We conduct an incentivized lab experiment to test participants' ability to understand the DA matching mechanism and the strategyproofness property, conveyed in different ways. We find that while many participants can (using a novel GUI) learn DA's mechanics and calculate its outcomes, such understanding does not imply understanding of strategyproofness (as measured by specially designed tests). However, a novel menu description of strategyproofness conveys this property significantly better than other treatments. While behavioral effects are small on average, participants with levels of strategyproofness understanding above a certain threshold play the classical dominant strategy at very high rates.
Authors
Yannai A. Gonczarowski, Ori Heffetz, Guy Ishai, Clayton Thomas
- Acknowledgements & Disclosure
- The authors thank Keren-Or Barashi Gortler, Itamar Bellaiche, Yehonatan Caspi, Yael Cohen, Gabriela Cohen-Hadid, Ayala Goldfarb, Michael Khalfin, Ido Leshkowitz, Josef Mccrum, Shenhav Or, Yonatan Rahimi, and Ohad Weschler for excellent research assistance; Eric Budish, Ben Enke, Nicole Immorlica, David Laibson, Markus Mobius, Assaf Romm, Shigehiro Serizawa, Ran Shorrer, Alex Teytelboym, and Leeat Yariv for helpful discussions; participants at the Stanford Institute for Theoretical Economics (SITE) 2023 Experimental Economics, SITE 2023 Market Design, WZB Berlin Matching Workshop, Crown Family Israel Center for Innovation (ICI) 2024 Academic Conference, Virtual Market Design Seminar, 2024 Marketplace Innovations Workshop, 8th Solomon Lew Conference on Behavioral Economics (Tel Aviv), 1st Annual Chicago Booth Market Design Conference, EC 2024, Communicating Clearly to Market Participants Workshop (Stony Brook), and seminar participants at Bar Ilan, Cornell, the Hebrew University, and Microsoft Research for comments that significantly improved the paper; and Adam Chafee and his team at the Cornell Business Simulation Lab for their help with running the experiment. Pre-registration of our experiment can be found at https://aspredicted.org/7eq7e.pdf. A one-page abstract of this paper appeared in the proceedings of EC 2024. The authors gratefully acknowledge research support by the following sources. Gonczarowski: National Science Foundation (NSF-BSF grant No. 2343922), Harvard FAS Inequality in America Initiative, and Harvard FAS Dean’s Competitive Fund for Promising Scholarship. Heffetz: Israel Science Foundation (grant No. 2968/21), US-Israel Binational Science Foundation (NSF-BSF grant No. 2023676), Cornell’s S.C. Johnson School, and Cornell’s Center for Social Sciences. Ishai: Barbara and Morton Mandel Doctoral Program, Bogen Family, and Federmann Center for Rationality. Thomas: NSF CCF-1955205, Wallace Memorial Fellowship in Engineering, and Siebel Scholar award; part of his work was carried out while in Princeton’s Department of Computer Science. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.
- DOI
- https://doi.org/10.3386/w33020
- Pages
- 102
- Published in
- United States of America
Table of Contents
- Introduction 3
- Experimental Design 9
- Setting (All Treatments) 10
- Descriptions (by Treatment) 11
- Training Rounds and Training Score (by Treatment) 14
- Real Rounds and Ranking Behavior (All Treatments) 17
- Strategyproofness-Understanding Test (All Treatments) 18
- Exit Questions and Cognitive Score (All Treatments) 21
- Results 21
- Sample 21
- Main Result #1: We Can Teach DA 22
- Main Results #2 and #3: Understanding DA Is Not Understanding SP, But We Can Teach SP 24
- Overall Measure of Strategyproofness Understanding 25
- Sub-Measures of Strategyproofness Understanding 25
- Main Result #4: Strategyproofness Understanding Predicts Straightforward Behavior 29
- Mean Levels of Straightforward Behavior 29
- Relation Between Strategyproofness Understanding and Straightforward Behavior 30
- Additional Results and Robustness 32
- Variation Across Samples 32
- Robustness to Controls 34
- Other Relations Between Outcome Variables 34
- Related Literature 35
- Conclusion 37
- Additional Results 41
- Sample Collection, Earnings, and Duration 41
- Full Training Score Distribution 42
- Sub-Measures of % SP-U 43
- Ranking Patterns Beyond % SF 43
- Relationship Between % SF and (Sub-Measures of) % SP-U 49
- Prolific vs. Cornell Samples, Cognitive Score, Attention Score, and their Mediating Effects on Main Results 56
- Robustness of Main Results to Adding Controls 60
- Supplemental Information and Analysis 62
- Comparison of Analysis and Findings With Pre-Registration 62
- Demographics 63
- Detailed Performance in Training Questions 69
- Null Training Common to All Participants 69
- DA Mechanics Training 69
- SP Property Training 74
- Null Treatment Training 74
- Cognitive and Attention Scores 75
- Additional Joint Distributions of Outcome Variables 79
- Univariate Regressions and Correlations Between Main Outcome Variables 79
- Relationship Between % SP-U and % SF: More Details 80
- Relationship Between % TR and % SP-U 80
- Relationship Between % TR and % SF 82
- Experiment Materials 84
- Description Screenshots 84
- Scenarios in DA Mechanics Training Rounds 84
- Randomization of Setting Components in Real DA Rounds 86
- Experiment Procedures 96
- Prolific 96
- Cornell Business Simulation Lab (BSL) 99
- Full Screenshots 102
