The World Bank Economic Review, 39(2), 2025, 341–361 https://doi.org10.1093/wber/lhae026 Article Chiefs, Courts, and Upholding Property Rights: Quasi-Experimental Evidence from Sierra Leone Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Henry Musa Kpaka Land disputes are unavoidable and costly to resolve in the formal courts in contexts with weak property rights and low state capacity. In order to relax the pressure on strained formal courts, many countries permit parallel informal dispute-resolution forums. This paper studies the extent to which one such forum—Chiefdom Land Committees (CLCs)—in Sierra Leone is able to resolve land disputes. This paper constructs a data set of ligated cases at local courts across the country and implements a difference-in-difference design to estimate the effect of the CLCs on land caseload in the formal courts. Contrary to the policy goals, this paper finds that on average, chiefdoms with CLCs have higher land caseload in the formal courts three years on. By adopting the CLCs, chiefdoms plausibly made land issues more salient, but, instead of providing final resolutions, CLCs are conduits for the formalization of land disputes. JEL classification: D02, D04, D74, D78, O17, P16 Keywords: state capacity, property rights, dispute resolution, informal institutions, difference-in-difference 1. Introduction In most of sub-Saharan Africa, property rights over land are less protected (World Bank 2007). Land ownership and usage rights are less individualist and are based on customary laws that are typically unwritten (Toulmin 2009; Boone 2013). Even where formal laws allow individual rights, these are often poorly defined. This generally discourages investments, depresses agriculture productivity (Besley and Ghatak 2010; World Bank 2007; Feder and Feeny 1991; Deininger and Chamorro 1999) and many times gives rise to land disputes, some of which can lead to violent conflicts (van der Haar and van Leeuwen 2019; van Leeuwen and van der Haar 2016; ; Huggins 2009). Pervasive lack of capacity in state legal institutions further hinders upholding any rightful claims to land. Claimants that take land disputes to the formal courts find that the court systems lack resources, are typically not trustworthy, and are costly to access (Deseau et al. 2019; Price 2018; Logan 2017). As a claimant do you then go to informal parallel dispute-resolution forums? But these forums are typically captured by local elites who may not be impartial arbiters in land dispute resolution (Unruh and Turray 2006; Goldstein and Udry 2008; Hartman, Blair, and Blattman 2018). Henry Musa Kpaka is the Minister of Agriculture and Food Security, Freetown, Sierra Leone; his email address is henrymkpaka@gmail.com. The research of this article was financed by the International Growth Center (IGC). The au- thor thanks Professors Stephane Wolton, Joachim Wehner, and Elliott Green for insightful comments and feedback on earlier drafts of this paper. A supplementary online appendix is available with this article at The World Bank Economic Review website. C Crown copyright 2024. This article contains public sector information licensed under the Open Government Licence v3.0 (https://www.nationalarchives.gov.uk/doc/open- government- licence/version/3/). 342 Kpaka In order to help claimants and to protect and uphold property rights, one approach states pursue is to focus on building the capacity of state legal institutions and make them more trustworthy by providing people with truthful information about improved service delivery (Acemoglu et al. 2020). Another is to educate individuals to privately and peacefully resolve their disputes while minimizing state engagement in dispute resolution (Blattman, Hartman, and Blair 2014; Hartman, Blair, and Blattman 2018). This paper investigates whether a hybrid state–customary dispute-resolution mechanism can be a viable option to resolve land disputes, and by extension, help uphold land property rights. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 This paper examines this question in the Sierra Leone context where land disputes in rural areas are a thorny issue for the government and the formal court systems. Rights over land in these areas are based on unwritten customary rules and norms that vary in different parts of the country. Customary rights are legally recognized, although such rights are not adequately protected as properties and associated rights are not registered (Barrows 1974; Unruh and Turray 2006; Renner-Thomas 2010; Johnson 2011; Ryan 2018). At the same time, the formal court system in Sierra Leone is marked by long delays, high costs, and mistrust. This is especially the case for the lowest courts-the local courts-that are grossly underresourced, lack trained personnel and are unable to meet the demand for justice in their jurisdictions (Suma 2014; Park 2008). Local courts are critical for upholding and protecting rights for land in rural areas. Most land cases that make it to the superior courts in the provinces originate from the local courts and remain unresolved for years. Furthermore, various analyses of the causes of the decade-long civil war in the country point to economic resources, such as land, captured by the traditional authorities that alienated the youths and other vulnerable groups (Abdullah et al. 1997; McIntyre, Aning, and Addo 2002; Richards 2005; Fanthorpe 2006), which undermines traditional modes of dispute resolution. As part of the 2015 National Land Policy, the government proposed the introduction of Chiefdom Land Committees (CLCs) to administer rural land in an equitable manner, and to help resolve land disputes among community members through third-party mediation. The policy does not change, but instead builds on the existing customary laws that govern land administration and land dispute resolution. The CLCs can be best viewed as a hybrid state–customary order in the spirit of Boege, Brown, and Clements (2009), as the CLCs get legitimacy and support from the state. The state also saw the CLCs as a way to emphasize principles of fairness, such as deliberation and inclusiveness, in customary ways of land dispute resolution. The policymakers expected the CLCs to reduce land cases that end up in the formal court system, through two possible channels: preventing disputes in the first place because of better land administration, and mediation through CLCs for disputes that arise anyway. The focus of the empirical analysis is whether this hybrid state–customary land-dispute-resolution forum has any effect on land cases litigated at the formal courts. This study surveyed all the chiefdom administrations in early 2019 and found that only 51 out of 149 chiefdoms had adopted CLCs. This study uses this variation in the policy adoption, and 10 years of local court records (from 2009 to 2018) in a difference-in-difference design to estimate the average treatment effect (ATT) of CLCs on both the likelihood of observing land cases, and the volume of land cases litigated at the formal courts. The results suggest that on the extensive margin, formal courts in non-compliant chiefdoms were just as likely to hear land cases as those in chiefdoms with CLCs. However, at the intensive margins, formal courts in chiefdoms with CLCs saw higher land caseloads on average (δ = 1.76, s.e = 0.71) than formal courts in non-compliant chiefdoms. This effect is large given that the mean number of cases in the formal courts is 1.28 per year for the study period. The program effect holds after a series of robustness checks. This paper exploits the time series nature of the outcome variable to conduct falsification tests and show that for each year prior to the policy implementation, there was no effect on the number of land cases in the formal courts. In addition, because the policy targeted land disputes, the study uses the number of other civil cases (non-land) as a placebo outcome and the results show that the policy had no effect on these types of cases. Finally, the paper The World Bank Economic Review 343 uses the total number of all cases, and the case types, in a triple difference specification, and the result is consistent with the difference-in-difference estimation. The paper explores three possible channels to further understand this observed effect. The first channel assesses whether the land reform and activities of the CLCs brought to the surface pent-up land concerns among vulnerable groups like strangers1 or from chiefdoms that are more prone to have land disputes, for instance because they have relatively high degrees of unemployment or are close to urban centers (Nuhu 2019; Lombard 2016). This analysis does not find support for this argument. The policy did not seem to Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 have had any effect on chiefdoms that adopted CLCs and have a higher proportion of strangers, higher unemployment, or are closer to major towns. Secondly, the paper explores whether, rather than solving land disputes, the CLCs are instead conduits of land cases to the formal court system. This can happen via interaction with CLCs, where the CLCs refer cases to the formal courts. In addition, a potential drawback to informal or alternative dispute resolution is that resolutions are often perceived as less final (Crook 2004), so that people that are unsatisfied with CLC resolution may proceed to formal courts with their land cases. The third channel explored is whether the observed effect is a result of the CLCs making land issues more salient. The idea is that the CLC activities may educate people of their rights over land, and those that can respond, for example the more educated, or people with relatively higher income, can take their land cases directly to the formal courts. There is merit to this mechanism because, despite the stated policy goals, the data collection revealed that the actual number of land cases in the local courts is a magnitude lower than non-land cases. This does not necessarily mean that the prevalence of land disputes is low, but it is more likely that people do not seek justice or may go to other dispute-resolution mechanisms. By making land dispute resolution more salient, the policy may prompt people to bring their disputes to the