Popularity, Social Networks and Student Outcomes

Last registered on January 23, 2024


Trial Information

General Information

Popularity, Social Networks and Student Outcomes
Initial registration date
January 22, 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
January 23, 2024, 1:17 PM EST

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



Primary Investigator

Columbia University

Other Primary Investigator(s)

Additional Trial Information

On going
Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Social networks shape an array of decisions, cognitive and non-cognitive skills for students that have long lasting consequences in their lives. Isolated students, due to lack of social connections, often exhibit deficiencies in multiple dimensions including behavioral, personality and academic skills. This project employs a novel randomized desk mate matching strategy based on initial classroom network characteristics to improve student cognitive and non-cognitive outcomes.

Linking isolated students to popular students can lead to development of cognitive and non-cognitive skills through better integration, positive peer effects and enhanced classroom experience. However, if popular peers are saturated in their socialization capacity, their inaccessibility may result in further ostracization. As a result, it might be better to link two isolated students to each other to form more stable friendships. A priori, it is not clear whether an optimal strategy requires linking isolated individuals to each other as desk mates or linking isolated individuals to popular individuals. The project divides the students in the classroom into three types by their measure of popularity (high, medium, and low). It then employs a two-level randomization design where firstly, desk mate type is randomized between high, medium, and low for each individual and secondly, the proportion of low to high and low to low matches are varied to alter the overall classroom environment. Using the randomization design, the project uncovers the optimal mixing strategy for improved social integration, and development of non-cognitive skills and academic capabilities.
External Link(s)

Registration Citation

Bhargava, Palaash. 2024. "Popularity, Social Networks and Student Outcomes." AEA RCT Registry. January 23. https://doi.org/10.1257/rct.12813-1.0
Sponsors & Partners

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

Request Information
Experimental Details


The intervention randomly matches individuals to each other as desk mates based on their level of popularity (constructed through nominations over several dimensions of classroom social network). Further, it varies the proportion of high to low and low to low popularity matches across classrooms within the same school-grade to alter the overall classroom match environment. Students are made to sit with their prescribed desk mates for a duration of 2-2.5 working months during regular classroom instructions.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
The project measures outcomes over 4 main dimensions.
1. Number and strength of social network ties.
2. Classroom experiences.
3. Personality and behavioral traits.
4. Academic performance and capabilities.
Primary Outcomes (explanation)
1. Number and strength of social network ties: Both in-degree nominations to and out-degree nominations from an individual are measured over friendship networks, recess networks, lunch networks and help networks. Further, a count of how many individuals nominate the individual as an ideal desk mate is obtained. Teacher perceptions about level of isolation are also documented.
2. Classroom experience: Self-reported responses are recorded on a scale of 1-10. Questions include an individual's comfort in asking questions and doubts in class, perceptions of bullying, willingness to participate in classroom discussions and the level of excitement to interact with peers and teachers.
3. Personality and behavioral traits: A combination of self-reported responses and hypothetical games are employed to elicit an individual's level of Big-5 personality traits, risk tolerance, patience, pro-sociality, trust, optimism, morality, competitiveness and grit. Games and questions are modeled from Jackson et al. 2023, Bhargava et al. 2023, Shan and Zoelitz 2023, Falk et al. 2018.
4. Academic performance and capabilities: Whenever available, school administrative records are used to obtain an individual's GPA. Further, in line with Keller et al. 2023, information on an individual's academic self concept is elicited through the use of self-reported responses to questions such as where do they think they stand within the classroom in terms of academic performance and how comfortable are they with subject materials and exams.
All outcomes are evaluated independently but wherever necessary and if interpretation allows, responses are clubbed into social indices either through PCA or through simple averages. All variables are standardized to facilitate comparisons.

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Based on the popularity score constructed from baseline measure of in-degree nominations, individuals are divided into three categories: low (L), medium (M) and high (H). Categorization is relative to the distribution of popularity score within the classroom an individual belongs to, i.e. bottom 33% of the distribution is categorized as low, top 33% as high and the remaining as medium.

The design then relies on a two-level randomization. In the lower level, individuals within the same classroom are randomly matched to each other, thus, forming pairs of the following type: LL, LM, LH, MM, MH, HH. In this case, a match to an H type desk mate is considered treatment 1, a match to an L type desk mate is considered treatment 2 and a match to an M type desk mate is considered control. A comparison of means method is employed where outcomes of individuals are compared across treatment 1 (match to H type), treatment 2 (match to L type) and control (match to M type) while controlling for baseline characteristics and classroom fixed effects. The primary population for which treatment effects are evaluated are the L type students. However, the project also tests the treatment effects on all students and also on M type and H type students separately.

In the upper level, with the focus being L type students, the proportion of LL, LM and LH type matches are varied across classrooms within the same school-grade. Three type of classrooms are created: C1 (60% (20%, 20%) of L type individuals are matched to L (M, H) type desk mates (respectively)), C2 (20% (20%, 60%) of L type individuals are matched to L (M, H) type desk mates (respectively)) and C3 (33.33% (33.33%, 33.33%) of L type individuals are matched to L (M, H) type desk mates (respectively)). For the remaining individuals in each class, pairs are formed randomly. Let's call C1 classrooms: Lopsided towards LL classroom (LopL), C2 classrooms: Lopsided towards LH classroom (LopH) and C3 classrooms: Balanced classroom (Bal).

By comparing outcomes of individuals across the three types of classroom (where one can consider LopL and LopH as treatment and Bal as control), treatment effects of classroom level changes (for e.g. better classroom management, general perception of amicability of high type peers etc.) induced by types and proportion of matches are obtained.

The lower level obtains treatment effects of the direct match whereas the upper level obtains treatment effects of overall match types. In order to further distinguish direct effects from global effects, interactions between the two treatment levels are used.

A linear in means comparison is also employed where the treatment (popularity level of desk mate) is considered to be continuous. Finally, to account for spillovers, structural assumptions are made about spillover procedure and Horvitz-Thompson estimators (Forastiere et al. 2024, Aronow and Samii 2017) are used to uncover direct vs indirect effects.

The seating plans stay in place over 2 - 2.5 working months, thus, giving desk buddies appropriate amount of time to form social connections with each other and each others' friends. Endline surveys are conducted at the termination of the seating plan.
Experimental Design Details
Not available
Randomization Method
Randomization is done in the office by the PI on a computer.
Randomization Unit
The randomization takes place on two levels (students within the classroom and classrooms with the same school-grade)
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
320 classrooms.
Sample size: planned number of observations
On an average 30 students per classroom, thereby providing us with a sample size of 9600 students.
Sample size (or number of clusters) by treatment arms
130 balanced classrooms, 95 lopsided towards LH classrooms and 95 lopsided towards LL classrooms. Within classrooms, randomization into direct treatment type (match to a specific type of desk mate) follows the strategy laid down in experimental details.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Institutional Review Boards (IRBs)

IRB Name
Monk Prayogshala
IRB Approval Date
IRB Approval Number
IRB Name
Columbia University
IRB Approval Date
IRB Approval Number