Nearly two months after Afghans cast their presidential ballots, the electoral battle continues with the possibility of spiraling into violence. With what some dub the Kerry Agreement and the recent sequel Kerry II, the presidential candidates Ashraf Ghani and Abdullah Abdullah pulled back from the brink, but disputes between the parties continue in a winner-take-all battle amid the UN-supervised national IEC audit.

Why have the candidates continued to fight? There was almost certainly fraud on both sides as supporters took advantage of Afghanistan’s insecurity and institutional deficits and found varying ways to “rock the vote.” However, the mere presence of fraud rarely matters; the fraud must be great enough to change the results. The preliminary results of the June 14 run-off show Ghani ahead with 4.5 million votes to Abdullah’s 3.5 million, about 56 percent to 44 percent. Does Abdullah really believe that he can overcome a million-vote difference?

I argue, with a couple assumptions, that Abdullah can reasonably believe that he can still win. As such, the parties will continue to fight over every vote and escalate when necessary, further threatening the stability of Afghanistan. This high-stakes game of electoral chicken will likely continue until the two sides collide or until the United States, desperate for some semblance of stability, can persuade a candidate to accept defeat.

Industrial-scale fraud

Two assumptions are necessary for Abdullah to believe that victory is still possible. First, he must believe that accusations of industrial-scale fraud by Ghani must be founded. The accusations focus on the preliminary turnout numbers: 8 million valid votes in the run-off versus 6.6 million in the first round. The change in turnout is certainly sizable. Breaking the total down by province shows that the change in turnout is varies considerably across provinces. Two provinces, Khost and Paktika, stand out as they account for 511,000 new votes, 38 percent of the total increase in turnout.

Ghani made heavy gains in these two provinces. Ghani picked up 544,000 votes to Abdullah’s 34,000; this also includes votes that went to other parties in the first round. Ghani made large gains in other provinces, such as Kabul and Kandahar. But those could be largely explained by Ghani picking up the large number of third-party votes up for grabs in those provinces. Graph 1, which shows Ghani’s vote gains by changes in turnout, highlights how Khost and Paktika are outliers.

This explains why Abdullah’s statements have focused in on these provinces. Ghani has argued that the gains were legitimate. The Afghanistan Analyst Network reported a possible surge in Loya Paktia (Paktya, Khost, and Paktika), but at least one observer disagrees. The surge is especially suspicious given the estimated populations of these provinces. According to the 2013-2014 population figures from the Central Statistics Organization (CSO), the run-off turnout represents 72 percent of Khost’s total population and 96 percent of Paktika’s. These numbers are especially suspicious since the CSO’s breakdown by age estimates that 50 percent of Afghanistan’s population is old enough to vote, and a March survey found that 10 percent of adults were not likely to register to vote in the election. In the first round no provincial turnout exceeded 50 percent of the population. The CSO population estimates are based on an incomplete 1979 census, so they are almost certainly wrong. But as they stand, those numbers would support Abdullah’s assumption of industrial-scale fraud.

Still, even if we assume that Abdullah is right about Khost and Paktika and that the additional turnout is indeed due to fraud, it is not enough for Abdullah to catch Ghani. If we only allow Ghani to gain the available third-party votes and invalidate the rest, this only disqualifies 477,000 votes for Ghani, leaving him with a comfortable lead. To make up the million-vote difference we would have to disqualify Ghani’s extra votes from all nine provinces labeled in Graph 1, which seems a much less reasonable prospect. Why then would Abdullah risk destabilizing the country just so he could lose by half a million votes?

Baseline fraud from 2009

The previous section looked at “industrial scale fraud,” but what of the regular every day fraud? This brings us to the second assumption, that the baseline fraud in this election is comparable to that of the 2009 election. With this additional assumption, Abdullah may honestly believe that he can win this.

The previous section took the first-round votes as valid and focused on the additional turnout. Unfortunately, there are no available measures of how fraudulent the first round was. Election officials rushed to get the first-round results out quickly. As the first-round outcome, an Abdullah-Ghani runoff, was long clear, the two parties were less aggressive in finding fraudulent votes. A report by the Afghanistan Analysts Network (AAN) wrote that “timelines appear to have trumped transparency and rigor. They conclude, “Interestingly, we have now reached the end of the first round, without a clear picture of how clean or how fraudulent the vote actually was.” Table 1 shows that in the 2014 first round, the final certified results barely changed from the preliminary results.