courts, and hence increase land caseload in treatment chiefdoms. There is some evidence to support both of these latter explanations. For instance, chiefdoms that adopted the CLCs and potentially have higher income or secondary education tend to drive the main result. Similarly, of the chiefdoms that adopted CLCs, the ones that said they referred land cases to either the local or magistrate’s courts also had higher land caseloads in the local courts. These findings suggests that these hybrid forums are more likely a link for state formalization of disputes rather than forums of resolution. This paper directly contributes to the debate about how to uphold and secure rights over land in con- texts where property rights are weak (Crook 2004; Toulmin 2009; Collins and Mitchell 2018; Unruh and Turray 2006). While informal alternative dispute-resolution mechanisms, primarily provided by non-state actors, are the predominant way people address everyday disputes (Logan 2017), the evidence presented here suggests that for land disputes, even when the state sanctions informal dispute-resolution channels, people might still prefer to bring their cases to the formal court system. This may be because resolution in the formal systems might be perceived as final or at least more permanent than they will get from the informal system. This is similar to the finding from Ghana as Crook (2004) shows from his survey of litigants. It also contributes to the broader debate about how to provide critical public goods, such as access to justice, in developing countries. To this end, the merits and demerits of having plural justice systems is dis- cussed in the literature (D’Aoust and Sterck 2016; Tamanaha 2011; Swenson 2018; Chirayath, Sage, and Woolcock 2005). While evidence exists that they help extend access to justice (Price 2018), little evidence exists about their direct impact on the capacity of formal justice systems. In the Sierra Leone context, and with land cases, I show a preference for formal courts, where people bring their land cases to the formal courts perhaps because they are unsatisfied with resolutions from the informal mechanism or because 1 Strangers are people born outside the chiefdom and have no claim to land. They must seek permission from a paramount chief to access land for cultivation and livelihood (Tangri 1976; Fenton 1951). 344 Kpaka informal forums motivated them to directly seek justice in the formal courts. In a similar vein, Acemoglu et al. (2020) show that by providing truthful information about formal courts in Pakistan, people switch from informal forums to the state forums in addressing disputes. They argue that “motivated reasoning,” where a positive experience with the state courts, not only changes people’s beliefs about the state institu- tion, but this mechanism also encourages citizens to use the formal state processes. My findings in relation to Acemoglu et al. (2020) suggest a potential trade-off between investing in informal dispute-resolution channels versus working to build the capacity of formal state courts. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Finally, this paper is among the first to make a direct empirical link between a land reform and land disputes litigated in formal courts. The vast majority of the literature on the impact of land reforms focuses on economic outcomes such as investment and productivity (Besley and Ghatak 2010; World Bank 2007; Feder and Feeny 1991; Deininger and Chamorro 1999), while ignoring reform effect on land disputes. The rest of the paper proceeds as follows: The Context Section describes the land tenure system and the policy intervention to extend land tenure, and to prevent and resolve land disputes. The subsequent section describes the data and empirical strategy of the research. The results section presents the main results starting with the average treatment effect. In the Section on Robustness Checks, the paper conducts placebo tests and other sensitivity analyses. Next, the paper offers possible explanations for the main result findings, and the final section discuss why the results of the study matter for how to protect and uphold property rights in state with weak capacity and customary land tenure system. 2. The Context 2.1. Land Tenure in Sierra Leone In Sierra Leone, the institution that governs land in rural areas is customary law. These are undocumented informal laws that vary from one locality to the next (Kanu and Henning 2019; Oredola-Davies 2006). In the absence of formal land demarcation, registration, and titling, claims and counterclaims over rural land have given rise to considerable disputes over farm land in rural areas in post-conflict Sierra Leone (Unruh and Turray 2006). These disputes can range from minor altercations among rival claimants, over particular pieces of land, to violent clashes against foreign land deals in the chiefdoms. As custodians of rural land, paramount chiefs (PCs) are key players in preventing and resolving disputes over land. No meaningful land transaction is completed in the chiefdoms without the stamped approval of the PCs, even when it involves the state (Ryan 2018; Johnson 2011; Bottazzi, Goguen, and Rist 2016). However, paramount chiefs may not be impartial arbiters in resolving disputes over land in rural areas. Land is frequently used as a political and economic tool by chiefs. The capture of critical resources such as land is argued to have formed the foundation of the decade-long civil war in the country (Jackson 2007; Unruh and Turray 2006; Sawyer 2008; Richards 2005). 2.2. The 2015 National Land Policy Reform A key policy reform effort after the war aimed to secure rights over land for the rural populations, par- ticularly for women and the youth, who had been marginalized by local elites in the pre-war era (Sawyer 2008). The 2015 National Land Policy is the result of a second attempt at this goal. The first effort at this goal was initiated in 2003 immediately following the end of the war. But the effort did not succeed at the cabinet level because opponents of the policy argued it had not consulted enough with relevant stakeholders like the paramount chiefs and land-holding families (Government of Sierra Leone 2015). What eventually became the 2015 National Land Policy was extensively consulted across the country, starting as early as 2011. The policy itself is a bundle of interventions in land governance and usage at various levels in the country. In terms of governance, the policy proposed a National Land Commission that is decentralized at the various administrative levels of the country. The Chiefdom Land Committee The World Bank Economic Review 345 (CLC) represents the National Land Commission at the chiefdom level. The focus of the paper’s analysis of the policy is at this level. The CLCs are headed by paramount chiefs, and are to be comprised of landowners and “persons ordinarily resident” in the chiefdoms. The policy also suggests that membership of the CLC must “respect gender, ethnic diversity, and social political dynamic.” Under the policy, the responsibility for communal land is now vested in the CLC. The policy states, “It [the Chiefdom Land Committee] shall vet/approve all land transactions and perform all other functions relating to the disposal of communal land presently Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 performed by the Chiefdoms Councils” (Government of Sierra Leone, 2015, p. 72-73). While the policy did not change the existing dual land tenure system in the country, the intended goal of vesting land decisions in a diverse committee of locals was to reduce the overwhelming influence of the paramount chiefs in land decisions in their chiefdoms. Recognizing that the formal court system is overwhelmed and lacks the capacity to handle most civil disputes, the CLCs were to also act as new forums for dispute resolution for land disputes. It states, “At the chiefdom level the CLC...will develop and maximize opportunities to formalize the application of Alternative Dispute Resolution mechanisms such as negotiation and mediation to reduce the number of cases that end up in the court system” (Government of Sierra Leone 2015, p. 76). In rural areas, the local courts are courts of first instance for civil cases in the formal court system in Sierra Leone (Government of Sierra Leone 1991). Most land cases that get into the formal court system start here. These local courts, which were headed by paramount chiefs until 2011, gained a reputation for being biased and favoring only the local elites in power (Sawyer 2008). With the CLCs, policymakers hoped to ensure fairer outcomes in land dispute resolution, and to reduce caseloads in the formal courts. People in rural areas can report land disputes to identified members of the CLC, who can then bring them to the wider committee for deliberation. Prior to the CLCs, only the paramount chief and a fewer lesser chiefs received land dispute claims at the chiefdom level. 2.3. CLCs as Hybrid Forums Theoretically, the CLCs can be best viewed as a “hybrid order” (Boege, Brown, and Clements 2009). CLCs get legitimacy from the state, unlike various other traditional channels of land dispute resolution. In addition, the CLCs combine customary norms with formal state processes, such as open deliberation, impartiality, and representation of interest groups in land administration and dispute resolution. The expectation is that these “Weberian” state features will help prevent land capture by local elites and lead to a more equitable access to land, which would prevent land disputes. The disputes that inevitably arise can also be resolved satisfactorily through CLC mediation. However, despite these goals, the policy does not specify the actual implementation or day-to-day running of the CLCs. Importantly also, there was no means of enforcement, punishment, or reward for chiefdoms adopting CLCs. In fact, the policy was launched by the president without much administrative resource and finance to implement it (Conteh 2015). It was largely left to the chiefdoms to adopt CLCs and implement them in their own way. The result is a variation in what CLCs do and how they are run. One exception to this variation is that almost all of the CLCs addressed land disputes in the community. While this paper cannot distinguish between these two mechanisms of the policy, the analysis aims to understand the extent to which the policy impacted land caseload in formal courts. 