Environmental Disasters and Subnational Conflict: A Study on the Effects of Environmental Disasters and Environmentally-Induced Migration on Subnational Conflict

Abstract: 

As climate change progresses, the alteration of weather patterns will invariably follow, leading to an increase in the frequency and magnitude of natural disasters. However, the effects that changes in weather patterns will have on human population and conflict is not yet clear. This study argues that the occurrence of an environmental disaster will induce intrastate migration among those who reside in the affected area. Environmentally-induced migration, in turn, will lead to the presence of low-level civil conflict within the state as large numbers of migrants rapidly enter a host community and alter resource availability, economic conditions, and cultural composition. Through the use of a negative binomial regression, this study analyzes the relationship between environmentally-induced migration and the onset of conflict at the subnational level in 27 Indian states from 1956 to 2002 and 98 sub-provincial units on the island of Java in Indonesia from 1994 to 2003. The results indicate areas struck by environmental disasters and probable environmental migrant destinations are associated with conflict in India; however, the significance of migration locations decreases as control factors are added into the equation while disasters remain significant. These results differ from the findings in the Indonesia case where neither disasters nor probable migration locations are strongly associated with conflict. This indicates disasters are related to causes of conflict at the subnational level, but also that institutions of the state play a major role in the occurrence of conflict following disasters.

Table of Contents: 

    Introduction

    Climate change is altering environmental conditions, such as land and water availability, as well as leading to an increase in the frequency and intensity of severe environmental disasters such as floods, droughts, heat waves and hurricanes (IPCC 2007). An increase in human migration is expected to occur as environmental disasters become more severe and frequent, and as land degradation leaves many in the world searching for new means of sustenance. As of 2008, there were 36 million environmentally displaced people with at least 20 million displaced due to environmental disasters (IDMC 2009). Will the increase in environmentally-induced human migration due to environmental disasters increase the occurrence of civil conflict at the subnational level? It has been argued that the increase in environmentally-induced migration may lead to an increase in the onset of civil conflict (Homer-Dixon, 1994; Reuveny, 2007; Swain, 1996); however, these previous studies have been based mostly on case analysis and rely heavily on theoretical arguments rather than empirical testing to determine the relationship between environmental migration and civil conflict. Furthermore, much of the previous literature concerning the environment and conflict looks at conflict among states rather than within them at the subnational level of analysis. This study argues that environmental disasters and the subsequent environmental migration will increase conflict at the subnational level because of increased resource scarcity and tensions among groups. If environmental disasters and subsequent migration increases localized subnational conflict, it is important to understand why and under what circumstances this occurs in order to prevent escalation to widespread conflict and intrastate war.

    This study will analyze the relationship between environmental migration and the instances of intrastate conflict in hopes of determining if environmental disasters and environmentally-induced migration increases conflict through the use of a negative binomial model. The results indicate various conditions under which environmental disasters are more likely to increase conflict in both the areas in which they occur and in the areas surrounding these disaster locations.

    The remainder of this study will review previous work regarding climate change, environmental migration, and conflict. The following section reviews previous research and develops a theoretical framework relating environmental disasters, migration, and conflict. From this theory, hypotheses are derived and a research design is outlined. The study uses empirical analysis to test the relationships between environmental disasters and subnational conflict, and the conclusion considers these relationships and evaluates their implications.

    Literature Review

    Climate Change and Migration

    In recent years, scholars have become interested in the potential impact climate change could have on both the lifestyles of the human population and upon migration patterns. Environmental disaster migration differs from the economic variant because this type of migration is typically forced on the individual. Voluntary migration is often due to economics. Decisions of where to reside are based on economic, social, and institutional reasons. Forced migrants, on the other hand, are pushed away from their locations due to civil wars, political reasons, and environmental catastrophes (Swain 1996). Environmental disasters leave an area devastated and thus force residents from their homes. This migration may be long-term or short-term depending on the rebuilding process following the disaster.

    Beniston (2010) argues that climate change in the 21st century has the capability to threaten food and water, security as well as air quality, and rainfall patterns, which will require the human population to make a variety of adjustments. The alteration of precipitation patterns and loss of arable land could potentially have devastating effects on cultures that rely primarily on agriculture for livelihood; this is especially true in less developed countries where there is a lack of technological and economic resources to provide alternate lifestyles for inhabitants of the affected areas (Reuveny 2007). Many of the poorest countries in the world can be expected to be most drastically impacted by the change in climate. The loss of agricultural land compounded by the rapid pace of population growth in these countries will further deplete resources such as land and water (Beniston 2010; Nordås & Gleditsch 2007). Changes such as these are “likely to undermine the capacity of states to provide the opportunities and services that help people to sustain their livelihoods,” affecting weak or unstable states in particular; this means that it will be difficult to provide aid and alternate resources for citizens who are dependent on land (Barnett & Adger 2007).

    Alteration of the environment and weather patterns will also increase the frequency and magnitude of large-scale environmental disasters, including events that occur over the long term and rapidly occurring events (Beniston 2010). Extreme droughts, heat waves, and fires will become more likely, posing threats to societies dependent on agricultural output and may cause an increase in rural to urban migration (Beniston 2010; Nordås & Gleditsch 2007). These types of events are long-term disasters and will cause migration over time as residents begin to struggle economically, and then to choose to leave their community. Rapidly occurring environmental disasters such as floods, cyclones, and hurricanes will contribute to the forced migration of large amounts of people. This migration is forced because of the magnitude of the disaster occurs quickly and devastates the community. Reuveny (2007) states these disasters are likely to cause mass migration from one region to another, citing Hurricane Katrina and subsequent migration as an example.

    Environment and Conflict

    Although scholars agree climate change is likely to lead to greater levels of migration, there is less agreement on the relationship between these environmental changes and the likelihood of civil conflict. Homer-Dixon (1994) states that increasingly scarce resources coupled with environmental degradation will lead to an increase in the occurrence of conflict around the globe. This conflict may occur for two reasons, according to Homer-Dixon: resource capture and ecological marginalization. Homer-Dixon (1994) describes instances of resource capture or ecological marginalization and conflict in history; however, he does not empirically test the relationship between an increase in scarcity and subsequent instances of conflict. Meier, Bond, and Bond (2007) disagree with Homer-Dixon when looking at the relationship between precipitation, vegetation, and pastoral raids in the Horn of Africa. In fact, it was found that conflict was not dependent on rainfall and vegetation but that it was most likely to occur when there were already patterns of aggressive behavior and a lack of a means of mitigation (Meier et al. 2007). Similar results were found in an empirical study regarding the drylands of Kenya in terms of scarcity and conflict (Adano et al. 2012). The authors found that the scarcity of resources during droughts caused lower levels of conflict, raids, and human deaths than times of abundant resources.

