By EL
The National Elections Watch (NEW), which provided independent observation monitoring of all polling stations across the country) published their report yesterday (27 June 2023) in which they stated that no presidential candidate was able to score 55%. NEW claims to have used a statistical tool called Process Results Verification for Transparency (PRVT) to validate the results published by the Electoral Commission for Sierra Leone (ECSL).
The NEW PRVT estimated that Maada Bio scored about 50.3% while Samura Kamara scored about 46.5% which could have led to a rerun of the presidential elections between the two candidates. NEW gave a margin of error of 2.7% for both the SLPP and the APC presidential candidates (Julius Maada Bio and Samura Matthew Kamara, respectively) thereby stating a lower limit of 47.7% and an upper limit of 53.1% for Julius Maada Bio and a lower limit of 43.8% and an upper limit of 49.2% for Samura Kamara.
NEW reported that they sampled data from 747 polling stations out of 11,832 polling stations across the country, which is about 6.3% of the total polling stations countrywide. The PRVT projection was thus largely based on this 6.3% sample size and 2.7% margin of error.
As a professional with several years of experience in complex and multi-dimensional data analyses and considering additional significant developments that took place ahead of the recent elections, let me point out possible areas where the NEW PRVT prediction could have gone wrong. I am ignoring other claims for now.
1. Sample Size and Representativeness: The PRVT projections by NEW is based on a sample size of only 747 polling stations out of 11.832 polling stations (about 6.3%) in the country. This small sample size raises concerns about its representativeness. If the sampled polling stations were not adequately distributed across different regions and constituencies, it may lead to biased results.
The concentration of the sample in certain areas may not capture the full diversity of voter preferences and behaviour in the country. NEW failed to mention in their report whether they sampled 6.3% of polling stations in every district, widely spread across each district or they arbitrarily took a total of 747 sample station results and analysed it. It is worth noting that uneven distribution of the samples taken can give a false impression of the election results.
2. Regional Biases and Alliances: Regional dynamics historically affect voting patterns in Sierra Leone significantly, which impact the election outcomes. The SLPP has its strongholds in the South-East of the country where they consolidated and increased votes in the just concluded elections. On the other hand, the APC has its strongholds in the North-West where SLPP made significant inroads. The SLPP alliance with the National Grand Coalition (NGC), in particular, favoured Maada Bio’s re-election bid. Let us go back to recent history. In the 2018 general elections, the NGC candidate, Dr Kandeh Yumkella, came third pulling a total of 174,014 valid votes cast, representing 6.86% of total valid votes. He also got 4 parliamentary seats. Kandeh comes from Kambia District, in the Northern part of Sierra Leone (which is in the APC stronghold). This was where he pulled much of his votes in 2018 and won all his parliamentary seats. Ahead of the 2023 elections, NGC formed a historic alliance with the ruling SLPP party where they agreed that the NGC party will not send a presidential candidate to throw all their support behind the SLPP candidate.
The PRVT projections may not have considered the above factors adequately. Failing to account for these regional variations in which the SLPP largely kept and even improved on votes from their stronghold, and their alliance with the NGC which fetched additional votes from the APC stronghold can lead to inaccurate estimates.
3. Defections and Role of Influential Politicians: The defections from APC and C4C as well as the influence of certain political figures could have substantially impacted the recent election results. The alliance between the SLPP and the NGC, along with the defections from the C4C party to the SLPP, suggest shifts in support and voter preferences in favour of the SLPP.
Let us particularly examine the case of C4C. In the 2018 general elections, the Coalition for Change (C4C) presidential candidate, Alhaji Samuel Sam-Sumana, had 87,720 of total valid votes cast, representing 3.46% of total valid votes. Sam-Sumana had much of his votes from the Kono District where he comes from. C4C also got 8 parliamentary seats, out of 9 and 1 went to APC. C4C also won the Kono City Mayoral election and had the Kono District Council Chairmanship. It is worth noting that in the 2012 elections, APC had all 9 seats as a result of Sam-Sumana’s influence in the district, but they lost all but ONE in 2018 due to the fallout with Sam-Sumana over his sacking as vice-president by Ernest Bai Koroma. Kono has always been seen as a swing district that can contribute to the outcome of elections in the country.
