A Look at the IGR Report: Sierra Leone’s Political Marathon, Volume 2

1. Introduction

Sierra Leone’s 2028 presidential election will be an open-seat contest. Section 46(1) of the 1991 Constitution bars any person from holding the presidency for more than two five-year terms, so President Julius Maada Bio cannot stand again.² The governing Sierra Leone People’s Party (SLPP) must therefore choose a successor while in office, a transition that has twice preceded a change of ruling party (2007 and 2018); commentators note that no ruling party has yet retained power after a two-term president since Sierra Leone returned to multi-party democracy in 1996.³ The opposition All People’s Congress (APC) must choose a challenger after two successive defeats. In both parties that choice belongs to delegates, not to ordinary members or the electorate.⁴

Into this setting, IGR launched Sierra Leone’s Political Marathon, a series intended to assess party preparedness for 2028 and stimulate issue-based debate although its Executive Director Andrew Lavali told Umaru Fofana of AYV Television this week that the intention was to strengthen internal party democracy.⁵ Its maiden edition, fielded in December 2025, surveyed 1,200 adults and found Jagaban leading among APC supporters and Kandeh Yumkella leading a fragmented SLPP field, with the First Lady fourth.⁶ At the first report’s validation meeting, leaders of both parties objected that a general-population sample could not predict delegate-run primaries.⁷ Volume 2 appears to answer that objection by claiming to survey delegates directly, and it reports the First Lady leading the SLPP and Jagaban leading the APC again. With 2,400 randomly surveyed Sierra Leoneans across the country and delegates categorized by higher and lower cadre, IGR’s data sources raised questions in the minds of many party faithful and certainly research analysts and critics.

This paper asks three questions:

1. Can a survey of “delegates” conducted before either party has opened, closed or certified its flagbearer contest measure delegate preference in a meaningful sense? (Validity).
2. What sources of error are most likely to have shaped Volume 2’s findings, and how severe are they? (Reliability).
3. What are the political, policy and social risks of publishing such findings at this stage of Sierra Leone’s succession cycle? (Consequence).

The argument is that Volume 2 suffers from a specification problem that no amount of fieldwork can cure: it measures a population that did not yet exist in certified form, choosing among a field that had not been declared. Secondary problems of recruitment, question design, self-censorship and disclosure compound this. The paper nonetheless takes seriously IGR’s position that early, imperfect data can expose ‘elite’ steering in opaque selection systems.

Section 2 sets out the political and institutional context. Section 3 reviews the relevant literature. Section 4 explains the analytical framework. Section 5 describes the IGR series. Section 6 presents the methodological findings, Section 7 the stakeholder critiques and counter-arguments, and Section 8 the wider implications. Sections 9 to 11 offer recommendations, limitations and conclusions.

2. Context: Political Parties, Delegates and the 2028 Succession

2.1. A two-party, delegate-centred system

Sierra Leone is formally multiparty but functionally a two-party system. International IDEA’s 2025 assessment found 17 registered parties but identified the APC and SLPP as the only dominant forces, with large, stable memberships recruited largely along ethnic and regional lines rather than ideology.⁸ The Bertelsmann Transformation Index similarly reports a public perception that citizens are either “red” (APC) or “green” (SLPP).⁹

In both political parties, presidential candidates are chosen by delegates elected through a tiered structure of constituency, district, regional and national conventions.¹⁰ The APC’s constitutional framework enumerates delegate categories for its National Delegates Conference (NDC), including fixed quotas from constituencies, the diaspora, the women’s and youth congresses, and other organs.¹¹ IDEA observes that the SLPP’s nationwide delegates have historically been the key voting bloc for any presidential aspirant.¹²

2.2 Elite influence over selection

The formal delegate system coexists with strong top-down influence. IDEA records that outgoing President Kabbah shaped the SLPP’s flagbearer choice before 2007, provoking the split that produced the PMDC, and that outgoing President Koroma selected the APC flagbearer at the October 2017 convention.¹³ IGR’s own Volume 1 interviews found that national party officials often sway delegates through persuasion, recommendation, coercion, or vetoes based on party rules.¹⁴ Volume 1 also found that only 26% of self-identified members or supporters had voted in a party primary.¹⁵

