Afghanistan Analysts Network – English

Political Landscape

Afghanistan’s 2019 Election (17): IEC tally process underway but figures remain murky

AAN Team 7 min

The IEC aims to announce the preliminary results of the presidential elections on 19 October 2019. It is however struggling with the logistical challenge of gathering and vetting even the most basic electoral data. There is also the question of how to deal with the aftermath of yet another instance of a patchily implemented, sophisticated anti-fraud system. While waiting for the first preliminary results, the AAN team takes another look at the evolving turnout figures and notes that the patterns remain remarkable and worthy of very close scrutiny.

The hangar where the Biometric Voter Verification (BVV) devices used at the 28 September presidential election are unpacked and their data transferred to the IEC data centre. Photo: Thomas Ruttig/AAN.

An internal struggle for transparency

The IEC has released four turnout updates since voting ended on 28 September 2019. Two of them were presented at an IEC press conference, one on 28 September on the evening of the election (for a recording of the press conference, see Tolonews’s 10 pm bulletin at the 48th minute) and the bulletin on 1 October. In both press conferences, detailed results were read out loud but not projected on a screen, provided in writing or posted on the commission’s website. Two other updates were released through the publication of tables with detailed data on the personal Facebook page of one of the commissioners, Mawlana Muhammad Abdullah, on respectively 29 September and 3 October.

Mawlana Abdullah told AAN that the two updates he shared on his private Facebook account should be considered official IEC figures, since he shared them in his official capacity. He, however, also said the other commissioners did not share his enthusiasm for making written records publicly available.

How the IEC data was gathered

AAN visited the IEC headquarters in Kabul on 3 October 2019 and spoke with commissioner Abdullah and several IEC employees about the tally process and the turnout updates. It seems that most figures given at the press conferences and shared by Mawlana Abdullah had been collected through phone calls with the provinces, in a relay up from the polling centre to the district-level and from the provincial managers to the IEC in Kabul. Mistakes with numbers were thus unavoidable. An example of how this ‘Chinese whisper’ method of collecting and presenting numerical data can produce errors, can be found in the third IEC update (see Table 2). On 1 October at the press conference read-out, the turnout in Jawzjan was given as 38,135 voters in 835 polling stations, despite the fact that the IEC had planned to open only 322. According to Mawlana Abdullah this kind of errors were numerous. “We had similar problem in other provinces, too,” he told AAN, adding that “our [IEC] information delivery and reporting was not up to standard.”

Not all figures had been collected by phone. AAN was also told that some figures had been collected via the online IEC reporting system between the provinces and the capital, and that when IEC reported the first update on 28 September the data was still uploading. (1)

Gathering and comparing the data

AAN has gathered the data of the four updates into a single table (see Table 1). To do this, the readouts from the press conferences have been double and triple checked by at least two AAN team members who listened to the recordings separately and have checked every entry twice.

Table 1: Four IEC turnout updates. Table by AAN

Table 1: Four IEC turnout updates. Table by AAN

Each update shows an increase in turnout, which in itself is not unexpected given that each time both more polling centres and more polling stations are included in the count. However, the high turnout numbers, compared to observations on the ground and coupled with the distribution of votes and turnout figures across the country, do raise pertinent questions.

Most numbers suggest a far higher turnout than observed by journalists and non-governmental organisations on election day – see for instance these AAN reports on the E-day observations from Kandahar, Takhar, Wardak and Balkh here; from Herat here, from Kabul here and from Zabul province here.  Most independent observer organisations also noted that turnout had been low. (2)

Table 2: The third and fourth IEC turnout updates by number of polling centres and polling stations, against IEC baseline figures. Table by AAN

Table 2: The third and fourth IEC turnout updates by number of polling centres and polling stations, against IEC baseline figures. Table by AAN

Particularly in several very insecure places turnout was reported to have been implausibly high. The provinces with the highest turnout in absolute numbers, after Kabul – in the fourth, and probably close to final, update, at least for now – included Nangrahar (255,052), Kandahar (193,741), Baghlan (187,340), Paktia (161,655), and Helmand (115,333) – next to Herat (129,993) and Daikundi (117,506).

The national average number of votes cast per polling station (PS) was 101 votes. The provinces that produced far more votes per polling stations, however, included Baghlan – a province with a heavy presence of the Taleban who had assaulted the provincial capital in early September 2019 (AAN reporting here) – with an amazing average of 183 voters per PS (187,340 votes in 1,023 PS) and Paktia with an average of 164 votes per PS (161,655 in 983 PS). Both had more votes per polling station than Bamyan (130 votes on average per PS, or 84,124 votes in 647 PS), while Baghlan had even more than Daikundi (178 votes per PS, or 117,506 votes in 658 PS). Daikundi and Bamyan tend to have high turnout figures, both in absolute and relative terms, due to the relative security, a high level of female participation and effective political mobilisation (although, as will be shown in a forthcoming report, even there turnout was lower than usual).

