328
(y=20, q=5/6)
Seller
’
s value
$20 [s]
$100 [w]
EU(str)=$0
[s]
$80
0, *
5/6, $80
EU(wk)=$33.33
[w]
$0
5/6, $20
1, $50
P(trade), E(price if trade)
Buyer
’
s value
Figure 4. An ex-ante incentive-efficient mechanism.
One of the basic contributions of mechanism design to economic analysis
is an extension of how we think about efficiency in markets and other systems
for economic transactions. In Figure 4, the probability of the mediator fail-
ing to achieve a mutually beneficial trade is strictly positive (it is (2/4)(1/6)
= 1/12). Ex post, after the traders have reported their types and one of them
has reported weakness, any such failure to realize mutually beneficial gains
from trade would be allocatively inefficient, as the object is not being allo-
cated to the person who can make the most valuable use of it. Such was the
classical view of efficiency that economists regularly applied before mecha-
nism design. But now we understand that, when the participation constraints
and informational incentive constraints are taken into account, no incentive-
compatible mediation plan can have a lower probability of such ex-post al-
locative inefficiency here than this mechanism in Figure 4.
When the expected profits are calculated ex-ante, before either indi-
vidual’s type is determined, the expected profit for each individual is
[0+0+(5/6)20+(1)50]/4 = 16.67 in this mechanism. In fact, there is no incen-
tive-compatible mechanism that could yield a higher ex-ante expected profit
for both individuals here. So in this sense (following Holmström and Myerson
1983), we may say that this mechanism is ex-ante incentive efficient.
Notice that this concept of incentive efficiency applies to the mechanism
for determining how the resource allocation will depend on people’s infor-
mation, not just to a particular resource allocation, and it takes incentive
constraints into account in asking whether any other feasible mechanism
could be better for these individuals. In ex-ante incentive efficiency, each
individual’s welfare is evaluated at the ex-ante stage, before anyone learns his
or her type.
The incentive-efficient mechanism in Figure 4 would not look so good to
the seller if he knew he was the strong type, however, because this plan never
allows the strong seller to get any positive profit from selling the object for
more than his value of 80. So if we care about individuals’ welfare as they
assess it in the game itself, when each knows his own type, then we need to
admit other mechanisms as incentive-efficient.
Figures 5 and 6 show other symmetric mediation plans that have y<20, and
so allow a strong individual to get a positive profit margin against a weak op-
ponent, but keep the incentive constraint binding with q = 25/(50–y), so that
the trading probability q is as large as possible given y. These mechanisms
are better for the strong types than Figure 4, and they are interim incentive
efficient, in the sense that there is no other incentive-compatible mechanism
that would be preferred by each possible type of each individual in the game.
329
(The word interim here refers to the fact that we are evaluating each indi-
vidual’s welfare at a time when he has learned his own type but has not yet
learned any others’ types; see Holmström and Myerson, 1983.) The mecha-
nism in Figure 6 (which is the best among these for the strong types), is the
solution identified by a natural generalization of Nash’s (1950) bargaining
solution, as defined by Myerson (1984).
Seller__’__s_value__$20_[s]__$100_[w]___EU(str)=$3.125'>(y=10, q=5/8)
Seller
’
s value
$20 [s]
$100 [w]
EU(str)=$3.125
[s]
$80
0, *
5/8, $90
EU(wk)=$28.125
[w]
$0
5/8, $10
1, $50
Buyer
’
s value
P(trade), E(price if trade)
Figure 5. An interim incentive-efficient mechansim.
(y=0, q=1/2)
Seller
’
s value
$20 [s]
$100 [w]
EU(str)=$5
[s]
$80
0, *
1/2, $100
EU(wk)=$25
[w]
$0
1/2, $0
1, $50
Buyer
’
s value
P(trade), E(price if trade)
Figure 6. An interim incentive-efficient mechanism (generalized Nash bargaining solu-
tion).
3.4 Dishonest equilibria and the revelation principle
The above analysis assumes that the mediator should get information by pro-
viding incentives for the traders to be honest. But even for mechanisms that
are not incentive-compatible, the traders may convey information by rational
strategies that form an equilibrium of their reporting game. For example,
consider again the simple split-the-difference plan from Figure 2. We saw that,
if the buyer were expected to be honest, then the seller would prefer to always
claim to be strong. But if the seller is expected to always claim “strong” in this
mediation plan, then the buyer would prefer to be honest, because an hon-
est admission of weakness could at least get 100–90 = $10 profit for the weak
buyer. Thus, the simple split-the-difference mediation plan has a reporting
equilibrium where the seller always reports strong but the buyer is honest.
With this equilibrium of this mediation plan, the outcome of the mediation
will depend on the traders’ actual types according to table shown in Figure 7.
Seller
’
s value
$20 [s]
$100 [w]
[s]
$80
0, *
1, $90
[w]
$0
0, *
1, $90
Buyer
’
s value
P(trade), E(price if trade)
Figure 7. Incentive-compatible mechanism equivalent to an equilibrium of split-the-
difference.
The plan in Figure 7 is itself an incentive-compatible mediation plan. In ef-
fect, the plan disregards the seller’s reported type and has trade occurring at