The Independent Election Commission (IEC) has published the first turnout figures which are backed up by biometric voter data. After weeks of ever greater turnout figures being successively announced, this latest update has reversed the trend: the IEC is now saying only 1.7 million ‘BVVed votes’ were cast, a decrease in reported turnout of almost one million from its previously-announced figure of 2.7 million. It has also delayed the announcement of preliminary results. In this dispatch, Ali Yawar Adili and Jelena Bjelica (with input from Thomas Ruttig) analyse the IEC’s latest turnout figures and calls for more clarity, especially in regards to why the IEC earlier included ‘non-BVVed’ votes.
In a follow-up dispatch, we will look at what lies behind the turnout figures, at issues to do with the Biometric Voter Verification (BVV) devices, the debate about which votes should be counted, and the consequences of delaying the announcement of the preliminary results.
First figures of biometrically-verified votes
After publishing a series of preliminary turnout figures (AAN reporting here), the IEC has now provided the first figures of what it calls ‘biometrically-verified votes’ (ara-ye biometrik shoda) that were cast in the election. The announcement was made at a press conference on 14 October 2019 attended by all seven Afghan commissioners and one of the two non-voting, international members of the IEC, along with the commission’s chief electoral officer and his two deputies (1) (see a two-part video here and here). As usual, the figures for each of the 34 provinces were read out, and not posted on the IEC website or in any other permanently-available format (see previous AAN report here complaining about this unsatisfactory procedure).
Commissioner Rahima Zarifi presented the statistics. Using an ambiguous turn of phrase, she said that 1,737,078 “biometrically-verified” (biometrik shoda) from 22,588 polling stations had been transferred to the IEC server. We assume Zarifi was referring to votes although she did not actually specify this, and refer to ‘BVVed votes’ or ‘biometrically-verified votes’ in this dispatch. Zarifi also said the biometric data from more than 3,900 polling stations “remains unprocessed” and that the statistics “remain subject to change.” (2)
A closer look at the turnout figures
AAN has compiled the data from the last three IEC updates on turnout figures into a single table (Table 1). We decided not to include the first two updates as the IEC did not provide the number of polling stations when reading out the first update, and the data provided in the second update was incomplete (all previous updates can be found in this AAN report and in detail about the first two updates in this AAN report). At least two AAN team members have double checked the readouts from the press conferences by listening to the recordings separately and checking every entry twice.
That said, it seems that the sum of the number of polling stations and votes given for each of the 34 provinces do not correspond with the totals (both sets of figures were given at the press conference). After adding up the figures that were read out, AAN arrived at a total of 1,737,051 ‘BVVed votes’ (27 fewer than the total announced by the IEC) from 22,538 polling stations (50 fewer than announced). The differences are marginal, but they cast a doubt over the reliability of the IEC figures, in general.
Table 1: Data from IEC; table by AAN.
Significantly, as Table 1 shows, the latest IEC update has a much lower turnout figure from the last one – almost one million fewer votes, from just under 2,700,000 to just over 1,700,000. This raises the question as to why, when and by whom one million votes not backed up by biometric data were previously included in the turnout figures.
The latest update also shows a steep drop in the number of average votes cast per polling station, from 101, as announced in the fourth update, to 77 – ie a drop of 24 per cent. This is likely to fall further once duplicate and underage votes are removed by the server. If this trend of BVVed votes continues in the remaining 3,980 polling stations, the final figure of BVVed votes can not be expected to be higher than two million.
At the 14 October press conference, IEC commissioner Mawlawi Muhammad Abdullah had already clarified, however, that even when the data from the remaining 3,980 polling stations are included, the resulting figure “will again not be final.” There will still need to be more, in his words, electronic “cleansing” of the votes, for example disqualifying votes cast by people who are underage, or where there are duplicate photos, stickers or thumb prints, or where a photo of a tazkera rather than the voter herself has been taken, as well as any votes cast before 7 am or after 5 pm on election day, ie before or after the opening of the polling sites (which would indicate manipulation and should be disqualified). Commissioner Abdullah said it was not clear what the percentage of such votes would be, and the IEC would “wait for the outcome of the server and the technology and, God willing, it will be small.”
The commissioners also explained why the IEC had presented four sets of turnout data – apparently trying to explain the nature of the figures published earlier and clarify what data would finally count. They said the first figures announced had been received from polling centre managers, district coordinators and heads of provincial offices by telephone on election day. From these updates, IEC had drawn what it called its earlier “estimated” figures, which had been “subject to change.” (For the earlier IEC announcements, see AAN reporting here and here). The commissioners did not say which of the five sets of turnout figures had been telephoned in.
There have also been some problems with the reliability of figures garnered from result sheets. Some of these figures, said IEC commissioner Awrangzeb, might be invalid as the sheets could contain – intentional or unintentional – mistakes. AAN was shown a screenshot of an example of such a digital sheet which has been transferred from a BVV device to the IEC’s digital centre (see photo below). It shows the ‘number of voters processed’ as 41, and the number of ‘valid votes’ cast as 135. The first row of figures seen in the picture below (without the hashtag symbol, ie the number of voters processed and registered voters) were automatically generated by the BVV device, commissioner Abdullah told AAN using an in-built, anti-fraud feature. The figures seen in the picture with the hashtag symbol before them (such as # of received ballots; # of unused ballots; # of valid votes; # of spoiled ballots and # of invalid votes) have been manually entered.
Again, it is not clear which of the five turnout updates included figures from processed result sheets which IEC member Awrangzeb said could contain “mistakes” in them. (3) Moreover, it is not clear if the turnout figures presented on 14 October, which, according to the IEC are ‘biometrically-verified’ represent the number of ‘valid votes’ or the number of ‘processed voters’.
