Daniel Waweru’s post is certainly provocative and there are a number of issues that we wish to raise in response. First, we dispute Waweru’s representation of an argument we made in an article published in African Affairs earlier this year. Apparently we ‘account for the upsurge in violence by appeal to elite fragmentation.’ This is an overly reductive version of our argument. The article was an attempt to identify some longer-term contributing factors to the ongoing political crisis, of which the violence was but a part. As Waweru acknowledges, this process of elite fragmentation is certainly part of the story but nowhere in that article do we claim that it is the whole story.
Our main concerns with Waweru’s piece do not, however, relate to what we or others may or may not have written. Rather it is the simplistic depiction of the nature of political violence in Kenya since the 1990s and the attempt to identify just one cause of it. First, the situation is bad enough without needing to be described as a single two-decade long episode of ethnic cleansing. Second, to reduce all political violence down to a majimboist agenda is a gross over-simplification. What about Mungiki? Or Mount Elgon? And where’s the state in this account? If it was, as claimed controlled by majimboists under Moi, why did Kenya retain its highly centralised state institutions? Third, the definition of majimbo (exclusive ethnic federalism) is not one that is universally shared. The term is constantly debated and there are countless operational definitions in contemporary Kenya. Fourth, while the Kalenjin are critically scrutinised, can the conclusion that ‘Kalenjin opinion leaders who support ethnic violence … lack effective internal constraints’ not be applied to other communities? Finally, ‘going directly to motives’ has no explanatory advantages if those motives are simply those that one individual, without any great evidential basis, chooses to ascribe to a whole range of actors. Waweru may see it as his aim to show that ‘the violence was actually unified in aim’ but without any compelling reason for doing so we are not convinced.

Daniel Branch (University of Warwick) and Nic Cheeseman (University of Oxford).

I’m not a doctor, so it’s an honour to have two actual doctors — and a soon-to-be-doctor — respond at length to my post. Let’s see if I can manage to remain unscathed.

1. So here’s the first issue:

First, we dispute Waweru’s representation of an argument we made in an article published in African Affairs earlier this year. Apparently we ‘account for the upsurge in violence by appeal to elite fragmentation.’ This is an overly reductive version of our argument.

Branch and Cheeseman’s complaint rests on the premiss that I credit them with the following achievement: [Branch and Cheeseman] account for the upsurge in violence by appeal to elite fragmentation
My defence is simple. Attention to the paragraph in which the sentence is located shows that the claim with which I credit them is a distant relative, at best, of the claim they employ in their complaint. Once it’s clear which claim is in play, it will be clear that there’s sufficient textual support to justify my actual attribution.

Here’s the relevant paragraph from my post:

Branch and Cheeseman account for the upsurge in violence by appeal to elite fragmentation. That’s a necessary rather than a sufficient condition. Remember that what’s wanted is an answer to why the violence crossed a certain threshold – why it escaped control of the state. Without an underlying capacity for violence, elite fragmentation need not have violent consequences, and it certainly need not have consequences so violent that the state struggles to control them. Appeal to a generalised diffusion of violence is nearer the mark, but it still underdetermines the quality of the violence in the Rift Valley: if elite fragmentation were sufficient to explain the escape of the violence from state control, then that would have happened in more than one place. It didn’t so, it isn’t. Capacity for violence matters; appeal to majimboist motives is sufficient to predict it.

Whatever else it does, the first sentence mentions elite fragmentation to the debate. The third sentence of the paragraph identifies what is to be explained: the fact that the violence escaped the state’s control. The fifth sentence gives a reason to think that elite fragmentation isn’t a sufficient explanation for the state’s lack of control. In that context, it ought to be clear that the first sentence of the paragraph doesn’t, as Branch and Cheeseman seem to suppose, lumber them with the claim that this particular outbreak of violence was caused by elite fragmentation. Rather, because I’m presupposing that it is common ground that the violence this time was different, and that one of the relevant differences was the state’s failure to get a grip, the sentence ascribes them the claim that elite fragmentation suffices to explain the state’s inability to get the violence sorted.

