Rumors have circulated since the weekend that Romney's 100,000-strong spike in Twitter followers consisted mostly of auto-generated Twitter accounts, not real users. We've done a new analysis that confirms it, using data from Twitter's API.
Irregularities of the spike day surge
First, almost all of the Twitter usernames on Romney's spike days followed a predictable pattern strongly suggesting that they were auto-generated. The usernames consisted of a first or last name plus a string of three to four random characters. 96% of Twitter usernames on the spike days had this property, compared to 50% on regular days.
Here's what the names and usernames look like:
Kimber Martin @martineksf
Levi Butler @butlerqbjf
Mark Fowler @markiffxy
Shirleen Bullock @bullockqoww
Quinn Perez @quinncjhu
Sommer Mcclure @mccluredkh
Juanita Wallace @juanitajft
Eliana Knapp @knappzwxc
Christopher Presley @PresleyTopher
Chris Cowan @Canadianmeat420
Andrew Walker @awwalker24
Nicolae Ionuţ @I_am_free2012
Rob Lutters @siempreNOkool
On the spike days of July 19th and 20th, thirteen times the usual number of Romney's new followers were were users who had just signed up to Twitter, shown in the graph above. That trend is also a huge red flag that new accounts were created for the specific purpose of following Romney.
But who is the culprit?
The fact that a large number of Romney's new followers are most certainly bots doesn't mean that his campaign organized them. Far from it: some people are speculating that the whole operation is so ham-handed that it must be the work of an opponent out to discredit Romney. Romney's campaign has denied buying the followers and asked Twitter to investigate.
We should add that Romney is far from alone in attracting fake followers. Internet search company PeekYou estimates that only 20-30% of politicians' Twitter followers are real.