Roughly 30% of the global population lives under customary laws and traditional governance structures, highlighting their widespread presence across continents (Baldwin & Holzinger, Reference Baldwin and Holzinger2019). Although traditional leaders – particularly village chiefs – are not subject to regular elections, they often enhance government responsiveness, facilitate collective action, and broker resources (Baldwin Reference Baldwin2016; Honig Reference Honig2017; Murtazashvili & Murtazashvili Reference Murtazashvili and Murtazashvili2016). One explanation is that village members possess the capacity to sanction their leaders, which may compel those leaders to act in the community’s interest. In particular, non-electoral sanctions have emerged as a prominent theoretical mechanism through which traditional leaders are held accountable (Paller Reference Paller2014; Wilfahrt Reference Wilfahrt2018; Carlson Reference Carlson2021; Mattingly Reference Mattingly2016).
This study builds on and contributes to existing research by formally testing the perceived legitimacy and feasibility of non-electoral sanctions through a novel survey experiment. It examines whether, and under what conditions, village members view sanctioning their chiefs as a justified response when leaders act against the interests of the communityFootnote 1. Scholars highlight two main mechanisms through which citizens may influence local leaders outside of electoral institutions. First, research emphasizes that leaders who are socially and economically embedded in their communities may be more responsive to citizen interests (Baldwin Reference Baldwin2016). In such contexts, villagers may withhold cooperation, voice public criticism, or even threaten noncompliant chiefs (Tsai Reference Tsai2007; Chen et al. Reference Chen, Pan and Xu2016; Baldwin Reference Baldwin2016; Zhong & Zeng Reference Zhong and Zeng2024). Second, scholars have pointed to the role of hierarchical political institutions. Honig (Reference Honig2022), for example, shows that hierarchy can generate horizontal accountability among chiefs by creating incentives for compliance with higher-level authorities. Yet, as Baldwin (Reference Baldwin2025) notes, little empirical work has examined whether such structures also enable downward accountability to citizens. Building on this gap, we theorize that villagers have two channels through which they may sanction chiefs: either by directly enforcing social norms themselves or by appealing to higher authorities who can sanction local leaders. We make a first step toward testing this theoretical claim by examining whether citizens view these sanctions as legitimate and feasible.
The study was conducted in Sierra Leone in April and May 2023, following pre-registered hypotheses (see Appendix C for the pre-analysis plan). In this context, chiefs wield considerable authority, face limited political competition, and have been described as capturing civil society (Acemoglu et al. Reference Acemoglu, Reed and Robinson2014). Scholars often characterize the institutional environment as lacking robust constraints on chiefs, with weak formal mechanisms of accountability. At the same time, village chiefs are deeply embedded in the local social and political fabric, making Sierra Leone a particularly relevant setting for examining informal mechanisms of accountability.
We conducted a survey experiment to test whether villagers would endorse sanctioning their village chief if the chief were hypothetically to steal from a community project. The experiment also examined the role of political intermediaries – specifically, community elders – as mediators shaping sanctioning preferences. Our findings show that villagers primarily support indirect sanctions, such as reporting to higher authorities. In contrast, many direct sanctions are not widely perceived as legitimate forms of political behavior. Those preferences differ substantially by gender, income, and voting rights, with men, lower-income respondents, and those with voting rights viewing, on average, a wider set of sanctions as legitimate and feasible. Notably, when community elders publicly criticize the chief, the range of acceptable sanctions expands, underscoring their role in coordinating local accountability and reinforcing social norms.
This study makes three contributions. First, it broadens our theoretical understanding of accountability mechanisms beyond electoral settings, complementing classic selectorate models of accountability (Mesquita et al. Reference Mesquita, Smith, Siverson and Morrow2005; Balasuriya Reference Balasuriya2023) and advancing discussions on the sources of chiefs’ authority (Baldwin & Ricart-Huguet Reference Baldwin and Ricart-Huguet2023). Categorizing sanctioning channels as direct or indirect offers a valuable analytical framework. Second, our study clarifies how chiefs in Sierra Leone may be held accountable by their communities (Acemoglu et al. Reference Acemoglu, Reed and Robinson2014; Baldwin & Holzinger Reference Baldwin and Holzinger2019), identifying the political tools available to citizens to hold leaders accountable. We show that indirect sanctions are generally seen as more legitimate and less costly than direct sanctions. Third, this research contributes to the growing literature on political intermediaries, with particular attention to community elders (Baldwin et al. Reference Baldwin, Muyengwa and Mvukiyehe2022). We demonstrate that community elders influence the range of sanctions considered as legitimate and feasible by village members.
