Alert Sign Dear reader, online ads enable us to deliver the journalism you value. Please support us by taking a moment to turn off Adblock on

Alert Sign Dear reader, please upgrade to the latest version of IE to have a better reading experience


Using analytics to identify voter fraud

Updated May 16, 2013

enter image description hereAs the vote count subsides, allegations of electoral fraud rise. Almost all major political parties in Pakistan have hurled allegations of voter fraud in the May 11 elections. Some activists have taken to the streets. They should, however, head to a computer lab to analyze precinct counts for voter fraud.

Statisticians have devised algorithms to identify certain types of voter fraud, such as ballot stuffing and manipulated counts. Other types of electoral fraud, such as harassment of candidates, voters, and poling staff do not necessarily require any statistical proof since such overt acts could be identified without the use of analytics.

Millions of Pakistanis have voted in the May 11 elections to show their support for the democratic process. Democracy, however, is not confined to casting votes. Dispute resolution through peaceful means is also a hallmark of the democratic process. Pakistanis may want to learn from others who have relied on scientific method, courts and the rule of law to resolve electoral disputes.

Recall the American presidential elections in 2000 in which President George W. Bush was declared victorious against the Democratic candidate, Al Gore. The election was decided arbitrarily by the US Supreme Court that halted the vote recount in Florida and upheld the decision by Florida’s Secretary of State, Katherine Harris, who declared Mr. Bush as the winner in Florida. The victory in Florida delivered Mr. Bush the presidency of the United States.

However, Mr. Gore was found to be the real winner of the US Presidential elections after the recount was completed months later. By that time, Mr. Gore had already conceded and Mr. Bush had begun his first tenure as the American president.

The important lesson to learn from the American example is that while Mr. Gore and Democrats believed that they were denied victory by the US Supreme Court, who decided to uphold a decision made by a partisan State official, Ms. Harris, the American people did not resort to destruction and violence. They accepted the Court’s decision even when most were not convinced by the verdict. Mr. Gore emerged as a statesman, winning a Nobel Prize for his efforts to protect the environment. Ms. Harris’s political career was later stalled by political scandals.

Some, not all, complaints of voter fraud in Pakistan can be evaluated by statistical methods. Instances of ballot stuffing and fake vote counts are rather easy to detect. This is possible because of human beings’ inability to come up with random numbers. Beber and Scacco in a recent paper have shown that numbers made up by human beings are likely not to follow a random order. They found that the last digit in a genuine vote count is likely to occur with equal frequency, which is not the case in forged numbers. Often made up numbers end with zero and thus forged vote counts have a higher frequency of zeros being the last digit. A study of the 2009 Afghan presidential elections by the US Naval Research Laboratory revealed that the vote counts for President Karzai reported zeros as the last digit more frequently, suggesting rounding by tabulators.

Beber and Scacco also suggested that the pairs of adjacent digits (e.g., 67 or 45) are more likely to occur in manipulated vote tallies than the pairs of distant numerals (e.g., 17 or 29). Other statistical rules, such as Benford’s law, could also be useful in detecting voter fraud.

The Election Commission of Pakistan should make every effort to earn the trust of the electorate. Transparency will certainly help the Commission in this regard. The Commission should therefore release all verified data at the polling booth (precinct) level for registered voters, discarded ballots, and votes polled by each candidate. This will enable political parties, academics, and NGOs to review results and test against voter fraud.