Domestic and international actors most frequently respond to threats of electoral fraud by sending observers to monitor and evaluate elections. However, little is known about the overall impact of these programs. Do observers actually deter fraud, or do political actors simply move fraud to other, unobserved polling stations?
Ghana has held regular presidential and parliamentary elections since 1992, but is still believed to experience widespread election fraud. Reports of suspiciously high turnout – even exceeding 100 percent – as well as other irregularities have led to legal challenges to nearly every presidential election since 1992. Accusations have ranged from ballot stuffing and record falsification to illegal voting by minors and foreign nationals (generally from neighboring Togo).
Since 2000, domestic election observers, led by the Coalition of Domestic Elections Observers (CODEO), have been assigned to polling places throughout the country. The assignment of an observer to a specific polling station is only known to the public when the polls open in the morning. CODEO’s trained observers wear official uniforms, have full access to the voting process, and take careful notes throughout the day of voting. They use an SMS system to transmit information to a central site about severe irregularities they may notice or any violence that may occur. The idea is that the presence of an observer will deter fraud, negligence, and violations, thereby promoting confidence in the electoral process.
The 2012 elections pitted two closely matched opponents, and pre-election polls had the election too close to call. The incentives to engage in fraud were high and the benefits from monitoring could have been substantial.
The researchers partnered with CODEO and assigned 1,000 of CODEO’s 4,000 election monitors in a randomized saturation experimental design. The researchers randomly assigned constituencies to receive observers at 30%, 50%, or 80% of their polling stations, and then randomly assigned the observers to the individual polling stations within the constituencies. This process – randomly assigning both the observers and the saturation rate – allows them to measure two types of effects: 1) whether fraud decreases at the observed polling places compared with control locations and 2) whether it increases at unobserved polling places in high saturation areas (compared to low saturation areas) because parties simply move their fraud elsewhere.
Although there are many forms of electoral fraud, the researchers focused on two of the most detectable and frequent indicators of fraud in Ghana –ballot stuffing and records of more votes cast than voters registered (labeled “overvoting”). Ballot stuffing can be identified because at the end of the polling day, the ballots are counted at each polling station in public. If there are more ballots than votes known to be cast, ballot stuffing occurred. Overvoting also uses objective data, and is measured as the number of persons allowed to vote compared with the number of voters registered at the polling station. Measuring fraud objectively gets around the possibly inconsistent reporting by observers and allows the researchers to collect data that easily detects fraud even at unobserved polling stations.
Using these techniques, the researchers measured both the direct and indirect effects of observers. The direct effect (lower fraud rates at observed polling places) appears to be quite substantial. Using a combined measure of overvoting and ballot stuffing, observed polling places were about 60% less likely to experience fraud than control stations.
The indirect effects from parties relocating fraud were smaller. Even in districts with many observers, where political parties have only a small number of unobserved polling places to commit fraud, the experiment observed only a modest (13%) increase in overvoting. Therefore, although political parties did relocate fraud in response to election observers, the experiment still suggests that electoral observers were a net positive, reducing fraud overall.
The caveat to this assessment is that political parties apparently relocated fraud much more effectively within areas where partisan political competition is low, areas that are commonly referred to as party “strongholds.” Whereas spillover effects were actually negative in competitive districts (observers deterred fraud even at unobserved polling stations), they were positive and very large in the strongholds of Ghana’s two major political parties. In these areas, the authors hypothesize that political agents are socially more deeply embedded and have better information and resources to move fraud to new locations.
Figure 1: Direct Effect of Observers on Electoral Fraud at Different Saturation Levels
Electoral fraud operates differently in constituencies with low levels of partisan competition (party strongholds) than it does in competitive constituencies. Political parties can more effectively relocate fraud within their strongholds, which means that electoral observers will have more impact in preventing spillover when deployed to strongholds.
Election observers are an effective way of reducing fraud, even in the long-term. Although parties in Ghana have had years to find ways of getting around election observers, CODEO’s monitors still reduce fraud overall.
However, electoral observers need to be concerned about spillover as well. There is clear evidence of spillovers to nearby polling places. Innovations that seek to reduce this effect, particularly in less competitive constituencies, could make more effective use of resources directed at election monitoring.