3. Data and Identification Strategy 3.1. Data This study combines two original sources of data for the analyses. A team of experienced enumerators were hired and trained to collect 10 years of administrative records for cases litigated at local courts 346 Kpaka across the country, and to conduct a survey of all the chiefdom administrations to understand whether, and how, chiefdom land committees were implemented. 3.1.1. Land Dispute Litigation In Sierra Leone, each chiefdom has at least one local court, and some chiefdoms have up to four in our data collection. Most rural people are encouraged to address all land disputes through the local courts, although they can also bring the case to higher courts like the magistrate’s courts. This makes local courts an ideal Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 place to capture litigation over land cases across the country. The research team collected information from all functioning local courts in each chiefdom. The research team visited 248 local court locations, but only 199 courts had any data. When aggregated, this includes 130 chiefdoms that had court data for some years in the study period. Panels (a) and (b) in supplementary online appendix fig. S1.1 show the locations of local courts with respect to population density and variation in costs across the country. At the local courts, the main source of information was the case logs. Each court has case logs where the court clerks are supposed to record all cases that are brought to that court each year. These records are all handwritten in a blue ledger that is provided by the government. Supplementary online appendix fig. S1.2 shows a typical case log. Critical information in the case log generally includes case number, the date the case was filed, the names of the plaintiff and defendant, the cause or claim, various cost items to file a case in that court, and the date of hearing. The team aimed to collect information from logs from 2009 to 2018. On average, 3.6 years of records per court, or 5.5 years of records per chiefdom, were collected, making a total of 711 court–year observations in the full data set. For each year record the data count/record the number of all civil cases in a programmed tablet. Next, the paper determined which of the litigations were over land. The type of case is easily obtained from the cause or claim of the filing, but the researchers also worked closely with the court clerk to identify case types. Each logged case also has a separate file with details on testimonies and witnesses, all of which helps determine what the case was about. The research team took photos of recorded pages where they are about land cases, as in the example in supplementary online appendix fig. S1.3(b). Panel (a) in fig. 1 shows the pattern of land cases for the study period. Panels (b) and (c) show the spread of cases post-treatment (2015–2018) and pre-treatment, respectively. Supplementary online appendix table S1.1 provides summary statistics of the outcome variables and other court-related variables. 3.1.2. Low Land Caseloads in the Local Courts It is worth noting at this point that despite the stated policy objective of reducing land cases in the local courts, it turns out that land cases are only a tiny fraction of cases in the local courts. During the pilot phase of this research, the team was told by many court clerks and community leaders that people are encouraged to resolve land cases at a much lower level before bringing them to the local courts. For example, land disputes within families are encouraged to be resolved at the family level. The data collected also confirm this claim. The average court in the full data set receives about 1.3 land cases per year, compared to 51.4 for non-land cases. It does appear that the policy goal with respect to caseload in the courts was not informed by much evidence. At best, one would expect that the policy would have no effect; however, as this study will show, the CLC presence did lead to an increase in land cases in the courts. Whether this is desirable or not very much depends on whether the emphasis is placed on dispute resolution, as intended by the Sierra Leone policymakers, or access to formal courts. The policy may have achieved the latter. 3.1.3. Data Attrition and Balance Concerns As can be seen from the maps in fig. 1, there are some holes in the data, especially in the pre-treatment period, where the data are available only for 58 local courts out of the 248 visited. Some courts also did not have consecutive years of data. This can potentially lead to biases in the average effect of the The World Bank Economic Review 347 Figure 1. Spatial Pattern of Land Cases and Treatment Locations. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Darker colors indicate higher numbers of cases. Dots are the locations of local courts. White spaces in panels a, b,and c indicate missing data. intervention if observations are not balanced between treatment and controls groups. In supplementary online appendix tables S1.2 and S1.3, with the exception of 2009, it is shown that there is some degree of balance in the proportion of observations for each year between treatment and control, especially for the latter years. In addition, at the chiefdom level, there is at least one observation per year between 2010 and 2018 for 130 chiefdoms. To further address potential concerns from data attrition and balance in the estimation, a restricted sample of the data from courts with consecutive years of observations that span at least one pre-treatment period for both treatment and control was used initially. This ensures balance in treatment and control. This data restriction was relaxed, and the full data set was used. The estimates were compared from the restricted sample to get a sense of bias implications for population average effects. 3.1.4. Chiefdom Administration Survey In early 2019, the research team conducted a survey of all 149 chiefdoms to investigate which chiefdoms had land committees and how they operated.2 We interviewed key officials on the chiefdoms adminis- tration who would know whether their chiefdom had CLCs. Most of the officials we interviewed would participate on CLCs if their chiefdom adopted the policy. See supplementary online appendix table S1.4 for the types of respondents we interviewed. Only 51 chiefdoms reported having CLCs. Tables S1.5, S1.6, and S1.7 provide information on membership selection, composition of CLCs, and case mediation activ- ities, respectively of the CLCs. Unsurprising also is the variation in how the policy is being implemented 2 Some chiefdoms were split up in 2017 and increased the total number of chiefdoms to 191. But by the time we conducted the survey, most of the newly created chiefdoms did not yet have a chiefdom administration, and more importantly local courts, which were the points of data collection. Hence we use the old administrative divisions in this study. 348 Kpaka for those that adopted it. For instance, about 36 percent of CLCs charge fees to hear land cases. In terms of membership composition, less than half of chiefdoms include vulnerable groups such as youth and women as shown in table S1.6. The way members of the CLCs are selected also varies as shown in table S1.5. One crucial exception to this variation is that almost all of the CLCs (over 98 percent as shown in table S1.7) play a mediation role in resolving land disputes among community members. In this paper, the main treatment is the presence of a CLC in the chiefdoms. This study also explores variation in how Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 CLCs are run, for instance whether chiefdoms with CLCs that refer cases to the court system are more likely to contribute to cases litigated at the local courts. 3.1.5. Other Data Sets The analysis of the drivers of the observed effect uses the 2015 census to obtain the chiefdom-level pro- portion of primary education attainment and proportion of employment in the non-agriculture sector. It uses the mean distance from the chiefdom centroid to the nearest major town, which was obtained from the Acemoglu, Reed, and Robinson (2014) data set. 3.2. Identification Strategy As noted above, some chiefdoms failed to implement CLCs three years after the policy. An implicit as- sumption this paper makes is that those chiefdoms that had CLCs in 2018 had them for some or all of the post-treatment period, whereas those chiefdoms that did not have CLCs by 2018 never adopted them. This allows me to create a control and treatment group and utilize the longitudinal nature of the court data in a difference-in-difference (DiD) design to estimate the causal effect, given by the average treat- ment effect (ATT). An immediate concern with the main treatment is that there might be some systematic reason why some chiefdoms have CLCs and not others. I show spatially in panel (d) in fig. 1 locations that adopted the policy. There is a slight density in the north of the country, where the incumbent govern- ment that passed the policy has a stronghold. However, such a selection is highly likely to be unrelated to the studied outcomes. Table S1.9 shows results from a linear probability model of potential determinant policy compliance. It suggests that the presence of mining activities in the chiefdom is positively associ- ated with CLC adoption, whereas distance to the nearest major towns3 is negatively associated with CLC adoption. I control for these with chiefdoms and court fixed effects. Difference-in-difference estimation is the most appropriate in the evaluation of large-scale policy pro- grams if longitudinal data exist for outcomes before and after the intervention for both the control and treatment groups (). The critical identifying assumptions in this design are that there is a counterfactual parallel trend in the outcome for the treatment and control groups, and that the allocation of treatment assignment is unrelated to outcomes at the baseline. When these assumptions hold, the DiD estimator removes biases in post-treatment-period differences between the treatment and control groups that could be the result of inherent differences between those groups, as well as biases from differences over time in the treatment group resulting from trends due to other causes of the outcome (Wooldridge 2002; Angrist and Pischke 2008). Visual inspection of fig. 2 appears to show that the parallel trend assumption holds for the restricted data, where data include only courts with five consecutive years of data. Parallel trends for unrestricted data are shown in panel (a) in fig. S1.4. The trend is stronger when all the data are used. Panel (b) in the same figure shows the trend for all cases. 