Bargaining

Introduction
 Bargaining is an integral part of our life – we almost do it daily

 Everyone bargains –
children (toys and games),
countries (tariffs, border),
superpowers (nuclear deals, arms),
corporate houses (valuation),
political parties (seat sharing)
couples (everything)

 Trick of bargaining (to get good results) is devise good strategies

2
Characteristics of Bargaining
 All bargaining situations have two things in common

1. Total payoff reached through successful implementation of bargaining

exceeds the sum of individual payoffs without bargaining.
Without the possibility of this excess value, or “surplus,” the
negotiation would be pointless.
If I don’t see anything more through bargaining negotiations, then
why shall I go for bargaining at all?
Both Kennedy and Krushchev must have understood the potential
hazard of a nuclear war looming and hence, successful bargaining is
the only way to achieve peace and that outcome is certainly better
than building up nuclear arsenal as a threat mechanism (we can only
think about the money involved only)!
3
Characteristics of Bargaining
2. Bargaining is not a zero-sum game – it’s all about dividing the
surplus
Both tries to get a larger share of the surplus giving a lesser share to
the other
This may appear to be zero sum, but behind it lies the danger that if
the agreement is not reached, no one will get any surplus at all!
This mutually harmful alternative, as well as both parties’ desire to
avoid it, is what creates the potential for the threats—explicit and
implicit—that make bargaining such a strategic matter.

 Before formal game theoretic analyses this was vaguely referred to as

“bargaining power”

4
Simple Nash Equilibrium
 Let’s reconsider the “divide the dollar” game. Two people - $1 –
moves sequentially

 A claims a share followed by B - only the shares must add up to

anything less than or equal to $1

 NE: Any pair (x, y) adding up to 1

 Experimental evidences anything not less than 0.25 is given to B by

A

 Not exactly the unique solution we are looking for

5
Approaches to Bargaining

 One approach views bargaining as a cooperative game, in which the

parties find and implement a solution jointly, perhaps using a neutral
third party such as an arbitrator for enforcement.

 The other approach views bargaining as a non-cooperative game, in

which the parties choose strategies separately and we look for an
equilibrium.

 However, unlike our earlier simple game of simultaneous

announcements, whose equilibrium was indeterminate, here we
impose more structure and specify a sequential-move game of offers
and counteroffers, which leads to a determinate equilibrium
6
Nash’s Cooperative Solution: Example
 Two SiliconValley entrepreneurs, Andy and Bill.
 Andy produces a microchip set that he can sell to any computer
manufacturer for $900.
 Bill has a software package that can retail for $100.
 The two meet and realize that their products are ideally suited to each
other and that, with a bit of trivial tinkering, they can produce a
combined system of hardware and software worth $3,000 in each
computer.
 Thus, together they can produce an extra value of $2,000 per unit,
and they expect to sell millions of these units each year.
 The only obstacle that remains on this path to fortune is to agree to a
division of the spoils.
 Of the $3,000 revenue from each unit, how much should go to Andy
7 and how much to Bill?
Nash’s Cooperative Solution: Example
 Bill’s starting position is that without his software, Andy’s chip set is
just so much metal and sand, so Andy should get only the $900 and Bill
himself should get $2,100.

 Andy counters that without his hardware, Bill’s programs are just
symbols on paper or magnetic signals on a diskette, so Bill should get
only $100, and $2,900 should go to Andy.

 One might suggest they “split the difference”

 Bill might offer to split the profit on each unit equally with Andy.
Under this scheme, each will get a profit of $1,000. Then $1,100 of
the revenue goes to Bill and $1,900 to Andy.
8
Nash’s Cooperative Solution: Example
 Andy’s response might be that they should have an equal percentage of
profit on their contribution to the joint enterprise. Thus, Andy should
get $2,700 and Bill $300.

 The final agreement depends on their stubbornness or patience if they

negotiate directly with one another.

 What about a third party arbitrator?

 Suppose the arbitrator decides that the division of the profit should be
4:1 in favor of Andy – that is, Andy should get four-fifths of the surplus
while Bill gets one-fifth

9
Nash’s Cooperative Solution: Example
 Suppose Andy gets x while Bill gets y

 Andy’s profit = (x-900)

 Bill’s profit = (y-100)

 So, by the arbitrator’s formula: (x-900)=4 (y-100)

 Total revenue is $3000: (x + y)=3000

 Solving: x = 2500 and y = 500

⇒ Andy’s profit = (2500-900) = $1600
and Bill’s profit = (500-100) = $400
10
General Theory
 Two bargainers, A and B, seek to split a total value v, which they can
achieve if and only if they agree on a specific division.

