For those following the Australian general election on Saturday, Scott Morrison’s victory came as a shock. The polls and the bookies had predicted a Bill Shorten/Labor win. This result echoed other recent reversals of the polls – namely those of Trump’s win over Hilary Clinton, and the Brexit vote.
None of these results came as a surprise, however, to Professor Bela Stantic. The director of Griffith University in Queenland’s Big Data and Smart Analytics lab has been analyzing social media. By doing so, he has an impressive track record of predicting the winners. In the case of the US Presidential election, he even accurately forecast how all but one state voted.
The Role of Social Media Sentiment in Predicting the Outcome
In the five days before the Australian election, Professor Stantic analysed 2 million social media posts relating to 50 key terms. The posts emanated from more than half a million unique accounts. His findings led to him predicting a Scott Morrison win. During the campaign, he was almost a lone voice contradicting what the polls were saying. But he was right. Despite the Coalition trailing in every major opinion poll for three years, the Coalition romped home. They can also form a majority government, a rare position to be in for recent Australian governments.
With polling and betting markets getting it so wrong, experts are increasingly turning to social media to assess voter sentiment.
Parallels with Crypto Sentiment
The analysis of social media sentiment is, of course, what Causality does in its Crypto Sentiment Alerts tool. There are some similarities between predicting the outcome of an election, and predicting the movement of cryptocurrency prices. In both cases, it is the departure from the traditional metrics and focusing on the cumulative effect of what the “punters” are actually saying that provides clues as to what the outcome will be. So in essence, people are more honest when they’re talking to their friends or to a focus group on social media than they are when they’re answering polls. Likewise, people say exactly what is going through their mind, and often, when responding to Cryptocurrency related news. In contrast, mainstream media reports news filtered by a pre-conceived set of ideas or principles.
When it comes to predicting movements in Cryptocurrency prices, traders need to be able to distinguish between the rhythmic ups and downs in social media sentiment from events that have a dramatic effect on prices. A key metric that allows this is sentiment signal. The sentiment signal is generated by a proprietary deep learning model. Values close to -100% are very strong negative sentiment signals. Values close to +100% are very strong positive sentiment signals. Used in conjunction with token price, volume data and other traditional metrics, early warning signs can be detected of any news or chatter that could prove to be market moving.
What Conclusions We Can Draw From the Election
What we can conclude from the election is that it was not just clever use of social media by the Coalition that made people change their vote. It was the noise generated by social media that helped shaped voters’ intentions. Towards the end of the election campaign, the volume of positive social media posts was higher for the Coalition than it was for Labor. So was social media engagement. Scott Morrison did a better job of attracting more likes and positive comments than did Bill Shorten. Hence, the engagement and the positivity was contagious. Similarly, when looking for cryptocurrency trading signals, it’s the upswell of sentiment – measured by the increase in volume of positive or negative sentiment, that corresponds to a market moving event, that can help traders make the right decisions.
Some Case Studies to Read
To see some real life cases on how our Crypto Sentiment Alerts tool has been able to predict significant moves in the cryptocurrency market, see these articles:
Also published on Medium.