‘Poll poison’?: Politicians and polling in the 2007 Kenya election

Abstract

The debate continues about the causes of the post-election violence that led to the most serious challenge to Kenya's national integrity since independence. However, there is one aspect of the pre-election period about which there is little disagreement: the far more frequent and widely disseminated results of opinion surveys by several research organisations than in any previous election year. Often controversial and provoking highly acrimonious exchanges, these polls and the alleged motives of those behind them became central features of the various campaigns. While their precise contribution to the violence cannot be determined, this analysis makes use of a selection of such results to consider this issue while arguing that: (1) they achieved considerable accuracy in tracking fluctuations in the candidates’ standings in what was a very fluid political landscape; (2) the reactions to them highlighted enduring and significant commonalities within the political class; and (3) this polling experience is certain to have longer-term effects on the country's political culture.

Keywords:

• public opinion in Africa
• pollsters
• poll results
• Steadman SPEC Barometer
• political culture
• political choice
• media credibility

In the countdown to the 2007 general election, pollsters were lauded by some as innovators, condemned by others as frauds, and accused by both [sic]1 political parties of bias. Whatever role they played, polling groups were more established and more influential during the last election than in any year prior.

‘Polling and the Kenyan Media’, Expression Today, March 20082

If the ‘Afro-pessimists’ are correct about the all too frequent deviations from ‘free and fair’ practices in African elections – which include vote buying, ghost voting, ballot stuffing, vote tallying falsification, intimidation and violence, to say nothing of pre- or post-election coups and even partial boycotts by key actors (for example, Cowen and Laakso 1997; Bratton and Posner 1999; Mehler 2007) – there would be little point in conducting voter-intention polls except to measure the ‘true’ distance from the eventual polling day ‘reality’. As Gyimah-Boadi (2007, 27) has argued: ‘Elections [in Africa] tend to be rigged, and incumbents have been keen to rig to the extent that they can get away with it – which has often fuelled post-election conflicts, and in some cases, triggered violence’.

Various other recent studies, however, have shown that African elections are steadily, if unevenly, improving in terms of their contribution to democratic development, even as the commitment of ordinary citizens to this goal has withstood such uneven progress (Bratton, Mattes and Gyimah-Boadi 2005; Lindberg 2006). While such advances appear to constitute a case for the increasing relevance (if not direct democratic utility) of voter intention (let alone exit) polls as part of this broad process, it is striking that in none of these works is there any reference to such instruments.3

As such, the polling experience in Kenya's 2007 election warrants both attention and further study. This is so for several reasons, even if the large-scale violence following the announcement of the presidential result by the (now disbanded) Electoral Commission of Kenya (ECK) had not occurred. First, there were far more polls, conducted by more firms, than in any previous pre-election period. Next, media coverage and outreach made it impossible for key political actors to ignore these poll results, thus ensuring that they became central subjects of the campaign. Finally, this prominence (and their general acceptance) means that such polls are likely to remain a fixture in Kenyan public life, warranting an assessment of their potential role in future elections as well as in the country's evolving political culture generally.

In considering them, however, it is necessary to note the particular context of the 2007 election, starting with the fact that it was the country's most competitive contest by far. In contrast to the 1992 and 1997 elections, for example, there were only two main candidates, representing the most deeply rooted divide in Kenyan public life: that between the Kikuyu and the Luo (Atieno-Odhiambo 2004). Moreover, the fact that it was Raila Odinga's ‘blessing’ that had assured Mwai Kibaki of the joint opposition's nomination in the ‘transitional election’ of 2002 (Oyugi, Wanyande and Odhiambo-Mbai 2003) – leading to a virtual walkover4 and the promise of tangible ‘change’ in the lives of ordinary Kenyans – meant that the former's challenge to the incumbent in 2007 was especially dramatic. Such grievances clearly fed into the tensions that surrounded this election and its disputed official outcome: a Kibaki win by about 230,000 votes out of some 9.8 million cast.

Just how contentious this official result was is evident in a Gallup (USA) poll conducted some six months after the election. It found that 57% of Kenyans believed that Odinga had won, with only 25% stating that they thought that Kibaki had; the remaining 18% were not sure.5 Given that Kibaki had been declared to have obtained nearly half of all votes cast (46%), it is evident that a substantial proportion of those who voted for the president were by this time doubtful that ‘their man’ had actually won.6

A few months earlier (in March), Kenya's major survey firm, the Steadman Group (now Synovate),7 released the results of its first post-election SPEC Barometer (‘Social, Political, Economic and Cultural’) surveys. It revealed that 65% of Kenyans blamed the announcement of the presidential result for the destruction and death that had occurred; it remains unclear, however, whether it was the identity of the winner that respondents had in mind, or rather, the entire performance of the ECK – or both. Only 23% attributed it to either land grievances or long-standing tensions between ethnic communities.8 No respondent cited opinion polls as the main cause of this violence.

Yet even before this civil explosion occurred (following the declaration of Kibaki's win, the scale of which no one could have imagined) several commentators wondered if the country's institutions would be able to cope with as close a result as the polls were suggesting. As one put it, barely a month before the election:

The opinion polls have largely calmed down and the wild swings of the past months are all but gone. The two leading candidates … are separated by not more than five percentage points, with Mr Odinga ahead in most of the polls. Taking into account [their] inbuilt inaccuracy, it then becomes clear that the distance between candidates is going to be short indeed.

Who or which institution has acceptance across the divide and could arbitrate a dispute over who has won a presidential election in Kenya today?

It does not help that Mr Kibaki, contrary to public opinion, saw fit to unilaterally appoint ECK commissioners rather than following an earlier agreement which required him to consult all political parties.9 (O. Obonyo, ‘Pollsters are only right when they endorse a voter's favourite’, Saturday Standard, 24 November 2007)10

Even without expecting the worst, however, given the attention they had received,11 it appears that these polls helped shape the public's reaction to the announcement of Kibaki's win in two ways. First, they encouraged its rejection by Odinga's Orange Democratic Movement (ODM) party. This was despite several published analyses accompanying these results showing how variations in voter turnout across the country's eight provinces could easily determine the winner in such a close race, and which, based on historical patterns, were more likely to favour the incumbent.12

Second, the very presence of these polls contributed to an almost intoxicating sense of irreversible democratic gain. Indeed, the heated debates they generated – without any serious effort to regulate, let alone, ban them13 – encouraged the assumption, held as firmly by Western diplomatic missions as by everyone else, that all the main players, like the vast majority of the public, were now fully committed to ‘numerical electoralism’: that obtaining the most votes, honestly cast and counted, is the only legitimate basis for public power, even when this requires the removal of a comfortably ensconced incumbent.

How the polls illustrated the changing fortunes of the main contenders, why they were so controversial, and what they revealed about Kenya's political culture even as they most certainly changed it, are the main subjects of this paper.

Opinion research in Kenya: The ‘market’, media and political context

While there was some polling during the period of Kenya's independence transition, it was only with the return to multiparty politics in 1992 that a domestic opinion survey industry gradually emerged (Hornsby 2002).14 Its full blossoming, however, awaited the departure of the autocratic President Daniel arap Moi.15 At the same time, such activity was an offshoot of Kenya's commercial dominance in East Africa, where market research of various types, drawing on techniques and expertise from many of the multinational corporations represented there, had become established practice. In this less constrained atmosphere, several companies came to prominence: Strategic Public Relations, Infotrak-Harris, Consumer Insight, and, especially, the Steadman Group.16 The trajectory of the polling industry in Kenya is thus itself a quite precise ‘barometer’ of the prevailing governance environment.17 Before looking at a selection of these poll results, five points necessary to appreciate their place in this election may be noted, beyond the passionate polarisation between the two leading campaign teams, noted above.

One is the high level of awareness that Kenyans came to have of the polls, and the considerable credibility they earned, both reflecting the outreach of and public trust in the media.18 During 2007, for example, respondents interviewed for Steadman's SPEC Barometer reporting whether they had ever heard of the company's polls rose from just under 50% to over 75% between March and November (with those in urban areas slightly more familiar with them than their rural cohorts). Likewise, the media have been consistently rated as among Kenya's ‘most trusted’ institutions. Steadman's SPEC survey for December 2006, for instance, found that 82% gave the media a combined rating of ‘some’/'a lot of trust’, putting it in fourth place of more than a dozen professional and institutional categories so ranked.19

A second factor is the quite fluid nature of political choice in Kenya. This applies both to parties and individual leaders, as the actual ballot-menus presented to voters in the last four elections have varied considerably in terms of the parties fielding presidential candidates and the alliances of particular leaders in connection with such parties. At one level, this has been a function of the ease with which political parties have been formed and registered, with a concomitant effect on the largely transient (if not ethereal) nature of party ‘membership’ among their followings (Wanjohi 2003).20 More fundamental, however, is the minimal policy-contrast choices that parties offer, leaving ethnic identity as the principal basis for party alignment.21 Yet the fact that even the largest ethnic group (the Kikuyu) comprises less than a quarter of the population (and the three largest constitute hardly half of it), combined with a presidency-oriented, first-past-the-post or winner-take-all electoral system,22 means that coalition building based on ethnoregional alliances is essential in any bid for state power. It also means that short-term deals often underpin such alliances, with today's comrades-in-arms sometimes tomorrow's most bitter adversaries (and vice versa).

