While almost all of the polling industry is in shock at the stunning Trump victory in the US presidential election on November 8, 2016, as reported earlier on TCN, one prediction - with a very distinct Croatian flavour - got things almost exactly right. TCN chats to Vuk Vukovic, one of the founders of Oraclum, which - just like the 45th American President to be - bucked the trend with its pre-election prediction.
1. Stunning predictions - congratulations. You were SO against the grain. What was your reaction when you saw the predictions prior to publication?
Thank you! It was a shock, I have to admit. We saw already on Sunday that PA was going to Trump, and I personally couldn't believe. At one point we even started doubting our results in some states given that they were so different than what everyone else was saying (e.g. in PA or NH). But we stuck to our model and we were right!
2. You obviously have faith in your system, but please confess - were you a little nervous publishing a prediction so different to the established pollsters?
Oh absolutely. We're always nervous when we make our predictions. This one however was particularly scary given that all the US pollsters have so far done such a good job of predicting elections. I don't even blame them, it's the polls that are becoming more and more problematic in catching these kinds of sentiments. Not just in the US, but in the rest of Europe as well.
3. I assume you followed the election results as they came in. Was there a defining moment in the results when you realised just how right you were?
I would say it was when they finally called North Carolina and Florida. Soon after that Trump was holding a big lead in Wisconsin and Michigan, and closing the gap in Pennsylvania and New Hampshire (he won all four in the end). At that point it was becoming obvious to us that we were right, but we were still careful and wanted to wait until PA was called.
4. Why did you get it so right when other more established pollsters got it so wrong?
Because we used a different type of survey - a prediction survey. The established pollsters are all basically picking up these low-hanging fruit polling data, whereas we had to make our own database, which was a difficult process. It took us three weeks and we were only able to get about 450 people, some with help of friends sharing our survey on social networks, others with ads. We needed to get a bit more people in the key swing states and we managed to do it in the end.
The point was that we do not require a representative sample to make a good prediction. And I think this is the biggest problem the pollsters are currently struggling with - how to make a more representative sample. This is why our method goes beyond representatives and random sampling, and focuses simply on trying to find out how people estimate their local conditions and sentiments. And the people are pretty good at this.
5. For those not familiar with your methodology, what sets it apart from more traditional polling methods?
It's a wisdom of crowds concept (some also call it citizen forecaster). We ask our respondents what do you think who will win the election, by what margin, and how do you think others in your state will answer that same question. We also apply a network analysis of our respondents' friends on facebook (the survey is anonymous, so we don't know who they are or who they voted for but we can see who their anonymous friends voted for). This helps us uncover whether the people are living in their own little bubbles where everyone around them simply thinks the same way.
6. Having been proved right on the Trump presidency, how did you feel on a professional - and a private - level?
On a professional level it's great! It was the big win we were hoping for and preparing for a long time. When we started the company we said that we needed to perfect the method in order to make it shine at the US elections. We managed to do just that!
7. What next for Oraclum? With such spectacular success in the most followed election in the world, the phone will not stop ringing. Where do you go from here?
Yes, it's gonna be hectic over the next few days. But this comes at a perfect time for us as we are currently looking for financing and looking to expand our little start-up. Hopefully we will be able to turn it into much more than a pollster agency. Elections are just the first step. We want to test and see where else the prediction method might work.
To see the Oraclum prediction in full, and to learn more about Oraclum, click here.
