Is it possible to predict the upcoming Afghan elections on the basis of polling data? There are severe methodological challenges in collecting accurate survey data in a country like Afghanistan, and, given how young the country’s democracy is, not much past data on how well those predictions turned out. But surveys during the 2009 presidential election did prove to be fairly consistent with regards to the final vote count. And a number of surveys conducted in recent months have shown reasonable consistency. AAN guest author Matthieu Aikins (with input from the AAN team) takes a closer look at the data.
On Saturday, the consulting company ATR released a self-funded survey conducted using an algorithm that randomly generated cell phone numbers, which were then weighted in an attempt to create a representative sample of Afghanistan’s potential voters. It was conducted from 19-26 March and showed the following results:
|Candidate (top five only)
||Result excluding undecided/none
|Abdurrab Rasoul Sayyaf
|Gul Agha Sherzai
|None of them
According to the survey, less than two weeks before the elections Ghani and Abdullah hold commanding leads, at least among its sample—indeed, if we discount undecided and non voters (right column), then each are only about 10% away from the 50% threshold needed to win the election without a runoff vote – assuming that the undecided voters will divide their votes in the same way.
This is fairly consistent with polls by Glevum and Democracy International conducted in November and December 2013, which also showed Ghani and Abdullah as clear frontrunners:
|Candidate (top six only)
If these surveys are accurate in predicting the elections, then a Ghani-Abdullah runoff elections seems highly likely. But how accurate will they be? The surveys conducted prior to the 2009 presidential election were fairly consistent with the final results:
There is, however, a major caveat here, one that speaks to a dilemma that overhangs the entire election. These are the final certified results from 2009, which means they are the result of major adjustments by the Independent Electoral Commission and Electoral Complaints Commission intended to combat fraud. In the preliminary or raw results, Karzai actually took 54.6% of the vote, versus 27.8% for Abdullah. When the IEC refused to act in the face of widespread fraud, the ECC intervened and ordered a sampled audit of all ballot boxes that met certain criteria suggestive of fraud. The resulting process—opaque, confusing, and politicised—disqualified around a million votes, the bulk of them for Karzai. So the surveys might have predicted the ‘fair’ result. But it’s also possible that the survey data may simply have been in line with what was generally perceived as “reasonably fair” result, introducing the problem of circularity.
And this is a problem that looms over the present election as well. Afghanistan’s electoral system is so badly hampered by fraud, insecurity, and institutional weakness that there is no effective way of knowing what the ‘true’ vote is—that is, how many voters individually cast their ballot. Indeed, we don’t even know how many voters there are, as there is no recent census or voter rolls. Estimates range from 10 to 12 million, while some 21 million legal voter cards have been issued.
What’s more, the notion of the only legitimate ballot being one cast by an individual voter is itself problematic. As past elections have shown, Afghan voters, especially in rural areas, are mobilised through networks of patronage and solidarity. An entire village might decide how to vote, after its elders have negotiated with a candidate—and this might mean casting their women’s votes en masse.
Moreover, given the geographic distribution of insecurity in the country’s south and southeast, which is strongly correlated with fraud, rigorously disqualifying fraudulent votes might have the effect of disenfranchising largely Pashtun areas, which could in turn jeopardise Afghanistan’s fragile political order.
With these points in mind, it becomes clear that current surveys will be of limited use in predicting the final outcome of the first-round election, because so many other variables, most notably the IEC/ECC adjudication process, will intervene. Nevertheless, if we assume that the surveys do accurately capture something about their samples—in this case, a significant lead for Abdullah and Ghani—it is likely that the preliminary election results will differ from the survey results for three reasons.
First, over the final two weeks, the candidates may gain voter share from undecided voters or from each other. For example, the March ATR sample was also the first subsequent to Qayum Karzai’s withdrawal from the race, and, given his endorsement of Rassoul, it’s possible that his support will accrue to him on election day.
Second, voters may follow groups patterns that are underrepresented in the survey. That is, calling and asking individual Afghans about whom they plan to vote for may not accurately capture an election process that is mediated by patronage networks and community mobilisation. Many Afghans who vote on 5 April may well vote for a specific candidate because their community and elders have decided on them. This might strengthen candidates who rely on traditional networks as opposed to, say, young urban voters.
Third, there is the issue of fraud. Given how little has been done to address the underlying issues behind massive fraud in the 2009 and 2010 elections, this election will most probably see similar levels. In those earlier elections, around a million votes or a fifth of the total cast were found fraudulent and thrown out in a painfully contested and public battle that did lasting damage to both Afghan’s confidence in their own democracy and the country’s relationship with the international community.
Given the number of candidates, the complex and cross-cutting alliances behind them, and the lack of overt state sponsorship, fraud patterns this time may resemble 2010’s mosaic rather than 2009’s sharply-drawn battle lines. Either way, fraud will be concentrated in areas of the country where insecurity is highest, in particular the south and southeast but also, given the deterioration in security since 2010, many other parts of the country.
Because they will draw varying levels of support from such areas, it seems likely that fraud will be perceived as benefitting certain candidates more than others.The result will be a contentious and politicised adjudication process by the IEC and ECC, where, as in past elections, it will prove all but impossible to clean-up the election process ex post facto. A very chaotic electoral process, it seems, is the one thing that can be confidently predicted for the 2014 presidential election.
This article was last updated on 9 Mar 2020