The other side of this coin, however, is that electoral choice for members of those many communities without (at least any serious) presidential candidates of their own may be much less constrained.23 As such this uncommitted section of the electorate (like voter registration and turnout itself) can thus determine the outcome in a close contest in Kenya as elsewhere in the world.

A third factor is the absence of relevant empirical research about polls. Consequently their impact was often subject to quite fanciful conjecture. For instance, a poll showing one leader less popular than another, or not appearing at all,24 or who had lost even a few percentage points over a given period of time, was seen, nevertheless, as highly damaging. More specifically, it was assumed that such results could create a ‘bandwagon’ effect by depressing further the votes for any candidate whose bid appeared unviable, or who was even just losing ground.25 As a local columnist argued early in the election year:

It is absolutely unscientific and undemocratic to seek to pinpoint ‘majority opinion’ by throwing a few leading questions at individuals, chosen at random, who are likely to be totally ignorant of the social implications of the questions facing them.

In any case, what exactly is the social value of knowing in advance which candidate a correspondent will vote for? Indeed, isn't it dangerous?

In a society where real issues matter so little, figures such as Steadman tosses around can powerfully sway the mass as to whom to vote for. (P. Ochieng, ‘Does Steadman take intelligence into account?’, Sunday Nation, 8 April 2008)

And even after the election, a communications consultant took the same position:

Polls can create a sense of confidence that one candidate is winning and lead to despair on the other side. In places where voters are not sure if their candidate is winning, they can cross over. They shape opinion and that is why we fear manipulation by polling groups. (K. Njogu quoted in ‘Polling and the Kenyan media’, Expression Today, 2008, 37)

Yet two local examples raise profound doubts about this assumption. One is the national referendum on a draft constitution that was held two years before the election. Six weeks before that event, a Steadman poll found that about 13% more Kenyans were prepared to vote ‘No’ than ‘Yes’, and these results received wide attention (Wolf 2005).26 The other is the same company's repeated finding of around 10% support for Kalonzo Musyoka in the 2007 presidential race for nearly six months prior to the election (see below). This also attracted considerable publicity. Yet in neither case did the actual results fall outside the margins of error of their respective surveys (57 to 43% in the referendum, and 9% for Musyoka), suggesting, at the very least, no such ‘negative-bandwagon’ effect. Still, the fact that no one, least of all the pollsters, could say just what impact these polls might have, gave free rein to those prepared to assign whatever import to them that they were inclined to.

A fourth and related contextual factor also reflects the novelty of polls: the lack of familiarity that most public figures and their aides have with survey techniques.27 This encouraged inconsistency in the assessment of particular poll results, depending upon whether they produced joy or gloom. So, too, did the uneven capacity of media practitioners in reporting and interpreting them.28 Such inconsistency, as well as suspicion, was also fueled by the absence of any mechanism to monitor industry standards (acting under whatever authority). In this connection, the fact that only a handful of firms were engaged in such work meant a lower risk to professional reputation if any one of them produced results at great variance with those of the others, whatever the cause of such deviation.

A final factor, and perhaps most critical, was the widespread perception, reflecting a history of the both blatant and subtle use of state and ‘shadow-state’ power, that even the most established professionals in Kenya are sometimes unable to resist the pressures of ethnic loyalty or the lure of financial gain,29 let alone to ignore direct threats. In the context of such a highly polarised political contest, therefore, such assumptions were bound to affect the pollsters’ credibility no matter what results they produced (as illustrated below).

‘Poll position’: Presidential preferences in a fluid political landscape

The impact of opinion polls on the outcome of the 2007 presidential elections was controversial. While politicians sought to downplay the significance of opinion polls on the outcome of the presidential race, it was obvious that the polls were being taken seriously by political campaign strategists, voters and even the presidential contestants. Every time the various poll results were released, a storm brewed. Opinion polls conducted within the last three months before the 2007 elections confirmed the prediction that the elections would be a hotly contested affair between Kibaki and Raila. (Independent Review Commission (IREC) 2008, 62–3)

To put into perspective the reactions of those most affected by the polls, it is necessary first to examine the changes in the ratings of the main presidential candidates over the immediate pre-election period, and to suggest briefly why these occurred.

Starting with the opposition side, the two main figures were Raila Odinga and Kalonzo Musyoka, the latter being one among several long-serving KANU-era cabinet ministers who, led by his cabinet colleague Odinga, ‘rebelled’ at Moi's imposed choice of the political novice Uhuru Kenyatta as the party's standard-bearer in the 2002 election. In doing so – having taken over the obscure Liberal Democratic Party (LDP) for the purpose – they threw their weight behind Kibaki's third opposition presidential bid (Ndegwa 2003). Key to the alliance that underpinned its success was the promise of constitutional reform, which included the creation of the position of prime minister, loudly earmarked for Odinga. By late 2005, however, this alliance had split, mainly over this issue, after the Kibaki faction succeeded in scaling down the powers attached to this proposed office as well as to local government, both centrepieces of a new draft constitution. The version subjected to a national referendum in November 2005 therefore, was viewed by many as a betrayal of the National Alliance Rainbow Coalition's (NARC) most alluring 2002 pre-election promise (Oyugi 2006; Chege 2008; Whitaker and Giersch 2009).

The combination of the referendum's defeat and Kibaki's response – the sacking of those cabinet ministers who had campaigned against it – gave a significant boost to the opposition's standing, with the president's rating dropping by nearly half over the next quarter (by 18% from the 40% he enjoyed before the referendum).30 Yet this triumph also sowed the seeds of discord on the opposition side, since both Odinga's and Musyoka's eyes were henceforth on the same (presidential) prize.31 With the hopes of constitutional reform dashed, and Odinga now making his own ambitions known, his ratings began a slow but steady rise,32 even if it took an entire year for him to overtake Musyoka (in March 2007, by just 3%; ‘Kalonzo edged out of ODM race’, The Independent, 5 March 2007). In effect, in terms of an aggressive ‘pro-change’ alternative to Kibaki that cut across a wide swath of ethnic communities chafing under a perceived ‘Mount Kenya hegemony’, Musyoka was simply no competition, leaving his standing largely static throughout 2007.33

With tensions increasing between these two opposition leaders, the president's side assumed a far more coherent posture than it had in the immediate aftermath of the referendum's defeat, so that Kibaki was able to overtake Musyoka in mid-2006 and then enjoy a comfortable lead in the polls for an entire year to mid-2007.34 As such, it was only after Musyoka and Odinga had split, each identifying with and obtaining the nomination of his respective party (ODM-Kenya and ODM),35 that Odinga pulled ahead of Kibaki. This development suggested that many prospective voters needed to see the actual election menu before making up their minds, although the government's impressive macro-economic growth record was also an important factor to another, distinct section of the electorate for whom the Odinga–Musyoka tussle was irrelevant (Chege 2008, 128).36 This lead narrowed, however, after the Kibaki campaign's formal launch in October,37 with the race becoming extremely close from November, according to most of the polls, including Steadman's.

‘Poll-Paranoia’: The (il)logic of political response

Reaction to political opinion polls has become predictable – those favoured harp on them … as if they have won the election. Those the poll does not favour take a hard line and target the polling agency … Attacking the polling agencies is killing the messenger and failing to listen to the message. (M. Nyanchama, ‘Opinion poll: How you can make bad news better’, The Standard, 10 January 2007)

As suggested above, whatever doubts had existed about the potential accuracy of such survey instruments, the Steadman Group's performance in gauging almost exactly the outcome of the November 2005 constitutional referendum put most of them to rest. Henceforth, the political class, the media, and ordinary Kenyans were far more inclined to take such results seriously, even as that ‘No’/‘orange’ versus ‘Yes’/‘banana’ outcome laid down the battle lines of the forthcoming presidential contest.38 Given the election's high stakes, therefore, their very credibility, combined with an unnerving assumption as to their impact on the electorate, caused tensions to rise whenever a new set of results was released. And such tensions were heightened by two specific aspects of these results: the positions of all three main contenders at first fluctuated, and the gap between the top two became extremely narrow in the campaign's final stages.39 As a consequence, key players often found it necessary to engage in rather inconsistent rhetorical ‘gymnastics’ in responding to them.

Looking first at the opposition, during 2005 when the polls showed Musyoka leading Kibaki, Odinga – still identified with the (non-existent) position of prime minister – vigorously defended them against attacks by the president's supporters, terming polls ‘a normal part of any mature democracy’ (‘Mixed reaction to opinion poll on Kenyans’ choice for president’, Sunday Nation, 12 June 2005). But his tune changed abruptly when he began to trumpet his own presidential ambitions after the defeat of the constitutional referendum. Now he dismissed his opposition colleague's top ranking in Steadman's March 2006 poll as merely a reflection of ‘how handsome someone is’ (‘Raila is no longer content just playing the kingmaker’, Sunday Nation, 9 April 2006). Moreover, given the time it took for Odinga's rating to rise, his stalwarts became even more convinced that the polling results were being manipulated by pro-Kibaki forces, a view captured by this ‘alternative’ publication:

According to the [Strategic Public Relations] poll, Kalonzo Musyoka still remains the leading presidential candidate followed by President Kibaki with Raila at number three and William Ruto comes a distant fourth.

