Crime and Conflict Statistics from Baseline GoBifo/ENCISS/Decentralization Surveys

Transcription

1 Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Crime and Conflict Statistics from Baseline GoBifo/ENCISS/Decentralization Surveys Preliminary Analysis September Analysis for this paper was conducted by Elizabeth Foster, Evaluation Specialist, Institutional Reform and Capacity Building Project (IRCBP), with input from Pamela Dale, Consultant, Justice for the Poor (J4P), World Bank; Gibrill Jalloh, Research Assistant, and Ryann Manning, Field Research Coordinator, Justice for the Poor and Understanding Processes of Change in Local Governance (J4P/LG), World Bank; and Phillip Kargbo, Research Assistant, IRCBP. The paper was written by Ryann Manning with contributions from Elizabeth Foster and Pamela Dale. Comments are welcome, and should be addressed to and

2 Summary / Key Findings This memo presents some interesting findings from a preliminary analysis of crime- and disputerelated data from the GoBifo/ENCISS/Decentralization baseline survey. 1 Due to a number of considerations, these findings should be considered preliminary and inconclusive. 2 They are presented to spark discussion and to generate hypotheses and areas for future research including the IRCBP public services survey and the follow-up GoBifo/ENCISS/Decentralization survey, as well as the Justice for the Poor and Understanding Processes of Change in Local Governance qualitative research program rather than to draw conclusions about the experience and resolution of crime and conflicts in Sierra Leone. The GoBifo/ENCISS/Decentralization baseline survey, conducted between November 2005 and January 2006, sampled 2,799 households in Bombali and Bonthe districts. As part of a longer questionnaire, the survey asked respondents whether they had been the victim of or involved in specific crimes and conflicts. 3 (There were differences in how this was asked: for loan and money conflicts and for fighting, respondents were asked if they had been involved in such a conflict; for thefts and for witchcraft, they were asked if they had been a victim.) It also asked whether (and to whom) they had first reported the crime or conflict, 4 what was the outcome, and whether they were satisfied with how the matter had been resolved. Overall, 48% of respondents experienced at least one of the types of conflicts or crimes included in the survey. The rates of experiencing such crimes and conflicts ranged widely, from less than 0.5% for fighting with weapons to 25% for theft of livestock. In general, those with higher socioeconomic status were more likely to be involved in (or a victim of) a conflict or crime. Reporting rates also varied for different types of conflicts; overall, just over half were reported outside the family. One of the strongest determinants both of whether people reported the crimes and conflicts that they experienced, and to whom they reported them, is which chiefdom they live 1 This survey was jointly funded by the Institutional Reform and Capacity Building Project (IRCBP), the GoBifo Project, and the Enhancing the Interaction and Interface between Civil Society and the State to Improve Poor People s Lives (ENCISS) Project. The development of survey instruments and data collection were managed by the Evaluation Unit of IRCBP in close collaboration with ENCISS, GoBifo, IRCBP and Statistics Sierra Leone (SSL) staff, and with technical oversight and support from Professor Edward Miguel and John Bellows at the University of California, Berkeley, Dr. Rachel Glennerster of the Abdul Latif Jameel Poverty Action Lab at the Massachusetts Institute of Technology, and Dr. Yongmei Zhou of the World Bank. 2 Though the researchers have made every effort to ensure the validity of survey results, it is important to note that this survey was not originally undertaken as an analysis of crime and case trajectories, and is not nationally representative. It addresses just a small number of specific crimes and conflicts, which likely are not representative of the full range of conflicts. It also asks only about where cases were first reported, not where they were ultimately resolved. Further, confusion between the various institutions and enumerator bias may have impacted the coding of some responses. The analysis in this memo is not the result of developing hypotheses and undertaking research to test them, but rather running a series of regressions and selecting those which were significant and/or interesting. Tests of statistical significance do not account for the clustered nature of the sample. 3 Respondents were asked if they or a member of their household had ever had a conflict over a loan or other money business, or if they in the last 12 months had been a victim of theft of livestock, of household items, or of money, a victim of juju or witchcraft, or involved in fighting (with or without weapons). 4 The options given were: Village headman court, Section chief court, Local court (NA court), Paramount chief court, Magistrate court, Police, Secret Society, Ward Development Committee, Village Development Committee, NGO or human rights organization, Mosque/church, or Village Elders. 1

