The Dismal Science

A Picture of Democracy

How digital cameras and smartphones might reduce corruption in Afghanistan and elsewhere.

Will digital photo monitoring fix electoral corruption in developing countries?

Photograph by Behrouz Mehri/AFP/Getty Images.

With the deadline approaching for troop withdrawals from Iraq and Afghanistan, the U.S. is facing the unhappy prospect of handing off military and political control of both countries to thoroughly corrupt regimes. Of the 178 countries in Transparency International’s Corruption Perception Index, Iraq comes in at 175; Afghanistan is tied with Burma for 176th. The prognosis for a transition to honest and accountable government is pretty grim.

At the same time, some innovative experiments in the democratic process provide at least a glimmer of hope for nudging these countries toward more inclusive government. A study just released by Michael Callen and James Long, a pair of ambitious doctoral students at the University of California-San Diego, exemplifies these efforts. The two researchers evaluated whether low-end digital cameras could be a cheap and easy technological fix for election fraud in Afghanistan’s 2010 parliamentary elections. Their results are promising. At polling stations where locally reported vote counts were digitally photographed, reports of electoral fraud were as much as 60 percent lower, and the vote counts of politically connected candidates—the ones most likely to have rigged elections—were reduced by about one-quarter.

Callen and Long’s study is part of a larger effort to figure out how to tweak the political process to improve the functioning of democracy. Figuring out what does and doesn’t work well to ensure free and fair elections is a thorny problem. Clearly, you can’t just compare measures of electoral irregularities in countries that utilize electronic ballots to those that use pen and paper. Expensive, computerized voting systems are more common in richer countries, and countries that lack a commitment to the democratic process may be unwilling to invest in advanced election technologies, regardless of their economic means. So many political researchers have joined a movement within the social sciences to try to understand what works in electoral politics by running randomized trials in India, Benin, and elsewhere.

If computerized balloting and transmission of results would be too expensive for a country like Afghanistan, mobile-phone technology holds much greater promise. (Besides, given the amount of corruption, any fancy election equipment would just get stolen anyway.) For improving election accountability, authors Callen and Long took as inspiration a study of teacher absenteeism in India, which showed that simply having a student photograph his teacher each morning sitting with the other students cut teacher absences in half and boosted students’ test scores.

For their election study, Callen and Long applied the same camera-audit approach to crack down on a common form of election fraud. In Afghanistan, votes are counted and recorded locally, and the results posted at each polling station. To determine the election winners, these ballots and results are sent to a provincial station for aggregation, and finally on to the national election center in Kabul. Somewhere between the local polling stations and the election headquarters in Kabul, the numbers can mysteriously shift in favor of a candidate who has, for example, bribed a provincial official to rig the numbers in his favor.

Working with a USAID Development Innovation Venture grant, and partnering with Democracy International, a fair election NGO on the ground in Afghanistan, Callen and Long hired a team of Afghans to travel to 471 randomly selected polling stations scattered throughout the country and act as election monitors. As votes were being cast on Election Day, the monitors in 238 of these voting stations delivered letters to polling managers, informing them that posted results would be photographed upon announcement and then compared to the official vote counts announced in Kabul. The following day, the monitors ventured forth in all 471 locales, cameras in hand, to take digital photos of local election results. After that, each monitor canvassed community members for any reports that election materials had been stolen or tampered with. By comparing voting patterns and reports of tampering in the 238 communities where election officials had been informed of the camera audit, to those where no such forewarning occurred, the researchers were able to infer the impact on a couple of methods of election fraud.

Citizens reported rumors that results had been manipulated or election materials stolen in nearly 19 percent of “control” stations where no warning letters were delivered. There were such reports in only 7 percent of the 238 “treatment” stations. The impact was even stronger in communities where leading candidates had personal ties to provincial election officers, i.e., those in which changing results or substituting false ones would have been easiest.

Of course, with one avenue of electoral fraud shut down, candidates determined to circumvent the democratic process turned to other forms of cheating. Indeed, complaints of voting irregularities, such as candidates claiming that their supporters’ votes were somehow unrecorded or invalidated, were higher at polling stations where station managers knew they’d get camera audits.

But evidently it’s harder to stuff the ballot box than it is to change the results in transit to Kabul. Overall, the camera intervention reduced the vote shares of candidates whom the researchers identified as likely to have engaged in fraud—those who, for example, were connected to Afghan President Hamid Karzai through relatives or past work experience—by about 25 percent.

Why were cameras so effective at keeping candidates honest, where so often monitors armed with clipboards had failed before? It may be the discipline that it imposes on the monitors themselves—it’s harder to doctor a photo than it is to fudge the numbers. The presence of digital cameras also provides the threat of immediate feedback, since images can be broadcast instantly to monitors in Kabul. And it doesn’t seem to be simple beginner’s luck with the first trial in Afghanistan—in the past year, Callen and Long have run another pilot during the national election in Uganda last February, to similar effect, this time using smartphones to capture the images.

But the researchers view these successes as mere proof of concept for a much more ambitious effort. They see camera and cell-phone technology as enabling crowd-sourced election monitoring, where any local resident could photograph and send in images of results to election authorities. This would do away with the cost of official monitors and also provide a further safeguard against doctored photographs, since election auditors in the capital would potentially receive many images taken independently of the same display of results, and collected in real time to reveal problems and irregularities as they occur.

Such an approach has already shown some promise using SMS text messaging, and Callen and Long are working with Qualcomm to develop the software that would be required to make this sort of citizen monitoring work for anyone with a functioning Android cell phone.

Before we get too excited about smartphones as the savior of democracy, it’s worth remembering that corrupt candidates were able to deploy alternative tactics even on very short notice, in the hours after the monitors delivered their warning letters. If something like Callen and Long’s procedure became the norm, powerful candidates might simply persuade the government to prevent the photographing of results, or to stop posting them at all. In fact, at the end of the day, democracy advocates may worry that photo audits may simply shift power from small-time candidates to big-time ones that are best able to find alternative forms of vote manipulation. We’ll find out when electronic monitoring makes its next appearance at Afghan polling stations.

Callen and Long’s grander hope is that even well-connected politicians might be kept in check by a more ambitious form of crowd-sourced monitoring—and that, inspired by their efforts, local software developers will use their app or variants of it to stay one step ahead of the corrupt officials. It’s of course ridiculous to expect that this simple fix, or any other, will springboard Afghanistan to the top ranks of Transparency’s corruption index. But regime change does bring hope for gradual change. For example, Indonesia was at or near the bottom rungs before the Suharto regime fell in that country. Its ranking in 2010 was 110, putting it squarely in the middle of the pack after just a decade of reform. It’s the sort of citizen-empowering innovations like smartphone election monitoring that can give Afghans similar hope for the future.