A couple of months ago, we reported on the success of European researcher Nikos Askitas in performing real-time tracking of public opinion on referendums using a very simple comparison of hourly Google searches.

Askitas did a good job on the Irish referendum on same-sex marriage and on the Greek referendum on economic policy.

Thursday is the British referendum on the exit from the European Union, and it’s super-close in the polls.

Askitas has been tracking the Brexit referendum by comparing searches with the words “remain” and “leave.” Here’s his most recent summary, with a day to go before the vote:

He writes that “the remain side is still ahead but that the Brexiteers are catching up in a hurry, after their side lost a lot of traction in the aftermath of the tragic shooting death of Jo Cox.”

One challenge with this approach is figuring out which Google searches for these relatively common words are relevant. Askitas writes: “With some luck and adjustment we may be able to … clean the searches of irrelevant ones,” suggesting that these search data need some cleaning. Another question is the extent to which searchers are representative of voters. We have no real way of knowing.

That said, this is an interesting way of getting a real-time estimate, and these estimates did well in the Irish and Greek referendums. Moreover, it’s not as if conventional polls are perfect. I do think this work should be part of the conversation. And it’s good to put this up for public scrutiny, to motivate improvements on these methods for the future.