2.3 The SLPP calendar

The SLPP had not opened flagbearer nominations at the relevant time. Its then Secretary-General Sulaiman Banja Tejan-Sie told fact-checkers that the new national executive would set a date later and that this would not happen soon while the President remained the face of the party.¹⁶ Commentary from early 2025 reported that the party’s National Leader warned that anyone campaigning prematurely to become the flagbearer of the party would not be doing so in the interest of the party and maybe suspended.¹⁷ At the party’s 14th National Delegates Conference, President Bio set out five qualities the next SLPP flagbearer should have, without naming a candidate.¹⁸ In 2026 the party’s publicity secretary Moses Mambu said the choice would be made by delegates through voting rather than sentiment although his statement contrasts with the party chairman’s ‘‘thou shall be’’ analogy.¹⁹ Coverage through 2026 still described nominations as formally unopened, even as aspirants such as Musa Tarawally courted district delegates publicly.²⁰

2.4 The APC calendar

The APC published a revised internal-elections timetable in January 2026. It provided for ward, constituency, regional and special-organ conventions between March and May 2026, publication of the eligible NDC delegate list on 18 July 2026, objections by 25 July 2026, and an NDC on 9–12 August 2026.²¹ The party named Bai Bureh Hall, PortLoko, as the venue for the NDC.²² Article 56(f) of the APC constitution requires an opposition party to elect its presidential candidate at least two years before the anticipated presidential election. A commentary published on 6 October 2026 reported that the NDC had not been held on schedule, citing petitions, reruns and litigation, and argued that the two-year deadline would now be missed.²³ A DUBAWA fact-check dated 25 August 2026 confirmed that the APC was still determining its flagbearer.²⁴

2.5 The wider constitutional climate

The succession is unfolding amid institutional tension, including an opposition walkout from Parliament on 10 August 2026 over a disputed constitutional amendment bill.²⁵ Contested data in this climate is more likely to be used as a political weapon than debated on its merits.

3. Literature Review

Four bodies of scholarship bear on the IGR report: the comparative study of candidate selection, research on primaries and nominations in Africa, survey-error methodology, and evidence on how published polls and politically sensitive questions behave.

3.1 Candidate selection as an object of study

Rahat and Hazan’s framework classifies candidate selection along four dimensions: who may stand (candidacy), who chooses (the selectorate), how decentralized the process is, and whether candidates are voted on or appointed.²⁶ Their later book shows that the selectorate’s size and inclusiveness shape participation, representation, competition and responsiveness, and cautions that more participation does not automatically mean better democracy.²⁷ For present purposes the key insight is definitional. Any study of “who will be nominated” must first specify the selectorate precisely. In Sierra Leone, that selectorate is a certified body of convention delegates, not party supporters in general.

3.2 Nominations and intra-party politics in Africa

Work on Ghana is the closest comparative guide. Ichino and Nathan find that, where politics is clientelistic, local members push for primaries partly to extract benefits from aspirants, and that party leaders interfere strategically in nominations.²⁸ They also find that primaries carried an electoral bonus for the opposition but a penalty for the ruling party, which they attribute to intra-party conflict.²⁹ Seeberg, Wahman and Skaaning show across sub-Saharan Africa that nomination contests are a significant and understudied site of election-related violence, especially in dominant-party settings with intense internal competition.³⁰ Together these studies suggest that delegates’ stated preferences are entangled with patronage, leader signalling and conflict risk, and that early public labelling of front-runners is not a neutral act.

3.3 Survey error beyond sampling error

The Total Survey Error (TSE) paradigm holds that sampling error is only one component of a survey’s inaccuracy. Groves and Lyberg trace how the framework integrates errors of representation (coverage, sampling, nonresponse, adjustment) with errors of measurement (validity, response, processing).³¹ Biemer and colleagues stress that less-studied sources such as coverage and specification error often pose the largest risks to data quality.³² Applied to delegate polling, TSE implies that a perfectly executed interview programme can still be badly wrong if the frame of delegates is incomplete or the construct measured is not the one the headline claims.