    The exclusion of external factors, such as pre-existing aggressive behavior, from the study of climate change and environmental scarcity is the basis for many scholars’ arguments that climate change causes conflict; however, this places too much emphasis on the environment and not enough on external pressures (Barnett & Adger 2007; Butler & Gates 2012; Gleditsch 2012; Raleigh & Urdal 2007; Salehyan 2008). Salehyan (2008) argues that past scholarly research has relied too much on environmental determinism and notes that for every instance of environmental scarcity and onset of conflict there are many more instances where this was not the case and scarcities did not spur conflict. Gleditsch (2012) also states that the relationship between degradation and conflict is weak and that previous models from scholars that produce contrary findings fail to account for other variables. While a change in environmental conditions and resource scarcity may increase conflict, this will not occur on its own and factors such as governmental institutions, GDP, and property rights all need to be taken into account (Barnett & Adger 2007; Salehyan 2008). The importance of the state in the environment and conflict relationship has been demonstrated in a study on African range wars and conflict which analyzed the relationship between resource scarcity and conflict between pastoral groups (Butler & Gates 2012). The authors find that violence among pastoral groups is not strictly resource based, but has much to do with property rights. They conclude that analyses of climate change and conflict need to incorporate characteristics of the state (Butler & Gates 2012).

    Environmental Migration and Conflict 

    The projected increase in migration due to environmental factors and potential of environmental scarcities increasing conflict has led to a number of analyses that consider this relationship. Prior research on migration has found that migration can be a source of conflict among host-migrant groups (Weiner 1992). Economic and political factors force migrants from their homes due to political repression or environmental degradation, and host communities are often unwilling to accept these migrants. States are often unwilling to host migrants due to already scarce resources, and other economic, ethnic, religious or political reasons (Weiner 1992).

    In an analysis of 38 states that experienced environmental migration conducted by Reuveny (2007), 19 of the areas did not generate conflict. Of these 19, 14 of the areas involved intrastate migration (Reuveny 2007). It was also noted that conflict was most evident in LCDs and regions that depended on the environment for survival, as well as areas facing resource scarcity. A shortcoming of this study is that this data was not empirically tested; Reuveny’s study was an assembly of environmental migration cases and recordings of their outcomes, making it hard to form generalizations. Reuveny (2007) lists five possible reasons environmental migration may cause civil conflict, including ethnic tensions, distrust, socioeconomic fault lines, auxiliary conditions, and resident-migrant competition. Ethnic tensions have the potential to lead to conflict because these tensions can fuel perceived threats in host communities (Reuveny 2007; Weiner 1992).

    Socioeconomic fault lines occur among different groups, such as the rural and urban portions of a society. Benjaminsen et al. (2012) discusses socioeconomic fault lines and states that differences in lifestyles and economic activities created tension between farmers and pastoralists in the Sahel. While the study did not demonstrate that water scarcity and climate change produced conflict in the region, it illustrated that conflict can occur along the farmer-pastoral fault line because the groups get into conflicts over access to water for different agricultural purposes. These fault lines may also be prevented by better auxiliary conditions. Auxiliary conditions refer to governmental institutions and conditions of the state that could affect conflict onset. Raleigh and Urdal (2007) address these auxiliary conditions regarding the likelihood of conflict occurring as the result of climate change. Although their study does not involve migration, their findings shed light on the relationship between climate change and conflict. Understanding the manner in which a state operates and handles environmental impacts is necessary when evaluating conflict. Beyond a consideration of governmental institutions, the core values, management styles, and availability of resources are also salient factors to consider (Barnett & Adger 2007). Finally, competition between residents and migrants may occur as residents may see migrants as an economic burden and as a threat (Reuveny 2007). Competition in an area is most likely to occur when migrants enter a host community in large numbers as is generally expected following disasters.

    Although scholars note that the increase in frequency of violent storms will produce greater levels of migrants (Beniston 2010; Nordås & Gleditsch 2007; Reuveny 2007; Swain 1996), it is not yet clear that this will invariably increase the likelihood of civil conflict. Taking a sociological perspective on post-disaster behavior, Slettebak (2012) found that natural disasters may encourage solidarity among citizens and reduce conflict in affected areas. It was found that the occurrence of a natural disaster did not increase conflict; however, an increase in population did increase conflict in a region, but this increase was highest when no natural disaster had taken place.

    The literature concerning environmental factors and conflict, including environmental migration and conflict, has consisted of conflicting perspectives on the impacts of climate change on conflict. Furthermore, much of the previous work does not consider the effect of government institutions, government structure or the wealth of the nations, making it difficult to determine the relationship between environmental migration and conflict. The purpose of this research is to determine the relationship between environmentally-induced migration and civil conflict, and the data for this study focuses on disasters and subnational conflict in India and Indonesia. Therefore, a brief history of conflict causes and disaster management for each state is given.

    Case Histories

    Indonesia was originally colonized by the Dutch, and later controlled by Japan before receiving independence in 1949. After independence and subsequent sovereignty, Indonesia was characterized by political factions and communal violence, resulting in thousands of deaths (Schwarz 1997). Following this period of unrest, in 1967 General Suharto gained control and kept a tight authoritarian rule until he was ousted during a financial crisis in 1998 (Kingsbury 2004; Schwarz 1997).

    Suharto ended past ethnic, religious, and group disputes by discouraging conflict and diffusing ethnic or religious tensions through a military presence (Schwarz 1997; Tajima 2008). Military deterrence was also used to prevent rising crimes and localized violence through a campaign where between 5,000 and 10,000 suspected criminals were killed and displayed on the streets (Tajima 2008). Although military deterrence was effective in preventing large outbreaks of conflict, it did not end all conflict. Religion has frequently caused conflict in Indonesia as Muslims make up 90% of the population but often feel marginalized because many political positions are held by Christians. This lead to riots in the 1990s when Muslims called for theocratic rule (Schwarz 1997). Following the collapse of Suharto’s authoritarian rule, there was a surge in localized rebellions as many islands sought independence (Kingsbury 2004). Class conflict and ethnic disputes also became more prevalent because repression was no longer present, allowing groups to speak openly about their political, social, and religious ideologies (Kingsbury 2004).

    Since 1966, the Indonesian government has taken measures to mitigate environmental disasters through the use of the National Disaster Management Agency (Badan Nasional Penanggulangan Bencana). The purpose of this agency is to help families and victims of disasters rebuild and receive aid. As disasters have increased, the agency has increased its services by taking measures to handle environmental refugees and by improving local governments’ abilities to solve disaster related problems.

    India, a heavily populated and ethnically diverse area, gained its independence from British colonial rule in August of 1947. Shortly after, India became a federal republic and the States Reorganization Act was implemented, creating the current federal structure of India. States in India face a variety of challenges, including differing levels of autonomy from the government, widespread and unequal distribution of poverty, subnational terrorism, and booming population growth (Piazza 2009; Urdal 2008).

    The history of ethnic and religious diversity, spatially dispersed poverty, along with the structure of the state system, lie at the root of India’s conflict-related issues. The creation and subsequent division of states along ethnic and linguistic lines has proven to be particularly troublesome since many minority groups were placed in larger states where they were often marginalized, creating ethnic and cultural tensions which have led to conflict and separatist movements (Piazza 2009; Urdal 2008). Similarly, clashes between Hindus and Muslims, and ethno-linguistic division often related to intrastate terrorism and group hostility, have been prevalent in states that are highly diverse (Piazza 2009). Population growth and migration have also been frequent causes of conflict, especially during periods of rapid migration and growth. A salient example is that of the Nellie Massacres, which occurred following large-scale Bengali migration in the 1980s (Piazza 2009).

