The Ontario provincial election wrapped up last night with the Liberals unofficially narrowly missing a third-straight majority, as Dalton McGunity remains premier of Ontario despite being over a dozen percentage points down in the polls prior to the start of the election. The election Twitter tracker was also very accurate in the early measurements of the election showing how the Liberals had caught up in a previous article I wrote.
For measurement purposes, I am comparing the Twitter sentiment results to that of the popular vote in last night's election to see how accurate E.L.E.C.T is. Is it more accurate than the 19 out of 20 or 5% margin of error traditional standard political research pollsters predict? The numbers I've used are an average from seven different days since the debate as recorded on the Edelman E.L.E.C.T site (including the night of), up to October 5th, the day before the election.
E.L.E.C.T used Twitter conversation sentiments- the Liberals got an average of 38.4% of the positive conversations, the Conservatives 33.5% and the NDP 21.5%, while the Conservatives got 40.2% of negative sentiments, the Liberals 37.4% and the NDP 18.8% as an average over the seven different days tracked. Neutral conversation sentiments were 36% for the Liberals, 37.4% for the Conservatives and 18.4% for the NDP.
The unofficial actual results of the popular vote were the Liberals with 37.6%, the Conservatives with 35.4% and the NDP with 22.7%, all falling within the industry standard 19 out of 20 or 5% accuracy rate that research firms like Ipsos and Nanos have boasted for years.
However, this was a much larger sample size that included 84,095 conversations over seven days, despite the Conservatives winning share of voice over the Liberals everyday except for one, with an average of 39.3% share of voice over the Liberals' 35.3%.
The positive, neutral and negative sentiments shown above should paint the picture why the Liberals won the election despite losing average share of voice!
In the past there have been comments made to me that Twitter is generally left-leaning. However, the results didn't really show that except in the case of the Green Party which only got three percent of the popular vote, but an average 5.4% share of voice. Still, well within the 5% margin of error.
Further, perhaps voter turnout would have been higher from Canadians than the 49.2% record low recorded in this year's election if there was mobile and online voting.
From personal experience I've spent two hours in line trying to vote in Toronto- and yesterday I was simply too busy to exercise my patriotic duty, but could have done so if there was an online or mobile option.