Violence Looms Over Closely Contested Elections in Kenya

[Note: This analysis was conducted with open source data up to and through March 2, 2013.]

Since violence erupted after Kenya’s elections in 2007, a combination of wariness about the political process, the growth of corruption, and the spectre of inter-ethnic tensions continue to haunt the country. With much anticipation and attention on Kenya’s upcoming elections, this post uses Recorded Future data to analyze possible triggers of unrest and political instability.

Hindsight as a Hindrance

A key element of using massive amounts of open source data is to go in without confirmation bias; if the analyst is truly dispassionate, the data will reveal a trend, instead of the analyst trying to select data to support a hypothesis. In this case, the operating assumption is that Kenya will experience some level of violence — it is a matter of location, degree, driver, and breadth — not of certainty. Unfortunately, this assumption is supported by the data; there is little examination or example of stabilizing forces.

Before forecasting where and how unrest might be incited, it is important to objectively look back at the past several months — without the color of the looming elections. Time is a powerful analytical pivot point; in looking back at coverage of Kenya in 2012, there is a clear peak in late August, when there was a strong popular responses to the killing of a local cleric.

Locality Matters, as Does Proximity

While the cause of that event may be important, it is critical to understand its reach and preceding trend of protests. By categorizing the data by event, a clearer picture emerges — rather, that this series of protests in late August stand out.

Re-constructing this query to include more recent data (through March 2, 2013), show the cities, and categorize by organization reveals that there is a wide distribution of entities and localities. After digging into the data, it is clear that protests and rallies revolve around either political transparency, perception of corruption, and foreign intervention. Although ethnic tensions don’t feature prominently, it is certainly a recurring secondary theme.

Tensions: Incited or Incipient?

Atypicality of unrest is high in the town of Garissa. Not only was there a bombing in October of 2012, but a recent bombing highlights the volatility of this border town. Despite the violence, a key element is the cause — as a town on the border with Somalia, the role of al Shabaab (perceived or real) is undeniable.

With a query comparing coverage of candidates Uhuru Kenyatta and Raila Odinga, explore the interactive visualization below to evaluate if there is a pattern of communication about the topics and frequency of discussion. Try hovering over each of the event types and comparing distribution over time.

One pattern that emerges is that both candidates speak frequently about a location when there is high-profile incident of violence. The two differ, greatly, however, in their narrative — between political legitimacy and the role of foreign entities (ICC, al Shabaab, etc.). By looking at Garissa, Odinga has made numerous statements about the recent bombing, using it as an opportunity to highlight which changes should happen under his administration.

Conclusions: When, Where, and How to Watch

Given the anticipated closeness of the election results, the incentive to create leverage — to manipulate the polls — is very high. Paradoxically, both sides will seek to take advantage of any situation to do so, which could lead to one of two dynamics: either a manipulated, but still close result, or a grossly uneven one, the legitimacy of which would be questioned by the loser. It is clear that even though Kenya’s electoral process is under great scrutiny, the “window to watch” is after the polls close and the results (or rumor of results) begin to trickle in.

Although there are many drivers of possible post-election violence, this analysis focuses on those that can be evaluated, tested, and monitored through open source media coverage. The likelihood of post-election violence, sadly, is high — however, its impact varies greatly depending on three critical factors: geographic distribution, contestation of results, and rhetorical tenor.

• Contestation of Results encompasses a variety of political factors, foremost the legitimacy of the results. Whether there is evidence of manipulation or not, if either candidate openly and immediately questions the outcome — worse, if he calls his supporters into the streets — the popular perception of post-election stability will quickly erode.
• Rhetorical Reference can have many interpretations; here it applies specifically to the response to the speaker (e.g. the candidate) and the topicality. Whether Odinga or Kenyatta dominates discussion and which topic (electoral transparency, ethnic alliances, foreign intervention, etc.).

Depending on how one combines these combination of factors, there are 6 to 8 conceivable* scenarios (If you are a sharp analyst, this should catch your attention. Here, I tried to limit the discussion beyond scenarios with low/medium impact and low likelihood.), for which three primary triggers are worth watching closely considering for their outsized impact a social unrest scenario:

Scenario 1: Kenyatta is declared the winner by a margin. Border towns exhibit localized violence and, if driven by Luo-Kikuyu violence, escalate up to the national level.

Scenario 2: Kenyatta-Ruto alliance wins overwhelmingly and Odinga suspects manipulation, in response to which Odinga openly and immediately mobilizes supporters into the streets.

Scenario 3: Odinga is declared the winner and Kenyatta-Ruto alliance unsuccessfully use force to maintain political credibility. Tensions in the Jubilee Alliance (between Kalenjin and Kikuyu) echo 2007 violence in the Rift valley.

Again, given the attention that this election has already garnered, analyzing the situation real-time requires monitoring focused factors. For example, one dynamic that is difficult to understand is a self-fulfilling perception by observers and the Kenyan population of a repeat of previous post-election unrest. As a result, during and after the elections on Monday, this analyst will be watching the stream of information for pre-determined flashpoints.