Whatever Happened to the Electronic Returns?

One still-unanswered question relating to the August 2017 Kenyan general elections - a frequent source of opposition concern and a pseudo-justification for NASA's fake results – is “whatever happened to the KIEMS electronic results that the IEBC had in their possession before the official results were announced?”.

It is mostly well-understood that the Form 34As, Bs and C constituted the formal election results, and that the parallel electronic transmission and display system - which was a legal requirement as a control check and verification mechanism - did not constitute the official record of the election. But we know that the vast majority of the KIEMS devices worked (because we saw the portal being updated moment-by-moment) so there was a central polling station-by-polling station record somewhere of the electronic transmissions (represented by the red “Results” below). That data has never been released (as an excel, pdf or anything else), though the display portal had something derived from it in a different format for several months.

Slightly oversimplified, this summarises the process as I understand it and highlights two issues in red:

James Orengo and other NASA figures have been asking for this transmission data for months, claiming that it would show a different set of numbers to those in the portal. I believe that this is correct, though not for the reason they give. Even without assuming malpractice, the portal began to be fed purely with electronic transmissions, but as the election continued and paper forms started streaming in, IEBC officials updated the portal with the official paper results and filled in missing stations which had never submitted an electronic return, which is why the total kept changing for several days.

But there was a record, somewhere, polling station by station, of what came in from the KIEMS kits. Why might it be the case that the IEBC can’t or won’t release this data? It would of course be incomplete, as some stations never got their KIEMS Kits working, or they failed part-way though, while others would have results different from the forms for reasons such as tired clerks mistyping, but that is surely no obstacle to release?

One reason could be that the data set is entirely different, and the portal was indeed rigged, as NASA claim. But this was always implausible because the election results were declared based on paper forms, not the portal results. The paper forms were produced by the 290 constituency returning officers. There would be no point in rigging the portal if the forms were different, as it would have no effect on the actual results and it would be obvious that it had been done. We repeatedly checked the paper forms against the portal during the election and they always matched very closely.

[As an aside, the Supreme Court instructed that access to various data be offered the NASA petitioners on 28 August, but for reasons known only to the Commission, some of the court-ordered requirements were not met. Perhaps surprisingly, given NASA's later "open the servers" campaign, the SC did not require the IEBC to give access to the actual electronic results transmissions, only to the locations of each KIEMS kit, log-in records to the portal and access to scanned and physical forms].

So, why not release it? Another possible reason might be that it would show just how often the KIEMS systems failed and officials went fully manual, which might allow observers to triangulate specific polling stations and then look for patterns in those stations of unusually high votes for one candidate. Or perhaps that some results were materially amended on the paper forms but correct electronically (though on a sample basis they have not been found  by anyone in checks). But surely from the perspective of the IEBC's leadership's credibility, that is a reason to release it, not to hide it?

And there are not one but two missing data sets

The second data set is a record of the means by which voters voted in a particular polling station, containing in the electronic voter identification system (EVID) return for each polling station (marked in red as Statistics). That shows how many people were identified and how that process took place (but not of course how they voted).

In the October 2017 presidential rerun, which Odinga boycotted, the IEBC was willing to release a download from the electronic systems, which I got a chance to copy. This showed for each polling station by what means people had been identified before being allowed to vote. It looked like this:

Quite a bit of analysis was done on that data set in October-November 2017, looking for patterns as to why fingerprint voting was not universal and confirming exactly what Document and Alphanumeric search really meant. We were looking for evidence of mass voting by non-fingerprinted voters, which would suggest local officials allowed stuffing or topped up votes for their favoured candidate after polls closed. We concluded that the data set made sense in general, that there was prima facie evidence of malpractice in a very few constituencies, but no pattern of by-passing the EVID tool nationwide.

These were the numbers for the percentage of voters not identified using fingerprint authentication (by any one of the three possible methods) in October, by constituency nationwide:

As can be seen, the numbers are pretty steady and low outside Nyanza (where they were zero as no-one voted) and the old North-Eastern Province. A deep dive on North-Eastern concluded there was prima facia evidence of “top up voting” or other forms of stuffing in Eldas and Wajir West, while the results were inconclusive in Mandera North.[I have since discovered that the EU Observer Mission report found the same problem in Eldas. The EU also found several stations in Garissa where extra digits were added ton the forms, producing official results for Kenyatta several thousand votes larger than the electronic results]

Elsewhere in the country, though, where voting did occur despite the boycott, there was no pattern of large numbers bypassing the electronic systems – 97.2% of voters were eventually biometrically identified. In other words, the (very modest) pro-Kenyatta turnouts in October 2017 were plausible.

So, given it could be produced in October, why has IEBC (or OT-Morpho, the service provider) not released the same station by station analysis for August? This question was asked of an IEBC Commissioner and elicited the response: “We had requested for the data on the 8 August election but did not get it from Safran [OT Morpho].”  

But is this true, or is it just a cover story because release would raise more questions that it would answer? OT Morpho (which has gone through several ownership changes in the last few years) is itself under some suspicion, particularly as it has since been disbarred by the World Bank for procurement corruption in Bangladesh.

Would release of the data show large-scale bypass of fingerprinting in certain constituencies, which might suggest top-up stuffing for Uhuru and the Jubilee candidates? Though it will probably change little now, it is still important for Kenyans to know the answer, because of the corrosive effect such allegations have on the legitimacy of the incumbent government.

If there is a logical explanation for withholding the data, it needs to be given. If there is not, the truth needs to be told and the file shared.