Thursday, October 27, 2011

Can Google predict Ireland's next president?

Today Ireland is holding its presidential election, so I wondered if I could use Google Insights for Search to predict the results.

Insights for Search shows what terms Google users are searching for, in a given area and over a given period of time, as a proportion of all terms searched. We can use this to compare the popularity of presidential candidates in terms of the number of people searching them on Google. The candidates are:

Mary Davis
Sean Gallagher
Michael D. Higgins
Martin McGuinness
Gay Mitchell
David Norris
Dana Rosemary Scallon

Insights allows us to compare only five terms at a time:
An immediate problem is clear by the peak in searches for Dana Rosemary Scallon in mid-October. On October 12 Scallon made some puzzling comments during a television debate, complaining about allegations of a 'malicious, vile nature' made against a family member. This weird outburst put Scallon into the news for a few days, which we see reflected in the Insights graph. But the searches here do not imply electoral popularity.

Google Insights for Search cannot distinguish between popularity and notoriety.

Nevertheless, let's carry on. The lowest two search terms of the above five in the latest date available (October 25th, two days ago) are Scallon and Gay Mitchell. Replacing them with the other two candidates not mentioned so far, we get this:
Sean Gallagher far out in the lead. Yet here is the same problem as we had with Scallon - Gallagher has been in the news lately because of allegations of corruption and of ties with the deeply unpopular Fianna Fáil party. So is he attracting searches because people like him and want to learn more, or because people are itching to find out about grimy connections in his past?

Let's break it down one more time. Here is Sean Gallagher versus Michael D Higgins:
Gallagher attracts more searches for almost the entire period. Recent opinion polls have put Gallagher in first place, with Higgins some way behind him. Meanwhile bookmakers Paddy Power have Michael D Higgins as favourite at 1/5, compared with Sean Gallagher at 3/1.

So I'm intrigued and tempted by the idea that Google is predicting a win for Gallagher! But he may be merely infamous, not popular. And Google users might be using other terms I have not checked here: 'Higgins' instead of 'Michael D Higgins', 'Dana' instead of 'Dana Rosemary Scallon', 'Martin McGuiness' instead of 'Martin McGuinness'. The two-day lag in Insights data may also be enough to undermine the results.

This is a nice little experiment, albeit a clumsy one, to test Google's predictive abilities. I will come back to this in a few days.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.