 If no agreement is reached, A will get a and B will get b, each by

acting alone or in some other way acting outside of this relationship –
these are called backstop payoffs or BATNAs (best alternative to a
negotiated agreement).

 Hence, from the point of surplus core to bargaining, we must have –

a+b<v

 Surplus = (v – a – b )

11
General Theory
 Each player is to be given his BATNA plus a share of the surplus, a
fraction h of the surplus for A and a fraction k for B, such that –
h + k =1
 Hence, we must have the following payoffs –
x = a + h (v − a − b )
⇒ ( x − a ) = h (v − a − b )
y = b + k (v − a − b )
⇒ ( y − b ) = k (v − a − b )

 Therefore,
( x − a) h k ak k
= ⇒ y = b + ( x − a ) = (b − )+ x
( y − b) k h h h

 Also, (x + y) = v
12
General Theory
 We use these equations (expressions of Nash Formula) to solve for the
equilibrium –

13
Nash Formula
 The Nash formula says nothing about how or why such a solution
might come about.

 One might think of the Nash formula as a shorthand description of the

outcome of a bargaining process – then, h and k can stand for the two
parties’ relative bargaining strengths.

14
Nash’s Approach to Bargaining
 In all the games that we have studied so far, the players chose and
played their strategies separately from each other.
 We have looked for equilibria in which each player’s strategy was in his
own best interest, given the strategies of the others.
 Some such outcomes were very bad for some or even all of the players
(e.g., PD)
 Even in the scope for players to communicate, the best response
strategy was followed by every player (the only thing that they
followed was self interest)
 Joint good has to be an equilibrium outcome of separate action in such
games.
 What if joint action is possible?

15
Nash’s Approach to Bargaining
 What if the players might take all their actions immediately after the
agreement is reached, in one another’s presence?

 What if the players delegate the implementation of their joint

agreement to a neutral third party, or to an arbitrator?

 Nash modeled bargaining as a cooperative game.

 The thinking of a collective group that is going to implement a joint

agreement by joint action can be quite different from that of a set of
individual people who know that they are interacting strategically but
are acting non-cooperatively.

16
Nash’s Approach to Bargaining
 So, instead of searching for an equilibrium, in cooperative solution
people search for the best outcome and then devise a way to
implement it.

 Nash’s principles for unique outcome in bargaining –

1. The outcome should be invariant if the scale in which the payoffs are
measured changes linearly
2. The outcome should be efficient (implying, no available mutual gain
should go unexploited)
3. If the set of possibilities is reduced by removing some that are
irrelevant in the sense that they would not be chosen anyway, then
the outcome should not be affected (just like removal of non best
response startegies)
17
Explanation of Principle 1
 Remember risk preference –

 Linear transformation doesn’t change the attitude towards risk (risk-

neutrality)

 So, this principle conforms to the theory of expected utility – linear

transformation should have no effect on expected payoff calculations
and no effect on outcomes

18
Explanation of Principle 2
 In the example of A and B splitting v, the solution must lie on the line
(x + y = v)

 Generally, the complete set of logically conceivable agreements to a

bargaining game will be bounded above and to the right by the subset
of agreements that leave no mutual gain unexploited.

 So we can think of a curve y = f(x) that constitutes the complete set of

conceivable outcomes. The curve itself consists of the efficient
outcomes; there are no conceivable outcomes that include more of
both x and y than the outcomes on y = f(x); so there are no
unexploited mutual gains left. Therefore, we call the curve y = f (x) the
efficient frontier of the bargaining problem. (Think of PPF).
19
Nash’s Optimization
 Nash proved that the cooperative outcome that satisfied all three of
these assumptions could be characterized by the mathematical
maximization problem: choose x and y to –

maximize ( x − a ) h ( y − b) k subject to y = f ( x)
 Let, X = ( x − a); Y = ( y − b)

 Where, X + Y = ( x − a ) + ( y − b ) = (v − a − b ) = S

 Therefore, ( x − a ) h ( y − b) k = X hY k = X h ( S − X ) k

 FOC:

20
Nash’s Optimization Solution
 Solution:

 This boils down to the Nash formula –

( x − a) h k ak k
= ⇒ y = b + ( x − a ) = (b − )+ x
( y − b) k h h h

21
Nash’s Optimization

22
Variable Threat Bargaining
 We embed the Nash cooperative solution within a specific game, as the
second stage of a sequential-play game

 We assumed that the BATNAs (a and b) were fixed

 suppose there is a first stage to the bargaining game in which the

players can make strategic moves to manipulate their BATNAs within
certain limits

 After they have done so, the Nash cooperative outcome starting from
those BATNAs will emerge in a second stage of the game.