3 in. Controlling for chiefdom eliminates the significance of both religion and ethnicity in whether people reported crimes or conflicts, and nearly eliminates the significance of ethnicity in where people reported them. (Religion was not significant in determining where people reported.) Likelihood of reporting a crime was also statistically correlated with respondents age, gender, migration since the start of the war, group membership, primary source of information about government and politics, and with certain measures of both trust and involvement with local government. 5 These relationships are not always predictable; for instance, those who belong to more groups and who have run for the village development committee or attended a ward committee meeting are more likely to report a crime, but those with higher levels of trust in government and in outsiders are less likely to report a conflict. Education and socioeconomic status were not significant, and neither was a measure of ethnic heterogeneity by chiefdom. The two most popular institutions for reporting crimes are the village headman court and the village elders. A total of 85% of the crimes and conflicts named in the survey are reported first to these village-level traditional leaders, with 60.8% of crimes and conflicts first reported to the village headman court and 24.7% reported to village elders. These institutions remain the most popular for almost any subgroup of the population, broken down by chiefdom, religion, ethnicity, age, etc. Patterns of where people report crimes and conflicts are significantly related to chiefdom, ethnicity, age, education, select income and position measures, and migration. (Measures of trust, group membership, access to information, and involvement with local government were not significant.) Examining these data in detail reveal some interesting trends; for instance, likelihood of reporting a crime to the village headman court increases with age and with migration since the start of the war, but decreases with higher socioeconomic status. Higher socioeconomic status is also correlated with an increased likelihood to report to the section chief s court and to the police. The analysis also considered the rate at which respondents said the crime or dispute had been resolved satisfactorily. Just 41% of all conflicts are resolved satisfactorily, with significant differences between different types of conflicts. People who reported a crime or dispute were significantly more likely to be satisfied with the outcome. Likelihood of satisfaction also increased along with education, socioeconomic status, and several measures of participation in local government and access to information about government. Satisfaction is also correlated with ethnicity, but religion, age, and gender were not significant. Satisfaction also varied both with where the conflict was first reported, and with the type of outcome. The lowest rate of satisfaction was with the two formal / recognized courts (local courts and magistrate courts), with only 20% of people who went first to these institutions saying the matter had been resolved satisfactorily. 6 The highest rates of satisfaction were reported by those that went first to an NGO or human rights organization (67% satisfied) and to a church or mosque (63% satisfied), though the absolute numbers of people using these institutions were 5 Involvement includes running for office, attending meetings, and speaking with local leaders. 6 The village development committee (VDC) had a lower rate of satisfaction (0%) but only 1 person reported first to the VDC. 2

4 quite small. 7 Satisfaction also varied by outcome, with jail the least likely to lead to satisfaction (24% satisfied), lower than dismissed with no punishment (28% satisfied). Although tests of significance for this section do control for the type of conflict, the magnitude of difference in satisfaction may be influenced by differences in which types of conflicts are brought to which institutions, or by their relative intransigence (e.g., if the more difficult conflicts to resolve are also more likely to be brought initially to a court). 7 Only 3 people reported first to an NGO, and 8 to a church or mosque. Therefore, the margins of error on the satisfaction rate from these institutions will be large. 3

5 Background on Survey The survey covers the areas in Bombali and Bonthe districts where the GoBifo communitydriven development project is operating. 8 GoBifo selected Bombali and Bonthe for regional diversity (Bombali is in the northern and Bonthe in the southern province) and to target poorer districts with fewer other donor projects. Within these two districts, project organizers agreed that GoBifo would work in all wards in which the National Social Action Program (NSAP) was not scheduled to work in The team drew up a list of villages of a specified size 10 within each GoBifo ward, and twelve villages were randomly selected for inclusion in the sample. (In Bonthe Town, each section was considered a community and 4 sections were drawn from each ward. Makeni Town was not included in the sample, although GoBifo is operating there.) Within each village or community, twelve households were randomly selected for the survey. Survey enumerators followed a set pattern in selecting the respondent for each household, such that equal numbers of male non-youth, male youth, female non-youth and female youth were interviewed. The total sample size was 2,799 households. The data is representative of mediumsized communities in wards in Bombali and Bonthe where GoBifo operates, but not of the districts or the country as a whole. Explanation of Analysis Survey respondents were asked about a series of crimes and conflicts: whether they or a member of their household had ever had a conflict over a loan or other money business, and whether they in the last 12 months had been a victim of theft of livestock, of household items, or of money; a victim of juju or witchcraft; or involved in fighting (with or without weapons). If they answered yes they were asked a series of follow-up questions about to whom (if anyone) they reported the conflict outside their family, what the outcome was, and if they were satisfied with how the matter had been resolved. The options given for where conflicts could have been reported were village headman court, section chief court, local court (NA court), paramount chief court, magistrate court, police, secret society, ward development committee, village development committee, NGO or human rights organization, mosque or church, or village elders. 11 The relevant sections of the questionnaire are reproduced in Appendix A. 8 GoBifo is a pilot project sponsored by the Japanese Social Development Fund of the World Bank. 9 NSAP is another World Bank-supported program, administered by the National Commission for Social Action (NaCSA), formerly the National Commission for Reconstruction, Resettlement and Rehabilitation (NCRRR). For its community-driven program (CDP), NSAP selected wards that were classified as the most vulnerable by its services and opportunities mapping exercise; therefore, the GoBifo wards represent better-off communities in these districts. 10 Villages were included if they had 10 to 100 households in Bonthe, and 20 to 200 households in Bombali, based on census data. 11 Though most of these categories would be well-defined within a local context, there is also some ambiguity that might be understood or interpreted differently by different respondents. For instance, there may not be a clear distinction between the village elders (a loosely-defined group of usually male elders who participate in governance decisions and informal dispute resolution) and the village headman s court (an unrecognized court of the headman and some of his advisors which resolves conflicts on request), though the latter is probably a slightly more formalized body. It is also possible that respondents confused some of the categories; for instance, the local or NA (native authority) court is a formal, clearly-defined institution independent from any of the chiefs courts, but this may not be clearly understood by the average citizen. 4