3.3. Estimation The effect of the intervention was estimated by the following least squares regression model: yict = αc + βt + δ Ict + cXict + εict , (1) 3 The towns include provincial headquarter towns of Bo, Kenema, Makeni, and the capital Freetown. The World Bank Economic Review 349 Figure 2. Visual Inspection of Parallel Trends with Restricted Data. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Vertical axis is the number of land cases per year in each local court in treatment and control chiefdoms. where yict is the outcome variable for court i in chiefdom c at time t. When the outcome is a binary, a linear probability model was used. The variable α c is the court fixed effects, and β t is year fixed effects; Ict is an interaction dummy for treatment in chiefdom c, and post-treatment period t. The variables Xict are relevant court-level controls, such as costs of filing cases, and finally the error term is given by ε ict . The coefficient of interest is δ , which is the average treatment on the treated (ATT). 4. Results—Average Treatment Effects This section presents the main results of the analysis. It presents two sets of results, one with the restricted sample that has consecutive years of court data from 2013 to 2018, which spans three pre-treatment years, inclusive, and for the full data. The idea is to compare the result from the restricted data that are a balanced sample in treatment and control, with results from the full sample, where estimates may be biased as a result of data attrition. It begins with a graphical representation of a simple difference of the outcomes, and then proceeds with the estimation results. 4.1. Simple Differences Panel (a) in fig. 3 below shows results for the extensive margin. The outcome is an indicator variable that takes value 1 if the case type is a land case, and 0 otherwise. In this way, the outcome can be interpreted as the likelihood of observing land cases at the formal courts. It shows graphically that although there is a difference between treatment and control in the pre-treatment period, this difference does not change in the post-treatment period, suggesting that the program had no effect at the extensive margin. Panel (b) captures the intensive margin, which is the number of land cases in local courts per year. As noted earlier the land caseload in courts was low before the reform and picked up after, but generally, land-case levels in courts remain low. Panel (b) shows that the difference between control and treatment chiefdoms in the pre-treatment period grows in the post-treatment period. The change in difference is about 1.5 land cases. It shows a similar result for the unrestricted data in supplementary online appendix fig. S1.5. 350 Kpaka Figure 3. Simple Difference of Outcomes—Restricted Data. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Source: Author’s analysis from local courts’ administrative data. Note: In (a), bars are average of the indicator variable that take 1 if case type is land, and 0 otherwise between the control and treatment chiefdoms. In (b) bars represent simple average for the number of land cases for the treatment and control chiefdoms. Table 1. ATT for the Likelihood of Observing Land Cases—Restricted Data Land case (binary) Variables (1) (2) (3) DiD −0.02 0.07 0.20 (0.10) (0.20) (0.20) Fixed effects Year x x x Court x x x Outcome statistics Mean 0.4 0.35 0.34 Standard deviation 0.49 0.48 0.47 Observations Number of courts 56 39 31 Court-year pairs 272 224 208 R-squared 0.53 0.44 0.49 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Columns 1, 2, and 3 only include 45, 31, and 25 chiefdoms respectively. The treatment group makes up 26 percent, 20 percent, and 25 percent in columns 1, 2, and 3 respectively. Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Average Treatment Effect on the Treated (ATT). 4.2. Estimation Results with Restricted Data 4.2.1. Extensive Margin Table 1 shows results from estimating equation (1) above. The outcome is the likelihood of observing land cases at the formal courts. The model is fitted with year and court fixed effects. The average treatment effect on the treated (ATT) is given by the coefficient on the variable DiD. In column (1), the data are restricted to include only courts that have five consecutive years of data from 2014 to 2018. In column (2), the data are restricted to include courts with six consecutive years of data spanning 2013 to 2018, and in column (3) seven years of data from 2012 to 2018. The coefficients on the DiD estimator are not statistically significant (δ = −0.02, s.e = 0.1), but noticeable also is that the sign changes from negative in column (1) and makes a big positive jump in column (3) where δ = 0.2, s.e = 0.2. These models suggest The World Bank Economic Review 351 Table 2. ATT on Land Cases—Restricted Data Number of land cases Variables (1) (2) (3) DiD 1.47∗ 1.82∗∗ 1.95∗∗ (0.85) (0.80) (0.75) Year fixed effects x x x Court fixed effects x x x Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Outcome mean 1.53 1.26 1.26 Outcome standard deviation 3.26 2.77 2.81 Number of courts 56 39 31 Court-year observations 272 224 208 R-squared 0.48 0.44 0.45 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Columns 1, 2, and 3 only include 45, 31, and 25 chiefdoms respectively. The treatment group makes up 26 percent, 20 percent, and 25 percent in columns 1, 2, and 3 respectively. Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Average Treatment Effect on the Treated (ATT). that the intervention did not have a detectable effect on the extensive margins. A possible interpretation of this result is that the land administration channel through which policymakers expected that CLC intervention would help reduce the risk of land conflict made no difference. 4.2.2. Number of Land Cases Litigated at Local Courts Similarly, table 2 shows results from the main OLS estimation, but now the outcome variable is the number of land cases. The model is fitted exactly the same as in table 1. Columns (1) to (3) have the same restrictions. As can be seen from the table, the ATT of the intervention ranges from δ = 1.47, s.e = 0.85 to δ = 1.95, s.e = 0.75 with the tightest data restriction, which is statistically significant at the 5 percent confidence interval. Noticeable also is that the coefficients and standard errors are stable even as the sample drops. The magnitude of the ATT is also large. For instance, in column (3), it is about 70 percent of the standard deviation of the outcome mean. 4.3. Estimation Results with the Full Data Set In this section the data restrictions are relaxed and the full data set used to run models. On the one hand, the restricted data provide the most robust estimate, but it loses statistical power when looking at potential mechanisms. The data restriction was relaxed to make sure the estimates are not too different from the restricted sample. Table 3 shows the estimation results with the number of land cases as the outcome variable. Column (1) is estimated with year and chiefdom fixed effects without controls. Column (2) introduces court-level controls including the mean cost of using courts, the number of years of court records in each court, and the mean cost in each court. Column (3) adds court fixed effects. The results suggest that at the intensive margin, the intervention increased land cases in the formal courts by 62 percent of outcome standard deviation (δ = 1.76, s.e = 0.71) in column (1), which falls to 53 percent (δ = 1.46, s.e = 0.73) in the most stringent model in column (3). Notable here also is that magnitudes are similar to the results with the restricted data set in table 2. Supplementary online appendix table S1.10 presents the results table for the probability of observing land cases with the full sample. The coefficients are similar to those in the restricted model, but they are also not statistically significant. The fact the magnitude of the estimates from the restricted and unrestricted data for both of these outcomes are similar, assuages potential bias concerns resulting from data attrition. I proceed in the next section with a series of robustness checks of the results for the intensive margins. 352 Kpaka Table 3. ATT on Land Cases—Unrestricted Data Number of land cases Variables (1) (2) (3) DiD 1.76∗∗ 1.65∗∗ 1.46∗∗ (0.71) (0.67) (0.73) Year fixed effects x x x Chiefdom fixed effects x x – Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Court controls – x x Court fixed effects – – x Outcome mean 1.28 1.28 1.28 Outcome standard deviation 2.8 2.8 2.8 Number of chiefdoms 130 129 129 Court-year observations 706 689 689 R-squared 0.36 0.37 0.52 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Court controls include mean cost and number of record years. Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Average Treatment Effect on the Treated (ATT). 5. Robustness Checks To bolster the internal validity of the results above, this section includes three additional robustness tests. It starts with a formal placebo test, using a leads and lags analysis. Secondly, since the policy targeted land cases, it also uses civil cases at the local courts as a second placebo outcome. Finally, it uses the case type to introduce a third dimension of variation in a triple difference specification, where the outcome is now the number of all cases (land and non-land) in the courts. 