To establish a baseline level of fraud, we may be better off looking at the 2009 election. The 2009 election had similar vote totals to the first round of 2014 and was also highly contentious. Journalist Havana Marking wrote of the election, “The buzz ‘question’ … among the international community is just how much fraud is acceptable.” The deputy special representative of the United Nations in Afghanistan accused the UN of ignoring the election and was allegedly fired for it.

The number of votes disqualified in the 2009 election, shown in Table 1, are much larger. Karzai lost many more votes than Abdullah, resulting in Abdullah picking up 644,000 votes. Fortunately the country avoided a run-off battle as Abdullah decided to pull out of the run-off rather than legitimize what he saw as a fraudulent election. Beaulieu and Hyde (2009) explain that opposition parties often boycott votes with international observers in order to deny legitimacy to an incumbent who will still cheat, just more stealthily.

Abdullah did not pull out this time, and seems unlikely to give up the fight until the audit is complete. However, his stubbornness is not without reason. If Abdullah assumes that he can (1) expose massive fraud in Khost and Paktika and (2) see baseline gains similar to 2009, that adds up to a net gain of 1,121,000 votes. This is enough to overturn the results.

Table 1: Preliminary and Final Results in 2009 and 2014
Candidate Election Votes Vote Disqualified % of Vote Lost
Abdullah 2014 1st Round Preliminary 2,973,706 1,565 0%
2014 1st Round Final 2,972,141
Ghani 2014 1st Round Preliminary 2,082,417 -2,130 0%
2014 1st Round Final 2,084,547
Abdullah 2009 Preliminary 1,571,581 165,339 11%
2009 Final 1,406,242
Karzai 2009 Preliminary 3,093,256 809,349 26%
2009 Final 2,283,907

Every vote counts

Although Ghani enjoys a substantial lead, I have shown why Abdullah may still believe he can win. This helps explain why the electoral aftermath has been so confrontational. Abdullah is not fighting to damage Ghani’s legitimacy or to establish a norm of fair election, Abdullah is fighting to win. With the presidency on the line, then, it is not surprising that it has taken the parties so long to agree on an audit process. Without an accurate census, a voter registration database, or exit polls to compare to, the audit will have the final say. When every vote counts, it matters whether a fraudulent sample invalidates an entire ballot box, if a fingerprint can stand for a tick mark, or how similar check marks can look.

The United States has walked a narrow line on Afghanistan’s elections. The United States is clearly interested in the election leading to a stable government. According to the SIGAR quarterly report (p. 128), USAID programs supporting the 2014 elections disbursed over $60 million. However, the United States is has also been very sensitive to any accusations of meddling. Four months before the first round vote, the United States cut off funding for all election polling after questions of meddling arose. The United States maintained its hands-off stance even when Abdullah declared fraud just days after the June 14 election. However, once Abdullah’s response to the alleged fraud threatened Afghanistan’s stability, America’s non-intervention stance vanished as President Obama called the candidates and Secretary of State John Kerry headed to the airport.

Ultimately, the parties will likely continue to escalate until one side can be sufficiently incentivized to accept defeat, something that the United States is keen to do. The second Kerry agreement did just that by providing the losing side more influence in a unity government. These kinds of deals may significantly change Afghanistan’s government structure. As Daxecker’s (2012) research on election monitoring concludes, there are “potential underlying tensions between a commitment to democratic norms and an interest in political stability. . .” Changes to Afghanistan’s democratic norms may not be a bad thing however if it facilitates smoother election in the future.

For now, we can only hope that this deal and any others will pass the final test by remaining intact when a victor is ultimately named. At this point it may be a Pyrrhic victory. Any resulting government will still face Afghanistan’s other existential threat with a legitimacy handicap. In Afghanistan’s 2009 election Berman et al (2014) found that even local election fraud eroded government support. Still, the alternative to a political settlement is sobering.

Thomas Scherer is a Ph.D student in the Politics Department at Princeton University.