Online Beauty Contest Game

How to cite

Engelhardt, Robin. 2019. "Online Beauty Contest Game." AEA RCT Registry. AEARCTR-0003925. https://doi.org/10.1257/rct.3925-1.0

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

Robin Engelhardt*

* Primary investigator

Registry ID
AEARCTR-0003925
DOI
https://doi.org/10.1257/rct.3925-1.0
First registered
Last updated
Status
Completed
Trial dates
2018-01-18 – 2018-01-31
Countries
Denmark (Copenhagen)
JEL codes
Noncooperative Games, Repeated Games, Cooperative Games
Registry record
http://www.socialscienceregistry.org/trials/3925

Abstract

We examine the process of reasoning among subjects playing variants of the iterated beauty contest game on Amazon Mechanical Turk. Our findings are substantially different from the behaviors observed in laboratory and newspaper experiments reported in the literature. In general, we do not find any strong evidence in favor of higher-order reasoning in the first round as well as in the following iterations of the game, which puts into question the presence of any strategic thinking among the vast majority of subjects. In addition, our results suggest that players are only weakly responsive to competitive as well as to cooperative incentives.

Experimental design

Workers on Amazon Mechanical Turk accept our HIT (human intelligence task), provide informed consent, and wait until there are enough workers who have accepted the HIT to form random groups (grouped by arrival) of size 2, 4 or 8, respectively, depending on the treatment condition. When a group has been formed, instructions are displayed for 90 seconds. After pressing NEXT, turkers see a page where they have to enter into a form field an integer number between 0 and 100. When all turkers in a group have done so, a result page is displayed, where they can see their own guess, the guesses of the other players, the average and the 2/3 of the average as well as information about whether they have won a bonus in the current found and what their total payoff is for the time being. After this, the previous steps are repeated for a total of 8 rounds. Every time turkers enter a new number, they can see a list of the 2/3 of the average of the previous rounds. Turkers have 90 seconds to think about a number. After eight rounds, turkers are required to give feedback by answering the question: What strategy did you use while playing this game?, after which they are thanked for their participation.

Design details

Randomization method
Randomization of group formation by arrival time.
Randomization unit
Four treatments: 1: split, 2: low, 3: high, 4: control. Each treatment has randomized groups of 2, 4 or 8 players, respectively.
Clusters
1 platform: AMT
Observations
1454 workers on AMT
Arms
86 groups control-treatment, 104 groups split-treatment, 106 groups low-treatment, 107 groups high-treatment.
Primary outcomes
Investigate - the effect of group size, incentives and nudging on the process of reasoning in an experimental iterated beauty contest game among players recruited from Amazon Mechanical Turk. - the presence of higher-order reasoning and - the effects of competitive and cooperative incentives.

Cited by

No citation data available for this registration yet.

Keywords