How Information Display Affects Health Insurance Decisions

How to cite

Park, Iris SooJin; Sydnor, Justin; Wen, Yuxin. 2026. "How Information Display Affects Health Insurance Decisions." AEA RCT Registry. AEARCTR-0018968. https://doi.org/10.1257/rct.18968-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Iris SooJin Park* , Justin Sydnor , Yuxin Wen

* Primary investigator

Registry ID
AEARCTR-0018968
DOI
https://doi.org/10.1257/rct.18968-1.0
First registered
Last updated
Status
In development
Trial dates
2026-06-23 – 2026-06-29
Countries
Private
JEL codes
I13, D81
Registry record
http://www.socialscienceregistry.org/trials/18968

Abstract

This study examines how the format used to present health-insurance information affects plan choices, and how these effects depend on the type of financial-risk information most relevant to the decision. In an online experiment, participants are randomly assigned to view health-insurance plan comparisons using one of four display formats: standard plan features, a simple summary graph, a detailed distribution graph, or a scenario-based display. Participants then make a series of choices between pairs of health insurance plans that differ in premiums, cost sharing, and the distribution of possible out-of-pocket costs. The choice scenarios vary the structure of the decision problem, including dominance, variance, tail risk, and risk tradeoffs. The primary outcome is whether participants choose the benchmark-preferred plan in each scenario. The study tests not only whether alternative displays improve decision quality on average, but also whether their effects depend on the structure of the choice problem and the type of financial-risk information being communicated.

Intervention

The intervention varies the format in which health-insurance plan information is displayed. Participants are randomly assigned to one of four display formats: standard plan features only, a simple summary graph, a detailed distribution graph, or a scenario-based display.

Experimental design

Participants are randomly assigned to one of four versions of an online health-insurance plan-choice simulation. The versions differ in how information about plan costs and financial risk is displayed: a control condition with standard plan features only, a simple summary graph, a detailed distribution graph, or a scenario-based display. All participants make choices between pairs of health insurance plans across 5 situations that vary the tradeoffs between plans. We measure a) understanding of dominance, b) insurance literacy, and c) attitudes toward financial risk and use these as independent proxies for potential biases in plan selection.

Design details

Randomization method
Randomization is done by computer in Qualtrics using embedded-data random-number fields. Participants are randomly assigned to one of four study groups.
Randomization unit
Treatment assignment is at the participant level. Participants are also individually randomized to one of two plan-pair versions for each scenario type, to the Plan A/Plan B display order.
Clusters
2,000 individual participants. The treatment is not clustered.
Observations
2,000 participants.
Arms
Planned sample size is 500 participants per arm: * 500 participants in the control arm * 500 participants in the simple summary graph arm * 500 participants in the detailed distribution graph arm * 500 participants in the scenario-based display arm
Primary outcomes
For four of the five decision scenarios, there is a plan choice that aligns better with standard theory of first-order and second-order stochastic dominance. The primary outcome of interest for the choices in each of these four sets is whether the participant chose the "dominant" plan in this sense. In the fifth scenario, both plans are rationalizable. For descriptive purposes we will focus on analyzing the share who chose the more risk-averse option (more insurance coverage and higher premiums).
Secondary outcomes
We do not have secondary outcome metrics, but we do intend to conduct as a secondary endpoint analysis of the interaction between elicited risk preferences and choices in the tail-risk and risk-tradeoff plan sets. The question of interest is whether any of the visualization treatments increase the correlation between plan choice and measured risk attitudes in these two settings. As additional secondary end points we will analyze the interaction between choices in the situations with dominance and two measure that capture potential bias: a) failure to understand dominance in a separate dominance question and b) low insurance literacy as measured by inability to accurately estimate out-of-pocket costs given a spending scenario and plan details.

Cited by

No citation data available for this registration yet.

Keywords