The theory, according to researchers at Indiana University, is Twitter can actually be used to predict election outcomes. Really.
I’ve been asked during many elections if Twitter could either influence or predict the outcome of elections. My response has always been “no” on both counts. I typically follow that “no” with “not yet, anyway.” The idea is too few people use or rely on Twitter in shaping their opinions, and the nature of the flow of information over the service is such that a small subset of people are responsible for the majority of traffic. That’s hardly representative.
It was when I reviewed the aforementioned theory and tested it against some recent election results and data I’d gathered from online chatter during the election that I began to wonder if the theory may apply.
The theory has largely been tested only against data from elections in the United States where politics is more like an industry. In Canada, politics is often viewed more as a nuisance.
Inspired by the plausibility of this theory and remarks by Ipsos Public Affairs CEO Darrell Bricker (“I always find it ironic when people come up with analyses after the fact about how things are supposed to work but they never actually tested it when they had to put a prediction on the line.”), I have decided to test the theory. Let me be clear that this is an experiment to determine if, and to what degree, measuring online activity can predict the outcome of a Canadian election.
I’m starting with Nova Scotia where Premier Darrell Dexter pulled the trigger on an election last week. Election-related tweets are most often tagged with #NSvotes (59%). Other popular hashtags are led by #NSelxn13 (28%) and #NSelxn (2%).
Based on the first week of online activity, Twitter is predicting Nova Scotians will re-elect the NDP. The question is, how large will the margin of victory be?
Analysis of Twitter mentions of the leaders suggests a reasonably tight race (Dexter/NDP 41.7%, McNeil/Liberal 34%, Baillie/Conservative 23.3% and Percy/Green 1%). [UPDATE 130916: It's important to note that Premier Dexter enjoyed some additional mentions on September 7th, the day the election was called. As illustrated in the graph, that clearly juiced his number for the week and things become more tight in the mentions race for the remainder of the period during which Premier Dexter 39.7% of the mentions followed very closely by Stephen McNeil with 35.7%.]
However, analysis of party mentions suggests a wider margin of victory (NDP 45.3%, Liberal 32%, Conservative 22.3% and Green 0.4%).
I’ve also conducted some initial analysis of selected issues. Taxes dominated the first week of online chatter (27.8% of issue-identified activity), followed by the province’s economy (16.9%) and employment (14.8%).