6.1 Introduction

The acceptability of an election depends very considerably on the extent to which the public feel the officially announced election results accurately reflect the votes cast for candidates and parties. It depends, too, on factors such as the character of the electoral campaign and the quality of the voter register, but reliable counting and tallying is a sine qua non if an election is to be considered legitimate by its key assessors – the voters.

Counting and tallying during the period 27-30 December 2007 (and even thereafter) and the announcement of individual results were so confused – and so confusing – that many Kenyans lost whatever confidence they might have had in the results as announced. Rumours of rigging and fraud during the counting and tallying process spread like wildfire, and the consequences were tragic.

It was therefore of paramount importance that IREC scrutinise these processes closely in order to enable it to assess whether or not the ECK administered these key elements in such a way that the voters – and others – could have full confidence in them.

6.2 The integrity of the counting, tallying and result announcement system

Integrity in systems or processes such as electoral management refers to systemic safeguards, which aim at reducing the need for personal integrity. It might even be argued that systemic integrity is what separates acceptable management, and therefore safety, from disaster.

While integrity is necessary at all stages of the electoral process, nowhere is it more important than in counting and tallying. To maintain integrity, vote counting must produce results that are and are seen to be valid and accurate, and therefore acceptable to all stakeholders. Some examples of safeguards used to ensure the integrity of counting and tallying are:

* Ballot papers may be given difficult-to-counterfeit security features or require stamping/ signing by the presiding officer (PO)/party agents.

* In some cases, ballots are numbered on the counterfoil and on a detachable section of the ballot itself. When the voter returns with the completed ballot, the presiding officer checks that the counterfoil and ballot-paper numbers are identical, detaches the number from the ballot (to eliminate the possibility of later identification of the vote) and posts the ballot in the ballot box.

* Rules for rejecting ballots should be unambiguous while aiming to retain the voter’s intention. The decision to reject or not should be taken at the polling station level, although procedures for appeal should exist.

* If votes are counted at the polling station, the EMB must ensure there is no possibility of collusion among those present. This is normally done by ensuring the presence of agents of rival political parties. In some cases, however, the EMB ensures that the polling station staff is politically balanced (for instance, by asking political parties to submit names for such positions).

* In most cases, the number of names crossed out on the voter register is compared with the number of ballots in the ballot box. If party agents have copies of the voter register, they can verify the correspondence between these numbers independently.

* After the count, sensitive material is usually placed in tamper-proof bags for secure safe-keeping.

* Tallying is frequently conducted in duplicate to avoid arithmetical errors. For instance, the results at a polling station may be entered twice. If the computer identifies differences between the two entries, it rejects the results, which will then be scrutinised and re-entered.

* While it is always necessary – and unproblematic – for the EMB to announce preliminary (or provisional) results (that is, results which have not yet been formally approved as final by the relevant authority), adequate time must be allowed for the processing of complaints and appeals before the final results can be announced.

The above are only examples of how adequate rules and procedures can help ascertain and sustain the integrity of an electoral system. The following section examines this topic in relation to the 2007 general elections in Kenya.

6.3 Vote counting and tallying in the 2007 elections

After the close of voting, party agents at the polling station may check that the ballot-box serial numbers are identical to those registered at the opening of the poll. Once the physical space has been cleared, counting starts in the presence of these party agents.

Ballots are unfolded by the counting clerks (assistants) and sorted according to candidates’ names, the counting clerk/presiding officer showing each ballot to the party agents before placing it in the correct pile for the candidate in question. Valid votes are bundles in fifties and counted. Rejected and disputed ballots are kept separate and stamped, but disputed ballots are treated as valid until the returning officer (RO) has decided what they are: disputed ballots are bundled separately according to candidate.

The counting clerks count the valid votes for each candidate, and the presiding officer announces the result for each presidential candidate in relation to the presidential election, each parliamentary candidate in relation to the parliamentary election, and each civic election candidate in relation to the civic election. The counting clerks then complete two forms 16A, one for the presidential and one for the parliamentary election, and one form 8 for the civic election, and invite party agents to check the correctness of the figures and confirm that by signing the forms. The Presiding Officer is also required to sign the forms. Copies of the forms are then provided to the party agents.

If agents decline to sign these forms (which they may), the presiding officer must try to establish a reason for their refusal; and if they decline to sign, he must state accurately what occurred. Agents should then (in any case) be given a copy of the relevant form. Copies of the forms are to be affixed where they are easily accessible to the public.

All sensitive material is then packed, sealed and transported to the constituency tallying centre, accompanied by party agents who so wish (and can find transport).

In Kenya, neither the electoral regulations nor the various checklists and training materials are as clear and unambiguous as one would have hoped. The regulations are not precise about the order in which the three counts (presidential, parliamentary and civic) should be carried out, and they do not indicate how to deal with situations where 100% or more of the registered voters have voted (in itself a rather fluid concept, given the acceptance late in the day by the ECK of voting by double-registered voters and the use of “the black book”). However, in some – but far from all

– constituencies, presiding officers were instructed to indicate on form 16A how many voters came from these categories; this information is helpful for the assessment of actual turnout.

To what degree party agents were present or were turned away is difficult to document, since relevant material for such documentation has not been available to IREC. Information gathered during the IREC meetings and hearings indicates that there were problems at a number of polling stations.

In 2007, counting and tallying was a straightforward process, at least in principle. At the constituency tallying centre, the returning officer received the material from the polling stations, checked that all the material was there, and then assumed control – almost took ownership – of it. The returning officer then announced the results from each of the polling stations, as the presiding officers submitted them, and these results were then entered in the relevant cells in the huge form 17A.

Once this was done for all polling stations in the constituency (typically between 100 and 200), the results were added up (column-wise), and the resulting figures for presidential candidates entered on form 16, for parliamentary candidates on the last page of form 17A and for civic candidates on the last page of form 9. When this had been done, the results were announced orally, and Certificates of Results were issued for parliamentary winners (form 17) and civic election winners (form 18C). The entire process is shown in the process-chart below.

The complexity of the process is evident, and it can be no surprise that many errors were detected when IREC analysed a sample of constituencies in some detail.