The most amazing thing about the poll is that the research firm ignored the other ODM presidential candidates that are Najib Balala and Moses [sic] Mudavadi.

We have however reliably established that this was done on purpose and the whole idea of bringing Ruto into the fray was to further scuttle the ODM luminaries. (‘Narc-K's trick to kill Raila, Kalonzo deal’, The Independent, 22 May 2006)40

Even mainstream newspaper editorials occasionally made similar assumptions, including this one, in response to the same (March) Steadman poll:

Raila did not think much of the survey, as no one had been campaigning. Apparently, Raila thought the researchers were looking for the most handsome man.

What Raila & Co should realise is that they, too, can spring a surprise in the coming few months and attain top position in the ratings. After all, to some imaginative minds, one can compute research data with the eyes shut. We are not suggesting that Steadman would resort to such antics, but the main lesson from New York Times reporter Jason Blair is that nothing surpasses the power of imagination.

(‘A presidential beauty pageant?’, Daily Nation, 10 April 2006)

For his part, as long as he led, Musyoka and his allies lauded the polls and sought to make practical political use of them. For example, referring to the results of this (March 2006) poll (in which Musyoka outscored Kibaki and Odinga by 12% and 23%, respectively) at a church fund-raising meeting, a fellow-Kamba Machakos Town MP, Daudi Mwanzia, applauded Musyoka's ‘national appeal’ that it had supposedly confirmed (‘ODM will nominate me, claims Kalonzo’, The Standard, 3 April 2006). And later that year at a party retreat, when a heated debate arose over its forthcoming choice of a presidential candidate, ‘The Kalonzo team, citing all the opinion surveys so far conducted, including the Steadman, insist[ed that] he was the best placed’ (‘Rivalry straining ODM-K’, Kenya Times, 17 October 2006).41 Subsequently, as the jostling within ODM-Kenya increased, and with Kibaki enjoying an apparently unassailable lead (as shown in Figure 1, of 34% by March 2007), it was the president's side that gave ringing endorsements to the polls, leaving the opposition collectively to accuse Steadman of a pro-Kibaki bias.42 The fact that ODM-Kenya dominated the party ratings even well after the launch of NARC-Kenya43 made Kibaki's steady rise that much more perplexing for them.44

Figure 1.  Presidential voting intention poll* (November 2005 – December 2007).

In this context, returning from a US visit in March 2007, Odinga ‘revealed’ that while there he had discovered that Steadman was ‘illegally’ using the Gallup name for which it would soon be charged in court,45 adding for good measure (if somewhat contradictorily) that the company should be ignored since it ‘has no votes’ (‘No more ‘Tosha’: Raila returns and declares he won't be endorsing anyone for ODM-K's presidential ticket as Kalonzo roots for party democracy’, The Standard, 13 March 2007).46

However derogatory, comments of this nature hid a serious interest in the use of these survey instruments across the political spectrum. For example, shortly thereafter it emerged that Odinga's party had used several polls to calculate to a tenth of one per cent the vote-winning potential of each of ODM-Kenya's five potential candidates.47 More generally, it emerged that by mid-year both main campaign teams were commissioning their own confidential surveys, in some cases undertaken by the very firms being blasted in public,48 with the president also reportedly using the government's National Security Intelligence Service (NSIS) for the same purpose.49

After the split on the opposition side, and with less than five months to the election, reactions to the polls became even more strident. Continuing to trail both of his rivals badly, Musyoka claimed that State House was using Steadman to boost Odinga since the latter constituted ‘less of a threat’ to Kibaki (‘Still doubting Raila’, The Weekly Citizen, 2–8 July 2007).50 The logic of this view was evidently lost on Kibaki's supporters from September 2007, however, when Odinga overtook the president in the polls. Now it was the latter's supporters who launched the fiercest attacks on Steadman.

Some of these were (the by now familiar) technical objections, such as the ‘nonsensical inadequacy’ of a sample size of ‘only’ 2,000 for an electorate of some 14 million, thereby inviting ridicule from several commentators and newspaper readers familiar with sampling methodology.51 More ominous, however, was the allegation made in reaction to Steadman's mid-October poll that gave Odinga a dangerous 16% lead: that the company was complicit in the pursuit of a US government/Western ‘agenda’ to remove Kibaki from power, a ‘fact’ confirmed by the leading role of a (resident) American consultant analyst in its polling work.52 Indeed, this was the only time the president himself mentioned the polls (at a campaign rally in Mombasa), though in doing so he dismissed these results as ‘media lies’ rather than attacking the pollster (‘Storm over opinion polls’, Sunday Nation, 14 October 2008). Such presidential attention seemed to further encourage ad hominem accusations levelled at Steadman, with a group of pro-Kibaki Nairobi MPs calling for the deportation of Steadman's expatriate analyst during their meet-the-people campaign rounds that same weekend.53

Yet, given the fact that the company's CEO is himself from Kibaki's home (Nyeri) district and its founder and principal shareholder (Roger Steadman) is an expatriate whose career ‘linked him to known allies of President Mwai Kibaki’ (‘Polling and Kenyan Media’, Expression Today, March 2008, 39), accusations based on such a silent attribution of some political (or other) anti-Kibaki motive largely fell flat.54 The PNU barbs tossed at two other firms according Odinga somewhat larger lead margins were slightly less incredible, however, given the Luo identity of their top personnel.55

Considering the tangible advantages of wielding state power,56 however, and the visible determination of the Kibaki team to counter the challenge to the status quo that Odinga's presidential bid was seen to represent, there was always more anxiety (if not paranoia) on the opposition side.57 This was evident following release of the first of Steadman's two November 2008 polls showing that Odinga's lead had dropped seven points (from 11% to 4%).58 Even when the company explained this change as largely due to adjusting its provincial sampling-distribution procedure,59 his camp leapt to attribute the earlier lead he had enjoyed to a ‘State House plot’ intended to foster a false sense of invincibility, with Steadman now ‘scheming’ to depress his ratings in the final weeks of the campaign to add momentum to a late Kibaki ‘surge’.60

Subsequently, by late November, when Steadman put the two principal candidates in a statistical tie, ODM loyalists attempted to explain these results by ‘revealing’ that the company's chief (expatriate) analyst had been forced to ‘toe the line’ to secure the renewal of his work permit, having been personally taken to State House and ‘slapped by the First Lady’ to ensure he did this.61 As if to acknowledge the preposterous nature of this claim, however,62 a (more flattering) ‘news flash’ was subsequently circulated that this same analyst had resigned ‘in disgust’ at his employer's capitulation to state pressure, having rejected it himself.63 These suspicions were also evident in a full-page advertisement placed in a leading daily by the ODM secretariat that made a similar insinuation about Steadman's ‘working relationship’ with the Kibaki campaign (‘Pondering On Polls’, The Standard, 11 December 2007).64

Returning to Musyoka: following his split with Odinga and with his Steadman ratings stuck at around 10%, he might have been expected to ignore these aspersions. Yet, on the contrary, the company remained a standard rhetorical target as he declared repeatedly at his rallies that an election day ‘miracle’ would prove Steadman wrong and see him ‘passing right through the middle’ of his competitors to win (‘Polls “will be three-way race”: I will beat Raila and Kibaki, says ODM-K's Kalonzo’, Daily Nation, 15 August 2007).65 Such concerns with his poll standings led Musyoka eventually to request an audience with Steadman's CEO, ‘during which he sought an explanation for his static single-digit performance in recent opinion polls’ (‘Fury over polls … as Kalonzo takes Steadman CO to task over opinion results’, The People Daily, 8 November 2007).66 However, this meeting had no apparent impact on the regularity with which Musyoka's meagre and ‘erroneous’ ratings featured in his campaign speeches.67

The home stretch: Competing firms, competing results

The fact that the results of different polling firms sometimes varied (even taking their error-margin ranges into account) made it easier for those wanting to discredit their work to do so, a situation encouraged in part by the failure of media houses always to provide full details of (and then draw attention to) differences in sampling methodology. Table 1 reproduces the final results released by five survey firms two weeks before the election, comparing them with the official (disputed) results.68

Table 1. Final (mid-December) survey results (rounded figures in percentages).

Survey firm	No. districts/sample size	Kibaki	Odinga	Musyoka
Steadman Group	71/6,111	43	45	10
Strategic Public Relations	Unstated/2,405	36	46	17
Infotrak-Harris	Unstated/2,400	36	46	16
Consumer Insight	39/2,400	41	43	15
Gallup (USA)/Path Assocs.	Unstated/2,000	44	43	12
‘Actual’ (ECK) Results	–	46	44	9
*There were 71 districts when Kibaki assumed office; he created a number of additional ones during his first term.

