= Leader optimization in a sequential private-value all-pay contest
{title2=$b(v)\in\arg\max_{0\le b\le v}\{vF(b)-b\}$}
For a leader with nonnegative value $v$ and a continuous nonnegative follower-value <cumulative distribution function> $F$, a bid $b\ge0$ gives <expected payoff> $vF(b)-b$. Bidding zero guarantees nonnegative <payoff>, whereas a bid $b>v$ gives $vF(b)-b\le v-b<0$. Every <best response> therefore lies in $[0,v]$, and the optimization is
$$
b(v)\in\operatorname*{argmax}_{0\le b\le v}\{vF(b)-b\}.
$$
A fixed upper bound $1$ is justified only under an additional normalization of the follower-value support; it is not a general restriction on effort.
If $F$ is a <concave function>, this <objective function> is concave. At a differentiable interior optimum it satisfies $vF'(b)=1$; one-sided derivatives handle the endpoints. Flat intervals can give several <best responses>, as described in <best-response selection at a flat leader objective>. Every optimal bid also obeys the sharper inequality $b\le vF(b)\le v$, by comparison with zero effort.
Back to article page