6 For this analysis, each recorded conflict was treated as an observation. In analyzing the determinants of reporting a crime, a binary for whether or not each conflict was reported was regressed on various characteristics of the respondent, using a simple linear probability model and controlling for the type of conflict. The same technique was used to analyze satisfaction. For the analysis of where conflicts were reported, a Chi-squared test was used to test for correlation between various characteristics of the respondent and where the conflict was reported. This Chisquared analysis does not control for the type of conflict. Likelihood to Experience a Crime Overall, 48% of respondents experienced at least one of the types of conflicts or crimes included in the survey. Rates varied significantly by district and chiefdom, as shown in the chart below. (It is important to remember that these statistics reflect involvement in the specific types of crimes and conflicts addressed by the survey, and not overall levels of crime and conflict in these chiefdoms.) Likelihood to be Involved by District and Chiefdom Bonthe Bombali Bum Imperi Jong Kwamebai Krim Bonthe Town Biriwa Bombali Sebora Gbanti Kamaranka Gbendembu Ngowahun Makari Gbanti Safroko Limba Sanda Loko Sanda Tendaran Sella Limba fraction involved in conflict 5

7 The rates of experiencing different crimes or conflicts also ranged widely by type of crime or conflict, from less than 0.5% for fighting with weapons to 25% for Likelihood to be Involved by Type of Crime or Conflict theft of livestock. Across the board, households with higher socio-economic status were more likely to be involved in (or a victim of) a conflict or crime. Education levels, membership in a ruling family, and various proxies for income and position 12 were significantly and positively correlated with the incidence of crimes and conflicts. The analysis also tested several measures of involvement or linkages with local authorities, loan conflict stolen livestock stolen HH item stolen money fighting w/o weapon fighting w/ weapon witchcraft fraction involved in conflict including running for the village development committee (VDC), meeting a ward committee (WDC) member, attending a WDC meeting, or recently speaking to the village head, women s leader, or youth leader. Of these, the first three are significant, with those who are more involved being more likely to experience a conflict or crime. (The latter three are a poor measure of degree of involvement, as approximately 90% of respondents had spoken to each local leader in the past week.) Also significant and positively correlated with experiencing crimes or conflicts are being able to name the local councilor and the paramount chief, and citing the radio as the respondent s main source of information about government and politics. Membership in community groups was also correlated with crime and conflict, 13 with people who belong to more types of groups being more likely to experience these conflicts. Membership in three specific types of groups (credit or savings groups, social groups, and school-related groups) was also significant and positively correlated. One possible explanation is that group membership actually leads to certain types of conflicts e.g., being part of a credit or savings group may increase the likelihood that the respondent has experienced a conflict over a loan but group membership may also be correlated with other factors that make people more likely to experience the specific crimes and conflicts measured here. 12 These include access to an improved water source or a phone, owning a radio or cassette player, and raising sheep, or ducks. Also tested but found to be not significant were measures for being divorced or widowed, and for the quality of walls, floor, and housing. 13 The survey asked about membership in a number of groups and communal activities. Included in this analysis are 5 indicators: 1) whether the respondent worked on a communal farm in the past year, 2) whether the respondent had worked as part of a labor sharing group in the past year, 3) whether the respondent belonged to a credit or savings group, 4) whether the respondent belong to a social group, 5) whether the respondent belonged to a school related group. A variable for the number of types of groups in which the respondent was involved (between 0 and 5) was also used. 6

8 Likelihood to Report a Crime Overall, 53% of crimes and conflicts were said to have been reported to someone outside the family. Reporting rates varied for different types of crimes and conflicts, with the highest rates reported for fighting and for loan or money conflicts. The higher Likelihood to report by Type of Crime or Conflict reporting rates for loan or money conflicts may be partly explained by the fact that questions were asked in a different manner than for the other loan conflict stolen livestock types of crimes and conflicts. (See Appendix A for the exact questions stolen HH item used.) stolen money Reporting was also correlated with religion 14 and ethnicity, but this relationship disappears when you control for which chiefdom they live in. Chiefdom seems to be one of the strongest determinants of whether people reported the crimes and conflicts that they experienced. fighting w/o weapon fighting w/ weapon witchcraft fraction reporting conflict Likelihood to Report by District and Chiefdom Bonthe Bombali Bum Imperi Jong Kwamebai Krim Bonthe Town Biriwa Bombali Sebora Gbanti Kamaranka Gbendembu Ngowahun Makari Gbanti Safroko Limba Sanda Loko Sanda Tendaran Sella Limba reporting rate 14 Less than 1% of those interviewed reported being anything other than Muslim or Christian. 7

