The exact number requires to do operations that are non-trivial at scale: you have to check who accepted to share information with that version of the app (ther

The exact number requires to do operations that are non-trivial at scale: you have to check who accepted to share information with that version of the app (there were several) and take into account the accounts that have been deleted in the last four years; whether Facebook can access that particular information at all is unclear to me (I would believe not). Then you have to recreate a list of all their friends at the time (which is, again, not trivial with some friendships being deleted) and take care of duplicates -- doable, but not trivial at that scale. Whether people are American is also non-trivial: I’m not sure there was clear geographic entities to make that easy in 2013. People move and Facebook only has your location, not your passport. You then have people without a recognisable name, who forgot or who are too young to be in the voting registry. Do you count all the IDs, those who were still active in November 2016, those who match local registries? People who specialise in data (like people who code professionally) are precisely those who would have all those questions on the top of their mind and know that old, graph-based, editable datasets that need to be matched with another dataset are bad. Edit: I was a data scientist for Facebook and I can personally attest that most of those are genuinely hard, especially those that I intentionally overlooked.