I use an online experiment to study how people make complex decisions and how information can both change beliefs but also distort attention.
Experimental design
Participants repeatedly choose whether to give up money for a complex "good" whose total value is difficult to assess. Information provided during the experiment can both change their beliefs about the value of some of these attributes or simply draw more attention to them.
Design details
Randomization method
Randomization is done within Qualtrics.
Randomization unit
Participants are randomly assigned to different treatments. The value of attributes for each choice are randomly generated within each survey (and so differ across participants).
Clusters
800 Individuals
Observations
800
Arms
For each participant, two attributes of Option B are randomly chosen within the survey code. Call these the "target" and "alternative" attributes. There are 80 total decisions, which are divided into 4 intervals of 20 choices each. In each interval (except the first), participants can be informed about the value of at most one attribute.
Experiment 1 has 5 treatments:
Treatment 1 (N = 120): No information in intervals 2 and 3.
Treatment 2 (N = 120): Informed about the target attribute in interval 2, nothing in interval 3.
Treatment 3 (N = 120): Informed about the alternative attribute in interval 2, nothing in interval 3.
Treatment 4 (N = 120): Informed about the target attribute in interval 2 and the target attribute in interval 3
Treatment 5 (N = 120): Informed about the target attribute in interval 2 and the alternative attribute in interval 3
Independently, everyone is randomized in interval 4 to see no information, a reminder that Option A comes with a lottery, or a reminder about the lottery but with additional information about the objective odds of winning it. Also independently, one attribute (other than the target and alternative attribute) is "frozen" at its starting value for the whole experiment.
Experiment 2 is similar to Experiment 1, except for three things. First, there is no lottery associated with Option A. Second, no attribute is frozen. Third, the value of Option A is chosen such that the target attribute is always decisive in determining whether Option A or Option B pays more. Within this experiment, participants are randomly and independently informed (in each of intervals 2, 3, and 4) about nothing, the value of the target attribute, or the value of the alternative attribute. We plan to recruit N = 200 people for experiment 2.
Primary outcomes
The primary outcome is how much attention participants pay to each dimension of the problem, depending on whether they have been or are being informed of it.
Secondary outcomes
We will also look at whether participants chose the higher-value option.
Cited by
No citation data available for this registration yet.