Elections 2014 (52): The not yet officially announced results – electoral maths with unknowns

Share

After over five months, Afghanistan has finally an election result … kind of. In a remarkable step, the Independent Election Commission (IEC) announced the winner of the run-off on 21 September 2014, but did not release the results – apparently one of Abdullah’s unbending conditions for reaching a final agreement on the national unity government. Instead, the IEC said it had sent the results to both teams. The Ghani team immediately released the figures on social media and to the press, making them at least available for analysis. But the absence of an official public release remains a blemish on the election outcome. AAN’s Thomas Ruttig has taken a first look at the results and what they may tell us about both the election, the turnout and the audit (with contributions by Martine van Bijlert).

The decision not to release the results when the IEC ‘named’ Ashraf Ghani Ahmadzai as the new president, as some news agencies put it, means that the elections ended with the same lingering uncertainty that dominated them all along. Ghani will be sworn in on Monday 29 September 2014, and soon we will also know whether his bitter election rival and now partner in the government of national unity (GNU), Abdullah Abdullah, will become the quasi-prime ministerial Chief Executive Officer himself or nominate someone else. There is also the matter of another new official position, ‘the leader of the opposition’, detail of which was left vague in the Abdullah/Ghani political deal (text and analysis in our earlier dispatch here), but which appears rather redundant given the opposition will be part of the GNU. After that, as Ghani said earlier, it might take up to 45 days to put together the cabinet.

In the meantime, however, the Abdullah team has protested against the IEC disclosing the result and called it a breach of the GNU agreement.

What many Afghan voters and observers would still like to know, however, is how the results were arrived at.

Election maths with various unknowns

One central – and still widely open – question is how many votes were invalidated and where. However, the results released by the IEC to the candidate teams do not tell us what happened during the audit. IEC chairman Yusef Nuristani admitted in two press conferences that there was fraud by both sides and in the range of “hundreds of thousands” votes, but stopped short of providing more exact figures. And although the IEC has released lists of which polling stations were fully invalidated and which were recounted and/or partially invalidated (see, for example, a list of invalidated polling stations here and our analysis of the procedures here), we also do not yet know, for example, how many votes were invalidated at any given polling station. (1).

We have reported earlier about several other unknowns: (un)reliable populations figures (the official ones from the Afghan Central Statistics Office, CSO here); the striking difference between the assumed number of voters (13 million) and of distributed voter cards (21 million) and, as a result of this, the lack of a voter registry (see here, here and here).

So the results released merely represent the net outcome of all the recounts (including, in some cases, addition of votes that had previously not been counted) and invalidations. But it is possible, at least, to calculate the difference between the preliminary and final results. More detail is still very much needed.

It is also possible to draw some conclusions about the turn-out – but it needs to be bourn in mind that a vote found in a ballot box does not necessarily represent a real voter. There is the widespread practice of bulk or proxy voting, ie village chiefs or heads of families, for example, voting for the other (mainly female) members of their family or community which might or might not have been discovered by the audit. This is not only a phenomenon in the Pashtun south of the country. So when we talk about ‘votes cast’ or ‘turnout’ in this text, please read “vote found in a ballot box which was considered valid.”

The IEC result in figures

According to the IEC result sheet shared with both candidates on 21 September (see annex 1), Ghani received 3,935,567 votes (55.27 per cent) and Abdullah 3,185,018 votes (44.73 per cent). This is a difference of over ten per cent and almost 750,000 votes. The total number of valid ballots cast was given as 7,120,585.

Judging from the over seven million validated votes and their gender split, as provided by the IEC, this would mean that 2,640,267 women (37.1 per cent) had cast valid votes countrywide. The number of valid male votes was 4,480,318 (62.9 per cent). According to these figures, Ghani won more votes from both men and women. However, the difference in how the women’s vote divided between Ghani and Abdullah was much smaller (136,617 = 5.2 per cent) than the men’s vote (613,932 = 13.7 per cent). The women’s vote was around the 37 per cent average in almost all provinces. Ten provinces were reported as having had more than 40 per cent of votes cast by women (Daikundi the most with 44.6 per cent, and also Faryab, Samangan, Herat, Bamyan, Takhar, Sarepul, Paktika). Four provinces (Kandahar with 21.2 per cent, Helmand 13.1 per cent, Zabul 10.1 per cent and Uruzgan 7.4 per cent) are clearly below the average (for more see annex 2).

