughness?

Consistency Checklist
1 Is the Entrant’s strategy optimal given her beliefs?
2 Is Incumbent’s strategy a best response to the Entrant’s strategy?
3 Are Entrant’s beliefs correct given Incumbent’s strategy?

Sukanta Game Theory 9 November, 2019 15 / 28

Separating Equilibrium

Sukanta Game Theory 9 November, 2019 16 / 28

Separating Equilibrium

Tough drinks beer.

Wimp eats quiche.
Entrant infers the true type.
Degenerate beliefs (0 or 100%).
Entrant should ignore prior information. . . . . . and use strategic
information.

Sukanta Game Theory 9 November, 2019 17 / 28

Credible Signals

Why doesn’t Wimp drink beer & deter entry?

I It’s too costly
This is the key feature of “credible signaling”!
What if the signal (beer) were a bit less costly?
Suppose the wimp prefers “beer & out” to “quiche & enter”
The “beer signal” can’t work!
If both types drank beer, the entrant would face 50:50 odds, and
enter!
Both types of incumbent are then better off with quiche
Cheap beer destroys signaling value
Pooling equilibrium: both eat quiche, entrant enters

Sukanta Game Theory 9 November, 2019 18 / 28

Lessons

1 Costly signals can be used more credibly.

I Warranties are expensive for sellers of bad cars
I Extra capacity must hurt inefficient firms more
2 Delicate balance.
I Cheap signals → no persuasion
I Expensive signals → no profit

Sukanta Game Theory 9 November, 2019 19 / 28

Education as a signal
Potential employees can use their training or education as a signal,
especially if training is more costly for less competent students ...
B wants to hire W, but can’t tell (private info) whether W is:
I highly productive (HP) or
I less productive (LP)
HP workers are worth 200 to B, but LP workers are worth 100.
W knows his worth.
Competition from other employers forces B to pay the full amount of
W’s perceived worth at the time of hiring.
W can choose to receive education before trying for a job, and get a
diploma.
Education has no effect on productivity, but is observed by B - an
extreme (“credentialist”) view of education.
Assume that schooling costs a LP worker more than an HP worker:
120 against 60. Why? Extra tutoring, etc.
Sukanta Game Theory 9 November, 2019 20 / 28
A Bayesian equilibrium

B’s beliefs or expectations are crucial.

If B believes that all with the diploma are HP and all without are LP,
then she will pay 200 to graduates and 100 to others.
If W is HP and
I a graduate, then he earns net 200 − 60 = 140.
I Without the diploma, W earns net 100,
⇒ education is profitable for HP W.
If W is LP and
I has no diploma, then he earns net 100.
I With the diploma W earns net 200 − 120 = 80.
=⇒ education is unprofitable for LP W.

Sukanta Game Theory 9 November, 2019 21 / 28

Separating Equilibrium

Given B’s belief, B is behaving optimally.

Given B’s strategy, both type workers’ choices are optimal.
B’s beliefs are correct given LP and HP W’s choices.
Sukanta Game Theory 9 November, 2019 22 / 28
Signaling

Signaling in this example works

the market succeeds in separating LP from HP workers, despite the
informational asymmetries.
For education to serve as a signal in this example, it must be
I hard (costly) enough to obtain to deter LP W,
I but not so hard (costly) to deter HP W.

Sukanta Game Theory 9 November, 2019 23 / 28

Signaling need not work

A change in B’s beliefs can change the equilibrium considerably.

Suppose B believes instead that although a diploma implies HP,
without a diploma W could be HP or LP.
She can’t distinguish the two.
This can be self-confirming:
Suppose there are 40% LP and 60% HP, and B knows these
proportions.
B’s expected worth of W is 0.6 × $200 + 0.4 × $100 = $160, which is
the wage B offers W without a diploma.
W with a diploma (+ HP) is paid $200.
What is the new equilibrium?

Sukanta Game Theory 9 November, 2019 24 / 28

Will W obtain a diploma?

To HP W, the value of education is wage minus education costs:

200 − 60 = 140, i.e. less than the 160 W could earn without a
diploma
⇒ education doesn’t pay, even for HP W.
To LP W, the diploma payoff is 200 − 120 = 80, but no diploma
guves him 160,
⇒ no diploma for him, either.
None is educated, and B’s expectations are confirmed: 40% of the
uneducated workers will be LP.

Sukanta Game Theory 9 November, 2019 25 / 28

Pooling Equilibrium

Sukanta Game Theory 9 November, 2019 26 / 28

Pooling Equilibrium

A pooling Bayesian equilibrium.

Equilibrium without signaling, even though a signal (education) is
available to workers.
Signaling cannot be guaranteed to work .
Again, B’s expectations are crucial.

Sukanta Game Theory 9 November, 2019 27 / 28

Wasteful Expenditures as Signals

Expenditures — such as education — even if yield no direct benefit in

themselves, can serve as communication devices, signals.
Any observable expenditures that are cheaper for ”good” signalers
than for ”bad” signalers might work .
How can you credibly communicate the value of your product (brake
linings) to potential customers (car manufacturers), when you ’re sure
that they’ll be satisfied?
Extravagant expenditures may signal success (wining & dining, lavish
brochures, high-rent address, etc), if the potential buyer knows that
you are relying on continued sales to cover these apparently
unproductive costs
If your product were of low quality, you couldn’t cover your
promotional expenditures.

Sukanta Game Theory 9 November, 2019 28 / 28