Optimal compensation schemes in organizations with interpersonal networks

I propose a mathematically tractable model to study the optimal compensation scheme in organizations with arbitrary interpersonal networks, in which agents connected in a network help neighbors and work collectively to produce a team's output, and payment is an equity share linked to the position in the network. The optimal compensation is shaped by: (a) ``incentivizing the peripheral'' mechanism: agents with a smaller (degree) centrality should be paid higher since they receive relatively less help; and (b) ``incentivizing the central'' mechanism: agents with a larger centrality should be paid higher to help others more. Due to these two conflicting mechanisms, the relationship between centrality and optimal pay may not be monotonic. When the externality of help is not pronounced, all agents should enjoy the same pay, as in a flat organization. I provide a necessary condition for monotonically increasing relationships and rule out monotonically decreasing relationships. I verify the theoretical predictions using numerical and empirical analyses.