Strange patterns on my Facebook

by darribas

Today I was in for some self-data hacking so I went to Facebook (Settings -> Account Settings -> Download a copy of your Facebook data) and downloaded all the information they hold from me. I’ll save you the boring stuff but I’ll say it’s not too difficult to grab all the time stamps from your wall. This means you can look at how activity (posts, replies, etc.) is distributed across time, and you can get pretty fancy because it’s recorded down to seconds. However you don’t have to get that detailed to get interesting things to pop up. Here is the volume of activity over last year, measured by week:


You have interesting trends, maybe seasonal ups and downs, bla bla… But what is that peak towards the final weeks of the year that goes more than twice as high as the second busiest week? Let’s zoom into that particular week to see if it helps:


Mmm… the 7th. and 8th. of November look like the epicenter of this “burst of socialness” in my life. At that time, I was living 9 hours behind the good old Europe, so maybe one could argue actually those two are related and somehow part of those posts, likes, etc. on the 8th. are leftovers or people who “got late to the party”. Anyway, what could it be? What would make my Facebook activity go bananas to that level? In applied Economics (as in most fields that deal with data), whenever you are stuck, rule number one is bring even more data to see if they show you what you are not seeing. So here is this additional “dataset” I came across:


Now, for those academic geeks, that figure seems to me like the killer feature that gets the paper into the top notch journal of choice. For the rest of mortals reading this, I will just say thank you very much for being some part of my life and sticking to me over the years, even though “close” is probably not the best word to describe our physical locations. You are all somehow somewhere with me, though.

Instead of this, today I could have also replied every single of all those nice posts on my wall one by one. But if the pattern of last year shows up again (and I can tell it will, by midnight the count is on 65), who would want to spend hours writing back in boring Spanish and English when you can use that time to write in Python?


  • pandas made it so much easier, faster and fun to process the data from the Facebook wall into something that could be plotted.
  • Matplotlib made the plots possible and this piece of code made them fun.
  • My mom and dad made the DGP, which is pretty cool.