Afghanistan Analysts Network – English

Political Landscape

What to Watch in the Elections (1): Voter registration

Martine van Bijlert 8 min

Elections in Afghanistan can be overwhelming, surreal and utterly confusing, even for those who have followed the politics of the country for a long time. To bring those who were not here during previous elections up to speed, and to remind those who may have forgotten the details of what it was like, AAN’s Martine van Bijlert answers some crucial questions. She discusses voter registration, the securing of polling stations, election day fraud, and the laborious process to settle on an election outcome – in four installments. This one, the first, focuses on voter registration and the significance of the millions of extra voter cards not linked to voters.

Why should we be watching the voter registration?

Afghanistan does not have a voter registry; nor does it have any kind of reliable population database. Population figures released yearly by the Central Statistics Office – the latest can be found here – are estimates based on household listings. This means that nobody knows how many voters there are, where they live or where they may want to vote.

Voter registration has done little to address this problem. The quality of the database, the basis of which was created during the first voter registration drive in 2004, is very poor. Voters are not linked to locations (only their provinces). Moreover, with no effective controls in place, the registration process resulted in a massive over-distribution of voter cards (see below). Some people registered multiple times; blank voter cards were sold, often in bulk, to the highest bidders; and voter cards were hoarded through mass proxy registration, in particular of (ghost) female voters (see pp. 19–21 of this AAN report).

In conservative areas, voter cards for women were often provided without the obligatory picture and men were allowed to ‘proxy’ register – and later vote – for the women of their families. This made acquiring large numbers of voter cards easy, even for those without special relations to the Independent Election Commission (IEC), while others with inside help from IEC staff at the central or provincial level were able to acquire large numbers of usually blank cards.

Over-registration is largely driven by the fact that voter cards, like votes, are treated as a commodity. Candidates need large numbers of (copies of) voter cards to register their candidacy. Political brokers, at all levels, use cards to shore up claims of their ability to deliver vote banks. And anyone who wants to engage in massive ballot stuffing or large-scale proxy voting needs large numbers of voter cards in order to fill in the forms and make it look as if an actual vote has taken place. (When voters come to cast ballots, their card numbers are registered on a list; if the quantity of ballots in the box does not match the amount of voter card numbers on the list, the results are flagged as suspicious.)

What did this year’s voter registration look like?

Like previous rounds, this year’s voter registration was a top-up exercise, meant for the recently arrived (for example refugees returned from Pakistan or Iran), those who newly turned 18, those who had lost their voter cards, and those who for some other reasons did not register in the past. The registration drive, ending today, has so far distributed over 3.7 million additional voter cards, 65 per cent for men and 35 per cent for women. The latest figures can be found on the IEC homepage.

The additional 3.7 million cards bring the grand total of distributed cards to over 21 million. According to the latest CSO figures, Afghanistan’s total population was around 27 million in 2012 (25.5m settled and 1.5m nomadic), with half of the population under 15 (48.4 per cent) and around 60 per cent under 20 (59.1 per cent) (NRVA 2011–12 report, p. 169). This means that the maximum number of eligible voters – all Afghans over 18 – is around 13.5 million. So even if every single person of voting age had acquired a voter card – which is clearly not the case – there would be at least 7.5 million additional voter cards floating around, not linked to real voters. In practice, given the considerable number of people still not holding voter cards, the number of unlinked cards is probably closer to 10 million, although not all of these will still be available.

In practice, every registration round has combined genuine registrations, (mass) proxy registration and the sale of cards. It remains fairly easy to acquire cards without following the proper procedures. In the past, suspicious voter registration figures have been a good indicator of preparations for ballot-stuffing and mass proxy voting. This year’s registration figures have shown similar suspicious patterns – high female registration in conservative areas and relatively high registration figures relative to the population, particularly in insecure areas (for details, see below) – although the patterns have not been as pronounced as in, for instance, 2009.

An optimistic interpretation of the fact that these patterns are now less pronounced, is that – for some reason – there will be less fraud. An alternative interpretation is that the country is to a certain extent already ‘saturated’ and that many players already have all the excess cards they need. (With millions of excess cards floating around, the urgency to hoard even more may not be equally pronounced. On the other hand, they remain a free commodity.) The fact that the IEC tried to clamp down on the disbursement of female voter cards without pictures seems to have somewhat – but not fully – curtailed the practice of mass proxy registration.

What are these suspicious registration patterns and how bad are they?

The most important suspicious patterns are (a) unexpectedly high registration numbers, to the point of exceeding the total estimated number of voters, or even the estimated total population; and (b) implausibly high proportions of female registrations, particularly in conservative areas.

In every election, implausibly high proportions of women were registered in conservative areas. Where initially this was welcomed as a sign of progress, it soon became clear that the cards were meant to facilitate polling fraud. Although bogus female registration has taken place in all conservative areas – facilitated by the relaxation of rules that allowed cards to be issued without pictures, without the women themselves present and often in bulk to ‘family members’ – the level of organization has been consistently higher in the southeast. This pattern was most pronounced in the run-up to the 2009 elections when in Nuristan, Khost, Logar and Paktia, respectively 71 per cent, 68 per cent, 66 per cent and 62 per cent of the total registrations were female. In comparison, in Kabul – arguably the most cosmopolitan province in the country – female registration figures never exceeded 34 per cent. (For details on past female over-registration, see AAN’s report How to Win an Afghan Election.)

