Dr. Seema Shah

Numbers do not lie. One lawyer said as much as he stood before the Supreme Court last week. “You can’t argue with the arithmetic.” It is unclear whether or not he realised the import of his words, for a look at the numbers referred to (but never explained) in court is quite revealing.

As the bedrock of a free and fair election, the voter registry is a crucial part of any electoral process. A legally verified and finalized list of eligible voters is vital, not only because it provides a check on fraud but because it provides a necessary constant for societal self-reflection.

The constant the voter registry provides allows for an analysis of voter turnout, overall and geographically distinct voting patterns and trends over time. A sound and reliable voter registry can help a society understand its own political behaviors, track these trends over time and space, and plan for the future.

It is thus not hard to see why the voter register was at the heart of the Kenyan Supreme Court case. Indeed, simple mathematics can help us to determine the validity of the claims the Independent Electoral and Boundaries Commission (IEBC) made with regard to the voter register.

First, the IEBC’s provisional register, which was compiled after the close of the voter registration process on December 18, 2010, contained 14,337,399 voters.

In court last week, the Commission claimed that there is also a special register, made up of 36,236 Kenyans who were eligible to vote but whose biometrics could not be captured because of physical disabilities.

In addition, the IEBC explained that there were 12 trainees, who also were not initially registered. Together, the provisional register, the special register and the 12 trainees total 14,373,647 voters.

After the closure of the register on December 18, 2012, there was no legal way to add any more voters. Only a subtraction could have been countenanced to address instances of double registration or the identification of other ineligible Kenyans. Indeed, the IEBC claimed that it had deleted some 20,000 names from the provisional register, because those constituted cases of people who had registered more than once.

So, the provisional register less the 20,000 should equal 14,317,399. On top of this, the IEBC said that it had to add the special register and the 12 trainees, which should come to a total of 14,353,647.

Why, then, is the total number of registered voters as announced on March 9, 2013 by the IEBC actually 14,352,536? Actually, the figure announced by IEBC on March 9, 2013 and the 12 trainees together is equal to the principal register.

The “special” register does not seem to fit in at all. This final number, announced on March 9, is still 1,111 voters less than what the total should be, according to the IEBC’s own explanation.

What is more important, however, is that this explanation still does not clarify why ALL these voters were not gazetted in the principal register in February.

If it is true that the IEBC wrote down the names of everyone who registered in the “green book,” including those with and without biometrics, shouldn’t ALL the names have been included in the legally gazetted register?

Also, even if the IEBC can explain the 1,111 unaccounted for voters, what explains the fact that at the level of the polling station, the sum of the principal and special registers do not equal the green book number?

Out of the 22 contested stations, there was only one station in which the numbers added up. Second, what explains the political parties register, which was issued shortly before the election?

That list included 14,336,842 voters, which is completely out of the range of the other numbers. Why would the register become a moving target? Third, and perhaps more worrying than the plethora of registers, is the fact that even though additions and subtractions have impacted all areas in the country, there does appear to be a pattern with respect to some of the most striking changes.

Additions and subtractions were made to the register all over the country. Aggregating these changes by province shows that 68,836 voters were added to constituencies in Central Province and Rift Valley, while 14,122 subtractions were made in Nyanza and Coast provinces.

This pattern undermines the argument that the irregularities were clerical, random and scattered across the country. During the hearing of the petition challenging the validity of the election, one lawyer attempted to explain this by claiming that voter registration, as laid out in the Constitution, is a “continuous process.”

What he omitted from his explanation, however, is the fact that the Elections Act states that voter registration ceases for the 60-day decision 2013: Aggregating these changes by province shows that 68,836 voters were added to constitu- encies in Central Province and Rift Valley, while 14,122 subtractions were made in Nyanza and Coast provinces.

This pattern undermines the argument that the irregularities were clerical, random and scattered across the country. period before the election. Fourth, it is interesting to note that if the original numbers in the December register had been used, certain constituencies would have experienced a voter turnout in excess of 100 per cent.

For example, the voter turnout would have exceeded the registered voters in Pokot South (109%), Loima (106%), Samburu North (105%), Kajiado South (103%), and Sigor (102%). The revised numbers as per the February register make the turnout figures a bit more reasonable, with Pokot South at 93 per cent, Loima 80 per cent, Samburu North 88 per cent, Kajiado South 91 per cent and Sigor, 92 per cent.

They are still incredibly high when placed in the context of the history of turnout in these areas. All of this comes into much sharper focus when we look closely at certain key areas.

Take Turkana Central, where turnout was about 53 percent in 2002. In this election, the IEBC reported turnout in the constituency to have reached 74 percent, a jump of 21 percentage points.

Notably, this constituency also experienced 8,516 additions in voters between December and February, the most of any constituency in the country. When comparing Forms 34 from the following three polling stations in Turkana Central to the number of registered voters published in the principal register in February, there are some significant discrepancies.

In each of the following polling stations, the number of registered voters recorded on Form 34 is more than what the IEBC published. If the number published by the IEBC had been used, these polling stations would have shown voter turnout in excess of 100 percent.

This pattern can be seen in other areas as well. In a polling station in Tharaka, for instance, the original number of registered voters had been written over with another number on the Form 34.

The new number is larger, such that voter turnout does not exceed 100 percent. A look at the comments from the presiding officer on the form, however, is telling.

The presiding officer wrote, “The number of voters was higher than the registered voters. This was authorised by the RO* as he argued that they were registered at Constituency level” (*RO is Returning Officer).