3.4 Poll effects and sensitive questions

Whether published polls change behaviour remains debated, but recent experiments find real if modest bandwagon effects. A Dutch survey experiment with over 23,000 respondents found that emphasis on poll growth raised subsequent support, though simple poll levels did not.³³ A Danish national experiment found bandwagon movement strongest when a party was reported as gaining.³⁴ Delegate electorates are small, networked and exposed to leader cues, so the effect may be larger there than in mass electorates; that is a hypothesis, not an established finding.

On sensitive questions, Tannenberg’s analysis of more than 140,000 Afrobarometer respondents in 36 African countries found that respondents who believed the government had commissioned the survey gave systematically different answers on citizen-state questions in more autocratic settings.³⁵ List-experiment evidence from China found self-censorship of 16 to 22 percentage points on regime-support items.³⁶ Questions about the President’s spouse put to ruling-party delegates are plausibly sensitive in exactly this way.

3.5 The gap this paper addresses

There is little published work on polling party delegates in West Africa, and none, to the author’s knowledge, that evaluates such polling before delegate certification. This paper applies the frameworks above to fill that gap for one prominent case.

4. Analytical Framework and Method

This paper is a desk-based methodological assessment, not a re-analysis of IGR’s microdata, which are not public. It combines document analysis of IGR’s published findings and press coverage, party statements and calendars, and fact-checks, with a structured appraisal adapted from the TSE framework.³⁷

4.1 Error Sources Assessed

Six error sources were selected because each maps onto a distinct way a delegate poll could misstate convention outcomes.

4.2 Severity Rating

Each source is rated high where the error could plausibly reverse the reported leader, medium where it could materially change margins, and low where its effect is likely small. Ratings are judgments based on the literature and the evidence in Sections 2 and 5; they are not statistical estimates [with emphasis].

4.3 Source Selection and Verification

Primary party documents were used where publicly available. Otherwise the paper relies on national news outlets, fact-checkers, and opinion pieces which are flagged as such in the references including opinions of party leaders. All web sources were accessed on 15 October 2026.

5. The IGR Political Marathon Series

5.1 The Institution

I view IGR as a Freetown-based think tank established sometime in February 2013 and serves as Afrobarometer’s national partner in Sierra Leone, with extensive experience of nationally representative surveys.³⁸ Its June 2023 pre-election poll on the presidency for both political parties forecast incumbent President Bio at 56% and Dr. Samura Kamara at 43%, a forecast widely reported internationally.³⁹ That record gives the series credibility, and also raises the stakes of a contested release. This is of course without prejudice to the eventual outcome of the observer’s reporting.

5.2 Volume 1 (December 2025)

Volume 1 interviewed 1,200 adults face-to-face across all regions, drawn from 150 enumeration areas.⁴⁰ Seventy-four per cent identified with the APC or SLPP, split evenly at 37% each, and the survey did not distinguish card-carrying members from supporters.⁴¹ Preferences were measured by asking respondents to name three people they liked (“likability”) and three they would vote for as president (“electability”).⁴² Results for the leading names were as follows.

Coverage elsewhere reported different SLPP figures, including 47% likability and 32% electability for the First Lady among SLPP respondents, suggesting that party-specific and national bases were reported inconsistently in the press.⁴³

Volume 1 disclosed its main limitation: no public database of registered members or designated delegates existed, so the sample could not be restricted to either.⁴⁴ Party leaders at the validation meeting objected on precisely this ground.⁴⁵ IGR replied that party membership in practice extends beyond formal cards and that delegates are selected at community level to represent grassroots views.⁴⁶ Volume 1 also found that 52% of respondents opposed spouses or children of a president contesting immediately after that president’s term.⁴⁷

5.3 Volume 2 (2026)

In my research, I found that, Volume 2 surveyed respondents identified as SLPP and APC delegates and reported the First Lady and Jagaban as the preferred candidates in their respective parties. This has been confirmed by the head of IGR in a recent interview with Truth Media. Its sample size, fieldwork dates, delegate-identification method, question wording and response rates could not be verified from public sources at the time of writing. However, in the Same interview, Andrew Lavali made a broad statement on them as well as the organization’s policy adviser on AYV television. This analysis captured both as accurate reporting from IGR.

