PRVT Methodology

PRVT Methodology

Introduction 

National Election Watch’s (NEW’s) core mandate is to observe all public elections and provide to Sierra Leonean citizens, an independent and objective assessment of the processes and outcomes. For Sierra Leone’s 24th June 2023 elections, NEW employed the Process and Results Verification for Transparency (PRVT), an advanced and proven election monitoring methodology that allowed the coalition to independently and impartially assess the quality of voting and counting on election day, as well as the accuracy of official results of the Presidential elections announced by the Electoral Commission for Sierra Leone (ECSL).  

The PRVT methodology is routinely used by citizen election observation organizations around the world. More than 35 PRVTs have been conducted for national elections in 10 African countries, including recent elections in Nigeria (2023), Kenya (2022), and Côte d’Ivoire (2020). This is not the first PRVT in Sierra Leone, NEW has employed the exact same methodology in the past, in 2007, 2012 and  for both the first and second round Presidential elections in 2018. 

The PRVT methodology is well known to EMBs in Sierra Leone and other political actors. In all previous instances when PRVT methodology has been deployed in Sierra Leone, including in 2018, the results of the PRVTs have been consistent with the announced elections results. In 2018 in particular, the NEW PRVT accurately projected the results of the first and second round of the presidential elections. These projections were well received by stakeholders, including political parties, the electoral commission and international community.

PRVT Methodology 

For the 2023 elections, NEW trained and deployed 6,000 observers across the country covering every polling centre. Of these observers, 750 were specially trained and deployed to a statistically representative sample of polling stations across all the 5 regions and 16 districts in Sierra Leone using the PRVT methodology. 

The PRVT methodology involves deploying observers to a random, statistically representative sample of polling stations. On election day, PRVT observers stayed at their  polling stations for the entire day, observing the setup and opening, voting, counting, and closing processes. Throughout the day, they transmitted their observations on the process and at the end of the day the official results for their assigned polling station, as documented on the Results Reconciliation Forms (RRF). 

Because observers are deployed to a statistically representative sample of polling stations, the PRVT provides independent and impartial projections of the official results. While it is only the ECSL that declares the outcome of elections, the PRVT is based on the official results as announced by ECSL polling officials. If the official results are consistent with the PRVT projections, as they were for both the first and second rounds of the presidential election in 2018, then it gives greater confidence that the results as announced by the ECSL reflect the ballots cast at polling stations. However, when the official results are inconsistent with the PRVT projections, as is the case for the 2023 presidential election, then it provides evidence that the results do not reflect the votes cast at polling stations. 

Representative Sample

For 2023, the PVRT employed a statistically representative random sample of 750 polling stations with a total 213,164 registered voters. The size of the PRVT sample was determined to ensure that it included sampled polling stations in all five regions and all 16 districts of the country. The size of the sample is similar to those used for PRVTs in other African countries. It is larger than the sample size for the PRVTs for the presidential elections in 2018 that verified the outcomes of those elections. In 2018 the PRVT sample size was 506 polling stations while for 2023 it was increased to 750 to provide additional precision and confidence. NEW increased the sample size for the 2023 in order to reduce the margin of error for PRVT projections. Increasing the sample size for a PRVT from 1500 to 2000, or 1000 to 6000, would not have a significant impact on the margin of error and may render rigorous observer recruitment, training and data verification challenging. However, increasing the sample size from 506 to 750 kept a reasonable number of observers while allowing for increased precision in PRVT projections. 

The quality of PRVT projection does NOT depend on the percentage of polling stations covered but on the manner in which the sample is drawn. Polling stations must be drawn at random and the overall sample must be geographically representative.  NEW’s sample adhered to both of these requirements. 