The returning officer was then required to telephone or fax the information in form 16 (that is, the votes obtained by each of the candidates and the number of rejected votes) to the national tallying centre at the Kenyatta International Conference Centre (KICC) in Nairobi, to a specifically assigned “verification” table. The information was captured on a special, preprinted constituency-specific form and subsequently checked by calling back and verifying the returning officer’s identity and by rechecking the information. The hand-written form was then handed over to the IT department, where the key data were entered into the computer, and a printed version of the form was brought back to the table in question.

The returning officer was then to make his way to KICC immediately, even though some of the training material used the phrase, “within two days”. He had to bring with him all results and relevant tallying forms from the constituency, and in particular the original, statutory form 16. Upon arrival, he was to certify that the content of the printed form produced by the IT department was identical to what he had on the original, statutory form 16 for the constituency.

In the meantime, an ECK commissioner would normally have announced the results from the constituency once the fax/telephone message had been received and been data-captured. If there were no differences when these were checked against the original form 16, the results changed from “provisional” to “final” (that is, the official, ECK-approved presidential results from that particular constituency) because the original form 16 was now with the ECK. What was not foreseen was that the returning officer might have realised that there were inaccuracies – or even more serious mistakes – in his form 16. As the content of form 16 would already have been communicated to KICC (and also been announced), there was a serious problem and the announced result ought to be changed. The ECK had not prepared clear procedures or rules for this eventuality. IREC has learnt that commissioners disagree on whether or not changes to announced results were possible. The same is and was the case even for senior staff, including the Deputy Secretary, the IT manager and table team leaders.

This was extremely unfortunate, as it meant that such necessary – probably unavoidable – changes were not dealt with identically. Some were accepted and changed in the database more or less immediately, others were not, and others again were changed or changed back between the announcement of the presidential winner and 8 or 9 January 2008. This means that some of the results announced on 30 December 2007 or published on 9 January 2008 do not correspond with those on the original forms 16.

6.4 Points and issues raised

Numerous complaints in relation to counting and tallying were made by political parties, individual voters, civil society organisations (CSOs), domestic and international observers, the media and even ECK commissioners and staff. Most of the complaints fall neatly into one of the categories listed below. We look first at problems brought up in relation to the polling station level, then at those at constituency level and last at those at national level.

Polling station level:

Polling station complaints can all be connected to problems caused by ethnic (and therefore also often party-political) dominance in a particular area. This issue was brought to IREC’s attention in one or other form in many parts of the country:

* “Zoning”, that is, a conscious policy of keeping political opponents away from one’s territory, for instance by not allowing them to campaign. Where zoning is practised, free and fair elections cannot take place.

* Agents of rival political parties were expelled from polling stations, in some cases at the time of counting. This complaint was made by interlocutors from all major political parties and IREC heard it so often and so vigorously that there is no reason to doubt its correctness. Such behaviour – whether initiated by presiding officers, supporters of the dominant party or others – is unacceptable: it is a denial of the access to information and transparency which are preconditions for free and fair elections.

* Results from polling stations in areas dominated by one political party were not always reliable. The reason given to IREC is that ECK polling personnel and party agents from the dominant party in the area would obviously agree on the desirability of a good result for that party. And agents for other parties – if present at all – might not necessarily be reliable witnesses of what went on as they might have been bought or threatened. The party affiliation of party agents is in any case impossible to check, as form 16A does not ask that very important piece of information.

* Agents of rival parties were not allowed to accompany ballot boxes to the constituency tally centre. This is most likely true, but it has not been revealed whether this was to provide an unchallenged opportunity for ballot-box tampering – as the suspicion goes – or because of lack of transport for a considerable number of party agents. Prudent behaviour by presiding officers would of course be to provide transportation for as many agents as possible, and primarily with an eye to securing transport for as many rival parties as possible.

* In some polling stations, supporters of the main party were allowed to enter without being properly identified as party agents (as stated by a returning officer at the IREC hearings).

* High turnout in polling stations in areas dominated by one party is extremely suspicious and in the eyes of IREC is in itself a clear indication of likely fraud, most probably conducted through ballot stuffing, utilising local knowledge of who on the poorly kept voter register is absent, deceased or for another reason unlikely to appear to vote. It is

unconvincing to ascribe high turnout to the quality of voter education and motivation in this election or to the gratifying results of the voter registration drives in 2007, even in cases where there were organised efforts to get out the entire vote. IREC generally holds a 100% turnout where the voter register is of poor quality to be a clear sign of fraudulent presiding officer behaviour. Even in countries with compulsory voting, a 100% turnout is never achieved! The identification of a particular level of voter turnout to distinguish “suspicious” from “non-suspicious” turnout levels will always be arbitrary, and IREC does not find it useful for its purposes to engage in such an exercise.

Constituency level

* It was claimed in some cases that returning officers did not include all the 16A forms when the constituency tallying form (17A) was being completed. This element of the entire electoral process should have been conducted transparently, but IREC has found cases where a few forms 16A are missing and/or a few rows in form 17A have not been completed. Whether that was intentional or not and why party agents did not react to figures which were lower than their own tallies has not been investigated. It should be remembered that the omission from the constituency totals of the results from one or more polling stations will hurt an area’s dominant party more than its weaker rivals, so why should this be in the dominant party’s interest? In a constituency where party predominance varies from polling station to polling station, this reasoning obviously cannot be used. Nevertheless, some members of IREC were of the view that a conclusion such as this should be taken with caution because in areas where dominance exists but is not overwhelming, especially if support also follows locales, omission of particular polling station results could yield a significant advantage over a rival.

* Inflating the number of votes when completing form 17A. This has been claimed but not substantiated, and IREC has found no such evidence in the sample of constituencies scrutinised. Some members of IREC were of the view that this conclusion should also be taken with care, given the limitations of its chosen methodology, which did not include a review of the veracity of Form 16As. Additionally, where some ROs gave two Form 16s, and could not successfully explain the differences between the two, even under oath.