Institutional context
Sierra Leone’s chieftaincy system features strong vertical authority and limited formal accountability, patterns common in rural sub-Saharan Africa (Baldwin Reference Baldwin2025). The study took place in Southern Province’s Bonthe and Moyamba districts, where chiefs hold substantial power and checks are limited. It covers 12 chiefdoms and 77 villages. Chiefdoms, led by paramount chiefs elected by tribal authorities, are key governance units (see Appendix B for further information). Each chiefdom is divided into sections governed by section chiefs above village chiefs. Chiefs act as state clients and community patrons, mediating between rural populations and national authorities (Becorpi Reference Becorpi2018). They allocate resources, sustain patronage networks, and mobilize electoral support in exchange for state-backed resources (Becorpi Reference Becorpi2018; Labonte Reference Labonte2012). Though formally part of the state, chiefs sometimes compete with government institutions (Henn Reference Henn2022; Grieco Reference Grieco2024). Despite 2004 decentralization reforms, many rural residents prefer traditional leaders, who are seen as better defenders of customary land rights and more effective dispute resolvers (Fanthorpe Reference Fanthorpe2006; Sawyer Reference Sawyer2008). Village chiefs remain trusted conflict authorities despite concerns over corruption and nepotism (Ruppel & Leib Reference Ruppel and Leib2022; Sawyer Reference Sawyer2008).
Accountability at the village level is complex. Town chiefs, though formally elected by taxpaying members of landowning lineages, serve for decades and often for life (Bulte et al. Reference Bulte, Richards and Voors2018). In our sample, the last elections for town chiefs ranged from 1982 to 2023, averaging around 2012 – over ten years before data collection in April–May 2023. Such infrequency makes non-electoral sanctions particularly important. Yet, in Sierra Leone, the chieftaincy has been described as a colonial invention designed to make chiefs accountable upward rather than to their communities (Acemoglu et al. Reference Acemoglu, Reed and Robinson2014; Labonte Reference Labonte2012; Mihaylova Reference Mihaylova2023). Consistent with this, about 24% of villages in our sample have chiefs who were elected after their predecessors were suspended, reflecting the frequency of higher-level interventions, which Bulte et al. (Reference Bulte, Richards and Voors2018) argue are often strategic and self-serving.Footnote 2 Still, Voors et al. (Reference Voors, Turley, Bulte, Kontoleon and List2018) argue that communities can sometimes leverage these vertical ties to discipline their town chiefs.
Downward accountability is further limited by the exclusionary structure of rural politics. Only members of “chiefly families” – sometimes as few as 13% of households – are eligible for office (Labonte Reference Labonte2012). Labonte (Reference Labonte2012) also notes that, while challenging chiefs through public forums is theoretically possible, “non-elites are risk-averse in claiming their rights from elites, airing grievances in public forums, or demanding accountability.” This exclusion has historically fueled grievances and contributed to civil war (Richards Reference Richards2005; Humphreys & Weinstein Reference Humphreys and Weinstein2006; Mokuwa et al. Reference Mokuwa, Voors, Bulte and Richards2011), echoing Conteh’s (Reference Conteh2013) mention of the use of revolts and murders in extreme cases. Yet, paradoxically, satisfaction with chiefs is high: Casey et al. (Reference Casey, Glennerster and Miguel2012) report that 94% of households expressed approval in a comparable sample of villages. To date, however, there has been no systematic assessment of whether villagers can effectively discipline chiefs or mobilize higher-level authorities. This paper aims to fill this gap by empirically testing whether community members view a range of sanctions as legitimate and feasible, particularly when chiefs misappropriate community project resources, a common form of theft in Sierra Leone.
Finally, in 87% of villages, chiefs are supported by councils of elders, heads of major descent groups (Leach Reference Leach2022), who claim legitimacy as early settlers and control land rights. Elders advise on land allocation and coordinate with chiefs. Village chiefs in our sample are well embedded: 77% of respondents were born in the village, 73% of chiefs own farms there, and only 21% have formal jobsFootnote 3. Appendix B provides additional descriptive statistics on the institutional context.
Theory and hypotheses
Accountability requires that leaders adhere to established behavioral standards, with the risk of sanctions for failing to do so (Grant & Keohane Reference Grant and Keohane2005; Chen et al. Reference Chen, Pan and Xu2016). At its core, accountability serves to penalize the illegitimate use of power and rests on two foundational components: access to information and the capacity to impose sanctions (Grant & Keohane Reference Grant and Keohane2005). While access to information is a necessary precondition, this paper focuses on the second condition – the capacity to impose sanctions.
In electoral democracies, accountability is often exercised through periodic, competitive elections that allow citizens to select and remove leaders. In the absence of regular elections, however, electoral accountability is absent or weak (Baldwin Reference Baldwin2025). Instead, community members must rely on non-electoral accountability mechanisms (Baldwin Reference Baldwin2016; Baldwin Reference Baldwin2025). We focus here on the repertoire of non-electoral sanctions available to community members to sanction their village leaders when it does not behave in the community’s interests. Our theoretical framework is further illustrated in Appendix A.