How could so strange a thought have crossed my mind? Roll tape:

Kibaki was also well aware that the coercive power of many of the key security institutions of today’s state is as strong as – or stronger than – in earlier periods of post-colonial history. What Kibaki failed to realize was that without the elite alliance that had given the Kenyan state its initial strength, and against an opposition remarkable for its unity, the physical appearance of the bureaucratic-executive state was no longer a true reflection of its ability to demobilize opposition. The centre tried to hold as it had done in the past. But this time Kibaki and his supporters found that the centrifugal forces within the Kenyan political system could no longer be contained. (Branch and Cheeseman 2008: 5)

That paragraph appears to justify my ascription to Branch and Cheeseman of the claim that the violence could no longer be contained by the state because the elite had fragmented.

2. Second issue. Rather it is the simplistic depiction of the nature of political violence in Kenya since the 1990s and the attempt to identify just one cause of it. First, the situation is bad enough without needing to be described as a single two-decade long episode of ethnic cleansing. Second, to reduce all political violence down to a majimboist agenda is a gross over-simplification. What about Mungiki? Or Mount Elgon? And where’s the state in this account? If it was, as claimed controlled by majimboists under Moi, why did Kenya retain its highly centralised state institutions?

First, ‘intermittent’, as used in the post, suggests that the violence has proceeded in fits and starts.
Second, I’m not committed to the claim that all political violence in Kenya is majimboist in nature. In the first paragraph of the post, I claim, rather, that the main strand of political violence in multiparty Kenya is majimboist; a claim backed by ample evidence. Kimenyi and Ndungu (Kimenyi and Ndung’u 2005: 128-138) compare incidents of ethnic violence across the decade from 1991 to 2001. On their evidence, the vast majority of politically-motivated displacements, violent incidents and deaths since 1991 have come in the Rift Valley. Nor is their finding idiosyncratic: independent observers since the 1990s have consistently concentrated on political violence in the Rift Valley because that’s where it is at its most intense. Kimenyi and Ndung’u conclude that the bulk of the RVP violence was aimed at ethnic displacement (Kimenyi and Ndung’u 2005: 149-50; 154); a view that finds favour with a range of independent commentators.
The presence of both Mungiki and violence in Mount Elgon is entirely compatible with the claim that Kenyan political violence usually comes in the majimboist flavour.

There’s something to be gained from looking past the state. I might have expressed myself more clearly; the point was that the usual state-invoking explanations seem not to do a good job, because either they underdetermine the violence, or they’re neither necessary nor sufficient to explain the violence. There are different ways in which the state could have mattered. If any one of them is to be useful, it needs to pick out majimboist violence. That majimboists are competing for state power isn’t enough to pick out the violence, since other competitors manage to compete without violence on a majimboist scale; competition for state power underdetermines majimboist violence. Majimboist violence has happened both when majimboists were in power and when they were out of power; non-majimboists have had control of the state without engaging in majimboist violence. Control of the state is neither necessary nor sufficient for majimbo violence. Time to try something different.

Fifth. Suppose the majimboist in power wants to do two things: first, to maximise the effective power of the state while it is under majimboist control, and, second, to avoid an effective post-majimbo state which excludes majimboists. Then the retention of centralised institutions of the state is sensible. The central institution of the centralised Kenyan state is the Presidency. The violent gerrymandering has carved out a set of safe majimboist seats, without which it’s difficult to win the Presidency (Snow 2009: 8). Any future non-majimboist holder needs majimboist support. Institutions outside the Presidency have been systematically weakened, while the Presidency – imitating Sauron in this respect as in so many others – has sought to concentrate power in itself. So it is slightly misleading to say that the central bits of the state have been strengthened: rather the powers of the Presidency have been expanded at the expense of other parts of the state. While a majimboist held the Presidency, this was acceptable since, naturally enough, he sought to maximise his effective power; so long as any future non-majimboist holder of the Presidency is indebted to majimboists, this state of affairs is tolerable for the majimboist. The persistence of centralised institutions doesn’t tell against the majimboist account.

