Where in the world is my tweet: Detecting irregular removal patterns on Twitter

Area of Science:

• Social Sciences
• Political Science
• Computational Social Science

Background:

• Twitter data is increasingly vital for political science research.
• Understanding tweet removal rates, especially for politically charged content, is essential but under-researched.
• Technical limitations of Twitter's API and user self-censorship impact data validity.

Purpose of the Study:

• To develop and test a strategy for analyzing tweet removal rates on Twitter.
• To investigate potential distortions in Twitter's API data collection.
• To compare removal rates across different content topics, including political and non-political subjects.

Main Methods:

• Technical evaluation of Twitter's API stream properties to identify data loss.
• Collection of tweet data from contentious (e.g., terrorism, political leaders) and non-contentious (e.g., food) topics.
• Application of multilevel analysis to detect patterns in tweet removal rates.

Main Results:

• Twitter's forward stream API can result in an average loss of 2-5% of tweets.
• Significant variations in tweet removal rates were observed between different topic categories.
• Politically charged topics exhibit distinct removal patterns compared to non-contentious topics.

Conclusions:

• The study provides a novel strategy for obtaining valid samples of removed tweets.
• Findings highlight the importance of accounting for API limitations and topic-specific removal rates in Twitter data analysis.
• This research contributes to a more accurate understanding of information dynamics on social media platforms.