There has been a bit of controversy after Mashable wrote a post about using Twitter as a method of predicting the Iowa Caucus. It has been attacked from a few sides with some very, very good points. On this blog we have inferred that social media data will, at times, mimic professional polls. We also looked at national data to see if we could spot trends in the Iowa Caucus. As it turned out we were pretty close this time, but we do not expect to predict every primary and every election. In fact, we doubt we'll have the same results for the New Hampshire primary. Santorum will continue to get a lot of attention across social media as more of his past (and current) statements get discussed. However, recent polls show he may not do well in that state.
So, moving forward we are going to be more careful how we word our analysis of political social media data.
Here is why:
1) While tool vendors are trying very hard to be able to "geo-locate" a large quantity of social media data, they just are not there yet unless you spend a huge amount of time going through the data manually. Using national data to predict a primary or caucus will simply not work every time.
2) There is no way to filter the results by party or even by registered voters. There could be a lot of positive posts about a candidate, but we'll never know if they are the ones who are actually going to cast the vote.
3) Sentiment analysis is almost there, but not quite. A good analysis would include a discussion of positive and negative mentions. While some of the more expensive tools get close to automating sentiment, the language of social media is still rapidly evolving so it is hard for the tool vendors to keep up.
That all said, there is still great value in social media data for politicians. What social media DOES provide is a huge amount of unbiased data. People are not trying to "guess" the right answer or decide between answers they may not fully agree with. We've all taken polls where we've chosen the "least wrong" answer rather than not answering at all. If there is one thing people like to do on social media is express their opinions. It may be unstructured and untargeted, but almost every time expresses their true feelings.
1) We can look at the words people are using to describe a candidate, an issue or an election over time. We can see if a certain advertisement, speech, social media strategy or GOTV program has changed the conversation across social media.
2) We can spot emerging trends before they hit the mainstream media. Occupy Wall Street is a perfect example of this, as was Rick Santorum's sudden rise in Iowa.
3) We can spot negative statements that might damage a politician's reputation.
4) We can identify the top key influencers; people who post about a particular topic often to a large audience that interacts with the data (retweets, likes, comments etc). From this data we can build outreach programs to reach these important folks in hopes of having them amplify our messages.
Will Ensomo make predictions about elections in the next year? No, and we truly were not trying to with the Iowa Caucus. Our goal was, and still is, to compare social media data to the polls and the outcomes. We'll leave the predicting to the pundits and the pollsters. But we will keep on monitoring what is being said across the social media universe to show how useful social media monitoring and analysis can be in politics.