9 Likelihood of reporting a crime was statistically correlated with respondents gender, with men significantly more likely to report (60% versus 46% for women). This may result from the way the question was asked: to whom (outside your family) did you first report the most recent crime / dispute? (emphasis added). For crimes against the household as a unit (theft of livestock, etc.), the man of the household would most often be the one to report it to the authorities. Even in the case of a personal attack on a woman or on her property, she might report it to her husband or to the male head of household and he would be responsible for taking it elsewhere. It is interesting to note that gender is not a significant determinant of where the crime or conflict is reported, or whether it is satisfactorily resolved. (See later sections for a discussion of levels of satisfaction and where crimes and conflicts are reported.) Age was statistically significant, but with no clear patterns across different age groups (except a decline in reporting rates beginning around age 40) and no significant Likelihood to Report by Age difference between youths and non-youths. Migration since the start of the war was also significant, with those who lived elsewhere before the war being significantly more likely to report a crime or conflict (54% versus 50%). Education level and socioeconomic status were not significant, and neither was a measure of ethnic heterogeneity by chiefdom. The respondents level of knowledge about local government (knowing the name of the local councilor, section chief, about 20 about 25 about 30 about 35 about 40 or 45 about 50 to 60 definitely over reporting rate or paramount chief) was also not significant, but their primary source of information about the government is significant, with those whose get their information from the radio (as opposed to the village head, family and friends, or other source) being less likely to report. Group membership is statistically significant, and positively related to likelihood to report. Of the five types of community groups considered, three (communal farming, labor sharing, and school-related) were significant, with respondents who belong to these groups being more likely to report crimes or conflicts. Also significant was the total number of types of groups (from 0 to 5), with those belonging to more groups being more likely to report. Two measures of involvement or linkages with local authorities running for VDC and attending a WDC meeting were significantly correlated with a higher likelihood of reporting. Trust is significant, but in a somewhat surprising direction. We might expect trust to mirror group measurement and linkages with local authorities, such that those who had higher levels of trust like those who were more involved or connected would be more likely to take the step 8

10 of reporting a crime or conflict. Instead, the analysis showed that two of five measures of trust 15 trust in government and trust in outsiders were inversely related to likelihood to report, with those with the highest level of trust being substantially less likely to report a conflict. 16 One possible explanation is that trust in government and in outsiders is endogenous: a person who is a victim of a crime and reports it, but does not get satisfaction, may become less trusting. Reports from some enumerators also suggest that questions about trust may have been poorly understood by respondents. 17 Where Crimes are First Reported 18 The analysis also examined where (if anywhere) people said they first reported the crime or conflict, outside of their family. Across the board, the two most popular institutions for first reporting crimes and conflicts are the village headman court and the Where People Report by Age village elders. Overall, 85% of the crimes and conflicts named in the survey were reported first to these village-level traditional leaders, with about 20 about 25 about % reported to the village about 35 headman court and 24.7% reported to village elders. These institutions remain the most popular for almost about 40 or 45 about 50 to 60 definitely over 60 any subgroup of the population, broken down by chiefdom, religion, ethnicity, age, etc. village headman court village elders section chief court paramount chief court local court magistrate court police VDC It is important to remember that the NGO Mosque / church survey asked only about where the crime or conflict was first reported not where it was ultimately resolved. Thus, although the vast majority of respondents first reported to these village-level traditional leaders, an unknown number then pursued the matter to alternative institutions. It is also important to note, as mentioned above, that there may not be a clear distinction between the village elders (a loosely-defined group of usually male elders who participate in governance 15 The analysis tested five measures of trust: one based on objective questions, such as whether you would send someone with money to buy you something from the market; one based on how much the respondent said they trusted all people; one based on how much the respondent said they trusted various people in the community; one for how much they trusted outsiders; and one for how much they trusted government officials. See the appendices for the exact questions and indicators used. 16 Reporting rates are actually the highest for those with medium levels of trust. The overall relationship is negative, and the most striking feature is that those with high levels of trust are quite a bit less likely to report. 17 Gibrill Jalloh, a supervisor for this survey, said that the word trust proved difficult to translate into some local languages, and that many respondents expressed confusion over the meaning of the questions. 18 This analysis excludes conflicts over loans or money business as we cannot know to which institution they were first reported. 9

11 decisions and informal dispute resolution) and the village headman s court (a court usually involving the headman and some of his advisors, unrecognized by the formal legal system, which resolves conflicts on request). Patterns in where people report crimes or conflicts are significantly related to chiefdom, ethnicity, age, education, select income and position measures, and migration. (Religion, measures of trust, group membership, access to information, and involvement with local government were not generally significant. 19 ) Examining these data Where People Report by Quality of Housing in detail reveal some trends; for instance, likelihood of reporting a crime to the village headman court poor increases with age and with migration since the start of the war, average but decreases with higher socioeconomic status. Higher socioeconomic status is also good correlated with an increased likelihood to report to the section chief s court and to the police village headman court.6.8 village elders 1 section chief court paramount chief court One of the strongest determinants police NGO seems to be which chiefdom the respondents live in (see chart on next page). Controlling for chiefdom nearly eliminates the significance of ethnicity. local court magistrate court VDC Mosque / church The differences by chiefdom raise interesting questions about the reasons behind such patterns (though we also recognize that some of the differences could be statistical artifacts). For instance, why are 30% of crimes in Biriwa chiefdom first reported to the section chief court, while the rate tops 15% in just one other chiefdom in Bombali district and is just 8.3% for the district overall? Is that related in any way to the ethnic divisions that have driven recent conflict over the paramount chieftaincy election in Biriwa? Or why are more than 10% of crimes and conflicts in Makari Gbanti chiefdom reported to the police, while the rate throughout both Bombali and Bonthe districts is just 1.7%? Are the police there more accessible or more effective, or is there some other reason that residents of that chiefdom have greater trust or confidence in the police (or less in alternative institutions)? Exploring these questions could be an important area for future research. 19 Two exceptions are membership in labor sharing groups (which probably reflects differences by age) and having spoken to the woman s leader (which is probably a statistical aberration). 10