* Proper scrutiny of polling station results (forms 16A) at the constituency tallying centre might not always be relied on in areas where one party is dominant, e.g. questions should have been asked or investigations carried out in cases where polling stations had 100% turnout.

o Under no circumstances were returning officers allowed to announce partial results from the constituency tallying processes since that could slow tallying down and create confusion and misunderstanding. For various reasons, including calls from the national tallying centre requesting results, some returning officers nevertheless did announce partial results, which unavoidably – because this was not anticipated – caused considerable, and tragic, confusion and misunderstanding. Some of these cases – Molo, Juja, Kieni, Limuru, Lari, among others – received considerable attention, and it is clear that the returning officers in question themselves contributed to the misunderstanding and

o confusion around such results and their announcement. IREC has not been able to discover information demonstrating that the official presidential election results from these constituencies are not correct. It is, however, equally clear that the handling of these cases – in the constituencies as well as at the national tallying centre at KICC – contributed to the confusion, misunderstanding and eventual denial of the correctness of the outcome of the presidential election. It is difficult not to attribute this to a combination (not necessarily the same in each of these cases) of problems emanating from recruitment of temporary staff (including returning officers), their inadequate training, conflicting information from ECK headquarters and, finally, inadequate preparation for results handling and announcement at KICC (including insufficient briefing of temporary personnel and party agents and observers). One can also ask whether the system as such was not primarily to at fault (that is, those who designed the system, or allowed it to be used). The system provided for the manual completion of reams of badly designed sheets of paper with thousands of entries to be made and tallied and very little technical assistance, if any.
* Some returning officers did transfer constituency results (presidential as well as parliamentary) to the national tallying centre before they were entirely certain that they had established their correct results. Therefore, in some cases, different results from those previously transmitted were brought to KICC by returning officers. This created confusion (since this had not really been foreseen) as well as a need for correction. In some cases, necessary changes were not allowed, resulting in the ECK insisting on results it knew were wrong! At the same time, attempts to correct erroneous results (no matter how well it had been established that the figures in question were not correct) also aroused serious suspicion, when altered/corrected data forms were spotted at a verification table or on their way to the IT department.

* There were allegations that delays in transferring constituency results from PNU strongholds to KICC were to allow for controlling how the score for President Kibaki stood compared with the score for his primary challenger, Mr Odinga. IREC was presented with various arguments from the two sides (PNU primarily explaining the delays by adverse weather conditions and the complexity of the constituency tallying [large constituencies, many candidates at all three levels], ODM primarily arguing that the apparent unreachability of many returning officers in these areas and the general delay was so suspicious that there could only be one explanation – that “somebody” was up to “something”). The ECK denied that there were any delays, saying the time taken in the circumstances was normal and explicable on the basis of a variety of known factors. Ultimately, IREC could establish neither the cause(s) of the delays nor that they formed part of a comprehensive plot, in which the components and collaborators remain unidentified. Nor could any numerical consequences be ascertained. In some cases, the explanation may have been that returning officers decided to deal with results from parliamentary and civic elections first, because that was what the many candidates in those elections – who happened to be present and vocal – obviously wanted. A section of the members were of the opinion that, taken in their totality, the foregoing explanations still leave many questions unanswered, and do not sufficiently explain what were then

perceived as delays in full. By way of illustration, they argued that the time indicated for receipt of the Kieni results, for example, differed materially from the RO’s testimony of when he received them, in addition to other inconsistencies such as his reasons for not speedily alerting ECK that he had sent the wrong results in the first instance, taken with the curious time-consuming effort at saving costs by travelling with five other ROs in the same vehicle to Nairobi.

National level

* Official ECK results in a number of constituencies showed considerable discrepancies between presidential and parliamentary election turnout. Since this phenomenon was observed primarily in Kibaki strongholds, it was seen by many – political parties, domestic and international observers, ordinary voters – as convincing evidence of inflation of the Kibaki vote by ECK personnel and commissioners at KICC. However, the documentation provided is unconvincing, and no cogent evidence has been produced by any of those who claimed that these discrepancies demonstrated ECK rigging in favour of the President and PNU. Furthermore, such discrepancies were also identified outside Kibaki strongholds. However, simple reason and solid evidence from all countries with simultaneous elections make such discrepancies – in some conspicuous cases of more than 15 percentage points – most unlikely. IREC therefore decided to conduct an in-depth analysis of a sample of such cases, the clarifying result of which is presented in the subsequent section. Though the sample is statistically valid, a number of the IREC members were of the view that the inferences drawn from it should be taken with care since, on their own, the statistics can only be relied on to illustrate errors in computation and transmission of results and no more.

* The verification exercise by the ten teams at KICC was conducted badly, if at all. The ECK chairman admitted in his evidence to IREC that the presidential results in the computer database for no fewer than 32 constituencies currently differ from what is recorded on those constituencies’ forms 16. This is 32 out of 210, i.e. 15%!

* Furthermore, the IT department did not provide an adequate checking procedure for the accuracy of the sum of valid votes; it is also not difficult to spot errors in the summation of presidential and/or parliamentary candidates’ votes. This has contributed to the use of incorrect results by political parties, CSOs and observer missions in their attempts at analysing the election, leading to a number of erroneous conclusions. The ECK IT manager must take considerable responsibility for providing at least some of these incorrect figures to users of the ECK’s official information.

* It was alleged that figures in forms 16A and 16 were being changed/corrected at KICC, which was in itself seen by some as proof of results-tampering. However, IREC has not been provided with any instance where this appears to be the case. The unmistakably corrected forms, primarily the handwritten data-capture forms used at the tallying centre when receiving telephone calls or faxes – the existence of which is not in doubt – reflect many other problems, such as later realisation by a returning officer (stated under oath) that he had made mistakes in tallying, or the eventually provided total results for

constituencies instead of previously provided partial results. These are convincing signs of the various inadequacies in the planning and management of the tallying and results transmission process, but that is something different from what has been the accusation made by – in particular – ODM, civil society, and domestic as well as some international observer missions. Whatever the conclusion on this particular issue, it is clear that ECK procedures and problems at KICC were never fully understood by most of those who aired an opinion, including vocal commentators. Evidently, these interlocutors would have been much more useful to IREC’s investigation if they did not have the material inaccuracies alluded to above in their assertions. Some members are nevertheless of the view that such inaccuracy should only be taken to mean ignorance of the procedures. Any aspects of their information that is unaffected by these material inaccuracies has been valuable in adding up to the final conclusions of IREC on specific issues, such as suspiciously high turnouts.

