EC319 Economic Theory and Its Applications,
Part II: Lecture 5
Leonardo Felli
NAB.2.14
13 February 2014
Bilateral Trade (Chatterjee and Samuelson, 1983):
I Consider the following simple model of bilateral trade (double
auction).
I Two players, a buyer and a seller: N = {b, s}.
I The seller names an asking price: ps .
I The buyer names an offer price: pb .
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 2 / 38
Bilateral Trade (2)
I The action spaces:
As = {ps 0}, Ab = {pb 0}.
I The seller owns and attaches value vs to an indivisible unit of
a good.
I The buyer attaches value vb to the unit of the good and is
willing to pay up to vb for it.
I The valuations for the unit of the good of the seller and the
buyer are their private information of each player.
I Player i {b, s} believes that the valuation of the opponent
vi takes values in the unit interval.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 3 / 38
Bilateral Trade (3)
I The type spaces:
Ts = {0 vs 1}, Tb = {0 vb 1}
I Player i {b, s} also believes that the valuation of the
opponent is uniformly distributed on [0, 1]:
s = 1, b = 1.
I The extensive form of the game is such that:
I If pb ps then they trade at the average price:
(ps + pb )
p= .
2
I If pb < ps then no trade occurs.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 4 / 38
Bilateral Trade (4)
I The payoffs to both the seller and the buyer are then:
(ps + pb )
if pb ps
us (ps , pb ; vs , vb ) =
v 2 if pb < ps
s
and
(ps + pb )
vb if pb ps
ub (ps , pb ; vs , vb ) = 2
0 if pb < ps
I Players strategies: ps (vs ) and pb (vb ). We consider strictly
monotonic and differentiable strategies.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 5 / 38
Sellers Best Reply
I Consider now the sellers best reply.
I This is defined by the following maximization problem:
max Evb {us (ps , pb ; vs , vb ) | vs , pb (vb )}
ps
I Consider now the sellers payoff, substituting pb (vb ) we have:
(ps + pb (vb ))
if pb (vb ) ps
us = 2
v if pb (vb ) < ps
s
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 6 / 38
Sellers Best Reply (2)
I or
(ps + pb (vb ))
if vb pb1 (ps )
us = 2
v
s if vb < pb1 (ps )
I The sellers maximization problem is then:
Z pb1 (ps ) Z 1
(ps + pb (vb )
max vs dvb + dvb
ps vb =0 vb =pb1 (ps ) 2
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 7 / 38
Sellers Best Reply (3)
I Recall that by Leibnizs rule:
!
Z (y )
G (x, y )dx =
y (y )
= G ((y ), y ) 0 (y ) G ((y ), y )0 (y ) +
Z (y )
G (x, y )
+ dx
(y ) y
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 8 / 38
Sellers Best Reply (4)
I Therefore the first order conditions are:
dpb1 (ps ) 1 dpb1 (ps )
vs ps + pb (pb1 (ps )) +
dps 2 dps
Z 1
1
+ dvb = 0
pb (ps ) 2
1
I or from ps = pb (pb1 (ps )):
dpb1 (ps ) 1 1
(vs ps ) + vb p1 (ps ) = 0
dps 2 b
I which gives:
dpb1 (ps ) 1
1 pb1 (ps ) = 0
(vs ps ) +
dps 2
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 9 / 38
Buyers Best Reply
I The buyers best reply is instead defined by:
max Evs {ub (ps , pb ; vs , vb ) | vb , ps (vs )}
pb
I Consider now the buyers payoff obtained substituting ps (vs ):
(ps (vs ) + pb )
vb if vs ps1 (pb )
ub = 2
0 if vs > ps1 (pb )
I we then get
Z ps1 (pb )
(ps (vs ) + pb )
max vb dvs
pb vs =0 2
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 10 / 38
Buyers Best Reply (2)
I Therefore the first order conditions are:
(ps (ps1 (pb )) + pb ) dps1 (pb )
vb +
2 dpb
Z ps1 (pb )
1
dvs = 0
2 vs =0
I or
dps1 (pb ) 1 ps1 (pb )
[vb pb ] vs 0 =0
dpb 2
I which gives:
dps1 (pb ) 1 1
(vb pb ) ps (pb ) = 0
dpb 2
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 11 / 38
Equilibrium Characterization
I To simplify notation we re-write pb1 () = qb () and
ps1 () = qs ().