    Disaster management in India is relatively new. The National Disaster Management Authority (NDMA) was created in 1999 following a major earthquake in Gujarat to mitigate disaster effects. In 2005 the Disaster Management Act was implemented to create a holistic approach to disaster management, including the creation of the Natural Disaster Response Force to deal with disaster cleanup and aid.

    Theory

    The occurrence of an environmental disaster will increase conflict in the region in which it occurred as well as induce intrastate migration among those who reside in the affected area. This environmentally-induced migration will lead to increased presence of low-level civil conflict in adjacent states.

    Environmental disasters, for the purpose of this study, refer to disasters that occur quickly and cannot be prevented by the human population. These disasters include floods, wildfires, hurricanes, tornadoes, earthquakes, tsunamis and cyclones. Environmental disasters are distinct from the phenomena of environmental degradation which occurs over a long period of time and where the environmental alteration is not immediately felt. Examples of environmental degradation are problems such as drought, loss of land due to poor management, and increased water scarcity, all of which gradually occur overtime. Environmentally-induced migration is the migration of people produced by environmental disasters. Environmental disasters are likely to spur large numbers of migrants due to the devastation experienced in the areas in which they occur, often destroying infrastructure and forcing many to leave the area in order to find shelter and economic means to survive. Low-levels of civil conflict refer to smaller, localized conflicts within the state. These conflicts include inter-communal violence, riots, protests, and other forms of general conflict that result in at least one fatality but are not associated with the high casualty rates seen in full-blown civil wars.

    Environmental disasters are likely to increase conflict in the areas in which they occur as well as in neighboring areas. When a disaster strikes, the community is devastated and thus faces a number of obstacles. Many times, resources that are necessary for survival, such as food, water, and shelter, are severely depleted. As resource scarcity increases, so will competition among people in the affected community for said resources. This sudden scarcity and competition will, in turn, result in increased frustrations in the community, culminating in conflict (Brancati 2012). The tensions and frustrations over resource scarcity are also likely to increase tensions between groups. Different ethnic and religious groups are more likely to engage in conflict because the need to secure resources is so crucial. Furthermore, the rapid onset of scarcities in a community is likely to lead to an increase in crimes and looting which in turn will result in clashes with shop owners and local police (Brancati 2012). Along with localized conflict, there is likely to be an increase in dissatisfaction with the government in regards to providing disaster relief, and this is more likely to occur in states that lack a well-developed disaster management plan. As government dissatisfaction increases, so will the amount of protests and riots, which will increase the likelihood of conflict within the region in which the disaster occurred.

    Environmental disasters are also causes of large-scale migration because, unlike environmental degradation, these disasters are immediately devastating to the community and force large numbers of residents to seek communities to support them following the disaster. When facing environmental disasters, residents find themselves with three options: voice concerns and mitigate effects, accept changes in lifestyles, or leave their community (Reuveny 2007). Individuals will choose to migrate if they find the benefits of migrating greater than the cost of staying in the affected area (Weiner 1992). This is most likely to be found in less developed countries where there is a lack of technological innovation to overcome the destruction of environmental disasters and where the environment is the main source of sustenance (Reuveny 2007).

    Environmental degradation will not produce the same migration patterns in communities. Although environmental degradation may cause migration as economic profits decline due to a decrease in environment quality, degradation migration will not occur at the same magnitude as that of environmental disasters because the effects of environmental degradation occur slowly and are not immediately present. Migration induced via degradation will occur in a more gradual fashion due to the lower sense of urgency associated with environmental degradation when contrasted with that of environmental disasters. Thus, outflow from one area to another will occur over a long period of time as residents gradually become affected by degradation. This study assumes those affected by environmental disasters are rational actors who will conduct a cost-benefit analysis, leading to the decision to migrate in order to prosper. Migrants coming from environmental disaster-stuck regions will be forced to move from their home community due to loss of land, resources, and structural surroundings that provided the basis for both economic prosperity and living space (Swain 1996). The benefit of moving to an area not affected by a disaster and being able to thrive will be greater than the cost of remaining in the location in which the disaster occurred, causing residents to make the decision to migrate.

    The expectation of this study is that migrants from environmental disasters will migrate to relatively proximate locations within the state that were unaffected by the disaster, such as neighboring provinces or districts. This is because the disasters leave residents searching for immediate means of sustenance making nearby locations ideal destinations. Furthermore, migration caused by environmental disasters will consist of intrastate migration because this move is easier to make than one which involves crossing international borders, as immigration regulations and international tensions are not an issue in intrastate migration. The financial and cultural burden of moving to a new location within a state is also significantly less than that of interstate migration. Common cultural practices and similarity in lifestyle, along with the reduced cost of the move itself are all reasons for migrants to choose to stay within their state rather than engaging in interstate migration. While varied forms of intrastate migration are not uncommon, the magnitude of migrants produced by environmental disasters will have a different effect on community relations than other forms of migration, where migrants come into an area at a much slower rate. Migration associated with environmental disasters will create large number of migrants within the state; thus in this study the researcher argues that this rapid mass migration will increase the likelihood of low-level civil conflict in the host community (Reuveny 2007).

    Migrants and Conflict

    Large numbers of migrants coming into a host community will place the community under new pressures and cause an alteration in norms. The introduction of new groups, an increase in resource scarcity, and increased job competition are all potential sources of conflict in the host community (Brancati 2012). In subnational regions, there can exist a variety of ethnic and religious groups. Thus, migrants may be of distinct ethnicities, speak different languages, or follow religions that differ from those in the host community. As a result, residents of the host community may protest against the influx of migrants of a different background or could lash out, committing acts of violence (Swain 1996). Similarly, language groups are likely to be sources of conflict because the differences are akin to that of ethnic and religious differences. Different languages represent differences in lifestyle and culture that which may be threatening to the host community. Migrants of different ethnic, language and religious backgrounds not only make the host community feel as though their security is being threatened, but also that the culture the community is built around is under threat as well (Reuveny 2007).

    Ethnic, language, and religious tensions are likely to increase overtime, putting a strain on the groups in the community. As tensions build, the community becomes increasingly unstable. Eventually, a trigger or flash point (a small dispute or disagreement in the community) can serve as the outlet for this building tension and conflict may follow (Weiner 1992). For example, the city of Ahmedabad in the Indian state Gujarat experienced two months of Hindu-Muslim conflict as the result of tensions between the groups that quickly escalated into conflict following an incident when train cars were set aflame (Chatterjee 2009). These type of incidents are more likely to occur in homogeneous communities than in those that are primarily heterogeneous because homogenous communities are more likely to place value on cultural and linguistic characteristics of the community and may feel threatened by the introduction of new groups (Weiner 1992).

    Migrants may also be seen as a financial threat because large numbers of people entering into a host community can undermine economic stability. Migrants become a financial threat to the community because new pressures on the job market are introduced due to the influx of individuals seeking employment. Locations of high unemployment and low job availability will be especially unwelcoming to migrants because those in the community may already be struggling financially (Weiner 1992). Even in areas where high unemployment is not prevalent, large numbers of migrants will still threaten the host community economically as migrants begin competing against those from the host community for jobs and take positions in the economy that normally would have gone to a resident of the host community. Along these lines, if migrants come from an area of a different socioeconomic class, clashes along class lines are likely to take place. By placing communities of different economic activities and lifestyles together, tensions are likely to increase in the host community. Benjaminsen et al. (2012) demonstrate this in their study, observing that the juxtaposition of herders, who allowed their livestock to graze, and farmers, who were dependent on different crops, resulted in clashes between the two groups. Similarly, migrants of the same socioeconomic background could just as easily create conflict between hosts and migrants. This is especially true in less-developed countries, where local communities are more likely to rely on agriculture as a primary source of self-sustenance. This creates an increase in the scarcity of resources as hosts and migrants compete for land in order to produce crops to ensure economic well-being.

