Big Data is a Big Deal: The Greferendum as a case study

I started looking at the Greek referendum and Google Trends early last week after a chat with my good friend Nikos Tachtsidis, a staffer at the Greek Presidency, in Athens. How could we see what’s going on? Polls are unclear and also not necessarily unbiased or trustworthy. I started thinking about how I could put my ear on the ground to hear what’s happening.  I looked at Google Trends but my first attempt I thought did not reveal anything of substance not good enough to use it for forecasting. Then I started looking closer. By Friday I had locked in my prediction with close friends: it would be over 60% NO. Most did not believe me but I asked them to remember I told them so.

On Sunday I had a tweet back and forth with Mark Carrigan:

and at 11:52 AM I wrote my coauthor Klaus F. Zimmermann:

Dear Klaus,

I am currently calling it for a NO based on Google Trends. The searches for no or yes are relevant to the referendum and the numbers suggest more than 60% No.
Caveat: The poor and the old do not surf and most of the Greek internet activity comes from Athens.



Shortly before voting closed I tweeted:

I called N. Tachtsidis shortly after.  He had just seen the first exit polls which were talking about a slight advantage for NO of one or two percentage points. I insisted and reiterated to him: “I see 60% or more for NO”. He was in disbelief. Shortly afterwards I could tweet what I had just heard on Greek TV:

Nikos called, he was excited, I upped the steaks: “It is going to be more than 60% for NO” I said. This morning Final Result: 61.31% NO.

The hard part where every actor has to decide what to do with this result is just ahead of us. Mr. Tsipras was suitably and quite intelligently humble in his first public appearance after the result was known last night. Now the creditors need to swallow their pride and transcend themselves as well. In the meantime Mr. Samaras and Mr Varoufakis have resigned. All quite positive developments one might argue.

This entry was posted in comment, data and tagged , . Bookmark the permalink.