International election monitors have observed a number of high-profile, highly suspect victories over the past few years: Hamid Karzai’s in 2009, Alexander Lukashenko’s in 2010, Vladimir Putin’s earlier this month. Observers have flagged ballot inconsistencies and missing vote totals and voter intimidation. But hardly anyone seems deterred.
“It’s very hard to conduct free and fair elections in a new democracy, even in old democracies,” said James Long, a doctoral candidate at the University of California, San Diego. “This is kind of curious because for 20 years, elections have been observed by international and domestic monitors. You’ve probably seen Jimmy Carter, NDI, IRI, the U.S., and the EU — they have spent a lot of time and energy trying to get at this problem.”
So why does it seem not much has come of it?
Election monitoring hasn’t changed significantly over these 20 years. International observers fly into a country. They drive around on Election Day, hitting as many polling places as they can. Then they reconvene, assess the election, and declare it “free and fair” or not. The whole process is a meaningful show of international support for democratic elections in countries that are still figuring out how to hold them. And while observers aren’t tasked with intervening — literally policing and preventing chicanery — it’s thought that their mere presence encourages better behavior.
But these missions don’t collect or produce any of the kind of data that might satisfy a social scientist — or that could put hard numbers to the prevalence of fraud and their own roles in reducing it.
“This is all based on observational data,” said Clark Gibson, Long’s adviser, and the chair of UCSD’s political science department. “We don’t know actually if any of this reduces fraud. We think it does. The conventional wisdom is that this stuff detects fraud and supposedly reduces it. But we have no measurement of this. On a scientific basis, if you want to do science, it fails on really almost all criteria.”
Clark then starts to list these criteria: The sample size in election monitoring is usually small. Missions only have so many monitors to dispatch. And they’re limited by the geography and safety concerns of a country. As a result, the sample isn’t nationally representative or truly random.
“The next issue for science is: What’s the intervention?” Gibson said. “If this is a drug trial — which is what we’re basically trying to copy in our work — you have to have a clear measure of the injection of fraud-reduction juice.”
OK, that’s a metaphor. But it’s not entirely clear what election monitors are doing — what type of cure they’re applying — to reduce fraud. Could their mere presence scare corrupt poll captains into respect for democracy? Or are they doing something more specific? Some monitors linger in a polling station for five minutes, some for an hour. Some speak to local officials, others don’t. The intervention itself is inconsistent. And how does anyone know that seemingly upstanding poll workers don’t revert to shenanigans the moment monitors leave?
“We thought about that and said, ‘Can science help us do better?’” Gibson asked. “Can we actually reduce fraud? Can we develop a tool to reduce fraud better that’s really rigorous, and use the best social science?’”
• • • • • • • • • • • • • • •
Gibson, Long, and his co-investigator Michael Callen began thinking about this near the same time the United States Agency for International Development began mulling a related question. Election monitoring isn’t just hard to measure; it’s relatively expensive. In an era of shrinking budgets (when most Americans say they want to cut first from foreign aid), USAID has been trying to figure out how to solve intractable international problems with a lighter, and less costly, touch.
In 2010, USAID launched a grant program to seed new approaches to old challenges, like maternal health and election fraud. Development Innovation Ventures was seen as a sort of DARPA for foreign aid. Its latest grant recipients are deploying cheap, pre-fabricated toilets in Kenyan slums, solar micro-grids in rural India, and low-cost surgical materials in Africa to halt post-partum hemorrhaging.
“A couple [of ideas] are aimed at big, hulking problems that we’re struggling with as a world to do something about,” said Maura O’Neill, USAID’s chief innovation officer. “And we’re looking for dramatic reductions in price — not just incremental [ones].”
O’Neill spoke about this Wednesday in Washington at the Center for Strategic and International Studies. Long was also on hand to present the results of what looks to be one of the program’s first scalable successes — a more effective, and cheaper, election-monitoring system tested during 2010 parliamentary elections in Afghanistan.
He and Cullen first had envisioned something grand, armies of monitors blanketing the country. But academic-scale money wouldn’t pay for that; “How do you improve an election in Afghanistan with $100,000?” Long asked. “We got smaller and smaller and smaller until it became a five-minute visit and a photograph with a digital camera.”
That is, essentially, their main intervention.
In describing what the researchers came up with, Gibson refers back to the not-so-scientific method of the traditional international monitoring mission. “So,” he said by phone from California, “we do all the opposite of what they do.”
