Brief 48: Strengthening Local Political Accountability Through Information in Uganda
EGAP researchers: Kristin Michelitch, Guy Grossman
Geographical region: Africa
Research question: How do transparency initiatives that are disseminated early on in an electoral term affect subsequent politician performance?
Preparer: Catlan Reardon
The study takes place in Uganda – an electoral authoritarian regime with a highly decentralized local government system. The National Resistance Movement (NRM) led by President Museveni maintains control of the national executive and legislature. Nevertheless, there is a fair amount of party variation at lower levels depending on the district. Local governments control the allocation of a large share of resources and public goods such as education and healthcare and can make local laws. Citizens elect representatives in first-past-the-post electoral constituencies to serve at the local government. Citizens possess limited information about their politician’s job duties, as well as performance in such duties, and base votes on handout distribution or other non-performance criteria (e.g. ethnicity ). As a result, politicians shirk their job duties and pander to citizens’ other criteria. Further, the electoral cycle is five years long and politicians are often considered absent outside of election season.
Since 2009, a local civil society organization called Advocates Coalition for Development and Environment (ACODE) has produced annual scorecards on politicians’ job duty performance, which are distributed to the incumbents themselves, district officials, and party representatives at meetings in the district headquarters. The performance information did not trickle down into the hands of citizens.
The authors set up a study to test the argument that if the performance information were additionally disseminated to citizens, politicians would improve their performance. The motivating theory revolves around the idea that if the performance information is disseminated early in the electoral term to citizens, politicians will fear the entry of challenger candidates and ultimately voter sanction. They would improve job duty performance to cater to the new criteria citizens have to hold them accountable. Further, the authors posit that the effect may be stronger in competitive constituencies over “safe seats” given that it takes time for opposition to mount successful campaigns in an electoral authoritarian regime.
The authors implement a field experiment to test whether politician performance information distributed to Ugandan citizens early in the electoral term improves their local politicians’ subsequent performance. The study takes place in 20 districts with around 400 local government politicians. The intensive dissemination intervention (“treatment condition”) involved distributing scorecard information – along with a range of general civic education information – directly to citizens at community-wide meetings. ACODE further sent periodic text messages to meeting attendees with scorecard information. District politicians were also invited to the community meetings and sent text messages whenever scorecard information was sent to their constituencies.
While all politicians had their scorecard disseminated yearly in the standard way at district headquarter meetings, the authors randomly allocated half the politicians to the intensive dissemination treatment. As such, the intensive dissemination intervention seeks to assess the effects of citizen access to performance information on incumbent performance above and beyond the information provided through ACODE’s standard accountability programs. Primary outcome measures were drawn from ACODE’s scorecard reports, original surveys conducted both pre-treatment and post-treatment with district bureaucrats and politicians, unannounced audits of schools and clinics, as well as administrative (budgetary) data.
The authors find that distributing politician performance scorecards to politicians’ constituents significantly improves subsequent performance in competitive constituencies, i.e. when politicians have a real fear of future electoral upset. This implies that electoral pressure is likely a necessary condition for transparency campaigns to shift politician behavior. The authors find that the treatment effect is primarily driven by an increase two types of job duties—in monitoring of service delivery as well as engagement with lower levels of government. In addition, they find that politicians in the treatment condition spend significantly more time in helping their constituent schools with applying for small grants—a novel, behavioral intervention implemented by the researchers in collaboration with the District Education Offices. Interestingly, the authors find that the effects kick in immediately in the middle of the electoral term, defying the conventional wisdom that politicians often wait until an election year to ramp up performance.
The authors also explore how the interventions influence development projects and service delivery in each constituency. They find that treated politicians implement more development projects than non-treated politicians. Again, this result holds only in competitive constituencies. Yet, they do not find a shift in the total development project spending or in tangible service delivery as measured through unannounced audits of health centers and schools. This result reveals that transparency campaigns (given a sufficient level of competition) can be highly effective, but only within domains over which politicians have direct control and not in areas that involve many different actors at different branches and levels of government.
The study implies the following for policy makers looking to conduct a successful transparency initiative to hold politicians accountable for job duty performance:
- The transparency initiative should start early in the electoral cycle, contrary to many other politician accountability initiatives that disseminate performance information to citizens directly prior to elections. Disseminating performance information early on motivates politicians to increase effort to deter challenger candidates’ entry and woo voters. If the initiative occurs directly too close to elections, there is no opportunity for politicians to improve performance and no opportunity for potential challenger candidates to identify poor performers and run for office to unseat them.
- The seat can’t be too safe. There needs to be at least some minimum level of pressure from a rival political party—in the study’s context (Uganda) at least a vote margin of 22 points, which is not a remarkably high threshold. The higher the level of competition in a constituency, however, the larger the effect of transparency.
- The civil society organization should be local, non-partisan, credible, and get buy-in from key stakeholders (e.g., politicians, local politician associations, and political parties). If the organization is seen to be partisan or slip-shod in the creation of a performance scorecard, politicians could derail the initiative. It is likely more feasible for civil society to introduce transparency initiatives at the subnational rather than national level. For the initiative to be sustainable and taken seriously by politicians, the local organization should not be seen as transient.
- Elected politicians can improve the performance of their job duties and outcomes that are under their direct control. Since, for example, public service delivery is a function of many government branches and levels, even when legislators improve job duty performance following a transparency initiative, public services may not improve in tandem. Increasing accountability of many actors may be necessary for improvements in public services to be seen.