Community interests refer to collective welfare and adherence to established social norms, including equitable allocation of resources, fair use of customary taxes, provision of public goods, protection of vulnerable community members, and adherence to local moral expectations. We focus specifically on one illustrative instance: the management of a community development project in which funds are misappropriated or stolen, a common issue in Sierra Leone (Anti-Corruption Commission of Sierra Leone 2024). Chiefs are expected to advance community interests in such projects – for example, by ensuring that funds are used for the intended public goods, reporting transparently to community members, and allocating resources equitably.
Sanctions are actions that impose costs – or withhold benefits – from a leader following a perceived violation (Meng et al. Reference Meng, Paine and Powell2023; Ostrom Reference Ostrom2005). Sanctions are not reducible to actions alone: their effectiveness depends on the social conditions under which they can be mobilized (Ostrom Reference Ostrom2000). We distinguish three interrelated but conceptually distinct dimensions of sanctioning capacity. First, sanction norms refer to the socially prescribed scripts that dictate what kinds of responses are considered appropriate following a violation (Ostrom Reference Ostrom2005). They are shared expectations about how community members should respond when leaders fail in their duties (Ostrom Reference Ostrom2005). Second, legitimacy concerns whether a given sanction is broadly recognized as rightful and appropriate within the community. A sanction norm may exist, but if citizens perceive it as biased, disproportionate, or captured by particular groups, its legitimacy is undermined and enforcement becomes unlikely. Third, feasibility refers to the practical ability to carry out a sanction. Even when a sanction is normatively prescribed and considered legitimate, it may not be feasible if it requires resources, coordination, or authority beyond the citizens’ reach. Sanctioning capacity thus involves: (1) the existence of norms that render sanctions socially meaningful, (2) legitimacy that makes them acceptable, and (3) feasibility that makes them practically implementable and enforceable. In this study, we focus on legitimacy and feasibility as two necessary but not sufficient conditions for non-electoral accountability.
Finally, we categorize the repertoire of sanctions according to their pathway of enforcement. Direct sanctions are those that citizens can impose on themselves. Chiefs who are embedded in their communities are more likely to provide public goods, as their interests align with those of their constituents (Baldwin Reference Baldwin2016), and they are motivated to earn the moral standing of citizens (Tsai Reference Tsai2007; Baldwin Reference Baldwin2025). Anthropological studies document various citizen strategies to pressure leaders (Arnall et al. Reference Arnall, Thomas, Twyman and Liverman2013), echoing Scott’s “weapons of the weak” (Scott Reference Scott1985), but the full scope of legitimate sanctions remains underexplored. Potential direct sanctions include public blame (Labonte Reference Labonte2012) or restricting chiefs’ access to communal institutions. In farming and fishing communities, collective labor depends on trust (Bulte et al. Reference Bulte, Richards and Voors2018), so chiefs neglecting communal interests risk losing labor support. Appendix A further discusses the set of sanctions considered in this study and their mentions in the literature.
On the other hand, indirect sanctions operate through intermediaries – community elders or higher-level chiefs such as section and paramount chiefs – who impose sanctions on town chiefs. These mechanisms rely on the vertical structure of traditional authority in Sierra Leone, where village chiefs are formally subordinate to section and paramount chiefs. This hierarchy is both codified in customary law and broadly recognized within communities, providing institutional oversight: severe violations of community interests by a village chief can lead to suspension or removal. These formal hierarchical ties create an institutional pathway through which citizens may strategically pursue indirect sanctions.
Overall, town chiefs are likely to be more responsive to some sanctioning agents than to others, making the pathway of enforcement critical for understanding accountability. This distinction matters because chiefs are not equally responsive to all forms of pressure: they may be more attentive to sanctions imposed by authoritative intermediaries than to citizen mobilization, as often stated in Sierra Leone (Labonte Reference Labonte2012; Bulte et al. Reference Bulte, Richards and Voors2018). While sanctions also differ in the distribution of their social costs (some primarily burdening the chief, others imposing collective costs), our focus here is on the enforcement pathway, which conditions accountability dynamics. This framework echoes existing research on accountability in other non-electoral contexts (Chen et al. Reference Chen, Pan and Xu2016; Anderson et al. Reference Anderson, Buntaine, Liu and Zhang2019) and complements selectorate models of accountability (Balasuriya Reference Balasuriya2023). Finally, for downward accountability to be effective, some sanctions require social coordination and, consequently, the resolution of inherent collective action problems. This is another necessary condition, which is beyond the scope of this paper.
