To analyze legislators’ activity on Twitter, researchers obtained 1,366,349 public tweets from each of the 2,056 members of the national legislative bodies in Australia, Canada, New Zealand, the UK and U.S. who had a Twitter account and posted at least once between Jan. 1 and June 30, 2019. The study includes legislators who were in office for just part of the six-month period. Tweets are only included in the analysis if they were produced while a legislator was in office. Researchers used the public Twitter Application Programming Interface (API) to collect the tweets and stored them in a database. Of the 2,180 legislators with a Twitter account, 2,056 legislators tweeted at least once during the time period analyzed here.
Researchers at the Center created legislator rosters by hand, manually identifying all sitting members in each country’s national legislatures and then searching for their Twitter accounts. These lists were continuously updated to account for elections, resignations, legislators changing parties and other events. Legislator accounts in the database include official, verified legislator accounts as well as any unofficial accounts that belong to the legislator, such as personal or campaign accounts. Accounts for honorary and non-voting members, such as Congressional representatives of U.S. territories and hereditary members of the British House of Lords, are also included.
Activity and engagement analysis
This report analyzes both legislator and audience behavior on Twitter. To measure activity, researchers took two separate approaches. First, they identified the share of all legislators in each country who tweeted at least one time on a daily, weekly or monthly basis. Second, they calculated the total number of tweets from each account during the study timeframe and then took the median across all legislators’ monthly totals. These measures are calculated separately for each of the five countries. Using the median helps to account for a small number of very active users of the platform, whose frequent tweeting skews the data. In both cases, the measures aggregate across both original tweets and retweets.
Original tweets are those that an official creates directly. This category includes quoted tweets (in which a user simultaneously retweets and adds their commentary to a tweet posted by someone else), reply tweets (in which a user takes part in a threaded conversation stemming from an initial tweet) and tweets that do not include any links to other tweets. Original tweets make up 46% to 74% of legislative tweets across the five countries in this study. The Twitter API provides engagement data, such as numbers of likes and retweets, for these original tweet types.
Retweets are tweets that come from other Twitter users and are retweeted by a legislator with no additional edits or comments added. We do not examine engagement data for these posts, but we do include retweets when examining each legislator’s total number of tweets, the percent of their tweets that are retweets, and each legislator’s overall mentions, hashtags and emoji use.
Engagement – the number of likes and retweets that legislators’ Twitter audiences provide – is also measured at the median. The main text of this analysis shows the median number of likes and retweets for original tweets from legislators in each of the five countries who have Twitter accounts and who have tweeted at least once during the study period. Researchers identified each legislator’s median original tweet, in terms of both likes and retweets, and then identified the median legislator based on these measures for each country, which provides the estimates shown here.
The analysis of likes and retweets excludes tweets that were retweets. It only includes original tweets, as researchers could not determine how much of the engagement for a particular retweet could be attributed to the legislator who retweeted it.
Defining the ‘engaged’ or ‘heavy’ Twitter user among legislators
Examining a small group of high-volume tweeters – a technique used in the body of this report – does not capture the broader set of activities that may define highly engaged users, such as following, liking and being followed by others. Still, understanding those who produce more tweets provides us with useful insight. For this report, we used the top 25% of tweeters by median total volume across the periods in which legislators were in office as a measure for users with the highest levels of engagement and compared them with less prolific legislators.
Counting hashtags, mentions and emojis
This analysis also includes counts of mentions, hashtags and emojis. Because mentions and hashtags are offset from normal text via punctuation (“@” and “#,” respectively), researchers used regular expressions to extract these from the tweet text and build term frequency matrices. Researchers also excluded forms of text that might contribute noise to the data, such as contractions (e.g. “’s” in “@realdonaldtrumps’s” or “#canada’s”) and emails (e.g. “firstname.lastname@example.org”). Since a prolific tweeter repeatedly using a hashtag or mention could inflate the count, these measures are normalized to reflect the percentage of legislators by country who have used each of the top hashtags or mentions in that country.
To count emojis, the research team used an additional Unicode emoji v13.0 (2020) lookup file from Kaggle to extract emojis from tweets, again using term frequency matrices and normalizing by legislator.