No. 19 / July 2004 Local Conflict in Indonesia: Incidence and Patterns Local conflict characterizes many countries such as Indonesia, but research has tended to focus on large- scale, `headline' conflicts. Using a unique dataset that maps conflict incidence across all of Indonesia's 69,000 villages, and combining quantitative and qualitative methods, the study applies an empirical framework to analyze potential links between local conflict and poverty, inequality, shocks, ethnic and religious diversity, and community-level associational and security arrangements. Positive correlations with conflict include unemployment, inequality, natural disasters, changes in sources of incomes, and clustering of ethnic groups within villages. Background. Major intrastate conflicts, such as civil time to map conflict incidence across all of Indonesia's wars, and the resulting set-backs for economic and 69,000 villages and neighborhoods, this study political development, have received increasing research investigates factors that appear to heighten the likelihood attention in recent years. A growing literature has sought of violent local conflict.2 Over 7% of Indonesia's lowest to investigate conditions that affect the probability of administrative tier (rural desa and urban kelurahan) in the countries experiencing large-scale violence, in particular census reported local conflict during 2002, suggesting the likelihood of civil war. However, many developing that conflict is significant across the archipelago, rather countries are affected by high levels of communal and than being concentrated in a few (headline) areas. inter-communal conflict that does not take the form of a Almost one-quarter of reported conflicts involved civil war, but nonetheless results in significant casualties, fatalities, about one-half injuries, and about one-third destruction of livelihoods and property. In Indonesia, for material damage. Total fatalities from conflict were example, only Aceh would fall under the Collier- reported to be 4,869, injuries 9,832 and material damage Hoeffler1 definition of a civil war, while high profile Rp 771 billion ($91.4 million). As discussed below, conflicts in East Timor, Central and West Kalimantan, PODES data likely underestimates local conflict. Central Sulawesi, West Papua, and the Malukus have also captured the headlines. Pervasive and widespread local Conceptual Framework. Measuring conflict incidence raises two fundamental issues: (i) the exact definition of conflict is not only a barrier to development but in some cases threatens to escalate into larger incidents of unrest conflict and the threshold at which an incident is defined or even fully-fledged violent conflict. Countries as violent conflict within a geographic jurisdiction; and (ii) the conditions under which respondents might undergoing difficult political and economic transitions appear especially vulnerable. misreport (or perceive differently) instances of conflict. Reporting biases are of special concern for statistical Yet to date, systematic evidence on the presence of local evaluations if these are themselves associated with some conflict across a large sample of small geographic units, of the factors posited to be associated with conflict. and on the potential underlying factors that might be associated with its incidence, has been limited. The The paper defines local conflict in terms of its incidence and impact measured within a locality (desa and extent of local violence across a country such as Indonesia (and variations in conflict type, incidence and kelurahan), while the causes of local conflict may be impact) suggests that intra-country empirical endogenous or extra-local. The Indonesia statistical agency has adopted a definition of local conflict beyond investigations offer an invaluable starting point to gain a deeper understanding of conflict dynamics. Using a some threshold of violence within a given locality in the unique dataset compiled by the Government (the Village past year, that may in turn have been associated with loss of life, serious injury, or property damage. Responses are Potential Statistics, PODES), which seeks for the first compiled from village leaders and central government statistical agents (mantri statistik ) posted at the sub- Economic deprivation or wealth disparities may increase district level. grievances. Economic shocks could lead individuals or groups to transgress on the property of others, thereby The complexity of conflict ideally demands mixed triggering conflict escalation. Out-migration may denote quantitative-qualitative approaches. Given the challenges lack of economic opportunity within a village, or in capturing local conflict as a discreet incident, the paper indirectly lead to a decline in local social capital as explicitly integrates an in-depth qualitative conflict individuals invest less in community associations as they mapping exercise conducted in eight sub-districts, as well expect their future to be elsewhere. Structural as comparisons to other quantitative data. Comparing the transformation away from (particularly subsistence) quantitative data with that gathered from a large-scale ethnographic World