* Access to the national tallying centre by presidential agents and by observers was handled extremely unprofessionally. It therefore became an awkward issue, causing all sorts of suspicion and justified accusations of lack of transparency. But it appears primarily to have been a sign of unprofessional process-management and inept public relations, not circumstantial evidence of fraudulent ECK misconduct. A number of the commissioners also pointed out that, given the difficulty of proving fraud, the unanswered questions around who prevented the observers (in the first instance) and agents (throughout the process) from accessing the tallying centre until the night of the 29th of December 2007 still cast a shadow of doubt on the reasons for either implementing transparency measures that the parties had been notified by letter about nonchalantly or the possible reversal of that transparency requirement.

* The strange circumstances surrounding the final announcement of the result of the presidential election, the handing over of form 18 and the low-key swearing-in ceremony at State House all contributed to the flow of rumours of ECK malfeasance before, during and after election day. Mr Kivuitu, the ECK chairman, has himself stated that he was not too happy about the situation and there can be no doubt that the very last part of the electoral process contributed significantly to the eruption of post-election violence. But while Mr Kivuitu was not happy, he did not intervene and request the necessary time to investigate the various claims in sufficient detail so that a proper solution could be found. The overnight “audit” exercise carried out by some party stalwarts and a couple of national observers was not conducted in such a way that it put the issues to rest.

In the final argument, ODM persisted in contending that, inasmuch as there has been no adequate refutation of such a plot, given the alterations and inconsistencies in the results and the documentation, a finding of fraud is indicated. PNU and the ECK submit that there has been no evidence that any of the alterations and inconsistencies were intended fraudulently to benefit any candidate or in fact had such result. Therefore they contend for a finding that the explanation must be human error.

It was not possible to attain consensus as between the members of IREC on this issue. Nor was it necessary. There is indeed consensus in respect to item (e) of the Terms of Reference, relative to the integrity of results, especially in relation to the presidential election: The conduct of the 2007 elections in Kenya was so materially defective that it has been, and will remain, impossible for IREC to establish true and reliable results for the presidential and parliamentary elections.

Therefore, although there is room for honest disagreement as to whether there was rigging of the presidential results announced by the ECK on 30 December 2007, the answer is irrelevant.

6.5 Analysis of the 2007 counting and tallying process

IREC soon realised that it was necessary to study in considerable detail how polling station counting results were documented (form 16A) and how such results were transmitted to the constituency tallying centre, where they were recorded and tallied (on form 17A) before the constituency results were announced. Subsequently, the constituency presidential results were transmitted to the national tallying centre, first by telephone or fax for fast announcement in the national media, and then the original form 16 was brought to KICC by the returning officer in person, whereupon that constituency’s presidential results, until then provisional, became final.

IREC’s intention to analyse in considerable detail how results floated upwards from polling stations via constituency tallying centres to KICC was also due to (1) claims by various stakeholders that they had identified potential problems, which needed to be investigated using basic source data, and (2) IREC’s own observation that in a number of official ECK constituency results (both presidential and parliamentary) the sum of votes for all the various candidates did not tally with the number of valid votes, as it obviously should, no matter what.

At IREC’s request the ECK promptly supplied photocopies of the material requested, which for this purpose was primarily the files containing forms 16A and 17A from a number of sample constituencies which evidenced various types of problems and suspicions. The ECK also made it clear that IREC, if it so wished, might have full access to the original files, but that was not considered necessary (apart from a few very specific checks). The reasons for that decision are presented below.

IREC requested the relevant material from nineteen constituencies in several tranches for these analyses, described in more detail in annex 6.A. The analytical approach was extremely simple, as an Excel spreadsheet was constructed for each of the two different elections in each of the nineteen constituencies (in one case, only the presidential election was included). The entire content of the forms16A was then transferred to the relevant spreadsheet, with candidates in the columns and polling stations in the rows, which resulted in the spreadsheets for all practical purposes being a replication of what the forms 17A for those constituencies should have looked like.

A comparison between the forms 17A and the IREC spreadsheets then allowed a closer scrutiny of discrepancies and the drawing of conclusions. In this latter part of the analysis, ECK in-house data transmission forms, forms 16 and other ECK material were also used.

The main result of this analysis is that the transfer of data from forms 16A to form 17A – especially when one considers the simplicity of the exercise – in many cases suffers from a very low level of precision and reliability. The tallying in forms 17A is also erroneous in many instances (“many” here meaning in relation to what one would expect, once again considering the simplicity of the arithmetic involved).

In some instances it was also discovered that results announced in the constituencies, which were more or less identical with the results of IREC’s exercise, had not been accepted for announcement at the national tallying centre at KICC, while the less accurate aggregated result had been.

Here are the three main observations:

(1)
This kind of audit is cumbersome, especially because some forms 16A apparently never made it to the constituency tallying centre (which can explain why they are not included in the forms 17A), while for unknown reasons others are not currently available in the constituency files. Though some of the members placed a premium on the story told by statistics, a number of others felt that (time and other resources permitting) a deeper investigation of some of the reasons why statutory forms were missing, for example, would have shed more light on the issues.

(2)
Almost all parliamentary and presidential election results for the constituencies sampled are erroneous, which means that very few of the officially published figures are actually accurate. In one constituency (090 Kirinyaga Central) IREC even discovered that the parliamentary candidate with most votes (when properly transferred and aggregated) had not been declared the winner, as that honour was enjoyed by the candidate with the second highest number of votes (and the returning officer from the constituency in question has in his evidence accepted the correctness of IREC’s analysis). Some IREC commissioners were of the view that this was one of the cases where a scrutiny and recount would have reinforced the statistical finding, especially given the narrow margin. This was, however, not done by IREC.

