Political Polarization and Social Media

Last registered on April 23, 2025
View Trial History

Pre-Trial

Trial Information

General Information

Title
Political Polarization and Social Media
RCT ID
AEARCTR-0004850
Initial registration date
October 12, 2019

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
October 15, 2019, 9:54 AM EDT

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

Last updated
April 23, 2025, 6:13 PM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Country
Argentina
Region
Buenos Aires Metropolitan Area

Primary Investigator

Name
Ernesto Schargrodsky
Affiliation
Universidad Torcuato Di Tella
Contact Primary Investigator

Other Primary Investigator(s)

PI Name
Rafael Di Tella
PI Affiliation
Harvard University
Contact Investigator
PI Name
Ramiro Galvez
PI Affiliation
Universidad Torcuato Di Tella
Contact Investigator

Additional Trial Information

Status
Completed
Start date
2019-09-20
End date
2020-12-31
Keywords
Electoral, Other
Additional Keywords
Political Polarization, Social Media, Electoral Debates, Twitter, Stress
JEL code(s)
P16, D72
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
A widespread concern exists regarding social media contributing to political polarization. Social media sites, by promoting the creation of “echo chambers” (situations in which beliefs are amplified or reinforced by communication and repetition inside closed systems), are believed to distort the salience of some beliefs and perceptions, potentially leading to polarized views. However, given that signing-up in a social network is not mandatory, a valid argument is that the presence of polarization patterns documented in social networks may arise in part because of their users being particularly prone to polarization. In other words, patterns observed in social networks may be simply the consequence of self-selection. Disentangling the magnitudes of these effects --polarization effectively caused by exposure to social media versus selection bias-- remains an open question.

We will conduct an experiment with participants from two populations: participants with and without Twitter accounts. Participants will randomly watch either a live stream of Argentina’s 2019 presidential debate or a placebo video. Also randomly, during the debate Twitter users will be asked to turn off their phones, to use their Twitter accounts as they wish, or to post content on their Twitter accounts. In this way, we plan to measure and compare two things. First, the way in which different degrees of access to social media affect polarization indices and stress levels. Second, the differential reactions on these variables for Twitter and non-Twitter users. This evidence will help to disentangle the roles of exposure and selection in the polarization observed in social media.
External Link(s)

Registration Citation

Citation
Di Tella, Rafael, Ramiro Galvez and Ernesto Schargrodsky. 2025. "Political Polarization and Social Media." AEA RCT Registry. April 23. https://doi.org/10.1257/rct.4850-1.1
Sponsors & Partners

Sponsors

Experimental Details

Interventions

Intervention(s)
To avoid experimenter demand effects, the study is described in detail in fields that will not become public until after the trial end date.
Intervention (Hidden)
See Experimental Design (Hidden)
Intervention Start Date
2019-10-13
Intervention End Date
2020-03-31