Expectations
We theorize that non-electoral sanctions may take two forms: citizens can appeal to higher authoritiesFootnote 4 to sanction a village chief, or they can directly sanction the chief themselvesFootnote 5.
Hypothesis 1: Village members view sanctions against chiefs who violate community interests as legitimate and feasible.
Our pre-registered main hypothesis stated that when undemocratic village leaders do not act in the interest of their communities, citizens sanction them through a variety of social, economic, and political channels, with a preference for lower-cost options (Hypothesis 1; see Appendix G). Although the hypothesis originally concerned sanctioning behavior, our design does not observe sanctions directly. Instead, we test whether citizens view different sanctions as legitimate and feasible. We focus primarily on the extensive margin, namely, whether citizens endorse any sanction, while also considering the intensive margin of sanction severity as a secondary outcome. By contrast, preferences for lower-cost sanctions are not tested in the main analysis and are only supported with qualitative evidence from five focus groups in Appendix K, so this aspect receives less emphasis.
We also investigate the role of village elites, the council of elders, in driving sanctioning behavior of other community members. The council’s sanctions against the chief might serve as a sufficient deterrent, reducing the need for further community action. Prior research has revealed that the pressure exerted by community elders effectively shapes and influences the behavior of village chiefs (Baldwin et al. Reference Baldwin, Muyengwa and Mvukiyehe2022). Consequently, when community elders have already taken action, the efficacy of community members’ pressure is diminished, leading us to expect their abstention from participating in such actions. The wording of pre-registered Hypothesis 2 was simplified for clarity; content and predictions remain unchanged (see Appendix G).
Hypothesis 2: community members will be less willing to view sanctions as legitimate and feasible if the council of elders has already blamed the chief for their action.
Additional hypotheses from the pre-analysis plan are outside this paper’s scope and are addressed separately in another project.
Research design
Examining whether and how community members sanction leaders who act against their interests raises methodological challenges. First, the lack of official records on sanctions complicates tracing and analyzing such incidents. Second, effective sanctioning mechanisms may deter violations of community interests, making theft and sanctions rare and hard to study. Third, the region’s history of civil conflict, tied to the exclusion of local voices (Peters Reference Peters2011), may discourage political actors from discussing sensitive behaviors, complicating data collection. In such a setting, a survey experiment provides a strong justification because it allows us to study sanctioning responses to off-equilibrium violations that might never be directly observable. Consequently, we use a novel survey experiment designed to measure individual preferences for non-electoral sanctions.
Communities were randomly selected from villages located near mangrove resources, with eligibility based on population size. This sampling frame was defined by the goals of the broader research project, which focused on forest-dependent livelihoods and community-forest relationships. Although this paper does not analyze forest use directly, selecting such villages ensured that local chiefs were embedded in the community, a key theoretical scope condition for this study. Villages with more than 200 households were excluded, as chiefs in larger communities are often less integrated. Villages with fewer than 20 households were also excluded due to sample size requirements in other parts of the project.
In each village, we randomly selected 12 household heads using a two-step process: (1) a full household listing, and (2) a random sampling.
Household survey experimentFootnote 6
We implemented a vignette survey experiment using five hypothetical scenarios to identify sanctions deemed legitimate by community members and assess how the behavior of community elders shapes sanctioning preferences.
The main scenario reflects a common issue: misallocation of Non Governmental Organization (NGO)-led development resources, such as cookstoves, agricultural inputs, or mini-grids. Elite capture of such resources is well-documented across sectors like education (Mbiti Reference Mbiti2016; Reinikka & Svensson Reference Reinikka and Svensson2002), health (Azfar & Gurgur Reference Azfar and Gurgur2008; Njong & Ngantcha Reference Njong and Ngantcha2013), and development (Carlson & Seim Reference Carlson and Seim2020). While difficult to trace, these leakages can be curbed by accountability mechanisms (Carlson & Seim Reference Carlson and Seim2020; Anti-Corruption Commission of Sierra Leone 2024). Our vignette features a chief who appropriates part of an NGO project’s funds. Though many NGOs distribute goods rather than cash,Footnote 8 we simplified the scenario by focusing on funds for clarity. Neutral language was used, and enumerators were instructed not to use the word “steal” in local translations to avoid bias.
Each participant received one randomly assigned treatment. Randomization was conducted independently within each village. We also block-randomized by gender to improve the precision of estimates, given the modest overall sample size (907 respondents). Table 1 outlines these conditions. The control presents a well-managed project. Treatment 1 introduces minor misappropriationFootnote 9. Treatment 4, a larger sum. These gauge citizens’ sanctioning repertoire and whether responses vary by severity. The inclusion of a control arm serves to capture baseline sanctioning attitudes, rather than sanctions triggered by a specific act of misappropriation. Even without evidence of wrongdoing, some respondents considered sanctions legitimate, likely reflecting a general skepticism toward chiefs. This baseline is essential for identifying the marginal effect of misappropriation on sanction preferences. In addition, our design leverages the comparison between Treatment 1 (minor misappropriation) and Treatment 4 (major misappropriation) to examine the intensive margin of sanctioning, ensuring that inferences are not solely dependent on control–treatment contrasts. Treatments 2 and 3 introduce the council of elders’ behavior, allowing analysis of how elite dynamics shape preferences.