They first hired local researchers in Afghanistan, which eliminated the perceived influence of international observers, and deployed them in a random sample of 471 polling stations across the country. The project was able to hit 19 of 34 Afghanistan provinces (compared to about nine for the international observers, among whom the largest mission had a $10 million budget).
Callen and Long were essentially designing the kind of randomized control trial you might see on a new drug line, but with Afghan polling stations: some were controls, and others received thefraud-reduction juice, if you will. At the end of the day, the Afghan monitors showed up and photographed the vote tallies posted at every polling station. Half of the stations, however, had received a letter warning them ahead of time that an observer would turn up to do this.
“One of the easiest ways to steal an election is just to alter the tally forms where all the votes are aggregated,” Long said. Of course, you can buy votes and steal ballots. “But you can also just take the tally that comes from a constituency, just erase people’s names, and add a bunch of zeros to a vote total. It’s just a very cheap, easy way to impact vote totals.”
In comparing the vote tallies at the end of the day between the control group and the polling stations that had been warned they would be monitored, the researchers found an impressive difference: That one letter reduced the vote counts for the powerfully connected candidates most likely to rig an election by 25 percent, compared to similar candidates at the other polling stations.
The letter also reduced the theft of election materials by about 60 percent.
“If we reduced it by 3 percent, with this design — and we really knew it was statistically significant with this design — we’d be very happy,” Gibson said. “But 25 percent and 60 percent are ridiculously high.”
• • • • • • • • • • • • • • •
The researchers then took the idea to Uganda, during the presidential election there last year (this part of the work was funded by the National Science Foundation). In Uganda, instead of digital cameras, the project used smart phones. San Diego-based Qualcomm helped the university design an app allowing monitors in the field to photograph vote tallies and automatically send them for analysis to the university server in San Diego. (Qualcomm also supplied the phones — several hundred of them — and Gibson pauses here to note with pride that the local monitors in Uganda returned every one of them.)
The researchers made one other adjustment in Uganda. They sent two letters: one warning of the pending photograph, and the other adding a more-menacing reminder about the local penalty for election fraud.
“But it was not from international people,” Gibson said. “It was, ‘This is the law, and it’s your law!’ This is what you know already, we’re just reminding you of the law.”
This time, the project hit 1,000 polling stations in a truly nationwide sample. And the results were equally promising, again finding a preliminary but statistically significant reduction in fraud.
“The grand vision is developing these tools so that citizens can take ownership of it, do it, and the whole point is to demand legal, fair elections from their own government,” Gibson said. “And we’ll get out of the business.”
This is exactly what USAID wants. The ventures program isn’t just looking for cheap, scalable ideas. “We want to start with the exit strategy at the beginning,” O’Neill said. How can the agency foster these ideas, and then ultimately hand them off? Who else can manage them? Local citizens? Local governments? Private-sector entrepreneurs? Gibson and Long are researchers, not election monitors. Ultimately, they envision civil-society groups utilizing these tools with open-source software (next they’ll head to another country to further test the concept, although Gibson demurs on exactly where that will be — “we don’t want to be gamed,” he said).
Development Innovation Ventures’ vision sounds almost too good to be true: suddenly, the U.S. is solving massive problems on a tiny budget, and then handing the solution to someone else to manage. This electoral-fraud research also invites some predictable criticism. The findings suggest the existing system of international elections monitoring might be phased out in favor of something as simple as a brief letter and an iPhone photo. Plenty of people don’t want to hear that, including within USAID.
“At first, a number of people said, ‘Oh that’s just taking a photograph, we took a photo 15 years ago, there’s nothing new about that,” O’Neill said. “Other people said, ‘This is just a piece of election fraud, it’s not all of election fraud. It isn’t a comprehensive view of solving this problem.’ Those are natural reactions to disruptive technologies.”
But O’Neil counters that this isn’t just a photograph. The real innovation here is the scientific design.
Gibson acknowledges that this idea is a finger in the dyke of election fraud. The project is increasing the fairness of elections, not abolishing all the forces that make them unfair.
“You make cheating more expensive, which is what we want,” he said. “Some people say, ‘They’ll just do something else [to cheat].’ Yeah, that’s the point. You want cheating to be harder and harder. We had our own Tammany Halls and Chicago Machines. The 19th century was riddled with cheating. But you want to make it more expensive, more and more difficult.”