I The two differential equations that define the best reply of the
seller and the buyer are then:
1
[qs (ps ) ps ] qb0 (ps ) + [1 qb (ps )] = 0
2
1
[qb (pb ) pb ] qs0 (pb ) qs (pb ) = 0
2
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 12 / 38
Equilibrium Characterization (2)
I Solving the second equation for qb (pb ) we obtain:
1 qs (pb )
qb (pb ) = + pb
2 qs0 (pb )
I Differentiating the previous equation with respect to pb we
get:
qs (pb ) qs00 (pb )
0 1
qb (pb ) = 3
2 (qs0 (pb ))2
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 13 / 38
Equilibrium Characterization (3)
I Substituting the equations for qb (ps ) and qb0 (ps ) above into
the first differential equation we get:
qs (ps )qs00 (ps )
qs (ps )
[qs (ps ) ps ] 3 + 1 ps 0 =0
[qs0 (ps )]2 2qs (ps )
I This is a second-order differential equation in qs () that has a
two-parameter family of solutions.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 14 / 38
Equilibrium Characterization (4)
I The simplest family of solution takes the form:
qs (ps ) = ps +
I Substituting in the second order differential equation we
obtain:
62 9 ps + (6 + 2 ) = 0
I Then the values = 3/2 and = 3/8 solve the
second-order differential equation.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 15 / 38
Equilibrium Characterization (5)
I In other words:
3 3
qs (ps ) = ps
2 8
I From the second differential equation we then get:
3 1
qb (pb ) = pb
2 8
I The definitions of qs () and qb () then imply:
2 1 2 1
ps = vs + , pb = vb +
3 4 3 12
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 16 / 38
Equilibrium Characterization (6)
I This is the (unique) Bayesian Nash equilibrium of this game.
I Notice now that it is efficient to trade whenever:
vb vs
I However in this double auction game trade occurs whenever:
pb ps
or
2 1 2 1
vb + vs +
3 12 3 4
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 17 / 38
Inefficient Trade
In other words, in equilibrium trade occurs whenever:
1
vb vs +
4
vb 6
vs = vb
1 .............................................................
..
..
..
Trade ..
..
..
..
..
..
..
1
..
vb = vs + 4 ..
..
..
6 ..
q -
(0, 0)
. 1 vs
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 18 / 38
Bilateral Trade (Myerson and Satterthwaite 1983):
I The key question is whether the inefficiency we found in the
double auction can be eliminated by choosing a different
trading mechanism.
I The revelation principle provides the right tool to get an
answer to the question above.
I Obviously, there is no principal, but the two parties at an
ex-ante stage, before they learn their private information act
as the mechanism designer.
I Assume further that this is a pure bilateral contract transfers
cannot involve a third party.
I In jargon the mechanism has to be budget balancing.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 19 / 38
Unavoidable Inefficincy
Theorem (Myerson and Satterthwaite 1983)
When parties to a contract bargain in a situation of bilateral
asymmetric information efficiency of trade cannot be achieved.
There always exist configurations of the parameter values that
yield no trade when trade is efficient.
The reason is that the inefficiency is necessary to induce separation
of the different types of buyer and seller.
In other words, the incentive compatibility constraints of the
parties imply that efficiency is no longer guaranteed.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 20 / 38
Dynamic Bayesian Games
I We now consider dynamic games in which the information of
the players is not perfect.
I These can be games of incomplete information like the static
games we considered so far.
I A key feature of the games of incomplete information is that
at least one of the player does not know a defining feature of
the other player that affect his/her own payoff.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 21 / 38
Games of Imperfect Information
I In a dynamic setting a different source of uncertainty may
arise.
I A player might not know the action chosen at an earlier stage
of the game by one of the opponents.
I Games that have this feature are known as games of imperfect
information.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 22 / 38
Games of Imperfect Information: Example
1
b
,@
A , @ B
, @
2
,
, @
r, @r
...............................
,J
@
, J
@
C , J D C
@ D
,
r, Jr r @r
J
@
(2, 0) (1, 1) (1, 3) (4, 1)
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 23 / 38
Games of Imperfect Information: Example (2)
I Player 2 does not observe the action chosen by player 1 when
taking her decision of which action to choose.
I Notice that this very simple game of imperfect information is
a very familiar game.
I The reason is that clearly since player 2 does not observe
player 1s action choice a1 {A, B} she cannot make her
choice of action a2 {C , D} depend on player 1s action
choice.
I Label the dotted line an information set (this is the set
containing both the history A and the history B of the game.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 24 / 38
Games of Imperfect Information: Example (3)
I Player 2s information at that stage of the game is such that
she cannot distinguish whether the game is at any of the
nodes that are in the same information set.
I In other words player 2 when taking her decision does not
know if the history of play is A or B.
I This implies that player 2 has to choose the same action at all
the decision nodes that are contained in the same information
set.
I In other words, player 2 has to choose the same action
following both the history A and the history B.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 25 / 38
Games of Imperfect Information: Example (4)
I We can then write the normal form associated with the
extensive form we just described as:
C D
A 2, 0 1, 1
B 1, 3 4, 1
I This in nothing but a standard normal form game that we
extensively analyzed at the beginning of the course.
I In particular it is immediate to check that the unique Bayesian
Nash equilibrium of this game of imperfect information is
(A, C ).