Bank study on conflict,3 suggests agriculture may increase uncertainty over incomes, may necessitate new institutions (e.g., for property rights) as some weaknesses in the PODES data, particularly that traditional/informal mechanisms are no longer able to PODES significantly under-reports levels of conflict. cope with increasing land scarcity and new economic The qualitative research found conflict in 21 of 25 structures. Such transformations could also stretch purposively selected villages in two provinces (East Java traditional modes of community level conflict resolution, and NTT). These included mass sickle fights, the burning either because these institutions cannot cope with new of witch doctors, and large-scale inter-group land situations, or because of growing tensions between formal disputes. Only 4 of the same villages reported that they and informal institutions. While formal land tenure may had experienced conflict in PODES. Likewise, PODES not necessarily denote security of tenure, the absence of under-reported the impacts of conflict (in terms of deaths, formal property rights for private agricultural land may injuries and material damage) in the research villages. make this a trigger for violence, especially as land grows There may be two explanations for this. The first more scarce. Similarly, property rights over natural concerns how respondents understand what is meant by resource rents from forests and mining are often ill- conflict. The second relates to the incentives for village defined, and highly contested, again precipitating or heads to accurately report conflict. Without clear escalating conflict. knowledge of the survey's purposes, village heads may understate the problem, since conflict is often seen as a The existence of various ethnic, religious, and/or sign of failure, with village harmony valued. The key linguistic fault lines could enhance the likelihood of question, however, is whether under-reporting is biased conflict. The presence of multiple ethnic groups has (levels of under-reporting differ across locations) or often been associated with lower growth and a higher systemic (underreporting is similar across sites). The probability of conflict or even civil war. Social sanctions study suggests that the latter is more likely, but that more against shirking may also be more difficult across ethnic research (perhaps by comparing different data sources on groups, hence weakening cooperative outcomes. more areas) needs to be conducted. Different groups may operate under different rules systems that allow for the peaceful resolution of conflict; Factors Affecting the Likelihood of Local Conflict. but when these competing rules systems come into Observers have emphasized different factors that may be contact, conflict escalation can result. Collier4 argues that associated with conflict and/or which are important to rather than ethnic diversity, ethnic dominance (e.g., a mediate tensions in the Indonesian context. They include, single group accounting for 45-90% of total population) but are not limited to, poverty, inequality, income may be a better predictor of conflict risk. Local conflict shocks/uncertainty, unemployment and inequitable in Indonesia is often triggered across jurisdictions, so that development, the effect of industrialization, measures of group "clustering" (or to use a more decentralization, lack of clarity in land rights, regional normative term, segregation) may be associated with economic disparities and natural resource management., conflict. In addition, conflict can result from social dynamics of group friction (e.g., across religious and ethnic groups); Stewart5 has argued that horizontal inequalities-- and institutional factors (e.g., the extent to which conflict differences between groups in political opportunities, is effectively mediated or, conversely, sanctioned by the economic assets, employment and incomes, and social state apparatus, including the police, and the legitimacy access--play an important role in determining when and and effectiveness by which power/leadership is exercised where violent conflict will take place. Varshney's6 work in a community). The study's hypotheses are derived on India has argued that fragmented social structures, from the available international literature and the on- where participation in associational structures and going qualitative work. These factors and the mechanisms informal networks consolidates identities and interests by which they could be associated with conflict are along, rather than across, religious and ethnic lines, briefly reviewed before presenting the results. explains variation in violence and peace in otherwise 2 similar areas. He argues that especially in urban areas, distinguishing between rural and urban areas. Estimates day-to-day (quotidian) interactions are insufficient to were calculated with and without Aceh. provide a bridge across ethnic/religious groups that stops conflicts from escalating into communal violence. The study found that poverty by itself has very little Communal violence is less likely where associations can correlation with conflict. Changes in economic conditions, on the other hand are