5.4 What changed between volumes

The APC result is stable across both samples, which lends it some robustness. The SLPP result is not. The First Lady moved from fourth among SLPP-leaning citizens in Volume 1⁴⁸ to first among delegates in Volume 2. Three explanations fit that pattern equally well on the available evidence:

1. Delegates, who are closer to party leadership, may genuinely prefer a candidate whom the wider base does not, which would itself be a significant finding about intra-party democracy [Genuine ‘elite’–mass divergence].
2. A delegate sample accessed through structures close to the presidency would over-represent her supporters [Sampling artefact].
3. Delegates may be more reluctant than ordinary citizens to name alternatives to the President’s spouse to an interviewer [Measurement artefact].

Volume 2 can only adjudicate among these if it discloses how delegates were identified and approached.

6. Findings: A Methodological Assessment

In this study, the author has identified six (6) error sources. Two of the six error sources are rated high severity, and both arise from timing rather than fieldwork quality. The remaining four are rated medium.

6.1 Specification Error: measuring a contest not yet called (high)

A convention poll presumes a declared field and a defined electorate. Neither existed. The SLPP had not opened nominations and had discouraged early campaigning;⁴⁹ the APC’s NDC had slipped past its scheduled dates.⁵⁰ What Volume 2 can measure, therefore, is name recognition and perceived proximity to power among politically active respondents, not convention vote intention. Rahat and Hazan’s framework makes the point sharply: without a fixed candidacy rule and selectorate, there is no defined selection event to forecast.⁵¹ Volume 1 itself worked from a list of 26 rumoured aspirants across both parties.⁵² The report’s own miscalculations of percentage and figures raises additional concerns for IGR.

6.2 Coverage Error: an uncertified frame (high)

Coverage error arises when the list from which a sample is drawn differs from the target population.⁵³ For the APC, the eligible NDC list was due on 18 July 2026 and open to objection until 25 July 2026,⁵⁴ and the process was subsequently disrupted by petitions, reruns and court action.⁵⁵ Any delegate poll fielded before certification, or using unofficial lists, will include people who will not vote and omit people who will. For the SLPP, no flagbearer convention had been convened, so its delegate body for that purpose was undefined. IGR’s own Volume 1 acknowledged that no public delegate database existed.⁵⁶ Unless Volume 2 documents a certified frame, this error alone could reverse the reported leader.

6.3 Sampling and Recruitment: access through gatekeepers (medium)

Where delegate lists exist, party secretariats hold them. IGR found that national officials exert heavy influence over delegates,⁵⁷ and the Ghanaian evidence shows party leaders intervening strategically in nominations.⁵⁸ If access ran through officials or networks aligned with particular aspirants, the sample will tilt toward them. Public declarations of delegate support, such as Moyamba delegates pledging to Hon. Musa Tarawally while he accused rivals of trading on the President’s favour, show how organised and partisan delegate networks already are.⁵⁹ Without a random draw from a certified list and a disclosed response rate, recruitment bias cannot be ruled out.

6.4 Question Design: mentions are not votes (medium)

If Volume 2 used Volume 1’s format of naming three liked and three electable figures,⁶⁰ it measured multi-mention salience rather than first preference. Mention-based items favour the best-known names, and a person can be named by many delegates without being anyone’s first choice. Conventions are decided by first-preference ballots among declared candidates. A valid instrument would present the declared field and ask for a ranked first choice.

6.5 Response Error: fear, loyalty and inducement (medium)

Questions about the President’s spouse put to ruling-party delegates are politically sensitive. Afrobarometer evidence shows that respondents who fear government sponsorship shade their answers on citizen-state questions in less democratic settings,⁶¹ and list experiments elsewhere reveal self-censorship of 16–22 points on regime-support items.⁶² Commentary on Sierra Leonean conventions describes delegates swearing oaths to aspirants and collecting money from several candidates without voting for all of them.⁶³ Each mechanism pushes stated preference away from the secret ballot, and the first two would plausibly inflate support for the incumbent’s preferred figures.

