The evolution of South African politics (as seen through Twitter)

I’ve been researching South African politics on Twitter for several years now. As we start a new year, I thought it worth wrapping up some of the things that I’ve observed into a single post. A previous post of mine drew quite a bit of attention including a dox attack on myself and my family. The attack got me thinking back over the past few years and how politics in our country has changed in that time. In doing so, I realised that there is quite a clear thread to be drawn that brings us to where we are now, and I realised that I had many datasets to illustrate the points along the way. Here then are my thoughts (and this really is just my unpaid interpretation) on how South African politics have evolved over the past few years, drawing on analyses of millions of tweets.

2012: The age of (relative) innocence

My first piece of research into the South African Twittersphere was done in 2012 in collaboration with a couple of folks (message me on Facebook if you’d like a copy of the paper). The dataset that we used was a bit different to all subsequent datasets in that it attempted to look at the entire South African Twittersphere; not just within the context of politics. This holistic view makes it a good jumping on point for contrasting what the SA Twittersphere looked like then versus now. We also used a slightly different methodology in that analysis compared to all subsequent analyses: rather than constructing a network based on who interacted with whom via retweets and @mentions, we looked at who followed whom (I moved away from this methodology in subsequent research because collecting all this follower data is too intensive and actual interactions carry more useful information than passive following links). The 2012 analysis gave us a unique snapshot of the overall Twitter picture at that time. The network was broken up into these five main communities:

Community breakdown of the 2012 South African Twitter follower network
Community breakdown of the 2012 South African Twitter follower network, showing the proportion of users that fell into each of the top five communities (as identified using the Louvain modularity algorithm).

Unsurprisingly, sportsmen, television celebrities and news media had the largest communities since most people use Twitter to keep up to date with their favourite public personalities. What was surprising is that there was little political debate at the time compared to today. The DA were early adopters of Twitter and had their own community that they engaged with but the ANC still appeared to be the de facto home for most black voters and it still enjoyed a relative moral high ground (although this had already started being eroded by the scandals surrounding President Zuma and then Youth League leader, Julius Malema). President Zuma even had his own prominent account, @SAPresident, which seems unthinkable today given the preoccupied, distant commander in chief that we now have (his last tweet was in 2013).

…at this point in our country’s politics, the ANC was still the de facto home for black voters and the ANC still enjoyed a relative moral high ground

It’s difficult to harken back to such a time nowadays but the EFF and the far left sentiment that they brought into the mainstream did not exist to a large degree in our political discourse in 2012. Indeed, @Julius_S_Malema was still a parody account that had not been handed over to its namesake yet.

Here’s what the overall follower network looked like then:

South African 2012 Twitter follower network. Node size based on number of followers
South African 2012 Twitter follower network. Node size is based on number of followers. The differently coloured regions represent distinct communities as identified using the Louvain modularity algorithm.

Sticking with 2012, below is an image of the same overall follower network as above but in this version I’ve sized the user nodes based on a metric of influence known as ‘betweeness centrality‘ instead of based on how many followers they have. This allows us to see which users, despite perhaps having fewer followers, were the gatekeepers of the flow of information between communities. In this view, we can see social media stalwarts such as Khaya Dlanga, Simphiwe Dana, Mandy Wiener, Ferial Haffajee and Stephen Grootes were very prominent even back then.

The most dramatic changes occurred in the DA camp. In some ways, it’s difficult to recognise the DA of today compared to 2012. Musi Maimane was less prominent at that time than Lindiwe Mazibuko and strategist, Ryan Coetzee; neither of whom are with the party anymore (Coetzee went on to play a role in UK politics where he most recently directed the campaign for the UK to stay in the EU and Mazibuko went on to study and work at Harvard University). This gives us a brief vignette of just how much the DA has changed in the intervening years.

The most dramatic changes occurred in the DA camp. In some ways, it’s difficult to recognise the DA of today compared to 2012.

South African 2012 Twitter follower network. Node size based on betweeness centrality
South African 2012 Twitter follower network. Node size based on betweeness centrality.

2014: President Zuma splits the ANC’s support on Twitter

I collected about a million tweets in the three months leading up to the 2014 national elections. Another colleague and I analysed and wrote up a conference paper on this research (also available on request). That piece of research clearly showed that four main communities of roughly equivalent size dominated the election discussions on Twitter. More than half (52%) of all users talking about the elections fell into one of these four communities which collectively generated 85% of all tweets about the elections. Two of the communities were very easy to describe: one related to liberals and included the DA accounts in it; the other related to the EFF. However, the last two communities (the large yellow and green ones below) were a bit more challenging to figure out:

Four communities dominated the election discussions in 2014. Two of them clearly related to the EFF and liberals (including the DA) but the other two were more difficult to describe initially.

To understand what defined the two remaining communities, I started by looking at their network structure (not shown). The large yellow community above demonstrated a high level of interconnectedness between its members which tells us that they were particularly engaged. Members of this community were talking and debating back and forth between each other about the upcoming elections. This community also lacked any key influencers or mouthpieces. Instead, the conversation was decentralised and distributed among its members. This was clearly a community built on passion and ‘street-level’ engagement. At first, for lack of a better description, I labelled them “Young, black influentials” based on what little I could glean by scanning through some of the accounts involved.