5.1. Placebo Tests 5.1.1. Leads and Lags ATT I exploit the time series nature of the caseload in courts to conduct a falsification test by estimating the ATT on the treated group for each year data are available. I use both the restricted and full data set. The full data set allowed me to compare treatment and control chiefdoms as far back as 2010. To do this, I use a variation of equation (1) above to estimate the following model: 2018 yict = αc + βt + δt (Ici ) + cXict + εict , (2) t =2009 where δ t is now the ATT for each year t from 2014 to 2018 with the restricted data, and 2009 to 2018 with the full data set. This formulation allows me to show a formal placebo test for the intervention by showing that the ATT for each pre-treatment year is not statistically different from zero. I present a graphical representation of the results below in fig. 4 for the restricted sample. The diamond shapes are standardized coefficients with error bars from running (2) above. From fig. 4, the ATT is statistically indistinguishable from zero for 2014, the first pre-treatment year. In the post-treatment period, the magnitude of ATT gradually increases and dips slightly again in 2018. Supplementary online appendix fig. S1.6 shows the same analysis with the full sample. Equation (2) uses both chiefdom and court fixed effects models in columns 1 and 2 respectively from supplementary online appendix table S1.11. Using the model with the chiefdom fixed effects, the results show that prior to the intervention in 2015 the ATT for each year is statistically zero. Post intervention, the coefficients generally trend positive and are different from zero for two of the three years after the intervention. The picture is similar with the more restrictive model with the court fixed effects. Although the coefficients for the The World Bank Economic Review 353 Figure 4. Formal Falsification Test—With Restricted Sample. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: The y-axis represents standardized coefficients from estimating (2) using the restricted data as shown in model (3) in supplementary online appendix table S1.11. Error bars are for 95 percent confidence intervals. yearly ATT are not statistically significant at the 5 percent level, it is easy to see why the average effect over the three years is statistically significant. 5.1.2. Civil Cases as Placebo The second placebo test was conducted by using the number of other civil cases (non-land cases) as the outcome variable. This additional placebo test addresses possible concerns that courts in treatment chiefdoms are in general more active. If this is the case, one would expect to find an effect even for non- land cases, even though the policy did not target non-land cases. Table 4 presents the results using the full sample. Column (1) estimates are with year and chiefdom fixed effects without controls. Column (2) introduces court-level controls including mean cost of using courts, and number of years of court records in each court. Column (3) adds court fixed effects. From table 4, as expected the coefficients on the DiD estimator are all not statistically different from zero. Although not statistically different from zero, the estimates on civil cases are large and positive. A potential concern may be that part of the effect captured is a result of differential trends in conflict in these chiefdoms. This a reasonable concern given chiefdoms self-selected into the treatment. To alleviate this concern this paper includes a triple difference-in-difference in the next section. 5.2. Triple Difference Estimation Finally, I use the number of all cases (both land and non-land) as the outcomes and introduces a third dimension of variation, using the case types in a difference-in-difference-in-difference (DiDiD) design to help further isolate the treatment effect (Cancian and Levinson 2005). The idea here is to show that the intensity of all cases in formal courts in treatment chiefdoms is higher for courts that recorded any land cases. The indicator variable for land cases is used as treatment. This third dimension of variation addresses potential concerns that the ATT from the difference-in-difference may have resulted from chief- 354 Kpaka Table 4. Placebo Test with Civil Cases Number of other civil cases (non-land) Variables (1) (2) (3) DiD 34.65 39.58 36.30 (23.39) (30.02) (38.39) Year fixed effects x x x Chiefdom fixed effects x x – Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Court controls – x x Court fixed effects – – x Outcome mean 51.38 51.38 51.38 Outcome standard deviation 65.2 65.2 65.2 Number of chiefdoms 130 129 129 Court-year observations 706 689 689 R-squared 0.58 0.60 0.71 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Court controls include mean cost and number of record years. Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. doms that have higher numbers of land disputes adopting the treatment. To do this, it uses the following specification: Yict = αc + βt + γa + δ Ica + ψ Iat + ρ Ict + φ (It × Ic × Ia ) + εict , where Yict is number of cases of all types. The mean number of all cases from the local courts for the study period has normalized the outcome of diving. I do this to bring the order of magnitude closer to that of the number of land cases to make comparison easier. The variable γ a is the case-type fixed effects, and γ a is 1 if the case type is a land case, α c is the court fixed effects, and β t is year fixed effects. The variable Ica is an interaction of treated chiefdom and a dummy for case type, Iat is an interaction dummy for case type and the post-treatment period, and Ict is an interaction dummy for post-treatment and treatment. This is essentially the DiD from equation (1) above. The triple difference is given by It × Ic × Ia , where It is a dummy for the post-treatment period, Ic is a dummy for treatment, and Ia is a dummy for case type, which is 1 when any land case is observed. The variable φ is the estimate for the triple difference. Table 5 shows the result of my estimation. The outcome for all three models is the number of all cases (land and non-land). Supplementary online appendix table S1.12 shows the same results without the normalization. Column (1) presents results for the DiD, and column (2) presents results for the triple difference (DiDiD) with the unrestricted data. In column (3) the results are for the DiDiD with the re- stricted data with a balanced sample in treatment and control. The coefficient on the DiD in column (1) is not statistically significant. In columns (2) and (3) the coefficients on triple difference estimator (DiDiD) are statistically significant at the 5 percent and 1 percent levels, respectively. The interpretation is that treatment chiefdoms with local courts that recorded any land case saw on average a higher number of cases overall. And like with the main result, this increase is larger than the mean number of cases. This result suggests that indeed the intervention had an impact on land cases litigated at the local courts. 6. Possible Explanations for the Observed Effects The analyses so far suggest that the increase in the number of land cases litigated at the local courts is causally linked to the intervention. This effect is also large: over 50 percent of the standard deviation of the mean annual land cases per court in the study period. This finding is the opposite of the stated policy objective, and my research hypothesis going into the study was also aligned with that of the policymak- The World Bank Economic Review 355 Table 5. Triple Difference Estimation Number of all cases (normalized) Variables (1) (2) (3) DiD (post × CLC chiefdom) 0.74 0.17 −0.18 (0.65) (0.44) (0.13) DiDiD (post × CLC chiefdom × case – 1.03∗∗ 1.56∗∗∗ type) Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 (0.48) (0.57) CLC chiefdom × case type (land) – −0.78∗ −1.48∗∗ (0.42) (0.63) Post × case type (land) – 0.07 −0.19 (0.18) (0.23) Case type (land) – 0.11 0.39 (0.12) (0.33) Year fixed effects x x x Court fixed effects x x x Outcome mean 1.00 1.00 1.03 Outcome standard deviation 1.26 1.26 1.34 Number of chiefdoms 130 130 45 Court-year observations 706 706 272 R-squared 0.71 0.72 0.75 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. ers. However, does this opposite effect suggest that the policy was undesirable? To start with, my data suggest that policymakers may have misjudged the land caseload in the formal courts. Panels (a) and (b) in supplementary online appendix table S1.4 show that, although civil cases have been trending upwards, land cases were in fact a small part of this increase. In this period the typical court received about 1.3 land cases compared to about 51.4 non-land cases. Could people be bringing their land cases to the formal courts for the “right” reasons? For example, perhaps the effect of the CLCs is to empower people about their land rights, which prompts them to seek justice. In this section, I explore possible explanations for the observed effects. This section uses the unrestricted data to gain statistical power. As shown in the main results sec- tions, the estimate obtained from using the unrestricted data is similar to that from the restricted data. Furthermore, the explored ideas provide suggestive evidence for the drivers of the observed effect of the intervention. 6.1. CLCs Expose Pent-Up Land Concerns from Vulnerable Groups or Chiefdoms Potentially Prone to More Land Disputes A big part of the land reform policy, and activities of the CLCs, was to administer land in a more eq- uitable manner in the chiefdoms. By aiming to loosen the grip of the local authorities over land in the provinces, policymakers hoped to extend land tenure to vulnerable groups such as women, youths, and strangers. Observers of the policy reform suggest that the reform would be beneficial to strangers across the chiefdoms, because strangers are generally landless and must rely on chiefs to use land in the chief- doms (Sawyer 2008). This study investigates how this group responds to the policy. Similarly, the policy may have also played out differently in chiefdoms with higher levels of unemployment or because they are close to major towns. Peri-urban areas are known to be prone to land disputes (Nuhu 2019; Lombard 2016). 