KYLE FINDLAY analyses over 370,000 tweets with hashtags relating to the #FeesMustFall movement collected since it first popped onto the social media radar on Thursday 15 October, when the movement began at the University of the Witwatersrand.
The #FeesMustFall movement has enraptured the entire country since it started at the University of the Witwatersrand (Wits) at the back end of the week before last. The resulting social movement has been unprecedented in its scale and earnestness. In addition to taking to the streets, students (and many other South Africans) took to social media to voice their frustration and articulate their demands.
We have been collecting tweets with the hashtags relating to the movement since it first popped onto the social media radar on Thursday 15 October when the movement first started at Wits. From then until the morning of Monday 26 October, we collected over 370,000 tweets, which I’ve analysed below. The analyses that follow are my own interpretation but I do so under the banner of Code4SouthAfrica, a non-profit organisation dedicated to liberating South Africa’s data for the benefit of all.
Figure: Time series showing the hourly volumes associated with the #FeesMustFall discussion. Conversation clearly peaked on Wednesday 21 to Friday 23 Oct.
Who were the main communities in the discussion?
The first thing we did with our dataset of 370,000 tweets was connect people together if they interacted with each other via @mentions and retweets. This gives us a ‘conversation map’ of the entire #FeesMustFall topic. It incorporates 97,669 users who in some way weighed in on the #FeesMustFall discussion, where each node represents a Twitter user and we draw a link between those users who interacted:
Figure : The overall #FeesMustFall conversation map, incorporating 97,669 users. Each user is represented by a separate dot, or “node”, and users that interacted with each other via @mentions and retweets are connected together. Nodes are coloured based on the community that they belong to, leading to regions of the same colour, collectively representing distinct communities. Node size represents the level of influence of that user based on their number of interactions.
Figure : A close up of the centre of the conversation map. We see several distinct communities in different colours e.g. Rhodes Must Fall community members in green, Wits SRC and Daily Vox community members in pink, news media in blue, etc.
We can tell a few things from this conversation map straight away just by looking at it. For one, the conversation was very inter-connected – very few users sit on the periphery of this map; most are pulled into the jumbled hairball at the centre. There was clearly a lot of back-and-forth between users on Twitter which lead to this jumbled hairball, which is unsurprising given the impassioned nature of the topic.
Next we can see that there are actually distinct groups of people, or ‘communities’, that made up the conversation. Users are coloured based on the community that they belonged to as identified using a community detection algorithm. They collectively appear as regions of solid colour on the conversation map. We can quantify these communities based on how many members each contained and the proportion of tweets that each community generated in the entire discussion:
Figure : Breakdown of the top 10 communities in the #FeesMustFall conversation. They collectively represented 64% (62,271) of all users in our dataset talking about the movement and they generated 77% (289,326) of all tweets.
The single largest community in our data formed around @douniatee and is clearly visible at the top of the conversation map. Eleven percent of all users talking about #FeesMustFall in our dataset fell into this community but they only generated 3% of all tweets. This is because most of these users simply retweeted a single tweet of @douniatee’s as evidenced by the half moon of users formed around her node, thus her “influence” mostly came from passive retweeting rather than through real, two-way conversations:
We can also see that, because she sits off to the side of the main conversation map along with her retweeters, she mostly spoke to an independent audience, rather than tapping into the larger South African conversation. This is unsurprising as she is based in New York City and seems to be tapping into the community around the #BlackLivesMatter hashtag. One might go so far as to say that that picture represented the international face of the movement on Twitter.
The second largest community is probably actually the first South African-specific community of note. This community formed around the official Rhodes Must Fall account and related activists. It represented 9% of all users talking about the #FeesMustFall movement and these users generated an unprecedented 21% of all tweets in our dataset.
In addition, we can also see the nucleus of the movement that formed around Wits University and accounts such as the Wits Student Representative Council (SRC) account, Daily Vox (self-billed as ‘South Africa’s youngest news portal’) and related activists. They were very vocal and it appears that their ‘vocalness’ contributed to the initial momentum of the movement. This chart shows the proportion of tweets per hour that came from each of the top 10 communities and we can see that the ‘Daily Vox, Wits SRC and activists’ community in yellow really drove the initial spikes in conversation before it was picked up by the ‘Rhodes Must Fall’ community in earnest from Monday 19:
Figure : Time series showing the proportion of hourly tweets generated by the top 10 communities in our data (as well as all other users in grey i.e. “Everyone else”)
We then also see geographical distinct communities of students and activists forming in Cape Town and Johannesburg.
News media communities are split between Power987News – which has often fallen into communities related to the Economic Freedom Fighters (EFF) in previous analyses) – and African National Congress (ANC) and Gupta-related ANN7 and City Press. Surprisingly, we see Radio702 falling into this community too. One must be careful not to draw too much inference from these community memberships though since community membership is based on interactions, which can be either supportive or antagonistic and which can change as different influencers enter the fray. This news media community stands in contrast to that containing news outlets Eyewitness News, eNCA and News24.
Finally, we see that the ANC, Democratic Alliance (DA) and EFF just scrape into the top 10 communities, with the DA and ANC communities being drawn together through an certain amount of overlap in the users interacting with both. The EFF, in contrast, stands as its own, distinct community.