    Employment and land scarcities are not the only forms of deprivation a host community may feel upon receiving migrant flows. In addition, water, food, and housing may all also become scarce as migrants move into the host community. The availability of water, food, and housing will decrease in the host community due to the influx of migrants; however, these scarcities could be further exacerbated by disasters that caused the migration flow to the host community. Environmental disasters and the destruction caused by them will likely affect more than one region since neighboring communities often depend on water and food from nearby locations. Therefore, when storms contaminate water, destroy supply lines, and devastate food crops, the neighboring communities will also be impacted by a shortage of resources (Brancati 2012).

    These scarcities have the potential to trigger conflict. Scarcities lead to conflict because they create frustration in the host community, especially when an item that was once available becomes suddenly scarce upon the influx of migrants (Brancati 2012). Scarcities are likely to result in conflict because they encourage competition among groups. Violence and theft between groups may become more pervasive in the community as a result of this competition to attain resources for survival. The presence of a disaster may facilitate violence and theft because the rule of law is likely to be disrupted in the state as the government tries to rebuild and alleviate the area struck by the disaster (Brancati 2012). Along with fighting among themselves, groups may take some of their scarcity-induced frustrations out on the ruling government. As groups find themselves facing an increase in scarcities, they may become frustrated with governmental efforts to alleviate the problem. In turn this frustration is likely to lead to communal protests and riots.

    State Characteristics, Environmental Mitigation, and Conflict

    Characteristics of the state in which environmentally-induced migration occurs will also have an impact on the host-migrant relationship and the likelihood of conflict. Government institutions are crucial in this equation of migration and conflict; thus, the likelihood of conflict is influenced by the state (Barnett & Adger 2007). In cases of environmental degradation, the support of the government (or lack thereof), in addition to the level of dependence on the environment, and the presence of social networks in affected areas, will all have impacts on the probability of conflict occurring due to environmental changes (Reuveny 2007).

    Regions with higher levels of development are better equipped to handle the pressures and problems associated with environmental disasters; thus, developed countries will have a lower likelihood of the onset of conflict following environmental disasters. This is due to a number of reasons, with a primary factor being the level of infrastructure development of a given country. Developed areas are at a technological advantage when it comes to enduring and rebuilding after environmental disasters. Because of technology, the infrastructure of a developed country will be better suited to withstand the stresses of an environmental disaster, allowing for more residents to remain in their home community following disasters. On the other hand, less-developed countries do not have this advantage and an environmental disaster is more likely to destroy large portions of communities, forcing residents to migrate to neighboring areas. A more developed state also has greater economic capacity to provide monetary assistance in the rebuilding of the community and in assisting those most affected by the disaster. Higher levels of development better enable the state to rebuild areas devastated by environmental disasters, allowing for migrants to have the opportunity to move back to their original community at a faster rate. Conversely, lesser-developed states do not have these advantages and are less likely to distribute aid to facilitate rebuilding of communities. Developed countries may also have more clearly defined property rights, while many lesser-developed states often lack well-defined property rights or the institutions to enforce them (Barnett & Adger 2007). As a result, there are likely to be more disputes and small-scale protests in less-developed states as there will be more conflict over claims to land and property.

    Environmental disasters occur quickly and without warning, leaving the communities in which they occur devastated. As a result, those affected make the rational choice to migrate using a cost-benefit analysis. As environmental disasters occur swiftly, they produce large numbers of migrants who relocate to neighboring provinces within the state. The sudden influx of migrants generates a number of problems in the host community that exacerbate conditions conducive to the outbreak of conflict. Those in the host community may feel religiously or ethnically threatened, feel as though the migrants are creating economic burdens, or begin to experience land and water scarcities that did not previously exist in their community. As a result of these pressures, host communities experiencing rapid and large migrant inflows are likely to experience greater levels of subnational civil conflict such as riots and protests.

    Hypotheses

    Hypothesis 1

    H1: The greater occurrence of environmental disasters, the greater the likelihood of conflict in the area in which the disaster occurred.

    Disasters are likely to cause conflict in the area in which they strike because of the stress disasters can place on a community. Loss of necessary resources such as food, water, and shelter along with the devastation of the community will increase tensions in the community. Disasters will lead to an increase in conflict where they occur because of increased competition among people to secure resources and due to potential frustration over a perceived unsatisfactory government response.

    Hypothesis 2

    H2: The greater the occurrence of environmentally-induced migration following environmental disasters, the greater the likelihood of the occurrence of conflict in adjacent subnational units.

    Environmentally-induced migration is likely to increase conflict occurrence because of the rapid population increase in a given unit. Environmental disasters leave areas devastated, forcing those located within the region to search for new communities in which to live. As large numbers of migrants flow into a new unit, the likelihood of conflict will increase. Resources will become increasingly scarce, as economic well-being is likely to decline and cultural tensions are likely to escalate.

    Hypothesis 3

    H3: The less developed a subnational unit, the greater the likelihood of civil conflict following environmentally-induced migration following environmental disasters.

    Less-developed units are more likely to experience the negative effects of environmentally-induced migration. This is because less-developed units lack the technology and financial means to mitigate the effects of environmental disasters. Monetary aid, reconstruction assistance following disasters, and the existence of infrastructure capable of withstanding disasters are all less likely to be present in less-developed regions. As a result, environmental disasters in these less developed units will lead to the greater outflow of migrants for longer periods of time because the government lacks the resources to rebuild quickly. Furthermore, the government will also be unable to alleviate the pressures put on migrants in terms of financial needs, creating more competition in host communities for jobs and resources, making conflict more likely to occur.

    Research Design

    The purpose of this study was to examine the relationship between environmentally-induced migration and intrastate conflict in India from 1956-2002 and in the provinces of Java in Indonesia from 1994-2003. The unit of analysis for this study was district-year, which consists of the subnational units within both India and Indonesia. For the purposes of this study, states in India and provinces in Indonesia were both referred to as units in order to establish a uniform term and avoid misconceptions. These two countries were chosen because they are both highly populated regions of the world that experience a wide variety of environmental disasters, making them good choices to determine the effects of environmental disasters upon conflict occurrence. Furthermore, these two cases are used because there exists extensive research on low-level conflict at the subnational level for both states. Employing two different datasets to conduct this study does come with limitations; however, the goal is to analyze the relationship between environmentally-induced migration and civil conflict at the sub national level. The ability to generalize findings will be limited as the cases cover different time periods and, while both cases measure similar variables, there is variation in the form of measurement used between the two cases. The difference in time periods of the studies limits generalization of findings because these two cases experienced differences in extraneous factors that may have an effect on conflict. For example, the majority of the data from India was collected during the Cold War period, which could have an effect on conflict, while the data for the Indonesia case was collected following democratization and a severe economic depression (Tadjoeddin and Murshed 2007). Furthermore, these case studies were produced for two different purposes. Urdal’s (2008) study was conducted in order to determine causes of local political violence in India due to population and resource pressures. Tadjoeddin and Murshed (2007) were interested in the social and economic determinants of “everyday violence” in Indonesia and how the change in regime and economy, as aforementioned, played a role in conflict occurrence.

