Mixed Strategies and Context: Evidence from a lab experiment

How to cite

Ahmad, Husnain; Banuri, Sheheryar. 2020. "Mixed Strategies and Context: Evidence from a lab experiment." AEA RCT Registry. AEARCTR-0005550. https://doi.org/10.1257/rct.5550-1.3000000000000003

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

Husnain Ahmad* , Sheheryar Banuri

* Primary investigator

Registry ID
AEARCTR-0005550
DOI
https://doi.org/10.1257/rct.5550-1.3000000000000003
First registered
Last updated
Status
On going
Trial dates
2020-03-12 – 2021-11-30
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/5550

Abstract

Due to COVID 19, the lab experiment needed to be halted part way through. We resumed data collection in Nov 2021.

We study the effect of a game’s contextual framing on how subjects play the game in a lab. Subjects play two strategically equivalent games across treatments, namely; hide and seek, and penalty kicks. The games have identical unique Nash equilibria in mixed strategies, with the only difference between them being the context in which the game is presented. We test whether subjects play is closer to theoretical prediction in the penalty kick game, which is a more familiar context for subjects and is closer to games played in the field. Data from past experiments show that while subjects play in accordance with theory in the field, they fail to do so in the lab. Our experiment attempts to determine whether this discrepancy is due to the difference in context between lab and field games. It can also be seen as a test of the underlining assumption of strategic equivalence in games with unique mixed strategy equilibria.

Intervention

Variation in the level of context in a game.

Full context involves a video game like PK environment, partial context involves descriptions drawn from football, and no context uses de-contextualised hide and seek.

Experimental design

The context of the game will be varied with three levels: no context, written context and visual context.

Design details

Randomization method
In lab with lottery.
Randomization unit
Individual randomisation for treatment. Within treatment random matching of players into pairs by computer.
Clusters
n/a
Observations
150 choice of pure strategies per individual. Additionally 4 computer "frequencies" (or mixed strategies) per individual.
Arms
Targeted sample size of 200 individuals (100 pairs). EDIT: Dure to COVID related closure of our lab mid-way, we have increased the sample size. 88 (44 pairs) - Pre-COVID. 212 (106 pairs) - Post-COVID As the experiment remains unchanged, we expect to pool the sample for a total size of 300 (150 pairs, 50 in each treatment).
Primary outcomes
- Player choices in each round. - Player set "frequencies" for computer to play multiple rounds on their behalf.

Cited by

No citation data available for this registration yet.

Keywords