Does election monitoring reduce fraud? International observers use monitoring as a core component of the effort to ensure fair elections in transitioning democracies and post-conflict countries. Yet scholars and policymakers know little about the actual impact and scale of monitoring given weak tests and flawed research designs. In an effort to overcome these challenges, this project employs more robust methodologies and tests new hypotheses to evaluate the effect of monitoring during the February 18, 2011, Presidential and Parliamentary elections in Uganda.

The project advances knowledge and understanding of election monitoring in several ways. By taking advantage of cellular phone technology and an application designed specifically for this project to facilitate the recording and transmission of vote tally results, the principal investigator and collaborators will randomize the presence or absence of monitoring across polling stations and district counting centers throughout Uganda. By creating a random sample of 800 locations, this study is the first to evaluate the impact of election monitoring at such a scientifically rigorous level. The PI will test whether or not this intervention reduces fraud in polling stations and district centers, whether incumbents have greater control over fraud than do challengers, and whether there are differences in fraud levels between polling stations and district centers. An important benefit to the scientific community is the provision of an experimental estimate of the causal effect of monitoring on election fraud.

This project makes several broader contributions. The study provides evidence to academics, policymakers, and citizens about the promotion of free and fair elections. The project should also make an important contribution to citizen welfare, in that robust evidence now indicates that enfranchising disadvantaged groups raises political responsiveness, whereas manipulating elections leads to popular resentment and often election-related violence. Moreover, because Uganda resembles many developing democracies along such key dimensions as institutional weakness, social diversity, and poverty, the results of this project will suggest how well its fraud monitoring technology is likely to work in other settings. If effective, the cellular technology used in the project would furnish a cost-effective alternative to conventional election monitoring. It could provide much greater coverage at multiple levels, and could also be easily adapted to other service delivery and corruption monitoring uses.

Project Report

This project detected and reduced electoral fraud during Uganda’s 2011 Presidential and Parliamentary Elections in 2011. Electoral fraud challenges many developing countries, including those that are close allies of the United States. Reducing electoral fraud and thus increasing political accountability has been the object of many U.S. funded programs, such as democracy and governance projects and election observer missions. Despite the efforts of these prior programs, none can be sure they have any impact, as they do not rigorously measure what they do or possible outcomes using a scientific research design. In contrast, we used a randomized controlled trial design – the gold standard of science – to detect and reduce fraud. We also used Ugandan citizens in our efforts to create a robust election monitoring system that can be duplicated in Uganda as well as elsewhere. We trained and deployed 600 Uganda citizens to 1000 polling stations on the day of the election. In one third of the polling stations, our Ugandans researchers delivered a letter - our treatment in the experiment -- to polling managers stating that the researcher would take a photo of the posted vote tally the next day, this we called the "Monitoring" letter. Another third of the sample received the same letter that similarly announced the monitoring of tallies and also reminded the manager of the penalties under Ugandan law for committing fraud (up to five years in prison), the "Punishment" letter. In the last third of the sample, the 'control' for the experiment, researchers did not deliver a letter. For both treated and untreated subsamples, researchers successfully took photographs using a smartphone with a customized application created by QUALCOMM Inc. of the posted tallies so that the treated and controlled stations could be compared. After the photos were taken, we transcribed the thousands of numerical entries captured in the hundreds of photos (one photo each for the presidential count, another for the parliamentary, for each station; hundreds of candidates). The project did detect and reduce fraud in significant ways. The "Monitoring" letter reduced the number of missing tallies by almost 11%. That is, some polling station managers did not even bother posting their results, which is required by law. The letter made 11% more managers post their tallies. The "Punishment" letter increased the number of posted tallies by 10%; when both messages where given, 12% more tallies were posted. Using adjacent digits as the fraud indicator (used to gauge if people just made up numbers for vote totals), "Monitoring" decreases adjacent digits by 8%, "Punishment" by 6%, and "Both" by over 7%. The letters also affected the incumbent President Museveni’s actual vote total, using only the sample for which we have photographs. All the treatment letters reduced the number of votes Museveni received as recorded in the tallies. The "Monitoring" letter reduced Museveni’s average of 307 votes at a polling center by about 40; the "Punishment letter by nearly 29, and "Both" by nearly 60 votes, a decrease of nearly 20%. Under all three treatments, however, the decrease in votes for Museveni is less than the effect found in photographed tallies. This result could be interpreted as an attempt by officials in higher echelons of the electoral administration to add votes back to Museveni’s counts that had been reduced by the treatment letters earlier in the aggregation process, although we have no additional evidence of this. Over the life of the project, the PI and his team helped create the application; created and field tested the application; constructed a random sample of the total number of Uganda polling stations; traveled to Uganda and trained over 600 Ugandan researchers in the use of randomized controlled trials in research, the use of smartphones, and the use of the application; deployed the Ugandans across the country, monitored and collected the resultant messages that were sent via the network to the PI’s server on election day; coded the resultant thousand of photos; analyzed the data collected; prepared scientific papers and delivered talks to policymakers, including at Qualcomm based groups, USAID staff (to design more effective electoral fraud reduction strategies), the Brookings Institution. The work was also covered in the press by the Economist and Miller-McCune. Overall, then, the intervention successfully reduced fraudulent electoral practices in Uganda, an ally of the US but known to have poorly administered elections. The treatment letters: 1) decreased the illegal practice of not publicly posting tallies, 2) reduced the number of sequential digits found on tally sheets, and 3) reduced the number of votes for the incumbent President Museveni, the candidate most likely to rig. The data also show that the decreases in polling center votes for Museveni caused by letters is systematically related to the inclusion of additional votes at the leve

National Science Foundation (NSF)
Division of Social and Economic Sciences (SES)
Standard Grant (Standard)
Application #
Program Officer
Brian D. Humes
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of California San Diego
La Jolla
United States
Zip Code