Learning Effects in Strategy-Proof Mechanisms in School Choice

Last registered on August 14, 2024
View Trial History

Pre-Trial

Trial Information

General Information

Title
Learning Effects in Strategy-Proof Mechanisms in School Choice
RCT ID
AEARCTR-0012834
Initial registration date
August 14, 2024

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
August 14, 2024, 3:52 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Primary Investigator

Name
Sabrina Herzog
Affiliation
Düsseldorf Institute for Competition Economics - Heinrich-Heine-University Düsseldorf
Contact Primary Investigator

Other Primary Investigator(s)

Additional Trial Information

Status
In development
Start date
2024-08-15
End date
2024-11-30
Keywords
Behavior, Education
Additional Keywords
Learning, Matching Mechanisms
JEL code(s)
C78, C91, D47, I29
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Improving the matching of students to schools has been the subject of much research in recent years. Since school choice can have a long-lasting educational impact on individuals’ lives, the mechanism determining the matching should be chosen with care. One property that researchers and policymakers have agreed on being desirable is that the matching mechanism in an educational setting should be strategy-proof in order to give applicants fair chances to submit their preferences over schools. Faced with a strategy-proof matching mechanism, it is a (weakly) dominant strategy for applicants to submit their true preferences. However, empirical evidence has shown that people fail to understand strategy-proofness and instead try to strategically manipulate their preference submissions.
Two common mechanisms that fulfill this property are the Deferred-Acceptance (DA) mechanism and the Top Trading Cycles (TTC) mechanism. The discussion around which mechanism is best to implement often favors the DA mechanism. Much of the research is concerned with making strategy-proofness clearer to applicants and to adjust the application procedure to facilitate truthful preference submission. What is less clear, however, is how good individuals are at learning how to participate in either mechanism. What if it turns out that individuals are much better at learning how to participate in the TTC than the DA mechanism? Since parents often have to make school choice decisions multiple times (e.g., for several levels of education such as kindergarten, primary school, secondary school, etc., or if they have more than one child), the comprehension of learning effects allows for a broader overall understanding of how people behave in these decisions.
This project aims to investigate learning effects in a laboratory experiment and to re-evaluate the DA and TTC mechanism in terms of comprehensibility and learning behavior. In 20 rounds, students are matched in groups of four and have to decide on their preference submissions in an artificial school choice setting where they compete over four schools. The matching outcome is determined by either the DA or the TTC mechanism. In addition to comparing learning effects in the DA versus TTC mechanism, measured as increases in truth-telling rates, in a second treatment variation advice is added on the dominant strategy (submitting preferences truthfully). This allows for an analysis of the extent to which improvements are due to learning versus advice. A subsequent survey on subjects' strategies, and economic and socio-economic preferences is intended to help deepen the understanding of the findings.
External Link(s)

Registration Citation

Citation
Herzog, Sabrina. 2024. "Learning Effects in Strategy-Proof Mechanisms in School Choice." AEA RCT Registry. August 14. https://doi.org/10.1257/rct.12834-1.0
Sponsors & Partners
Experimental Details

Interventions

Intervention(s)
Four experimental treatments: Deferred-Acceptance mechanism (with and without advice) and Top-Trading-Cycles mechanism (with and without advice) (see experimental design)
Intervention Start Date
2024-08-15
Intervention End Date
2024-11-30

Primary Outcomes

Primary Outcomes (end points)
Truth-telling rates in Deferred-Acceptance mechanism (with or without advice) and Top-Trading-Cycles mechanism (with or without advice)
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The laboratory experiment will be conducted in a between-subject design with four treatments. In each treatment, students are randomly assigned to groups of four and are asked to submit an application form to a centralized school admission authority in 20 rounds of an artificial school choice problem in which they compete for four schools with one seat each. Students are randomly reassigned to groups of four after each round of matching. To simplify the experiment for students, each student is assigned a color and remains with that color throughout the experiment. Preferences over schools are induced by the payoffs that students can receive when they are matched to a school. One round out of 20 is randomly selected to be payoff relevant at the end of the experiment.

The matching of students to schools is determined by either the Deferred-Acceptance (DA) mechanism or the Top-Trading-Cycles (TTC) mechanism (1st treatment variation). In addition to the preference rankings submitted by the students, the mechanisms determine the matching outcomes using priority rankings that prioritize the four students at each school. The priority rankings are not known to the students. The priority rankings of students at schools as well as the induced preference order of students over schools change in each round, so that students face a new matching problem in each round. Following Chen & Sönmez (2006), preference orders are determined by a utility function that includes artificial utilities for school proximity and school quality, and a random factor to capture diversity in tastes. For the priority ranking, each student is prioritized at one school, while the priority over the other students is randomly determined for all schools. In each round, each student is prioritized at a different school, to support equal chances of being assigned to the more preferred schools.

The aim of the experiment is to measure learning effects over the 20 rounds in both mechanisms by assessing the truth-telling rates in each round. By comparing the evolution of truth-telling rates in the DA vs. the TTC mechanism, the two strategy-proof mechanisms, can be evaluated in terms of their learning effects. In addition, students may receive advice on the (weakly) dominant submission strategy, i.e. to submit preferences truthfully (2nd treatment variation). This allows to analyze the effect of learning compared to the effect of advice (in both mechanisms). Students receive feedback on their own matching outcome in each round.

After the repeated school choice problem, students in all treatments are asked to complete a survey. I collect the following measures: their submission strategies, understanding of the optimal strategy (truth-telling), perceived fairness, perceived efficiency, and perceived understandability of the mechanism, trust (in general, in the mechanism, and in institutions), IQ (proxied through highest school-leaving certificate, last math grade and three items of a Cognitive Reflection Test), risk preferences, loss aversion in risky choices, and some demographic characteristics.

Experimental Design Details
Not available
Randomization Method
Subjects who sign up for the experiment randomly receive a seat number before entering the laboratory. Otherwise, randomization is done by a computer.
Randomization Unit
experimental sessions
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
planned are five clusters per treatment
Sample size: planned number of observations
up to 256 students in four treatments
Sample size (or number of clusters) by treatment arms
up to 64 students in DA without advice, up to 64 students in DA with advice, up to 64 students in TTC without advice, up to 64 students in TTC with advice
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
With a total of 256 students (64 subjects per treatment), I can detect an effect of 20 percentage points at conventional power levels (one-sided chi^2 test, 5% significance level, 80% power). This is based on the assumption that in the DA without advice and TTC without advice treatments, 60% of the students play truth-telling in the first rounds while in the last rounds of DA without advice and TTC without advice, as well as in the first rounds of DA with advice and TTC with advice, 80% of the students play truth-telling, respectively. If the truth-telling rates are generally lower, e.g., 55% in the first rounds of both mechanisms without advice, and 75% in the last rounds of DA without advice and TTC without advice, as well as in the first rounds of DA with advice and TTC with advice, I can still detect the same effect size with a power of 77%.
Supporting Documents and Materials
IRB

Institutional Review Boards (IRBs)

IRB Name
German Association for Experimental Economic Research e.V. Institutional Review Board
IRB Approval Date
2024-06-18
IRB Approval Number
IyGUKThE
Analysis Plan

Analysis Plan Documents

Analysis plan "Learning Effects in Strategy-Proof Mechanisms in School Choice"

MD5: 3b41610162f15440e5bfbfbaa1a9c0c2

SHA1: 242c4ee6e6554abff1b5c361111a02cbefb9b0a5

Uploaded At: August 12, 2024