Description of the control and treatment conditions
Notes: Enumerators read the scenarios to participants in local languages (Sherbro, Mende, or Krio) using tablets.
Appendix E shows the balance across 13 covariates. F-tests are insignificant at the 5% level, indicating balance. However, Treatment 1 recipients are slightly less trusting of the chief and less employed than controls; Treatment 2 recipients are older than those in Treatment 3. We control for these imbalances in robustness tests (Appendix J).
Appendix F reports comprehension rates: 95% understood the amount taken, and 92% understood the elders’ behavior.Footnote 10
After reading the scenario, enumerators asked respondents five questions:
-
• Outcome 1: Agreement with the chief’s behavior (1–5 scale);
-
• Outcome 2: Whether citizens should respond (binary);
-
• Outcome 3: If yes, specify actions (open-ended);Footnote 11
-
• Outcome 4 (main): Number of sanctions deemed legitimate from a list of nine. Six direct: (1) blame the chief, (2) threaten the chief, (3) refuse collective labor, (4) refuse to work on the chief’s farm, (5) refuse marriage to the chief’s family, (6) refuse to pay local tax. Three indirect: (7) complain to an elder, (8) complain to the section chief, (9) complain to the paramount chief. Item order was randomizedFootnote 12,Footnote 13 ;
-
• Outcome 5 (main): Number of those sanctions respondents feel able to personally undertake.
The list of sanctions was developed from a literature review and field consultations with two groups (10 and 20 participants). The literature provided examples of citizen-led sanctions and identified sources of chiefs’ authority, which we reframed as potential leverage points. Key informants reviewed and refined this list for local plausibility (See Appendix A for a formal definition of each sanction). The open-ended responses (Outcome 3), categorized in Appendix H, confirmed coverage of most relevant actions. Additional responses – for example, “mobilize citizens” (3%), “report to police” (4%), “go to court” (2%), and “report to NGO” (2%) – were excluded due to low frequency.
We distinguish between perceived legitimacy (Outcome 4) and self-reported ability (Outcome 5) for two reasons. First, given the influence of chiefs in southern Sierra Leone, respondents may hesitate to express direct intentions; therefore, questions about legitimacy provide a less confrontational means to gauge preferences. This approach helps address concerns about social desirability bias. Second, little is known about non-electoral sanctions in traditional institutions; the legitimacy question maps the repertoire of politically acceptable responses. The ability question captures individual agency to act on these preferences. Appendix D details the measurement strategy and summary statistics. To test Hypotheses 1 and 2, we aggregate counts of direct and indirect sanctions.
Because the legitimacy and efficacy measures referenced that “some members of the village reacted in this way,” it is possible that this phrasing introduced subtle demand effects by signaling socially acceptable responses. Nevertheless, it is unlikely to vary systematically between Treatment 1 and Treatment 4, where only the quantity of money taken changes. We develop sensitive analysis to demand effects in the analysis section.
Empirical strategy
We estimate the effect of providing information about leaders’ malevolence on attitudes towards the legitimacy of sanctioning behavior with an average treatment effect estimand. As there is covariate balance between the control and the treatment group, we use the following estimator for the respondent
$j$
:
$$GS{I_j} = {\beta _0} + {\beta _1}{Z_j} + {\gamma _g} + {\gamma _v} + {\varepsilon _j}$$
Where
$GS{I_j}$
, the outcome variable, is the number of direct or indirect sanctions chosen by the respondent
$j$
,
${\beta _1}$
is the Average Treatment Effect, and
${Z_j}$
is a dummy variable indicating whether the participant
$j$
belongs to Treatment 1 or the control group, and to Treatment 4 or the Treatment 1 for the Hypothesis 1 or the Treatment arm 3 or 2 for the Hypothesis 2.
${\gamma _g}$
and
${\gamma _v}$
are gender and village-fixed effects accounting for the block randomization strategyFootnote 14. We use robust HC2 standard errors (Aronow & Middleton Reference Aronow and Middleton2013). The robustness checks include the pre-analysis plan strategy (excluding experimental block fixed effects), clustering standard errors at the village level, adding covariates that were not initially balanced, and excluding participants who did not fully understand the scenario (Appendix J). We also conducted p-value corrections for multiple hypothesis testing (in the main text) and performed a sensitivity analysis to account for potential experimental demand effects. Results shown in the main text for Hypotheses 1 and 2 are robust to all those alternative strategies. The experiment is well-powered when comparing Treatment 1 with the control group, or Treatment 4 with Treatment 1. Power reaches 51% when comparing Treatment 3 with Treatment 2.