correlated. act as bridges across groups. Unemployment is universally closely associated with Local institutions and leadership could in principle higher conflict rates, which is not surprising as provide the mechanisms for reducing or mediating unemployed young men may be especially associated conflict in line with Varshney's hypothesis. On the other with local conflict. High in- and out-migration for hand, these same institutions may also be instrumental in work/school related reasons in rural areas (though not in mobilizing individuals to engage in conflict, increasing urban areas) have similar positive associations with the capability for collective action to defend group conflict. Land variables indicate that the share of interests. The analysis distinguishes between political agricultural land converted to other purposes is associated institutions, religious institutions, and informal religious with higher levels of conflict, whereas the share of titled and adat (traditional customs) organizations. The land has no correlation with conflict. The presence of a introduction of the democratic village council (BPD) is a natural disaster is also clearly associated with higher centrally mandated institutional change aimed at conflict. Mining and forestry as main sources of increasing accountability of village leaders. By the time household incomes did not significantly affect conflict, of the PODES fielding 80% of villages had instituted a except in East Java for villages that depend primarily on BPD. Female leadership may also affect the ability to mining. avert violent conflict, to the extent it may be indicative of a community's openness to broader participation of Much of the headline conflict in Indonesia is reported as typically excluded groups. clashes between ethnic groups. The 2000 Census for the first time collected self-reported ethnicity (over 1,000 Religious institutions could form a bridge to overcome different ethnic groups were reported) and the study used differences between their followers. More formal this data to investigate the correlation between conflict institutions, organizing larger groups of people, have a and ethnicity. There is no evidence that increased ethnic potentially more important role to play in avoiding diversity affects conflict, but there was some evidence conflict. The presence of adat institutions is probably that lack of dominance or high fragmentation, defined as indicative of an increased adherence to traditional laws no single group having more than 40% of a sub-district and customs. While at times these local institutions may was associated with higher conflict. As expected, higher be more successful in solving community problems, they clustering of ethnic groups within villages across a sub- may also clash with centrally mandated conflict district is associated with higher levels of conflict. The resolution mechanisms, providing the potential for main counter intuitive finding is that higher horizontal conflict escalation. inequality, measured as the difference between educational attainment of large ethnic groups, is Legal and institutional enforcement weaknesses associated with lower levels of conflict. Further research (inadequate policing, ineffective courts and judicial is needed to investigate what explains this correlation. systems) that allow for high levels of crime tend to also create space for violent conflicts to take place; incidents Institutions matter for mediating potential conflict. The of crime, when severe, may also constitute conflicts in study investigated the correlation between the presence of themselves. Public and communal security arrangements institutions and conflict, while recognizing that their very may therefore reduce the likelihood of conflict. presence may result from conflict. We find that the However, these are especially subject to endogeneity density of places of worship and the presence of the bias, as their existence may be in response to a worsening democratically-elected BPD are associated with lower security situation. levels of conflict. Religious groups and adat organizations, however, are associated with higher levels Results and Conclusions. The study conducted a of conflict. One possible explanation may be that their quantitative analysis to determine which locally-specific presence reflects the differences in norms about conflict factors are correlated with the presence of conflict. It resolution. The qualitative work indicated that often derived proxies for local poverty, inequality, diversity, varying expectations between different groups in the economic shocks, land tenure, and community political, community, or between the community and the state, associational and security arrangements. A logit model about how tensions should be resolved resulted in was applied and estimated at the village level, conflict. The presence of a female village leader is associated with lower conflict in rural areas, but this is 3 still a rare phenomenon (only 2% of rural communities effective proxies for the effectiveness of local institutions had a female village leader). The presence of security that might help mediate conflict. The study in its present posts