Imposters among the believers… fakers and spammers “following” party leaders

In the race to spread their political messages and win over voters during this year’s election, all of the leaders and their parties are making a mad dash to see who can gain the most number of followers, “Likes”, retweets, and comments, etc. on social media.  For many in the mainstream media, these absolute numbers are being reported with baited breath, as if they can somehow serve as an indicator of leader or party popularity among voters. But in the world of social media, a raw number of followers alone can be very misleading. This is especially true on Twitter, the social media version of a soapbox, where most messages and accounts are free and open to the public.  To shed some light on these often reported numbers, we decided that it is time to investigate who actually follows our party leaders; how many are “actual” and “loyal” followers and how many are “imposters”.

If we simply look at the raw number of “followers”, all five party leaders have been steadily gaining support on Twitter, adding hundreds of followers each day. As of April 15th, Stephen Harper had the highest number of followers (~126k), followed by Ignatief (~90k) and Layton (~80k), who are now neck and neck in their number of Twitter followers. Racing to keep up in the rear is Duceppe (~54k) and May (~21k).

But out of the five leaders, Elizabeth May has gained more followers percentage wise (55%) than any other party leader, relative to how many followers they had on March 29th. This is not surprising if we consider that May and her supporters have been especially active on the Internet, encouraging people to participate in online discussions around last week’s debates and other election issues. Also, May’s style of using Twitter and other social media is very conversational in nature which could also have contributed to her gaining many more followers. On the other side of this scale are Duceppe and Harper who gained the least number of new followers (percentage-wise) relative to how many followers they had on March 29th. This may suggest that they might be approaching the upper limit of online support. (See the chart below.)

The Loyal Supporters

Next, we tried to find out how many of each leader’s followers are actual supporters, undecided, or are simply spams or fake accounts?  For the purpose of this test we define “loyal” or “dedicated” followers to be those who follow only one of the party leaders.  To estimate the number of actual party supporters (“loyal” or “dedicated” accounts) for each leader, we counted how many users only follow a single candidate (See the chart below).

Interestingly, Harper has the highest number of loyal followers among the five party leaders; out of ~126k users who follow Harper’s Twitter account, 51% don’t follow any other party leaders. Duceppe came in second with 39% of loyal followers. While Ignatieff, Layton, and May have about the same number of “loyal” followers, estimated in the 27-29% range.

The Undecided

Next we considered only users who follow any two leaders at the same time, but don’t follow the other three. Here we made an assumption that these users are either undecided between two leaders or follow one leader to support and another to attack. In either of the two cases, the number of shared accounts between any two leaders can be used to assess the level of competition between any two leaders through the eyes of Twitter users. The chart below shows for each party leader, the percentage of his/her followers who also follow another leader.

Based on the counts, Harper’s main competition is Ignatieff (11% of his followers also follow Ignatieff); for Ignatieff – it’s Harper, for Layton – it’s Ignatieff (and then Harper), for Duceppe – Layton, and for May – primarily Ignatieff and Layton. These results are not very surprising; they reflect the rivalries that we all saw on display last week during the leaders’ TV debates. (Except maybe for the Green party leader who criticized all three main parties; however, based on the chart, she competes for followers more with Igtanieff and Layton than with Harper.) What is interesting is that we can see, based on Twitter data which only represents a very small proportion of the Canadian population (skewed towards a younger demographic), that we still can detect underlining trends of this election. Another interesting observation based on this data is that when we added up all of the percentages for each leader, Ignatieff had the largest proportion of followers that also follow other leaders (~29%) which puts him in the most contested position, being challenged for 15% of his followers from the right by the CPC, and 8%  from the left by the NDP.

Imposters among the Believers

Finally, the most interesting result came up when we try to find “fake” and spam accounts from among the “loyal” supporters of each of the leaders. Due to the open nature of Twitter, there are a lot of different bots (automated programs) that can easily create fake user accounts and start following real accounts in an attempt to disseminate various types of information or simply increase one’s number of followers. One of the main characteristics of bots is that they tend to “follow” many accounts, but very few would follow them back. In order to assess how many bots follow each party leader, we calculated the Follower-Friend Ratio (FFR), for every “loyal” follower, as defined earlier, for each of the 5 party leaders. FFR is a ratio of the number of followers to the number of friends (accounts that are being followed back). For example, if a user follows 100 accounts, and only 50 accounts follow that user back, then their FFR = 50/100 = 0.5 which is considered low. Most of the “human” accounts have a ratio close to 1 (following as many accounts as they follow back) [1-2]. Celebrity-type and news accounts would have FFR higher than 2. For example, Harper’s FFR = 10 and Ignatieff’s FFR = 8. [3] For the purpose of our analysis, we flagged all accounts with FFR less than 0.1 as potential bots. [4] Below is the chart with the results.

Of all the leaders, Duceppe has the highest number of followers which exhibit bot-like behaviour; 41% (!) of all of his “loyal” followers have FFR<0.1. Elizabeth May has the least number of bot-like followers (only 9.6%). Harper, Ignatieff, and Layton have between 17% to 22% of such accounts. (It should be noted that there are likely some real users with FFR<0.1 who were swept in with the bots accounts, for example, people who just joined Twitter and didn’t have enough time to “collect” followers. But even if we allow for a 5% or even 10% error rate, the results are still very convincing.)

This suggests that of the five party leaders, May pays the most attention to who follows her on Twitter and Duceppe the least. It also suggests that whoever is maintaining May’s twitter account is actively culling and weeding the Green Twitter follower base to prevent spammers from taking root and that Duceppe’s Twitter account is open to all followers whether they be “human” or “bots”. Further analysis is necessary to confirm whether or not this is a case of a computer-savvy supporter or detractor creating fake accounts to follow Duceppe’s account and driving his numbers up, or is it simply the case that Duceppe’s social media team decided not to clean up their Twitter account from bots and spammers who can easily and quickly take over one’s followers’ list.

In sum, even when considering only “loyal” users and non-bots, Harper is still in the lead in terms of the number of followers on Twitter, and May leads in two other categories: (1) the total number of followers gained relative to how many followed her at the beginning of this election campaign and (2) she also has the least number of bot-like accounts who follow her. Ignatieff has a lead of only a few percentage points over Layton in all categories.

This analysis raised some interesting questions for all political parties. Should leaders and parties actively cull their list of followers on Twitter to remove fakes and spammers- type accounts? Or, should it be a free for all that is open to all?

Stay tuned for our continuing analysis of how the election is being played out on Twitter and other social media platforms.

 

* Sreejata Chatterjee assisted with the data collection and analysis, and Philip Mai contributed to the writing and analysis for this post.

Footnotes:

[1] Chu, Z., Gianvecchio, S., Wang, H. and Jajodia, S. (2010). Who is tweeting on Twitter: human, bot, or cyborg? Proceedings of the 26th Annual Computer Security Applications Conference (ACSAC ’10). ACM, New York, NY, USA, 21-30 : 10.1145/1920261.1920265

[2] There are also some more intelligent bots that may have a FFR ratio close to 1. These were not considered in this analysis.

[3] Try measuring your FFR at http://tffratio.com

[4] This test was only performed on public accounts. Followers with unavailable or private profiles were not included into the analysis due to the restricted access to their user data; as a result, the final counts may overrepresent bots and spammers since their profiles are almost never private; but because it’s skewed in the same direction for all of the five party leaders, it still shows a general trend.

 

This entry was posted in Uncategorized. Bookmark the permalink.

Comments are closed.