Results
The experimental conditions represent a clear breach of community norms. As Figure 1 shows, the proportion of respondents who disagree with the chief’s behavior increases with the magnitude of the offense, reaching 70% in Treatment 1 (
$\beta = 0.7$
,
$p \lt 0.001$
) and 80% in Treatment 4 (
$\beta = 0.1$
,
$p \lt 0.001$
). However, this proportion does not vary with the elders’ behavior (
$\beta = 0.01$
,
$p = 0.377$
). Furthermore, the proportion of respondents who believe that citizens should take action against the town chief also increases with the magnitude of the offense, reaching 74% in Treatment 1 (
$\beta = 0.74$
,
$p \lt 0.001$
) and 86% in Treatment 4 (
$\beta = 0.12$
,
$p \lt 0.001$
). This suggests that respondents want the norm to be enforced. Yet, this proportion likewise does not vary with the elders’ behaviorFootnote 15 (
$\beta = 0.03$
,
$p = 0.277$
).
Estimated treatment effects: disagreement with chiefs’ behavior and citizens’ actions.
Notes: The figure presents estimated average treatment effects for two outcomes: (i) disagreement with the chiefs’ behavior (a five-point Likert scale rescaled to range from 0 to 1) and (ii) whether citizens believe they should take any action (binary). The comparisons shown are Treatment 1 vs. Control, Treatment 2 vs. Treatment 3, and Treatment 4 vs. Treatment 1. Dots represent point estimates, thick bars indicate 95% confidence intervals, and thin bars indicate 90% confidence intervals. Treatment 1 corresponds to a scenario where chiefs steal a very small portion of development funds; Treatment 2 corresponds to a scenario where elders blame the chief after misappropriation; Treatment 3 corresponds to a scenario where elders do not blame the chief; and Treatment 4 corresponds to a scenario where half of the funds are stolen.
A wide range of sanctions is seen as legitimate
As shown by Figure 2, respondents support sanctioning leaders via multiple channels and adjust their responses based on the magnitude of the offense, confirming Hypothesis 1. Comparing Treatment 1 (minor theft) to the control reveals a significant increase of 3.04 (
$\beta = 3.04$
,
$p \lt 0.001$
) in the number of sanctions deemed legitimate (see Figure I.1 for breakdowns by sanction). Disaggregated effects show increases of 1.2 for direct sanctions and 1.9 for indirect sanctions. Comparing Treatment 4 (major theft) to Treatment 1, the number of legitimate sanctions rises by 0.54 (
$\beta = 0.54$
,
$p \lt 0.001$
), 0.27 for direct and 0.27 for indirect sanctions. Respondents are more likely to endorse a range of sanctions as legitimate when the theft is more severe. These findings remain robust when the alternative outcome – the number of feasible sanctions – is usedFootnote 16.
Estimated treatment effects on the number of legitimate and feasible sanctions.
Notes: The figure presents estimated average treatment effects for two outcomes: (i) total number of legitimate sanctions, the number of direct and indirect legitimate sanctions against the village chief and (ii) the total number of feasible sanctions, the number of direct and indirect feasible sanctions against the village chief. The comparisons shown are Treatment 1 vs. Control, Treatment 2 vs. Treatment 3, and Treatment 4 vs. Treatment 1. Dots represent point estimates, thick bars indicate 95% confidence intervals, and thin bars indicate 90% confidence intervals. Treatment 1 corresponds to a scenario where chiefs steal a very small portion of development funds; Treatment 2 corresponds to a scenario where elders blame the chief after misappropriation; Treatment 3 corresponds to a scenario where elders do not blame the chief; and Treatment 4 corresponds to a scenario where half of the funds are stolen.
Figure I.1 in Appendix I further explores sanction preferences. The most commonly supported actions – endorsed by about 75% of respondents in treatment groups – involve appealing to higher authorities (elders, section chiefs, or the paramount chief). Blaming the chief directly or refusing to work on their farm are also frequently chosen, but to a lesser extent. In contrast, the other direct sanctions – such as physical threats, tax refusal, or rejecting marriage ties – are rarely seen as legitimate, despite occasional references in the literature (Richards Reference Richards2021; Bulte et al. Reference Bulte, Richards and Voors2018).