is associated with lower levels of conflict in rural form is only effective in identifying underlying areas as is the distance from a police post. associations with conflict. These should provide a stimulus to qualitative and quantitative work to identify A more systematic mapping of conflict should also pathways and directions of causality. In a mixed stimulate work on assessing the impacts of conflict in quantitative-qualitative framework, future refinements of Indonesia, not just in high profile conflict areas but the quantitative evidence and methodological frameworks beyond. An appreciation of the pervasive costs of could in the future lead to better hypothesis testing and conflict to individuals, firms, and society, and the the derivation of potential policy findings. This study is incidence of these effects, will alert policy makers to the but an initial presentation of the results and implications real costs of conflict. For example, while the study of the 2003 PODES on conflict. The authors hope that it suggests that it is not the poorest communities that will stimulate further debates and consideration, both of experience conflict, the costs of conflict may fall the causes of conflict in Indonesia, and of the disproportionately on poor and vulnerable households. methodological tools that can be used to measure and Future work along the lines presented in this paper will understand local conflict in development more generally. need to focus on a number of areas, including: (i) improved methodologies and instrument implementation to measure the incidence of local conflict; (ii) better 1World Bank. 2003. Breaking the Conflict Trap: Civil War and empirical proxies at the local level to measure potential Development Policy. Washington, DC. factors associated with conflict; and (iii) increased 2P. Barron, K. Kaiser and M. Pradhan. 2004. Local Conflict in attention to empirical research design to begin to address Indonesia: Measuring Incidence and Identifying Patterns, draft issues of causality. (anticipated as a World Bank Policy Research Working Paper). 3Barron, P., R. Diprose, D. Madden, C. Smith, and M. With its unparalleled national coverage, the PODES Woolcock. 2004. "Do Participatory Development Projects Help initiative provides an unprecedented insight to the Villagers Manage Local Conflicts?" Social Development incidence of local conflict beyond those areas that have Papers: Conflict Prevention & Reconstruction, CPR Working tended to capture the headlines. PODES relies on Paper No. 9. Washington, DC: World Bank. village/neighborhood leadership as key respondents, 4Collier, Paul. 2001. "Ethnic diversity: An Economic under the assumption that they are the most informed and Analysis." Economic Policy, pp. 128-66. 5 able to report conflict at the local level. The approach has Stewart, Frances. 2002. "Horizontal Inequalities: A Neglected the potential to provide a systematic mapping of conflict Dimension of Development." QEH Working Paper No. 81. by administrative area, something well beyond the ability Queen Elizabeth House, University of Oxford: Oxford, 6 of household surveys, given sampling requirements. As Varshney, Ashutosh. 2002. Ethnic Conflict and Civic Life: Hindus and Muslims in India. New Haven: Yale University the study points out, there may be certain factors that can Press. cause underreporting of conflicts. In future enumerations of the survey, the statistical agents who implement the This CPR Dissemination Note was prepared by Ian Bannon survey could be encouraged to cross-check the data on (CPR), based on the report by Patrick Barron (EASES), Kai conflict with other key respondents, while at the same Kaiser (PRMPS) and Menno Pradhan (EASPR) (2004) Local time gaining a better understanding of their reporting Conflict in Indonesia: Measuring Incidence and Identifying incentives. Additional work might compare these results Patterns. The work was supported by DfID, AusAID and Dutch trust funds. This Note was also published as Social with more targeted surveys of household respondents in a Development Note No. 95 and is part of a series intending to subset of localities. The study recommends that the local disseminate good practice and key findings on conflict conflict module be retained in the 2005 PODES, but that prevention and reconstruction. The series is edited by the CPR it consider some refinements to the module in terms of Unit in the Social Development Department of the definitions/thresholds, sequencing of questions, and the Environmentally and Socially Sustainable Development training of statistical agents. Network of the World Bank. CPR Dissemination Notes are distributed widely to Bank staff and are available on the CPR The study has presented a number of potential measures website http://www.worldbank.org/conflict and can also be for local level inequality, diversity, and group inequality. requested via e-mail at cpr@worldbank.org Future qualitative work might gain a better understanding of group dynamics at the local level, and how these could be mapped in local indicators (for example from the Population Census), and then mapped across datasets. Similar efforts should be made in devising simple but 4