12 Where People Report by District and Chiefdom Bonthe Bum Imperi Jong Kwamebai Krim Bonthe Town Biriwa Bombali Sebora Gbanti Kamaranka Gbendembu Ngowahun Bombali Makari Gbanti Safroko Limba Sanda Loko Sanda Tendaran Sella Limba village headman court section chief court paramount chief court police NGO village elders local court magistrate court VDC Mosque / church Satisfaction with Outcome The analysis also considered the rate at which respondents said the crime or dispute had been resolved satisfactorily. Just 41% of all conflicts are resolved satisfactorily, with significant differences between different types of conflicts, ranging from 29% for a stolen household item to 60% for loan conflicts and for physical fighting without a weapon. District and Satisfaction Rate by Type of Conflict chiefdom were again significant, with people in Bonthe significantly more likely to be satisfied with the outcome of a conflict, and with significant loan conflict stolen livestock differences among chiefdoms (see chart in discussion section showing stolen HH item involvement, reporting, and stolen money satisfaction by district and chiefdom). People who reported a crime or dispute were significantly more likely to be satisfied (47% versus 33%). Likelihood of satisfaction also increased along with socioeconomic fighting w/o weapon fighting w/ weapon witchcraft satisfaction rate 11

13 status and several measures of involvement in local government and access to information about government. Satisfaction is also correlated with ethnicity, but religion, age, and gender were not significant. Satisfaction Rates by Education Level and Housing Quality None Primary Higher education satisfaction rate poor average good housing quality satisfaction rate Satisfaction also varied both with where the conflict was first reported, and with the type of outcome. The lowest rates of satisfaction was with the two formal, recognized courts (local courts and magistrate courts) 20 and the highest rates were reported by those that went first to an NGO or human rights organization and to a church or mosque, though the absolute numbers of people using these latter institutions were quite small. 21 Again, this is based just on where people first reported the crime or conflict, which is not necessarily where it was resolved. village headman court section chief court Mosque/church Satisfaction Rate by Where Reported local court paramount chief court magistrate court police VDC NGO village elders no one satisfaction rate 20 Actually, the village development committee (VDC) received the lowest rate of satisfaction, with 0%, but only one (1) person reported first to the VDC. 21 Only 3 people reported first to an NGO, and 8 to a church or mosque. Therefore, the margins of error on the satisfaction rate from these institutions will be large. 12

14 Satisfaction also varied by outcome, 22 with jail the least likely to lead to satisfaction (24% satisfied), even lower than dismissed with no punishment (28% satisfied). One consideration here is that several of the questions (about fighting and about loan or money conflicts) asked not whether the respondent had been a victim, but whether he or she had been involved. A person who instigated a fight or failed to repay a loan will likely have a very different definition of a satisfactory outcome than the victim of a theft. dissmissed with no punishment Satisfaction Rate by Outcome fine jail stolen item returned other satisfaction rate It is also interesting to highlight that in 37% of the conflicts, the primary outcome is other neither a fine, nor jail, nor stolen item returned, nor dismissed without punishment. What are these other types of outcomes? One likely candidate is a formal apology or other symbolic act of repentance or restitution, but we do not know how common that is, nor how it compares with other outcomes in terms of satisfaction. Although tests of significance for both satisfaction by where reported and satisfaction by type of outcomes do control for the type of conflict, the magnitude of difference in satisfaction does not, and may be influenced by differences in which types of conflicts are brought to which institutions. 23 Another factor may be the relative intransigence of different conflicts; e.g., perhaps the most difficult to resolve are also more likely to be reported first to the courts. On the other hand, the variations in satisfaction could be related to differences in how these different institutions resolve crimes or conflicts, and might suggest that people are more satisfied with reconciliatory approaches than with the retributive or adversarial approaches adopted by courts. Moreover, many groups in Sierra Leone consider it a failure to have to bring a case to court (rather than solving it within the community ), so satisfaction may be lower for anything resolved in a court, regardless of how well the courts discharged their duty. Discussion The findings presented here should be considered preliminary and inconclusive. Although the researchers have made every effort to ensure the validity of survey results, the survey was not originally undertaken as an analysis of crime and case trajectories, and is not nationally 22 This analysis does not include conflicts over loans / money business as the possible outcomes for this type of conflict were coded differently. 23 When regressing satisfaction on type of punishment controlling for type of crime, the ranking of satisfaction by institution is the same, and the differences in predicted satisfaction rate are actually slightly larger than indicated in the figure. 13

