and if ti = T, i gets
f-2+τ 1
I (v — x) - exp
Jo c
(-X)
dx + exp
(1—t) (v
V — C.
Thus i is indifferent between all ti E (O, —c/2 + T] U {T}, and Fi and Fj are mutually
best responses.
A.3 Proof of Lemma 3
Parts a), b(i) and c) follow directly from the analysis in the main text. It remains
to prove part b(ii). If there is an equilibrium in mixed strategies, the equilibrium
strategies must exhibit similar properties as in the case of complete information. In
particular, for waiting times ti and tjL in the support of the mixed strategies, it has
to hold that
Fi (Ji) = 1 — (1 — bi) eχp
(C;)
(10)
FjL (tjL ) = ~ 1 — (1 — Pl⅛l )exp f—-jτ^ , (11)
pL L ∖ c /_
where the constants bi and bjL correspond to the mass points at zero, Fi (O) and
FjL (O), and remain to be determined. The factor 1/p; in FjL takes into account the
probability p; that i faces a rival with cost cl. It has to hold that O ≤ bi, bjL < 1, and
min (bi, bjL ) = O : if i (j;) concedes immediately with strictly positive probability, jL
(i) strictly prefers a concession in ε > O, ε infinitesimally small, to a concession in O.
Assumption 1 implies that no tjH < T with Fi (t jH ) < 1 can be part of jH’s
equilibrium strategy: jH won’t choose any tjH < T in the support of Fi. In turn,
for any ti < T, we must have FjH (ti) = O, and thus i strictly prefers ti = T to all
ti E (—c/2 + T, T). Moreover, Fi must be continuous on (O, T). To see why, suppose
that i concedes in ti E (O, —c/2 + T] with strictly positive probability. Then, there
are δ > O, ε > O such that j; strictly prefers tj = ti + ε to any tj E (ti — δ, ti), hence
i is strictly better off by choosing ti — δ∕2 instead of ti. Therefore, possible mass
28