Coalitions, incentives, and risk sharing (Q1308816)

The paper investigates the possibility that in a principal agent problem it may be preferable, from the standpoint of the principal, to have cooperative rather than non-cooperative behaviour on the part of the agents. \(P_{ij}(e_ 1,e_ 2)\) denotes the joint probability of the (stochastic) output pair \((x^ 1_ i,x^ 2_ j)\) being realized when effort pair \((e_ 1,e_ 2)\), a point in a subset of \(\mathbb{R}^ n\times \mathbb{R}^ n\), is chosen by the agents \(n=1,2\); \(w^ n_{ij}=\) payment to agent \(n\) when output pair \((x^ 1_ i,x^ 2_ j)\) is realized; \(V_ n(w^ n_{ij})- G_ n(e_ n)\) is the additive separable utility function of agent \(n\); \(\phi_ n\) is \(V^{-1}_ n\); \(v^ n_{ij}\equiv V_ n(w^ n_{ij})\) and \(\bar u^ n\) is agent \(n\)'s reservation utility level. The non-cooperative version is: (I) \(\text{Min}_{\nu^ 1,\nu^ 2} \sum_ i \sum_ j P_{ij}(e_ 1,e_ 2)\Bigl[\sum_ n \Phi^ n(\nu^ n_{ij})\Bigr]\equiv\theta\) (say) s.t. (1) \(e_ 1\in\arg_ e\max \sum_ i\sum_ j P_{ij}(e,e_ 2)\nu^ 1_{ij}- G_ 1(e)\), and similarly for agent 2, and (2) \(\sum_ i \sum_ j P_{ij}(.)\nu^ n_{ij}- G_ n(e_ n)\geq \bar u^ n\) for each \(n\). The cooperative version of the problem is: (II) \(\text{Min }\theta\) (as above) s.t. (3) \((e_ 1,e_ 2)\in\text{arc max}_{e_ 1',e_ 2'} \sum_ i\sum_ j P_{ij}(e_ 1',e_ 2') \Bigl(\sum_ n \nu^ n_{ij}\Bigr)- \sum_ n G_ n(e_ n')\), (4) \(\sum_ i \sum_ j P_{ij}(e_ 1,e_ 2)\Bigl(\sum_ n \nu^ n_{ij}\Bigr)- \sum_ n G_ n(e_ n)\geq \sum_ n \bar u^ n\); (5) \(v_ 1'(w^ 1_{ij})= v_ 2'(w^ 2_{ij})\) for all \(i\), \(j\). (1) says that agent choice is a Nash equilibrium while (3) says that agent choice is the collusion solution presuming transferable utility. (2) and (4) ensure that the contract is (weakly) preferred to no contract. (5) ensures that optimal risk sharing between agents is attained, so the only reason for side contracts is to enforce coordinated effort choice which is the focus of attention. It is shown (Theorem 2) that under appropriate assumptions, for each \(e_ 1\), \(e_ 2\), \(\theta\) is smaller in (II) than in (I), i.e. the principal can implement any effort pair with less costs under agent side contracting than under no side contract. Extensions to non-transferable utilities and team production (individual outputs are non-observable, only joint output is) are provided.