Research‎ > ‎Other data‎ > ‎

Electoral Forensics

The study of 'electoral forensics' has become increasingly popular in the last decade. Like medical autopsies conducted after death arising from suspicious circumstances, statistical techniques can be used to analyze election returns reported in local wards, polling districts, or constituencies. 

The aim is to detect local outliers to the usual patterns of election results, such as in voter registration, voting turnout, vote shares cast for the incumbent, blank or invalid ballots cast, or other anomalies in the official results. Where outliers cannot be explained satisfactorily by other factors, these appear to indicate cases of irregularities arising from practices such as electoral fraud, ballot-stuffing, malpractice, bribery, or vote tampering. 

These techniques are most effective for capturing any problems occurring on or just after polling day.  This is a valuable way to detect the most obvious types of voting anomalies which suggest illegal practices, irregularities and fraudulent manipulation of the outcome, particularly when statistical analysis is supplemented by other types of evidence, including the reports of observer missions, election watch NGOs, or journalists. 

The techniques are unable to detect various types of malpractice occurring earlier during the electoral cycle and prior to polling day, however, such as those arising from media bias in campaign coverage, electoral laws, partisan redistricting, or restrictions on party registration or ballot access for candidates. 

Moreover there is little agreement among experts about the most appropriate mathematical techniques which can be used to analyze overt fraud with any degree of reliability, such as the conditions under which it is appropriate to apply Benford's first digit or second digit 'laws', or to analyze rounding of the final digit in local polling district returns. Debate about the reliability of these techniques continues among statisticians and analysts.

Afghanistan case-study

An illustration concerns the Afghanistan election results, where visual mapping and other techniques have been used to detect possible voting anomalies.

On September 8 2009, the Afghan Election Complaints Commission (ECC)  declared that it had found ‘clear and convincing evidence of fraud’ in the 2009 presidential elections.  It ordered the Independent Election Commission to conduct an audit and recount of polling stations nationwide. More than 3,300 polling stations and over 1.5 million ballots came under this order. The audit looked for polling stations which had either (i) 600 or more vote totals, or (ii) where any presidential candidate receiving 95 percent or more of the total valid votes cast (provided that more than 100 votes had been cast at the station).  

The map below shows the density of polling stations per district affected by the ECC audit criteria of election irregularities. Thus irregularities were found to be most pronounced in the south, southeast and east. 
Given the findings of the audit and the electoral complaints process, 19 percent of all votes cast on August 20 were excluded from the final presidential vote tally. For more details, see the NDI's


For more resources about the general approach of electoral forensics, see: 
  • Bernd Beber  and Alexandra Scacco. 2012. ‘What the numbers say: A digit-based test for election fraud.’ Political Analysis 20(2): 211-234.
  • Joseph Deckert, Mikhail Myagkov and Peter C. Ordeshook. 2011. ‘Benford’s Law and the detection of election fraud.’ Political Analysis 19: 245-268.
  • Hausmann Ricardo and Rigobon Roberto. 2011. ‘In search of the black swan: Analysis of the statistical evidence of electoral fraud in Venezuela.’ Statistical Science 26(4): 543-563 
  • Walter Mebane's website with several related papers:
  • Mikhail Myagkov, Peter C. Ordeshook and Dimitri Shakin. 2009. The Forensics of Election Fraud: Russia and Ukraine. New York: Cambridge University Press.
  • The Pew Center on the States. 2012. Electoral Administration by the Numbers: An Analysis of Available Datasets and How to Use Them. The Pew Charitable trust: Washington DC.
  • Center for Electoral Forensics with discussion blogs