356 Kpaka To investigate this, I use the 2015 census data to generate a binary variable that takes value 1 for chief- doms with an above-median proportion of strangers in the population and 0 otherwise. For employment, the variable is 1 for chiefdoms with below-median employment levels, and 0 otherwise.4 For proximity to major towns, I use geospatial data from Acemoglu, Reed, and Robinson (2014) to create a binary variable that takes value 1 for chiefdoms with above-median nearest distance to major towns, and 0 otherwise. Descriptive statistics for these binary variables are given in supplementary online appendix table S1.8. These variables are interacted with the DiD estimator and the interaction term shows any differential Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 impact of the policy. I present the results of these analyses in supplementary online appendix table S1.13. Each column corresponds to the models in table 3. As can be seen, the coefficients on each of the interaction terms are not statistically significant. It is also notable that the main ATT remains significant in columns (1) (δ = 2.03, s.e = 1.05) and (2) (δ = 1.80, s.e = 0.95). In column (3) the main ATT becomes insignificant, but the magnitude is comparable to that of the main result (δ = 1.83, s.e = 1.11). These results suggest that the impact of the policy is not driven by a response to the policy from vulnerable groups or chiefdoms. 6.2. CLCs as Conduits of Land Cases in the Formal Courts Research elsewhere suggests that because alternative dispute-resolution channels are often not able to provide final resolutions to cases, litigants are more likely to take land cases to the formal court system (Crook 2004). It is also possible that in this case study too, instead of resolving land cases, CLCs are either directly or indirectly conduits of land cases to the local courts. The direct channel may include instances where CLCs refer cases to the courts, either local courts or magistrate’s courts. In the chiefdoms survey, about 43 percent of the CLCs reported sending land cases to local or magistrate’s courts. Another option, which is not directly captured in my data, is that people who may be unsatisfied with resolutions from CLCs may still take their disputes to the formal court system. This is limited by the availability of other channels to justice, for example, the existence of a magistrate’s court in the chiefdom or close by. To test these ideas, this study creates an interaction term of DiD estimator and a binary variable that takes value 1 for a chiefdom with CLCs that make case referrals to either a local or magistrate’s court, and 0 otherwise. This variable is a post-treatment measure, so the coefficient on the interaction term does not explain the ATT, but it shows whether there are differences in impact between chiefdoms with CLCs that send cases to the courts and those with CLCs but do not send cases. Furthermore, since the locations of formal courts impact people’s ability to bring cases there, a triple interaction with the DiD estimator, CLC case referrals, and a binary variable that takes value 1 if the chiefdom also has a magistrate’s court, has been created. Model (1) in fig. 5 shows the coefficient plot from column (1) in supplementary online appendix table S1.14. The interaction term is indeed significant at the 95 percent level, and it suggests chiefdoms with CLCs that make land-case referrals to the formal courts have about 1.79 more land cases litigated at the local courts than those that have CLCs but do not make case referrals. Also important is that the coefficient on the DiD estimate is now negative and not statistically significant. This suggests that the main impact may be driven by the CLCs that make case referrals. Model (2) is a triple interaction of DiD estimator, CLC case referrals, and the presence of a magistrate’s court. This coefficient is significant at the 99 percent level, and the magnitude of 2.97 is more than double the mean number of land litigation cases for the study period. In addition, the coefficient on the main effect is no longer significant. These results provide correlational evidence for the argument that CLCs are indeed direct conduits of cases that end up in the formal court system. 4 As time series data for employment at the chiefdom level are not available, the assumption I make here is that the 2015 figure represents a stock that captures the relative difference in employment across chiefdoms. The same assumption holds for the proportion of strangers in the population. The World Bank Economic Review 357 Figure 5. Coefficient Plot of CLC Land-Case Referral. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: The x-axis represents standardized coefficients from columns (1) and (2) in supplementary online appendix table S1.14. Model (1) shows coefficients of the interaction of CLC case referral and the DiD estimator. Model (2) is similar but with a triple interaction with the presence of magistrate’s courts. Error bars are for 95 percent confidence intervals. 6.3. Issue Salience Finally, I explore the role of issue salience in explaining the observed effect. The argument is that the presence and activities of CLCs make land issues more salient. For instance, the CLCs may educate people about their land rights. People that are able to respond may decide to bring their land disputes directly to the formal courts, including the local courts in the chiefdoms. The ability of people to respond in this way will depend on their level of education and incomes (Logan 2017). Although not a direct test of the issue salience argument, I investigate whether the effect is stronger in chiefdoms with high levels of education and income. I again rely on the 2015 census data and create a binary variable that takes value 1 for chiefdoms with above-median secondary-level education attainment, and 0 otherwise. For income, I use the proportion of non-agriculture labor in each chiefdom as a proxy for relative income difference among chiefdoms.5 A binary variable that takes value 1 for chiefdoms with an above-median proportion of non-agriculture labor and 0 otherwise has been created. Descriptive statistics for these binary variables are give in sup- plementary online appendix table S1.8. Panels (a) and (b) in fig. 6 show results for education and non-agriculture labor respectively. The coeffi- cients on the interaction terms are the same and are both statistically significant. They suggest that a chief- dom whose secondary-education attainment is above median, or that has above-median non-agriculture labor, and adopted CLCs has about 1.90 more land cases litigated in the local courts. The effect for the baseline category is no longer statistically significant. This suggests that the ATT is driven by chiefdoms that have relatively higher levels of education or potentially higher income. My interpretation is that these are the chiefdoms with people that are able to respond to the introduction of the CLCs. 5 Non-agriculture labor is used widely in development economics as a proxy for income in developing country contexts (Gollin, Lagakos, and Waugh 2014; Haggblade, Hazell, and Reardon 2007). 358 Kpaka Figure 6. Response to Issue Salience. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: The x-axis represents standardized coefficients from supplementary online appendix table S1.15. Error bars are for 95 percent confidence intervals. 7. Conclusion Lack of capacity in state legal institutions means informal dispute-resolution channels handle the vast ma- jority of everyday civil disputes in many developing countries. Previous work shed doubt on the viability of these informal forums to address land disputes. The central question of this research is whether an in- formal channel supported by the state can help resolve land disputes and thus reduce land caseload in the formal court system. This case study from Sierra Leone suggests that instead of resolving land disputes, the informal forums are more likely conduits of land cases to the formal courts. It is plausible that the involvement of the state through the CLCs raised the salience of land issues, but as resolutions through informal channels may be perceived as not final, people that are unsatisfied with outcomes from these channels may still pursue a more final resolution in the formal court system. My analysis in fact shows that the CLCs themselves do direct case referrals of land cases to the formal courts. Another plausible channel is that people may have responded to the salience of land disputes by sidestepping the informal forum altogether and bringing land cases directly to the formal courts. As my analysis suggests, this would be the case for those who can respond, such as the more educated and the relatively more resourced. This finding from Sierra Leone matters for our understanding of how to protect and upload property rights under customary land-tenure regimes. The results raise questions about how to address poor le- gitimacy in customary land-dispute-resolution mechanisms, and whether effective linkages can be made between customary and non-customary property rights. It does seem in this case that state institutions cannot easily be replaced by informal channels of dispute resolution. The case suggests that investment of resource to build informal dispute-resolution mechanisms should accompany efforts to build the capacity of the formal court system as well. The positive side of this unexpected result is that informal dispute-resolution channels may be more important for providing access to justice through the formal systems. This happens when people are empowered and are able to seek justice. Emphasizing this positive aspect of informal dispute-resolution channels in policy reforms, while also investing in state capacity, probably provides the best chance to uphold land property rights in contexts with weak property-right regimes. The World Bank Economic Review 359 Data availability The survey data on chiefdom administration, as well as the data set created from the local court records are available upon request to the Author. Data used from secondary and other sources are noted in the paper, and available with the various authors. References Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Abdullah, I., Y. Bangura, C. Blake, L. Gberie, L. Johnson, K. Kallon, S. Kemokai et al. 1997. “Lumpen Youth Cul- ture and Political Violence: Sierra Leoneans Debate the RUF and the Civil War.” Africa Development/Afrique et Développement 22(3/4): 171–215. Acemoglu, D., A. Cheema, A. I. Khwaja, and J. A. Robinson. 