Search the Entire SiteSearch
Genocide and related crimes against humanity are devastating in their scale and scope; in the enduring scars for survivors and their families and the long-term trauma they cause in societies; and in the economic, political, and social costs and consequences, often extending far beyond the territory in which they were committed.
Working to prevent future genocides requires an understanding of how these events occur, including considerations about warning signs and human behaviors that make genocide and mass atrocities possible.
We know from studying the Holocaust and other genocides that such events are never spontaneous. They are always preceded by a range of early warning signs. If warning signs are detected and their causes addressed, it may be possible to prevent catastrophic loss of life.
This assessment identifies the risk—the possibility—that a mass killing may take place. On average, one or two countries experience a new episode of mass killing each year. But relative infrequency does not make the brutality less devastating for victims: a mass killing, by our definition, is 1,000 or more civilians deliberately killed by armed forces (whether government or non-state), over a period of a year or less, because of their membership in a particular group. Virtually all cases of genocide include mass killings that meet this definition.
The United States Holocaust Memorial Museum’s founding charter, written by Holocaust survivor Elie Wiesel, mandates that our institution strive to make preventive action a routine response when warning signs appear. Wiesel wrote, “Only a conscious, concerted attempt to learn from past errors can prevent recurrence to any racial, religious, ethnic or national group. A memorial unresponsive to the future would also violate the memory of the past.”
The Museum’s Simon-Skjodt Center for the Prevention of Genocide was established to fulfill that vision by transmitting the lessons and legacy of the Holocaust, and “to alert the national conscience, influence policy makers, and stimulate worldwide action to confront and prevent genocide.” The Simon-Skjodt Center’s Early Warning Project works to fulfill this aspect of the Museum’s mandate by using innovative research to identify early warning signs. In doing so, we seek to do for today’s potential victims what was not done for the Jews of Europe. One of the Simon-Skjodt Center’s goals is to ensure that the United States government, other governments, and multilateral organizations have institutionalized structures, tools, and policies to effectively prevent and respond to genocide and other mass atrocities.
The more governments and international organizations develop their own early warning tools and processes, the better our Early Warning Project can help serve as a catalyst for preventive action.
In many places, such violence is ongoing—in countries such as Burma, Syria, and South Sudan. These cases are well-known. But this risk assessment’s primary focus—and the gap we seek to fill—is to draw attention to countries at risk of a new outbreak of mass killing. We use this model as one input for selecting countries for more in-depth research and policy engagement. The Simon-Skjodt Center focuses on situations where there is a risk of, or ongoing, large-scale group-targeted identity-based mass atrocities and where we believe we can make the most impact based on a combination of factors. These factors include the ability for Simon-Skjodt Center staff to conduct rigorous field work in the area (or a pre-existing level of staff expertise in the area), opportunities for effective engagement with the community at risk, and the need to draw attention to cases where policy, media, and public attention on the case are lower than merited by the level of risk.
Preventing genocide is of course difficult. In deciding how to respond, policy makers face an array of constraints and competing concerns. The choice to prevent one potential tragedy often takes a back seat when policy makers are confronted by multiple ongoing conflicts. But we know from the Holocaust what can happen when early warning signs go unheeded. We aim for this risk assessment to serve as a tool and a resource for policy makers and others interested in prevention. We hope this helps them better establish priorities and undertake the discussion and deeper analysis that can help reveal where preventive action can make the greatest impact in saving lives.
Simon-Skjodt Center for the Prevention of Genocide
The Early Warning Project’s Statistical Risk Assessment uses publicly available data and statistical modeling to produce a list of countries ranked by their estimated risk of experiencing a new episode, or onset, of mass killing.
Policy makers face the challenge of simultaneously responding to ongoing mass atrocities, such as those in Burma, South Sudan, and Syria, and trying to prevent entirely new mass atrocity situations. A critical first step toward prevention is accurate and reliable assessment of countries at risk for future violence. Earlier identification of risk broadens the scope of possible preventive actions. This report aims to help identify countries meriting preventive actions.
In essence, our statistical model identifies patterns in historical data to answer the question, which countries today look most similar to countries that experienced mass killings in the past, in the year or two before those mass killings began? The historical data include basic country country characteristics, as well as data on governance, war and conflict, human rights and civil liberties, and socioeconomic factors.
