Does the nature of experience affect choice?

How to cite

Diekert, Florian; Goeschl, Timo; König, Christian. 2018. "Does the nature of experience affect choice?." AEA RCT Registry. AEARCTR-0003355. https://doi.org/10.1257/rct.3355-5.0

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

Florian Diekert* , Timo Goeschl , Christian König

* Primary investigator

Registry ID
AEARCTR-0003355
DOI
https://doi.org/10.1257/rct.3355-5.0
First registered
Last updated
Status
Completed
Trial dates
2018-09-27 – 2019-09-27
Countries
United States of America ()
JEL codes
C92, D83, Q54
Registry record
http://www.socialscienceregistry.org/trials/3355

Abstract

Both strategic and natural uncertainty (risk) determine whether agents receive an economic reward in many environments. Moreover, many environments require repeated decisions. When periods are stochastically independent, do agents respond to an adverse outcome? And does it matter for their response whether the strategic or the natural risk materialized to cause the event?

We design an experimental choice situation that contains both strategic and natural uncertainty and repeat the situation once, after feedback on first-round outcomes. Observing the behavior of experimental subjects in both rounds allows us to investigate whether the experience of a zero-payoff outcome in the first choice situation affects participant's behavior in a second choice situation and whether the source of the zero-payoff matters. Payoffs are determined by joint play between the subject and a co-player in a game of chicken and by the outcome of a lottery draw. While the rounds are independent in statistical (random rematching) and payoff terms (random determination of payoff-relevant round), casual observation and evidence from cognate experiments suggests that choice behavior is likely to differ depending on whether the zero-payoff event materialized or not. This behavioral response conceivably also differs depending on whether natural or strategic uncertainty is seen as the reason for the zero-payoff event. Our experiment is designed to detect the presence and nature of such behavioral changes. Furthermore, we ask whether age, gender, or educational status can explain observed choices and changes in choices. Finally, we study whether a change in the perception of what caused the outcome mediates the effect of experience on the change in choices.

Intervention

Participants are recruited through an online platform and are asked to make two payoff relevant choices as well as answer questions about their beliefs on other participant's play as well as on the outcome of an urn draw. Participants are either matched with a computer co-player or a human co-player. In the latter case, they are randomly re-matched with another co-player for their second choice.

Experimental design

We match participants with a co-player to play a game of chicken. The eventual payoff of the players depends on their play in the game and on the draw of a lottery. That is, when a "red ball" is drawn in the lottery, or when both players choose "action B" in the chicken game, they receive a payoff of zero. When a participant chooses "action A" in the chicken game, and a "green ball" is drawn in the lottery, she receives a payoff of 1 USD and when a participant chooses "action B" in the chicken game, while her co-player chooses "A", and a "green ball" is drawn in the lottery, she receives a payoff of 3 USD.

We play two rounds of this game and randomly rematch players (perfect stranger matching). After the first round, full information about the outcome of the lottery and the choice of the matched co-player is provided. Participants receiving a payoff of zero will therefore know exactly whether the event can be attributed to the unfortunate realization of the natural uncertainty, or the strategic uncertainty, or both. In round two, participants are tasked to again choose between A and B. Finally, we ask participants not only to make payoff-relevant choices, we also elicit their perceptions about the cause of the outcome. We do so at two points in time: first, after they have made their first choice but before they know the outcome, and second, after they have learned the outcome from the first round and before they make their second choice.

Design details

Randomization method
Computer
Randomization unit
Individual
Clusters
900 observations per treatment (human condition or robot condition), so 1800 in total.
Observations
1800
Arms
900
Primary outcomes
Do participants change their choice between the first and the second round?

Cited by

No citation data available for this registration yet.

Keywords