(3)
In all ten constituencies displaying large discrepancies between presidential and parliamentary election turnout which were selected for analysis, the discrepancies are reduced to the low level one would expect on the basis of simple reason or from comparison with other countries conducting simultaneous elections. The table below demonstrates how the turnout discrepancy apparent from the official ECK election results virtually disappears when the far more reliable IREC data are used.

There is no reason to doubt that the picture is the same in the constituencies not included in this analysis. The overall conclusion is, therefore, that conduct of the results transfer from polling stations to constituencies, the tallying in constituencies, the transfer of constituency-level presidential election results and the tallying at national level is – generally speaking – of incredibly low quality: it is actually not acceptable.
Constituency Presidential / parliamentary turnout as % of total registered voters (based on ECK data) Presidential / parliamentary turnout as % of total registered voters (based on IREC data)
090 Kirinyaga Central 15.9 0.7
017 Kaloleni 14.7 0.9
144 Kajiado North 12.8 0.1
104 Limuru 11.5 0.6
203 Bomachoge 10.3 1.1
Average: 13.1 Average: 0.7
138 Molo 3.2 NA**
100 Juja 3.2 0.3
095 Maragwa 1.7 1.3
182 Bondo 1.0 -2.9
051 North Imenti 0.0 1.2
043 Saku 0.0 4.0
034 Wajir North -0.2 0.0
082 Kieni -1.0 0.3
105 Lari -1.6 1.5
Average: |1.3| Average: |1.4|
155 Malava -11.0 -0.2
009 Changamwe -11.3 NA*
070 Machakos Town -15.0 2.5
052 Central Imenti -19.9 -0.8
066 Masinga -22.4 -2.3
Average: -15.9 Average: |1.5|

*Forms 16A for the parliamentary election for Changamwe have not been re-entered, so this constituency is not available for this analysis.

** Form 17A for the parliamentary election in Molo has not yet been identified so this analysis cannot be performed for this constituency.

This scrutiny of the handling of results-transfer and tallying has not indicated any particular or discernible party bias in the demonstration of incompetence by constituency tallying centre staff, by national tallying centre contract staff at KICC, or by ECK permanent staff and commissioners.

6.5.1 The random nature of the errors affecting the presidential election in the eighteen constituencies analysed

The analysis of the results in the eighteen selected constituencies clearly indicates the poor quality of the tallying process conducted by the ECK, and the ubiquity and magnitude of the errors in the translation of results from polling stations to form 17A, in the addition in that form, and even in the relatively simple task of transcribing the results calculated on the forms 17A to the statutory forms 16.

It is not within IREC’s mandate to conduct a new tally of the vote for all the 27,555 polling stations. This would not only be beyond our capacity, but the results of the exercise could be doubted on account of the unreliability of the basic data. Even though the sample used was designed with other purposes in mind, it might, however, be of interest to conduct a brief analysis of the impact of the errors on the numbers of votes obtained by the main candidates. This has been done in the table below.
CANDIDATE MWAI KIBAKI RAILA ODINGA KALONZO MUSYOKA
CONSTITUENCY VOTES ACCORDING TO DIFF VOTES ACCORDING TO DIFF VOTES ACCORDING TO DIFF
IREC ECK IREC ECK IREC ECK
KALOLENI 16,816 16,879 63 21,748 21,716 -32 1537 1,535 -2
BONDO 156 148 -8 56,022 53,202 -2,820 9 7 -2
MALAVA 18,152 14,712 -3,440 26,355 19,891 -6,464 346 340 -6
CHANGAMWE 15,151 9,366 -5,785 29,648 17,706 -11,942 8,368 5,347 -3,021
MASINGA 2,341 1,038 -1,303 387 238 -149 33,659 21,225 -12,434
MACHAKOS TOWN 4,954 3,187 -1,767 1,515 812 -703 55,785 41,295 -14,490
KIRINYAGA CENTRAL 55,097 52,866 -2,231 601 580 -21 59 53 -6
LIMURU 48,302 48,389 87 2,988 2,934 -54 443 436 -7
JUJA 101,003 100,390 -613 13,948 13,752 -196 4,986 4,409 -577
MOLO 75,314 75,261 -53 22,510 23,268 758 263 222 -41
MARAGWA 56,429 56,439 10 471 460 -11 577 564 -13
NORTH IMENTI 90,572 78,684 -11,888 3,677 3,370 -307 862 802 -60
KIENI 72,354 72,054 -300 612 580 -32 62 58 -4
LARI 49,280 49,276 -4 458 457 -1 70 32 -38
CENTRAL IMENTI 58,811 43,410 -15,401 1,585 1,108 -477 185 130 -55
KAJIANDO NORTH 49,055 49,038 -17 28,921 25,330 -3,591 3,260 3,664 404
BAMACHOGE 26,176 26,967 791 20,043 18,070 -1,973 224 177 -47
WAJIR NORTH 5,084 52,44 160 5,166 4,988 -178 30 29 -1
TOTAL 745,047 703,348 -41,699 236,655 208,462 -28,193 110,725 80,325 -30,400

What the table clearly shows is the random nature of the errors in these constituencies: no candidate can be said to have benefited. Table 1 shows that all three main candidates lost a significant number of votes on account of the errors, and those errors that negatively affect the candidates most were even concentrated in their strongholds. Kibaki lost 15,401 votes in Central Imenti and 11,888 in North Imenti, two of his main strongholds. Raila Odinga lost 11,942 in Changamwe and 6,464 in Malava, two of the constituencies where his support was stronger. Kalonzo Musyoka lost 14,490 votes in Machakos Town and 12,434 in Masinga Constituency, two of his main strongholds. Some of the commissioners did point out that, in sum, these random “mistakes” did result in significant gains or losses by certain candidates. This is a conclusion that they admit means little working with a small sample but insist would possibly have significant effects if all the results were analysed in similar fashion and with greater rigour, which IREC did not have the wherewithal to do.

6.6 Assessments by observer groups of the counting, tallying and result announcement processes

The various domestic and international observer groups experienced problems in general with understanding the counting, tallying and transmission processes and the possibilities for them not being conducted strictly according to the various sets of instructions and guidelines issued. This might have led to assessments based on insufficient understanding of these processes, their regulatory foundation and the actual impact of problematic recruitment procedures, unclear rules and insufficient training.

