Results are in: DIV-funded study on election fraud in Afghanistan finds success, grabs attention of Qualcomm
University of California, San Diego (UCSD) researchers find a 60% reduction in the theft of election materials and a 25% drop in votes for the most well-connected candidates when polling stations are warned that they will be monitored. The researchers have shared their lessons within and outside USAID, including in a December 15 post on Foreign Policy’s Afpak Channel.
In 2010, Afghanistan held elections for its lower house of parliament that were marred by fraud: ultimately, over 1 million votes were invalidated and a resulting power struggle nearly destroyed the country’s political institutions. Of the 182 countries ranked on Transparency International’s 2011 Corruption Perceptions Index, Afghanistan is ranked 180th—placing it ahead of only North Korea and Somalia, which tie for last.
Systemic corruption in government institutions profoundly undermines any development effort in the areas under rule. “Efficient and resilient governance systems are essential for ensuring basic public services, fostering trade, attracting private investment, and managing aid flows,” writes the Center for Global Development. All development goals drive an imperative to find ways of making corruption more difficult and more expensive – and Afghanistan is one of the hardest environments in which to institute change.
In the Afghan election’s midst, two researchers from the University of California, San Diego used funding from USAID’s Development Innovation Ventures to measure how candidates and polling officials would react to the knowledge that their vote counts would be photographed. The results showed a 25% reduction of votes for the candidate most likely to influence the count, and a 60% reduction in the theft of vote tallies and other election materials. In the paper, Callen and Long write that the candidates found ways to circumvent the monitoring measure in predictable ways depending on the strength of their connections to election officials. The results inspired the use of the election monitoring tool in Ugandan elections, and Qualcomm is interested in implementing the approach in upcoming Arab Spring elections.
The experiment made use of the “gold standard” in analytical techniques
Callen and Long carried out a randomized control trial1, a method that uses treatment and control groups the same way the FDA tests pharmaceuticals, to compare the behavior of candidates and election officials who were warned about possible election monitoring (the “treatment” group) to those who received no warning (the “control” group):
- The pair designed a photo “quick count” system using digital cameras to snap photos of vote tallies from polling stations and compare them to the tallies used by provincial vote-aggregating centers. The tallies were meant to be carbon copies of one another: any deviations would indicate election fraud.
- Out of a sample of 471 polling centers, researchers sent letters to a randomly selected half, giving an official notice to the polling station manager on election day that observers would take photos of the final tallies. In reality, the team took photos of the tallies at all 471 stations.
- Callen and Long compared the photos taken of the “treatment” group who had received notice of the election monitoring and the corresponding results used at the provincial level to those who had received no warning and the corresponding provincial-level tallies
DIV-funded research turns up lessons learned
The results from the research left three major takeaways for practitioners interested in election fraud:
- New technology can help improve the fairness and transparency of elections: Digital and cell phone cameras can become a tool in the hands of citizens to combat fraud at election polls.
- Monitoring can make it difficult and expensive to cheat: Candidates reacted to the announcement of the monitoring mechanism by trying to commit fraud in other predictable ways depending on their influence over election officials. Targeting vote tallies is one of the easiest ways to commit fraud, and the researchers’ obstacle drove up the cost of cheating, but access to election officials remains cheaper than other cheating mechanisms such as ballot-stuffing. An eye towards the incentive structures of election officials could be one avenue for building an impenetrable system for cheaters.
- There is a need for rigorous analysis of democracy assistance tools: Callen and Long’s study contributes to a small but growing body of research using scientific methods to assess democracy and governance assistance strategies. The knowledge gained from this body of work can help drive efforts’ effectiveness up and costs down.
Researchers share lessons at USAID and beyond
At a visit to USAID headquarters in Washington DC on November 13-17th, Callen and Long met with staff from across the agency to discuss the results of their work, the applications to USAID programs, and possible avenues for further research and collaboration. The team will return again in January to meet with senior advisors, and to follow up on initial discussions to re-test their intervention with government assistance. Coverage on USAID internal and external communication channels has helped circulate the findings around the agency and out to the missions around the globe.
Media coverage of the results started with an early article from Slate magazine, a post on Foreign Policy’s Afpak Channel, and more will appear over the coming weeks. Callen and Long have presented their findings at economics conferences, and Qualcomm’s adoption of the photo “quick count” system helps achieve the Development Innovation Ventures’ objective to have widespread adoption of its successful investments. In the year since the grant was awarded, the DIV-funded research on election fraud in Afghanistan is already helping the agency accomplish its goal to test new solutions and scale proven successes.
 The link to the "Randomization" page on the Jameel-Poverty Action Lab (J-PAL) website offers one definition of randomization. Others include the 3ie Impact evaluation Glossary and the World Bank's Impact Evaluation in Practice document. See also USAID's Evaluation Policy, pgs. 4 and 18.