15 representative. It also addresses just a small number of specific crimes and conflicts, and asks only about where cases were first reported, not where they were ultimately resolved. Moreover, confusion between the various institutions and enumerator bias may have impacted the coding of some responses, and the analysis presented here was the result not of developing and then testing specific hypotheses, but rather of running a series of regressions and selecting those which were significant and/or interesting. 24 This paper is therefore presented to spark discussion and to generate hypotheses and areas for future research including for the IRCBP public services survey and the follow-up GoBifo/ENCISS/Decentralization survey, as well as the Justice for the Poor and Understanding Processes of Change in Local Governance qualitative research program rather than to draw conclusions about the experience and resolution of crime and conflicts in Sierra Leone. Among the many issues and questions raised by these findings, we would like to highlight a few: The importance of chiefdom in determining whether and where crimes and conflicts are reported and whether people are satisfied with the results suggests that behaviors and outcomes vary with factors that are not institution-specific (e.g., the characteristics of local courts versus paramount chiefs courts) but location-specific (e.g., the characteristics of the local courts and paramount chiefs courts in X chiefdom). 25 Involvement, Reporting, and Satisfaction by District and Chiefdom Bonthe Bum Imperi Jong Kwamebai Krim Bonthe Town Biriwa Bombali Sebora Gbanti Kamaranka Gbendembu Ngowahun Bombali Makari Gbanti Safroko Limba Sanda Loko Sanda Tendaran Sella Limba fraction involved satisfaction rate reporting rate 24 In addition, tests of statistical significance do not account for the clustered nature of the sample. 25 The chart below shows differences by chiefdom and district in likelihood of being involved in (or a victim of) the specified crimes and conflicts, the likelihood of reporting, and the likelihood the matter was resolved satisfactorily. An earlier chart highlighted differences in where crimes or conflicts were reported. 14

16 The differences in satisfaction by institution may be explained by differences in which conflicts are reported to which institution, or may reflect differences in the way these institutions resolve crimes and conflicts, and how effectively they serve people s needs. Differences must also be understood in relation to the fact that the survey asked only where crimes and conflicts were first reported, not where they were ultimately resolved. Nearly half of crimes are not reported to any authority outside of the family. This could indicate that resolution was sought through alternative mechanisms not considered in the study (such as women s or youth groups, or direct mediation between families), or may result from calculations concerning the expected costs and benefits of pursuing justice within the available structures. We would also like to note just a few of the many unresolved questions: How do the crimes and conflicts discussed here compare with other types of crimes and conflicts experienced by people in Sierra Leone? How would the results change if we were asking respondents not only about where they first reported a crime or conflict, but where (if anywhere) they pursued the matter beyond that? Where were these crimes or conflicts ultimately resolved or abandoned? Readers familiar with the landscape of conflict and justice in Sierra Leone will likely have additional issues that they would raise. We welcome any comments or suggestions. 26 Conclusions Although the findings of the GoBifo/ENCISS/Decentralization study cannot be considered conclusive, they do provide a starting point for an informed discussion of local level justice in Sierra Leone s provinces. We hope that this study, accompanied by additional quantitative and qualitative research into the incidence and resolution of conflicts, will better equip international institutions, government agencies, and community-based organizations to understand and facilitate access to justice in Sierra Leone. 26 Please address any correspondence to efoster@ircbp.sl, pdale@worldbank.org, and rmanning@worldbank.org. 15

17 Appendix A: Survey Questionnaire Reproduced below are the questions of the GoBifo/ENCISS/Decentralization survey that were discussed in this paper. Questions on Crimes and Conflicts SECTION 9: Conflict Resolution Capacity (Selected Questions) 1. Have you or a member of your household ever had a conflict with someone over a loan or other money business? (1 = Yes, 2 = No) 1e. IF YES to 1, In your opinion, has the conflict been resolved satisfactorily? (1 = Yes, 2 = No) Question A. 1=Yes 2=No 3. During the last 12 months, has someone stolen any livestock from you? 4. During the last 12 months, has someone stolen any household items? 5. During the last 12 months, has someone stolen any money from you? 6. During the last 12 months, have you been involved in any physical fighting without weapons? 7. During the last 12 months, have you been involved in any physical fighting with weapons (such as a club, machete or gun)? 8. During the last 12 months have you been a victim of witchcraft (juju)? IF NO, GO TO NEXT ROW B. IF YES, How many times? C. To whom (outside your family) did you first report the most recent crime/dispute? 01=Village headman court 02=Section chief court 03=Local court (NA) 04=Paramount chief court 05=Magistrate court 06=Police 07=Secret Society 08=WDC 09=VDC 10=NGO 11=Mosque/church 12=Village Elders 13=No one (skip to E) D. What was primary outcome of the most recent crime/dispute? 1=Fine 2=Jail 3=Stolen item returned 4=Dismissed with no punishment 5=Other E. In your opinion, has the matter been resolved satisfactorily? 1=Yes 2=No 16

