Please fill out this short user survey of only 3 questions in order to help us improve the site. We appreciate your feedback!
Choices in ambiguous settings
Last registered on May 04, 2018


Trial Information
General Information
Choices in ambiguous settings
Initial registration date
May 03, 2018
Last updated
May 04, 2018 10:43 AM EDT
Primary Investigator
Washington University in St Louis
Other Primary Investigator(s)
PI Affiliation
University of Graz
PI Affiliation
Washington University in St. Louis
Additional Trial Information
In development
Start date
End date
Secondary IDs
In decision theory, people may behave differently when faced with unknown risks compared to known risks. We call unknown risks "ambiguity". One possibility is that a person may be averse to ambiguity and, therefore, prefer to face a known risk, all else equal. A famous experiment proposed by Ellsberg in 1961 provides a concrete experiment in which many people indeed choose a known risk over an unknown risk. In fact, this behavior is not consistent with traditional models of decision making and has motivated a literature that captures such "ambiguity averse" behavior. Our experiments are designed to test (i) to what extent this behavior can be interpreted as aversion to ambiguity in experiments closely related to Ellsberg's experiments, and (ii) to what extent people might behave differently when we describe how and why to avoid the unknown risk through a "hedging" process that involves basing their choice on the outcome of a coin flip.

External Link(s)
Registration Citation
Kuzmics, Christoph, Brian Rogers and Xiannong Zhang. 2018. "Choices in ambiguous settings." AEA RCT Registry. May 04. https://doi.org/10.1257/rct.2580-1.0.
Former Citation
Kuzmics, Christoph et al. 2018. "Choices in ambiguous settings." AEA RCT Registry. May 04. http://www.socialscienceregistry.org/trials/2580/history/29100.
Experimental Details
Lab experiments.
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
The proportion of subjects betting white after YG and GY preliminary draws in (1).
The proportion of subjects with “Ellsbergian” behavior before and after we provide the arguments above in (2).
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
The full profile of preferences in (1).
The proportion of Ellsbergian behavior in (2) after various combinations of observing the arguments and conditional on whether the coin is physically flipped by experimenters or, instead, a mental device employed by the individual subject.
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
In an Ellsberg-inspired environment, we do two things:
1) We elicit preferences over acts in an environment where Ellsbergian behavior is inconsistent with most models of ambiguity aversion.
2) We elicit preferences over acts when subjects have been exposed to Raiffa's hedging argument.

Experimental Design Details
In an Ellsberg-inspired environment, we do two things: 1) We elicit preferences over acts in an environment where Ellsbergian behavior is inconsistent with most models of ambiguity aversion. 2) We elicit preferences over acts when subjects have been exposed to Raiffa's hedging argument. In both cases we use a risky urn and an uncertain urn, implemented as physical boxes filled with 100 ping pong balls each. The risky urn has 49 white balls. The uncertain urn has 100 balls, each of which is either green or yellow. One ball will be drawn from each urn. We identify the available acts with colors; each act generates a binary lottery that "wins" if the corresponding color is drawn. The acts are (white, green, yellow). In (1), we do the following. Before eliciting preferences over (white, green yellow), we draw and display two balls, with replacement, from the uncertain urn. Before those draws, we elicit preferences using the strategy method for each of the four possible combinations of preliminary draws. Any profile that bets on white following preliminary draws of GY and YG is state-wise dominated by another act and is thus inconsistent with most models of ambiguity aversion. In (2) we describe a coin toss, after which a subject can bet green if the coin lands heads and yellow if the coin lands tails. This mixed act generates a state-independent 50% chance of winning and so dominates the risky urn (betting white). We present, through videos, either one or two arguments, where one argument describes the hedge as dominating a choice for white, and the other argument observes that conditional on either heads or tails, the resulting lottery is still ambiguous. We have in total 10 treatment groups, including the control group. The treatment groups are designed to test the following hypothesis: (1) Subjects in 4 groups are asked to place their bets without being exposed to any arguments. In the other 6 groups, subjects are asked to choose after they watch the videos explaining the arguments. (2) In the 4 groups in which subjects are asked to place their bets before observing any arguments, we will expose them to the arguments in a questionnaire and ask for their “advice for future participants”. We will be able to directly observe any changes within-subject. (3) Subjects in 4 groups are asked to choose from the 3 options described above. In the other 6 groups, subjects have one more option: coin flip for green/yellow ball. If they choose this option, their bet will depend on a coin flip conduct by experimenters. (4) We have two symmetric videos, one arguing in favor of hedging, and one arguing against hedging. In some treatments we only show the subjects one of the videos, and in some treatments we show them both videos in a different order.
Randomization Method
All subjects will be recruited from the lab's registry. Subjects are randomly allocated to one of the treatments via the experimental session they sign up for.
Randomization Unit
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
Sample size: planned number of observations
350 undergraduate students
Sample size (or number of clusters) by treatment arms
11 treatments, with the following sample sizes:

1) Observe preliminary draws: 30

For the remaining, the coding is as follows:
A=exposed to arguments before incentivized decision
N=not exposed to arguments before incentivized decision
"3"=coin flip by experimenter is not an option; the only options are white/green/yellow
"4"=a fourth option to bet according to an experimenter-implemented coin flip is available
V0=video describing virtual coin toss to implement lottery between green/yellow bets
V1=video with argument in favor of using coin toss
V2=video with argument against using coin toss

2) A3 V0,V1 sample size=10
3) A3 V0,V1,V2 sample size=10
4) A4 V1 sample size=10
5) A4 V2 sample size=10
6) A4 V1,V2 sample size=15
7) A4 V2,V1 sample size=15
8) N3 V0,V1 sample size=20
9) N3 V0,V1,V2 sample size=20
10) N4 V1,V2 sample size=10
11) N4 V2,V1 sample size=10
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Supporting Documents and Materials

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
IRB Name
The Washington University in St. Louis Institutional Review Board
IRB Approval Date
IRB Approval Number
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports, Papers & Other Materials
Relevant Paper(s)