Arrangements to have Sam-Sumana and his party to join the opposition APC for the 2023 elections did not go well. While Sam-Sumana left his party to start covert arrangements with APC, the rest of his party members were left behind, dissatisfied with what was called a selfish attitude. This left the party without leadership. In the end, influential stakeholders of the C4C party defected to the SLPP ahead of the 2023 elections, including but not limited to: the Kono City Mayor, the Kono District Council Chairman, and 4 influential parliamentarians. Because Sam-Sumana did not get what he wanted from APC, he did not play any active part to help APC gain votes in Kono District in the just concluded elections.
Another influential politician, who claims to have come from Kono, is the First Lady of the Republic of Sierra Leone, Madam Fatima Bio. In addition to the mass C4C defections to SLPP, Madam Fatima Bio had some good influence on the electorates in Kono.
As a result of the above factors, Kono went in favour of the SLPP. These human behaviours are factors which the PRVT may not have fully accounted for in its projections that could affect the validity of their claim.
4. Historic Election Results Trend: The trend from 2007 to 2018 indicates significant changes in voter preferences and party performance. In the 2007 first-round presidential elections, the APC presidential candidate got 44.34% while the SLPP presidential candidate got 38.28%. In the run-off, the APC presidential candidate got 54.62% while the SLPP presidential candidate got 45.4%.
In the 2012 presidential elections, where Ernest was seeking his second term re-election, he got 58.7% of all valid votes in the first rounds, while the SLPP presidential candidate got 37.4%. There was no run-off. Hence, Ernest came from 44.34% in the first rounds in 2007 to a whopping 58.7% in 2012. The dynamics then in 2012 are similar to the 2023 dynamics: defections from the SLPP during that period as well as APC making significant inroads in the South-East due to influential politicians who declared for Ernest by then.
In the 2018 presidential elections, the APC presidential candidate got 42.7% while the SLPP presidential candidate got 43.3%; In the run-off, the APC presidential candidate got 48.2% while the SLPP presidential candidate got 51.8%. Since that election, the SLPP made significant moves to consolidate their South-Eastern votes and penetrate into the North-Western votes which resulted in good progress in districts such as Falaba, Koinadugu and Kambia in the APC strongholds. Also, the SLPP significantly outperformed APC in Kono, a district APC commanded in 2012 and which was in the hands of C4C in 2018.
5. Voter Registration: A comparison of the 2018 and 2023 voter registers revealed that the number of registered voters in the SLPP stronghold of the South-East increased by a significant 9.6% while it only increased by 3.9% in the APC stronghold of the North-West. Poor and biased statistical sampling that is inconsistent with these realities can produce an erroneous projection of the results.
6. Other Factors: The PRVT projections solely rely on statistical analysis and do not account for other contextual factors that may have influenced the election outcomes. Factors such as campaign strategies, voter turnout, voter education, and campaign finance could have also had a substantial impact on election results. Also, from voter registration to a few days before the elections, the APC was very undecided as they called for a boycott, postponement of the elections and even demanded the resignation of the ECSL Commissioner. These factors were not taken into consideration in the PRVT analysis and this can affect the accuracy of their projections.
In conclusion, it is sufficient to say that the NEW PRVT projections could be significantly erroneous and is therefore misleading to fall for it 100% at first glance. The choice of a margin of error of 2.7% makes statistical sense but considering the voter behaviours from 2007 to date, or even dating back from the two elections won by the late Alhaji Dr Ahmed Tejan Kabba, it can lack merit. These historic voting patterns might suggest a much wider margin of error that could be between +/-3 and +/-6 as opposed to the lower 2.7% used by NEW. The NEW predictions would have been much different if the margin of error chosen was different and in line with historic trends. For instance, a margin of error of 5.5% could have given a limit within which the ECSL result holds true.
Also, the NEW PRVT sample size of 6.3% is too small to draw meaningful conclusions. Additionally, another thing that can affect its merit is the sample distribution. If this was distributed in a skewed manner, it can erroneously discredit the result of the election which can be very misleading. It is also important for NEW to publish the full results they sampled stating the respective polling stations where they were taken from.
Since NEW claims to have had elections observers at all polling centres, I want to recommend that in future, they should sample at least a 20% of all stations proportionately to the voter register, evenly distributed across the district, and carefully choose the margin of error and confidence interval for their PRVT analysis. It will be necessary also to consider voter behaviours leading to the elections as these are human factors that have affected election outcomes in the country.
Discrediting the elections result based on the grounds they presented is a very dangerous game.
I rest my case!
I remain, EL.