    The data for the India case covers the 27 most populous units in India. This data set was originally used for Urdal’s (2008) study on the causes of subnational conflict in India. This data set is ideal for the purposes of this study because the level of conflict under consideration is that of smaller localized events, rather than large-scale violent outbreaks and civil war. Since the purpose of this study is to determine the impact of environmentally-induced migration on intrastate conflict, this dataset was a logical choice. The data on Indonesia covers cities in four of the six units on the island of Java, including Banten, Central Java, West Java, and East Java. The two excluded units were Jakarta, as its metropolitan characteristics make it vastly different from the other Javanese units, and Yogyakarta, due to a lack of available data. This data was originally used for Tadjoeddin and Murshed’s (2007) study on the social and economic effects on routine violence in Javanese provinces. This data varies slightly from the data used for India in terms of how it measures conflict, but still deals with low-level subnational conflict. The study on routine violence in Java is helpful to this study because it consists of low levels of violence, similar to the data on India. This data will be useful in analyzing the effects that environmentally-induced migration has on localized civil conflict. Because this data is not inclusive of all units in Indonesia and India, it is difficult to generalize within these states and the lack of availability of rural data also limits generalizability.

    The dependent variable for this study is the onset of low-level civil conflict. The data source for conflict in India was used by Urdal (2008) for his analysis of subnational conflict. From this study, the conflict data from the political violence events variable will be used. This dataset is a count measure of subnational conflict in India determined by presence of inter-communal violence, organized violence, riots, and political assassinations. The measure of conflict for the units of Java in the Indonesia case comes from the Tadjoeddin and Murshed (2007) dataset. The conflict in this study includes cases of “routine violence,” which consist of state-community violence, economic violence (such as land or natural resource conflicts), and popular justice (group brawls and political party violence). Routine violence is measured in this dataset by the number of violent incidents which occurred in a given year.

    Independent variables for this study include environmental disasters, migration, and level of development. Environmental disasters will be operationalized via the inclusion of disasters that strike quickly. These disasters include floods, storms (cyclones and hurricanes), earthquakes, tsunamis, and landslides. Long-term environmental issues such as land or water degradation, drought, heat/cold waves, and deforestation are not included in this study because these disasters will produce migrants over a longer period of time as opposed to disasters which produce more migrants in a short period of time. A binary variable will be used to record the occurrence of an environmental disaster where ‘1’ represents the presence of an environmental disaster and ‘0’ represents no environmental disaster. The data on environmental disasters was gathered from Emergency Events Database (EM-DAT) complied by the Centre for Research on the Epidemiology of Disasters (CRED). The EM-DAT database covers the occurrence, effects, and locations of disasters from the 1900s to the present and has been collected from a vast number of sources including research institutes, UN institutions, media outlets, and government agencies.

    There exists a lack of data in terms of intrastate migration patterns for both India and Indonesia. To account for migration, a proxy variable will be used to indicate units adjacent to areas in which an environmental disaster occurred. This variable will be a binary variable in which ‘0’ indicates the state is not adjacent to a location in which an environmental disaster took place, and ‘1’ indicates the unit is adjacent to an environmental disaster-struck unit. Units adjacent to those in which an environmental disaster took place are inferred to be probable migration locations because of their close proximity. Migrants are not expected to migrate long distances following a disaster. This is because the cost of migrating to a nearby state both financially and culturally is typically lower than migrating longer distances within the state or migrating internationally. These factors collectively provide a plausible justification for the employment of this measure as a proxy for the presence of environmentally-induced migrants.

    The level of development in both the Indonesia and India datasets is operationalized via the level of education for a subnational unit; however, the method of determining educational levels varies slightly between the two studies. In the case of Indonesia, education level is measured by the average mean years of schooling for the population above the age of 15. For India, the educational level has been measured by the number of literates per one-thousand above the age of seven.

    The inclusion of extensive control variables is not absolutely necessary for this study because natural disasters constitute a natural experiment as they are random and not brought on by outside forces caused by human activity. In order to firmly grasp the circumstances under which natural disasters increase the likelihood of conflict, this study will control for a variety of exogenous factors that potentially lead to conflict. The population of the unit in both India and Indonesia will be taken into account. Population density is used in India to account for the forces exerted by high concentrations of population. In Indonesia, the log of the population is used to control for population in each unit. Previous conflict in both India and Indonesia will be controlled through the use of a binary variable in which a ‘1’ denotes a unit that experienced five or more violent conflicts in the previous year and a ‘0’ indicates there were less than five instances of conflict (Urdal 2008). Religious heterogeneity is also controlled in India as it is a potential cause of conflict (Urdal 2008), and employs a measure of the percentage of heterogeneity per unit. While the Indonesia data does not have measures for religious heterogeneity, a control for human development in these units is used. The measure for this variable is the overall level of development of each unit consisting of three factors: health, education, and income per capita (Tadjoeddin & Murshed 2007).

    The model used for both the Indian and Indonesian cases is a negative binomial regression. This model was chosen due to the fact that the dependent variable is constructed as accounts of violent events in both datasets. The negative binomial model is appropriate for this study because the dependent variable is a count variable that is widely dispersed. The negative binomial model allows for the interpretation of the sign and significance of the coefficient on the independent variables; however, the strength of the coefficient cannot be ascertained due to the non-linear nature of the analysis. The strengths of the coefficients on the independent variables for the India case are ascertained via the use of Incident Rate Ratios. Ratio values above 1 indicate the greater the likelihood of the conflict, while values less than 1 indicate a decreased likelihood of conflict occurring.

    Analysis

    India

    In the first analysis of the effects of environmental disasters on conflict in India, the negative binomial model was used to test hypotheses 1 and 2 without any control variables. This was done to isolate the effects of the primary independent variables, considering that the nature of environmental disasters constitutes a natural experiment. Hypothesis 1 posits that areas where environmental disasters strike are likely to experience greater conflict. In this model, the coefficient for areas in which disasters occurred is both positive and significant at the p≤.000 level. Refer to Table 1. This finding is consistent with expectations of the study since disasters increase scarcities and tensions in communities. Hypothesis 2 posits that the rise in environmentally-induced migration due to environmental disasters will increase conflict in adjacent subnational units. This model shows that the coefficient for units adjacent to areas in which a disaster occurred were positive and significant, although the significance level, at p≤.10 level, did not achieve the criterion level of p≤.05. This is also consistent with expectations that a rapid rise in population increases conflict in surrounding areas because those in the host community begin to feel threatened and experience sudden deprivation of resources.