The change from the preliminary results – which were announced by the IEC chairman on 7 July 2014 while negotiations over a more extensive audit were still on-going and which drew immediate protests by the Abdullah team – is 852,142 votes. On 7 July, turnout was put at 7,972,727 (after 136,766 votes had been invalidated). That represented 61.54 per cent of the 13 million that is assumed to be the maximum number of voters. The number of male voters was given as 5,057,613 (or 62.3 per cent) and female voters 3,051,880 (37.6 per cent). Ghani’s result was given as 4,485,888 votes (56.44 per cent) and Abdullah’s as 3,461,639 votes (43.56 per cent). This means that the audit brought Ghani’s share of the vote down by 1.17 per cent and boosted Abdullah’s by 1.21 per cent.

It is difficult to tell, based on these figures, exactly how many votes the IEC invalidated in total, given that some recounts during the audit also unearthed votes that had previously not been counted. But as a net result of the audit, Ghani lost 550,321 votes (64.6 per cent of the total number of votes cancelled) and Abdullah lost 276,621 votes (35.4 per cent). The ratio of ballots cancelled was roughly 2:1 in favour of Ghani – ie for every votes that Abdullah lost in the audit, Ghani roughly lost two. (2) While the audit did not change the overall result, or indeed significantly change the percentage figures, it did reveal just how many fraudulent votes the IEC had initially determined as valid.

Provincial distribution of votes after the audit

According to the figures made available, Ghani won a majority of votes in 18 of the 34 provinces, while Abdullah won in 16. Each won five of the ten provinces the CSO gives as the most populous (although population figures in Afghanistan are not reliable – more of which later): Ghani won majorities in Kabul, Nangrahar, Kandahar, Kunduz and Faryab, while Abdullah took Herat, Balkh, Ghazni, Takhar and Badakhshan. Six of these provinces were also among the ten provinces with the highest audited turnout: Kabul (AG), Herat (AA), Nangrahar (AG), Balkh (AA), Faryab (AG) and Takhar (AA); the others are Paktia, Paktika, Khost (all AG) and Baghlan (AA).

In seven of the ten most populous provinces, the results were clear (see annex 3). This was to be expected for Nangrahar and Kandahar (respectively 77.5 per cent and 82.4 per cent for Ghani), with its almost entirely Pashtun population, for Faryab (65.6 per cent for Ghani), predominantly Uzbek and with the strongest position of the ethno-political Jombesh party (its leader, General Dostum, is Ghani’s running mate for the post of first vice president), and the Jamiati stronghold of Badakhshan (77 per cent for Abdullah). In the ethnically and politically mixed provinces of Balkh, Herat and Ghazni the balance of votes between Abdullah and Ghani was still around 2:1 in favour of Abdullah (respectively 66.6 per cent, 69.2 per cent and 64 per cent).

Kabul (51.1 per cent for Ghani versus 48.9 per cent for Abdullah), Kunduz (57 per cent for Abdullah vs 43 per cent for Ghani) and Takhar (52 per cent for Ghani vs 48 per cent in favour of Abdullah) – all with mixed populations – were much closer runs.

Both Kabul and Kunduz also experienced low turnouts. In Kabul’s case, the reported turnout was particularly low, at only 43 per cent, ranking it 27th among the 34 provinces (but because of its huge population, it still produced the highest number of actual votes). This pattern of relatively low turnout was also seen in earlier elections. In Kunduz’ case, this year, the low turnout can by explained with repeated periods of fighting and the presence of a large number of armed groups (see here, here and here); it was 16th among the 34 provinces in terms of turnout and 28th in terms of percentage of voters registered).

Khost, Paktia and Paktika had extremely high reported turnout, more than 100 per cent compared to their official Afghan population estimates even after the audit. Paktia had 11 per cent more votes than people of voting age according to the CSO population, Khost 32 per cent and Paktika 40 per cent more. Ghani won all three with sky-high majorities of 89.5 per cent of the vote or more. This created an outcry from the Abdullah camp, alleging they represent mass fraud (3), and turned into a major reason for its demand for an UN-supervised audit. The counter argument fielded was that the CSO figures for the three provinces are too low (they put Khost at number 16 of all provinces only, Paktia at 18 and Paktika at 26). This is possible, too, but neither version can be verified as long as there is no census.

