In our last post, we discussed a few of the challenges the polling industry is facing. Now let’s imagine a polling technique that never asked you anything. Nobody called or texted you. It didn't even ask you to fill out an online questionnaire. Instead, it just read the content you posted publicly online. Using your own opinions expressed on Twitter, Facebook, or whatever social platform you use, this 'pollster' deciphers what your leanings might be, using your own words, without asking you one question. This is exactly what Eyesover does.
Of course, there are challenges with this approach just as there are with traditional polling. For example, in the same way landlines can cause age range distribution issues, the issue can arise with social media data. The major difference is where the phone problem is only getting worse, social media usage is rapidly increasing across all age ranges and according to the Pew Research Center, there is now only a 15-16% drop in usage between the 18-29 age group and the 50-64 age group on Twitter and Facebook.
But there is still no question that there are many reasons why polling from online sources can and should be incorporated into public opinion tracking.
The most obvious reason is the massive amounts of data is available online. While a traditional poll might be produced from a sample of 500-2000 respondents, a poll derived from online sources can be based on hundreds of thousands, if not millions of respondents - a concept that would have seemed ludicrous to statisticians (and the firms that would have to pay for a poll that size!) not that long ago. This combined with the rapidly improving distribution of online users will tend to reduce the potential for sampling errors.
Another concerns that we often hear about online data is the potential for spam, bots, and fake accounts to skew the data. The Eyesover system eliminates the potential for such skewing by not only blocking bots, but we aggregate mentions by social media username. In other words, our data reflects individuals not simply mentions.
Does it work? You be the judge.
Eyesover’s forecast in the 2015 Canadian Federal Election was on a par with traditional polling and we also fared quite well in the 2016 US Presidential Election as we saw an under-reported base of Trump support throughout the campaign and reported the Electoral College was up for grabs heading into election night.
Our results show that the power of online, real-time polling has too much potential to ignore!
The Eyesover Team