Election Prediction

It is hard to keep up with politics back in Australia, especially with the ridiculous game of polls leading up to the September 7th elections in 2013. Elections tend to be less about policies and more about how people feel about the candidates.

Twitter is a great resource to get an insight into what people are saying in bursts of 140 characters. So I decided to get a sense of what people are thinking about two main candidates by looking at Tweets. Instead of relying purely on the number of tweets or followers for a candidate (which was used to predict the US election outcome), I opted for a sentiment based approach – i.e. how positive or negative are the tweets mentioning the candidate. I call it Tweet Sentiment Index (TSI).

Tweet Sentiment Index for Australian Federal Election - 2013

The chart above shows a snapshot of the TSI over 6 days before the September 7th election for Tony Abbott (blue) and Kevin Rudd (red). As you can see Kevin Rudd is consistently trailing Tony Abbott. Even though Kevin Rudd (with 1.4M followers) has more followers than Tony Abbott (272K followers), the tweets mentioning Kevin tend to have more of a negative sentiment. The most telling bit is that both candidates have a negative TSI, which is very typical of Australian irreverence towards its politicians.

Now for the actual election result – As predicted by TSI, Tony Abbott won the election by a 17-seat 3.6% two-party swing. [ ABC Election Result,  wikipedia ]

So in conclusion, Tweet Sentiment Index works. A more subtle point is – its important to look at what is being said, not just how many people are saying it.