The last unknown community in green above seemed to be centred around official ANC accounts. It was less interconnected and therefore less engaged. Many users simply retweeted content from official ANC accounts without engaging in the same level of debate and conversation as the yellow community had. This last community seemed more like a broadcast community, passively consuming and passing on whatever party content was put out there by official ANC accounts and senior members who were essentially preaching to the choir. As a side note, this was the last time that I would see an overtly ANC-aligned community in a dataset. From this point onwards, Black Twitter would fracture into a myriad sub-communities.

…this was the last time that I would see a clearly ANC-aligned community in a dataset… Black Twitter would fracture into a myriad sub-communities.

My first thought was that the “Young, black influentials” community represented a new constituency that sat outside of the main political discourse. Unlike any of the other partisan communities, who each shrilly supported their chosen party, this community was far more even-handed. It debated the pros and cons of the ANC, DA and EFF equally. It was a great community to see emerge and seemed to me to represent the hope for a mature future political discourse. On further digging (using Latent Dirichlet Allocation topic models for the methodologically curious), I came to realise that this community’s natural home had been the ANC but they were torn over their historical allegiance to the party and their poor estimation of its current leader, President Jacob Zuma. They were young, their vote was up for grabs and they passionately engaged with each other as they tried to figure out where their political home now lay.

How had this community come to be? My hypothesis is the following: 2014 was the first real test for Zuma’s ANC since taking power in 2009. Coming off the back of repeated scandals, real damage had been done to President Zuma and his ANC administration’s reputations, which had lead to a crisis of confidence among members of this Twitter community. They still had an affinity for the ANC but the internal conflict created by these scandals lead them to consider other political parties. Indeed, it is from this young, engaged community that I believe the DA and especially the EFF gained many of their new voters in the 2014 general elections and 2016 municipal elections.

Based on my unpacking of these two initially confounding communities, I updated their descriptions to “Pro-ANC, anti-Zuma” and “Pro-ANC, pro-Zuma” respectively. The two communities had roughly the same number of Twitter users in each, hence we could really say that Zuma had split the party in two, at least on Twitter (although we know that Twitter is not representative of the national population, where we saw the ANC lose fewer voters overall).

Here then is what the final network map looked like for the 2014 elections, covering roughly one million tweets:

Twitter interaction network based on 981,878 tweets covering the period of 3 March – 12 May 2014
Twitter interaction network based on 981,878 tweets covering the period of 3 March – 12 May 2014

2015: The disenchanted youth coalesce as Woke Twitter

Appearing to have been made politically homeless by President Zuma’s ANC post the 2014 general elections, a large swathe of the electorate found itself adrift within the South African political milieu. Coupled with steady increases in access to the internet and platforms like Twitter, 2015 saw many engaged, young South African voters enter the debate without a natural political home nor a political or social narrative to define themselves by.

Appearing to have been made politically homeless by President Zuma’s ANC post the 2014 general elections, a large swathe of the electorate found itself adrift within the South African political milieu.

This state did not last long though. Movements such as Rhodes Must Fall, Fees Must Fall and the EFF’s own brand of socialism, catalysed frustration over a lack of an empowering narrative into something new. Young, disenchanted black South Africans were given a new voice. In the process, many ideologies that had never really gone away were rediscovered and reinterpreted, particularly those relating to Africanism and nationalism.

The ANC has always been a party of multiple competing ideologies, where the non-racial ideals of Mandela, Luthluli and Tambo had eventually won out in the past. However, views of black nationalism espoused by the ANCYL’s Anton Lembede in the 1940s, the Pan Africanist Congress (PAC) splinter group in the 1960s and elements of the Black Consciousness Movement of the 1970s re-emerged as part of this process. These ideas were revisited within the context of Ghana’s Kwame Nkrumah’s Pan Africanism and Frantz Fanon’s writings on decolonisation. Old heroes such as Patrice Lumumba, Julius Nyerere, Samora Machel, Thomas Sankara  and other African anti-colonialists were lionised by a new generation. All these African ideas further cross-pollinated with those of the Black Lives Matter movement in the USA. These factors (and more I’m sure) came together to give many young South Africans an empowered new narrative and self image. It would seem that, as far as many born free South Africans were concerned, these ideas resonate much more than those that won the 1994 liberation but which have now stuttered to a halt under the current ANC. Thus, we saw the coalescence of feelings that had sat latent in many young South African into Woke Twitter in its many incarnations.

…many ideologies were rediscovered, particularly those relating to Africanism and nationalism.

This network map summarises conversations around the initial Fees Must Fall movement in late 2015, covering about 370,000 tweets (see my article on Fees Must Fall at the time here and my post on the 2016 Fees Must Fall movement here). Many different interest groups were already evident at this stage:

Twitter interaction network from the 2015 #FeesMustFall dataset based on over 370,000 tweets in October 2015
Twitter interaction network from the 2015 #FeesMustFall dataset based on over 370,000 tweets in October 2015

2016: The age of partisanship & propaganda

Going into 2016, South Africa’s political landscape sat along a fairly clear continuum between liberal- and socialist-leaning groups as the below network map of 1.5 million South African politics-related tweets from this post summarises. Things still made some kind of sense.