6.6 Disclosure: what cannot be checked (medium)

AAPOR’s Code requires researchers to disclose, when results are released, enough about sponsorship, data collection and sampling to permit independent review and verification.⁶⁴ At minimum, readers need the frame source; number of delegates contacted and interviewed by party and region; response rate; weighting; margin of error; fieldwork dates; full questionnaire; and funder. Without these, the alternative explanations in Section 5.4 cannot be distinguished.

6.7 Summary of findings

7. Stakeholder Critiques and Counter-Arguments

Criticism of Volume 2 can be expected from five constituencies, and IGR has credible replies to some but not all of them.

7.1 The sharpest SLPP objection

As observed, volume 1 found majority public opposition (52%) to a president’s spouse or child standing immediately after that president’s term.⁶⁵ Some commentators also dispute the First Lady’s eligibility under SLPP rules and nationality requirements, though these claims are contested and come from opinion writing.⁶⁶ A report placing her first among delegates while its own earlier public data cut the other way will be read either as evidence of elite capture of the delegate body or as evidence of a skewed sample. IGR needs to show which.

7.2 The sharpest APC objection

With the NDC delayed and compliance with Article 56(f) in doubt,⁶⁷ the identity of APC delegates is itself contested. Naming a favourite before the list is settled risks inserting an independent research body into an internal dispute that may end in court. Since lower level elections are ongoing, IGR not being a party member cannot even guess the delegate list. From a social and political analyst’s point of view, it does not only show inaccurate findings, it is dangerous to do so.

7.3 The strongest case for IGR

IGR’s position deserves a fair hearing on three points. Firstly, the APC finding is consistent across two different samples, which suggests it reflects something real.⁶⁸ Secondly, IGR openly disclosed Volume 1’s frame limitation, which many partisan polls do not.⁶⁹ Thirdly, in systems where selection is dominated by elites and only about a quarter of supporters vote in primaries,⁷⁰ early public data may be one of the few checks on decisions made behind closed doors. On this view, the remedy for imperfect data is better data, not silence.

This paper accepts the third point in principle. It holds, however, that the case for publication is strongest when findings are framed as sentiment rather than prediction, and when methods are fully disclosed.

8. Discussion: Political, Policy and Social Implications

A delegate poll released before delegates are certified can shape the contest it claims to measure whether IGR agrees with this or not. IGR spokespersons including its head, Andrew Lavali, have categorically denied that the organization’s report does shape the contest for leadership in both political parties. There is a real risk in doing such a report. Four channels make this risk concrete in Sierra Leone.

8.1 Signalling and the bandwagon

Experimental evidence shows that reported momentum can draw support toward a front-runner.⁷¹ Delegates are a small, networked selectorate whose members often depend on patronage and respond to leadership cues.⁷² A published “delegate favourite” can therefore act as a coordination signal. If delegates infer that a candidate enjoys leadership backing, a poll may become self-fulfilling regardless of its accuracy.

8.2 Intra-party conflict and its escalation

Comparative work identifies nomination contests as a significant source of election-related violence in Africa, especially where internal competition is intense.⁷³ Sierra Leone’s own record includes splits and court cases tied to candidate selection, among them the SLPP split that produced the PMDC ahead of the 2007 elections.⁷⁴ Premature front-runner labels can harden factions before rules and lists are settled, raising the cost of losing and the incentive to litigate.

8.3 Credibility of independent research

IGR’s value to Sierra Leonean public life rests on perceived neutrality, built partly through its Afrobarometer partnership and its 2023 election forecast albeit the controversy that followed the final results, according to reports.⁷⁵ If Volume 2 is seen as favouring the governing party’s establishment candidate, or as interfering in the opposition’s disputed process, that perception could suffer. Partisan dismissal of credible future research would be a loss for evidence-based debate generally.

