I used the tool that Anil mentioned, Twee-Q, and it reported that 83% of my retweets are of men. I was stunned to see that number, so I decided to investigate further.
First, I checked the gender breakdowns of whom I follow. Followerwonk reported 22% female, but also 33% undetermined. I only follow 191 people on Twitter, so I decided to do my own count. I first eliminated group accounts and bots, unless I knew there was a single person behind the group account, as with @WiserEarth (female) or @wikistrategies (male). That brought my following number down to 172.
Of the individual accounts I follow, 62% are male, 38% are female. This is better than what Followerwonk reported, but still not 50-50.
Also, based on this breakdown, I was still disproportionately amplifying the voices of men. However, I didn’t trust the 83% number. First, I didn’t trust that Twee-Q was accurately determining which voices were of which gender, especially given the discrepancy I saw with my Followerwonk numbers. Second, I didn’t think retweets alone were an accurate representation. I sometimes added my own commentary and used a citation rather than a retweet.
I decided to manually count all of my retweets and citations (not counting conversation, only amplification) in 2014. I’m not a prolific tweeter, so this was easily doable.
In 2014 so far, 74% of my retweets and 60% of my citations have been of men. In other words, 66% (two-thirds) of my retweets and citations are of men. The gender breakdown of whom I cite roughly maps to the breakdown of whom I follow, but I definitely retweet more men.
My initial reaction to these numbers was surprise, then rationalization. I won’t bother going into either — I don’t think they’re particularly important. The reality is that there are some well-documented implicit biases in society, and I’m statistically likely to suffer from all of them, no matter how enlightened I think I am.
The true measure of enlightenment is what you do with that self-awareness. I thought Anil’s metrics were interesting, but his followup experiment was inspiring. I’d like to try a similar experiment as a followup; I’m just not sure what that experiment should be. Any ideas?