The voter registration exercise starts and runs right through the month of November. As we speak, the long awaited Biometric Voter Registration (BVR) kits necessary for the exercise arrive later this month and voter education is underway. While these two features are important cogs in the election process, Kenyans should pay more attention to the actual number of voters registered as this influences the quality of the election. A sound voter registration exercise is the precursor to a successful election.
To get a better sense of the risks and challenges of the voter registration process for the 2013 election, Mzalendo has analysed the performance of voter registration at a constituency level for the 2010 referendum, looking at actual numbers of voters registered compared to the number of people eligible to vote extrapolated from the 2009 census data.
When compared to a national level of registration of some 61.5% with a broadly even gender split, the performance in a number of constituencies was surprising, and in a number cases worrying. These results fell into two broad categories:
1) More people were registered to vote in some constituencies than the 2009 Census suggests is technically possible
There are a number of constituencies where the number of people registered to vote was significantly HIGHER that the maximum number of people eligible to vote (as extrapolated from the census). Clearly such instances are deeply worrying. Significant results include:
Mvita – 65,769 voters were registered, extrapolating the 2009 census data gives a figure for the maximum number of eligible voters as being 49,515. In other words, 16,000 more people registered to vote than was technically possible. The picture is even more interesting when you look at the gender splits: 98% of eligible women and 169% of eligible men were said to have registered
Kisumu Town West – 90,069 voters were registered, extrapolating the 2009 census data gives a figure for the maximum number of eligible voters as being 79,387. That means more than 10,000 more people registered to vote than was technically possible. As with Mvita, there is an interesting difference with the gender splits 97% of eligible women and 129% of eligible men were said to have registered
2) There are significant differences in the registration rates of the different genders in the same constituency
Both of the greater than 100% registration rate examples cited above showed significant gender bias, with significantly more men than women registering. This trend was seen in a number of other constituencies, including:
Westlands – 91% of eligible men, 62% of eligible women registered
South Mugirango – 61% of eligible men, 45% of eligible women registered
Kamukunji – 87% of eligible men, 55% of eligible women registered
Bomachoge – 79% of eligible men, 56% of eligible women registered
And there were a few instances where the gender bias skewed the other way, though these tended to be in constituencies with very low registration levels, for instance:
Ijara – 38% of eligible women, 25% of eligible men registered.
An unsurprising, but still important further insight that came from this analysis was the very low levels of registration that took place in a number of constituencies, particularly those which are rural and arid. These included:
Mandera West (15% of eligible voters registered)
Mandera Central (18% of eligible voters registered)
Turkana North (20% of eligible voters registered)
Lagdera (19% of eligible voters registered)
As a result of this work Mzalendo has three questions for the IEBC:
1) What steps are you taking to ensure that there are no instance of inflated voter numbers?
2) How will you ensure that where possible the gender bias in registrations in some constituencies is reduced
3) How will you work to increase the levels of voter registration, particularly in the arid, rural constituencies?
A Note About the Data
In order to build a view on the number of eligible voters in each of the constituencies by year we took the following approach.
We took the 2009 census data broken down by each age banding (by year) split by gender and by constituency. This wasn’t easy, as we weren’t able to obtain a soft copy of the data. Instead we had to buy a hard-copy of the census report, scan the 200+ pages, run the output through OCR and then manually fix any lines of data that had not been read correctly.
We have uploaded a copy of this file should anyone else be interested in using it – please find it here:
Having got this constituency level data we then set about using it to establish the size of the potential registered voter population for each constituency.
We did this, very simply, by looking at the following population cohorts:
2009 Potential Voter Population: All 18+ in 2009 Census
2010 Potential Voter Population: All 17+ in 2009 Census
2012 Potential Voter Population: All 15+ in 2009 Census
A couple of caveats:
1) This does not factor for migration between constituencies
2) This does not factor for the death rate of the higher age cohorts over these years. The strong youth bias in population size reduces the impact of this and by taking the population however the age cut-offs for each year reflect the census data of 28th August. Voter registration will be taking place in November, so there will be a tranche of people turning 18 in September and October 2012 which will at least partially offset this mortality.
Whilst these mean that it is not possible to get a perfect view of the eligible voters in each constituency, it is sufficiently robust for some of the huge discrepancies identified to be certain.
In order to get the relative % of registered voters to actual eligible voters, these figures were then compared to the constituency level data on registered voters split by age, taken from the report: INTERIM INDEPENDENT ELECTORAL COMMISSION (IIEC) VOTERS’ REGISTRATION STATISTICS – GENDER AGE DISTRIBUTION 2010 BY CONSTITUENCY. You can see a copy here:
This has been an intensely manual process, and should you spot any errors in the files used please do let us know so that we can rectify.