The biggest variations are in support for Kibaki: 44% (Gallup USA) versus only 36% (Strategic and Infotrak-Harris). But they are also evident in the figures for Musyoka ranging from just 10% (Steadman) to 17% (Strategic); those for Odinga are in greater agreement (43 to 46%).69

What drew most attention, however, was that just one week earlier all four local pollsters had given Odinga a nearly identical lead (between 3 and 4%). Now, with his margin by the two ‘Mount Kenya’ firms (Consumer Insight and Steadman)70 down to just 2%, while rising to 10% according to the two ‘Nyanza’ firms (Strategic and Infotrak),71 some concluded that late-hour ethnopolitical loyalty had trumped professional integrity, and that the pollsters accepted the notion that their products could somehow change enough voters’ minds to make a difference as the electorate prepared to cast real ballots. But, once again, such ‘doctoring’ convictions appeared stronger on the ODM side, whose general suspicions were bolstered by such late-hour revelations as the ‘secret’ drafting of a large number of administrative police to serve as PNU polling station agents (and ‘private’ security officers) in ODM's Nyanza stronghold (‘Mysterious police trips fuel poll rigging claims’, The Standard, 26 December 2007).72

In this tense atmosphere, and with such a narrow gap reported by even just two firms, Kibaki's campaigners could cautiously predict his forthcoming triumph while stressing the critical importance of voter turnout. For example, cabinet minister Martha Karua, campaigning for Kibaki in their native central province in the final week, warned that ‘the polls have shown just how close this election will be’, even as she claimed that they were being ‘staged-managed by Raila strategists’, though adding that ‘in the world where opinion polls show the incumbent was neck and neck with his closest challenger the incumbent emerges the winner’ (‘We are confident of winning – Karua’, The People on Sunday, 23 December 2007). Meanwhile, Odinga reassured his supporters ‘not to worry’ since they were ‘headed to State House’, leaving Musyoka merely to repeat his assertion that Steadman ‘will be shocked when the winner is named’ (‘Steadman polls now questioned’, Sunday Standard, 9 December 2007). At the same time, Musyoka seemed to be contradicting his own advice when he ‘urged Kenyans to ignore the polls’, though his accusation that the government was ‘planning to use the Provincial Administration to rig the General Election’ constituted an altogether different reason for their supposed irrelevance (‘Kalonzo alleges plot to rig polls’, The Standard, 2 December 2007).73

Conclusion: Lessons from Kenya's 2007 election polling experience

In their societies of origin, voter intention, exit and more general opinion polls are largely taken for granted, and with few dissenting voices,74 are considered as supporting democracy in two main ways. First, they give political First, they give political ‘merchants’ and their prospective ‘customers’ a better idea of the former's electoral prospects, thus encouraging the campaign teams to discover and address more precisely key bases of electoral support, while allowing voters to make more ‘strategic’ choices. Second, and more broadly, they bolster the legitimacy of political pluralism itself by allowing for the far more frequent expression of competing views on leadership preference, policies and other matters by the ‘voiceless citizen’ than can either elections or direct, personal representations of one sort or another, especially in contrast to well-organised ‘special interest’ lobby groups (Monroe 1975; Manza, Cook and Page 2002).

Yet, in light of the several factors discussed at the outset, such positive and largely uncontroversial contributions elsewhere are far less certain in a setting like Kenya. In this context, lessons relating to two polling issues that arise from the 2007 Kenyan election may be revisited: the reactions of the political class to them, and their contribution to the post-election violence.

Lesson one: ‘Poll-ution’ versus political opportunity

The high stakes in this election, combined with its extreme competitiveness, served to highlight two related aspects of Kenyan political society that encouraged, at best, an ambivalent attitude towards the country's maturing survey research industry among most key political actors. One is the quite fluid, if not often opportunistic, alliances they often forge. Among numerous possible examples, three may be mentioned: the long-suffering opposition leader Odinga first ‘co-operating’ with Moi's KANU and then actually joining the then ruling party before the 2002 election; Uhuru Kenyatta supporting Kibaki's re-election bid in the last election after being his main competitor five years earlier; and Moi himself backing Kibaki in 2007 after fronting Kenyatta against him in 2002. An additional example from several months before the 2007 election when such alliances were still far from cemented, and that indicates the growing impact of opinion polls, is provided in this account:

The latest of these meetings is between two political party bigwigs, read LDP de facto leader Raila Odinga and Ford Kenya chairman Musikari Kombo.

This is against the background that a cloud of uncertainty is hanging over the head of LDP leading light, Kalonzo Musyoka. In the recent weeks, and especially after an opinion poll by the famous Steadman International [sic] revealed that Kalonzo is the most popular presidential candidate in the forthcoming general elections, wheeling and dealing has been the order of the day.

These came barely a few days after Raila had dismissed the opinion poll as a gimmick that was based on how handsome somebody was without reading the people's moods in the current ultra volatile political scenario. (‘Raila, Kombo in new alliance’, The Independent, 24 April 2006)

In this vein, it appears one scholar could well have had Kenya in mind when he identified the foundations of such volatility in ‘third-wave’ multiparty systems, even if he apparently gave no thought to such polls in doing so:

In mature democracies, politics are considerably less autonomous of voters and their preferences. It would be inconceivable for the socialist and communist-dominated trade unions in a country like France to switch their support to a right-wing party, and even if they did, their members would be unlikely to follow suit. Ideological divisions and long-standing cleavages shape the actions of politicians, who seek the support of voters within a fairly narrow and well-established political space in which economic and social policies dominate. In most multiparty systems that have emerged out of the third wave of democratization, however, political cleavages are not well-set, and identity politics typically trump ideology. As a result, political alliances are more fluid and changing, and individual politicians have a greater degree of autonomy in the deals and alliances they make to gain political power. (Van de Walle 2006, 84)

In sum, given the proclivity for such flexibility within the Kenyan political class, public opinion constitutes, at the very least, an annoying constraint. And beyond election-related polls, survey findings on public issues generally, by giving voice to the inchoate (and thus largely silent) public, are a threat to the sway that elites have tended to enjoy over the masses. As Bratton and others (2005, 347) suggested in another context:

As well as being materially underdeveloped, sub-Saharan Africa suffers acute scarcities of public information. Especially in rural areas, the everyday lives of ordinary Africans are inward-looking, absorbed in community affairs. When information about national and international public affairs is hard to come by, knowledge becomes key.

For Kenyans, thanks to the polls, such ‘information’ has come to include what they think about their own living conditions and the suitability of aspirants for public office, for example, and the connection between the two.

A second relevant aspect of Kenyan political culture is the nature of public communication. As has been noted in different contexts (Parkin 1975; Haugerud 1993), this is but one in a larger set of control tools employed by those in (or aspiring to) positions of authority. Even in the current multiparty era, therefore, ruling party and opposition leaders alike tend to address their rally audiences in modes that negate the very pluralism they enjoy in doing so, employing such ‘congregational’ phrases as, ‘You all support us in (name of party) don't you?’, ‘Everyone agrees that (whatever issue is mentioned) don't they?!’, and so on. Viewed thus, polling's generic lexicon – slicing up a population into segments of different experiences or views – is inherently devoid of the sweeping hyperbole that typically fills such speeches.75 Their efforts to ‘manage’ poll results, exploiting those that flattered them and discrediting those that did not – without appearing too contradictory, or self-serving and thus insecure in the process – thus revealed the value of and need for a quite new set of communication skills.

Lesson two: Expectations and violence

In the aftermath of the post-election crisis, two key questions were raised regarding the polls’ contribution to it, one on each side of the partisan divide. The first was whether, without them, the ODM side would have accepted the outcome or at least not have resorted to anything like the scale of violence they did in rejecting it. Given the attention paid especially to the flawed performance of the ECK in maintaining communication links with its field officers, the anomalies in the vote-reporting and tallying exercise, and the speed and manner of Kibaki's furtive swearing-in,76 few have suggested any direct causal link77. Yet several prescient warnings well before the election included the following:

The just released Steadman Group opinion polls appear to have split the nation down the middle. With the division between the two antagonistic camps – that is, the ‘Yes’ and the ‘No’ to the opinion poll – the great nation of hardworking people has been taken back to the Banana versus Orange scenario of November, 2005.

My final submission is that we need, as Kenyans, to appreciate that as much as we want to have opinion polls conducted, we must be vigilant that they should not be allowed to divide the country. (J. Gathogo, ‘Popularity polls posing big threat to our unity’, Kenya Times, 8 April 2007)

And afterwards, there was widespread agreement that the polls’ impact went beyond simply helping to energise the largest voter turnout effort in Kenyan electoral history:78

The opinion poll figures contributed to the frenzy, and the verdict of the presidential election being ‘too close to call’ by the most respectable of the Poll Houses, Steadman Associates [sic] helped animate a determination by voters that saw an unprecedented turnout throughout the largely peaceful polling day on December 27. (Kenya National Commission of Human Rights 2008, 31)79

As allegations of vote-rigging stoked protests and ethnically charged riots across this East African country Sunday, President Mwai Kibaki was declared the winner of the closest presidential election in Kenya's history and swiftly sworn in for a second term.