Screenshot of a digital sheet transferred from a BVV device to the IEC digital centre
The number of polling stations and BVV devices
The IEC has also given figures for the polling stations which opened and those that remained closed on election day. It said it had planned to open 29,586 stations, but 3,018 stations had not opened. In another strange turn of phrase, the commission said it had received figures “as though” the election had been held in 26,568 stations, ie not that the election ‘had been held’ in 26,568 stations. Doubt remains especially because the IEC has yet to provide a list of the open and closed stations, or indeed, any other detail about their locations.
There are also only inconsistent figures on the number of BVV devices used on election day. Tolonews quoted Commissioner Zarifi saying on 17 October that, so far, 186 BVV devices and 495 memory chips, (of presumably other devices), as well as 172 printers were missing. This represents 0.03 per cent of the open 26,568polling stations. She said, however, she thought “this will not be a major challenge on the way to getting election results.” (4)
More concerning is what IEC commissioner Mawlana Muhammad Abdullah told 1TV on 16 October, that from some provinces, more BVV devices had been brought to the IEC than were polling stations. He did not specify how many devices more and from which provinces or whether or not it indicated that the additional BVV devices had been used fraudulently.
The IEC chief electoral officer, Habib ul-Rahman Nang, also mentioned ‘quarantined’ biometric devices, tweeting on 16 October that his technical team had“uploaded [the] entire bio-metric data into the [IEC’s] server. As per the decision of the respected commission, my technical team is now uploading data from the ‘quarantined’ bio-metric devices which will be completed by tonight.” He gave no figures. Almost two weeks earlier, however, on 4 October, Tolonews had reported that “around 2,300 devices” (7.8 per cent) had been quarantined, quoting a source in the IEC. In its 19 October update, the IEC said 2,888 (or around 11 per cent of open polling stations) had biometric devices which had been quarantined. Reasons for quarantining BVV devices include technical defects (broken screensand other malfunctions) and irregularities, such as lacking the photo of the result sheets from election day. The IEC said quarantined devices were now being reviewed in the presence of candidate agents and observers and, if biometric data was found on them, this would be transferred to the server.
The IEC said during a 19 October press conference that it had, in total, assigned 34,958 biometric devices for election day, of which 34,767 had been returned to headquarters, but 199 were still missing and a review was underway. Again, these figures do not add up: 34,767 retrieved and 199 missing devices makes a total of 34,966, not 34,958. The announcement also fell short of specifying how many biometric devices were actually usedon election day. There was also no mention of whether or not there were other biometric devices whose data have not been transferred yet.
Conclusion: An uncertain period
Since the election day on 28 September, the IEC has been struggling to determine the level of turnout and only in hindsight, said the presented figures were subject to change. The latest fluctuation in the turnout figure, by almost one million, or one third of all votes, requires an explanation from the IEC. It needs to explain whether those one million votes were fraudulent or not and why they had included them in their original figures in the first place. It also needs to answer questions as to how many votes were cast without biometric data capturing due to a malfunction of the devices, and how many and which polling stations and centres were open on election day.
The new figures bring the turnout, which was already the lowest of any election yet – down still further. However, there is no legal minimum threshold for turnout in Afghanistan, so the election is technically legitimate with any figure, however low it is. Even so, the next president will be ruling on a weaker mandate. A winner, if one emerges from the first round, supported by possibly only around one million voters – among at least 15 million Afghans of voting age – will need to be aware that the vast majority of Afghans above the age of 18 could not vote, chose not to vote, or chose not to vote for him.
Edited by Thomas Ruttig and Kate Clark
(1) They were: IEC chair Hawa Alam Nuristan, deputy chair Sayyed Esmatullah Mal, IEC secretary Musafer Quqandi, Muhammad Hanif Daneshyar, Awrangzeb, Mawlana Muhammad Abdullah, Rahima Zarifi, internal commissioner Ivilina Aleksieva-Robinson, IEC chief electoral officer Habiburrahman Nang, his deputy for operations Muhammad Osman Frotan and deputy for strategic communications Ali Shah Mesbah. It is not clear why the second international commissioner Ahmed Issack Hassan was absent.
(2) AAN visited the IEC on 3 October 2019 and learned then that the two sets of data were being transferred to the IEC digital tally centre. First, the result sheets that have been transferred to Kabul in the ‘temper evidence bags’ are scanned by the data tally centre. Second, the memory cards from the biometric devices are uploaded by the digital centre. These memory cards contain the photo of the result sheet and the manually-entered results (transmission of this data began on election day from the provinces that had internet connections; see AAN’s reporting here). The memory cards also contain the biometric voter data, including photos, of them, their tazkeras, voter registration stickers, thumbprints, as well as personal details such as their names and fathers’ names. This data is being uploaded onto the IEC’s central server, which according to an official from the digital tally centre, is also linked to the digital tally centre.
(3) An IEC commissioner explained the criteria for deciding to audit a polling stations and ordering a recount. The first criteria, he said, is when the number of valid votes in the result sheet differs from the number of voters processed. Second is when the server software triggers duplicate votes. He also said that before the announcement of the preliminary results, several steps should happen – the recount of votes from the polling stations, which had been subjected to an audit, by using QR code based on “a clear mechanism”; the transfer of the recount forms from the provinces to the IEC headquarters and; the processing of the recount forms at the national tally centre.
After that, he said the IEC would “arrive at the announcement of the preliminary results.” This explanation seems to indicate that an extended process before this announcement can be expected, probably rather counting in weeks than days.
(4) The number has been changing. A day before, on 16 October, 1TV quoted IEC commissioner Awrangzeb as saying that more than 100 BVV devices in some of the provinces had been missing.
This article was last updated on 9 Mar 2020