The main consequence of this incomplete comprehension of the entire counting, tallying, reporting and announcement process has been a tendency to deliver verdicts on the process which do not give a fully reliable picture of what actually happened after the original counting of votes at the polling station level.

Three problems in particular have been difficult for observers and observer missions to comprehend:

o The announcement of partial presidential results in some constituencies. Such announcement was not foreseen in the general regulations, in the guidelines for returning officers or in their training, but nevertheless it happened. The returning officers in question were actually announcing might well have been more or less difficult to

o understand, but that does not necessarily make the subsequent announcement at the national tallying centre at KICC of the final result (that is, including all polling stations in the constituency, which by necessity must have produced more votes for at least some of the candidates) a proof of rigging or fraudulent behaviour as some observer groups (domestic as well as international) have claimed in their reports. Some political parties and individual commentators have also been keen to see this as an indication of ECK malfeasance, generally accusing some unnamed person or persons at the ECK of allotting a substantial number of votes to one of the presidential candidates. IREC has not been able to substantiate such accusations. Counsel acting for ODM did not suggest to any of the ECK witnesses, from the chairman to the returning officers, that they were party to or witnessed any such malfeasance. Some of the commissioners are of a materially different view. In their opinion, the fact that partial results were expressly forbidden taken together with the lack of data on, for example which 111 polling stations the Juja RO referred to in his exculpatory letter, still leave reasonable doubt as to whether the results were not only partial as alleged, but also the percentage of the overall total from those polling stations at that stage.
* The very noticeable discrepancy between turnout in presidential and parliamentary elections was interpreted by observer groups as well as one major political party in particular as clear evidence of rigging in favour of the incumbent president. However, IREC’s analysis of counting and tallying in ten constituencies with huge turnout discrepancies demonstrates convincingly that the discrepancies are probably due to human error and general incompetence, difficult working conditions at constituency tallying centres, pressure from KICC-based ECK staff on returning officers to provide fast results, pressure from candidates and incumbents eager to know their own electoral fate, lack of training and unclear messages as to when and how erroneous constituency results might be corrected. The direction of the changes also indicates that other explanations which have been suggested are less likely to be true. The conclusions here are based on a study of forms 16A and 17A in presidential as well as parliamentary elections in a sample of constituencies. They fully explain the discrepancies, which disappear completely. What these constituencies’ ballot boxes may contain is not relevant for this analysis. The dissenting view of some of the commissioners is that, while some of the discrepancies can be attributed to human error, the available evidence cannot be taken to have fully explained the discrepancies (given lingering doubts about Form 16As) both in the sample as well as the universe of constituencies overall.

* Poor understanding of the procedures at KICC (even among staff members, temporary as well as permanent) unavoidably contributed to confusion and misunderstanding of procedural issues. These concerns, such as what the proper procedures were for the correct transmission of constituency-level election results, if and where party agent signatures were required, the requirements for valid data entry (and – in particular – correction to results already entered), the inadequately designed data entry forms with no space for corrections and for proper authentication of such corrections and the time of such corrections – and inadequate and misunderstood instructions on how to correct already announced results.

These problems and the lack of proper understanding of the process reduce the usefulness of the observer reports The problems at KICC and in the transmission of election results from constituency tallying centres to the national tallying centre must, however, be seen within the context of the other key factors in the entire electoral process: only this will allow a reliable and comprehensive assessment of the quality of the entire electoral process. Also see section 4.4.

6.7 Advice received by the ECK in relation to the counting, tallying and result announcement process

Concerns about the counting, tallying, transmission and announcement of results are not new in Kenya. In 1992, on the occasion of the first multiparty elections, an International Republican Institute (IRI) pre-election report noted that:

“... the electoral law does not stipulate the mechanism for transmittal of constituency results to the ECK in Nairobi [and] urges that this information be transmitted in … a timely way” (p. 23).

The Report of the Commonwealth Observer Group was more drastic in its evaluation:

“given … the poor communication between the ECK and the returning officers and between returning officers and presiding officers, the lack of coordination and inconsistencies in dealing with clear-cut problems, we can only conclude that neither the polling day arrangements nor … the counting processes were adequately designed or carried out to meet the specific situations and needs which the Kenyan electoral environment required” (p. 38).

The situation had not much improved by the 1997 general elections. A joint report of the Institute for Education in Democracy (IED), Catholic Justice and Peace Commission and the National Council of Churches in Kenya recommended that:

“... it is vital to have a speedy counting exercise, with results verified by all parties and announced immediately after the count is completed. This is the only way in which public confidence in the result can be ensured. Unfortunately, this was not the case in 1997” (p. 82).

What happened then in 2002? Charter Hall in downtown Nairobi was to serve as a centralised ECK results centre but this did not eventuate, since events overtook the results process. The media tracked results as they were tabulated at the constituency tally centres and reported them to the public. By the afternoon of 28 December 2002 it was clear that Mwai Kibaki was heading for victory and the KANU presidential candidate, Uhuru Kenyatta, conceded defeat in the early afternoon the next day, publicly taking up the role of leader of the official opposition.

However, when the ECK called a press conference that afternoon, 29 December 2002, to announce that it did not yet have sufficient official results, the commission was pressured by the large crowd of NaRC supporters to declare Mr Kibaki president. Apparently glass was smashed, flower-beds trashed and the ECK chairman punched in the back. Bowing to public pressure, several hours later the ECK announced – on the basis of unofficial results – that Mr Kibaki was indeed the winner and president-elect.

Official results were announced only on 3 January 2003. At that point:

“... [the ECK chairman stated that t]here was a great deal of pressure from NaRC leaders and their supporters that the declaration of the results be made nevertheless. Between the demands of the law and those of the people, the ECK chose to obey the latter. It declared Hon Mwai Kibaki the new president. It was a political rather than a legal decision. … ECK was convinced that just like in 1997, there was a possibility of insecurity, and further delay in the announcement of results would have resulted in chaos (only 10 forms 17 had been received when the ECK chairman announced the results)” (ECK Report on the 2002 elections, p. 86).

