As in many developing countries, in Sierra Leone the process of selecting candidates often caters to the elite and lacks direct input from ordinary voters, which raises questions about representation and accountability. In response, researchers designed an evaluation to measure the efficacy of an intervention that informs party executives about the qualifications and policy visions of “aspirants," or potential candidates, during the primary selection stage, as well as which aspirant the local voters would most like the party to select to be the candidate for their constituency. In a separate but related evaluation, researchers built on a successful pilot to explore the link between a poorly informed electorate and weak accountability pressure in the general election. Promising early stage results from both evaluation can be found below. Further results are forthcoming.

Policy Issue 

In developing countries, elected government officials often do a poor job of performing their core responsibilities of providing public goods, developing legislation, and representing their constituents at the national level. One challenge is that regional party strongholds often ensure that the locally dominant party’s candidate wins the general election, which weakens competition and reduces accountability pressure on elected officials. In such strongholds, competition between aspirants of the same party during the primary stage is likely an important source of selection and accountability pressure. The problem is that primary elections are often opaque, controlled by powerful elites, and exclude the average voter whose priorities may differ from elites. Could a more transparent and inclusive primary selection process, one that provides reliable information on aspirant qualifications and seeks systematic input from regular voters, lead to the election of more accountable political leaders?

Another concern is that accountability breaks down when voters have poor information about politics.  For example, data from the previous election shows that 64 percent of voters could not name a single job responsibility of elected Members of Parliament (MP) and only 3 percent had any idea of how much public money MPs were allocated in the annual constituency facilitation fund.  Evidence from an earlier pilot showed that political debates between MP candidates from rival parties were highly effective in educating voters, informing their vote choices, and increasing accountability pressure on elected MPs.  However, it is not clear whether such debates can be effectively delivered at scale.

Context of the Evaluation 

Sierra Leone’s Parliament is comprised of 132 constituencies, each of which elects one MP to represent local residents in the national government. Voting patterns tend to reflect historic relationships between ethnic groups and the two major political parties—the All People’s Congress (APC) and the Sierra Leone People’s Party (SLPP).

Previous research suggests that elected MPs underperform. For example, only 36 percent of discretionary public funds controlled by MPs could be verified as being spent on projects to develop their constituency.[1] Additionally, on average MPs made only four public statements during more than 50 sittings of Parliament, and held only one meeting with their constituents during their first year in office.

While the APC and SLPP differ in how they select candidates to compete in the general election, they are similar in that rank-and-file party members, and the voting public more generally, do not directly participate or formally vote in either party’s primary selection process.  So voters choose a party to represent them, and the party chooses a candidate to represent it: lack of information, poorly aligned interests, and incomplete contracting can create problems at both of these stages. 

Details of the Intervention 

Researchers evaluated two strategies designed to enhance political accountability. The first focused on how political parties select candidates, during the primary stage, and the second scales up an information campaign targeting voters in the general election.

In the first, researchers conducted a randomized evaluation to test the impact of greater citizen voice and informative party conventions on the selection and accountability of MP candidates. Both major political parties participated in the experiment, which covered 92 party-races randomly assigned to one of two groups.

  • Comparison: Status quo;
  • Party Convention and Voter Report: Aspiring candidates presented their qualifications and platforms as well as answered questions during constituency-level party conventions. The conventions featured town-hall debates that were broadcast on local radio. Researchers then asked a representative sample of registered voters which aspirant they would like the party to select to run in the general election, and shared this information with party leadership.

Researchers partnered with Search for Common Ground (SFCG), a trusted nonpartisan media presence in the country, with whom they partnered for another evaluation in 2012.   SFCG provided training and moderated debates between the aspirants. The conventions were successfully carried out in 43 of 46 constituencies.  The research team also conducted detailed interviews to gather information about the number and characteristics of aspirants and selected candidates.

