Point and shoot elections

Members of the international community met this past week in Bonn, Germany to discuss Afghanistan’s future in the shadow of a NATO withdrawal oftroops. At the conference, key policymakers, from the United States to Afghan PresidentHamid Karzai, expressed the consensus that corruption is one of the biggest stumbling blocks to efforts at rebuilding and stabilizingthe country.

In a time of belt-tightening in aid budgets in the United States and Europe, andweariness at a lack of significant progress in Afghanistan’s corruption outlook,donors may prove less and less willing to provide development assistance thatis then lost to graft. Similar to post-conflict and poor countries elsewhere,Afghanistan’s government agencies lack accountability. Service delivery can beseverely compromised because of graft, in turn fueling mistrust of thegovernment.

Another example of malfeasance is the widespreadelection fraud perpetrated during the 2010 election for the Wolesi Jirga,Afghanistan’s lower house of parliament. The voting itself and the subsequent dubious adjudication process provide a stark illustration of howcorruption can destabilize political institutions. The Afghan Electoral ComplaintsCommittee (ECC) had to delay the induction of parliament after adjudicatingnearly 6,000 allegations of malfeasance. Nine members have lost their seats evenafter serving for nearly a year.

With an eye towards understanding howinstitutions like the Wolesi Jirga could be strengthened through cleanerelections, we developed and evaluated a new approach to policing electoralcorruption for the 2010 races. It involves the implementation of a photo "quickcount" of election results. Specifically, we took photographs of tally sheetsfrom polling stations right after voting concluded, and compared them to whatshould be carbon copies of tallies from the same polling stations later in theaggregation process. We then took differences in results for specificcandidates as evidence of rigging. We implemented our project with funding fromthe newly established Development Innovation Ventures (DIV) at the United States Agency forInternational Development (USAID), a unit that embodies the organization’srenewed enthusiasm for improving development through rigorousevaluation procedures.We partnered with Democracy International (DI), the largest internationalorganization monitoring elections in Afghanistan.

To evaluate the effectiveness of thistechnology, we randomly announced monitoring in about half of a sample of 471polling centers. This sample spanned 19 of the 34 provincial centers in allregions of Afghanistan and was drawn from a universe of 5,897 polling centersscheduled to open on election day. We deployed a team of Afghan researchersthat delivered letters to polling center managers during voting on election day,announcing that the team would return the following day to photograph thetallies; teams visited the other polling centers without providing any priorwarning.

Through a comparison of these"treatment" polling centers and the "control" centers that were unaware thatour researchers would photograph results, we found that our program worked insignificant ways to decrease electoral corruption. Specifically, the monitoringprogram reduced vote counts by 25 percent for the candidate our team deemed mostlikely to rig the vote (generally the candidate with strongest links toofficials in the Election Commission or President Karzai, and those with ahistory of working in the government) and reduced the theft of vote tallies andother election materials by about 60 percent. In the study,we also found that candidates react to undermine the effort, and that they doso in a way that is predictable based on their connections to officials in theelection commission.  Specifically,candidates with a connection to the Provincial Elections Officer moved theirfraudulent activity in the direction of manipulating the returns form inpolling centers that did not receive a letter. By contrast, candidates lackingthis connection committed fraud by altering the count before the form wasposted.

We assessed the effect of the programusing a Randomized Control Trial (RCT), the most robust form of programevaluation. In a RCT, researchers estimate the effect of a program on keyoutcomes of interest (in our case, election fraud) by first identifying apopulation of potential beneficiaries and then randomly assigning the programto a subset (usually half). The half receiving the program are "treatments" andthe remaining half are "controls."  Themethod is therefore a straightforward adaptation of the approach used inmedical drug trials, only applied to questions of governance and institutions.A comparison of outcomes in the "treatments" and "controls" metes outeverything else that was going on in parallel with the program. For example,because we randomly assigned "treatments," we did not need to worry aboutwhether international monitors might be creating the change that we attributedto photo quick count. Additionally, one might worry that the effect we documentedis due to a selection of polling centers where fraud was less likely. But oneof the core strengths of RCTs is the ability to remove such a "selection bias"from our estimates of program effect. Because polling centers were selected bya random number generator, we can summarily rule out this concern.

We draw three important lessons from ourstudy. First, these results provide a convincing proof of concept that theapplication of new technologies can improve the fairness of elections and helpbattle corruption. In Afghanistan, we implemented the program using simpledigital cameras. In February of 2011 we replicated the experiment in Ugandausing smart phones and an application developed by Qualcomm to similar effect. Ultimately, webelieve this approach can be implemented via crowd-sourcing (essentiallyencouraging average people to document the process, as cell phones and evensmart phones become more accessible in the developing world), which woulddramatically reduce costs and increase coverage as citizens mobilize to policeelections.  

Second, while corrupt candidates surely willdevelop their own innovations to undermine fair electoral processes, making theaggregation process impermeable will greatly increase the difficulty of theirtask. If the election returns form posted at the polling center must match thereturns form that enters the official count in the capital, a major avenue offraud is shut off to candidates. More generally, we need to worry more aboutconnections between candidates and officials at the lower and middle echelonsof election commissions. Such officials can use their position and influenceover the aggregation and vote-counting process to dramatic effect. Reflectingthis, known affiliates of candidates should not be allowed to staff thecommission. Similarly, punishments for using such positions to favor a givencandidate should be serious, and these officials should be monitored. While a variety of evidence demonstrates corruption in Afghanistan’selectoral commission, the country is not unique in this regard — mostdemocratizing countries fail to establish truly independent election managementbodies and suffer fraud as a result.

Last, and most importantly, we only havescientific evidence of the effectiveness of a small numberof democracy assistance strategies. This is an area ripefor experimentation,which we encourage the international policy community to take seriously becauseof its clear importance for stability and welfare in fragile states likeAfghanistan. While clean elections will not solve all of the country’s problems,helping to reduce corruption and strengthen confidence in institutions like theWolesi Jirga will pay important dividends as foreign donors exert less and lessinfluence over Afghanistan’s future, and Afghanistan must take moreresponsibility for its own future.

MichaelCallen is a post-doctoralresearcher at the Institute on Global Conflict andCooperation at the University of California, San Diego.  James Long is a doctoral candidate in political scienceat the Department of Political Science, University of California, San Diego. Mohammad Isaqzadeh is an assistant professor of politicalscience at the American University Afghanistan, and provided researchassistance for the study.