Kandahar and Helmand had an average of 123 and 124 votes per PS, respectively (Kandahar 193,741 votes in 1,567 PS; Helmand 115,333 in 937 PS); Kunar and Khost an average of 115 and 113 votes (Kunar 67,383 in 586 PS; Khost 94,485 in 834 PS); while Nangrahar matched the national average of around 101 votes per PS (255,052 votes in 2,529 PS). Kabul province, on the other hand, had an average of 96 votes cast per PS (459,473 votes in 4,762 PS). Baghlan, on 3 October, still had 45 polling centres that were reported as disconnected from the central IEC (see Annex 1).

These provinces that had a high average of votes per polling station, also showed an above average relative turnout, compared to the national average (which was 27.9 per cent) and compared to the relative turnout in 2018. See Table 3.

Table 3. Key figures for the 2019 presidential election, based on the IEC’s update of 3 October 2019 and turnout comparison with the 2018 Wolesi Jirga election. Table by AAN

Table 3. Key figures for the 2019 presidential election, based on the IEC’s update of 3 October 2019 and turnout comparison with the 2018 Wolesi Jirga election. Table by AAN

Conclusion: The need for close scrutiny

The turnout figures as reported by the IEC, based on data provided by the provinces, do not match the impressions of observers. Moreover, they show remarkably high turnouts, both in relative and absolute numbers, in provinces that are known to be insecure and where voting was expected to be curtailed.

The tally process may well end up showing up large numbers of results sheets that show major discrepancies between the BVV vetted voters and the number of votes that were cast. These differences could be much larger than can be logically explained by the handfuls of voters that were allowed to vote after they found their names not in the voter lists. (3) This work of tallying and vetting seems to be ongoing, as the IEC reported recently on its Facebook page, but so far, the IEC has only provided information on the number of the result sheets it has processed, not the number of votes.

The tally process is already proceeding slower than planned (the latest update on how many PS are in which stage of the process can be found here). This gives rise to speculations that the IEC will not be able to meet its goal of announcing the preliminary results on 19 October. Mawlana Abdullah said as much when talking to the media on 8 October, but he was contradicted on the same day by IEC chair Hawa Alam Nuristani in a “joint press conference” of the IEC and the Election Complaints Commission.

Two German engineers from Dermalog, the company contracted to design the BVV process, have arrived in Kabul to assist the IEC in getting the central server up to speed. Claims by an IEC member that there had been cyber attacks on the server, were later denied, see here and here). Meanwhile the head of the IEC’s IT section has been introduced to the attorney general for questioning about the reports of registered voters missing on the voters lists on election day (see here).

In the worst-case scenario, the IEC may be forced to start checking ballot boxes in order to reconcile the various discrepancies, since the votes as counted may include BVV votes, non-BVV votes that are backed up by actual ballot papers with QR-codes, and non-BVV votes that exist only on the result sheets.

Edited by Martine van Bijlert


(1) The fact that data was shared while it was still uploading explains some of the differences between the first and the second IEC update, such as the number of votes going up while the number of polling centres remained the same, as noted by AAN in this earlier report. For example, Panjshir reported a turnout of 10,144 from all its 96 polling centres on 28 September, while on 29 September from the same number of centres 19,205 votes were reported. In Daikundi and Kunar provinces, the same thing happened (see Table 1).

(2) The director of Transparent Election Foundation of Afghanistan (TEFA), Muhammad Naim Ayubzada, told AAN that TEFA’s 5,200 observers had collected data in all 34 provinces and had found the participation level to be very low. He believed that whatever number or statistics the IEC had announced on the turnout and level of participation could not be realistic because “they collected the data by phone.” Free and Fair Election Forum of Afghanistan (FEFA) spokesperson Samira Rasa told AAN that FEFA’s had fielded 6,000 observers in all 34 provinces and said that the “level of participation in the presidential election had been very low compared to previous elections.” The Afghanistan Civil Society Forum organisation (ACSFo), which had 4,647 observers in the country’s 23 central, northern and western provinces (in 2,644 PCs or 19,700 PSs) also observed low participation.

The Afghanistan’s Independent Human Rights Commission (AIHRC) – which also observed the election – in its press release stated that “the presence of citizens at polling stations was low. For instance, Kunduz province had low voter turnout due to threats, especially for women. For example, in the Itarchi School of Argo of Badakhshan district, from morning to 12:42 in the afternoon, only 13 people voted. But in Bamyan, Daikundi, Herat and Helmand provinces participation among women voters was high.”

(3) Additionally, Kabul daily Hasht-e Sobh reported on 8 October 2019 that 22 BVV devices had been quarantined so far. IEC quarantined 16 BVV devices from Badakhshan, three from Laghman and three from Kunduz province. According to the IEC, Hasht-e Sobh reported, these devices had no memory cards or had other technical issues.


Annex 1: Overview of open and closed PCs and PSs on 28 September by province, received from commissioner Mawlana Abdullah on 3 October 2019.


IEC Mawlana Abdullah Presidential election