Economic Rationality under Cognitive Limitations: The Effect of Sequential Elimination

Participants are randomly assigned to one of three treatments, each distinguished by the choice procedure used in the first part of the experiment.

Participants are randomly assigned to one of three treatments: Sequential Elimination, the Direct Procedure, or the Minimum-Time Procedure. All treatment conditions employ closely comparable interfaces and instructions to isolate procedural differences. For each decision problem, participants select an option from a vertical list of options on the screen's left side. Each treatment begins with a practice trial. Example screenshots of the treatments are included in the Supplementary Materials.

In Sequential Elimination, participants make their choice by sequentially eliminating the options they do not prefer until only one option remains. To eliminate an option, they click on it, moving it to a "Trash" box on the right side of the screen. Until submitting their choice, participants can reinstate any eliminated option by clicking on it in the Trash box.

In the Direct Procedure, participants make their choice after sequentially examining all options. To mark an option as examined, they click on it, moving it to a "Choice List" on the right side of the screen. After all options are in the Choice List, participants select their preferred option from that list. In the instructions, this condition is referred to as “Sequential Examination” to control for potential naming effects.

In the Minimum-Time Procedure, participants make their choice directly from the initial list of options. They must spend at least 35 seconds on each decision problem before submitting their choice. This minimum duration is calibrated from pilot data to match the average time spent under Sequential Elimination. A timer on the right side of the screen displays the time elapsed since each problem appeared.

2025-12-08

2026-01-31

Choice consistency

Our principal measure of economic rationality is choice consistency, defined as a binary variable equal to 1 if a participant’s choices satisfy the Generalized Axiom of Revealed Preference (GARP), a necessary and sufficient condition for consistency with preference maximization (Afriat, 1967; Varian, 1982). Guided by our theoretical framework, we pre-register the following hypothesis:

Hypothesis 1: Sequential elimination improves choice consistency relative to the direct procedure among individuals with cognitive limitations.

We test this hypothesis by contrasting Sequential Elimination and the Direct Procedure treatments among low-IQ participants (those scoring at or below the sample median), who serve as our primary proxy for individuals with cognitive limitations. To capture the extent of deviations from rationality, we pre-register the Houtman–Maks Index (HMI, Houtman and Maks, 1985) as a discrete metric. The HMI is defined as the minimum number of choices that must be removed to achieve consistency, with removed choices typically interpreted as mistakes. As robustness measures, we employ the number of GARP violations and stricter variants of these measures that impose first-order stochastic dominance (FSD). These measures offer complementary assessments of sequential elimination, varying in stringency and granularity. The Supplementary Materials report the minimum detectable effects for all measures, providing a comprehensive account of design sensitivity.

References:
Afriat, Sidney N. 1967. “The Construction of Utility Functions from Expenditure Data.” International Economic Review 8 (1): 67–77.
Houtman, Martijn, and Julian Maks. 1985. “Determining All Maximal Data Subsets Consistent with Revealed Preference.” Kwantitatieve Methoden 19 (1): 89–104.
Varian, Hal R. 1982. “The Nonparametric Approach to Demand Analysis.” Econometrica 50 (4): 945–973.

Effects by attentional deficits (mechanism); response time and deliberation time; elimination behavior.

To isolate the underlying mechanism, we examine whether Sequential Elimination improves economic rationality (as captured by choice consistency and the HMI) relative to the Direct Procedure among participants with higher attentional deficits (those scoring at or above the sample median on the ASRS inattention subscale), as our secondary proxy for cognitive limitations.

As a supplementary analysis, we test for treatment effects in the full sample and using the full ASRS scale. Additionally, we contrast Sequential Elimination with the Minimum-Time Procedure to investigate whether the effect of the former persists when response times are comparable. We also measure deliberation time as the interval until the final click on any option, to evaluate whether imposing a simple time requirement influences deliberation and, in turn, choice consistency. Finally, we analyze elimination behavior under Sequential Elimination to explore systematic patterns that may account for its effect.

