Not logged in.

Contribution Details

Type Bachelor's Thesis
Scope Discipline-based scholarship
Title Shapley-Based Core-Selecting Payment Rules
Organization Unit
  • Raphael Haemmerli
  • Sven Seuken
  • English
Institution University of Zurich
Faculty Faculty of Business, Economics and Informatics
Date 2022
Abstract Text Recent research has found Shapley-based core-selecting payment rules to perform well in combinatorial auctions, but we still lack a satisfactory explanation for their good performance. We investigate the hypothesis that low local manipulability can explain the good performance of Shapley-nearest, one Shapley-based payment rule. To this end, we define a local manipulability metric that measures the expected influence bidders can exert on their own payments. In order to measure the local manipulability, we develop a method for finding the derivatives of core-selecting payment rules in any domain. We provide analytical and experimental results for the well-known LLG domain and for the L3G domain, a slightly more complex domain we introduce. We show that Shapley-nearest does not have a particularly low local manipulability compared to other core-selecting payment rules. Further, we show that local manipulability generally is not a good predictor for the performance of payment rules. Lastly, we introduce a novel core- selecting payment rule that we call SVCG-nearest. It is a hybrid between Shapley-nearest and Quadratic, the payment rule used most commonly in practice. We find SVCG-nearest to have good performance and show that it has very low local manipulability.
PDF File Download
Export BibTeX