 This type of game is called variable-threat bargaining.

23
Variable Threat Bargaining
 Starting from original BATNAs (point P), Nash solution is Q
 If, player A can raise his BATNA (point P1), Nash solution Q’ – better
for A and worse for B

24
Variable Threat Bargaining
 Hence, a strategic move that improves one’s own BATNA is desirable

 Example: If you have a good job offer in your pocket—a higher

BATNA—when you go for an interview at another company, you are
likely to get a better offer from that employer than you would if you
did not have the first alternative.

 Question – is P’ the only option?

 If player A can make a strategic move that reduces player B’s BATNA
and moves the game’s backstop point to P2, the Nash solution starting
there leads to the same outcome Q’ that was achieved after A
increased his own BATNA to get to the backstop point P1
25
Variable Threat Bargaining
 This alternative kind of manipulation is equally in player A’s interest

 Example: You are already working and want to get a raise. Your chances
are better if you can make yourself indispensable to your employer so
that without you his business has much worse prospects; his low
outcome in the absence of an agreement—not offering you a raise and
your leaving the firm—may make him more likely to accede to your
wishes.

 More dramatically, if player A can make a strategic move that lowers

both players’ BATNAs so that the game’s backstop point moves to P3,
that again has the same result as each of the preceding manipulations.
This particular move is like using a threat that says, “This will hurt you
more than it hurts me.”
26
Variable Threat Bargaining
 In general, the key for player A is to shift the game’s BATNA point to
somewhere below the line PQ.

 The farther southeast the BATNA point is moved, the better it is for
player A in the eventual outcome.

 As is usual with threats, the idea is not actually to suffer the low payoff
but merely to use its prospect as a lever to get a better outcome.

27
Alternating-Offers Model I: Total Value
Decays
 We move back to the more realistic non-cooperative game theory and
think about the process of individual strategizing that may produce an
equilibrium in a bargaining game.

 Our standard picture of this process is one of alternating offers.

 Player A makes an offer. Player B either accepts it or makes a

counteroffer. If he does the latter, then A can either accept it or come
back with another offer of his own. And so on….

 Thus, we have a sequential-move game and look for its rollback

equilibrium.
28
Alternating-Offers Model I: Total Value
Decays
 To find a rollback equilibrium, we must start at the end and work
backward.

 But where is the end point?

 Why should the process of offers and counteroffers ever terminate?

 Perhaps more drastic, why would it ever start?

 Why would the two bargainers not stick to their original positions and
refuse to budge?

29
Alternating-Offers Model I: Total Value
Decays
 It is costly to both if they fail to agree at all, but the benefit of an
agreement is likely to be smaller to the one who makes the first or the
larger concession.
 The reason that anyone concedes must be that continuing to stand firm
would cause an even greater loss of benefit.

 This loss takes one of two broad forms –

1. The available pie, or surplus, may decay (shrink) with each offer
2. time has value and impatience is important, and so a delayed
agreement is worth less

30
Alternating-Offers Model I: Total Value
Decays
 A fan arrives at a professional football (or basketball) game without a
ticket. He is willing to pay as much as $25 to watch each quarter of
the game. He finds a scalper who states a price. If the fan is not willing
to pay this price, he goes to a nearby bar to watch the first quarter on
the big-screenTV.
 At the end of the quarter, he comes out, finds the scalper still there,
and makes a counteroffer for the ticket. If the scalper does not agree,
the fan goes back to the bar.
 He comes out again at the end of the second quarter, when the scalper
makes him yet another offer. If that offer is not acceptable to the fan,
he goes back into the bar, emerging at the end of the third quarter to
make yet another counteroffer.
 The value of watching the rest of the game is declining as the quarters
31 go by
Alternating-Offers Model I: Total Value
Decays – Rollback Analysis
 At the end of the third quarter, the fan knows that, if he does not buy
the ticket then, the scalper will be left with a small piece of paper of
no value.
 So the fan will be able to make a very small offer that, for the scalper,
will still be better than nothing.
 Thus, on his last offer, the fan can get the ticket almost for free.
 Backing up one period, we see that, at the end of the second
quarter, the scalper has the initiative in making the offer.
 But he must look ahead and recognize that he cannot hope to
extract the whole of the remaining two quarters’ value from the
fan.

