VOTING WITH YOUR TWEET: An experiment in political forecasting
If we were conducting a poll, we would disclose how many people we reached and where they were. Since we’re basing our predictions on a Twitter feed, we want to provide similar data on who Tweets, when, where, and about what.
Three pieces of information are particularly relevant:
- The Total Tweets per Day figure shows how many raw messages we gathered each day since we started collecting data. We are averaging about 10-20,000 messages a day and that sounds like a lot, but it works out to an average of just 40-50 messages per district
- The Total Tweets per District map shows that this average doesn’t mean much. Some districts get a lot of attention from Twitter users, others very little. This is especially true for national figures like Democrat Nancy Pelosi and Republican John Boehner.
- Regardless, we can see from the Daily Share by Party figure that the Democrats and Republicans tend to receive about equal attention from the Twitterverse. Republicans might get slightly more–but as we pointed out elsewhere, we also have slightly more Republican-leaning districts in our sample.
- Even if tweets are evenly distributed on average, they aren’t evenly distributed within each district. Incumbents receive far more attention from Twitter users than their challengers
The graphics require a modern browser such as Chrome, Safari or Firefox to render properly.
We’ll add more descriptive data soon, including summaries of what people are talking about in each district. But we’re already seeing lots of variation in how Twitter is used in each Congressional race and who receives the most attention. Also, be sure to read the FAQ if you are interested in how the data is generated.
*Notes: Third-party candidates are not accounted for in this experiment, nor are districts where candidates from the same party are competing for the same seat. For a full explanation of why some districts have no data, please see the FAQ.
Credits: Mark Huberty, Doctoral Candidate, Political Science, UC Berkeley;
Len De Groot, Interactive Data & Design instructor at kdmcBerkeley;
Hillary Sanders, Research assistant