An external consultant, Michael Yard, evaluated the 2002 elections, noting that there were many problems that did not receive public attention because of the huge margin of Mr Kibaki’s victory. He pointed to several deficiencies and made a number of recommendations, among which the following:

* lack of a communication protocol, weakness and slowness of communication system

* lack of quality controls

* need for the ECK to provide a steady stream of information to the press

* need for the development of an election database that would allow for communications, logistical planning, material delivery tracking, production of master ballots, production of vote count forms for polling stations and constituencies

When the ECK itself evaluated the 2005 referendum:

“It was noted that initial results to the public were first released through other sources, e.g. the media. ECK was not fast enough to release the results. The reason for this was noted as poor network as some polling stations were not networked” (ECK Evaluation of the 2005 Referendum, p. 30).

The recommendations made to solve the problem:

“The law should be reviewed to allow preliminary results by SMS, email, fax, etc. ECK to install ICT equipments to hasten the process” (ECK Evaluation of the 2005 Referendum, p. 30).

Nevertheless, the delivery of provisional results to the ECK command centre was seen as very effective owing to the provision of mobile and satellite phones. It was recommended that this be extended to presiding officers to ensure smoother and faster delivery of results to the returning officers and subsequent delivery to the ECK HQ (ECK Evaluation of the 2005 Referendum, p. 32). It is also worth quoting again from the ECK’s Report on the National Referendum Evaluation Workshop, because some of the recommendations are closely related to the problems experienced in 2007:

“To speed up the tallying process, the Commission should consider procuring computers for this purpose. Those to do the tallying must be computer literate … There should be a mechanism for verification during counting and tallying of votes … [A]dvance planning for communication protocol to be utilized in the 2007 general elections [should] commence immediately” (p. 26). “The plenary stressed that the plans to decentralize the computerization process … should be aggressively pursued. In addition, the plan to procure/hire 210 computers for the constituency level should commence immediately in readiness for the 2007 general elections. … It was suggested that there is need to brief the commissioners on ICT activities regularly to enable them [to] understand the technological advancements and make informed decisions” (p. 36).

In early 2007, the ECK received the advice of two competent external advisors: Michael Yard, through the International Foundation for Electoral Systems (IFES) project; and Paul Anderson, through the United Nations Development Fund (UNDP) basket funding programme. In March 2007 IFES offered to design a computer program for results at the national and constituency levels and proposed methods for improving the transmission of results from polling station to constituency. It was assumed that results would be tallied at constituency level using UNDP-provided laptops and a WAN (wide area network) to be set up by the ECK. However, as was pointed out in Yard’s report of August 2007:

“... there is still no plan for timely transmission of results from Polling Station to Constituency, with some ECK staff still favoring physical transfer of result forms as the only viable method. It is strongly recommended that ECK make plans for use of more efficient transmission methods for provisional results, including mobile phones (voice) and SMS messaging. Without such a plan it is easy to envision a repeat of a 2002 scenario in which the political parties, the media, and even the public know the results of the elections (before the ECK announces them). In 2002 this led to widespread threats of violence and even some cases of violence with growing mistrust of ECK.”

In the case of Anderson and the UNDP, the effort was concentrated on providing a Media Elections Result Centre (MERC), building on the 2005 experience. It seems that most of the advice was not accepted and/or was poorly implemented. On the tallying and transmission of results, the commission plenary endorsed a simplified plan proposed by Ayub Imbira, the ECK IT manager. The ECK failures in the implementation of Mr. Imbira’s plan have been presented in some detail in section 3.8 and need not be repeated here.

Both consultants had suggested the use of progressive (partial) result announcement on a systematic basis, assuming that adequate methods for transmitting the information would be adopted. This is the procedure used in almost all countries to fill the information void in the time lapse between the closing of the poll and the availability of final results. According to the proposals, constituencies could have begun transmitting progressive (partial) results when they had received results from at least ten polling stations and suspended the transmission once 90 per cent of the polling stations’ results had been received at KICC. This would have meant that the MERC could have begun issuing progressive (or partial) results during the morning of 28 December 2007, with results arriving from all constituencies.

However, the advice was rejected, the laptops were not used for tallying at constituency level, no WAN was set up and the ECK opted for the method that had been used – and proved slow – in 2002. The distinction the ECK introduced between “provisional” and final results was legally sound but conceptually fuzzy. If returning officers had proceeded as expected, they would have completed their forms 17A and 16 at the constituency. Then they would have faxed or telephoned the results as entered on form 16 to KICC, which then would have proceeded to announce these results, terming them “provisional” because legally they were not valid. If everything had gone according to plan, the provisional results would have been exactly the same as the final – after all, the only thing the returning officers had to do after telephoning or faxing was to deliver the physical form 16 to the KICC. In most cases, the time saved was only a few hours.

The plan had, however, several conceptual and practical flaws. First, the announcement of “provisional” results had to wait until form 17A was completed at the constituency level and form 16 derived from it, and this took several additional hours. This, even in the smallest of constituencies, would add several hours of delay before form 16 could be completed and in the meanwhile the media would be announcing results collected at the polling stations. Second, if for some reason just one single polling station was delayed, this would delay the sending of “provisional” results until the problem causing the delay had been solved. Third, the provisional results would arrive from small constituencies first and only later from bigger ones, which would create some bias in the sequence of results, as argued in Professor David Throup’s analysis (even though this argument is not unchallenged). Last, and most importantly, if the returning officers did not perform well, because they were either inadequately recruited or poorly trained – or for any other reason – any difference between “provisional” and “final” results might be interpreted as manipulation, as actually happened in a number of cases. What happened between 27 and 30 December 27 is largely the consequence of the ignoring of adequate advice and reversion to the inadequate 2002 procedures. As to the MERC, and according to Margie Cook of the UNDP:

“… we got to the KICC to find that the ECK had not arranged for internet access even though this was one of their responsibilities and I went beyond the bounds of my responsibilities as a UNDP project manager and promptly authorized the expenditure of some hundred dollars to buy internet access for the ECK and have it installed immediately, otherwise the journalists would 135

not have been able to access the website at all. But we also found that the ECK failed to facilitate the links from the ECK results database to the graphic interface at any stage. Therefore no results were displayed on the screens in the KICC and the equipment was not used for the purpose for which it was brought” (Hansard record of IREC’s interview with Margie Cook).

