“1, 2, 3...Vote”: Designing Voting Games in Uganda

MIT GOV/LAB and Twaweza interviewed 1,200 Ugandans after national elections to understand how citizens conceptualize politics and make voting decisions.

(Header: Ugandans voting on February 19, 2016. Photo Credit: AFP).

This post was also featured on Twaweza's website.


Loaded down with 75 kilograms (165 pounds) of coins, and 20 stacks of hypothetical candidate profile cards and 10 Tupperware containers masquerading as ballot boxes, MIT GOV/LAB and Twaweza set off to study citizen engagement and voting in Uganda. Packing for a field study involving interactive voting games, pictorial supplements to our survey, and financial incentives with over 1,000 respondents is no small feat.

Over the course of three weeks, MIT GOV/LAB and Twaweza interviewed 1,200 citizens throughout the northern and central regions of Uganda. The survey took place after the national election to understand how citizens conceptualize politics and make voting decisions in the context of a national election. We designed the survey to answer several theoretical questions about political behavior, citizen engagement, and public opinion.

To investigate how citizens engage with the government during elections, we aimed to replicate the real conditions under which voters participate in elections, particularly the costs of voting and the role of social influence in determining vote choice. In order to simulate the types of decisions voters face in the ballot box, we used conjoint analysis, a survey method originally developed by market researchers to examine how consumers make purchasing decisions between competing products.

Designing Voting Games

Adapted from similar research in Tanzania, implemented during their own presidential elections in 2015, we developed a survey game in which respondents were asked to decide between two hypothetical candidates running for Parliament in their constituency. Survey enumerators would provide citizens with a profile of two candidates that was comprised of six aspects of a candidate’s profile. Some aspects focused on the candidate’s identity (ethnic identity, religion, and political party) while others were focused on their performance in office. Based on these profiles, respondents were asked which candidate they would vote for if these were the candidates running in an election in their community.


Figure 1: Pictorial representations of candidate profiles. Sample candidate profile, similar to the ones used in the field.

For more details about the motivation and logistics behind this research design, please see Leah Rosenzweig’s guest post here. And, for new tools new tools to implement conjoint analysis in developing country contexts, click here.

In real life, citizens must make costly decisions if they would like to vote, including taking time off work, sometimes waiting in line for hours, and traveling to the polling station. To replicate the cost of voting, we asked respondents to pay to vote in our conjoint game. For each round of conjoint decisions, respondents had to pay 500 shillings (about 15 cents in US dollars) if they wanted to cast a vote. The payments were meant to tangibly capture the cost of voting.

In addition to understanding which candidate attributes voters prioritize in the ballot box, we also wanted to better understand the role of social influence in voting. In other words, how do community norms and social pressures from peers impact individual voting behavior. If your peers are watching how you vote, do you make different electoral choices? To accomplish this, we randomly assigned respondents to one of three groups. In the first group —the control group— respondents completed the conjoint survey game in private. Respondents filled out a secret ballot and placed the ballot in a ballot box. Their voting decision was kept secret, even from the enumerators.

In the other two groups, respondents played the voting game in public. In each group, we randomly selected four community members and one community leader, either a teacher or an LC1, or Local Council 1, official (similar to a village chief). As teachers at government schools, these community members were both public figures and government representatives, yet were unelected. LC1s , on the other hand, are elected officials. Seeing if community members responded differently to these two types of leaders could shed insight on how citizens engage with elected and unelected government officials.

Group members then discussed the hypothetical candidates together and voted in front of their peers and a community leader. In all versions of the conjoint game, respondents not only had the option of voting for Candidate A or Candidate B, but also had the option of abstaining from voting. Voting for either Candidate A or B was costly, yet abstaining was free, as it would be in real life.

Community Influence in Voting

Comparing voting patterns and abstention rates between private respondents, the “teacher group” respondents, and the “LC1 group” respondents will allow us to better understand the role of social influence in voting. While many surveys examine voting decisions and political behavior in isolation, this design simulates the inherently social nature of politics. With this method, we will also be able to understand how particular actors shape political behavior in unique ways. Does the presence of an elected official, such as an LC1, influence how people discuss and engage with politics differently from the presence of a teacher, who is less engaged with politics, but serves as an informal community advisor? Stay tuned for the answer to this, and our other research questions in our next post on our conjoint analysis research in Uganda.

Blair Read is a first year PhD student in the MIT Political Science Department. Previously she was a GOV/LAB field manager and research support associate based in Uganda and Tanzania. She can be reached at bmread@mit.edu.