A close look at the form shows a three-digit number beginning with 12. The last digit is not clear, but whatever it actually was, it clearly had to have been less than 166, the number written over the original figure.

This clearly leaves many questions unanswered. It’s now a waiting game, as the public anticipates the Supreme Court’s detailed judgment. We will have to see how the Court managed to resolve the existence of no less than six different registers and constantly changing voter totals – and then decide if we can also resolve it ourselves.

Dr Seema Shah was part o
f the Africa Centre for Open Governance observation team during the hearing of the presidential election petition.

After observing the Supreme Court’s scrutiny of election results from the polling stations and the constituencies, I was surprised that the report presented to the judges barely scratched the surface of what we found.

On the first day of the pre-trial conference, before the hearings even began, the Supreme Court judges ordered judiciary staff to conduct a scrutiny of Forms 34 and Forms 36.

The court also ordered a re-tallying of 22 contested polling stations, alleged to have had serious problems. The order was a welcome one, as it affirmed hope in the independence of the court and seemed to indicate the judges’ commitment to fully understanding the myriad problems alleged by petitioners.

Soon after the process began, however, it became clear that it was fraught with problems. Security during the scrutiny was severely lacking. Judiciary employees, as well as agents for the petitioners and respondents, were initially divided into eight stations.

Each group was tasked with scrutinising all submitted Forms 34, which consisted of manually entering all the numerical data from Form 34 into a spreadsheet. All data from the eight scrutiny stations would then be transferred onto one central computer using flash drives.

It is unclear when these flash drives were issued and what the protocols were to secure them overnight. There was also little security around the central computer, which was intermittently surrounded by one group of people or another.

After these issues were brought to the attention of the judicial staff, armed guards were brought in. While that helped secure the room overnight, it did little to secure the main computer receiving all the data.

Omitted from the report

Our observation notes covering just one day of scrutiny showed 64 missing forms 34 from 14 different constituencies. The report to the judges, on the other hand, showed that Forms 34 from only 10 constituencies could not be found.

Notably, our notes show several instances in which the number of votes cast exceeded the number of registered voters. Those were not the only omissions.

On many forms, the numbers did not add up. For instance, the number of votes cast, as recorded, was not always the sum of valid and rejected votes. There were also differences between the aggregate number of valid votes as written out in numerals and in words. Which result was announced, the one in words or the one in numerals?

We also noted multiple copies of the same form, some of which contained identical figures and others of which included non-identical figures. Some forms were missing results for certain candidates, including instances where all candidates were not listed, or were listed with no corresponding result.

Often, figures were missing from the documents, and the numbers were illegible or had been changed without an authorising counter-signature. How did the judges end up receiving a partial report of the scrutiny?

Flawed methodology

The methodology for scrutinising the Form 36 – the document used to collate results at the constituency level — was also flawed and failed to show important discrepancies.

Our analysis showed that in some cases, the numbers for a particular polling station, as recorded on Form 36, were different from what was recorded on the corresponding Form 34.

There was no way to identify the problematic polling stations without using a polling-station level scrutiny of Form 36. Moreover, the methodology failed capture problems like missing polling stations on Form 36.

Based on our observation, the judiciary review also failed to highlight important differences between Forms 34 and Forms 36. It did not show, for instance, that in Isiolo North, the total number of votes calculated for Uhuru Kenyatta from all Forms 34 was 17,675.

On Form 36, Kenyatta is reported to have won 18,489. Where did 814 extra votes come from? In Turkana North, the Form 34 total for Kenyatta was 3,567, but Form 36 showed Kenyatta to have won 3,507 votes, which is 60 less votes than what was on the primary document.

Such discrepancies can be found for almost all the candidates’ results.It is also worth noting that it was impossible to fully observe the scrutiny process, because each station simultaneously reviewed multiple constituencies.

This meant that observers had to somehow keep an eye on all the different constituencies at the same time. Since there were only 10 observers from each side, it is not hard to see how being able to keep up with all the forms was difficult.

Also, each station was equipped with a large screen, which was meant to enhance transparency by showing the data being entered by judicial staff. Since more than one constituency was being entered at each station, though, not all data entry was transmitted to the screen.

Turnout beyond 100 per cent

It is now clear that the judiciary staff never carried out a re-tallying of the 22 contested polling stations as ordered. Instead, they simply reviewed and entered the data from the contested stations’ Forms 34 and Forms 36 into its spreadsheet.

In this way, then, it was no different from the general scrutiny of the forms. Inexplicably, its report on these stations highlighted only five as problematic. This was surprising, given that a simple calculation using the recorded figures showed four important anomalies.

First, in 16 polling stations, voter turnout as calculated using Form 34 and the principal register exceeded 100 per cent. The largest recorded turnout in this category was 301 per cent.

Second, in 18 polling stations, voter turnout as calculated using Form 36 and the principal register exceeded 100 per cent. The largest recorded turnout in this category was 450 per cent.

Third, and even more striking, was that there were two polling stations with voter turnout in excess of 100 per cent when using the green book, which the IEBC argued was the actual, complete register. One polling station in this category showed a 238 per cent turnout.

Lastly, it is only in one polling station that the sum of registered voters in the principal register and the special register equaled the number recorded in the green book.

Since the Respondents explained that the principal and special registers (as well as 12 trainees) together totaled the green book, the observed discrepancies are highly problematic and clearly undermine the Respondents’ claim. It will be interesting to see how the Supreme Court judges explain this when they release their judgment in less than 10 days’ time.

Dr Seema Shah was an observer for the Africa Centre for Open Governance during the scrutiny of results documents ordered by the Supreme Court