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 26 / 38
Games of Imperfect Information: Example (5)
From the point of view of the solution of this game it is irrelevant
whether:
I player 1 chooses a1 before player 2 and player 2 does not
observe player 1 choice when taking her own,
I player 1 and player 2 take their decisions simultaneously and
independently,
I player 2 chooses a2 before player 1 and player 1 does not
observe player 2 choice when taking her own.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 27 / 38
Games of Imperfect Information: Example (6)
In particular the same normal form is also associated with the
following game of imperfect information:
2
b
,@
C , @ D
, @
1
,
, @
r, @r
...............................
,J
@
, J
@
A , J B A
@ B
,
r, Jr r @r
J
@
(2, 0) (1, 3) (1, 1) (4, 1)
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 28 / 38
Games of Imperfect Information: Example (7)
Notice that the imperfect information makes a big difference in the
characterization of the normal form associated with the extensive
form of the game as well as the set of equilibria:
1
b
,@
A , @ B
, @
2 , @ 2
,
r, @r
,J
@
, J
C1 , D C @ D2
@
J 1 2
,
r, Jr r @r
J
@
(2, 0) (1, 1) (1, 3) (4, 1)
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 29 / 38
Games of Imperfect Information: Example (8)
I Clearly the game differs in the strategies of player 2.
I In particular the normal form associated with this game is:
C1 , C2 C1 , D 2 D 1 , C2 D1 , D2
A 2, 0 2, 0 1, 1 1, 1
B 1, 3 4, 1 1, 3 4, 1
I The set of equilibria of this game is also different:
[A, (C1 , C2 )] and [B, (D1 , C2 )] are Nash equilibria while
[A, (C1 , C2 )] is a Subgame Perfect equilibrium.
I Imperfect information can therefore reduce the set of
equilibria of the game.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 30 / 38
Harsany Trasformation
I We consider now the relationship between the games of
incomplete information and the games of imperfect
information.
I A game of incomplete information is a game in which at least
one of the players does not know whom he/she is playing
against.
I In the sense that, say, player 2 is uncertain about player 1s
payoff function and/or player 1s strategy space.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 31 / 38
Harsany Trasformation: Example
I For example, player 1 knows exactly which game is played at
any time.
9
I With probability player 2 thinks the game is Game A:
10
L R
U1 1, 1 0, 0
D1 0, 0 6, 6
1
I while with probability she thinks the game is Game B:
10
L R
U2 1, 0 0, 10
D2 0, 0 6, 6
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 32 / 38
Harsany Trasformation (2)
A game of imperfect information is instead a game in which at
least one player does not know the full history of the game.
Theorem (Harsanyi 1967)
It is always possible to re-write a game of incomplete information
as a game of imperfect information.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 33 / 38
Harsany Trasformation (3)
This is done by
I introducing a new player denoted Nature N .
I assuming that N is the first player to choose its action.
I assuming that N choice of action follows the believes of the
other players in the game.
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 34 / 38
Harsany Trasformation: Example (2)
I In the example described above assume that N s action space
is {A, B}.
I Further, N is the first player to choose its action.
I N randomizes in the decision of its action: chooses A with
9 1
probability 10 and B with probability 10 .
I Then we assume that player 1 observes the move of N while
player 2 does not observe the move of N .
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 35 / 38
Harsany Trasformation: Example (3)
In other words here is a game of imperfect information:
N
b
c
9 A c B
c 1
10 c 10
r
2 c
cr
.......................................
%e %e
L%. eR L% eR
% e % e
r 1 er
.....................
% r 1 er
.....................
%
U1 E D1 U1 L D1 U2 L D2 U2 L D2
E L L L
E
L L L
E L L L
r Er r Lr r Lr r Lr
(1, 1) (0, 0) (0, 0) (6, 6) (1, 0) (0, 0) (0, 10) (6, 6)
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 36 / 38
Harsany Trasformation: Example (4)
I The normal form associated with this game is
L R L R
U1 1, 1 0, 0 U2 1, 0 0, 10
D1 0, 0 6, 6 D2 0, 0 6, 6
A B
I Player 2 when choosing her action has to take into account
the move of N and this yield the expected payoffs:
2 (U1 , U2 ; L) = 9/10 2 (U1 , U2 ; R) = 1
2 (D1 , D2 , L) = 0 2 (D1 , D2 , R) = 6
2 (U1 , D2 ; L) = 9/10 2 (U1 , D2 , R) = 6/10
2 (D1 , U2 ; L) = 0 2 (D1 , U2 , R) = 64/10
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 37 / 38
Harsany Trasformation: Example (5)
I Therefore the unique Bayesian Nash equilibrium of this game
is:
[(D1 , D2 ), R]
I Notice that the definition of strategy for player 1 is a fully
contingent plan that has to specify an action choice for every
action of N .
I This is equivalent to the definition of strategy according to
which the strategy has to specify an action choice for every
type of the player.
I Every type corresponds to a move by the player N .
Leonardo Felli (LSE) EC319 Economic Theory and Its Applications 13 February 2014 38 / 38