These groups represent the 10 main constituencies in the #FeesMustFall discussion on Twitter as identified by our algorithms.
When did it evolve into a national movement?
The #FeesMustFall campaign started on the Wits University campus amongst Wits students before evolving into a national movement. By looking at our time series and the relative usage of ‘X must fall’ hashtags, we can see when the inclusive #FeesMustFall hashtag really emerged as distinct from any university-specific hashtag (such as #UCTFeesMustFall or #UJFeesMustFall. This chart shows us the prevalence of the various university-specific variants of the hashtag and the overall hashtag:
Figure : Times series showing the usage frequency of the various “fees must fall” hashtags
What is clear from the above graph is that #FeesMustFall was used consistently from the beginning. However, it all but swept away all other hashtags from Wednesday 21 onwards as the movement gathered steam nationwide, likely spurred on by the confrontation between students and police at parliament in Cape Town. Also worth noting is that, aside from the Wits University variant of the hashtag in orange, other notable variants came from Stellenbosch University in green and the University of Cape Town (UCT) in yellow.
This raises the question of each university’s contribution to the overall discussion. From this chart we can see what kind of volumes each university represented and when they joined the conversation:
Figure : Times series showing hourly tweet volumes by university
Again, what is clear is that Wits spearheaded the discussion and consistently contributed to it, followed in volume by UCT and Stellenbosch, which both really kicked into gear on Monday 21, while the University of Johannesburg (UJ) and the University of the Western Cape (UWC) spiked on Thursday 22 and Friday 23, respectively.
Political parties wade in
The ANC has been criticised for its delayed and distanced response to the movement, with some accusing it of not taking the movement seriously. Does the data support this?
Figure : Times series shown hourly tweet volumes from official political party accounts
If we look at the tweets coming from official party accounts @MyANC_, @Our_DA and @EconFreedomZA we see that no party really picked up on the movement until Monday 21 when the DA was quick off the mark. It joined the conversation a day before the EFF and two days before the ANC. The ANC was the last to join the conversaion, only doing so on Wednesday 21 but when it did, it did so with a vengeance, generating the most tweets of all three parties from then on.
The picture looks a bit different though if we include key party member accounts within each party’s counts. For the ANC, we include Fikile Mbalula and Blade Nzimande’s accounts, for the DA we include Mmusi Maimane and Helen Zille’s accounts, and for the EFF we include Julius Malema and @sentletse’s (self-described leader of the EFF Twitter army) accounts:
Figure : Times series shown hourly tweet volumes from official political party accounts and the accounts of party leaders/influencers
Now we see that someone within the ANC did actually pick up on the conversation very early in the process. Digging into the data, we see that it was actually Mbalula who tweeted on Thursday 15:
Who or what must fall?
We’ve seen how the movement evolved from a Wits University issue into a national issue as evidenced by the ultimate domination of the #FeesMustFall hashtag over university-specific variants. Similarly, we saw the nature of the movement evolve from simply demanding that university fee hikes be abolished to the resignation of Higher Education Minister Blade Nzimande into calls for the downfall of the ruling ANC government:
Initially, the call was for Wits vice-chancellor Adam Habib to step down. However, come Monday 19, as the movement started to gain national traction, its demands evolved into a call for the downfall of Nzimande. This in turn evolved into calls for the downfall of the ruling ANC party on Wednesday 21. Was this perhaps related to the party’s tardy response to the movement or perhaps their response to the students banging on the door of Parliament during Finance Minister Nhlanhla Nene’s interim budget speech, which was met by an intermediate level of force by police?
What is clear is that by the weekend, most calls were for the downfall of the ANC over both President Jacob Zuma and Nzimande specifically.
Where to from here?
What we have witnessed in the past two weeks is the birth of a social movement within South Africa; one born out of a desire for a better life through education by young students, but one which quickly evolved to lay the ultimate blame for their lot in life at the door of the ruling ANC. Through this process, students (mostly) behaved in a peaceful and ordered fashion, clearly exhibiting restraint and strategic decision making on numerous occasions. What is also clear is that this is not a purely racially motivated movement, given to racial partisanship as much of our recent political and social rhetoric has focused on. Indeed, much of the most shared media clearly highlights the role of all races involved, with students of all races making their presence felt on the front line. Just take a look at the top 10 most shared images whose authors and content cross the racial lines from all directions (none more dramatically so than @douniatee’s already mentioned photo of a white human shield which was shared five times more than the next most popular photo):
We clearly have a national youth movement on our hands that has demonstrated restraint and intelligence and which draws wide support from all sectors of society, as it would because its members are the children of our MPs, our police officers and our own. What is unclear is whether the movement has built up enough momentum to keep going forward. If it does, and the data seems to imply that it might, we are in for some very interesting times in our country as it feels like, for the first time in a long time, we have a broad-based movement with the moral high ground that everyone can get behind. DM
Acknowledgements: My thanks to Clint Albertyn for helping to collect the data.
Caveats: a) the 380,000 documents does not necessarily represent every single tweet generated, and b) network community membership is not a confirmation of political ideology nor membership.
Main photo: Incoming Wits SRC president Nompendulo Mkhatshwa leads a group of students outside the Union Buildings. (Greg Nicolson)