Complexity and Decision Making under Risk
How to cite
Jung, Uihyeon; Choi, Syngjoo; Lee, Eungik. 2025. "Complexity and Decision Making under Risk." AEA RCT Registry. AEARCTR-0016214. https://doi.org/10.1257/rct.16214-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
This study investigates how complexity affects decision-making under risk, focusing on risk attitudes and rationality. Participants make a series of portfolio allocation decisions varying in complexity, with some receiving state-payoff feedback (information treatment). The study examines whether complexity-induced behavioral attenuation stems from cognitive uncertainty, which reflects optimization difficulties. Additional variables are measured to explore their roles in shaping decision-making. In a follow-up experiment, a social learning treatment is introduced, providing participants with the average choices of previous participants.
Intervention
This lab experiment will be conducted with 300 participants aged 18 or older, recruited from four-year universities located in Seoul, Korea. It will be conducted over Zoom and will last approximately two hours.
Experimental design
The first experiment is sixty portfolio allocation problems based on Choi et al. (2007). Thirty baseline problems are allocation across two assets, and thirty complex problems are allocation across four assets, which contain two additional assets that are just convex combinations of baseline assets. Participants can make their decisions either by dragging the investment bars or directly entering their allocations in the investment box. By between variation, we introduce state-payoff information treatment, which automatically computes and shows their state-payoff determined by their investment decisions. If participants who receive information treatment also exhibit behavioral attenuation, we can conclude that attenuation was an individual-specific inherent phenomenon, caused by not knowing how to optimize, rather than computational complexity.
We elicit cognitive uncertainty in every 5th round, 12 rounds in total. Also, after the experiment, we additionally measure two individual variables: cognitive ability and objective uncertainty. Cognitive ability is measured by 12 problems from Raven Advanced Progressive Matrices. Objective uncertainty is defined as the portfolio choice. More specifically, participants solve 6 problems which give a target-state payoff, and properly allocate investment as closely as possible to satisfy the state payoff. These two measurements show ancillary yet meaningful explanations of how decision under complexity, risk attitude, and rationality are associated with these two measurements.
The second experiment is identical in structure to the first, except for the social learning treatment: participants are shown the average investment allocations made by previous participants. For each information treatment condition, we provide separate averages based on whether the source participants exhibited behavioral attenuation or not. This design allows us to investigate if the social learning treatment helps reduce cognitive uncertainty, then its effect differs depending on the social learning treatment conditions.
Design details
Cited by
No citation data available for this registration yet.