Optimum/near-optimum incentive policies for stochastic decision problems involving parametric uncertainty (Q1064303)

This paper deals with a stochastic two-person, nonzero-sum game. One player, called the leader, has access to the decision of the other player, called the follower. Such an information scheme is often referred to as an (inverse) Stackelberg problem or an incentive problem. With respect to the uncertainties it is assumed that there is a ''state of nature'', which is a parameter in both cost functions and about which both players have imperfect information. It is assumed though, that the follower's information regarding the state of nature is nested into that of the leader (the leader knows more). Apart from this uncertainty, the leader has incomplete knowledge of (a parameter in) the follower's cost function. Under various conditions, convexity of the follower's cost is an essential one, nonunique linear incentive schemes for the leader exist, such as to obtain the most desirable outcome of the game. This nonuniquess, due to the fact that the leader knows more about the state of nature, is exploited by the leader in order to make the follower's optimal reaction minimally sensitive to variations in the unknown (to the leader) parameter in the follower's cost. The theory has been elucidated explicitly for the scalar problem (all appropriate spaces are one-dimensional). The extension to the vector case is indicated. The paper concludes with an example related to an incentive design problem in a divisionalized firm.