Pj(aj; a∙,αk,ak) = Pr^
hjmj + hε (ηj + aj + εj -
aj) > hkmk + hε(ηk + εk) 1
hk + hε J
hj + hε
= Pr {hj +ε hε (aj j >
hkmk + hε(ηk + εk) hjmj + hε(ηj + εj
hk + hε
hj + hε
).
Define the random variable
ζj ≡
hkmk + hε(ηk + εk)
hjmj + hε(ηj + εj)
hk + hε
hj + hε
Our independence and normality assumptions imply that the prior distribution of ζj is
normal with mean
zj ≡ mk
mj
(3)
and variance13
2 ≡ ( hε V ( 1 -1 ^ ( hε V ( 1 1 ʌ (4)
∖ hk + hε / ∖ hk hε∕ ∖hj + hε / ∖ hj hε /
We denote this distribution by ψj (∙) with c.d.f. φj∙ (∙). The probability of j’s nomination
given ak = ajk is then
Pj(aj;aj, ak, ak) = Pr {c∙ < h + h (aj- aj) J = φj ɑ + h (aj - ajf)
(5)
The marginal impact of aj∙ on j's expected payoff given ak = a*k is
∂Πj (aj ; aj ,ak ,ak) h hε ∖
—∂a— = φj hι+τ-ha (aj - aj) )
τ^zw Wj
hj + hε
- c (ai) .
Note that if c' (0) ≥ ψj (0) hh+hε Wj, then aj = 0.14
The first-order conditions for an equilibrium with aj1 , a2j > 0 are
c' j ■ (0) h1hWW1,
(6)
(7)
c' (aj) = φ2 (0) rh-i-W2.
h2 + hε
Note that each of these conditions is a function of only one of the effort levels: in equi-
librium the principal correctly anticipates effort levels, which implies that given k indeed
chooses akj , j ’s nomination probability is independent of the level of ajk. Our assumptions
13 Since the prior distributions of ζ1 and ζ2 have the same variance, we can simply denote this variance
by σ2 , not using any subscript.
14 Since aj is j’s effort additional to the effort he would have exerted in the absence of the nomination
contest, it may well be that c' (0) > 0.