In 2008, statistician Nate Silver upended reporting on elections by calling primary results on his blog FiveThirtyEight with uncanny accuracy. His predictive model considers a wide number of factors outside of polling data, such as economic trends and state-by-state demographic factors. As we tread further into the 2012 Presidential Election season, an increasing number of data-driven services and enterprises are trying to best Silver’s model by using a variety of approaches and datasets.
The latest is Twitter, which launched the Twitter Political Index on Wednesday. Developed in partnership with Topsy Labs, the Twitter Political Index aims to gauge how the candidates are faring among users of the service. The Index scrapes over 400 million tweets a day to perform sentiment analysis on posts that mention the election or candidates. The approach has definite advantages—sentiment on Twitter could reveal developing trends quicker than pollsters or reporters—but is also limited by demographics, since Twitter users are not necessarily representative of the general voting population.
In an interview with USA Today, Topsy Labs’s chief scientist Rishab Ghosh likens the project to a “barometer” of public enthusiasm towards the respective candidates, “not a replacement for opinion polls.” In that respect, the Index provides unique insight into how sentiment is shifting among the Internet’s ever-expanding hive mind.
Twitter is far from the only one applying predictive models to new and unique datasets to try and call the horse race before it’s done. As detailed at GigaOm, a number of organizations such as InTrade and Yahoo! News are taking innovative approaches to tracking and visualizing developing political sentiment.
Watch video of Nate Silver speaking at the 2012 Data Science Summit.
About the Author