The experiment consists of three parts. In the first, participants make economic decisions under their assigned treatment. In the second, they complete a series of cognitive tasks. In the third, they complete a set of questionnaires assessing attitudes toward inconsistency, attentional traits, and demographic characteristics.

The experiment is conducted online via the Qualtrics platform. Participants are recruited from Prolific and must meet eligibility criteria (approval rate ≥ 95%, prior submissions ≥ 10, aged 18 or older) and provide informed consent.

In the first part, participants must pass initial instruction checks (with two attempts allowed for each) to proceed. They then face twenty-five risky decision problems under their assigned procedure, including twenty-four for assessing economic rationality and one comprehension check. Adapted from Kim et al. (2018), each problem presents eleven distinct options in random order, with each option representing a lottery yielding one of two monetary prizes with equal probability. The comprehension check problem contains eleven options ranging from (0, 0) to (111, 111) in strictly increasing order; selecting any option other than the strictly dominant (111, 111) indicates a lack of comprehension. Participants who fail this decision-problem check are excluded from the final analytic sample.

In the second part, cognitive ability is measured using IQ scores obtained through the ICAR test (Condon and Revelle, 2014), selective attention using the Stroop task (Stroop, 1935), and working memory using the Sternberg task (Sternberg, 1966). In the third part, participants rate how at ease they are with a choice inconsistency scenario on a scale from 0 (least at ease) to 10 (most at ease), where higher scores indicate a less negative attitude. The scenario illustrates the attraction effect, a well-documented form of choice inconsistency (Huber, Payne, and Puto, 1982). This part also includes questionnaires assessing attentional deficits using the Adult Self-Report Scale (ASRS v1.1; Kessler et al., 2005), which contains an embedded item to verify instruction following. Participants failing both the decision-problem check and this instruction check are not eligible for compensation. Finally, participants report demographic characteristics, including age, gender, and education level.


References:
Condon, David, and William Revelle. 2014. “The International Cognitive Ability Resource: Development and Initial Validation of a Public-Domain Measure.” Intelligence 43: 52–64.
Huber, Joel, John W. Payne, and Christopher Puto. 1982. “Adding Asymmetrically Dominated Alternatives: Violations of Regularity and the Similarity Hypothesis.” Journal of Consumer Research 9 (1): 90–98.
Kessler, Ronald C, Lenard Adler, Minnie Ames, Olga Demler, Steve Faraone, EVA Hiripi, Mary J Howes, Robert Jin, Kristina Secnik, Thomas Spencer, et al. 2005. “The World Health Organization Adult ADHD Self-Report Scale (ASRS): a short screening scale for use in the general population.” Psychological Medicine 35 (2): 245–256.
Kim, Hyuncheol Bryant, Syngjoo Choi, Booyuel Kim, and Cristian Pop-Eleches. 2018. “The Role of Education Interventions in Improving Economic Rationality.” Science 362 (6410): 83–86.
Sternberg, Saul. 1966. “High-Speed Scanning in Human Memory.” Science 153 (3736): 652–654.
Stroop, J. Ridley. 1935. “Studies of Interference in Serial Verbal Reactions.” Journal of Experimental Psychology 18 (6): 643–662.

Randomization is implemented via Qualtrics.

Individual

No

N/A

We target 600 participants (200 per treatment arm) in the final analytic sample. Anticipating a 5% exclusion rate due to failing the decision-problem check, we will recruit 630 participants, with additional recruitment if needed to reach the target.

200 participants per treatment arm. Based on the median split, this yields an internal target of 100 low-IQ participants per treatment arm.

Power is evaluated via simulation using pilot data. We target 80% power at a two-sided 5% significance level for the primary test on choice consistency, comparing Sequential Elimination with the Direct Procedure among low-IQ participants. We simulate 10,000 independent datasets per candidate effect size, with 100 low-IQ participants per treatment arm, preserving the empirical distribution of choice consistency observed in the Direct Procedure. Under this design, the minimum detectable effect of Sequential Elimination on choice consistency is a log-odds ratio of 0.782 (corresponding to an odds ratio of 2.186), implying an increase of 19.3 percentage points (from 0.412 to 0.605).