Second, building on a previous study, researchers evaluated the impact of disseminating video tapes of debates between candidates from rival political parties, testing whether the strategy could be implemented cost-effectively at a much larger scale. This subsequent experiment asks whether debates can be effective in educating voters, informing their vote choices, and increasing accountability pressure on elected parliamentarians when taken to scale. Researchers also explored the impact of incentives on encouraging candidates to participate in debates, the role of private sector cinema halls in disseminating the debates, and voter willingness to pay for political information.

SFCG filmed and screened debates between MP candidates in 44 out of the 45 constituencies originally scheduled. Candidates from both major and emerging parties, as well as independent candidates, participated in the debates.

Results and Policy Lessons 

Preliminary Results

Evaluation 1: Party convention and voter reports

Overall, preliminary results suggest political parties in Sierra Leone were willing to experiment with more democratic selection methods to identify parliamentary candidates. The new process built greater political knowledge for both party leaders and voters, and substantially increased representation as measured by the preferred choice of local voters progressing to the general election. 

Selection into politics: Researchers found strong positive selection into politics on education and wealth: while 43 percent of voters have never been to school, 80 percent of aspirants have some university schooling.  And while only 11 percent of voters have a bank account, nearly all aspirants have one.  Party leaders are situated in between the two, which suggests that they might play a valuable role in helping voters screen aspirants on their technical qualifications.

Local information constraints: Prior to the debates, party leaders were surveyed about their own first choice aspirant and whom they thought would win a popular vote. About 90 percent of party leaders assumed that their preferences will coincide with that of voters. This was incorrect as only 55 percent of presumed shared preferences coincided.  Overall, leaders correctly guessed the local choice 52 percent of the time. In about 34 percent of races, not a single party leader surveyed correctly guessed the local choice. This clearly suggests there is room for polling to improve representation by alleviating information constraints.

Voter learning: Voters in constituencies where party conventions were held and broadcast over local radio had more political knowledge than voters in control “politics as usual” races.  They were more likely to know which aspirants were better educated, had more public office experience, and had been involved in more local development projects. This suggests that town hall meetings can be an effective tool for educating voters about the range of potential candidates they can choose from.

Where did parties choose to experiment? The political parties chose Parliamentary races from all around the country to participate in this experiment.  Overall, they tended to select more races from their respective regional strongholds, which constitute 52 percent of the experimental sample, and fewer from competitive areas (28 percent) or their weak holds (17 percent).  This suggests that political parties preferred experimenting in their strongholds.

Increased representation: The combination of party conventions and voter polling enhanced the likelihood that the person who local people wanted to represent them was selected to be the party’s candidate in the general election. This increased representation, as captured by this measure, from 37 percent to 61 percent, for a sizeable 65 percent increase.

Evaluation 2: Debates scale-up

Knowledge and behavior: Debates at scale remained effective in increasing voters’ political knowledge, but did not impact voting behavior.

Willingness-to-pay for information: Voters displayed high willingness to pay for political information: for example, roughly 30 percent of voters were willing to travel two miles (at a cost of roughly $1) to attend a debate screening, as measured by vouchers that were redeemed in person at local cinema halls.

Private sector participation: Cinema halls hold promise as a viable dissemination partner for political information: over 90 percent of cinemas that were given a free copy of a debate played it as part of their for-fee programming.

Candidate participation: Without incentives, candidates were largely unable or unwilling to coordinate with each other to participate in a debate. However, this was easy to shift with a relatively small change in incentives: a $200 guarantee for an hour of free airtime on a local radio, conditional on at least one candidate showing up, increased candidates participation in a live on air debate to 70 percent of races (as opposed to 10 percent without the incentive).

Data analysis is ongoing and further results are forthcoming for both evaluations.


[1] Bidwell, Kelly, Katherine Casey, and Rachel Glennerster. "Debates: Voting and Expenditure Responses to Political Communication." Stanford GSB Working Paper No. 3066, May 2018.