3. Onward to the third question. Third, the definition of majimbo (exclusive ethnic federalism) is not one that is universally shared. The term is constantly debated and there are countless operational definitions in contemporary Kenya.

I take it that the point here is that my definition — characterisation, actually — of majimbo is unsound, incorrect or otherwise dodgy because it’s not universally shared, and because there continues to be debate about its meaning. Disagreement dictates that my definition of majimbo is in some serious (but unspecified) trouble. That’s not very plausible. It’s easy to think of lots of other terms whose definitions are contested with some heat: justice, equality, diversity, and the like. Participants in arguments in which those terms occur don’t generally have to bring their opponents to a truce on the definitional front before making their positive argument. There doesn’t seem to be any pressing need to impose a stronger requirement in this case.
Both sides agree on the dictionary definition(s) of the word; there’s also substantial agreement on non-dictionary definitions. Yet there is a dispute. So it is unlikely that the definition of the word is what’s at issue. In fact, the dispute revolves around a contested case: the ethnic cleansing of the last twenty years. Proponents of majimbo want a definition of the word which excludes that case. Opponents of majimbo want a definition that doesn’t. Proponents of majimbo argue, therefore, that what they mean by their use of the word majimbo is given by the lexical definition of the word. Opponents want to say that looking up the dictionary meaning isn’t a terribly useful way of determining what majimbo proponents intend. What could resolve that dispute? Well, it would end if someone or other turned up with a theory of how to determine the meanings of words. The meaning of majimbo-as-used-by-its proponents would follow; we could all inspect the shiny new meaning of the word; it would be clear who was right, or whether both were wrong.
Two pleasant consequences. First, the Branch and Cheeseman objection doesn’t get off the ground because it relies on the false premiss that the dispute is about the meaning of the word, rather than about a method for finding its meaning. Second. In general, from the fact that there is a dispute about how to determine the meaning of a word, it just doesn’t follow that a user of the word doesn’t know the meaning of the word, for two reasons. First, the fact that there’s a dispute doesn’t establish that none of the disputants is right. And second, there’s simply no requirement that one has to settle disputes about how to find the meaning of a word before being credited with knowledge of the meaning of a word: for any given language, there are lots of native speakers who lack the formal training necessary to engage disputes about the semantics of their language, but whom it would be absurd to regard as incompetent users of their language.
It’s possible, then, that I’ve got the meaning of the word wrong, but Branch and Cheeseman’s argument doesn’t show it.

4. Fourth, while the Kalenjin are critically scrutinised, can the conclusion that ‘Kalenjin opinion leaders who support ethnic violence … lack effective internal constraints’ not be applied to other communities?

First, the questions are independent. The focus on majimboist violence is justified by its duration, intensity and reach, on all of which counts it stands out. The status of internal restraint in other communities is independent of these facts. Second, there’s reasonable evidence that non-Kalenjin communities have restrained themselves. Kimenyi and Ndung’u (Kimenyi and Ndung’u: 152-3) explain Kenya’s lack of a proper civil war partly by appeal to precisely this restraint. They think, for instance, that the Gikuyu experience of Mau Mau, and the high opportunity cost to Gikuyu business of violence, have restrained retaliation. I’d add that the first Kibaki administration’s extraordinarily violent repression of Mungiki is proof that there are internal restraints – even if only pragmatic ones – on the use of ethnic militias in the Gikuyu political elite.

5. And finally… Finally, ‘going directly to motives’ has no explanatory advantages if those motives are simply those that one individual, without any great evidential basis, chooses to ascribe to a whole range of actors. Waweru may see it as his aim to show that ‘the violence was actually unified in aim’ but without any compelling reason for doing so we are not convinced.

This can’t be quite right, since the motives ascribed may have explanatory power even if the evidential basis for their ascription is thin. The explanatory power of the attributed motives is independent of the evidence given for adopting them. The objection confuses the reasons we have for believing that this motive has explanatory power and actual possession of explanatory power. So an argument that succeeded in showing that there was no good reason to ascribe this motive doesn’t show that the motive, if ascribed, would not do the explanatory job. (And nor does the failure of one attempted explanation by motive call into question the general strategy, as Branch and Cheeseman seem to suppose.)