Indeed, if the differences of 803,000 votes in Ghani’s favour from these two provinces (340,000 in Khost, 243,000 in Paktia, 220,000 in Paktia) had been invalidated altogether, this would have put Ghani and Abdullah level country-wide. However, not all votes in Loya Paktia were fraudulent. Even with an average turnout of around 50 per cent of their officially reported population (which might be too low), the three provinces would have come up with around 360,000 voters which still would have been sufficient for Ghani to win.

In contrast, Helmand, Nimruz, Zabul and Uruzgan in the south, Pashtun and assumed to be pro-Ghani, all highly insecure and prone to manipulation, reported low or very low turnouts – as in previous elections. Farah (65,226 votes) and Nimruz (with 29,886) had comparatively unimpressive turnouts, given that at least parts of their populations are relatively progressive. Helmand (with 57,549 votes) ranked behind even little Nuristan), and Zabul (20,200 – although 66 per cent of its originally reported vote was disqualified) and Uruzgan (19,193) were even further behind.

Also from other provinces, very high, although not as extreme turnout figures were reported. Kunar (94.4 per cent turnout, i.e. based on the CSO determined voting age population), and Wardak (64.7 per cent) were won relatively clearly by Ghani (87.5 per cent vs 12.5 per cent and 62.6 per cent vs 37.4 per cent), Daikundi (87.4 per cent turnout), Samangan (76.6 per cent) and Ghor (83.3 per cent) clearly went to Abdullah (77 per cent vs 23 per cent; 76.1 per cent vs 23.9 per cent; 71.8 per cent vs 28.2 per cent). The result in Baghlan (78.7 per cent per cent turnout) produced a comparatively close with 57.1 per cent vs 42.9 per cent in favour of Abdullah. Nuristan (85.3 per cent turnout) and Panjshir (86.2 per cent) were both clearly won by Abdullah, with 67.3 per cent vs 32.7 per cent and 93.5 per cent vs 6.5 per cent, but did not carry much weight for the countrywide result because of small their populations. Also the high turnout figures for Kunar, Wardak, Nuristan and, to a lesser degree, Ghor and Baghlan were surprising, given the high degree of insurgent or other illegal armed groups’ activity.

In Loya Paktia, the vote was probably high because a combination of different factors played out differently in each province: genuine mobilisation on ethnic or other grounds, mass proxy voting, including on behalf of women and ballot stuffing, as well as intimidation and other forms of strong-arming of voters in areas where warlords or other armed groups (whether illegal or under nominal government control, like the ALP) have the say. (These factors, to varying degrees, were also present in many of the other provinces with unexpectedly high reported turn-out.)

The provinces most affected by the audit

The province most affected by the audit in absolute terms was Paktika, which lost 121,726 (30.1 per cent) of its total number of votes. In relative terms, Zabul was most affected, losing 66.3 per cent of all its votes (39,654 of its 59,854 preliminary votes). Both provinces had largely voted for Ghani. Other provinces that lost relatively large parts of their votes were Nuristan (27.6 per cent – AA majority), Wardak (24 per cent – AG), Badghis (19.1 per cent – AA) Ghor (18.8 per cent – AA), Kandahar (15.7 per cent – AG), Paktia (14.7 per cent – AG) and Baghlan (13.7 per cent – AA).

In absolute terms, Ghani lost most votes in Paktika (104,106), Wardak (50,008), Khost (46,753), Kandahar (44,982), Paktia (43,021) and Zabul (38,756). In these same five provinces, Abdullah lost respectively 17,620 (Paktika), 6,543 (Wardak), 830 (Khost), 5,280 (Kandahar), 5,954 (Paktia) and 898 votes (Zabul). Abdullah lost most votes in Ghor (46,592), Badakhshan (29,323), Baghlan (22,597), Nuristan (17,746) and Paktika (17,620). In these same five provinces Ghani lost respectively 15,120 (Ghor), 966 (Badakhshan), 29,945 (Baghlan), 4,589 (Nuristan) and 104,106 (Paktika).