8.4 Gender, dynasty and democratic norms

The SLPP finding also intersects with two normative debates. On one hand, a woman leading a major party’s field would be notable in a political system where women remain under-represented at the top. On the other, Volume 1 recorded majority public resistance to family succession, and IDEA documents a history of outgoing presidents shaping their successors.⁷⁶ A report that cannot distinguish genuine delegate enthusiasm from ‘elite’ steering risks lending legitimacy to the latter, or unfairly discrediting the former.

8.5 Policy relevance

The episode exposes a regulatory gap. This assessment found no specific Sierra Leonean code governing the publication of intra-party polls, and the parties do not publish delegate registers in a form researchers can sample from. Both gaps are addressable by the parties themselves, by the research community, and potentially by the Political Parties Regulation Commission (PPRC) through voluntary guidance.

9. Recommendations

The findings point to a minimum standard for delegate polling: certified frame, declared field, random draw, confidential first-preference ballot, and full disclosure.

9.1 For IGR

1. Publish a methods annex for Volume 2 covering frame source, contacts and completes by party and region, response rate, weighting, margin of error, fieldwork dates, full questionnaire and funder, consistent with AAPOR disclosure standards.⁷⁷
2. Reframe the findings as “early sentiment among party activists” until certified delegate lists exist.
3. Re-field after nominations close and delegate lists are certified, drawing a random sample from the official lists.
4. Ask for a ranked first choice from the declared field, rather than open multi-mention items.
5. Use a confidential, self-completed mode and, for sensitive items, indirect techniques such as list experiments.⁷⁸
6. Report Volume 1 and Volume 2 side by side and test the competing explanations in Section 5.4.

9.2 For the SLPP and APC

1. Publish certified delegate registers promptly and on the timetables already announced if this reports is something interesting and helpful to party preparation.⁷⁹
2. Consistent with 1 above, grant independent researchers equal, documented access to those registers.
3. Respond to unwelcome findings with data and process transparency rather than rebuttal alone.

9.3 For regulators, media and civil society

1. To support a transparent process supported by these interests, develop voluntary guidance on publishing intra-party polls, including minimum disclosure items.
2. Report headline names alongside sample size, fieldwork dates and frame caveats.
3. Support further research on delegate behaviour in West Africa, a gap identified in Section 3.5.

10. Limitations

This assessment has four main limitations. Firstly, Volume 2’s full text and microdata were not available for public comments, so its design is inferred from the author’s account and from Volume 1; if Volume 2 sampled from a certified list using a random draw, my sections 6.2 and 6.3 would need revision. Secondly, several sources are national news reports and opinion pieces, as flagged in the references; party constitutions in force were not all publicly accessible, and both political parties’ constitutional texts were only found late on in the analysis but not verified. Thirdly, severity ratings are reasoned judgments rather than estimates. Fourthly, the evidence on bandwagon effects and self-censorship comes largely from mass electorates outside Sierra Leone, and its application to small delegate bodies is an inference.

11. Conclusion

IGR and its leadership has contributed a lot to our progress on policy design as a country especially in the last decade. It has been criticized on some of its work. This work is done because there is need for an independent professional perspective on a subject as sensitive as elections and the presidency, from a standpoint of experts, to contribute to a debate on the reports.

The three objections that prompted this paper hold up under scrutiny. Neither party had formally called its flagbearer contest; the APC delegate list was not final; and a delegate poll without a certified frame cannot claim to predict a delegate vote. These are problems of specification and coverage, which the TSE literature identifies as among the most consequential and least visible sources of survey error.⁸⁰

IGR’s defence, that early data enriches debate and that formal membership understates real party activity, is legitimate for a sentiment study. It does not justify headlines that name winners. On the available evidence, Volume 2 tells us who is most visible and best networked among party activists in 2026. It tells us little yet about who will carry either banner in 2028.

The broader lesson concerns Sierra Leone’s democratic infrastructure. Credible intra-party polling requires parties willing to publish their selectorates and researchers willing to wait for them. Building both before the 2028 conventions would serve the parties, the research community and voters alike.