Random Experiment on Relative Feedback with University Students

How to cite

Macías, Cristian. 2024. "Random Experiment on Relative Feedback with University Students." AEA RCT Registry. AEARCTR-0014016. https://doi.org/10.1257/rct.14016-1.1

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

Cristian Macías*

* Primary investigator

Registry ID
AEARCTR-0014016
DOI
https://doi.org/10.1257/rct.14016-1.1
First registered
Last updated
Status
Completed
Trial dates
2022-01-09 – 2024-12-31
Countries
Spain (Madrid)
JEL codes
C93
Registry record
http://www.socialscienceregistry.org/trials/14016

Abstract

Our research analyzes the impact of relative feedback on the academic performance and motivation of university students at one of the leading universities in Madrid. After the students take a midterm exam, the treatment group is given their percentile rank in relation to the class average grades. At the end of the course, we aim to determine whether the feedback's impact on the treatment group improves their grades or, conversely, if it leads to overconfidence and worsens their results.

Intervention

Our research analyzes the impact of relative feedback on the academic performance and motivation of university students at one of the leading universities in Madrid. After the students take a midterm exam, the treatment group is given their percentile rank in relation to the class average grades. At the end of the course, we aim to determine whether the feedback's impact on the treatment group improves their grades or, conversely, if it leads to overconfidence and worsens their results.

As a post-trial update, we wish to register that, in addition to the pre-specified average treatment effect analyses, we have conducted a series of heterogeneity analyses. These include subgroup interactions by baseline performance (low initial grades), socioeconomic indicators (scholarship status, low parental education), self-reported expectations, relative age (quarter of birth), gender, and academic history (number of enrollments, exam round, and age). We also explored variation across different PLAN assignments and interactions with resource indicators such as car ownership, driving license, and private classes. These analyses were performed using the same specifications as the main models, with robustness checks including PLAN fixed effects, individual controls, and alternative outcome definitions (e.g., grade improvements between midterm and final exam).

The results indicate that average treatment effects remain small and mostly insignificant, but important heterogeneity emerges. In particular, students with weaker academic histories (older, more exam attempts, or multiple prior enrollments) tend to benefit more from percentile feedback, consistent with a motivation-through-information mechanism. Similarly, students with lower expectations or disadvantaged family backgrounds display differential responses, though often imprecisely estimated. These findings, while exploratory and not pre-registered, enrich the interpretation of the intervention’s impact and highlight the importance of considering subgroup variation in the design of information-based educational interventions.

Experimental design

The aim is to analyze whether knowledge of their percentile rank throughout the semester among students at Universidad Rey Juan Carlos in Madrid improves their final course grades. To achieve this, a database has been compiled over several semesters, comprising a sample of over 600 students. The class was divided into a treatment group, who received their percentile rank along with their midterm grade, and a control group, who only received their exam grade. To understand the sample, questionnaires were administered at the beginning of the course, asking for relevant data and their estimated final course grade. Additionally, at the end of the course, another questionnaire was administered to determine whether awareness of their percentile rank stimulates additional motivation or, conversely, leads to overconfidence.

As a post-trial update, we wish to register that, in addition to the pre-specified average treatment effect analyses, we have conducted a series of heterogeneity analyses. These include subgroup interactions by baseline performance (low initial grades), socioeconomic indicators (scholarship status, low parental education), self-reported expectations, relative age (quarter of birth), gender, and academic history (number of enrollments, exam round, and age). We also explored variation across different PLAN assignments and interactions with resource indicators such as car ownership, driving license, and private classes. These analyses were performed using the same specifications as the main models, with robustness checks including PLAN fixed effects, individual controls, and alternative outcome definitions (e.g., grade improvements between midterm and final exam).

The results indicate that average treatment effects remain small and mostly insignificant, but important heterogeneity emerges. In particular, students with weaker academic histories (older, more exam attempts, or multiple prior enrollments) tend to benefit more from percentile feedback, consistent with a motivation-through-information mechanism. Similarly, students with lower expectations or disadvantaged family backgrounds display differential responses, though often imprecisely estimated. These findings, while exploratory and not pre-registered, enrich the interpretation of the intervention’s impact and highlight the importance of considering subgroup variation in the design of information-based educational interventions.

Design details

Randomization method
Randomized experiment conducted using statistical software
Randomization unit
University students from different degrees and subjects across various academic years
Clusters
604 students
Observations
1500 students
Arms
300 students in the control group and 304 in the treatment group
Primary outcomes
Determining whether relative feedback helps improve university students' grades

Cited by

No citation data available for this registration yet.

Keywords