The findings point towards significant fraud in familiar places and towards more mass fraud in favour of Ghani than Abdullah. However, they show a much less clear-cut divide on Pashtun/non-Pashtun or north/south grounds. There were also significant numbers of fraudulent votes for both candidates in many of the ‘high fraud’ areas, even for the candidate not perceived to be dominant, for example for Ghani in Ghor and Abdullah in Paktika.

What about the audit?

The audit results have not been released by the IEC and have not been accepted by Abdullah and his team. Abdullah finally walked away from the audit at the very end, with his spokesperson claiming that it was “a joke,” even though he had initially demanded it and experts from both teams had been involved in its design and amendments.

This has left the status of the audit itself and its findings somewhat hanging in the air. Most observer missions have yet to finalise their reports, and some of them may not ever be made public. However, although there is a consensus that the audit had its problems, there is no consensus that it was so flawed that it can be dismissed.

The National Democratic Institute (NDI), for instance, in its statement on 24 September 2014 said:

In addition to the appraisal of reforms needed to prevent the recurrence of abuses in future elections, the Afghan people deserve a full and public account of the 2014 electoral process, including timely release of all data related to the runoff audit. While electoral fraud as well as certain problems in the audit process make it impossible for any official results to precisely reflect the votes cast, evidence was not unveiled that would cause the outcome to be reversed. The review of the 2014 electoral process should include a thorough inquiry into the vulnerabilities that allowed high levels of fraud to take place, and the identification and prosecution of those responsible for it.

The European Union Electoral Assistance Team (EU EAT), on the other hand, maintains that the real level of fraud was not uncovered by the audit at all, stating that:

The audit procedures, negotiated with both sides and therefore inevitably complex, have been at times inconsistently and hastily applied under high political tension. This has contributed to an imperfect effort to separate fraudulent votes from clean votes.

The EU EAT concludes that, even after this full audit, questions remain on the electoral process and on the final outcome, in particular as the audit failed to bring full clarity on final results. The political agreement reached should now be the start of a new era. The EU EAT bases its position on internal calculations, which it has not shared, despite repeated requests. It however also “regrets that no precise results figures have been published.”

Another case of strong rejection of the audit combined with claims that, however, are not substantiated came on the Foreign Policy website on 23 September by Chris Mason:

The U.S. Center for Naval Analysis (CNA) said in July that it

was mathematically impossible for Ghani to win, given Afghan demographics and the initial 46 percent to 32 percent first-round vote spread, according to sources familiar with the analysis. According to sources who reviewed the private report, the top experts in statistical analysis in the United States used every known computer model of election balloting and concluded that a Ghani victory was scientifically impossible. In simple terms, there is no mathematical doubt that Abdullah Abdullah won.

It concludes that “[t]he fraud perpetrated on the Afghan people this time was on an epic scale […] exceeding two million fake votes.”

Hyperbole aside (“top experts”, “every known computer model”, “scientifically impossible”), it would, of course, be useful if the reports on which such assertions were bases could be shared so they could be scrutinised and inform the debate.

The audit invalidated (a net result of) 826,942 votes. It did not find the two million fraudulent votes Abdullah had been claiming and, although there were clear indications of ballot stuffing and crude manipulation (more in some places, less in others), there was no evidence of widespread ‘industrial-scale’ fraud. The Abdullah observers were present during most of the audit – and would definitely have shouted out loudly if they had come across such cases. (Remember the brawls about even small differences of opinion.) The Abdullah team also did not give any indication that they saw general flaws in the audit’s methodology – if this had been the case, they should never have been part of it in the first place. Either the fraud did not exist on the scale that was claimed or it was sophisticated enough to escape the audit. It is a pity that neither the auditors nor those still claiming mass fraud, have provided details and contributed to clarity.

What could (and should) have been done

It remains true that, had a census, as stipulated by the 2001 Bonn agreement, been carried out and standardised ID cards distributed for all Afghans early on in the Bonn process when security conditions were much more conducive for reaching all parts of the country, a better election with a reliable result could have been organised. On the basis of such ID cards, a more reliable voter registry could have been built.