2020. “Trust in State and Nonstate Actors: Evidence from Dispute Resolution in Pakistan.” Journal of Political Economy 128(8): 3090–3147 Acemoglu, D., T. Reed, and J. A. Robinson. 2014. “Chiefs: Economic Development and Elite Control of Civil Society in Sierra Leone.” Journal of Political Economy 122(2): 319–68. Angrist, J. D., and J.-S. Pischke. 2008. Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton Univer- sity Press. Barrows, R. L. 1974. “African Land Reform Policies: The Case of Sierra Leone.” Land Economics 50(4): 402–10. Besley, T., and M. Ghatak. 2010. “Property Rights and Economic Development.”In Handbook of Development Eco- nomics, 4525–95. Elsevier. Blattman, C., A. C. Hartman, and R. A. Blair. 2014. “How to Promote Order and Property Rights under Weak Rule of Law? An Experiment in Changing Dispute Resolution Behavior through Community Education.” American Political Science Review 108(1): 100–20. Boege, V., M. A. Brown, and K. P. Clements. 2009. “Hybrid Political Orders, Not Fragile States.” Peace Review 21(1): 13–21. Boone, C. 2013. “Land Regimes and the Structure of Politics: Patterns of Land-Related Conflict.” Africa 83(1): 188– 203. Bottazzi, P., A. Goguen, and S. Rist. 2016. “Conflicts of Customary Land Tenure in Rural Africa: Is Large-Scale Land Acquisition a Driver of ‘Institutional Innovation’?” Journal of Peasant Studies 43(5): 971–88. Cancian, M., and A. Levinson. 2005. “Labor Supply Effects of the Earned Income Tax Credit: Evidence from Wis- consin Supplemental Benefit for Families with Three Children.” Technical report, National Bureau of Economic Research. Chirayath, L., C. Sage, and M. Woolcock. 2005. “Customary Law and Policy Reform: Engaging with the Plurality of Justice Systems.” Collins, A., and M. I. Mitchell. 2018. “Revisiting the World Bank’s Land Law Reform Agenda in Africa: The Promise and Perils of Customary Practices.” Journal of Agrarian Change 18(1): 112–31. Conteh, S. 2015. “Now That the Sierra Leone Land Policy 2015 Has Been Officially Launched – An Open Letter to President Koroma.” Accessed April 14, 2020. https://sierralii.org/content/now- sierra- leone- land- policy- 2015- has- been- officially- launched- %E2%80%93- open- letter- president. Crook, R. C. 2004. “Access to Justice and Land Disputes in Ghana’s State Courts: The Litigants’ Perspective.” Journal of Legal Pluralism and Unofficial Law 36(50): 1–28. D’Aoust, O., and O. Sterck. 2016. “Who Benefits from Customary Justice? Rent-Seeking, Bribery and Criminality in Sub-Saharan Africa.” Journal of African Economies 25(3): 439–67. Deininger, K., and J. S. Chamorro. 1999. Investment and Income Effects of Land Regularization: The Case of Nicaragua. Available at SSRN 636200 Nov 30. Deseau, A., A. Levai, and M. Schmiegelow. 2019. “Access to Justice and Economic Development: Evidence from an In- ternational Panel Dataset.” Technical report, Université catholique de Louvain, Institut de Recherches Economiques et …. Fanthorpe, R. 2006. “On the Limits of Liberal Peace: Chiefs and Democratic Decentralization in Post-War Sierra Leone.” African Affairs 105(418): 27–49. Feder, G., and D. Feeny. 1991. “Land Tenure and Property Rights: Theory and Implications for Development Policy.” World Bank Economic Review 5(1): 135–53. 360 Kpaka Fenton, J. S. 1951. Outline of Native Law in Sierra. Government Printer. Goldstein, M., and C. Udry. 2008. “The Profits of Power: Land Rights and Agricultural Investment in Ghana.” Journal of Political Economy 116(6): 981–1022. Gollin, D., D. Lagakos, and M. E. Waugh. 2014. “The Agricultural Productivity Gap.” Quarterly Journal of Economics 129(2): 939–93. Government of Sierra Leone. 1991. The Constitution of Sierra Leone. ———. 2015. “National Land Policy of Sierra Leone.” Haggblade, S., P. B. Hazell, and T. Reardon. 2007. Transforming the Rural Nonfarm Economy: Opportunities and Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Threats in the Developing World. Intl Food Policy Res Inst. Hartman, A. C., R. A. Blair, and C. Blattman. 2018. “Engineering Informal Institutions: Long-Run Impacts of Al- ternative Dispute Resolution on Violence and Property Rights in Liberia.” Technical report, National Bureau of Economic Research. Huggins, C. 2009. “Land in Return, Reintegration and Recovery Processes: Some Lessons from the Great Lakes Region of Africa.” Uncharted Territory: Land, Conflict and Humanitarian Action, 67–91. Jackson, P. 2007. “Reshuffling an Old Deck of Cards? The Politics of Local Government Reform in Sierra Leone.” African Affairs 106(422): 95–111. Johnson, O. E. 2011. “Reforming the Customary Land Tenure System in Sierra Leone: A Proposal.” International Growth Centre (IGC) Working Paper, (11/0558). Kanu, E. A., and C. H. Henning. 2019. “An assessment of land reform policy processes in Sierra Leone: A network based approach (No. WP2019-04).” Working Papers of Agricultural Policy. Logan, C. 2017. “Ambitious SDG Goal Confronts Challenging Realities: Access to Justice Is Still Elusive for Many Africans.” Lombard, M. 2016. “Land Conflict in Peri-urban Areas: Exploring the Effects of Land Reform on Informal Settlement in Mexico.” Urban Studies 53(13): 2700–20. McIntyre, A., E. K. Aning, and P. N. N. Addo. 2002. “Politics, War and Youth Culture in Sierra Leone: An Alternative Interpretation.” African Security Studies 11(3): 6–15. Nuhu, S. 2019. “Peri-urban Land Governance in Developing Countries: Understanding the Role, Interaction and Power Relation among Actors in Tanzania.”In Urban Forum, Volume 30, Springer, pp. 1–16. Oredola-Davies, F. 2006. “Land and Pro-Poor Change in Sierra Leone.” https://citeseerx.ist.psu.edu/document?repid =rep1&type=pdf&doi=f282f5d40e652b5628f973b8a508aa6f87279ec5. Park, A. S. 2008. “Consolidating Peace: Rule of Law Institutions and Local Justice Practices in Sierra Leone.” South African Journal on Human Rights 24(3): 536–64. Price, C. 2018. “Alternative Dispute Resolution in Africa: Is ADR the Bridge between Traditional and Modern Dispute Resolution.” Pepp. Disp. Resol. LJ 18: 393. Renner-Thomas, A. 2010. Land Tenure in Sierra Leone: The Law, Dualism and the Making of a Land Policy. Author- House. Richards, P. 2005. “To Fight or to Farm? Agrarian Dimensions of the Mano River Conflicts (Liberia and Sierra Leone).” African Affairs 104(417): 571–90. Ryan, C. 2018. “Large-Scale Land Deals in Sierra Leone at the Intersection of Gender and Lineage.” Third World Quarterly 39(1): 189–206. Sawyer, E. 2008. “Remove or Reform? A Case for (Restructuring) Chiefdom Governance in Post-Conflict Sierra Leone.” African Affairs 107(428): 387–403. Suma, M. 2014. Sierra Leone-Justice Sector and the Rule of Law. Review by AfriMAP and the Open Society Initiative for West Africa, Accessed January 7, 2016. http://issat.dcaf.ch/esl/layout/set/print/content/download/48039/7587 86/file/Sierra% 20Leone% 20Justice.pdf. Swenson, G. 2018. “Legal Pluralism in Theory and Practice.” International Studies Review 20(3): 438–62. Tamanaha, B. Z. 2011. “The Rule of Law and Legal Pluralism in Development.” Hague Journal on the Rule of Law 3(1): 1–17. Tangri, R. 1976. “Conflict and Violence in Contemporary Sierra Leone Chiefdoms.” Journal of Modern African Studies 14(2): 311–21. Toulmin, C. 2009. “Securing Land and Property Rights in Sub-Saharan Africa: The Role of Local Institutions.” Land Use Policy 26(1): 10–19. The World Bank Economic Review 361 Unruh, J. D., and H. Turray. 2006. “Land Tenure, Food Security and Investment in Postwar Sierra Leone.” FAO LSP WP 22. van der Haar, G., and M. van Leeuwen. 2019. “War-Induced Displacement: Hard Choices in Land Governance.” Land 8(6): 88. van Leeuwen, M., and G. van der Haar. 2016. “Theorizing the Land–Violent Conflict Nexus.” World Development 78: 94–104. Wooldridge, J. M. 2002. “Econometric Analysis of Cross Section and Panel Data.” MIT Press, Cambridge, MA 108. World Bank G. 2007. World Development Report 2008: Agriculture for Development. World Bank Group Washing- Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 ton, DC, USA. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Supplementary Online Appendix Chiefs, Courts, and Upholding Property Rights: Quasi-Experimental Evidence from Sierra Leone Henry Musa Kpaka S1. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Table S1.1. Outcome and Court-Level Control Descriptive Statistics Variable Obs Mean Std. dev. Number of land cases 711 1.28 2.80 Land case (binary) 711 0.40 0.49 Other civil cases 711 51.38 65.21 Total cases 711 52.66 66.14 Mean court cost (Le) 711 5,857.93 10,297.27 Number of record years 706 5.48 2.89 Source: Authors calculation from local court administrative data. Note: Not all courts houses had records for each year of the study. Figure S1.1. Local Court Locations and Cost. Source: Author’s analysis from local courts’ administrative data, author’s chiefdom administration survey and 2015 Population and Housing Census. Note: The darker colours on the map represent more intensity of courts per population and higher costs of using courts. Table S1.2. Observations by Year Control group Year Observations Percentage 2009 12 2.47 2010 14 2.89 2011 22 4.54 2012 25 5.15 Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 2013 30 6.19 2014 40 8.25 2015 51 10.52 2016 72 14.85 2017 93 19.18 2018 126 25.98 Total 485 100 Source: Authors calculation from local courts’ administrative data. Note:The later years had more records at the local court. Figure S1.2. Typical Case Log at the Local Courts. Source: Local court records. Note: Typical case log in the local court system. Figure S1.3. Example Land Cases. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Source: Local Courts records. Note: Example of title of cases logged at the local courts. Figure S1.4. Visual Inspection of Parallel Trends. Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Vertical axis is the number of land cases per year in each local court in treatment and control chiefdoms. Horizontal is years, and vertical line in each chart indicates year of intervention. Figure S1.5. Simple Difference of Outcomes—Unrestricted Data. Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: In (a), bars are average of the indicator variable that take 1 if case type is land, and 0 otherwise between the control and treatment chiefdoms. In (b) bars represent simple average for the number of land cases for the treatment and control chiefdoms. Figure S1.6. Formal Falsification Test—With Restricted Sample. Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: The y-axis represents standardized coefficients from estimating (2). The model with a diamond on the error bar is with court fixed effects from model (2) in table S1.11, and the other is with chiefdom fixed effects. Error bars are for 95 percent confidence intervals. Table S1.3. Number of Observations by Year Treatment group Year Observations Percentage 2009 2 0.9 2010 6 2.71 2011 9 4.07 2012 7 3.17 Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 2013 7 3.17 2014 14 6.33 2015 20 9.05 2016 35 15.84 2017 50 22.62 2018 71 32.13 Total 221 100 Source: Authors calculation from local courts’ administrative data. Note:The later years had more records at the local court. Table S1.4. Respondents Interviewed Interview respondent Freq. Percentage Chiefdom administrator 22 14.77 Chiefdom secretary 8 5.37 Chiefdom speaker 25 16.78 Chiefdom treasury secretary 61 40.94 Chiefdom council member (councilor) 1 0.67 Paramount chief 5 3.36 Section chief 6 4.03 Religious leader (imam, pastor, etc.) 1 0.67 Regent chief 4 2.68 Village elder or notable 16 10.74 Total 149 100 Source: Author’s chiefdom administration survey. Note:Category of respondents interviewed during field work was diverse. Table S1.5. Chiefdom Land Committee and Member Selection Variable Obs Mean Std. dev. CLC members selected by chiefdom administration 51 0.45 0.50 CLC member selected by PC/regent chief 51 0.53 0.50 Chiefs appoint community members to the CLC 51 0.31 0.47 CLC members selected by other community members 51 0.39 0.49 CLC members are “elected” 51 0.12 0.33 People volunteer to be on CLC 51 0.06 0.24 Source: Author’s chiefdom administration survey. Note:Chiefdom Land Committee (CLC) Paramount Chief (PC). Table S1.6. Composition of Chiefdom Land Committees Which groups are represented on CLC in this chiefdom? Variable Obs Mean Std. dev. Section chiefs 51 0.67 0.48 Village headmen 51 0.51 0.50 PC/regent chief 51 0.57 0.50 Chiefdom speaker 51 0.39 0.49 Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Landholding families 51 0.41 0.50 Local council representatives 51 0.27 0.45 Youth representative 51 0.45 0.50 Women representative 51 0.45 0.50 Religious leaders (imam/pastor) 51 0.35 0.48 Traditional healers 51 0.04 0.20 Members of secret society 51 0.04 0.20 Source: Author’s chiefdom administration survey. Note:Chiefdom Land Committee (CLC) Paramount Chief (PC). Table S1.7. Mediation Role of Chiefdom Land Committees Variable Obs Mean Std. dev. Does CLC hear land dispute cases between people in community 51 0.98 0.14 Number of land disputes per month 50 0.29 0.22 Does CLC charge to hear land cases 50 0.36 0.49 Amount charge (Le) 12 87,083.33 82,254.00 Does CLC refer cases to local/magistrate’s court or village level 51 0.43 0.50 Source: Author’s chiefdom administration survey. Note:Chiefdom Land Committee (CLC). Table S1.8. Summary Statistics for Hypothesis Testing Variables Obs Mean Std. dev. Above-median secondary education 154 0.43 0.50 Above-median stranger population 154 0.45 0.50 Above-median non-agriculture labor 154 0.40 0.49 Chiefdom has magistrate’s court 149 0.13 0.33 Below-median employment 154 0.37 0.48 Above-median distance to nearest town 149 0.30 0.46 Source: Data on secondary education, stranger population, employment, and non-agriculture labor calculated using 2015 census data. Data on median distance to towns (km) taken from Acemoglu et al. (2014), and these include Freetown, Bo, Kenema, and Makeni. Location of magistrate obtained by the author from local court survey. Note:Author’s calculation from various data sources. Table S1.9. Determinants of Policy Compliance CLC compliance Variables (1) (2) (3) (4) (5) (6) (7) PC tenure years (by 2016) 0.00 – – – – – 0.00 (0.00) (0.00) Number of ruling houses – −0.00 – – – – −0.01 (0.02) (0.02) Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Chiefdom primary edu. attainment – – −0.30 – – – −0.89 (1.18) (1.49) Chiefdom non-ag labor – – – 0.17 – – 0.01 (0.41) (0.50) Mining license 1930 – – – – 0.25∗ – 0.27∗∗ (0.13) (0.13) Ln(distance to nearest major town) – – – – – −0.18 −0.23∗ (0.11) (0.13) Amalgamation 0.01 0.02 0.01 0.01 0.02 0.01 0.04 (0.09) (0.11) (0.09) (0.09) (0.08) (0.09) (0.10) Population density (2015) 0.00∗ 0.00∗ 0.00∗ 0.00 0.00∗ 0.00 0.00 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Ethnolinguistic fractionalization 0.13 0.14 0.15 0.10 0.05 0.20 0.12 (0.25) (0.25) (0.25) (0.27) (0.25) (0.25) (0.28) District fixed effects X X X X X X X Observations 149 149 149 149 149 149 149 R-squared 0.19 0.19 0.19 0.19 0.22 0.21 0.24 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Chiefdom Land Committee (CLC); Paramount Chief (PC). Table S1.10. ATT for Likelihood of Observing Land Cases—Unrestricted Data Land Case (Binary) Variables (1) (2) (3) DiD 0.09 0.10 0.06 (0.13) (0.15) (0.16) Year fixed effects x x x Chiefdom fixed effects x x – Court controls – x x Court fixed effects – – x Outcome mean 0.4 0.4 0.4 Outcome standard deviation 0.49 0.49 0.49 Number of chiefdoms 130 129 129 Court-year observations 706 689 689 R-squared 0.52 0.52 0.59 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Court controls include mean cost and number of record years. Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Average treatment on the treated (ATT); Difference-in-Difference (DiD). Table S1.11. Formal Placebo Test Number of land cases Variables (1) (2) (3) Treatment × 2009 1.48 – – (1.12) Treatment × 2010 0.68 −0.33 – (1.04) (0.39) Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Treatment × 2011 −0.37 −1.56 – (1.40) (0.96) Treatment × 2012 0.47 −0.33 – (1.22) (0.47) Treatment × 2013 1.84 0.18 – (1.65) (1.17) Treatment × 2014 1.40 −0.64 0.45 (1.55) (0.76) (0.75) Treatment × 2015 1.71 0.10 1.23 (1.20) (0.81) (0.76) Treatment × 2016 2.60∗∗ 0.68 0.88 (1.18) (0.85) (0.89) Treatment × 2017 3.33∗ 1.67∗ 3.30∗∗ (1.76) (1.01) (1.62) Treatment × 2018 2.56∗∗ 1.08 2.47∗ (1.01) (0.81) (1.34) Chiefdom fixed effects X – – Year fixed effects X X X Court controls X X X Court fixed effects – X X Outcome mean 1.28 1.28 1.53 Outcome standard deviation 2.8 2.8 3.3 Number of chiefdoms 129 129 45 Court-year observations 689 689 267 R-squared 0.38 0.52 0.50 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Court controls include mean cost and number of record years. Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Table S1.12. Triple Difference Estimation Number of all cases (land and non-land) Variables (1) (2) (3) DiD 39.11 9.11 −9.25 (34.00) (23.32) (6.62) DiDiD – 54.36∗∗ 82.06∗∗∗ (25.11) (30.26) Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 Treatment × case type (land) – −40.90∗ −77.80∗∗ (22.04) (33.10) Post-treatment × case type (land) – 3.54 −9.78 (9.69) (11.94) Case type (land) – 6.03 20.71 (6.45) (17.21) Year fixed effects X X X Court fixed effects X X X Outcome mean 52.66 52.66 54.23 Outcome standard deviation 66.14 66.17 70.65 Number of chiefdoms 130 130 45 Court-year observations 706 706 272 R-squared 0.71 0.72 0.75 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Difference-in-Difference (DiD); Difference-in-Difference- in-Difference (DiDiD). Table S1.13. Pent-Up Land Concerns from Vulnerable Groups Number of land cases Variables (1) (2) (3) DiD 2.03∗ 1.80∗ 1.83 (1.05) (0.95) (1.11) DiD above-median stranger population −1.02 – – (1.13) Above-median stranger population −2.39∗∗∗ – – (0.35) DiD × below-median employment level – −0.50 – (1.21) Below-median employment level – 2.95∗∗∗ – (0.95) DiD × below-median distance to major town – – −0.52 (1.35) Below-median distance to major town – – 1.19∗∗∗ (0.33) Chiefdom controls X X X Year fixed effects X X X Court fixed effects X X X Outcome mean 1.28 1.28 1.28 Outcome standard deviation 2.8 2.8 2.8 Number of chiefdoms 130 130 130 Court-year observations 706 706 706 R-squared 0.52 0.52 0.52 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Chiefdom controls include 2015 census population density and distance to the nearest major town. Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Difference-in-Difference (DiD). Table S1.14. CLCs as Direct Conduits of Land Cases Number of land cases Variables (1) (2) DiD −0.20 −0.45 (0.95) (0.92) DiD × refer cases to courts 1.79∗∗ – (0.78) Downloaded from https://academic.oup.com/wber/article/39/2/341/7705880 by The World Bank user on 02 May 2025 DiD × refer cases to courts × magistrate’s court presence – 2.97∗∗∗ (0.98) Refer cases to courts −1.50 −1.62 (1.09) (1.31) Magistrate’s court presence – 0.03 (0.98) Year fixed effects X X Court fixed effects X X Outcome mean (full sample) 1.28 1.28 Outcome standard deviation (full sample) 2.8 2.8 Number of chiefdoms 44 44 Court-year observations 218 218 R-squared 0.51 0.52 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Chiefdom Land Committee (CLC). Table S1.15. Issue Salience and Response Number of land cases Variables (1) (2) DiD 0.25 0.25 (0.54) (0.54) DiD × above-median non-ag labor 1.90∗∗ – (0.85) Above-median non-ag labor 0.28 – (0.83) DiD × above-median secondary education – 1.90∗∗ (0.85) Above-median secondary education – 2.85∗∗∗ (0.54) DiD × magistrate’s presence – – Magistrate’s court presence – – Chiefdom control X X Year fixed effects X X Court fixed effects X X Outcome mean 1.28 1.28 Outcome standard deviation 2.8 2.8 Number of chiefdoms 130 130 Court-year observations 706 706 R-squared 0.52 0.52 Source: Author’s analysis from local courts’ administrative data and author’s chiefdom administration survey. Note: Chiefdom controls include 2015 census population density and distance to the nearest major town. Robust standard errors in parentheses clustered at the chiefdom level. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Chiefdom Land Committee (CLC); Difference-in-Difference (DiD). C Crown copyright 2024. This article contains public sector information licensed under the Open Government Licence v3.0 (https://www.nationalarchives.gov.uk/doc/open- government- licence/version/3/).