Information, Discretion, and Overshoot in Organ Allocation: An Experimental Test of Out-of-Sequence Offers

How to cite

Dutcher, Glenn; Green, Ellen; Schold, Jesse; Stewart, Darren. 2026. "Information, Discretion, and Overshoot in Organ Allocation: An Experimental Test of Out-of-Sequence Offers." AEA RCT Registry. AEARCTR-0019341. https://doi.org/10.1257/rct.19341-1.0

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

Glenn Dutcher* , Ellen Green , Jesse Schold , Darren Stewart

* Primary investigator

Registry ID
AEARCTR-0019341
DOI
https://doi.org/10.1257/rct.19341-1.0
First registered
Last updated
Status
In development
Trial dates
2026-08-11 – 2027-08-11
Countries
Private
JEL codes
C90, C91, D82, D91, D47, I11
Registry record
http://www.socialscienceregistry.org/trials/19341

Abstract

This study tests whether human decision-makers in a simulated organ allocation setting deviate from rational behavior when given discretion to reorder a transplant waitlist ("out-of-sequence" offers). Participants act as an organ allocation agent choosing whether to list a donor organ and, if listed, whether to promote identified high-acceptance-probability candidates ahead of their default waitlist position. The design varies the amount of information about candidate types (none, partial, or full) across three simulated organ-quality markets (high, median, and low quality), each calibrated so a rational, profit-maximizing benchmark strategy is well defined. The study measures whether participants over-reorder the queue (promoting candidates beyond the point that improves placement) or over-list organs (listing when a rational benchmark would decline), relative to the model's optimal predictions.

Intervention

Participants make a series of incentivized decisions in the role of an organ allocation agent. In each round, they decide whether to list a simulated donor organ and, if listed, whether and how to reorder a 4-candidate waitlist before extending offers. The amount of information they have about which candidates are likely to accept the organ is varied (none, partial, or full), and the underlying organ quality is varied (high, median, or low).

Experimental design

We will use a mixed between/within design. Participants are randomly assigned to one of three simulated "market" conditions (manipulated between subjects) and complete a sequence of incentivized allocation decisions under four information conditions (manipulated within subjects) that vary how much candidate-level information is revealed before the decision. The within-subjects conditions are presented as contiguous blocks of five periods each (20 periods total), with block order counterbalanced across participants. In each period, participants make an entry decision (whether to process a simulated case) and, conditional on entry, a costly sequencing decision over candidates. Participant behavior is compared against an optimal benchmark and against pre-specified alternative decision rules, to test (a) whether participants take the costly action more than the benchmark prescribes, (b) whether they enter more than the benchmark prescribes, and (c) whether any such departures are stable across the session or attenuate with within-session experience (order and learning effects).

Design details

Randomization method
Random assignment performed by the platform: participants are randomly assigned to one of three market-quality arms; within each market arm, participants are assigned round-robin to one of four block-order sequences.
Randomization unit
Individual participant.
Clusters
150 individual participants (no clustering)
Observations
150 participants x 20 periods each = 3,000 observations
Arms
50 participants in each of 3 market arms (High, Median, Low), 150 total.
Primary outcomes
(1) Re-ordering count (number of promotions performed), compared to the model's optimal promotion count k*; (2) Entry/listing decision (list vs. decline), compared to the model's optimal entry decision; (3) Unconditional waste (probability the organ goes unplaced, counting non-listing as waste); (4) Expected waitlist rank of the accepted candidate (E[W|accepted]), a measure of departure from the default fairness-based queue order.
Secondary outcomes
Expected stopping position/sequence position of the accepted offer; number of high-priority candidates "leap-frogged" (contacted after a lower-priority candidate was promoted past them); realized profit relative to optimal profit. We also are interested in how the order of the information condition leads to differential learning.

Cited by

No citation data available for this registration yet.

Keywords