25.12.2013 Views

Job Market Paper - Personal Web Pages - University of Chicago

Job Market Paper - Personal Web Pages - University of Chicago

Job Market Paper - Personal Web Pages - University of Chicago

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Kim: Endogenous Choice <strong>of</strong> a Mediator<br />

g(σ 1,1 (w|w)) ∈ (0, 1) if σ 1,1 (w|w) = 1 2 , and g(σ 1,1(w|w)) = 0 if σ 1,1 (w|w) < 1 2 .27 Then, player 1<br />

(the vetoer) <strong>of</strong> a weak type would expect to get the pay<strong>of</strong>f <strong>of</strong> zero if he disagrees, and G µ 5/17,0<br />

is<br />

played. But he would also get the pay<strong>of</strong>f <strong>of</strong> zero if he had voted for the alternative µ 1,0 . Under<br />

this belief ¯q 1,1 (w) = 1 and equilibrium strategies, a strong player 2 does have an incentive to not<br />

veto, satisfying (T7). For the alternative mediator µ 1,0 , after a veto on µ 1,0 , the players’ beliefs are<br />

restricted to a credible vote belief that satisfies (ii) <strong>of</strong> Definition 2. Lastly, µ 1,0 satisfies (T3) for<br />

both types <strong>of</strong> player 1:<br />

U 1 (µ 1,0 |w) = 3.6 ≥ 0.85 · (1 − ψ ′ 1,1(w)) · ( ∑ r∈R<br />

σ 1 (r|ww) ∑ d∈D<br />

µ 5/17,0 (d|r 2 , ˆr 1 )u 1 (d, ww)),<br />

where the right-hand side is always less than or equal to 3.6 for any ψ ′ 1,1 (w) ∈ [0, 1] and for any<br />

ˆr 1 ∈ R 1 , and<br />

U 1 (µ 1,0 |s) = 0.6 ≥ 0.85 · (1 − ψ ′ 1,1(s)) · ( ∑ r∈R<br />

σ 1 (r|t) ∑ d∈D<br />

µ 5/17,0 (d|r 2 , ˆr 1 )u 1 (d, sw)),<br />

where the right-hand side is always less than or equal to 0 for any ψ ′ 1,1 (s) ∈ [0, 1] and for any<br />

ˆr 1 ∈ R 1 because u 1 (d, sw) ≤ 0 for any d.<br />

When the status quo is G µ 5/17,0, there is an equilibrium ratification <strong>of</strong> µ 1,0 because (ii) <strong>of</strong><br />

Definition 2 is satisfied. Thus, µ 1,0 is a ratifiable mediator against µ 5/17,0 even though it is not<br />

individually rational for the weak type with passive updating. That is, a weak type cannot veto<br />

the alternative without giving away his identity; and if he gives away his identity, then there is<br />

a disadvantage to vetoing. Because the mediator µ 5/17,0 is not immune to the ratification <strong>of</strong> an<br />

alternative, the unique ex ante incentive efficient mediator µ 5/17,0 when ¯p(s) = 0.15 is not threatsecure.<br />

The reasoning behind this example is straightforward. A weak type who would receive more <strong>of</strong><br />

a peace pay<strong>of</strong>f in the status quo finds a switch to the alternative µ 1,0 unpr<strong>of</strong>itable, which puts a<br />

higher probability on war, and therefore the weak type gets a lower pay<strong>of</strong>f. Such a player would<br />

veto the alternative µ 1,0 in favor <strong>of</strong> the status quo if the status quo is played with the prior beliefs.<br />

But the ratifier can credibly infer that the vetoer is a weak type as a result <strong>of</strong> the veto. Also,<br />

conditional on the weak type vetoing, a strong type has the advantage <strong>of</strong> not vetoing.<br />

So, the<br />

vetoer can credibly infer that the ratifier is a strong type. Such inferences from a vote outcome<br />

27 For a given vote belief ¯q·,i, there is more than one equilibrium to G µ 5/17,0<br />

under the ¯q·,i-revised beliefs. However,<br />

more than one equilibrium does not present a problem.<br />

42

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!