Impartial elections are an integral part of any functioning democracy. Transparent electoral processes can be undermined, however, by a number of corrupt practices, including ballot box stuffing, intimidation of voters, and fabrication of vote totals. These practices are especially prevalent in post-conflict states, which are plagued by instability and weak institutions. While efforts to improve oversight in such states are well intentioned, countries often lack the resources to accurately monitor the electoral process. Political elites may particularly benefit from this corruption, though little evidence exists on the subject. This study examines how politicians exploit connections to election officials in order to change vote totals, and how a simple photo monitoring protocol may mitigate such fraud.
In Afghanistan, democratic institutions are struggling to create transparent infrastructure after decades of occupation and conflict. The United States oversaw Afghanistan’s 2005 parliamentary elections and 2009 presidential elections, but despite this oversight, the elections were marked by fraud and corruption. Afghanistan is particularly vulnerable to fraud due to its new and weak democratic institutions, and polling stations in uncontrolled areas that have virtually no oversight. This complicated political structure has made fraud detection difficult. Many assume that politically powerful candidates benefit substantially from fraud. This study, conducted during the 2010 parliamentary elections, focuses on aggregation fraud (fraudulent tallying of votes) because it is especially indicative of collusion between the candidate and election officials. Researchers pilot a low-cost method of detecting fraud and evaluate its effectiveness in an unstable political climate.
To measure fraud, researchers collected information on vote aggregation forms. After votes were cast in September of 2010, polling stations tallied the votes and wrote the totals on an aggregation form. A copy of the form was posted at the polling center, while polling center officials sent the original to a Provincial Aggregation Center. Individual votes stayed at the local center, and there were no further checks of the initial vote total. Officials at the Provincial Center lastly sent forms to the National Aggregation Center. Officials handling forms without oversight could change vote totals.
Researchers first collected voting data from polling stations by photographing the Declaration of Results tally posted at individual polling stations. They then compared those to the numbers posted online by the National Aggregation Center in October (once all votes were tallied). To control for a series of variables, the authors conducted a baseline survey to households surrounding the 450 polling centers in the initial survey sample. (Researchers added 21 polling stations to the sample after receiving additional funding, though these centers do not have baseline survey data.)
To measure the effect of monitoring on election official behavior, researchers randomly selected 238 of the 471 stations to receive explanatory letters about the monitoring process, informing them that their station would be monitored. The research team recorded 347 polling station pre-aggregation results, 146 of which were from the test sample. While research staff visited all 471 polling centers, many centers had damaged forms posted, or no form at all.
In addition to data collected on the incidence of aggregation fraud, researchers also collected information about 57 candidates’ political power (connections to polling officials, relation to the Karzai regime, incumbency, etc.), in order to measure the effect of candidate connections on fraud.
Results and Policy Implications
Fraudulent vote tallies were found in a majority of observed polling centers. Researchers recorded differences between polling center aggregation forms and the National Center forms in 78 percent of the photographs. In addition, research staff encountered damaged or missing forms at 13 percent of the centers. The monitoring announcement did, however, reduce the incidence of damaged materials from 19 percent in the control sample to 8 percent in the treated sample.
Politically connected candidates received more fraudulent votes on average than other candidates. More importantly, announcing the monitoring reduced votes for politically powerful candidates by 25 percent. When stations were told about the vote count, connected candidates averaged a drop of 6.5 votes per intervention, while unconnected candidates did not experience any significant vote changes. Results were strongest for candidates who knew a provincial aggregator (see graph).
Aggregation fraud undermines election legitimacy. This fraud is particularly prevalent when there is a relationship between the candidate and election officials. The study finds, notwithstanding, that photographing tallies and comparing them across the aggregation process significantly diminishes fraud. The newly termed “photo quick count” is both accurate and inexpensive relative to other methods, and will only become more cost-effective and accessible as cell service increases. While it is an uphill battle to ensure fair elections in an unstable state, monitoring techniques help build the transparent processes required for democracy.