This time around, the provinces with the highest rates of female registration are still largely in the conservative southeast: with the exception of Daikondi, 48 per cent, the highest were in Paktika (47 per cent), Nuristan (44 per cent) and Paktia (43 per cent). Kabul had 33 per cent female registrations.

In terms of total over-registration, the first indications surfaced in the run-up to Afghanistan’s first presidential elections in 2004, when the number of distributed voter cards (10.5 million, which does not include the cards handed out in Pakistani refugee camps for out-of-country voting) exceeded the number of estimated voters (9.8 million). In 2005 and 2009, respectively, 1.7 million and 4.4 million cards were added. By 2010 – which saw a very limited top-up exercise – the voter cards in circulation had reached a highly implausible 17.5 million. The IEC however had based its operational planning in 2010 on an estimated 12.5 million actual voters, while the head of UNAMA Staffan de Mistura floated the figure of 10.5 million, implying that up to 7 million cards in circulation (40 per cent of the total) were not linked to actual voters.

This time around, over 3.7 million additional voter cards were distributed, which brings the grand total to over 21 million. The maximum number of eligible voters, as discussed above, is probably around 13.5 million.

The IEC has extended voter registration in 41 provincial registration centres until 72 hours before polling day. At least in Kabul the demand seems to be greater than the IEC’s capacity to deliver, with still long lines on the last day of registration.

But does the over-registration substantially affect the integrity of the election?

The over-registration does several things: first, it adds to the fluidity and confusion of the process, which in turn breeds ambiguity, opacity and enormous scope for improvisation and manipulation. Second, it effectively facilitates large-scale fraud by putting large numbers of voter cards in the hands of those who wish to stuff the ballot boxes. And third, it confuses the calculations with regard to the voter turnout.

What about the voter turnout?

With all these mismatched figures, getting a solid readout of the turnout on election day becomes impossible. In the first few elections, turnout was calculated simply by looking at the relationship between the number of distributed voter cards and the number of ballots cast. Soon, however, it became increasingly obvious that both figures – voter ‘turnout’ and ballots ‘cast’ – were so inflated as to have become practically meaningless.

In 2009, the IEC apparently introduced a new way of calculating the turnout. According to a 2010 IEC fact sheet released in response to confusion over the figures (see here), the calculation was based on the “maximum number of voters the IEC prepared for”, taking into account local population estimates by the Central Statistics Office and past turnout figures per district. In 2009, the number the IEC prepared for was approximately 15.3 million. In 2010, however, the IEC explained that while its provisional turnout figures would be “based on the number of voters it planned to accommodate, approximately 11.4 million,” the final turnout figures – that is, after finalisation of the count, the audits and the disqualifications – would be “expressed as a percentage of voters who could have voted and did.” This would take into account only the polling stations that had opened on election day and that had not been excluded in any of the audits. It is a method that sugar coats the turnout figures, as it eliminates voters from areas that lacked polling stations from the statistics (apart from depriving them the right to vote).

‘Turnout’, calculated in this manner, basically becomes the percentage of ballot papers used in polling stations that were not grossly affected by the main two problems of the election: insecurity and mass fraud (although the latter is only the case if the IEC does indeed rigorously disqualify votes found to be fraudulent). This can be seen as an attempt to extrapolate how Afghanistan would have voted if circumstances had been better, but it’s obviously not an accurate reflection of how the elections went. According to a later fact sheet, however, the IEC dropped this idea and returned to using the number of voters that it had prepared for as its guiding figure (although it upped the number of voters to 12 million, as opposed to the 11.4 million mentioned earlier).

Neither method of calculation gives an accurate picture of what happens on election day. This is not a criticism of the IEC; under the current circumstances, it is very difficult to conceive of a method that would. But the problem is that such figures, once released, are widely quoted as if they are solid and credible.

But the confusion did not end there. The absolute number of ballots cast on election day in 2010 fluctuated throughout the process, as the IEC scrambled to make sense of its data. In its first estimate, released two days later on 20 September 2010, the IEC indicated that 3,642,444 votes had been cast. A day later the figure was increased to 4,332,871 votes – the IEC explained that more polling centres had returned results than had originally been reported as having opened. When the full preliminary results were released, the IEC indicated that 5.6 million ballots had been cast, out of which around 1.4 million had been disqualified.

Confusion also surrounded the number of ballots that were finally counted as valid:

The Commission would like to clarify that the total number of valid votes reported in the IEC press release of the 20th October is 4,265,347 while the number of valid votes reported by the tally center is 4,271,908. The difference of 6,561 votes roughly corresponds to the votes “invalidated” because the candidates had been excluded from the final list of candidates, while their names and details still existed on the ballot papers. (See IEC fact sheet here.)

Neither of these figures, however, corresponded with the final 2010 turnout summary where the total number of valid votes was given as 4,030,227; the number of invalidated votes (votes for excluded candidates) as 186,367, and the grand total as 4,216,594. The fact that settling on the final figures has been hard is indicative of how confusing the process was.

The IEC has not yet announced how it intends to calculate the (proportional) turnout this time.


Elections female registration fraud polling stations turnout voter cards voter registration votes


Martine van Bijlert

More from this author