Afghanistan Analysts Network – English

Political Landscape

What the preliminary results tell us (1): Kabul provincial council

Martine van Bijlert 5 min

With (international) attention focused firmly on the complexities surrounding the Presidential vote, the struggle for a fair outcome in the provincial council elections continues.

The competition was more localised and the level of organisation of the fraud more limited than in the Presidential elections, but the margins needed to win were also much smaller. Many unsuccessful candidates without money or access to the electoral apparatus have travelled to Kabul to register their complaint or to lobby their case. The list of allegations is by now familiar: various forms of ballot stuffing on polling day, manipulation of the local count, tampering with result sheets, and manipulation during data entry and tallying.
Of course not everyone who complains would have won in a clean election. But several of those who came out on top in the preliminary count results would not have been successful without manipulating the process. An analysis of the results as posted on the IEC website, focusing on unusual or implausible vote patterns, provides some clear clues on how – and in some cases at what stage – the manipulation took place.

Let’s tackle the most complex one first: the Kabul constituency with its 524 candidates (for 29 seats) and its 3410 pages of results. Due to low turnout and the high number of candidates the Kabul results consist of hundreds of pages full of zeroes and low numbers, in which any result over 10 votes (per polling station per candidate) stands out like a foreigner in an Uruzgani village. So here are some examples of what I would consider obvious fraud:

In polling centre 101062 (Suraya Girl’s School in Taimani) candidate numbers 373 and 447 received the following number of votes in the centre’s twelve polling stations:

– nr 373: 185 / 205 / 205 / 0 / 0 / 1 / 0 / 1 / 0 / 0 / 0 / 1
– nr 447: 115 / 95 / 95 / 0 / 0 / 0 / 0 / 0 / 0 / 0 / 0 / 0

Note how the votes of the two candidates add up to round numbers (300 / 200 / 200) and how they have virtually no votes in the other stations. Their shared totals make up more than half of the total votes in those three stations (respectively 509, 464 and 419 votes) with the rest of the votes spread in low numbers over the other candidates. The vote pattern indicates that a real vote took place in this polling centre, but that the polling staff at three stations was recruited to cast extra votes for the two candidates and that they have done so by dividing two or three books of 100 ballots between the two candidates per station.

A similar pattern is found in polling centres 101065 and 101267 in Kabul city (the same two candidates), polling centres 111420, 114456 and 114458 in Charasyab and Bagrami (candidates 248 and 280 – in 114456 they receiving respectively 1200 and 1100 votes!), 109389 in Shakar Dara (candidates 373, 398 and 402), 109387 also in Shakar Dara (candidates 373, 398, 402 and 446), 110407 in Paghman (candidates 234 and 402), and so on. In each of these polling stations the ballots shared between the candidates add up to numbers dividable by 100. In all cases the candidates had (next to) zero votes in the other polling stations.

Other candidates simply arranged to be given the full ballots books on their own. This happened in polling centre 115484 in Sorobi (200 votes each in two polling stations and 300 votes each in three polling stations, out of seven, for candidate 219); 115516 (200 votes each in two polling stations, out of three, for candidate 219), 107380 in Estalef (300 votes in one station for candidate 169; 200 votes in two polling stations and 400 in another for candidate 324), 103341 in Mir Bacha Kot (400 votes each in two polling stations, out of four, for candidate 324), 101301 (100 votes each in three polling stations and 200 each in two stations for candidate 228), and so on. Again, in all cases the candidates had (next to) zero votes in the other polling stations.

None of these obviously fraudulent outcomes would have triggered the IEC algorithms, had they been implemented, as all boxes contained less than 600 votes and less than 95% for a single candidate. And all candidates mentioned above (except 446) made it into the provincial council (seeprovisional list). If there have been no specific complaints against these candidates or polling stations, the votes will remain untouched and counted.

This is only the most obvious fraud. There are many other candidates with remarkably high numbers of votes. This could of course signal local popularity or a high level of voter mobilisation, but it could also point towards more clever or less organised ballot stuffing, or to the moving around of votes between candidates in the data entry centre – as has been alleged by many unsuccessful candidates. You would however need to look at the ballot boxes to find out what happened (are the ballots really there, were they filled in by the same people, etc.).

Candidates with unusually high number of votes – with no obvious pattern – in certain stations include, among others, candidates 108 (in 101330), 180 (in 115486), 258 (multiple stations), 297 (gathered 629 of his 1052 votes in 109389), 442 (gathered 366 of his 474 votes in 107347), and so on.

The allegations that during data entry and tallying votes are added or moved around in order to change which candidates are elected are consistent and detailed. Unsuccessful candidates quote the numbers of votes that they received (and the numbers of votes that have thus been stolen from them) based on the initial polling station counts. Unfortunately none of them seems to have properly documented their data. Their allegations are however corroborated by a large number of credible reports of candidates receiving phone calls from IEC staff, informing them that they did not have enough votes to win, and offering to fix that – for a price – as was the case during the 2005 elections.

The data entry process is surrounded by all kinds of safeguards (including double-blind data entry) but there are always staff members with authorities to check, correct and overrule. If they start selling their services the network can cascade from there. Observers often seem to assume that fraud needs to be centrally organised and ordered, but my impression is that it is much more chaotic than that, with random people – often IEC staff – realising they have something of value to offer, and an expanding network of ad hoc middlemen and -women looking for takers (and vice versa).

Another strong indicator of manipulation – whether during polling, counting or data entry – is the discrepancy between votes cast in the Presidential and provincial council election. Many people have been doing this analysis on the provincial level, but it is the polling station level that counts.

Let’s take for instance polling centre 101077 (mosque of the Polythechnic) where a total of 2886 votes were cast in the provincial council elections and only 1237 in the Presidential elections. The additional provincial council votes were cast in four of the nine polling stations. The beneficiaries of these votes are all familiar (candidates 248, 280, 373, 398 402 and 447) and have all come out of the election successful.

114456 (Bagrami High School), which was already discussed above, is even more pronounced: 3271 in the provincial countil and 1353 in the Presidential election (2300 votes were cast for 248 and 280 alone). The inverse can also be found, for instance in 102470 (Deh Sabz, Pole Charkhi Primary School), where 2942 votes were cast in the Presidential elections (2869 for Karzai) and 1003 in the provincial council elections.

There is much more where all of this came from for whoever wants to do their own analysis. And I apologise for the level of detail, but that is for the few other nerds who would happily spend their day off poring through hundreds of pages of vote results.


Democratization Elections Government


Martine van Bijlert

More from this author