Projects to carry out a census and establish a voter registry had been on the table early on and the latter again in preparation for the 2014 elections; they included a UNDP project, supported by the old IEC (before its members were changed). These projects were rejected by the government in favour of introducing the e-tazkera, an electronic ID card that would also serve as a tamper-proof voter card. The government decided on this course of action, even though it was clear from the start that there was not enough time to reach even the majority of the population, let alone all of it. As was expected, the e-tazkera project was not finalised in time for the election, and the old and easy-to-forge voter cards remained in use: 21 million have been officially distributed – far more than the total voting population – and many more illegally printed in neighbouring countries.

A new voter registry for the 2014 election was made impossible by a government which lacked political will or indeed interest (being able to muddy the election waters in 2009 had been useful to the president). It was also able to play different donor countries with interest in getting the e-tazkera project against each other. On the census, donors did not push for it in the early post-Taleban years, fearing getting bogged down by the highly charged issue of whether the ‘ethnicity question’ should be asked. But it would probably have been possible, at least, to pull together data collected in the meantime collected by household surveys, vaccination and other programmes by government, UN and other agencies, and NGOs (for many years, the anti-polio campaign covered almost all of the country). The CSO population figures could then have been verified and, if needed, adjusted.

The full distribution of the e-tazkera again figures in the GNU agreement, as a task in preparation for next year’s parliamentary election (plus possibly the district council election, needed for the constitutional loya jirga that is, as stipulated in the GNU agreement, too, to be held within two years’ time). Although it seems unrealistic to get it done by then (before the election, the Ministry of Interior had estimated the e-tazkera would need a decade to implement), this and the census – for which pilot projects have been done – need to remain priorities for the Afghan government.

All this is still in the future. For now, if the two partners in the GNU agreement are serious about the electoral reform they have agreed to, their first step should be to push for the official public release of the full results of both the election and the audit. The figures need to be scrutinised and conclusions drawn for an urgently needed improvement of the quality of Afghan elections. Transparency is an indispensable prerequisite for this and, in fact, for a democracy.

(1) In order to get an idea about the dimension of invalidated votes, I have made the following calculation: Based on UNAMA’s press statement of 22 September 2014 that the IEC had completely invalidated the votes of 1,206 polling stations (of altogether 22,828, i.e. 5.3 per cent of them), the number of maximally 13 million voters (see how Afghan election-related figures, including this one, were arrived at here) and the IEC’s official turnout figure (7,120585, i.e. 54.77 per cent of 13 million), this would give us around 396,300 votes disqualified in those boxes (600 votes maximum in each x 1,206 boxes = 723,600; 54.77 per cent of this = 396,300). This would amount to almost half of the difference (852,142) between the IEC’s turnout figures in the preliminary result and in the end result.

There also was a much higher number of polling stations (5,316) that have been partly disqualified and of 1,618 others that were “deferred to further decision-making” already after the first five (of eight) IEC sessions. Even if most of them had been ruled valid by the IEC in their final verdicts, the remaining invalidated votes could run into the 100,000s also. Summing up both categories (fully invalidated and partly invalidated) could probably bring us into the region of the difference of 852,142 votes between round 1 and 2.

(2) In the first round, on 5 April, Abdullah, coming out of it as the leading candidate, scored almost the same percentage (44.94 per cent), representing 2,973,707 votes (100,000 less), based on a reported turnout of 6,892,816 –200,000 votes less than in the run-off. Ghani had 2,082,417 votes (31.47 per cent). There were still six other candidates who united 1,562,000 votes on themselves, most of which might have been transferred to one of the contenders in the run-off.

Annex 1: original result sheet sent out by the IEC and distributed by the Ghani team

Annex 2: the male-female vote, calculation by AAN, based on the IEC figures in annex 1

Annex 3: Provincial results by percentage (sorted by biggest winning margins of Ghani)

No information given of number of invalid votes cast or votes invalidated during the audits.

Endnotes:

Revisions:

This article was last updated on 9 Mar 2020

Tags:

election Gender Ghani presidential provinces Result

Authors:

Thomas Ruttig

More from this author