Appendix J. 6 tests the hypotheses using the proportion of direct and indirect sanctions chosen. It reveals a statistically significantly larger effect size for the indirect-sanctions outcomes. Appendix K includes a formal statistical test that confirms the preference for indirect sanctions. This evidence is also supported by the open-ended survey question. A random subset of respondents was asked how they would respond in such a situation, and all indicated an indirect approach: they would complain to a tribal authority, an elder, or the section/paramount chief. The low incidence of direct sanctions can be equally explained by interpersonal conflict avoidance and by villagers’ high baseline satisfaction (94%) and trust in their town chiefs – who are often the most accessible and responsive authorities – rather than an unwillingness to act.
Elders’ involvement increases perceived legitimacy
Contrary to expectations, respondents do not view fewer sanctions as legitimate when elders have already blamed the chief. Instead, they are more likely to regard a wider range of sanctions as legitimate in such cases. Comparing Treatment 2 (where elders blamed the chief) and Treatment 3 (where they did not), the number of legitimate sanctions increased by 0.37 (
$\beta = 0.37$
,
$p = 0.028$
), driven primarily by a 0.24 increase in indirect sanctions. Appendix J.6 robustly reveals a statistically significantly larger effect size for the indirect-sanctions outcomes. While direct sanctions also rose, the effect was not statistically significant. These findings are robust when restricted to villages with a council of elders (around 90% of the sample). In the Treatment 2 (where elders blamed the chief) vs. Treatment 1 (which made no mention of elder behavior) comparison, effects are similar: 0.38 for total sanctions (
$\beta = 0.38$
,
$p = 0.028$
), 0.24 for direct (
$\beta = 0.24$
,
$p = 0.028$
), and 0.14 for indirect (
$\beta = 0.14$
,
$p = 0.13$
).
We identify the following two mechanisms: the information-legitimation mechanism, where elders’ actions against the chief legitimize villagers’ grievances, and the covering mechanism, where collective actions reduce retaliation risk. Appendix M displays formal evidence from statistical tests and from open-ended responses and suggests that both mechanisms are at play, with elders’ influence playing a key role in shaping villagers’ willingness to challenge the chief and maintain social order. These findings contribute to a growing body of research demonstrating that traditional elites can promote mobilization and cooperation by rewarding such behavior (Goist & Kern Reference Goist and Kern2018), while also shaping public opinion (Yekple & Mitkov Reference Yekple and Mitkov2024) and influencing political behavior (Brierley & Ofosu Reference Brierley and Ofosu2024).
Legitimacy of sanctions differs by social status
As shown in Figure 3, treatment effects vary systematically across subgroups. Comparing Treatment 1 to the control, men display larger increases than women in both direct and indirect sanctions considered legitimate, confirming that gender moderates sanctioning responses. Income further shapes responses: higher-income respondents (Q3–Q4) report weaker increases in indirect and direct sanctions deemed legitimate than poorer respondents, with a difference of about 1.0 for both direct and indirect sanctions between the top and lowest quartiles. This may suggest that wealthier individuals are more tolerant of small theft. Voting rights for the town chief are also a strong moderator: respondents with voting rights show significant increases in both direct and indirect sanctions considered legitimate, with large (about 1.0) and statistically significant differences between these two groups. By contrast, comparisons between Treatment 4 and Treatment 1, and between Treatment 2 and Treatment 3, yield no systematic heterogeneity across subgroups. These findings suggest that socioeconomic and political status influence how citizens view the legitimacy of sanctions against chiefs.
Heterogeneity analysis of the estimated treatment of effect on the number of legitimate sanctions by gender, income, and voting rights.
Notes: Panel A presents the average values of total, direct, and indirect sanctions considered as legitimate in the control, Treatment 1, and Treatment 4 conditions, with the dots representing these averages. The bars illustrate two standard errors. In Panel B, the figure displays the average legitimate sanctions per sanction type and treatment condition. Since the outcome in this case is a binary variable, no standard errors are shown. Treatment 1 corresponds to the scenario where the chiefs steal a very small portion of development funds, while Treatment 4 represents a scenario where half of the money is stolen.
Our experimental design does not allow us to draw substantial conclusions from the heterogeneity analysis. However, if town chiefs are aware of and responsive to these potential sanctions, differences between subgroups may help explain why current local political choices tend to favor men and citizens with traditional political rights (insiders).
Limitations
We test the robustness of our results to experimenter demand effects, which occur when subjects infer the expectations of researchers and adjust their behavior accordingly (Mummolo & Peterson Reference Mummolo and Peterson2019). Our survey experiment was embedded within a one-hour household survey primarily focused on understanding livelihood activities and deforestation patterns in these communities. This design provides strong reason to believe that respondents were unlikely to discern the experimenter’s intentions, thereby reducing the likelihood of such effects. Nonetheless, our empirical strategy addresses this concern through a bounding estimation strategy. Previous studies have found demand effects to be either negligible (Mummolo & Peterson Reference Mummolo and Peterson2019) or relatively small (de Quidt et al. Reference de Quidt, Haushofer and Roth2018). We apply a bounding estimation strategy, accounting for these effects using the 0.1 to 0.3 standard deviation range reported by de Quidt et al. (Reference de Quidt, Haushofer and Roth2018). The results are displayed in Appendix J. Sensitivity analysis of Hypothesis 1 demonstrates that the results are robust to all experimenter demand effect sizes, while for Hypothesis 2, small experimenter demand effects could explain the findings. Furthermore, because our design does not include a condition where elders’ behavior is described in the absence of chief misappropriation, we cannot fully disentangle the independent effect of elders’ actions from the possibility that longer vignettes themselves cue participants to reflect more carefully.