32
Alternating-Offers Model I: Total Value
Decays – Rollback Analysis
 If the scalper asks for more than $25—the value of the third quarter to
the fan—the fan will turn down the offer because he knows that he
can get the fourth quarter later for almost nothing, so the scalper can
ask for $25 at most.
 Now consider the situation at the end of the first quarter. The fan
knows that if he does not buy the ticket now, the scalper can expect to
get only $25 later, and so $25 is all that the fan needs to offer now to
secure the ticket.
 Finally, before the game even begins, the scalper can look ahead and
ask for $50; this $50 includes the $25 value of the first quarter to the
fan plus the $25 for which the fan can get the remaining three
quarters’ worth.
 So, through full BI, the ticket will change hands for $50
33
Alternating-Offers Model I: Total Value
Decays – General Case
 Two bargainers, A and B. Suppose A makes the first offer to split the
total surplus (v)
 If B refuses the offer, the total available drops by x1 to (v - x1)
 B offers a split of this amount
 If A refuses B’s offer, the total drops by a further amount x2 to (v -
x 1 - x 2)
 A offers a split of this amount
 This offers, counter-offers continue for 10 rounds assuming
(v - x1- x2 – x3 - … - x10 ) = 0 and the game ends

34
Alternating-Offers Model I: Total Value
Decays – General Case
 If the game has gone to the point where only x10 is left, B can
make a final offer whereby he gets to keep “almost all” of the
surplus
 Left with the choice of that or absolutely nothing, A should
accept the offer
 Roll back one period – A must offer at least x10 to B else B will
refuse
 So, on round 9, A will offer a split where he keeps x9 and leaves
x10 to B
 Then on the round before B will offer a split where he gives x9 to
A and keeps (x8 + x10)
35
Alternating-Offers Model I: Total Value
Decays – General Case
 Working backward, on the very first round, A will offer a split
where he keeps (x1 + x3 + x5 +x7 + x9) and gives (x2 + x4 + x6 + x8
+x10) to B.

 This offer will be accepted.

 If each player has a positive BATNA which can easily be

incorporated (like, instead of x10, give a so that B is left with x10 –
a)

36
Alternating-Offers Model I: Total Value
Decays – General Case
 Remember: Later stages are never reached, game solved in 1st
stage; foresight)
 The other important point to note is that gradual decay (several
potential rounds of offers) leads to a more even or fairer split of
the total than does sudden decay (only one round of bargaining
permitted).
 In the latter, no agreement would result if B turned down A’s
very first offer; then, in a rollback equilibrium, A would get to
keep (almost) the whole surplus, giving B an “ultimatum” to
accept a measly cent or else get nothing at all.
 The subsequent rounds give B the credible ability to refuse a very
uneven first offer.
37
Experimental Evidence
 Decay Model – Ultimatum Game

 Only one chance of bargaining: player A makes an offer and, if B does

not accept it, the bargaining ends and both get nothing

 Structure of the experiment:

A pool of players is brought together; They are paired; one person in
the pair is designated to be the proposer (the one who makes the offer
or is the seller who posts a price) and the other to be the chooser (the
one who accepts or refuses the offer or is the customer who decides
whether to buy at that price). The pair is given a fixed surplus, usually
$1 or some other sum of money, to split.

38
Experimental Evidence
 Results:
The most common offer is a 50:50 split. Very few offers worse than
75:25 are made (with the proposer to keep 75% and the chooser to
get 25%), and if made, they are often rejected.

 Interpretations:
1. The players cannot or do not perform the calculation required for
rollback – but college students can easily do these calculations
2. The payoffs of the players include something other than what they
get out of this round of bargaining. Self esteem or pride that prevents
them from accepting a very unequal split. Also, the participants in
the experiment are gathered in a room, the anonymity of pairing
cannot be guaranteed and value relationships. Fairness.

39
Experimental Evidence
 Solutions:

1. The point about ongoing relationships can be handled by stricter

procedures that visibly guarantee anonymity
2. The fear of rejection can be removed by considering a variant called
the dictator game. A only proposes and B has to accept. The results
are little more unfair but not more than 70:30. Hence, left with
fairness.
3. In one variant, the participants were given a simple quiz, and those
who performed best were made proposers. This created a sense that
the role of proposer had been earned. When the dictator game was
played with earned rights and with stricter anonymity conditions,
most proposers kept everything but still 5% offered 50:50 split.