Primary Outcomes

Primary Outcomes (end points)
(1) Differences in measures of political polarization of participants across treatment arms.
(2) Differences in participants’ positive and negative perceptions of politicians across treatment arms.
(3) Differences in participants’ measures of affective polarization, institutional views, and political preferences of participants across treatment arms.
(4) Differences in salivary cortisol levels of participants across treatment arms.
(5) Heterogeneous effects across pre-treatment sociodemographic characteristics, ideology and beliefs.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
To avoid experimenter demand effects, the study is described in detail in fields that will not become public until after the trial end date.
Experimental Design Details
Two types of participants will be recruited, active “Twitter users” and “non-Twitter users”:
- Twitter users. We define a Twitter user to be active if her/his account was activated before June 20, 2019, and if she/he tweeted, retweeted, or answered tweets at least three times in the period from June 20, 2019 through September 20, 2019 (recruitment will start after this date).
- Non-Twitter users. This group includes people that do not have a Twitter account at all.
All participants must be between 18 and 70 years old and must be qualified to vote in the 2019 Argentine presidential elections. Participants will be recruited by our team and by a professional recruitment agency.
Participants will be invited to attend the Universidad Torcuato Di Tella campus (located in Buenos Aires, Argentina) on October 13, 2019 during the first 2019 Argentine presidential debate at 9pm. During recruitment, they are going to be informed that they will have to watch images in screens and that they will be required to provide two saliva samples (they are also going to be informed that all of their data will be handled anonymously). Finally, they are not going to be told explicitly that the experiment will overlap with the presidential debate.
We expect around 540 participants to take part in the experiment but effective turn out is uncertain. Eventually, we might reproduce the experiment at future presidential or mayoral 2019 electoral debates.
Upon arrival, non-Twitter users will be randomly assigned to two different groups:
- Control group. These participants will watch a placebo video, and will be requested not to use their cell phones.
- Treatment group 1: These participants will watch a live stream of the presidential candidates’ debate and will be requested not to use their cell phones.
Upon arrival, Twitter users will be randomly assigned to four different groups:
- Control group. These participants will watch a placebo video, and will be requested not to use their cell phones.
- Treatment group 1: These participants will watch a live stream of the presidential candidates’ debate and will be requested not to use their cell phones.
- Treatment group 2: These participants will watch a live stream of the presidential candidates’ debate and will be allowed to use their cellular phones and Twitter account as they usually do.
- Treatment group 3: These participants will watch a live stream of the presidential candidates’ debate and will be requested to produce content on Twitter in the following way: In addition to the Twitter accounts that they usually follow, they will also be asked to follow a Twitter account managed by us. From that account, we will be sending partisan and neutral messages related to the presidential debate. Users will be asked to interact with these messages by retweeting, liking, or commenting these tweets. Some of these messages will be retweeted from non-anonymous partisan accounts supporting or opposing the presidential candidates, others will be neutral (e.g., a tweet simply saying "I am enjoying this debate" or “This debate is boring”), or retweets from a fact-checker NGO (Chequeado). In addition, the screen in which the debate will be streamed will also contain a Twitter timeline showing the messages the account sends.
As participants arrive to the university facilities, they will first fill out a short questionnaire (which will include questions on sociodemographic characteristics, social network usage, beliefs, and betrayal aversion) and provide a first saliva sample (which we will use to measure base level salivary cortisol).
Then participants will watch their respective videos and, in the case of treatment groups 2 and 3 (Twitter allowed and Twitter encouraged participants), interact with their cellphones.
Once the videos end, participants will fill out a second questionnaire and provide a second saliva sample. This second questionnaire will include questions regarding political polarization, perceptions of politicians, affective polarization, participatory intention, institutional views, and electoral preferences.
Upon answering this questionnaire and providing the second saliva sample, each participant will be awarded a supermarket voucher as a token of gratitude and will leave the University premises.
Randomization Method
As participants arrive to the University premises, Twitter users will be instructed to go to a counter and non-Twitter users to go to another counter. Then, Twitter and non-Twitter users will be separately randomized into placebo and treatments.
Randomization will be made by order of appearance in the following way:
- Using an office computer, we will randomize whether the first arrived non-Twitter user will be assigned to the placebo group, or to treatment group 1. The following one will be assigned to the other group, and so on.
- Using an office computer, we will randomize the order of assignment of Twitter users into the placebo group, the treatment group 1, the treatment group 2, and the treatment group 3. Then, we will use that sequence to assign Twitter participants by their order of arrival..
Randomization Unit
Participants
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
We aim to have about 540 participants (360 Twitter users and 180 non-Twitter users)
Sample size: planned number of observations
We aim to have about 540 participants (360 Twitter users and 180 non-Twitter users).
Sample size (or number of clusters) by treatment arms
1) Non-Twitter users in placebo group: about 90 participants (one half of non-Twitter participants).
2) Non-Twitter users in treatment group 1: about 90 participants (one half of non-Twitter participants).
3) Twitter users in placebo group: about 90 participants (one quarter of Twitter participants).
4) Twitter users in treatment group 1: about 90 participants (one quarter of Twitter participants).
5) Twitter users in treatment group 2: about 90 participants (one quarter of Twitter participants).
6) Twitter users in treatment group 3: about 90 participants (one quarter of Twitter participants).
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Supporting Documents and Materials

Documents

Document Name
Pre-treatment Survey (in Spanish)
Document Type
survey_instrument
Document Description
File
Pre-treatment Survey (in Spanish)

MD5: 440dc1d6c7b96eb480a799bcd6e723e2

SHA1: 424ebf21b3cc9001c19225f04a6043b15f7808f8

Uploaded At: October 12, 2019

Document Name
Post-treatment Survey (Twitter users) (in Spanish)
Document Type
survey_instrument
Document Description
File
Post-treatment Survey (Twitter users) (in Spanish)

MD5: 2bf06da179cd6a61649fce5c5fe696db

SHA1: fcfbb13f735ef826162315e94dfc3cd4397c3361

Uploaded At: October 12, 2019

Document Name
Post-treatment Survey (non-Twitter users) (in Spanish)
Document Type
survey_instrument
Document Description
File
Post-treatment Survey (non-Twitter users) (in Spanish)

MD5: 5d159a61d84771ab9d3cc8ea57664a7d

SHA1: a8c712c8aaf7825ec4d5287a632c672d44581dab

Uploaded At: October 12, 2019

IRB

Institutional Review Boards (IRBs)

IRB Name
Comité Institucional de Ética - Instituto de Neurología Cognitiva (INECO), Buenos Aires, Argentina
IRB Approval Date
2019-09-17
IRB Approval Number
N/A
Analysis Plan

Analysis Plan Documents

Analysis Plan

MD5: e627530600ee7b606a04ac78639b0df9

SHA1: 305a66632e171fc8a32025a7d979906f00a29942

Uploaded At: October 12, 2019

Post-Trial

Post Trial Information

Study Withdrawal

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

Request Information

Intervention

Is the intervention completed?
Yes
Intervention Completion Date
October 14, 2019, 12:00 +00:00
Data Collection Complete
Yes
Data Collection Completion Date
October 14, 2019, 12:00 +00:00
Final Sample Size: Number of Clusters (Unit of Randomization)
We randomized at the individual level.
Was attrition correlated with treatment status?
No
Final Sample Size: Total Number of Observations
Twitter users: 447 participants.
Non-Twitter users: 241 participants.
Final Sample Size (or Number of Clusters) by Treatment Arms
Twitter users: 110 Twitter-allowed participants, 113 Documentary participants, 112 Debate-Only participants, 112 Twitter-Interact participants. Non-Twitter users: 122 Debate-Only participants, 119 participants Documentary participants.
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials