A new study by a researcher at the Woodrow Wilson School of Public and International Affairs employs a method that clearly identifies and maps patterns of vote-rigging by province in the initial round of Afghanistan's recent presidential elections, and which may help detect electoral fraud in other countries.

The study, “Mapping Election Fraud in Afghanistan,” by Woodrow Wilson School post-doctoral researcher Nils Weidmann, uses a statistical method devised by political scientists Bernd Beber and Alexandra Scacco, which is based on analyzing last digits and their frequency as reported in election results. For example, if a polling station in Afghanistan reported a total number of 304 votes cast, Weidmann only included the last digit, four, in his analysis. Thus, repeating this procedure for each polling station in Afghanistan’s provinces provides a sample of digits from zero through nine.

How this method works in practice to detect election fraud is that in fair and unmanipulated election results, each of the digits zero through nine should occur with equal frequency across all polling stations. However, fraudulent elections often deviate from this pattern; for example, in Nigeria’s 2003 elections, Beber and Scacco found that the digit zero occurred as a last digit much more often than others, which indicates that mechanisms other than individual voting were involved in generating the reported results.

In Afghanistan’s presidential elections, uncertified election results Weidmann obtained from Afghanistan’s Independent Election Commission (IEC) exhibit similar characteristics to the results from Nigeria. He found that across all polling stations, the last digit zero occurs significantly more often. Deviations from the uniform distribution in last digits are particularly visible in Afghanistan’s southern provinces – an indication of electoral fraud. (Initial results indicated incumbent president Hamid Karzai won by large margins in these provinces.) Weidmann detected fraud in 10 of 34 of Afghanistan’s provinces - almost one-third.

Weidmann even found that when factoring in the order by the Electoral Complaints Commission (ECC) of Afghanistan to identify and correct instances where more people voted than were actually registered, or one candidate received suspiciously high vote shares, evidence of electoral fraud remains even for the corrected, certified election results. In these results, last digit patterns of polling station turnout are suspiciously non-random in three provinces. Thus, Weidmann concludes, the ECC’s guidelines to detect fraud are insufficient.

“The study results are important even with the second round of Afghanistan’s presidential elections cancelled, first, because we know where fraud primarily happened, which may help understand the mechanisms involved in making election fraud possible,” said Weidmann. “In addition, the IEC’s and ECC´s criteria for identifying fraudulent stations seem to be insufficient, as evidence of fraud remains even in the corrected results.”

Weidmann also notes that his results point to a particular type of fraud, that of manipulating return sheets, which explains the suspiciously high number of polling stations reporting results ending in zero. Other types of election manipulation - for example, by bussing people to different polling stations - would not be detectable by examining digit distributions of results. This presents an opportunity to identify and correct the problem if Afghanistan’s government and the international community have the means and the political will – and as well provides a method to potentially prevent and detect fraud in other countries’ elections. 

Maps and charts illustrating the results of the study are available online.