As Kibaki took the oath of office on the green statehouse lawn, gray smoke rose from the sprawling tin-shack slums that are the strongholds of his main challenger, opposition firebrand Raila Odinga, who had been leading in opinion polls and refused to accept results he said were fixed.

A close legal adviser to Odinga said the opposition leader would not challenge the results in court, which could take years, but would ‘take our case to the court of public opinion’, the streets. (Washington Post Foreign Service, 31 December 2007)

The second question was whether these same polls had convinced the Kibaki side that stealing the election was probably the only way they could retain power, and that if they elected for this option, the amount of theft required would be tantalisingly small.80 While the answer to this question is equally elusive, the view of one member of a US (International Republic Institute) election observer team (in his testimony before the US Senate in the wake of the violence) is instructive:

Although polls indicated that the election was too close to call and that the temptation to engage in fraudulent practices by both sides was therefore very high, most Kenyans as well as the international community believed that the leaders of both of the two largest parties – the Party of National Unity and the Orange Democratic Movement – and the Electoral Commission of Kenya would rise to the occasion. (Barkan 2008)

Both questions can also be – and have been – posed with regard to the future. First, will losers in a close election who have led in the polls be encouraged thereby to resort to violence to claim their ‘rightful’ prize? Second, will those trailing in the polls and their supporters either (a) resign themselves to (and thus ensure) their own defeat; or (b) attempt to abort the election if they are too far behind to be realistically hopeful; or (c) try to steal it if that margin is small enough to put victory within reach?81 To these, another less controversial question may be added: will pre-election poll ratings, given their enhanced credibility, encourage more informed bargaining between key players before election day, and more strategic choices by voters when it arrives?

However inconclusive the answers to these questions, it seems that when the campaign battle lines are so intense, where there is such a backlog of socio-economic grievances, and where the integrity of those responsible for overseeing the electoral process is subject to doubt, close results (especially in the absence of a runoff provision) are more likely to be rejected if they fail to replicate those of late-hour pre-election opinion (let alone carefully executed exit) polls; whether such doubts are accompanied by unrest depends on many other factors, however, not least the verdict of credible election observers, as well as the option of recourse to a judicial review process broadly perceived as fair.

A preview post-script: A 2008 poll on the 2007 polls

A: Are you the white man from Steadman?! Your work is very good!

Q: Yes I am. But why do you say that?

A: I think you must be very close to God!

Q: Why do you say that?

A: Because you tell us Kenyans something, and then we find out it is true!

Q: I don't think that's because we're close to God; it's only because we're close to Kenyans, because whatever we say, we find that out from Kenyans.

EN: Q & A (in Swahili) with a petrol station attendant, initiated by the attendant; Nairobi, 10 February 2009

As the immediate post-election crisis recedes, Kenyans continue to express contrasting views regarding the contribution of opinion polls to their country's democratic development. A preliminary investigation of this issue was undertaken through several questions in the last (December) Steadman SPEC survey of 2008, referring specifically to that election. Awareness of at least some opinion polls during 2007 was acknowledged by 69%. Additionally, more than half (58%) stated that they thought these polls had influenced their voting in one way or another: by getting a better idea of their preferred party and candidates’ chances of winning, or by encouraging (or discouraging) them to contribute to campaign efforts or even going to vote at all.⁸¹ Finally, whereas a similar majority (55%) said that they feel such polls strengthen Kenya's developing democracy, just 12% believed that they weaken it, while the remainder saw them as having no impact, or were uncertain as to what this impact might be (17% and 15%, respectively). Of relevance here too is the finding of an earlier (March) 2008 SPEC survey that found that Kenyans view opinion poll researchers with considerable trust, just 3% claiming to have ‘no trust at all’ in them.82

Such tentative findings invite further research. In the meantime, several tentative conclusions may be offered regarding this aspect of Kenya's evolving political society. One is that while the polls probably contributed to the post-election violence in the two ways noted at the outset, they were not among its proximate causes, any more than was having the election itself.83 In any case, even if there had been no published polls, it is likely that the results of those undertaken confidentially by the contending campaign teams would have leaked out, provoking even more tensions and misinterpretation than those that accompanied the ones that were released and disseminated.

Yet the potentially divisive (if not actually destabilising) impact of particular findings should also not be ignored. Such potential explains, for example, the general aversion in Kenya to releasing results that are correlated with ethnic identity, even when such demographic data have been obtained. More specifically in the 2007 case, the suppression of the International Republican Institute's exit poll that showed Odinga winning by somewhere between 3% and 6%,84 in the context of first, the violence, and then, the African Union-led mediation talks,85 was allegedly justified by some weighty actors on just such grounds.86 And in the context of the current grand coalition government that capped those talks and brought the post-election violence to an end, it appears to explain the annoyance (if not anger) expressed by several cabinet ministers to polls during 2008 comparing the performance of leading public figures, especially the president (Kibaki) and the prime minister (Odinga), in which the latter repeatedly scored higher.87

Yet whatever their risks, though, and however irritating if not downright ‘subversive’, some may find them, it is hard to imagine this opinion poll genie being put back entirely in its bottle88 even if the next election might well see some restriction, whether by industry-self or statute, in the form of some clear embargo time-period before the election that prohibits the release (if not the conduct of) polls.89 Whether their role will increasingly conform to the normal, if not banal one played in their place of origin (that is, in mature, liberal democracies), however, remains to be seen. For like other good governance ‘imports’ (elections and multiparty politics, human rights conventions, anti-corruption measures, and so on), we should expect that their actual functions in ‘alien’ settings will reflect those settings to a significant degree. The use and impact of such instruments thus call for continuing examination, as their local ‘domestication’ remains very much ‘in progress’. The coming years – and the polls themselves – will add clarity to this issue, among the many others that Kenya now faces.

But domestication will also depend upon the further deepening of democracy itself. Yet if there is one conclusion that emerges from the entire 2007 election experience, it is that such a unidirectional path of political change in Kenya cannot be taken for granted, whatever the state of public opinion as captured by the polls.

Note on contributor

Thomas P. Wolf (DPhil in Comparative Politics, University of Sussex) is an independent consultant in Nairobi, having undertaken assignments for the Ford Foundation, Transparency International, UNDP and DANIDA among other agencies, but is best known for his association with the Steadman (now Synovate) Group's opinion survey work. Earlier he taught at the University of Nairobi, and served as Democracy/Governance Adviser for USAID/Kenya. His published works include ‘Immunity or accountability?: Daniel Toroitich arap Moi, Kenya's first retired president’ in Legacies of power (eds. Southall and Melber 2006).

Notes

1. Such accusations came from all three main parties. Subsequent (minor) errors of fact and grammar appearing in quotations are generally ignored.

2. Captions have been retained in press citations to indicate the gist of the stories themselves.

3. In their exploration of opinion survey work in Africa, for example, Bratton and others cite only a single pre-election poll, conducted in Kinshasa in the Democratic Republic of Congo in 2001 (2005, 52). Mattes (2007), reviewing the challenges of conducting opinion research in ‘emerging democracies’, makes no reference to such polls whatever. For his part, Lindberg, referring to the utility of post-election polls, laments: ‘We can measure elites’ behaviour, but we lack reliable indicators of voters’ acceptance of the electoral process and its results, since that would require a kind of voter survey or behavioral data that does not exist in Africa – yet’. He then acknowledges the emerging Afrobarometer data set, but fails to mention that it does include questions that assess the most recent election in each country (Lindberg 2006, 43, 199). For comparative data from the Asian polling experience, see Mangahas (2000).

4. Kibaki took 63% of the vote as against 32% for Kenyatta.

5. ‘Grand coalition government’, PowerPoint, 8 September 2008 (www.gallup.com/poll/111622/Kenya).

6. Curiously, this finding (among others released at a well-attended media briefing in Nairobi) attracted little attention. The survey did not include any questions about actual voting in the election.

7. The company was bought by Synovate – a leading international market research firm – in July 2008.

8. The remaining 12% cited a variety of other reasons. It also found that 61% had ‘no confidence at all’ in the ECK, contrasting with only 9% who held this view two weeks before the election. The Constitution of Kenya (Amendment) Act of 2008 (passed on 16 December) brought the life of the ECK to an end, replacing it with an Independent Interim Electoral Commission (the IIEC, with a mandated life-span of 18 months or within three months of the ratification of a new constitution). It is tasked with re-establishing credibility of the country's electoral machinery, beginning with the creation of a new (and reliable) voters’ roll. This followed a key recommendation of the Independent Review Commission (IREC), to restructure the ECK radically (2008, 153–5).

9. Whether such lagging poll figures had any impact on Kibaki's decision to make such unilateral appointments must be left to conjecture (but they may have affected his campaign expenditure; see Willis 2008).

10. See also, B. Muluka, ‘Country stares at signs of a major breakdown’, Saturday Standard, 2 February 2008 (who here also made reference to his dire predictions in an earlier piece); A. Ahmednasir, ‘Are our courts ready for general election fall-out?’, Sunday Standard, 4 November 2007. Less than three weeks before the election – in the wake of at least one campaign-related killing – all the main presidential candidates also acknowledged such dangers and pleaded for peace (‘Poll: Kibaki and Raila in appeal for peace’, Daily Nation, 7 December 2007).