    A second negative binomial model for India was run including the control variables—literacy rate, population density, religious heterogeneity, and prior conflict. Refer to Table 2. When these control variables were added, there was minimal effect on the results of hypothesis 1. For areas in which disasters struck, the coefficient remained positive and significant at the p≤.000 level. The inclusion of control variables did alter the findings for the test on hypothesis 2 to a small degree. The inclusion of control variables increased the significance of units adjacent to disaster-struck areas, and the coefficient remained positive and significant at the p≤.01 level. In using literacy rate as a proxy measure of development, hypothesis 3 tested the notion that the less developed a unit, the more likely conflict following environmentally-induced migration was to occur. The coefficient for literacy was negative and was significant at the p≤.000 level. This met expectations, as less developed areas are likely to face greater struggles following disaster in terms of rebuilding and scarcity of aid. The other control variables in this model illustrate alternative reasons a unit experiences increased conflict following disaster. The coefficient for population density was both positive and significant at p≤.01 when tested, suggesting that the more populous a unit is, the more likely it will be that disasters increase conflict. Furthermore, it is clear that the level of religious heterogeneity is a major factor when it comes to conflict. The religious heterogeneity coefficient was both positive and significant at the p≤.000 level. This finding was consistent with the expectation of this study that environmental disasters will increase tensions among groups and that environmentally-induced migration will cause juxtaposition of these groups, thus increasing conflict. Interestingly, conflict did not increase in units where there were five or more violent incidents in the previous year. The coefficient for prior conflict was positive but insignificant in India.

    To determine the impacts of natural disasters, adjacent units, literacy rate, and other control factors on conflict in India, an incident rate ratio (IRR) analysis was conducted. Refer to Table 3. In this model the IRR tests the impact of the independent variables upon the likelihood of conflict by establishing ratios relative to 1. In this model, it was seen that when all other variables are held at constant, a one-unit increase in disaster increased conflict likelihood by 90%. Adjacent units also increased percentage of conflict likelihood. A one-unit increase in adjacent units means conflict becomes 39% more likely. Literacy rates when all other variables were held constant decreased conflict likelihood by .2% and population density increased conflict likelihood by .2%. Of all these factors, religious heterogeneity had the largest increase in conflict likelihood. A one unit increase in religious heterogeneity, with all other variables are held at constant, increased conflict likelihood by 289%.

    Indonesia

    The tests and models for Indonesia proceed in the same manner as those for India. As in the models for India the first model tested hypothesis 1 and 2 without the inclusion of any control factors. Refer to Table 4. In testing hypothesis 1, the coefficient for units in which a disaster occurred was positive and insignificant in its impact upon conflict. There were similar results for the test of hypothesis 2; the coefficient for units adjacent to disasters was negative and also insignificant. These results were inconsistent with expectations and indicated a difference in post-disaster management between India and Indonesia.

    The next negative binomial model used for Indonesia included the following control factors: mean years of schooling, log of population, human development index, and prior conflict. Refer to Table 5. In this model, the inclusion of control variables had minimal effect on the coefficients of both units that experienced disasters and the units adjacent to these areas. The coefficient for disasters remained insignificant and was negative, while the coefficient for units adjacent to disasters remained insignificant and was positive. Thus, neither disasters nor the areas near these disasters were significantly related to increased conflict. When testing hypothesis 3 in this scenario, the level of development was measured by mean years of schooling and it did have an effect on conflict. The coefficient for mean years of schooling was positive and significant at the p≤.000 level. These results are similar to results found in the case for India, indicating that the more developed a unit is, the better equipped the unit is to handle conflict. Other control variables in the model also illustrated significant sources of conflict in Indonesia. The coefficient for the log of population was positive and significant at the p≤.000 level. This finding was again similar to that of India and showed that population was an important causal factor when considering conflict occurrences. Similar results were found for the human development index which took healthcare, education, and income per capita into account. The coefficient for human development index was negative and significant at the p≤.000 level, indicating once more that the less developed a region was, the more likely it would be involved in conflict. Unlike the case for India, prior conflict was of importance in Indonesia. The coefficient for prior conflict population was positive and significant at the p≤.000 level.

    Logit Models

    To further test the robustness of the results of the negative binomial model, a logistic regression model was conducted after converting the dependent variable into a binary variable for the presence of conflict or no conflict in each unit following disaster. This measure is distinct from the measure of conflict used for a count measurement because the dependent variable in logit models was a strictly a binary measurement. The logit model allows for a different operationalization of the dependent variable and serves as a way to check for robustness in this study. The results in the logit models resemble the findings using the negative binomial model, indicating the results in this study are robust.

    In the case of India, the first logit model consisted of the dependent variable measuring conflict onset and the two main independent variables measuring disaster and adjacent unit. Refer to Table 6. The results from the test vary only slightly from the original negative binomial model. The coefficient for units in which disaster occurred was both positive and significant at p≤.000. Similarly, adjacent units were also significant in relation to onset of conflict. The coefficient for adjacent units was both positive and significant at the p≤.05 level, which was slightly more significant that what was found in the negative binomial model. For the next hypothesis test, the control variables were reintroduced and a logit analysis was conducted. Refer to Table 7. As in the first logit model, the coefficients and levels of significant were similar to the findings of the negative binomial model for India that incorporated all of the control variables. Disaster and adjacent units in this case were both positive and significant at the p≤.000 level, making adjacent units slightly more significant in this case. Literacy also continued to be negative and significant at the p≤.05 level in this model. Population density was salient in this model as it was in the negative binomial model. The coefficient for population density remains positive and significant at the p≤.000 level. Interestingly, religious heterogeneity lost some of its significance in the logit model. The coefficient for religious heterogeneity was positive but only marginally significant at the p≤.10 level. As in the earlier models, prior conflict continued to remain positive but insignificant in this model.

    As in India, a logit model for Indonesia was used to model the dependent variable and the two main independent variables. Refer to Table 8. The results of this model were consistent with earlier results. The coefficient for disaster struck regions was positive and insignificant and the coefficient for units adjacent to disasters in negative and also insignificant. A second logit model for Indonesia was then considered, incorporating the independent variables as in the negative binomial models. Refer to Table 9.

    In this second model, the coefficients for the two main independent variables, disaster struck regions and adjacent states, were both positive and insignificant. Mean years of schooling in this model indicated results similar to those found in the negative binomial model. The coefficient for mean years of schooling was positive and significant at p≤.000. The log of population also remained a salient factor when considering conflict in Indonesia. The coefficient for log of population was both positive and significant at the p≤.000 level. The coefficient for the human development index remained negative; however the significance declined slightly and was significant at the p≤.05 level. Prior conflict also became slightly less significant in this model. The coefficient remained positive but the significance was now at the p≤.01 level.

    After examining models for both India and Indonesia, it is clear there are conflicting results. While disasters and adjacent units in India were significantly related to conflict, this was not consistent with the findings from the models for Indonesia. One possible reason for this discrepancy could be use of disaster management in these states. While Indonesia has planned for disasters since sovereignty, India did not fully consider disaster management until 1999, indicating disaster management policy was an important consideration when evaluating environmental disasters and conflict. Because of this, the India dataset was split into two time periods, pre-disaster management plans (1956-1999) and post-disaster management plans (1999-2002).

    In the pre-disaster management implementation model, results reflect what has been seen in earlier models. Refer to Table 10. Both disasters and units adjacent to disasters are significant in the relationship to conflict. However, when a model for post-disaster management plan implementation was run, the results differ. Refer to Table 11.

