Revisiting Information Aggregation in Networked Markets

How to cite

Deck, Cary; Jun, Tae In. 2026. "Revisiting Information Aggregation in Networked Markets." AEA RCT Registry. AEARCTR-0017639. https://doi.org/10.1257/rct.17639-1.0

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

Cary Deck* , Tae In Jun

* Primary investigator

Registry ID
AEARCTR-0017639
DOI
https://doi.org/10.1257/rct.17639-1.0
First registered
Last updated
Status
In development
Trial dates
2026-01-15 – 2026-06-01
Countries
Private
JEL codes
C9, D8, G1, G4
Registry record
http://www.socialscienceregistry.org/trials/17639

Abstract

This study involves a laboratory experiment examining the ability of markets to aggregate information. Prior to trading, traders share information based on a network structure, which is exogenously imposed by the researchers. This project closely follows Halim, Riyanto, and Roy (2019, JoF) - hereafter HRR - who considered 4 network structures. Reexamination of their data indicates that RE is a nominally better predictor of market prices than is PI only in the Circle network (among those they considered). The experiment follows the procedures of HRR and uses a modified version of their software. The main modification is the introduction of another network structure (Mobius Ladder) that determines who traders share information with in their group. Other modifications are minor (e.g. final payments now denoted in US$, running 2 groups at once rather than 3). New data is also collected for a Circle as HRR. No new data is collected for other networks used in HHR.

Intervention

The experiment involves a group of 8 traders who observe public and private information about the value of an asset. Prior to trading, traders share private information based upon an exogenous network structure. Markets are run using two network structures: Circle and Mobius ladder.

The prior information (PI) and rational expectations (RE) models each make predictions about what prices should be observed in the market based on the information held by each trader. The experiment is designed to determine which model's prediction better matches observed data for each network.

Experimental design

The experiment involves two treatments using a between subjects design. Subjects are placed in either a Circle or a Mobius Ladder network that determines how private information about asset value is shared before the trading period begins. Each market group consists of 8 people. Two market groups in the same treatment will be conducted concurrently (i.e. a session will involve 16 people).

The experiment procedures (including directions, survey questions, number of periods, parameter values, etc.) follow those of HRR. The only deviations involve minor adjustments given the change in subject population and where the experiment is being conducted) such as text reflecting that final payoffs are in US$.

Design details

Randomization method
computer
Randomization unit
session (which consists of two groups of 8 subjects in the same treatment). Randomization of sessions to treatment is conditional on equal number of sessions per treatment.
Clusters
16 groups (8 groups for each of 2 treatments)
Observations
128 subjects (8 groups of 8 subjects for each of 2 treatments).
Arms
128 subjects (8 groups of 8 subjects for each of 2 treatments).
Primary outcomes
Analysis will be based on observed market prices. Specifically, the mean absolute deviation between observed prices and the prediction of the RE and PI models will be calculated. Following Plott and Sunder (1988, Ecma) and Corgnet, et al. (2023, Man Sci), a Wilcoxon signed rank-sum test will be used to determine which model, if any, best describes the data. The tests will be conducted separately for each treatment.
Secondary outcomes
Other relevant analysis in HRR will be replicated. A Wilcoxon signed rank-sum test will be conducted on the combined data for the Circle network from the experiment and HRR.

Cited by

No citation data available for this registration yet.

Keywords