Many of the problems were due to the ECK conviction that:

“Margie, this election will not be a competitive election ... the public (will) not believe the press and Kenyans all know that the only authority with results is the ECK” (Hansard record of IREC’s interview with Margie Cook).

The ECK was wrong on both counts.

6.8 Statistical evaluation of results

For two reasons 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 first reason is that the IREC analysis of tallying etc. in nineteen sample constituencies demonstrates convincingly that there are so many more or less erroneous constituency results (refer annex 6.A as well as section 6.5 above) that one cannot rely on any figures from the ECK. It should also be understood that the IREC figures – though very much better – are also not all completely reliable in cases where there were noticeable problems with the availability of all relevant data sources (forms 16A, supplementary information from 17A, etc.).

The second reason one should not conduct any statistical analysis of results from these elections is that the official ECK election results (published on the website and elsewhere) have not been cleaned of mistakes of a purely arithmetical nature, for example, checked to establish whether the sum of votes for candidates equals the number of valid votes cast in a constituency. The consequence of this separate blunder is that the turnout percentages provided by the ECK are not necessarily correct.

For these two reasons, IREC decided that it was not worthwhile to conduct any more profound statistical analysis of the available turnout data etc. The results would be almost useless for IREC’s purposes as they could not form a basis for useful conclusions.

Furthermore, the 2002 general elections also displayed discrepancies in a number of constituencies between presidential and parliamentary turnouts. It cannot be ruled out that some of the same problems would have been found in 2002 as in 2007 if a proper analysis of the transfer of data from polling stations to constituencies and subsequent tallying at the constituency level had been conducted while the original source material was still available.

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.

This observation is valid not only in relation to the 2007 elections but probably also in relation to the 2002 elections – in any case nothing could be done without a repetition of the exercise of reproducing form 17A for all constituencies to be included in such analysis.

An example: The very high turnout in some constituencies – e.g. more than 90% in the presidential contest in five constituencies – has aroused suspicion (and for good reason) of ballot stuffing, in particular because of the miserable state of the voter register in Kenya. However, other constituencies, such as Bondo, cross the 90% line if all votes are entered and added correctly, so the problem is not restricted to the constituencies identified as high-turnout constituencies on the basis of current official ECK results.

IREC has felt it necessary to abstain in its conclusions from relying on the results of the two exit polls which were conducted on election day at sampled polling stations. Exit polls should always be used with the utmost care, if the purpose is to forecast accurately the result of an election. A whole array of methodological concerns and sampling issues must be considered, and even then exit polls are in general more useful as a means of assessing the impact of the various explanatory factors which individually and in their interaction produce the eventual election result.

IREC also had the opportunity to familiarise itself with Professor Throup’s analysis and interpretation of the December 2007 election. Interesting as it was, IREC does not find it possible to rely on that analysis, inter alia because it too relies on ECK data, which have not been properly scrutinised (see above). It is doubtful whether a data set of the quality necessary for a study of the December 2007 elections will ever be available. Professor Throup’s analysis did, however, influence IREC’s thinking and made it ask some very pertinent questions – as did the exit poll results. Some of the commissioners nevertheless doubted the value, overall, of either interlocutor.

6.9 The integrity of the counting, tallying and announcement of results

IREC’s analysis of nineteen sample constituencies – which we have no hesitation in using for drawing conclusions about the entire operation – has led it to the irrefutable conclusion that the ECK was not able to manage the counting, tallying and results announcement processes in such a way that it secured the integrity of the electoral process at either the presidential or the parliamentary level. IREC has only sporadically concerned itself with the civic elections, but we believe that the situation is not much different at that level. If one – be it a voter, a candidate, a media representative, a party leader, an election observer – cannot trust the accuracy of the election results published by an EMB, then nothing is left and the political system loses credibility as well as legitimacy.

It is not IREC’s mandate to attempt to establish who won the presidential election, and it would in any case be extremely difficult, even if all ballot boxes were to be opened for that purpose. It would, inter alia, also require one to ascertain if some of the ballot papers did not belong there and also if some of them had been marked contrary to assisted voters’ intentions.

During its analysis conducted for another purpose, IREC established that a parliamentary candidate did not obtain the seat in the National Assembly that he had won by securing more votes than any other candidate in his constituency. If that can happen – and if the ECK can also get away with allocating a Nominated Seat to a party without using a recognised and documentable seat allocation method – then the conclusion can only be that the election management system as it functioned in the 2007 elections is unacceptable. It did not live up to the basic international standards of transparent, free and fair elections which the ECK Chairman had forcefully stated was what the ECK aimed for. It may also be worth noting that the ECK itself has stated that they “worked harder” in 2007 than in previous elections.

Recommendations

* IREC recommends that the ECK integrate the various descriptions of the entire counting and tallying procedure into one document – and one document only – which will then be the principal description and must be adhered to. The need for such descriptive regulations does not depend on possible changes in the counting and tallying system.

* IREC recommends that without delay ECK start having developed an integrated and secure tallying and data transmission system, which will allow computerised data entry and tallying at constituencies, secure simultaneous transmission (of individual polling station level data too) to the national tallying centre, and the integration of this results-handling system in a progressive election result announcement system.

* IREC recommends that the media must have full access to this new system, which will not be a problem if it is properly constructed. This will assist the media in obtaining fully reliable results at high speed from all over the country and will also place the ECK in the driver’s seat in relation to providing the media with fast and reliable data.

* IREC recommends that ample time be allowed for verifying provisional results, so that they are declared final/official only once there is no risk that errors may still be found or non-frivolous objections raised. Most countries allow one to two weeks for this – there must be sufficient time to check the provisional results, which are given status as final results only when all objections have been considered, all checks and rechecks conducted and the final verdict issued by the proper authorities. Given a clear explanation of what a provisional result is, there is no problem in making voters understand that election results are so important that they can be declared final only once they have been properly scrutinised and checked.