18 Questions used to create independent variables SECTION 1: HOUSEHOLD DETAILS 1A. Household Identification SUPERVISOR: Fill in the Chiefdom and Village Codes from your list: 5. Chiefdom Code 1B. Respondent Identification 3. Respondent Gender (1 = Male, 2 = Female) 4. Respondent s year of birth (YYYY) 1C. Household Characteristics 1. Religion (see RELIGION Codes) 2. Ethnicity (see ETHNICITY Codes) 4. What is the present marital status of the head of household? (1 = Married, monogamous, 2 = Married, polygamous, 3 = Informal/loose union, 4 = Divorced/separated, 5 = Widowed, 6 = Never married) 5. Is this household from a chief/ruling family (eligible to run for paramount chief? (1 = Yes, 2 = No) Building Materials For the main house/hut where the respondent sleeps, describe the following: 6. Roofing material (1 = Thatch (grass/straw), 2 = Wood, 3 = Corrugated iron (zinc/tin), 4 = Tarpaulin (plastic sheets), 5 = Cement/Concrete, 6 = Roofing tiles, 7 = Other) 7. Material of Walls (1 = Mud/mudbricks/wattle; 2 = Wood, 3 = Corrugated iron (zinc/tin), 4 = Stone/burnt bricks, 5 = Cement/sandcrete, 6 = Other) 8. Material of floor (1 = Earth/mud, 2 = Wood, 3 = Stone/brick, 4 = Cement/concrete, 5 = Other) 9. What is the primary source of drinking water for your household now (this week)? (1 = Piped indoors, 2 = Piped in compound, 3 = Public tap, 4 = Protected ordinary well, 5 = Unprotected ordinary well, 6 = Mechanical well, 7 = River/riverbed/stream, 8 = Neighbor s tap, 9 = Water vendor, 0 = Other) 12. Does any member of the household own any of the following items? CHOOSE ALL THAT APPLY (1 = Yes, 2 = No) 17

19 (A) Radio (B) Radio cassette 13. During the last 12 months, has any member of your household raised sheep? (1 = Yes, 2 = No) 14. During the last 12 months, has any member of your household raised poultry other than chickens (i.e. ducks)? (1 = Yes, 2 = No) 15. If you needed to make a phone call TODAY, would you be able to access a phone? (1 = Yes, 2 = No) SECTION 2: WAR EXPERIENCE 2A. Household War Experience 6. Were you living in this same village before the war started (i.e., in 1990)? (1 = Yes, 2 = No) Section 2B: Household Roster (L) What is the highest level of schooling EVER attained by this person? (see EDUCATION codes) SECTION 3: Productive Activities 3A. Agriculture Activities 11. In the calendar year 2005, did you work on a communal farm (this means a farm owned by the community where community members work on the farm)? (1 = Yes, 2 = No) 12. In the calendar year 2005, did you work as part of labor sharing gang (a group that works together on their own farms, rotating across each member s farm OR a group that works together for hire)? (1 = Yes, 2 = No) SECTION 4: Groups 1. Are you a member of any groups that do credit or savings (osusu, cooperatives, etc.)? (1 = Yes, 2 = No) 10. Are you a member of any social groups (i.e. one that organizes dances, sports matches, etc.)? (1 = Yes, 2 = No) 14. Are you a member of a Parent Teacher Association (PTA), Community Teacher Association (CTA) or School Management Committee (SMC)? (1 = Yes, 2 = No) SECTION 5: Trust 1. Suppose that someone in another village needs to pay you some money. If you and your family are unable to go (for example, you are sick and your family is away) would you ask someone in the community (not a household member) to go pick up the money for you? (1 = Yes, 2 = No) 1A. IF YES, who would you ask to collect the money for you? ENUMERATOR: Check that the respondent does not specify a member of the household (1 = Friend, 2 = Neighbor, 3 = Elder, 4 = Anyone in the community) 18

20 2. Has this ever happened to you (i.e. that you needed to ask a non-household member to collect money for you from another village)? (1 = Yes, 2 = No) 3. Suppose you were at a community meeting and you accidentally left your purse/wallet/some money on the bench. If you go back to get it one hour later, will it still be there? (1 = Yes, 2 = No) 4. Has this ever happened to you (i.e. that you left some money somewhere in the village)? (1 = Yes, 2 = No) 5. Tomorrow, if you needed to buy something from town or the market but were unable to travel there, would you give your money to someone from the community (not a household member) to buy the item for you? (1=Yes, 2=No) 5A. IF YES, to whom would you give your money to buy goods for you at town/market? ENUMERATOR: Check that the respondent does not specify a member of the household (1 = Friend, 2 = Neighbor, 3 = Elder, 4 = Anyone going to market) 6. Have you ever done this (i.e. given money to a non-household member to buy something for you at town/market)? (1 = Yes, 2 = No) 7. How much do you trust the following people? (1 = Trust totally; 2 = Trust somewhat; 3 = Neutral; 4 = Distrust somewhat; 5 = Distrust totally) (A) Local government officials (B) Central government officials (C) Chiefdom officials (D) Youth in this community (E) Women in this community (F) Men in this community (G) Elders in this community (H) People who recently joined this community (I) People from outside this community (J) Health workers (doctors, health officers, nurses) (K) Teachers SECTION 7: Access to Information 3. Are you able to name the Local Councillor from your ward? (1 = Yes, 2 = No) 3a. IF YES, specify name: SUPERVISOR: Correct? (1=Yes, 2=No) 5. Are you able to name the Section Chief for your section? (1 = Yes, 2 = No) 5a. IF YES, specify name: SUPERVISOR: Correct? (1=Yes, 2=No) 6. Are you able to name the Paramount Chief for this chiefdom? (1 = Yes, 2 = No) 6a. IF YES, specify name: SUPERVISOR: Correct? (1=Yes, 2=No) 19

