A new mechanism for the free-rider problem

Sujay Sanghavi, Bruce Hajek

Research output: Contribution to journalArticlepeer-review

Abstract

The free-rider problem arises in the provisioning of public resources, when users of the resource have to contribute towards the cost of production. Selfish users may have a tendency to misrepresent preferences-so as to reduce individual contributions-leading to inefficient levels of production of the resource. Groves and Loeb formulated a classic model capturing this problem, and proposed (what later came to be known as) the VCG mechanism as a solution. However, in the presence of heterogeneous users and communication constraints, or in decentralized settings, implementing this mechanism places an unrealistic communication burden. In this paper, we propose a class of alternative mechanisms for the same problem as considered by Groves and Loeb, but with the added constraint of severely limited communication between users and the provisioning authority. When these mechanisms are used, efficient production is ensured as a Nash equilibrium outcome, for a broad class of users. Furthermore, a natural bid update strategy is shown to globally converge to efficient Nash equilibria. Also, upper bounds are provided for the revenue that can be generated by any individually rational mechanism that ensures efficient production at any Nash equilibrium. It is shown that there exist mechanisms in our class that achieve each of the bounds. An extension to multiple public goods with interrelated valuations is also presented.

Original languageEnglish (US)
Pages (from-to)1176-1183
Number of pages8
JournalIEEE Transactions on Automatic Control
Volume53
Issue number5
DOIs
StatePublished - 2008

Keywords

  • Convex optimization
  • Free-riders
  • Game theory
  • Public goods

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Computer Science Applications
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'A new mechanism for the free-rider problem'. Together they form a unique fingerprint.

Cite this