Twitter interaction network based on several combined politics-themed dataset, incorporating over 1.5 million tweets between Jan-Jun 2016. Originally shared in this post:
Twitter interaction network based on several combined politics-themed datasets, incorporating over 1.5 million tweets between Jan-Jun 2016. Originally shared in this post.

However, by the time that the second wave of Fees Must Fall protests fully erupted in the latter half of 2016, the ground had shifted again. Where the Fees Must Fall movement in 2015 had given many young, mostly black South Africans a platform for their frustrations, the next several months had allowed the movement to solidify and fracture into many different interest groups, from those concerned with speeding along the realisation of a truly a non-racial society to those seeking a radical, nationalist – violent if necessary – revolution, with many views in between. This splintering was likely further facilitated by ever more South Africans with differing views joining Twitter after the events around Fees Must Fall popularised the platform as a venue for political and social change. It is no longer the preserve of students and the technologically privileged.

Also in the latter half of 2016, the municipal elections demonstrated the maturity of our politics as we moved from an essentially one party state (the ANC) to a multi-party state where coalition politics suddenly became vital (indeed, the DA historically wrested several key metropolitan areas away from the ANC due to the coalitions that they formed). Things were looking good for the stability of our democracy as the various political actors now appeared to be counter-balanced against each other. However, one can never under-estimate the lengths that vested interests will go to push their agendas.

Subsequent to the elections, we’ve seen our politics change gear yet again. Our media landscape has shifted away from (at least the pretense of) objectivity. We have the Naspers, Primedia, Times Media, Sekunjalo/Independent Media and Gupta Media empires, among others, all competing to control our national narrative. Both Sekunjalo/Independent Media (Cape Times, The Argus, IOL, The Star, The Mercury, Daily Voice, etc. and the African News Agency syndication service) and the Gupta Media (ANN7 and The New Age) empires have left the Press Council and are no longer beholden to its self-imposed code, further increasing the splintering of our media landscape. Combine these developments with the ongoing inquiry into the alleged capture of the SABC, our national broadcaster, and we no longer have a common media. Our media dynamics are starting to look as partisan as the United States’.

Our media landscape has shifted away from (at least the pretense of) objectivity.

In addition, there are parties who no longer feel it is necessary to work through established media channels at all (naively, because they do not think that their voices are represented; more cynically, because it does not suit their agenda). We have started to see the propaganda techniques of ad hominem attacks and obfuscation pioneered in Russia take hold in our own country. We have seen fake news arise in multiple ANC-related scandals such as the most recent “War Room” scandal (see here and response here) and the earlier City Sun scandal, and we have seen propaganda botnets used in a way that seems aimed at redirecting focus away from controversial individuals.

Example posts from the City Sun fake news website
Example posts from the City Sun fake news website

Politics in South Africa have never been dull but they have always been about ideologies and, at the end of the day, the ideas of good men and women tend to have won out. These things come in waves though and the current ideological era is near an end. In the absence of a (mostly) objective, self-regulated media, we all get to choose what we believe as some groups work overtime to re-frame our politics. We’ve always been a highly racialised society but we’ve also strived (or at least paid lip service) towards the post-race ideal of Tambo, Luthuli and Mandela. However, nationalist ideologies and calls for revolution by any means, including potentially violent, are now being given a national platform and a hardening stance against national reconciliation is being legitimised in the process. Similar to how Europe and the USA are facing hard moves to the right and resurgent nationalism, we are seeing a hard move to towards a new nationalism in South Africa.

Similar to how Europe and the USA are facing hard moves to the right and resurgent nationalism, we too are seeing a new nationalism in South Africa.

Where we go from here remains to be seen. What is clear is that, with the most recent attempts to steer the debate from state capture by President Zuma and the Gupta family to state capture by ‘white monopoly capital’ through the reintroduction of the ABSA lifeboat saga, we are going to see a further fracturing of the political landscape, and thus further undermining of our democracy, as big business is also drawn into the fray and our institutions continue to be whittled away at. This only serves to further the interests of those who benefit from obfuscation. It’s very important to emphasise that the issues they raise are often legitimate. In a more moderate political climate, we could explore and expose the nuanced complexities of these situations but, in the current climate, they will only be used for partisan self-interest and to effect greater polarisation in society.

The rules of the game have changed but I have faith in the resilience of our beautiful country as we continue to navigate the evils of the past and face up to the predatory agendas of the future.


  1. Kim

    I would love to know more about how you conducted your research. This is an excellent piece of writing. Like you, I believe in the strength of the average SA citizen, who has been though too much not to have a stake in our political debate. But I must admit I am afraid that they might not win out, simply because the picture is not clear enough yet. It’s still too hard for people to see.

  2. Pingback: Sudafricanos no quieren a su gobierno intervenga en redes sociales | Alter News Africa

Leave a Reply

Your email address will not be published.