21 16. What is your primary source of information about the government and politics? (01 = Radio, 02 = Newspaper, 03 = Friends/family, 04 = Chief/village headman, 05 = VDC/WDC/Local Council, 06 = Community notice board/posters, 07 = Religious leader, 08 = Teacher, 09 = Political party representative, 10 = Campaign/political rally, 11 = TV, 12 = Respondent s own observations, 13 = Other, specify: ) SECTION 8: Local Governance 1. Did you or anyone in your household run for Village Development Committee membership? (1 = Yes, 2 = No) 4. Have you ever had personal contact with a Ward Committee Member? (1 = Yes, 2 = No) 5. Have you ever attended a Ward Committee meeting? (1 = Yes, 2 = No) 13. In the last 7 days, have you spoken with the village head man? (1 = Yes, 2 = No, 8 = Headman is part of respondent s household, 9 = Don t know who the headman is) 14. In the last 7 days, have you spoken with the women s leader for this community? (1 = Yes, 2 = No, 8 = Women s leader is part of respondent s household, 9 = Don t know who the women s leader is) 15. In the last 7 days, have you spoken with the youth leader for this community? (1 = Yes, 2 = No, 8 = Youth leader is part of respondent s household, 9 = Don t know who the youth leader is) 20

22 Appendix B: Independent Variables Tested Basic Demographics Ethnicity (For each district, I kept only those ethnic groups that were more than 3% within the district, the others became part of other. In Bonthe this meant keeping Mende and Sherbro, and in Bombali Temne, Loko and Limba.) [Section 1C, question 2] Religion (Christian or Muslim only as less than 1% reported being anything else) [Section 1C, question 1] Ethnic Concentration Index by Chiefdom (sum of squares of shares of ethnic groups within the chiefdom) [based on Section 1C, question 2 and Section 1A, question 5] Age [Section 1B, question 4] o as a continuous variable o youth versus non-youth (with the cut-off at 25 or 35) o as a categorical variable (categories: about 20, about 25, about 30, about 35, about 40 or 45, about 50 to 60, definitely over 60. This creates approximately equally-sized groups and reflects the fact that many people do not know their age exactly and estimate to the nearest 5) Gender [Section 1B, question 3] Education Level (categories: none, primary or higher. Only 12% of respondents had any education past primary) [Section 2B, question L for respondent] Migration (was the household living in the same village before the war) [Section 2A, question 6] Position and Income Proxies HH from a ruling family [Section 1C, question 5] Head of the HH divorced [Section 1C, question 4] Head of the HH widowed [Section 1C, question 4] Have an improved (non-thatch) roof [Section 1C, question 6] Have improved (non-mud) walls [Section 1C, question 7] Have an improved (non-earth) floor [Section 1C, question 8] Housing quality index (average of above three) [Section 1C, questions 6-8] Use improved (non-river/riverbed/stream) water source [Section 1C, question 9] Own a radio [Section 1C, question 12] Own a radio cassette [Section 1C, question 12] HH raise sheep [Section 1C, question 13] HH raise poultry other than chickens [Section 1C, question 14] Have access to a phone [Section 1C, question 15] Trust Objective (fraction of objective trust questions answered positively) [Section 5, questions 1, 3 and 5] Trust all (how much they trust all categories of people normalized to be between 0 and 1) [Section 5, questions 7] Trust community members (how much they trust youth, women, men and elders in the community normalized to be between 0 and 1) [Section 5, questions 7] 21

23 Trust government officials (how much they trust local government officials, central government officials, chiefdom officials normalized to be between 0 and 1) [Section 5, questions 7] Trust outsiders (how much they trust people who recently joined this community and people from outside this community normalized to be between 0 and 1) [Section 5, questions 7] Groups Work on communal farm [Section 3A, question 11] Work as part of labor sharing group [Section 3A, question 12] Belong to credit / savings group [Section 4, question 1] Belong to social group [Section 4, question 10] Belong to school related group [Section 4, question 14] Total number of these groups involved in (between 0 and 5) [Section 3A, questions 11 and 12 and Section 4, questions 1, 10 and 14] Linkages Anyone in HH run for VDC [Section 8, question 1] Had personal contact with WDC member [Section 8, question 4] Ever attended WDC meeting [Section 8, question 5] Spoken with village headman in last 7 days [Section 8, question 13] Spoken with women s leader in last 7 days [Section 8, question 14] Spoken with youth leader in last 7 days [Section 8, question 15] Access to Information Correctly named Local Councillor [Section 7, question 3] Correctly named Section Chief [Section 7, question 5] Correctly named Paramount Chief [Section 7, question 6] Main source of information about government and politics (categories: radio, friends / family, village headman. All possible responses given by less than 1% of respondents each) [Section 7, question 16] 22