11. In addition to the countless media reports (see below), there were many panel and phone-in talk shows on radio and television during this period that examined both specific results and polling in general, in many of which the author participated, as did representatives of other survey firms.

12. For example, T. Wolf, ‘Secret weapon in presidential race’, Sunday Standard, 11 November 2007. Reflecting a concern that such a rejection was at least possible, the author received expressions of gratitude from several diplomatic missions following the publication of the above article, based on the hope that if the candidate leading in the final polls lost the election, he would be more inclined to accept defeat on this basis (The author had no such intention in writing the piece, however.).

13. The Statistics Act (2006) was initially seen as potentially restricting the work of polling firms, but has yet to prove so. Whether those in power ever contemplated imposing any such restrictions during this period is unknown; interviews with several senior politicians who served in the ninth parliament identified only one cabinet minister who floated the idea – but subsequently dropped it. The future use of this Act for such a purpose remains uncertain.

14. Under single-party rule, no such polling took place (as was apparently the case elsewhere in Africa).

15. The Afrobarometer project has so far refrained from undertaking a survey in Ethiopia on the grounds that the environment is insufficiently conducive to the public's free expression of views. The first Afrobarometer survey undertaken in Kenya (2003) was itself delayed until the Moi era had ended, on similar grounds (personal communication, Senior Afrobarometer staff, Michigan State University.) Analyses based on these first two surveys are provided by Wolf et al. (2004), and Logan, Wolf and Sentamu (2007).

16. Throup notes that this latter company's work was ‘widely considered as the most scientifically based and accurate of the Kenya polls’ (2008, 290). It began to conduct political surveys only in 2002, after obtaining Gallup International-affiliate status, though it had become prominent in the fields of media monitoring and market research over the previous two decades under its founder (Roger Steadman, a former Unilever employee). The Market and Social Research Association of Kenya (MSRA), established in 1996, has nine corporate members; of the political polling firms, only Steadman is among them.

17. Much of this recent Kenyan experience echoes that of polling's early history in the US, especially regarding the development of methodology and the media's – as well as the public's – response. In both settings, for example, pre-election polls were used to test methodology; according to George Gallup, this was ‘the only useful purpose served by election forecasting’. Such validation was thus seen as a major attraction for more lucrative commercial and market survey work, at least for organisations whose polls were confirmed by the actual election results. As Igo notes, ‘New clients viewed elections as litmus tests for methods’ (2007, 134–5).

18. Such trust seems largely a function of media plurality. The state's broadcasting monopoly ended in the 1990s: by 2007 only 23% of all radio-listening time and 32% of all television-watching time remained with it (Media Audience Survey, Media Monitoring Division, the Steadman Group, 2007).

19. Teachers scored the highest, at 90%, slightly ahead of doctors and nurses.

20. Whether the Political Parties Act of 2008 that took effect on 1 July of that year significantly alters this situation remains to be seen (Wolf 2008) even if, as of February 2009, only 47 parties had complied with its quite stringent registration provisions (‘List of Current Political Parties in Kenya’, Office of the Registrar of Political Parties, Nairobi). Of these, over 40 had been among the 168 registered as of December 2007. Several parties (including one already registered) have filed suits in the High Court challenging the Act's constitutionality, however, with their outcomes pending as of June 2009.

21. For a recent consideration of this issue, see, Bratton and Kimenyi (2008). One local columnist described Kenyan political parties and alliances as nothing more than ‘drifting tectonic plates temporarily bound by the glue of mutual gain that they guarantee our tribal headmen’ (K. Tanui, ‘Ruto should tread carefully’, The Standard, 13 March 2009). Yet such a generalisation should not be overstated, as sub-ethnic and ‘proto-ideological’ (or ‘class’) divisions also matter, the split within the Kikuyu community between the 1992 presidential campaigns of Kenneth Matiba and Kibaki being one example. Further, certain policy differences have come to matter more recently, such as those over-centralised versus more dispersed constitutional authority (Harneit-Sievers and Peters 2008, 136).

22. Such cross-ethnic alliances are also encouraged by the requirement that a successful presidential candidate obtains not less than 25% of the votes cast in at least five of the country's eight provinces.

23. MacArthur (2008) and Gona (2008) show how these factors played out in the 2007 election in Western and Coastal provinces respectively.

24. Such ‘non-appearance’ reflects questionnaire design. In questions such as, ‘Which cabinet ministers do you think have performed best over the last three months?’ no names are read out, but those certain to be the most frequently mentioned are listed/pre-coded on the questionnaire to ease the interviewers’ task of recording responses. Other names mentioned (usually at frequencies of at most 2%) are then recorded manually under ‘others’. Those who complained that ‘You never asked about me!’ were apparently unaware of this. For example, ODM ‘Pentagon’ member Joseph Nyagah sought to dismiss the polls as fraudulent on that basis, while threatening to have the polling industry subjected to statutory regulation (Interview, Family-TV, 7 August 2006), while later, the ODM-Kenya presidential nominee-aspirant Dr J. Ojiambo complained that ‘it was unfair for the Steadman Group to leave her out of its popularity ratings’ (‘Steadman polls fake, says Ngilu’, The Standard, 3 April 2007).

25. Such concerns echoed the early US experience: ‘Many Americans – from members of congress to the fabled “man in the street” – were suspicious of pollsters’ methods as well as their potential sway over elections … Opinion polls were routinely accused of creating a “bandwagon” effect and of muting or corrupting public discourse … [T]he pollsters often received voluminous reactions to their findings from the public through the post, including hate-mail’ (Igo 2007, 168–9). An example of empirical research on this ‘bandwagon’ issue in an established democracy is provided by Ceci and Kain (1982).

26. There were two polls, one six weeks and the other a week before the event. This difference was the same in both, leaving out the ‘undecided’ respondents. Wolf (2005) explains why the second of these polls was withheld until after the results were announced, and how this was done so that, whatever those results, its release was guaranteed.

27. A number of factors have been identified as responsible for ‘a relatively innumerate and skeptical political class of elected leaders, policy-makers, civil society leaders and news journalists’ in much of the developing world (Mattes 2007, 114). The requirements for random-sample survey research, with particular reference to Africa, are available at the Afrobarometer website (www.afrobarometer.org).

28. The author participated in several pre-election workshops for media practitioners of this nature. At the same time, the ‘democracy donors’ declined to fund the inclusion of a unit on survey methodology and the use of polling results in the curriculum of the National Civic Education Programme (‘NCEP-II’, or Uraia), reportedly because of budgetary constraints.

29. For example, it was asserted in a campaign corruption study that a Kenyan polling firm accepted a Shs 20m. bribe in exchange for ‘fake’ polling results (Campaign for Accountable Financing 2008, 55). Numerous attempts by the author to encourage the sponsoring organisation's executive director and the report's chief researcher to identify the firm, or at least indicate the type of evidence upon which this claim was made, proved fruitless. More generally, those unhappy with poll results frequently demanded to know who had commissioned them, since ‘he who pays the piper calls the tune’, notwithstanding the often-repeated fact that at least in Steadman's case, these surveys generally had multiple sponsors, sometimes as many as a dozen.

30. It should be noted, however, that as early as December 2003, in the first poll on the subject following the 2002 election (conducted by Strategic Public Relations), Musyoka outscored Kibaki: 31.5% to 27.9%. (The actual question posed was, ‘Who should run for president in 2007?’; ‘What Kenya says about Kibaki, Narc’, East African Standard, 27 December 2003).

31. Odinga first ‘gently’ announced his presidential aspirations on New Year's day (‘Kalonzo: Kenyans want me for Presidency’, Kenya Times, 4 January 2006). See also ‘Kalonzo's future in LDP uncertain: Differences over presidency could force MP out’, The Standard, 6 April 2006 ; ‘Raila is no longer content just playing the kingmaker’, Sunday Nation, 9 April 2006).

32. Note that in Steadman's December 2005 poll, Odinga was the presidential choice of only 3%. In the same survey, 61% still named him as their preferred prime minister were that office to be created; this compares to just 11% who identified (second-placed) Musyoka in this regard.

33. Suspicions arose in certain quarters that his refusal to back Odinga even after it became clear that the latter was the most credible challenger was part of his ‘real’ role in the campaign: that of a ‘spoiler’ on Kibaki's behalf. Such views were encouraged by the ‘secret visit’ he was reported to have made to State House early in 2006 (‘Kalonzo, Kibaki in secret meeting’, The Standard, 25 February 2006). Musyoka vehemently denied having had any such meeting (‘Angry Kalonzo now accuses OMD rivals: They want to block my presidential bid, he says’, Sunday Nation, 26 February 2006). Shortly thereafter, rumours circulated that Musyoka was ‘on the verge of quitting ODM and looking to strike a deal with Uhuru Kenyatta as his running mate’ (‘Kalonzo confident of win in LDP poll’, Sunday Nation, 2 July 2006).

34. The president ‘finally’ declared his interest in a second term on 26 January 2007 (‘I will be in the race again, says Kibaki’, Saturday Nation, 27 January 2007), which appears to account for the 9% rise in his rating over the next quarter.

