Mapping Election Fraud in Afghanistan
This page shows some figures and maps about the extent of vote rigging in the 2009 presidential election in Afghanistan.
Background: The units of analysis for the results below are polling stations. Stations were set up for a maximum of 600 voters each and belong to polling centers, which are located in provinces. Rather than votes for individual candidates, I analyze voting station turnout.
Method: Evidence of fraud was established according to Beber and Scacco's last digit method, which tests for deviations from a uniform distribution of the last digits in election results (turnout, in my analysis). In an unmanipulated election, the last digits 0 through 9 should occur with equal frequency. However, this is not the case for many provinces in Afghanistan. For example, this plot shows the distribution of last digits for polling station turnout across stations in the Helmand province. Clearly, the digit zero occurs much more frequently than any other. A chi-square test can be used to find out whether the observed frequencies of the last digits deviate significantly from the uniform.
The patterns identified can be compared to the criteria employed by the Electoral Complaints Commission of Afghanistan (ECC), see maps 2 and 3 below.
Analyses based on the initial, uncertified election result
(retrieved October 20, 2009 from http://www.iec.org.af/results/Index.html, 22853 stations)
- Figure 1: The distribution of polling station turnout. Note the large number of stations that reported a turnout at or above the maximum (600).
- Map 1: Provinces with non-uniform last digit frequencies (all polling stations). This indicates election manipulation in these provinces.
- Figure 2: Last digit distribution by province. Most of the deviating provinces exhibit a high number of zeros, but other patterns exist (for example, high number of digit "5" in the Urozgan province).
- Map 2: Suspicious polling stations, type A: spatial distribution of polling stations with 600 or more voters.
- Map 3: Suspicious polling stations, type B: spatial distribution of polling stations with overly high vote shares for one candidate (type B).
Analyses based on the final election result
(retrieved October 22, 2009 from http://www.iec.org.af/results/Index.html, 19474 stations)
- Figure 3: The distribution of polling station turnout. The large peak at around 600 has disappeared, as the results of these stations have been "corrected"/dropped.
- Map 4: Provinces with non-uniform last digit frequencies (all polling stations). Despite the fact that the certified results contain almost no suspicious stations anymore, evidence of manipulation remains for four provinces.
Bernd Beber and Alexandra Scacco's paper on detecting election fraud