    After implementing disaster management plans in India, the coefficients for the independent variables, disasters and units adjacent to them, were both positive, but no longer significant causes of conflict. This indicates disaster management did have an effect in controlling conflict in post-disaster scenarios. It should be noted however, the sample for the post-disaster management plan implementation time period was small and this study could further benefit by expanding the analysis to the present day to better understand the effects of India’s disaster management plans on conflict.

    Before further discussing the findings in the analyses conducted in this study, it is important to reiterate that this study lacks migration data for both India and Indonesia disasters and instead employed the adjacent units as a proxy measurement for probable locations where environmental migrants may find refuge. To an extent, this limits the ability to generalize the findings that environmentally-induced migration increased conflict; however, these findings do provide insight into the relationship between disasters and conflict.

    Discussion of Results and Implications

    The analysis of India and Indonesia revealed two different stories. In the case of India, both disasters and units near disasters had positive and significant impacts upon conflict. These findings were consistent with expectations of this study, but this was not the case for Indonesia where disasters were insignificant in the units they occurred in as well as adjacent units. Given the case histories of these two states, it is possible to speculate on the reasons disasters in India increased subnational conflict but did not in Indonesia. One possible reason is that Indonesian history is characterized by a militarily enforced peace and by fear tactics to reduce crime. However, it is far more likely the difference in conflict onset in the two states is due to the structure of disaster management. Since sovereignty in 1966, Indonesia has had a disaster management agency as the government was well aware of the large numbers of disasters experienced in Indonesia. As storms have increased in severity over time, the government has developed more strategies to deal with refugees and to rebuild. India, on the other hand, has not had comparable planning and infrastructure to handle disasters. India lacked a fully formed disaster management system until 2005 when the Disaster Management Act made the visions of the 1999 plan a reality.

    These results seem to indicate that disaster management and policy do have an effect in reducing conflict caused by environmental disasters. Because Indonesia has had plans for sending aid to disaster-struck regions and helping refugees find shelter since sovereignty the state is much more equipped to handle post-disaster situations in comparison to states such as India. Results also show areas that are less developed and have greater population pressures experienced greater instances of conflict in the aftermath of disasters. This indicates lesser developed units within states have are particularly prone to post-disaster conflict this could be due to the lack of infrastructure that can withstand disasters, forcing a greater amount of residents to migrate. Another possible reason is the lesser-developed units in state lack the local government strength to mitigate effects of disasters by providing shelter, food, and rebuilding aide.

    Conclusion

    The purpose of this study was to develop an understanding of the relationship between environmental disasters and subnational conflict. As climate change progresses, it is expected that there will be a surge in both the severity and frequency of devastating storms. While many scholars are unsure of the effects that changes in weather patterns and climate may have on the human population, this study has argued that in the aftermath of environmental disasters, conflict will increase in both the area in which the disaster occurred and in the host community receiving migrants from disasters. This study has illustrated that, in certain circumstances, changes in weather patterns can increase the likelihood of conflict and should be further studied to gain a deeper understanding of this relationship.

    This research could be improved by including more states and by the incorporation of intrastate migration data. Further research in this area would benefit from including subnational migration data, especially in the aftermath of disaster situations. This study relied on the proxy variable of adjacent units to gauge probable migration locations; however, proxies are not precise forms of measurement and, therefore, do not illustrate the entire causal relationship between environmentally-induced migration and subnational conflict. Furthermore, research regarding environmental disasters and subnational conflict would be enhanced by including more states in order to have a firmer understanding of why some states experience conflict and others do not following disasters. An obstacle to obtaining both subnational migration and subnational conflict information is the lack of data. Many of the countries expected to be impacted most by the change in climate and weather patterns do not currently have subnational data on migration and conflict, making it difficult to carry out this research.

    The results of this study indicate certain conditions in which environmental disasters will increase conflict. States that are unable to mitigate the effects of natural disasters and provide support for those living in the affected area are more likely to see an increase in conflict following a disaster than states with well-grounded disaster management systems. This increase in conflict is likely to be seen in both areas in which disasters occur and in communities hosting migrants from disasters. Furthermore, characteristics of areas in which disasters strike make these regions more vulnerable to conflict following disasters. Population density, level of development, and religious heterogeneity all contribute significantly to conflict in the aftermath of disasters. The findings of this study illustrate that more research on environmental disasters and the locations in which they strike would be useful for policy makers to better mitigate the effects of environmental disasters. This research and implementation of disaster management strategies will become increasingly crucial as the occurrence of environmental disasters increases and the resulting displacement of people becomes an ever-increasing threat to peace.

    References

    • Adano, Wario R., Ton Dietz, Karen Witsenburg, and Fred Zaal. 2012. “Climate Change, Violent Conflict and Local Institutions in Kenya’s Drylands.” Journal of Peace Research 49(1): 65-80. http://jpr.sagepub.com/cgi/doi/10.1177/0022343311427344.
    • Badan Nasional Penanggulangan Bencana. “Sejarah.” http://www.bnpb.go.id/website/asp/content.asp?id=2 (July 24, 2012).
    • Barnett, Jon, and W. Neil Adger. 2007. “Climate Change, Human Security and Violent Conflict.” Political Geography 26(6): 639-655. http://linkinghub.elsevier.com/retrieve/pii/S096262980700039X
    • Beniston, Martin. 2010. “Climate Change and Its Impacts: Growing Stress Factors for Human Societies.” International Review of the Red Cross 92(879): 557-568. http://www.journals.cambridge.org/abstract_S1816383110000342
    • Benjaminsen, Tor A., Koffi Alinon, Halvard Buhaug, and Jill T. Buseth. 2012. “Does Climate Change Drive Land-Use Conflicts in the Sahel?” Journal of Peace Research 49(1): 97-111. http://jpr.sagepub.com/cgi/doi/10.1177/0022343311427343
    • Brancati, Dawn. 2012. “Political Aftershocks: The Impact of Earthquakes on Intrastate Conflict.” The Journal of Conflict Resolution 51(5): 715-743.
    • Butler, Christopher K., and Scott Gates. 2012. “African Range Wars: Climate, Conflict, and Property Rights.” Journal of Peace Research 49(1): 23-34. http://jpr.sagepub.com/cgi/doi/10.1177/0022343311426166
    • Chatterjee, Ipsita. 2009. “Social Conflict and the Neoliberal City: A Case of Hindu – Muslim Violence in India.” Transactions of the Institute of British Geographers, 24(2): 143-160. DOI: 10.1111/j.1475-5661.2009.00341.x
    • EM-DAT: The OFDA/CRED International Disaster Database – www.emdat.be, Université Catholique de Louvain, Brussels (Belgium).
    • Gleditsch, Nils P. 2012. “Whither the Weather? Climate Change and Conflict.” Journal of Peace Research 49(1): 3-9. http://jpr.sagepub.com/cgi/doi/10.1177/0022343311431288
    • Homer-Dixon, Thomas F. 1994. “Environmental Scarcities and Violent Conflict: Evidence from Cases.” International Security 19(1): 5-40.
    • Intergovernmental Panel on Climate Change. 2007. “Climate Change 2007: Synthesis report. Contribution of Working Groups I, II, and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change.” Cambridge: Cambridge University Press.
    • Kingsbury, Damien. 2004. “Indonesia: Causes of Conflict and Resolution.” Teaching and Learning-VCE 12(2).
    • National Disaster Management Authority Government of India. “Evolution of The NDMA.” http://ndma.gov.in/ndma/evolution.html (Accessed July 24, 2012).
    • Meier, Patrick, Doug Bond, and Joe Bond. 2007. “Environmental Influences on Pastoral Conflict in the Horn of Africa.” Political Geography 26(6): 716-735. http://linkinghub.elsevier.com/retrieve/pii/S0962629807000820
    • Nordås, Ragnhild, and Nils Petter Gleditsch. 2007. “Climate Change and Conflict.” Political Geography 26(6): 627-638. http://linkinghub.elsevier.com/retrieve/pii/S0962629807000856
    • Piazza, James A. 2009. “Economic Development, Poorly Managed Political Conflict and Terrorism in India.” Studies in Conflict & Terrorism 32(5): 406-419. http://www.tandfonline.com/doi/abs/10.1080/10576100902831552
    • Raleigh, Clionadh, and Henrik Urdal. 2007. “Climate Change, Environmental Degradation and Armed Conflict.” Political Geography 26(6): 674-694. http://linkinghub.elsevier.com/retrieve/pii/S096262980700087X
    • Reuveny, Rafael. 2007. “Climate Change-Induced Migration and Violent Conflict.” Political Geography 26(6): 656-673. http://linkinghub.elsevier.com/retrieve/pii/S0962629807000601
    • Salehyan, Idean. 2008. “From Climate Change to Conflict? No Consensus Yet.” Journal of Peace Research 45(3): 315-326.
    • Schwarz, Adam. 1997. “Indonesia After Suharto.” Foreign Affairs 76(4):119-134.
    • Slettebak, Rune T. 2012. “Don’t Blame the Weather! Climate-Related Natural Disasters and Civil Conflict.” Journal of Peace Research 49(1): 163-176. http://jpr.sagepub.com/cgi/doi/10.1177/0022343311425693
    • Swain, Ashok 1996. “Environmental Migration and Conflict Dynamics: Focus on Developing Regions.” Third World Quarterly 17(5): 959-73. http://www.ncbi.nlm.nih.gov/pubmed/12321041.
    • Tadjoeddin, Mohammed Z., and Syed M. Murshed. 2007. “Socio-Economic Determinants of Everyday Violence in Indonesia: An Empirical Investigation of Javanese Districts, 1994 2003.” Journal of Peace Research 44(6): 689-709. http://jpr.sagepub.com/cgi/doi/10.1177/0022343307082063
    • Tajima, Yuhki. 2008. “Explaining Ethnic Violence in Indonesia: Demilitarizing Domestic Security.” Journal of East Asian Studies 8: 451-472.
    • Urdal, Henrik. 2008. “Population, Resources, and Political Violence A Subnational Study of India, 1956-2002.” The Journal of Conflict Resolution 52(4): 590-617.
    • Weiner, Myron. 1992. “Security, Stability, International Migration.” International Security17(3): 91-126.