35. Formally establishing the ODM as a party was seen as a way to take advantage of the now-popular ‘orange’ symbol assigned to the ‘No’ position in the referendum (see below), based partly also on its symbolic link with the recent ‘people's revolutions’ of Eastern Europe. Achieving a more equal balance of interests between the leading members of the ‘No’ campaign than existed within LDP was another aim. In seeking to register the party immediately post-referendum, however, it was found that the name had just been acquired by a politically ambitious lawyer, Mugambi Imanyara, making it necessary to register ‘ODM-Kenya’ as an alternative. After the split with Musyoka, the Odinga faction managed to ‘aquire’ the party from Imanyara (in exchange, some asserted, for him later to become a nominated MP) and thus to make use of the original ‘Orange’ name. For background, see http://en.wikipedia.org/wiki/Orange_Democratic_Movement; ‘The Orange Democratic Movement: Not out of the woods yet’, The News Review, 21 August–8 September 2007. One ODM ‘insider’ reported that in return Imanyara was to be made a nominated MP.

36. According to the International Republican Institute's exit poll, while only 8% of those claiming to have voted for Odinga described the country's economic situation as either ‘good’ or ‘excellent’, 61% of those who said they voted for Kibaki did so (Gibson and Long 2009, 5).

37. In September, his choice as an electoral vehicle for the coming campaign was finally announced: the Party of National Unity (PNU), just established for this explicit purpose. Two main factors accounted for this. First, his allies had never gained full control over NARC, leaving uncertainty over its nomination outcome. Second, although NARC-Kenya (registered a year earlier, see below) had been registered primarily for this reason, it was viewed as having too much of a ‘Mount Kenya’ hue, when a broader (ethnic) appeal was considered essential.

38. To assist illiterate voters, the ECK assigned these fruit ballot symbols to the opposing sides.

39. By November, the Kibaki/PNU campaign was out-spending ODM's presidential effort by more than 3 to 1 (‘Political campaigns media coverage and advertising expenditure’, Media Monitoring Division, the Steadman Group, 23 November 2007). In his analysis of campaign advertisements and expenditure, Willis notes the mounting concern in the Kibaki campaign over the president's laggard poll performance, resulting in the release of more aggressive media products (Willis 2008, 266).

40. Given the widely assumed association of several individuals in this firm with the Odinga campaign, such an insinuation lacked even minimal credibility.

41. The Odinga side brushed off these same results as ‘inconclusive’, and no decision was made on the issue (‘Revealed: The deals and threats at ODM retreat’, The Standard, 16 October 2006).

42. This contrasted starkly with the late 2005 to early 2006 period when Musyoka enjoyed top ratings (as shown in Figure 2), and Kibaki stalwarts attacked Steadman and, indeed, the polls in general (‘Kenyans want me: Public says I succeed Kibaki’, Kenya Times, 4 January 2006).

43. This party was officially launched at the beginning of June, ostensibly to put up pro-Kibaki candidates for several of the by-elections necessitated by the loss of five MPs in a plane crash two months earlier. Its main purpose, however, was to provide the president with an electoral vehicle for 2007, given his failure to gain control over the bitterly-divided NARC; see, ‘The big fallout’, The Standard, 5 March 2007; ‘Narc-K's ‘winning strategy’ for 2007’, The Standard, 10 April 2006. It subsequently won three of the five by-elections. As noted above, the Kibaki team later settled on another new party (PNU) that fielded 135 parliamentary candidates (and won 43 parliamentary seats, as opposed to ODM's 99). This was after the initial proposal to have Kibaki alone run on a PNU ticket while all his affiliated parliamentary candidates would contest on their respective party tickets was scrapped so as to provide him with a sufficiently large ‘team’ of fellow-PNU MPs and at least reduce the competition between such pro-Kibaki candidates, especially in those constituencies where such competition could hand victory to the opposition (which occurred, nevertheless, in about two dozen parliamentary races).

44. Some journalists also had difficulty with this ‘contradiction’.

45. Ironically, such a charge echoed another one made in the run-up to the referendum when several Steadman polls had shown the president's popularity in decline; see ‘Polling firm caught in name brawl’, Sunday Nation, 26 June 2005, and the rebuttal (published as a letter to the editor), ‘Gallup International has its network in Kenya’, Daily Nation, 6 July 2005. The fact that few would be expected to know that Gallup International (UK), a non-profit body established by George Gallup, has no legal connection with Gallup, USA (a for-profit company) gave some credibility to this accusation on both occasions.

46. Odinga made the comments in a radio interview (Classic-FM, 13 March 2007).

47. The results were obtained by employing 2002 election and 2005 constitutional referendum turn-out figures in combination with the poll findings of individual strength in particular areas, (‘Best man to beat Kibaki – Revealed: ODM secret report says Musyoka is best bet…’, Sunday Nation, 17 June 2007). Just how this was done was explained earlier by one of those who reportedly participated in the exercise: J. Okungu, ‘Emerging tribal arithmetic ahead of poll: Will it ruin or propel ODM-K to power?’, Kenya Times, 17 January 2007.

48. As the election approached, a number of parliamentary aspirants from various parties also privately commissioned constituency-based surveys, both before and after the party nominations.

49. At least one, conducted two months before the election, was reported to have confirmed Odinga's widening lead (‘Polls: State Intelligence report alarms Kibaki’, The Weekly Citizen, 15–21 October 2007).

50. This was a code phrase for Odinga's ethnic ‘baggage’, implying that a (Nilotic) Luo is unelectable in (predominantly Bantu: about 65%) Kenya. The Luo comprise about 12% of Kenya's population, and Musyoka's (Bantu) Kamba about 11%.

51. See J. Mwazemba, ‘Healthy respect for opinion polls wise’, Sunday Standard, 21 May 2006, and P. Mwaura, ‘Politicians shouldn't look down upon opinion polls’, Saturday Nation, 20 October 2007.

52. NTV 9pm News, 13 October 2007. I. Ndirangu expounded this ‘Western agenda’ argument: ‘Why the West will back ODM-K in '07 elections’, The People Daily, 4 December 2006. See also the personal attack on the author by PNU campaign activist and former Nairobi mayor J. Aketch: Letters, The People Daily, 16 October 2007; the former's reply (‘The truth about survey results’) was published by the same paper on 1 December 2007.

53. The ‘espionage’ claim was included in the news broadcasts of several media houses; the deportation demand was not (Confidential interview, Kenya Television Network reporter, 15 October 2007, Nairobi). Typical among the pro-Kibaki mobile phone texts circulating during this period was this one: ‘Busted: Steadman now infiltrated by CIA … US (now looking 4 a stooge President) proxy war against China! They openly interviewed a biased sample returning from the ODM Launch in a bid to influence the polls to get their stooge [elected]. Voters be [sic] ware!’ (15 October 2007).

54. Similarly misplaced was the demand made earlier by the assistant minister of information, K. Wamwere, that Steadman as a ‘foreign-owned company’ should be shut down by government, since ‘no foreign polling firm would be allowed to operate in England or other Western countries’ (‘Cross-Fire’, KISS-FM, 23 July 2006). He was evidently unaware that the company is Kenyan. Later, he attacked Steadman's September 2007 poll (in which Kibaki trailed Odinga for the first time) as ‘a tool for propaganda’ (‘Poll being used for propaganda, Koigi’, The Kenyan, 5–11 October 2007).

55. This applied to Strategic Public Relations and Infotrak; for example, one of the latter's founders, J. Okungu (see above), was widely identified as a member of the Odinga campaign team. However, the fact that neither of these firms’ public profile was equal to Steadman's meant that rather less general attention was paid to them.

56. The various abuses of state power in connection with the ‘Yes’ referendum campaign, for example, were still fresh in ODM minds (Kenya National Commission of Human Rights/Kenya Human Rights Commission 2006).

57. Such fears were bolstered by the president's unilateral appointment of 19 of the ECK's 22 commissioners whose terms expired in the months before the election (‘Kibaki picks five new ECK bosses’, Daily Nation, 30 October 2007) and by the clearly biased campaign coverage of the Kenya Broadcasting Corporation, violating its statutory obligation to neutrality and balanced coverage. The inconclusive investigation into the nighttime raid on ODM's campaign headquarters on 9 September during which ‘computers with vital information’ were stolen just added to such concerns (‘ODM to await statement on raid’, Daily Nation, 12 September 2008). Also feeding such anxieties was the reassignment of many ECK officials in various parts of the country less than two months before the election (‘Spotlight on ECK: ODM alleges manipulation as Commission moves 45 offices in Nyanza, Eastern and Central provinces’, The Standard, 5 November 2007).

58. The fact that PNU campaign actors now praised Steadman apparently just heightened such suspicions; see again, ‘Mixed reaction to opinion poll’, Sunday Nation, 11 November 2007.

59. Steadman justified its position by arguing that as long as the voters’ registers remained open (until parliament was dissolved), it did not want to pre-judge the final national distribution of registered voters. At least one other firm (Strategic Public Relations), however, had been using rolling ECK registration figures throughout.

