As we follow the Canadian Federal Election through social media here at Eyesover, we often have to remind ourselves that we are not a polling company. While polling companies are measuring vote intention through traditional methods, we are analyzing online comments for issues related to the election and determining if the comments have a positive or negative sentiment attached to them.
However, with new polls appearing almost daily during the election, it is hard to not turn our attention to polling results with the understanding that sentiment towards a party or leader on various topics would play a significant role in the formation of an individual’s vote intention. So we set out to use the Eyesover system to look for insights into voting intention using metrics such as sentiment and frequency analysis.
Since elections are based on one vote per individual, any social media data used in determining vote intention must reflect that reality. As a result, we used the average sentiment of all comments by an individual to ensure hundreds of comments from one individual do not outweigh one comment from hundreds of individuals. We also took the number of individuals discussing a particular party or leader into consideration since frequency of mentions has been shown in various research studies to be a reasonably good proxy of overall support for a party. The methodology we use to arrive at our Support Index is to multiply the normalized daily leader/party sentiment score by the percentage of individuals discussing that particular leader/party for the same time period.
So what were the results? If you’ve been following the polls, we’re sure you’ll recognize the patterns in the accompanying graph. After analyzing an average of 60,000 daily comments, we found the Conservatives and the Liberals in a close race with the NDP support decreasing over the past ten days.
The fact the online data has produced results so similar to traditional polls demonstrates
the possibilities for using the Eyesover system in conjunction with polling. The benefit for our customers comes from the fact our results are an accumulation of thousands of daily discussions and opinions pertaining to hundreds of, in this case, election-related topics. This provides customers with the ability to identify the underlying topics or issues where their positions are viewed positively by the public and those that are not – good information for any entity, private or public, to have.
The Eyesover Team