    Table 1: India: Simple Negative Binomial Model

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    0.6426
    (.1614)

    0.000***

    Adjacent Unit

    .2170
    (.1274)

    0.088*

    N= 1165 ***=P<.01 **=P<.05 *=P<.10

    Table 2: India: Negative Binomial Model with Controls

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    0.6452
    (.1616)

    0.000***

    Adjacent Unit

    .394
    (.1288)

    0.002**

    Literacy

    -0.0018
    (.0005)

    0.000***

    Population Density

    .0002
    (.0001)

    0.007**

    Religious Heterogeneity

    1.36
    (.3418)

    0.00***

    Prior Conflict

    .1182
    (.2125)

    0.578

    n= 1165 ***=P<.001 **=P<.01 *=P<.05

    Table 3: India: Incident Rate Ratio Model

    Variable

    IRR

    Coefficient
    (Standard Error)

    Disaster

    1.906

    0.6452
    (.1616)

    Adjacent
    Unit

    1.483

    .394
    (.1288)

    Literacy

    .998

    -0.0018
    (.0005)

    Population
    Density

    1.000

    .0002
    (.0001)

    Religious
    Heterogeneity

    3.895

    1.36
    (.3418)

    Prior
    Conflict

    1.125

    .1182
    (.2125)

    n=1165

    Table 4: Indonesia: Simple Negative Binomial Model

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    0.1461
    (.3560)

    0.682

    Adjacent State

    -0.0502
    (.2629)

    0.849

    n=980

    Table 5: Indonesia: Negative Binomial Model with Controls

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    -.0867
    (.5796)

    0.881

    Adjacent State

    .2125
    (.4542)

    0.64

    Mean Years Schooling

    0.3639
    (.086)

    0.000***

    Log of Population

    .6560
    (.1107)

    0.00***

    Human Development

    -0.1107
    (.0299)

    0.00***

    Prior Conflict

    .9273
    (.1759)

    0.00***

    n=294 ***=P<.001 **=P<.01 *=P<.05

    Table 6: India: Simple Logit Model

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    .6952
    (.1689)

    0.000***

    Adjacent Unit

    .3877
    (.1281)

    0.002**

    n= 1165 ***=P<.001 **=P<.01 *=P<.05

    Table 7: India: Logit Model with Controls

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    0.6708
    (.1689)

    0.000***

    Adjacent Unit

    .5158
    (.137)

    0.000***

    Literacy

    -0.0015
    (.0005)

    0.002**

    Population Density

    .0004
    (.0001)

    0.000**

    Religious Heterogeneity

    .5867
    (.3145)

    0.071*

    Prior Conflict

    .2842
    (.2181)

    0.193

    n= 1165 ***=P<.01 **=P<.05 *=P<.1

    Table 8: Indonesia: Simple Logit Model

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    0.8504
    (.5757)

    0.140

    Adjacent State

    -0.1806
    (.3566)

    0.613

    n=980

    Table 9: Indonesia: Logit Model with Controls

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    .7119
    (1.254)

    0.570

    Adjacent State

    1.108
    (1.161)

    0.340

    Mean Years Schooling

    0.7397
    (.1722)

    0.000***

    Log of Population

    1.0874
    (.2286)

    0.000***

    Human Development

    -0.1261
    (.0299)

    0.012*

    Prior Conflict

    1.6452
    (.6379)

    0.010**

    n=294 ***=P<.001 **=P<.01 *=P<.05

    Table 10: India: Negative Binomial Model 1956-1999

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    .6490
    (.1646)

    .000

    Adjacent Unit

    .3951
    (.1310)

    .003

    Literacy 

    -.0018
    (.0005)

    .000

    Population Density

    .0002
    (.0001)

    .010

    Religious Heterogeneity

    1.410
    (.3483)

    .000

    Prior Conflict

    .1194
    (.2139)

    .56

    n=1134 ***=P<.01 **=P<.05 *=P<.1

    Table 11: India: Negative Binomial 1999-2002

    Variable

    Coefficient
    (Standard Error)

    P>|z|

    Disaster

    .6739
    (.7406)

    .363

    Adjacent Unit

    .0677
    (.7072)

    .924

    Literacy 

    -.0061
    (.003)

    .040**

    Population Density

    .0003
    (.0002)

    .177

    Religious Heterogeneity

    -.9412
    (.0002)

    .535

    Prior Conflict    

    n=31 ***=P<.01 **=P<.05 *=P<.1