Sharing and following on Twitter
How to cite
HENRY, EMERIC; Zhuravskaya, Ekaterina; Guriev, Sergei; Marquis, Theo. 2022. "Sharing and following on Twitter." AEA RCT Registry. AEARCTR-0010219. https://doi.org/10.1257/rct.10219-1.1
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
This study aims at understanding news consumption and news sharing on Twitter. Specifically the purpose is to:
* Estimate parameters of the utility function that can explain the decision to share news (true or false). In particular how do individuals trade off veracity of the news / ideological bias of the news and potential virality of this news?
* Compare different policies that have been proposed to fight the circulation of fake news, in particular (a) making veracity salient (b) asking for confirmation clicks (c) proposing fact checking.
* Evaluate the impact on beliefs and voting intentions of exposing individuals to balanced content.
Two experiments will be conducted during the campaign of the 2022 midterm elections.
Intervention
The study will be composed of two experiments.
In the first experiment the participants will be offered 2$ to sign up to follow an account for a month. In the first treatment the account will stop being active, in the second treatment, the account will retweet the most popular news from both liberal and conservative media (balanced account) while in the third treatment the participant will be offered to follow either an account with only retweets from conservative sources or an account with only retweets from liberal sources.
In the second experiment, the participants will be offered a choice to retweet one of 4 tweets (2 true, 2 false). The treatments will reproduce different policies to reduce the circulation of fake news
Experimental design
Experiment 1
The survey will ask questions on (a) Twitter usage (2) socio-demographic information (3) beliefs on key policy issues (4) voting behavior. The individuals will be contacted 3 weeks before the elections. All participants will be recontacted one week after the elections and will answer questions on (1) beliefs (2) who they voted for in the elections.
Treatments: participants will be offered 2 dollars if they choose to follow a Twitter account we created (see details below) that offers a selection of information. We have written an algorithm that selects the most popular tweet (as in most retweeted) from left leaning mainstream media and the most popular tweet from right leaning mainstream media related to politics. The list of media is taken from Alcott and Gentzkow.
Treatment 1: account stops being active
Treatment 2: account retweets balanced combination of conservative and liberal news
Treatment 3: either a liberal or a conservative account.
Experiment 2
The second experiment aims at studying the drivers of sharing behavior of news. We will conduct an online survey of roughly 3500 American voting-age individuals on the Qualtrics online platform, restricting our sample to Twitter users.
The survey will have the following phases
(a) Socio-demographic characteristics
(b) Twitter usage
(c) Political preferences
Then, at the end of the survey, the participants will be asked whether they want to share one of 4 tweets or not to share (we refer to this as the retweet screen). The question will be:
Do you want to retweet? (randomized order)
• Tweet 1
• Tweet 2
• Tweet 3
• Tweet 4
• Not retweet
Tweet 1 to 4 will be selected in the following way. There will be 2 tweets from the mainstream media (one tweet left leaning, one right leaning) described above and 2 tweets based on information that was fact checked and shown to be false, also one from the left and one from the right (see methodology below).
After the choice of whether to retweet or not has been made, we will have two additional questions
Q1: How many times do you expect this tweet to be retweeted? (asked for each tweet)
Q2: How likely do you think the statement in this tweet is true? (asked for each tweet)
Q3: Do you think the information in this tweet favors the republicans or the democrats? (asked for each tweet)
Treatments
T1 (300 participants): the participants do not have the retweet screen. The goal of this treatment is to compare the answers to Q1, Q2 and Q3 with participants who don’t share.
T2 (700 participants): standard treatment with the retweet screen
T3 (extra click) (700 participants): Before seeing the retweet screen we will inform them that they will need to confirm their intention to retweet in the next screen.
T4 (make veracity salient) (700 participants): Ask question Q2 (how likely is the treatment true) before the retweet screen
T5 (fact checking) (700 participants): On the retweet screen give them the option to see fact checks of the tweets.
Design details
Cited by
No citation data available for this registration yet.