NEW drew its representative sample of 750 polling stations using interval sampling, a form of random statistical sampling. In order to ensure geographical representativeness, the PRVT sample was drawn using stratification. This means including in the sample of polling stations the same percentage of polling stations for every region and district as the percentage of polling stations in the complete list 11,832 polling stations. For example, there are a total of 1,106 polling stations in Bo District. These 1,106 polling stations represent 9.3% of all 11,832 polling stations in Sierra Leone. The NEW PRVT sample included 70 polling stations in Bo District, which is 9.3% of the total 750 polling stations in the sample. Since these percentages are approximately equal, the sample does not overrepresent or underrepresent Bo district in terms of assessing the quality of election day processes and projecting the results. As Table 1 below shows for every district the percentage of sampled polling stations closely matches the percentage of total polling stations and therefore the sample is geographically representative.

Table 1: Distribution of Polling Stations in 2023 NEW PRVT Sample by District

 

Once the number of polling stations per district were determined using stratification the polling stations within each district were selected entirely at random to prevent any possible bias.

PRVT Observer Recruitment and Training 

NEW recruited PRVT observers from its member organizations across the country. Seventeen verification teams drawn from NEW’s Strategic Management Committee (SMC) and individuals from the coalition’s wider civil society member organizations were sent to all districts across the country (Bonthe District divided to Mainland and Island) to physically locate the sample points/polling centres and verify the recruited observers. The verification teams assessed network coverage and accessibility of the sampled polling stations, and ensured that recruited PRVT observers were independent and nonpartisan, without any known preference for any political party or candidate, region, ethnic group or persons. Observers were required to reside near the sampled polling station where they observed, and needed to know how to read and write in English in order to understand key election documents and observation tools. 

A rigorous 2-day Training of Trainers workshop was conducted for the PRVT trainers at NEW’s Secretariat. Trainers were drawn from the cohort of NEW member organization leaders who made up the verification teams. PRVT observers underwent a rigorous full-day training on the NEW PRVT methodology, ECSL election day procedures, the observation checklist, and NEW’s reporting strategy, including how to send reports to the NEW Citizen Situation Room via coded SMS. Prior to election day, NEW carried out two national simulation exercises to test observers’ understanding of the PRVT checklist as well as NEW’s reporting systems. During the first simulation, observers were required to deploy to their assigned polling stations.

PRVT Data Collection

On election day, trained PRVT observers deployed to the statistical sample of polling stations, where they examined crucial aspects of voting and counting at assigned locations, reporting on strengths and irregularities that affected the quality of the election. After observing  the counting of ballots PRVT observers also recorded  the number of ballots cast and the votes for each candidate in the presidential election as announced by the polling officials and recorded by officials on the Reconciliation and Results Form (RRF #1). Of note, PRVT observers reported on whether or not party agents agree with the official results announced by the polling officials. PRVT observers did not count any ballots. Instead, they observed the process and reported on the process and the official results.

PRVTs use technology to be able to have an assessment of what is happening with the election in near-real time. At specific times during the day, observers sent their observation information to the NEW Situation Room via coded SMS. SMS reports were transmitted via an SMS gateway to a specialized database. NEW then carefully verified the reported observations, analyzed aggregate data, and distilled findings into observation statements providing accurate characterizations of elections, and confident verification of official election results. 

NEW carried out rigorous data verification procedures, using logical checks to identify and correct any inconsistencies in PRVT data submitted via SMS. NEW had three observers who reported data that was for more than one polling station, failed to report results data, or reported figures that were not possible, so all three were excluded from the results projection and results projections were announced based on 747 of 750 observers’ data. The integrity of the sample was not impacted by excluding these three observers’ data, and there is no way that their data could have changed the ultimate finding that no candidate reached the 55% threshold. NEW followed a similar practice in 2018, when the PRVT verified official results for both the first and second rounds of presidential elections. This is a common practice for citizen observer groups implementing PRVTs to exclude any data from their results projections that is illogical. As noted above, excluding several sample points does not negatively impact the integrity of the overall sample and PRVT projections remain accurate. 

