Experimental School Choice with Parents Conference: ; ;
Mikhail Freer, Thilo Klein, Josué Ortega · 2026
Read Experimental School Choice with Parents Conference: ; ; by Mikhail Freer, Thilo Klein, Josué Ortega (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-24615.txt.
Introduction
Freer and co-authors run the first lab school choice experiment where parents, not students, are the subjects. They compare deferred acceptance with two manipulable but possibly more efficient rules. All three are manipulated often, with no clear gap in truth-telling. Parents and students manipulate at similar rates, though students make more obvious errors. Deferred acceptance yields little justified envy but never a Pareto-efficient match, while rank-minimizing gains efficiency at a real stability cost.
Catalog record
| CREATOR | Mikhail Freer, Thilo Klein, Josué Ortega |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 112396 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.24615
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Experimental School Choice with Parents Conference: ; ; Mikhail Freer, Thilo Klein, Josué Ortega Abstract . We conduct the first laboratory school choice experiment in which parents – the relevant decision makers in the field – are the experimental subjects. We compare Deferred Acceptance (DA) with two manipulable but potentially more efficient alternatives: Efficiency-Adjusted Deferred Acceptance (EADA) and the Rank-Minimizing mechanism (RM). We find that all mechanisms are frequently manipulated, with no significant differences in truth-telling rates. Parents and students manipulate at similar rates, supporting the external validity of student-based experiments, though students make significantly more obvious errors, suggesting parents’ deviations are more deliberate. Despite widespread manipulation, the predicted welfare–stability tradeoff largely survives: DA never produces Pareto-efficient allocations yet generates little justified envy; whereas RM delivers substantial efficiency gains at a meaningful stability cost. EADA occupies a middle ground: its efficiency gains over DA are modest and imprecisely estimated yet double justified envy. Higher cognitive ability is associated with more deviations, and under EADA with worse outcomes. While DA does not induce truth-telling, it is the only mechanism in which manipulation never pays off and rarely changes outcomes. Keywords : school choice, experiments, parents, efficiency-adjusted DA, rank-minimizing. 1. Introduction The redesign of student assignment mechanisms is one of the most visible contributions of economic theory to public policy. Beginning with Abdulkadiroğlu and Sönmez 2003 , economists showed that school admissions procedures in several American cities induced strategic behavior, giving an advantage to families who better understood the allocation process. A central policy response has been the adoption of strategy-proof mechanisms such as the student-proposing Deferred Acceptance (DA) algorithm ( Gale and Shapley 1962 ) . DA achieves two important objectives. First, truthful reporting is a weakly dominant strategy, so sophisticated families cannot gain an advantage through strategic preference manipulation. Second, under truthful reporting, DA produces assignments that respect schools’ priorities: put differently, it does not generate any justified envy. These advantages have driven DA’s widespread adoption in school admissions around the world. As DA became the policy benchmark, attention naturally turned to the question of whether parents in practice behave in the truth-telling manner that theory prescribes. Although DA makes truth-telling a weakly dominant strategy, observed behavior may still depart from truthful reporting. Such departures can generate justified envy relative to agents’ true preferences, even though DA’s outcome is stable with respect to the submitted rankings. To study behavior under controlled conditions, economists have conducted laboratory experiments in which subjects rank artificially created schools, with monetary incentives tied to their assignment. These experiments allow researchers to compare subjects’ induced true preferences with their submitted rankings, and to measure the impact of any manipulations on the efficiency and justified envy of the corresponding assignments. The large experimental school choice literature, beginning with Chen and Sönmez 2006 , has consistently found that DA experiences substantially less manipulation than the Boston mechanism. These experimental findings also influenced policy decisions. In the words of Calsamiglia et al. 2011 , “it played an important role to convince the Boston school district authorities to replace the previous mechanism by one of the other mechanisms.” Nonetheless, despite being extensively replicated under a number of robust environments in the lab, the external validity of lab-based school choice remains unclear, as findings about students’ behavior in school allocation mechanisms may or may not generalize to those of actual parents. Differences in cognitive sophistication, among several other characteristics in which these two populations may differ, may be especially relevant for strategic behavior in matching mechanisms ( Basteck and Mantovani 2018 ) . Because parents and undergraduates may differ systematically along these dimensions, evidence from student-subject experiments need not reflect parental behavior. As Levitt and List 2007 note, “ there is no question more fundamental to experimental economics than understanding whether laboratory results generalize to naturally occurring environments ’’. Yet, despite influencing education policy worldwide, the school choice literature has not, to the best of our knowledge, directly tested this generalization from undergraduates to parents in a controlled experiment. 1 1 1 There is, however, a large literature beyond school choice that analyzes the representativeness of standard laboratory subjects. We discuss it in Section 2 . Our main contribution is to conduct the first laboratory experiment on school choice mechanisms in which parents, rather than students, serve as the experimental subjects. Using this field-in-the-lab setting directly addresses the external validity gap by observing how the relevant decision makers behave when confronted with different assignment rules. Studying parents allows us to examine behavior under markedly different demographic and decision-making characteristics than those of undergraduates. By comparing parents’ reports with those of students under otherwise identical experimental conditions, we provide the first direct evidence on whether the behavioral patterns documented in student-based studies extend to the population for whom these mechanisms are ultimately designed. A second contribution is to experimentally evaluate two alternative and relatively novel mechanisms alongside DA: Efficiency-Adjusted Deferred Acceptance (EADA, Kesten 2010 ) and, for the first time in any experimental setting, the Rank-Minimizing mechanism (RM, Featherstone 2020 ). These mechanisms represent different points on the efficiency-stability frontier. EADA is designed to deliver Pareto improvements over the DA outcome under truthful reporting, limiting justified envy. RM is also Pareto-efficient, but goes further by minimizing the sum of assigned ranks, at the cost of disregarding priorities entirely. RM therefore serves as a benchmark for rank-efficiency: it quantifies how much welfare can be achieved if one is willing to give up on schools’ priorities. Both mechanisms, in particular EADA, have received substantial interest from theorists because, although manipulable, the incentives to deviate are not obvious and thus they could deliver large efficiency improvements over DA if subjects report their preferences truthfully ( Troyan and Morrill 2020 ; Troyan 2024 ) . Moreover, while EADA does not fully respect priorities, it satisfies several weaker but meaningful stability notions (described in Section 2 ), which have made it a prominent mechanism in the literature and a candidate for real-life implementation in Flanders. Our contribution is to test these mechanisms in the lab and to document whether manipulations eliminate their efficiency guarantees under truthful reporting. Results preview. We recruit 324 parents and 216 students and assign them to one of five treatments: parents face DA, EADA, or RM. Students face DA under equal or lower stakes for comparison. We match them with their peers in groups of 18 subjects, competing for 18 seats distributed among seven schools (i.e. parents interact with parents, and students with students). Subjects are informed about the mechanisms and the ranking interface, tested on their mechanism understanding, and given a practice round in the ranking task. We induce preferences over schools using a convex payoff function, guaranteeing a payment of £55 if they are assigned to their top choice. The experiment is designed to mimic real-life school choice: there is substantial but incomplete information and parents only rank schools once. In particular, subjects know how many seats are available at each school, the number of competitors who rank each school as top choice, and their priority at each school. After ranking schools, we measure cognitive ability using incentivized Raven matrices, as Basteck and Mantovani 2018 , to detect differences in cognitive