Again, they’ve slightly misrepresented my argument: the point of which – the main conclusion of which – is that reform and accountability are incompatible. To get there, I needed to show that there’s a continuous history of violence in the Rift Valley which has caught the RV political class, hence the claim that the violence is unified.

The shape of the sub-argument on the way to that conclusion is thoroughly unmysterious. Spontaneous – more accurately, disorganized – violence is short-lived, rarely lethal, and unlikely to be selective. Past episodes of majimboist violence have been prolonged, lethal, and ethnically-selective. It’s safe to assume they were organised, which makes it sensible to wonder what the point of this organisation is; while the consequences of the violence make it sensible to assume that past episodes had ethnic cleansing as an intermediate end, at least. The recent episode of majimboist violence is so similar in important ways – in location of the violence; in the cast, both victims and assailants; and in results – that it’s safe to say that the motives, too, remain the same.
Start with the victims. Majimboist violence has been ethnically-selective since its inception. There’s substantial overlap in the sets of victims: quite consistently through the violence, the victims have been non-Kalenjin, and usually disproportionately Gikuyu. I’ve already mentioned Biwott and Lotodo’s exhortations to anti-Gikuyu violence in the post, but there’s no shortage of other stuff of that ilk, then or now: from Ntimama’s remark about envelopes; to the remarks often heard in those days about ‘opposition-free zones’; to Ntimama’s 2007 remarks about ODM’s redemption of Kenyans from their enslavement by Gikuyu; and on to the well-reported fact that Rift Valley led in recorded incidents of pre-election ethnic hate speech (KCHR 2008: 24). In the 90s, as Boone (see below) points out, Kalenjin and Maasai were urged to evict their non-KAMATUSA neighbours. In 2007, Lynch urges us to pay attention to both elite manoeuvring and to local understandings of history (Lynch: 567-8): a polite way of saying that were rekindled to encourage Kalenjin and Maasai to evict their Gikuyu neighbours. As one might expect, the victims were disproportionately Gikuyu in the 90s; a pattern repeated in the most recent round of violence. The selectiveness of the violence is not incidental to it: in 1992, there were severe attacks on Luo; in 2007, majimboist violence was directed elsewhere. If the ethnic selectivity of the violence can vary in this way, t’s safe to assume it’s intended; and equally safe to infer that the motives at work in the 90s aren’t impotent in 2007.

Majimboist violence, it has been noted, is consistent in its location. Prof. Catherine Boone’s exceptionally detailed analysis of the violence (“Politically-Allocated Land Rights and the Geography of Electoral Violence: The Case of Kenya in the 1990s” (unpublished)) picks out key areas of conflict in the 1990s: Nandi and Muhoroni; Trans-Nzoia and Mount Elgon; Uasin Gishu and adjacent areas; Molo, Kericho, Sotik and Narok. In the 1990s, as she says, politicians in all these encouraged their Maasai and Kalenjin constituents to dispossess and evict their non-KAMATUSA neighbours. Every single one of the areas she picks out also saw violence in 2007; indeed, Uasin Gishu was the epicentre of this round. Much the same goes for Throup and Hornsby’s analysis.
Comparing the Akiwumi Commission’s list of adversely mentioned persons against the Kenya National Human Right’s Commission’s Schedule of alleged perpetrators of post-election violence, one finds that the lists have at least three distinguished names in common: Jackson Kibor, Kipkalya Kones and William Ole Ntimama. Much of the rest of the Akiwumi list is a roll call of the senior membership of the KAMATUSA political class. Anderson and Lochery (Anderson and Lochery 2008: 334) report that culpability for the 2007 violence among ‘local-level Kalenjin politicians’ was widespread. This is more or less the pattern reported in Boone, in Throup and Hornsby and in countless other sources. Majimboist violence, it appears, has similar personnel, both at the level of individuals and at the level of groups, a smilarity which, taken together with the others, justifies the inference that the motives for the violence have remained the same.