Grading the AfPak experts

Now that the second round of elections has been canceled and Hamid Karzai officially declared the winner of Afghanistan's fraud-riddled ballot, it's time to assess how the AfPak Channel's experts did when they informally predicted the results of the presidential election the day before the August 20 polling in a parlor game we dubbed "The AfPak Crystal Ball."

After the fraud audit reduced Karzai's share of the vote by some one third of his ballots, the incumbent president was left with 49.7 percent, according to official results from the Independent Election Commission. Challenger Abdullah Abdullah wound up with 30.6 percent, and Ramazan Bashardost came in third place with 10.5 percent of the August 20 ballot. And according to the United States Assistance Mission to Afghanistan and the Afghanistan Independent Human Rights Commission, one third of registered voters turned out to cast their ballots.

Our experts jointly predicted Karzai would win 49.2 percent, Abdullah 31.2 percent, and Bashardost 10 percent even, with a 51 percent turnout rate.

As you can see, our experts were almost eerily close in their collective prediction about each candidate's share of the vote. Congratulations are in order for Martine van Bijlert, a regular AfPak Channel contributor and the co-director of the Afghanistan Analysts Network, who came closest to the actual tallies won by the candidates with her guess of 48 percent for Karzai, 31 percent for Abdullah, and 10 percent for Bashardost.

Our experts' accuracy on the candidates is a counterpoint to research done by a leading expert on leading experts, in which UC-Berkeley professor Philip Tetlock found in a quarter-century long study with nearly 300 ‘experts' that their predictions barely beat out a random forecast generator. Chalk one up for the AfPak Channel's hive mind.

What I find interesting but have no convincing explanation for is how our experts were so accurate in their communal forecast of the ballot percentages but were comparatively far off for official turnout (setting aside for the moment the question of whether these figures can be trusted). One guess at this is that our contributors were remembering the nearly 70 percent turnout from the 2004 presidential election; another is we underestimated the effectiveness of the Taliban's campaign of voter intimidation and the level of cynicism among Afghan voters disillusioned with the process, both of which depressed turnout this time around. Andrew Wilder, who founded Afghanistan's first independent policy research organization, wrote at the time that a good national turnout for the 2009 election would be between 40 and 45 percent -- and even that turned out to be somewhat optimistic.