For democracies to function well, citizens need to hold leaders accountable, yet in many African nations demand for accountability appears low. This research tests the theory that because people are loss averse, taxation pushes citizens to demand more from their leaders. In games that mimicked real life choices, “taxed” participants were 13 percent more likely on average to punish leaders than those who were not taxed, providing evidence that taxation induces citizens to demand higher levels of accountability from government leaders.
Many African countries still suffer from poor governance and corruption despite substantial efforts over the last decades to build functioning, democratic institutions. In many countries citizens’ demand for governmental accountability appears low, even though government mismanagement and corruption is detrimental to the wellbeing of citizens. However, little evidence exists on what makes citizens demand more from their governments. This research examines the role of taxation in increasing citizens’ demands for accountability. In contrast to previous work on taxation and accountability, this research provides evidence on whether taxation drives individuals to demand accountability from their leaders for non-economic, behavioral reasons. If citizens expect taxation to produce an overall gain, such as in services, benefits, or good leadership, the theory is that taxed citizens will perceive corruption as a producing a bigger loss than untaxed citizens. In turn, due to loss aversion, taxed citizens may be more willing to engage in costly actions like protesting or voting against corrupt leaders than untaxed citizens.
While overall tax revenues in Uganda have increased in recent years, direct taxation of poor citizens has actually decreased, with tax cuts often timed to win support among key electoral groups. The result is that poor individuals pay few direct taxes, and while even basic public services are difficult to obtain, there is little concrete outrage about poor governance. A common view among local political leaders, civil society organizations, and academics is that tax cuts have removed a critical link between citizens and government making citizens more disengaged from the political process.
This study was conducted in three low-income neighborhoods of Kampala, Uganda in 2012 and 2013. Kampala has seen significant political mobilization in recent years and its urban citizens have higher exposure to formal taxes than rural citizens. The population in this urban setting was therefore ideal for evaluating the norms and expectations surrounding taxation and accountability.
To test the theory that taxed citizens are more willing to punish poor performance by government leaders than untaxed citizens, even without any economic incentives to do so, the researcher designed laboratory experiments in which a “citizen” had to decide whether to pay to punish the “leader” based on how the Leader allocated a group fund. By varying whether this group fund came from a tax on the citizens or in the form of an outside “grant,” the researcher was able to compare how participants behaved in the different contexts.
In the Tax Game, the citizen was given a “wage” of 1,000 Ugandan Shillings (UGX); half of this was then taxed, doubled, and given to the leader as the 1,000 UGX group fund. In the Grant Game, the citizen received wages of 500 UGX, and 1,000 UGX was given to the leader as a non-earned group fund. In both games the leader then decided how to allocate the 1,000 UGX and, after observing the leader’s decision, the citizen could fine the leader 400 UGX by paying a fee of 100 UGX.
The study was run using a sample of 371 respondents over 18 sessions. For each session, IPA recruited volunteers from different neighborhoods in Kampala. Each participant was randomly assigned to a role (leader or citizen) and anonymously paired with others for the games. Each participant played either the Tax or Grant game then took a short survey.
Researchers hypothesized that citizens would be more likely to punish leader in the Tax Game than the Grant Game. In particular, the minimum acceptable share of the group fund that citizens would accept without punishing was expected to be higher among taxed citizens.
A second set of games was used to differentiate the proposed loss-aversion mechanism from alternatives, such as a stricter fairness norm. The new games were identical except that a third-party observer, rather than the citizen, could decide to pay to fine the leader if she disapproved of the allocation; this observer was not taxed and had no stake in the group fund. If taxation increases punishment because a different societal fairness norm applies to tax funds, the third-party observer should be more likely to punish when he sees the citizen being taxed. If loss aversion is the mechanism, the effect of taxation should disappear, as the observer has not been taxed and suffered a loss.
Citizens were significantly more likely to punish the leader in the game when they were taxed. On average the minimum acceptable share of the group fund was 13 percent higher among taxed citizens overall, and 30 percent higher among adult, wage-earning men; the subgroup who has the most exposure to taxation in Uganda. The results from the second set of experiments provide evidence that the treatment effect is due to loss aversion, rather than stricter societal fairness norms: a third-party observer was no more likely to punish the leader when he saw a citizen being taxed.
Together, the results suggest that taxation may provide a self-sustaining way to increase citizens’ demands on government. It also suggests that donors should consider the role of formal taxation, as well as more informal community contributions, when designing development interventions. Adding community contributions to external aid programs could give beneficiaries more ownership over projects and, this project suggests, make them more likely to hold local leaders accountable for how development funds are spent. However, further research is needed to test the conditions under which such a program would work.