40
Alternating-Offers Model II: Impatience
 Lets consider a different kind of cost of delay – money value does not
decay but players have a “time value of money” and therefore prefer
early agreement to later agreement

 Let’s start with – both players believe that having only 95 cents right
now is as good as having $1 one round later

 A player who prefers having something right away to having the same
thing later is impatient; he attaches less importance to the future
relative to the present

 Reasons for impatience – (1) rate of return and (2) risk that the game
doesn’t continue
41
Alternating-Offers Model II: Impatience
 Two players; zero BATNAs; divide the $1

 The two bargainers are in identical situations when each makes his
offer, because the amount to be split is always $1.

 Thus, in equilibrium the amount that goes to the person currently in

charge of making the offer (x) is the same, regardless of whether that
person is A or B.

 Suppose A starts the alternating offer process.

 He knows that B can get x in the next round when it is B’s turn to
make the offer
42
Alternating-Offers Model II: Impatience
 A must give B at least an amount that is equivalent, in B’s eyes, to
getting x in the next round implying A must give B at least 0.95x now.

 Player A will not give B any more than is required to induce B’s
acceptance.

 A offers B exactly 0.95x and is left with (1 - 0.95x)

 But the amount that A gets when making the offer is just what we
called x.

 Therefore, x = (1 - 0.95x), or x = 0.512.

43
Alternating-Offers Model II: Impatience
 Two things about this calculation should be noted.

 First, even though the process allows for an unlimited sequence of

alternating offers and counteroffers in the equilibrium the very first
offer A makes gets accepted and the bargaining ends. Because time has
value, this outcome is efficient. The cost of delay governs how much A
must offer B to induce acceptance; it thus affects A’s rollback reasoning

 Second, the player who makes the first offer gets more (0.512 and
0.488). Each player gets more when he makes the first offer than
when the other player makes the first offer. But this advantage is far
smaller than that in an ultimatum game with no future rounds of
counteroffers.

44
Alternating-Offers Model II: Impatience
 Suppose the two players are not equally patient - Player B still regards
$1 in the next round as being equivalent to 95 cents now, but A
regards it as being equivalent to only 90 cents now – A is more
impatient.

 Let x be the amount that A gets when he starts the process and y for
what B gets when he starts the process

 A knows that he must give B at least 0.95y ⇒ x = 1 - 0.95y

 Similarly, y = 1 – 0.90 x
 Solving: x = 0.345 and y = 0.690
 So if A starts (0.345, 0.655) and when B starts (0.31, 0.69)

45
Alternating-Offers Model II: Impatience
 Once again, each player does better when he makes the first offer than
when the other player makes the first offer, and once again the
difference is small

 But unequal rates of patience has strong impact – whether A moves

first or B moves first, impatient A gets a lot less than B (almost 1/3)

 We expect that the person who is willing to accept less to get it sooner
ends up getting less, but the difference is very dramatic – with equal
rates of impatience the split was almost 50:50

46
Alternating-Offers Model II: Impatience
 Now we generalize the model

 A regards $1/(1 + r) immediately as being equivalent to $1 one offer

later and let a = 1/(1+r)

 Likewise, for B we have b = 1/(1+s)

 Hence, if r>s, implying a<b; A is more impatient than B

 We repeat the divide the $1 game an solve for rollback

47
Alternating-Offers Model II: Impatience
 Suppose A’s payoff in the rollback equilibrium is x when he makes the
first offer; B’s payoff in the rollback equilibrium is y when he makes
the first offer.

 When A is making the offer, he knows that he must give B an amount

that B regards as being equivalent to y one period later (by) and A can
keep the rest ⇒ x = (1 – by)
 Similarly we will have y = (1 – ax)

 Solution:

48
Alternating-Offers Model II: Impatience
 Observation 1: (x + y) >1
 Remember, its (x, 1-x) if A starts and (y, 1-y) if B starts
 Payoff higher if I move first than if my opponent moves first

 Observation 2: The discount factors (r and s) are small numbers –

lets assume r=0.01 and s=0.02; hence, the term rs = 0.0002 which
one can ignore and we have –

 Observation 3: x and y in the approximate solution are the shares of

the surplus that go to the two players, and y/x = r/s ; that is, the
shares of the players are inversely proportional to their rates of
impatience as measured by r and s. Patience is advantage
49
Reference
1. Games of Strategy (3rd Edition) by Avinash Dixit, Susan Skeath and
David H. Riley Jr.; Viva-Norton [Chapter 18].

50