NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Decision-making Procedures, Rational Choice and Welfare
Initial registration date
January 15, 2020
May 29, 2020 9:05 PM EDT
Universidade Pompeu Fabra
Other Primary Investigator(s)
Additional Trial Information
This paper studies the impact of decision-making procedures on individuals’ rationality and welfare. We theoretically and experimentally study a specific decision-making procedure, sequential elimination in which a decision-maker has to sequentially eliminate options that are not preferred in the decision problem, one by one, until only one option remains. First, we provide a necessary and sufficient condition that sequential elimination results in consistent choice while direct decision-making may not. Then we experiment to examine the treatment effect of sequential elimination on the choice consistency by comparing choice data in three treatment groups: sequential examination, sequential elimination and procedure selection. We also have an incentive-compatible design in which participants can revise their choice to test decision makers' awareness of "mistake" in making choices. We also ask the participants' satisfaction with their choices and decision-making processes as criteria for welfare. We hypothesize that decision makers who use sequential elimination have a higher level of consistency, less likely to change their choices and are more satisfied with their decisions. We control for preference for consistency, preference for consequentialism and sunk cost effects,
Intervention Start Date
Intervention End Date
Primary Outcomes (end points)
The key outcome variables in this experiment are consistency scores measured by the choice data, as indicators of rationality. They include Generalized Axiom of Revealed Preference (GARP) violations (Varian 1982), Houtman-Maks' index (Houtman and Maks, 1985), Afriat’s (1967, 1972) critical cost efficiency index (CCEI) and swaps index (Apesteguia and Ballester, 2015). We will compare all consistency scores of individuals under different treatments.
Primary Outcomes (explanation)
We follow the literature to construct our outcome variables with choice data. The GARP violations index counts how many combinations of choices violate GARP. The Houtman-Maks' index finds the minimal subset of observations that needs to be removed from the data in order to make the remainder rationalizable. The CCEI measures the proportion of income that a person wasted by making the choice that violated revealed preference. The swaps index finds the minimized weighted sum of numbers of alternatives in a menu that must be swapped with the chosen one in order to make the choice of a consistent with the maximization of a preference.
Secondary Outcomes (end points)
Our secondary outcomes include individuals' satisfaction with their choice and decision-making procedures, choice frequencies, choice changes, awareness of mistake, first-order stochastic dominance and risk preference.
Secondary Outcomes (explanation)
We measure individuals' satisfaction by survey questions after their completions of choice tasks. Participants simply provide their subjective evaluations of satisfaction with their choice and decision-making procedures. We measure choice frequencies by computing the frequency of different choices in each menu in different treatment groups. We measure individuals' changes in choices by using an incentivized design. There are two blocks, A and B, each consists of the same 21 decision problems. Participants ex-ante do not know that Block A and B consist of the same decision problem before entering Block B. Depending on the treatments, in Block B, they are presented with their Block A choices and can make different choices by applying again the same procedure as in Block A. In the end, they are asked to only one of Block A and Block B for payment. We count the number of choices that differ in Block A and Block B as a measure of awareness of "mistake" in making choices. We also check the violation first-order stochastic dominance in every menu. We will estimate risk-preference using the algorithm provided in Halevy, Persitz and Zrill (2018), We will elicit preference of each individual based on the method provided in Apesteguia and Ballester (2015).
We design an online experiment via Qualtrics, which will take approximately 35 minutes. We recruit participants from the Prolific. The experiment consists of three sections. In Section 1, each participant has to make choices in economic decision problems, either with a randomly assigned decision-making procedure or with her preferred decision-making procedure. In Section 2, we have three tests to measure cognitive function, including the International Cognitive Ability Resource (ICAR), the Stroop Color and Word Test (SCWT) and the Sternberg task. In Section 3, we collect information about participants’ preference for rational choice and choice process. Lastly, we ask participants for their demographic information, including gender, nationality, age, English level, education level and field of study.
Experimental Design Details
We recruit participants who are aged above 18 and are English native speaker. We exclude participants whose major are in Economics to avoid experimental demand effects. In Section 1, there are two blocks, A and B, each consists of 21 decision problems representing a set of portfolio options associated with two equally probable unknown states. This is the most common design in studying decisions under risk (e.g. Choi et al., 2014; Carvalho et al., 2016; Kim et al., 2018). There are three treatment groups: sequential examination, sequential elimination and procedure selection. Participants firstly complete Block A without knowing that Block B consists of the same decision problems as Block A. In Block B, participants are presented with their Block A choices and can make different choices by applying again the same procedure as in Block A. In the end, every participant is asked to choose only one of Block A and B for payment. Then the computer randomly chooses one round from the 21 rounds from the chosen block with equal probability. The participant will get rewarded depending on the outcome of this chosen round.
In Section 2, The ICAR consists of 10 questions about matrix reasoning, and three-dimensional rotation (Condon and Revelle, 2014). The SCWT and the Sternberg task are standard tests in psychology to measure selective attentions and working memory capacity, respectively.
In Section 3, participants will have to imagine themselves in two sets of hypothetical scenarios and answer a few questions related to the scenarios. First, we present them with two inconsistent choices known as "attraction effects" and ask how comfortable they are. Participants who indicate that they are uncomfortable suggests that they have a strong preference for choice consistency. Second, to control for "consequentialism", we ask them to make a choice between two hypothetical trips in which they will be doing the same things: 1) The trip is planned by the subject herself; 2) The trip is planned by someone else; 3) Indifferent. Participants who select the third option suggests that they are more likely to focus on the consequence. Third, we will also ask a scenario of sunk cost fallacy because it may lead to higher satisfaction under the treatment conditions. I adapt the example from Arkes and Blumer (1985). Subjects are asked to imagine that they have spent £50 on a ticket for concert A and £100 on a ticket for concert B. They are also to imagine that they really prefer A to B, that they have just discovered that the events are mutually exclusive and that the tickets have no salvage value. They are asked which concert they would then choose to go to. Participants who choose concert B suggests that they are affected by the sunk cost bias.
The randomization will be done by the online survey tool Qualtrics.
Was the treatment clustered?
Sample size: planned number of clusters
Sample size: planned number of observations
Sample size (or number of clusters) by treatment arms
80 individuals treatment 1 (sequential examination), 80 individuals treatment 2 (sequential elimination), 100 individuals treatment 3 (procedure selection)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
INSTITUTIONAL REVIEW BOARDS (IRBs)
The Institutional Committee for Ethical Review of Projects (CIREP) at Universitat Pompeu Fabra
IRB Approval Date
Post Trial Information
Is the intervention completed?
Is data collection complete?