1 Axiomatic Bargaining Theory

1.1 Basic denitions

What we have seen from all these examples, is that we take a bargaining sit-
uation and we can describe the utilities possibility set that arises from it. In
general, we get a compact convex set that contains the utility point correspond-
ing to the disagreement outcome.
Nash, instead of starting with a real life bargaining situation, chose to take
the set of feasible utilities and the disagreement utilities as a primitive of the
model. In other words, he chooses to abstract from the details of the bargaining
situation and consider only the set of attainable utilities. This is what is behind
the following denition:
Denition 1 A bargaining problem is a pair S; d, where S IR
2
is a
compact convex set, d S, and there exists s S such that s
i
> d
i
for
i = 1, 2.
We must understand a bargaining problem as coming from some real life bar-
gaining situation. On the other hand, we do not know the bargaining situation.
There might be several bargaining situations that give rise to the same bargain-
ing problems. Two bargaining situations that induce the same pair S; d are
treated identically.
The assumption that the set of feasible utility pairs is bounded means that
the utilities obtainable as an outcome of bargaining are limited. Behind the
convexity assumption on S lies the idea that the set of agreements is a set of
lotteries over some set of physical outcomes and that individuals preferences
can be represented by a von Neumann-Morgenstern utility function. Players
can agree to disagree d S (the disagreement is a possibility), and there is
some agreement preferred by both to the disagreement outcome. This ensures
that the agents have mutual interest in reaching an agreement, although there
is a conict of interest over the particular agreement to be reached.
The set of all bargaining problems is denoted by B.
Denition 2 A bargaining solution is a function f : B IR
2
that assigns to
each bargaining problem S; d B a unique element of S.
Examples of Solutions:
(i) The disagreement solution. It assigns to each bargaining problemS; d B,
the disagreement point d. Since d S, it is a well-dened solution.
(ii) The player 1 dictatorial solution. It assigns to each bargaining problem
S; d the strongly ecient and individually rational point that maximizes
player 1s utility function.
1
(iii) The egalitarian solution. It assigns to each bargaining problem, the great-
est feasible point (s

1
, s

2
) that satises s

1
d
1
= s

2
d
2
.
(iv) The Kalai-Smorodinsky solution:
6
-
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
......................... .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.....................
s
1
s
2
Figure 1: Kalai-Smorodinsky.
(v) The Nash bargaining solution: It selects the unique solution to the follow-
ing maximization problem.
max
(s1,s2)S
(s
1
d
1
) (s
2
d
2
)
s.t. s
1
d
1
s
2
d
2
The Nash bargaining solution will be denoted by f
N
.
1.2 Nashs Axioms
Denition 3 The bargaining problem S

, d

is obtained from the bargaining

problem S, d by the transformations s
i

i
s
i
+
i
, for i = 1, 2 if
d

i
=
i
d
i
+
i
for i = 1, 2
and
S

= {(
1
s
1
+
1
;
2
s
2
+
2
) IR
2
: (s
1
, s
2
) S}.
2
Example: Consider S; d where
S = {(s
1
, s
2
) IR
2
: s
1
+ s
2
1 s
1
, s
2
0}
d = (0; 0)
Consider the following transformations:
s
1
2s
1
+ 1
s
2
s
2
+ 2
When we apply these transformations to S; d we get
S

= {(2s
1
+ 1; s
2
+ 2) : s
1
+ s
2
1 s
1
+ s
2
0}
d

1
= 2d
1
+ 1 = 1
d

2
= d
2
+ 2 = 2.
The rst property that we may want a bargaining solution to satisfy is
Invariance to equivalent utility representations (INV): Suppose that
the bargaining problem S

, d

is obtained from S; d by the transformations

s
i

i
s
i
+
i
where
i
> 0 for i = 1, 2. Then f
i
(S

, d

) =
i
f
i
(S; d)+
i
for i =
1, 2.
This axiom requires that the utility outcome co-vary with the representation
of preferences, so that any physical outcome that corresponds to the solution of
the problem S; d also corresponds to the solution of S

