Analysis of Country-"Withheld" Twitter Accounts

This repository contains data and analysis supporting the BuzzFeed News article, "An Inside Look At The Accounts Twitter Has Censored In Countries Around The World," published January 24, 2018. That article contains important contextual and methodological details. Please read it before continuing below.

Data

This repository contains the following three datasets:

• data/withheld-accounts.csv: This file contains metadata on all of the Twitter accounts BuzzFeed News observed being withheld in at least one country. You can find a more detailed description of the columns below. Unless noted otherwise, all data came directly from Twitter's API. Countries in which an account is withheld are explicitly indicated in Twitter's API.

• data/first-observations.csv: This file indicates the first date on which BuzzFeed News noticed that a particular account was being withheld in a particular country. To be clear: These dates do not not correspond to the dates on which Twitter began withholding that account in that country. All this file affirmatively says is that the account was first withheld on this date or earlier.

• data/twitter-transparency-reports/*.csv: These tables come directly from Twitter's semiannual transparency reports, specifically their reports on "removal requests".

Fields in the withheld-accounts.csv file

Field name	Description
user_id_numfix	This is a copy of the user_id field, but prefixed with an ' so that it is not converted into scientific notation by spreadsheet programs.
user_id	This is the account's Twitter-assigned identification number.
screen_name	This is the account's screen name, e.g., @screenname.
bio	This is the most recent profile bio that BuzzFeed News observed for this account.
withheld_category	This the account's most recent status, with regards to withholding, as categorized by BuzzFeed News in its most recent data collection. See below for possible values.
withheld_in_countries	This is a +-separated list of two-letter abbreviations corresponding to the countries in which BuzzFeed News observed this account being withheld in its most recent data collection.
withheld_ever	This is a list similar to withheld_in_countries, except for every country in which BuzzFeed News has observed this account being withheld at any point.
followers_count	This is the number of followers that this account had at the time of the most recent data collection.
following_count	This is the number of accounts this account followed at the time of the most recent data collection.
signup_date	This is the date on which the account was created.
first_observed	This is the date on which BuzzFeed News first observed this account being withheld. See above for important caveats about this date.

withheld_category values in withheld-accounts.csv

The withheld_category field in the withheld-accounts.csv file can be one of the following four values:

• withheld: As of the most recent data collection, this account was still withheld in at least one country.
• not_withheld: As of the most recent data collection, this account was no longer being withheld in any country.
• media_violation: This account was previously withheld in one or more countries but, as of the most recent data collection, was "temporarily unavailable" globally for violating Twitter's Media Policy.
• inactive: This account was previously withheld in one or more countries but, as of the most recent data collection, was no longer accessible on Twitter — because, for example, the account has been suspended or the account's creator has deleted it.

Analysis

This repository contains two analysis notebooks, both written in Python:

• notebooks/analyze-withheld-accounts.ipynb, which focuses on the data collected from Twitter's API.
• notebooks/analyze-twitter-transparency-reports.ipynb, which focuses on the data collected from Twitter's transparency reports.

Reproducibility

To reproduce the calculations, you'll need to do the following:

• Ensure that you have installed Python and the Python libraries listed in requirements.txt.
• Re-run each of the Jupyter notebooks in the notebooks/ directory. (Shortcut: run make reproduce; requires Python 3.)

Feedback / Questions?

Contact Jeremy Singer-Vine at jeremy.singer-vine@buzzfeed.com.

Looking for more from BuzzFeed News? Click here for a list of our open-sourced projects, data, and code.