Margin of Error

The margins of error for the PRVT’s projections are the window within which the true results for  all polling stations for the presidential election must fall. The official turnout rate,  vote shares for each candidate, or the rate of invalid ballots must fall within the PRVT’s margin of error for these results. The margin of error is not a measurement of quality and does not indicate the degree of “mistakes” in the data. Rather, it is a technical term used by statisticians. The margin of error is a measurement of precision – a smaller margin of error means a more precise estimate. The margin of error is determined by three key factors: the confidence interval ; the sample size (750 polling stations); and the degree of homogeneity or variance of the data received (i.e. the variance in vote shares between candidates or rather the differences in voting patterns in sampled polling stations).  NEW used a confidence interval of 95%, a standard confidence interval in social science. A lower confidence interval such as 90% would have a smaller margin of error, while a 99% confidence interval would have a larger margin of error. Even at a 99% confidence interval, the PRVT showed that no candidate would have reached the 55% constitutional threshold to avoid a runoff. 

 

Key PRVT Findings

In 2023, based on official results as recorded on the Reconciliation and Results Form for the presidential election (RRF #1)  at a statistically representative valid sample of 750 polling stations from across all five regions and 16 districts, the official final presidential results announced by the ECSL are inconsistent with PRVT data. The PRVT data clearly shows that no candidate met the threshold in the first round and that there should be a runoff election. 

For this election, the PRVT data demonstrates that the official results announced by ECSL are statistically impossible. 

The PRVT showed that:

  • Voter turnout was 77.3% with a margin of error of +/- 1.7% (between 75.6% and 79%) while ECSL’s official turnout is 4 percentage points higher than the upper limit of the margin of error at 83.0%.
  • Rejected, or invalid, ballots were 5.1% with a margin of error of +/- 0.6% (between 4.5% and 5.7%). 
  • The PRVT’s estimate is consistent with past elections. In 2018, rejected ballots for the first round were 5.2%, 4.7% in 2012, and 7.3% in 2007. 
  • ECSL’s rejected ballot rate of 0.4% represents a significant departure from the trend in Sierra Leonean elections over time.
  • The results announced for the top two candidates are statistically impossible and no candidate should have met the constitutional threshold of 55% to avoid a runoff. 
  • The candidate of the Sierra Leone People’s Party (SLPP) should receive 50.4% of the vote with a margin of error of +/-2.7% (between 47.7% and 53.1%) of the vote, while 
  • ECSL’s official result is 56.1%, three percentage points higher than the upper limit of the PRVT range.
  • The candidate of the All People’s Congress (APC) should receive 46.5% of the vote with a margin of error of +/- 2.7%  (between 43.8% and 49.2%) while ECSL’s official result is 41.2%, 2.6% percentage points lower than the  lower limit of the PRVT range. 

 

PRVT Communication

Throughout the 2023 electoral cycle, NEW  engaged the media, Election Management Bodies (EMBs), Political Parties, other civil society organizations and the public on its observation methodology and findings. NEW launched its 2023 election observation strategy on 19 May 2023 in a ceremony that included a wide range of election stakeholders and members of the international community. NEW invited the Chief Election Commissioner to launch the strategy, and he was represented by the Commissioner for the Eastern Region.  Ahead of election day, NEW hosted a breakfast event with media representatives on Friday, 23rd June to explain and answer questions about the PRVT, as well as to officially launch its PRVT Situation Room. NEW also held a series of press conferences – on 23rd June to release the pre-election statement, on 24th June to issue the mid-day statement, and on 25th June to issue the preliminary statement – to ensure the timely communication of observation findings with the public. Before, during, and after election day, NEW hosted visitors to its PRVT Situation Room, including representatives from the media, civil society, the diplomatic community, and international election observation missions. All statements can be found on NEW’s website.

 

National Election Watch (NEW) is a coalition of local and international organizations in Sierra Leone with a common objective of supporting credible and peaceful elections.