Conclusion
This paper investigates how villagers in Sierra Leone perceive the legitimacy and feasibility of non-electoral sanctions against village chiefs. Using a survey experiment, we find that indirect sanctions – such as reporting to higher authorities – are viewed as more legitimate and feasible than direct actions. The range of sanctions considered acceptable increases with the severity of the chief’s offense, and preferences vary by gender, income, and voting rights: men, lower-income individuals, and those with voting rights endorse a broader set of sanctions. Elders also play a key role; when they publicly criticize a chief, villagers expand the set of sanctions they deem acceptable.
These findings suggest that even where electoral mechanisms are rare or absent, villagers perceive avenues to discipline chiefs, primarily mediated through hierarchical institutions rather than direct action. While consistent with accounts emphasizing chief accountability to higher authorities (Labonte Reference Labonte2012; Bulte et al. Reference Bulte, Richards and Voors2018; Acemoglu et al. Reference Acemoglu, Reed and Robinson2014), our results show that community members retain some agency, leveraging hierarchical relationships or occasionally imposing direct sanctions such as refusing to farm on the chief’s land. Whether such sanctioning norms are effective in promoting downward accountability remains an open question.
The study also contributes to debates on accountability under traditional authority (Baldwin Reference Baldwin2025), highlighting that villagers recognize a wide range of legitimate sanctions and that elders function as political intermediaries, legitimizing grievances and reducing sanctioning risks (Baldwin et al. Reference Baldwin, Muyengwa and Mvukiyehe2022).
These results should be interpreted cautiously, as they reflect responses to hypothetical misappropriation and may not generalize to other types of offenses. Endorsement of sanctions may not translate into implementation, which depends on power dynamics or coordination. Future research could examine links between perceived legitimacy and actual sanctioning, explore other governance systems, and investigate how gender, wealth, and voting rights shape power and inequality.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/XPS.2025.10025.
Data availability
Funding was received from the green transition scheme of the International Growth Center. The data, code, and any additional materials required to replicate all analyses in this article are available at the Journal of Experimental Political Science Dataverse within the Harvard Dataverse Network, at: https://doi.org/10.7910/DVN/K8QXXV (Chazottes & Nabieu Reference Chazottes and Nabieu2025).
Acknowledgements
We thank Alphious Jalloh, Emmanuel Saffa, Jeremaiah J.P. Simbo, John Jusu, Kadiatu Kanneh, Margret Kawa, Mariama M Barrie, Mohamed Lawan, Samai Lahai, Sayo K. Mansaray, Sheku Bengura, and Tommy Barley for excellent research assistance. We are strongly thankful for the logistical support and feedback received from Niccolo Meriggi, and the Wageningen University teams in Sierra Leone. We have received valued feedback from Maarten Voors, Miriam Golden, Susanna Garside, Virginia Rocha, Daniel Goldstein, and the other attendees of the Political Behavior Colloqium at the European University Institute, and Michaël Aklin at EPFL Lausanne.
Competing interests
The authors have no competing interests to declare.
Ethics statement
The research protocol for this study received approval from the Office of Sierra Leone Ethics and Scientific Review Committee (SLESRC), under protocol number n020/04/2023. A copy of the official approval letter has been uploaded to the editorial management system (EM).
This research was conducted in adherence to the American Political Science Association (APSA)’s Principles and Guidance for Human Subjects Research.
Enumerators have first read an Informed Consent form to the respondents and ensure that they had read the entire informed consent, understood it, and agreed before proceeding. Respondents were informed that their participation was completely voluntary and that they were allowed to withdraw from the study at any time without any impact on their relationship with the academic institution. Respondents were also informed that they were able to abstain from answering any question. This informed consent document also informed respondents that there were no known risks or direct benefits of the study. Respondents were assured that no identifying information about them would be made public and any views they expressed would be kept completely confidential. A full copy of the informed consent is available by contacting the authors.
The studies did not use deception in this experiment. Nor did the survey experiment intervene in political processes. Respondents were given the field coordinator and the IRB’s contact information for any concerns or questions. Respondents were not compensated for their participation in the study.
Open access
