First Experiments in Estimating Age of Twitter Audiences

I’m working on an algorithm that will give me the estimates of the ages of people who tweet.  It’s still in the process of being refined, but I thought I’d share some preliminary results:

20130929 ages

Click for poster sized version!

I took the data from my Cable News Twitter Ratings for 9/29, and estimated the age of the sender of every tweet.  Some I could not, but a large percentage I could estimate.  The chart (you really want to click on it) gives the breakdown of the ages for each of the networks.

E! News and FXX (Totally Biased w/ W. Kamau Bell) tend to skew the youngest over all, while of the traditional networks CNN and HLN have younger correspondents in the Twitterverse than MSNBC, Fox News, and Real Time with Bill Maher on HBO.

Some important notes:

First, this represents ages of people who tweet, not necessarily people who watch.

Second, what’s being added up is the age per tweet, not per sender. If you’re 30 and send one tweet and I’m 50 and send two tweets, it will show 33% 30 year olds and 66% 50 year olds.  I will probably change the logic behind this to only count each sender once, but for now that’s what I’m doing.

Finally, the whole process is still being refined, and I expect better answers to come out of the program over time, but it’s still interesting to see the picture as it starts to form.