Opaque Admissions: Preference Intensity, Cutoff Information, and Reporting in Centralized Admissions

How to cite

Yurko, Anna. 2026. "Opaque Admissions: Preference Intensity, Cutoff Information, and Reporting in Centralized Admissions." AEA RCT Registry. AEARCTR-0019224. https://doi.org/10.1257/rct.19224-1.0

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

Anna Yurko*

* Primary investigator

Registry ID
AEARCTR-0019224
DOI
https://doi.org/10.1257/rct.19224-1.0
First registered
Last updated
Status
In development
Trial dates
2026-10-01 – 2027-10-01
Countries
Private
JEL codes
D47, D83, I21
Registry record
http://www.socialscienceregistry.org/trials/19224

Abstract

Centralized university admissions systems often ask applicants to submit rank-order lists, but applicants may have limited information about how these lists are processed. Even in systems that use strategy-proof mechanisms such as deferred acceptance (DA), applicants may observe admission thresholds and relative program competitiveness without understanding whether it is safe to rank ambitious options first. This experiment studies how participants report preferences in opaque settings, how historical threshold information and transparent DA guidance affect their reports, and how alternative assignment rules perform when applied to the resulting behavior.

The experiment uses a neutral prize-allocation setting that preserves the main strategic features of centralized admissions while reducing the influence of participants’ prior admissions experiences. Participants rank a small set of alternatives. They share the same ordinal preferences but differ in the intensity of their preferences, allowing the study to test whether submitted rankings respond to cardinal incentives. Participants make ranking decisions at different priority levels with no interim feedback in a controlled competitive environment.

The design has three treatments: an opaque mechanism with no historical thresholds, an opaque mechanism with historical threshold information, and historical thresholds combined with a transparent DA cutoff/menu description. The experiment tests whether participants report truthfully or use safety-oriented heuristics, and whether these distortions are insensitive to cardinal payoffs or vary with preference intensity in ways consistent with immediate acceptance (IA)-like reasoning or nonstandard preferences. Elicited mechanism beliefs are interpreted together with ranking decisions as complementary evidence on these possible explanations. The experiment also tests whether transparent DA guidance increases truthful reporting and weakens safety-oriented distortions. Finally, estimated reporting rules are used in a secondary fixed-behavior counterfactual that applies DA and IA to the same behaviorally generated reports and compares their welfare and priority-violation outcomes under realistic limited-information conditions.

Intervention

The intervention varies the information environment in which participants submit rank-order lists in a centralized assignment task. All participants face the same prize-allocation problem, share the same ordinal ranking but differ in cardinal preferences over three prize types, and receive information about the available prize types and capacities, the number of competitors, the distribution of priority scores, and their own payoffs and priority score. Treatments differ in the additional information provided about the assignment environment. In Treatment 1, the mechanism is opaque, and no historical thresholds are shown. In Treatment 2, the mechanism remains opaque, but participants observe historical admission thresholds. In Treatment 3, they observe the same thresholds and also receive transparent deferred-acceptance guidance using cutoff/menu language that explains how submitted rankings are processed and discourages the strategic demotion of a truly preferred option.

Experimental design

This is an individual-level laboratory experiment on reporting behavior in a centralized admissions problem. Participants complete a neutral prize-allocation task with three prize types, score-based priority, equal prize capacities, and possible non-assignment. All participants share the same ordinal preferences over prizes, but differ in the cardinal value of the middle prize. Subjects submit rank-order lists for several prespecified priority scores that are common across subjects but presented in random order with no interim feedback. One submitted list and score are randomly selected for payment. Experimental variation comes from three information treatments that differ in whether the mechanism is opaque, whether historical threshold information is shown, and whether deferred-acceptance cutoff/menu guidance is provided. The design isolates the effect of information on preference reporting by holding the competitive environment fixed through computerized opponents.

Design details

Randomization method
Treatment assignment will be randomized by computer at the individual-subject level, with approximately equal allocation across the three treatment arms. Payoff type will also be randomized by computer and balanced so that the high-middle-value and low-middle-value types are equally represented overall and, as closely as feasible, within treatment arms. The order of the priority scores will be randomized within subject.
Randomization unit
Individual participant.
Clusters
360 individuals.
Observations
360 participants * 12 priority scores = 4,320
Arms
120 participants per treatment arm
Primary outcomes
The primary outcomes measure reporting behavior in the rank-order-list task. The first primary outcome is truthful reporting: an indicator equal to one when the submitted rank-order list matches the participant’s true ordinal preference ranking. The second primary outcome is payoff-sensitive misreporting in prespecified score regions around the relevant thresholds. These outcomes will be compared across payoff types and treatments.
Secondary outcomes
Secondary outcomes are: 1. Low-score safety-oriented reporting, measured by whether the lowest value prize type is ranked first and by its position in the submitted list. 2. Overall reporting patterns, measured by the frequencies of the six possible rank-order lists over three prize types by priority score, as well as by distance from the truthful list. 3. Mechanism beliefs and comprehension, measured using responses to the mechanism-belief questions and performance on the incentivized comprehension task. 4. Benchmark fit under DA and IA, measured by report-specific regret. For each mechanism, regret is the difference between the highest expected payoff obtainable from any of the six possible reports and the expected payoff from the submitted report, conditional on the participant’s priority score and payoff type in the prespecified computerized-opponent environment. Outcomes include mean and median regret, the share of zero-regret reports, and the difference between DA and IA regret. 5. Fixed-behavior mechanism performance, obtained by applying DA and IA to the same empirically estimated reporting rules. Outcomes include total welfare, the share of participants left unassigned, the number or share of priority violations, and the share of middle value prizes assigned to participants who value it more highly.

Cited by

No citation data available for this registration yet.

Keywords