Clearing Markets via Bundles
- Michal Feldman ,
- Brendan Lucier
Symposium on Algorithmic Game Theory 2014 |
We study algorithms for combinatorial market design problems, where a set of heterogeneous and indivisible objects are priced and sold to potential buyers subject to equilibrium constraints. Extending the CWE notion introduced by Feldman et al. [STOC 2013], we introduce the concept of a Market-Clearing Combinatorial Walrasian Equilibium (MC-CWE) as a natural relaxation of the classical Walrasian equilibrium (WE) solution concept. The only difference between a MC-CWE and a WE is the ability for the seller to bundle the items prior to sale. This innocuous and natural bundling operation imposes a plethora of algorithmic and economic challenges and opportunities. Unlike WE, which is guaranteed to exist only for (gross) substitutes valuations, a MC-CWE always exists. The main algorithmic challenge, therefore, is to design computationally efficient mechanisms that generate MC-CWE outcomes that approximately maximize social welfare. For a variety of valuation classes encompassing substitutes and complements (including super-additive, single-minded and budget-additive valuations), we design polynomial-time MC-CWE mechanisms that provide tight welfare approximation results.