Decentralized Matching with Transfers: Experimental and Noncooperative Analyses

How to cite

WU, JIABIN; Zhang, Hanzhe; He, Simin; Zhu, Xun. 2020. "Decentralized Matching with Transfers: Experimental and Noncooperative Analyses." AEA RCT Registry. AEARCTR-0005488. https://doi.org/10.1257/rct.5488-1.2

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

JIABIN WU* , Hanzhe Zhang , Simin He , Xun Zhu

* Primary investigator

Registry ID
AEARCTR-0005488
DOI
https://doi.org/10.1257/rct.5488-1.2
First registered
Last updated
Status
Completed
Trial dates
2019-05-14 – 2021-06-30
Countries
China (Shanghai)
JEL codes
C71, C72, C78, C92
Registry record
http://www.socialscienceregistry.org/trials/5488

Abstract

We conduct one of the first laboratory experiments and noncooperative analyses of the decentralized matching market with transfers (Koopmans and Beckmann, 1957; Shapley and Shubik, 1972; Becker, 1973). Some theoretical predictions align with but some differ from experimental evidence. Stable matching, which coincides with efficient matching in this setting, is the most frequent outcome. Theoretically neither factor should matter, but experimentally whether equal split is in the core and whether efficient matching is assortative determine the rate of matching, efficient matching, and surplus achieved. We also study the bargaining process and categorize the reasons why participants end up unmatched. Finally and most interestingly, for the matched ones, experimental payoffs coincide with the equilibrium payoffs of a multiplayer extension of Rubinstein (1982) bargaining model, providing the cooperative game with a noncooperative foundation.

Intervention

In our matching experiment, the main intervention is the features of the surplus configurations. In total we have four different configurations. The configurations differ in two ways: the “assortativity level” and “whether equal split is in the core”. Surplus configuration is either assortative or non-assortative, and equal split is either in the core or not. This gives in total four (two by two) different types of configurations. We call them positive assortative (assortative, equal split in the core), negative assortative (assortative, equal split not in the core), mixed equal (non-assortative, equal split in the core) and mixed nonequal (non-assortative, equal split not in the core), respectively.

The secondary intervention is the balance of the matching market. We have in total two balance conditions: The balanced one has 3 subjects on each side of the market; the non-balanced one has 3 versus 4 subjects on each side of the market. For both the balanced and non-balanced markets, we perform the main intervention within the balance type.

Experimental design

In our experimental design, we use four different surplus configurations. Each surplus configuration represents a different matching market. We vary the configurations in two dimensions: (1) whether the stable matching pattern is assortative, and (2) whether equal split is in the core. We also design the surpluses in a way that the maximum total surplus that all agents can obtain is 200, the average total surplus that all agents can obtain is 180 if they are all matched randomly, and the minimum total surplus they can obtain is 160, which is constant across all four markets. Maximum total surplus is obtained only under stable matching. Hence, the rate of stable matching serves as a measure of efficiency.

We employ a within-subject treatment design. All the subjects play the four different matching markets (surplus configurations), but they play them in different orders. According to the Latin square method, we have in total four different treatment, which differ in the order of markets played by the subjects. The four treatments orders are described below.

1. Positive assortative, negative assortative, mixed equal, mixed nonequal.
2. mixed nonequal, positive assortative, negative assortative, mixed equal.
3. Mixed equal, mixed nonequal, positive assortative, negative assortative.
4. Negative assortative, mixed equal, mixed nonequal, positive assortative.

Design details

Randomization method
Within each experiment session, multiple treatment orders are implemented by randomization; the randomization is pre-determined. Subjects who sign up for the experiment receive a seat number randomly before entering the laboratory; the seat number determines the order of the four payoff configurations subjects will experience.
Randomization unit
Individual-level randomization
Clusters
We are aiming to collect about 36 subjects in each of the 4 treatment orders for the balanced markets, and 42 subjects for each of the 4 treatment orders for the non-balanced markets.
Observations
About 300 individuals, recruited via the subject pool of the Economic Lab of the Shanghai University of Finance and Economics.
Arms
6-7 matching markets (36-42 subjects) per treatment arm.
Primary outcomes
There are three main outcomes variables: 1. At market level, the rate of matching; 2. At market level, the rate of stable matching; 3. At market level, the rate of surplus achieved. 4. At individual level, the payoff gained if matched.
Secondary outcomes
1. Number of proposals 2. Detailed information of all the proposals by individual and matching group level, including the time a proposal is made/rejected/accepted; the proposed division of surplus of each proposal; and whether it is rejected/accepted by which player.

Cited by

No citation data available for this registration yet.

Keywords