60. This ‘psychological manipulation’ plot was spelt out in Sigliare, ‘The Steadman poll spell?’ (www.jaluo.com/wangwach/200710/con_sigliare). For an instant rejoinder, see Kambona, ‘URGENT reaction Edawrd's [sic] ‘Flipside of Steadman polls’ (www.jaluo.com/wangwach/200710/Ngoje-ogalo_Jarongo). Shortly thereafter, the author received an email from a local journalist claiming that ‘a source has revealed’ the details of a meeting Steadman had held with the Kibaki campaign team ‘where this Friday's poll results were negotiated’ in order to reduce the gap between the leading contenders to two points, and concluded by asking: ‘This may not be true, but does it bother you that it touches on your credibility?’ Steadman's response to such charges was to elaborate its sampling methodology in a press release (27 November 2007).

61. This claim was made in a widely circulated email message and repeated by a caller to Radio Umoja-FM on 12 November, who identified himself only as an employee of the government's intelligence service, while insisting that he was making this revelation ‘with the interests of the country at heart’.

62. The First Lady's recent history of publicity succumbing to fits of rage and physically accosting several individuals gave some credence to this claim, however; see, for example, ‘First Lady on the loose’, The Standard, 4 May 2005, which describes her midnight ‘raid’ on the Nation Media Group, during which she slapped a TV journalist from the Standard Group who had arrived to video-record the bizarre event.

63. The latter claim was circulated via mobile phone messages as well as over the Internet (http://www.wanjuguna.blogspot.com/2007/12/breaking-newstom-wolf-pulls-out-of.html). The author received some half-dozen phone calls from media houses seeking confirmation of this ‘breaking news’.

64. Such a full-page advertisement costs about KShs. 400,000 (that is, $6,000). This had followed the (uninvited) attendance of several ODM officials at Steadman's 7 December media briefing, where they distributed what were claimed to be the ‘true results before Steadman doctored them’, and the ‘fake’ results that they had expected Steadman to release. The latter showed Kibaki with a clear lead; they were thus at a loss to explain why the actual Steadman results showed otherwise.

65. Some saw his reference to such a ‘miracle’ as an expression of his evangelical Christian faith rather than simply rhetorical bravado or a lack of realism. At the same time, his post-election appointment as vice-president, raising his profile well ahead of the 2012 election, later encouraged some to believe that such a ‘miracle’ is still very much on the cards or even that it had already occurred – at least in part.

66. To minimise the possibility of any damaging rumours, the meeting was held in public (in the snack bar of a major Nairobi hotel), which is how the media came to know of it.

67. As noted above, however, it appears that publicising his low poll numbers had little effect on his eventual vote total. At the same time, a member of the ODM-K campaign team complained to the author that their candidate's laggard poll ratings were hurting fundraising efforts (Confidential interview, 15 September 2007, Nairobi).

68. Previous reports in the Daily Nation on these firms’ poll results had indicated that Infotrak-Harris sampled in some 30 districts, and Consumer Insight, between 40 and 50. For these three firms, the margin of error was just under +/− 2%; for Gallup, +/− 3%, and for Steadman (in its final, enlarged sample) +/− 1.25%. In addition, whereas Steadman was using a 65–35% rural-urban distribution (closely reflecting the national population), Infotrak-Harris was reported as using a 56–44 one; it is not clear what distribution the other three firms employed, though a senior manager from Consumer Insight told the author in November that his firm used a ‘more urban-based sample than you do at Steadman’.

69. Curiously, however, neither the Nation (at the time, or later) nor Expression Today's post-election consideration of the pollsters’ performance (‘Polling and the Kenyan media’, op. cit.) ever sought to explore such discrepancies.

70. Like Steadman, the ownership and most senior staff of Consumer Insight are from the Mount Kenya region.

71. No doubts were ever publicly raised by ODM about the work of these two latter firms.

72. They were transported at night in a fleet of privately owned Nairobi city buses, the entire operation first coming to light through media reports rather than via a government or PNU announcement. Once so known, it was justified on the grounds that the safety of regular party agents in this ODM bastion could not be guaranteed. Indeed, five administration police officers were killed by members of the public just before and after the election in several locations of Nyanza province (from press reports compiled by the Kenya National Commission of Human Rights 2008, 93).

73. Even after the election, Musyoka dismissed Steadman's polling work on several occasions, its proven (basic) accuracy notwithstanding.

74. At the same time, occasional concerns have been expressed that ‘transient’ or ‘ill-informed’ public opinion sometimes plays too great a role in such political systems (for example, Wesson 1988, 164–5).

75. In the same vein, Ndegwa identified a central shared feature of both ruling and opposition parties during the later Moi era: ‘Both share fundamental conceptions of leadership and of the relationship between the state and its citizens: a patriarchal, patronising, elitist and, ultimately, disempowering attitude towards the citizenry’ (1998, 209).

76. Although its terms of reference included an evaluation of the ECK's performance with specific regard ‘to assess the integrity of the results of the entire election with special attention to the presidential contest’ (IREC 2008, 9), on numerous occasions the commissions's chair explicitly excluded such an obligation from its mandate. In this vein, IREC's final report argued that ‘[It] is not advisable to conduct any form of statistical evaluation of results from the December 2007 elections in Kenya or to draw any conclusions on that basis, nor to conduct some kind of more advanced psephological analysis … The deplorable conclusion is thus that ECK turnout data and election results for individual candidates are so error-infected that they should not be used for any kind of statistical analysis’ (2008, 136). The European Union Election Observer Mission was of the same general opinion (2008, 3, 33–6). Such views evidently help to explain the (yet tobe adjudicated) suit filed in 2009 calling for the dissolution of parliament on the grounds that if the ECK's performance warranted its own demise, the election of the house itself lacked legitimacy; see ‘Aggrieved voter's case to be heard in Nairobi’, Daily Nation, 9 March 2009.

77. Some who testified to the commission investigating the election crisis (the Kriegler Commission) had a contrary view, however; see ‘Opinion polls blamed for election chaos’, Daily Nation, 1 July 2008.

78. The official overall figure is about 70%, but with significant variations. For example, in Musyoka's (Kamba) area of the eastern province, turnout approximated this national average, but was considerably higher in Kibaki's central province (between 80 and 90%), and higher still in most of Odinga's (Luo) Nyanza. This contrast remains valid even if some of the constituency figures in these latter two provinces may have been fraudulently inflated by the use of ‘ghost’ voters or ballot stuffing (as has been claimed, or at least surmised; IREC 2008, 8 (Harneit-Sievers and Peters 2008, 135).

79. Using the ECK's turnout figures, there is no significant deviation from the company's final poll, within the stated margin of error.

80. This is not to imply that without these ‘early warnings’ from polls, election theft would not occur in Africa or elsewhere. Case (2006) provides many examples, though without mentioning the use of fake or doctored polls, let alone electoral theft in anticipation of defeat indicated by (genuine) polls.

81. No question asked whether knowledge of these poll figures actually made them change their vote.

82. By contrast, 18%, 21% and 22% had a complete absence of trust in the judiciary, the president and the police respectively.

83. This view accords with that of IREC, which argued for ‘maximising journalists’ skills and knowledge vis-à-vis electoral reporting’ rather than imposing restrictions, even while acknowledging the contribution that the polls made to the tensions that fed the violence (2008, 63).

84. Gibson and Long (2009); see also, A. Halperin, ‘What's really going on in Kenya?’, Slate Magazine, 2 January 2008; and ‘Raila won’, Nairobi Star, 15 February 2008. It was apparently assumed that making public such additional evidence in support of ODM's claim of a stolen election, at least as regards the presidency, would sabotage the negotiations, or simply further raise the stakes should such an accommodation (a climb-down of some kind by the Kibaki side) not be achieved, possibly leading to even higher levels of renewed violence. An Institute for Education in Democracy exit poll reportedly had Odinga winning by 2%, but ‘due to delays in receiving the findings because of the violence we never released it’ (Confidential interview, 27 April 2009, Nairobi).

85. One source close to the Kofi Annan-led mediation talks confided that the better part of the 6 February session was spent arguing over these various polls, with each side selectively using them to prove who the ‘real winner’ was (Confidential interview, 6 February 2008, Nairobi).

86. In this vein, a Western diplomat remarked to the author shortly after the National Accord was signed, that ‘We really do not want to know who won the election now!’ (Confidential interview, 12 March 2008, Nairobi).

87. This unhappiness was expressed to the author privately, however, not via media comments. There was almost no reaction to the first hypothetical presidential candidate-preference rating, released early in 2009, however, which showed ODM's Odinga only slightly ahead of NARC-Kenya's M. Karua, 19% to 16%, followed by Musyoka at 14%, all far behind ‘none’, at 33%.

88. For example, all television channels now invite viewers of their evening news programmes to ‘vote’ and comment on a daily ‘opinion’ question via mobile phone text.

89. For this last election, all Kenyan polling firms individually set this cut-off date themselves. Spagenburg (2003) has provided a recent global review of this issue, but which includes only two African countries (Nigeria and South Africa, neither of which currently has any restrictions). Igo notes that on repeated occasions from 1932, (unsuccessful) attempts were made by the US Congress to regulate polls (Igo 2007, 169).