, d

.
In order to present the next axiom, we need another technical denition.
Denition 4 A bargaining problem is symmetric if d
1
= d
2
and (s
1
, s
2
) S if
and only if (s
2
, s
1
) S.
We can now state the symmetry axiom.
Symmetry (SYM): If the bargaining problem is symmetric, then f
1
(S; d) =
f
2
(S; d).
The next axiom is more problematic.
Independence of irrelevant alternatives (IIA): If S; d and T; d are
bargaining problems with S T and f(T; d) S, then f(S; d) = f(T; d).
The axiom relates to the (unmodeled) bargaining process. If the negotiators
gradually eliminate outcomes as unacceptable, until just one remains, then it
may be appropriate to assume IIA. On the other hand, there are procedures
in which the fact that a certain agreement is available inuences the outcome,
even if it is not the one that is reached.
Pareto eciency (PAR): Suppose S; d is a bargaining problem, s S, t S
and t
i
> s
i
for i = 1, 2. Then f(S; d) = s.
3
This axiom implies that the players never disagree (since we have assumed that
there is an agreement on which the utility of each player i, exceeds d
i
).
Note that the axioms SYM and PAR restrict the behavior of the solution on
single bargaining problems, while INV and IIA require the solution to exhibit
some consistency across bargaining problems.
Theorem 1 There is a unique bargaining solution f : B IR
2
satisfying the
axioms INV, SYM, IIA and PAR. It is the Nash bargaining solution.
Proof : We proceed in a number of steps
a) f
N
is well-dened.
b) We check that f
N
satises the axioms.
Invariance: Suppose that S

; d

and S; d are like in the statement of the

axiom. Then
S

= {(
1
s
1
+
1
;
2
s
2
+
2
) IR
2
: (s
1
, s
2
) S}
and d

i
=
i
s
i
+
i
i = 1, 2.
In other words, s

if and only if there exists s S such that

s

i
=
i
s
i
+
i
for i = 1, 2.
Therefore, if (s

1
, s

2
) S

, we have
(s

1
d

1
)(s

2
d

2
) = (
1
s
1
+
1

1
d
1

1
)(
2
S
2
+
2

2
d
2

2
)
= (
1
s
1

1
d
1
)(
2
s
2

2
d
2
)
=
1

2
(s
1
d
1
)(s
2
d
2
)
for some (s
1
, s
2
) S.
Now, (s

1
, s

2
) maximizes (s
1
d
1
)(s
2
d
2
) over S, if and only if
(s

1
d
1
)(s

2
d
2
) (s
1
d
1
)(s
2
d
2
) (s
1
, s
2
) S
if and only if

2
(s

1
d
1
)(s

2
d
2
)
1

2
(s
1
d
1
)(s
2
d
2
) (s
1
, s
2
) S
if and only if
(s

1
d

1
)(s

2
d

2
) (s

1
d

1
)(s

2
d

2
) (s

1
, s

2
) S

where s

i
=
i
s

i
+
i
.
Namely (
1
s

1
+
i
,
2
s

2
+
2
) maximizes (s

1
d

1
)(s

2
d

2
) over S

.
Symmetry: Let H(s
1
, s
2
) = (s
1
d
1
)(s
2
d
2
). Let (S; d) be a symmetric
bargaining problem. Assume that (s

1
, s

2
) S maximizes H over S, namely
(s

1
d
1
)(s

2
d
2
) H(s
1
s
2
) (s
1
, s
2
) S.
4
Since (S; d) is symmetric, d
1
= d
2
. Therefore,
(s

2
d
1
)(s

1
d
2
) H(s
1
, s
2
) (s
1
, s
2
) S. (1)
Since S is symmetric (s

2
, s

1
) S. Thus (1) means that (s

2
, s

1
) also maximizes
H over S. But since the maximizer is unique, it must be that (s

1
, s

2
) = (s

2
, s

1
)
which implies s

1
= s

2
.
IIA: Assume S T and that (s

1
, s

2
) S maximizes H over T. Namely
(s

1
d
1
)(s

2
d
2
) (s
1
d
1
)(s
2
d
2
) (s
1
, s
2
) T.
In particular,
(s

1
d
1
)(s

2
d
2
) H(s
1
, s
2
) (s
1
, s
2
) S.
Since s

S, the result follows.

PAR: Since H(s
1
, s
2
) is increasing both in s
1
and s
2
in the sense that if
s

1
> s
1
and s

2
> s
2
then H(s

) > H(s) (s
1
, s
2
) cannot maximize H if
there exist (t
1
, t
2
) S with t
1
> s
1
and t
2
> t
2
.
c) Finally, we show that f
N
is the only bargaining solution that satises all
four axioms.
Suppose that f is a bargaining solution that satises the four axioms. We
shall show that f = f
N
. Let S, d be an arbitrary bargaining problem. We
need to show f(S, d) = f
N
(S, d).
Step 1. Let f
N
(S, d) = (z
1
, z
2
). Since by the denition of bargaining
problem there exists (s
1
, s
2
) S with s
1
> d
1
and s
2
> d
2
, we have z
i
> d
i
for i = 1, 2.
Let S

; d

be the bargaining problem that is obtained from S; d by

the transformations s
i

i
s
i
+
i
, for i = 1, 2, which move the disagreement
point to the origin and the solution z = f
N
(S; d) to the point (
1
2
,
1
2
).
That is

i
d
i
+
i
= 0

i
z
i
+
i
=
1
2
for i = 1, 2
which yields

i
=
1
2(z
i
d
i
)
;
i
=
d
i
2(z
i
d
i
)
for i = 1, 2.
Since both f and f
N
satisfy INV we have
f
i
(S

, 0) =
i
f
i
(S, d) +
i
and
f
N
i
(S

, 0) =
i
f
N
i
(S, d) +
i
=
1
2
.
for i = 1, 2.
5
Hence, f(S, d) = f
N
(S, d) if and only if f(S

, 0) = f
N
(S

, 0). Since
f
N
(S

, 0) = (
1
2
;
1
2
) it remains to show that f(S

, 0) = (
1
2
;
1
2
).
Step 2: We claim that S

contains no point (s
1
, s
2
) for which s
1
+s
2
> 1,
s
1
0 s
2
0.
If it does let = s
1
+ s
2
1 > 0 and let
(t
1
, t
2
) = ((1 )
1
2
+ s
1
; (1 )
1
2
+ s
2
) = (1 )(
1
2
,
1
2
) + (s
1
, s
2
).
Since S

is convex, (t
1
, t
2
) S

. Now
t
1
t
2
= (1 )
2 1
4
+ (1 )
1
2
s
1
+ (1 )
1
2
s
2
+
2
s
1
s
2
= (1 )
2 1
4
+
(1)
2
[s
1
+ s
2
] +
2
s
1
s
2
(1 )
2 1
4
+
(1)
2
(s
1
+ s
2
) +
2
0
(1 )
2 1
4
+
(1)
2
(1 + )

1
4
[(1 )
2
+ 2(1 )(1 + )]

1
4
[1 + 2 (2 + 1)
2
]

1
4
+
1
4
[2(2 + 1)]
So t
1
t
2
>
1
4
for < 2

2+1
, contradicting the fact that
1
2

1
2
t
1
t
2
(namely
that f
N
(S

, 0) = (
1
2
,
1
2
)).
Step 3: Since S

is bounded, step 2 ensures that we can nd a rectangle

T that is symmetric about the 45
o
line and that contains S

, on the boundary
of which there is (
1
2
,
1
2
).
6
-
S
T
(1/2, 1/2)
Step 4: By PAR and SYM of f we have f(T, 0) = (
1
2
,
1
2
).
Step 5: By IIA we have f(S

, 0) = f(T, 0) so that f(S

, 0) = (
1
2
,
1
2
)
completing the proof. 2
1.3 Is any axiom superuous?
INV: The egalitarian solution satises PAR SYM and IIA. It does not satisfy
INV.
SYM: The bargaining solution dened by
arg max(s
1
d
1
)

(s
2
d
2
)
1
6
(d
1
, d
2
) (s
1
s
2
) S
where = 1/2 and 0 < < 1, satises INV, PAR and IIA but it is not
symmetric.
IIA: The Kalai-Smorodinsky bargaining solution satises INV, PAR and SYM
but does not satisfy IIA.
PAR: The disagreement point solution satises all the axioms except for PAR.
7