This report highlights findings from our Statistical Risk Assessment for 2020–21, focusing on:
We recognize that this assessment is just one tool. It is meant to be a starting point for discussion and further research, not a definitive conclusion. We aim to help governments, international organizations, and nongovernmental organizations determine where to devote resources for additional analysis, policy attention, and, ultimately, preventive action. We hope that this report and our Early Warning Project as a whole inspire governments and international organizations to invest in their own early warning capabilities.
Before discussing the results, we underscore five points about interpreting this Statistical Risk Assessment:
First, as a statistical matter, mass killings are rare. On average, just over one percent of countries see a new mass killing in any given year—that means one or two countries. Our risk model predicts a similar number of new episodes of mass killing, so the average two-year risk estimate produced by our model is just under two percent. Just three out of 162 countries have a two-year risk estimate greater than ten percent, and the highest-risk country, Pakistan, is estimated to have about a one in six chance of experiencing a new mass killing in 2020 or 2021.
Second, our model is designed to assess the risk of a new mass killing, not of the continuation or escalation of ongoing episodes. Much of the Simon-Skjodt Center’s work spotlights ongoing atrocities and urges life-saving responses. We focus here on the risk of new mass killing to help fill an analytic gap that is critical to prevention. This feature is especially important to bear in mind when interpreting results for countries that are currently experiencing mass killings, such as Burma/Myanmar and Syria (see Figure 1 and our website for a full list of these countries). For these countries, our assessment should be understood as an estimate of the risk that a new mass killing event would be launched by a different perpetrator or targeting a different civilian group in 2020 or 2021. (Our model estimates that having a mass killing currently in progress is associated with lower risk of another one beginning.) Regardless of their ranking in this assessment, cases of ongoing atrocities demand urgent action.
Third, for practical reasons, we only forecast mass killings within countries (i.e., in which the perpetrator group and the targeted civilian group reside in the same country). This risk assessment does not forecast civilian fatalities from interstate conflict. Situations in which large numbers of civilians are killed deliberately by an armed group from another country are not captured in our historical data or current forecasts. This decision does not involve a value judgment about the moral or practical significance of such atrocities, only a pragmatic judgment about what we are able to forecast reliably.
Fourth, readers should keep in mind that our model is not causal: the variables identified as predicting higher or lower risk of mass killings in a country are not necessarily the factors that drive or trigger atrocities. For example, a large population does not directly cause mass atrocities; however, countries with large populations have been more likely to experience mass killing episodes in the past, so this factor helps us identify countries at greater risk going forward. We make no effort to explain these kinds of relationships in the data; we only use them for their predictive value. An important consequence of the non-causal nature of these forecasts is that actions aimed at addressing risk factors identified in the model are not necessarily effective ways of mitigating the risk of mass atrocities; this assessment does not seek to evaluate atrocity prevention policy prescriptions. For example, although our model finds that countries coded as having severely limited freedom of movement for men are at greater risk of experiencing mass killings than are other countries, this does not imply that action to improve freedom of movement for men would help prevent mass killings.
Fifth, this assessment is based on available data reflecting conditions as of the end of 2019. Events that occurred in 2020 are not reflected in country risk estimates. Our assessment relies on publicly available data that is reliably measured for nearly all countries in the world, annually updated, and historically available going back many years. Because mass killing is rare, global data spanning decades are necessary to identify patterns. This means that some risk factors that might be useful predictors, but for which data meeting the above criteria are not available, are not included in the model (e.g., data on dangerous speech may be a useful predictor, but is not currently included due to a lack of data availability). Additionally, in situations where governments deliberately restrict access to international observers, such as in Burma’s Rakhine state or China’s Xinjiang region, existing data might not fully reflect conditions on the ground. In addition, updated data for 2019 were not available on two risk factors that we have used to produce previous assessments: regime type and regime duration, both from the Center for Systemic Peace. We found that dropping these two variables did not affect the overall accuracy of the model by most measures, but it may account for shifts in specific countries’ risk-ranking.
Our model generates a single risk estimate for each country, representing the estimated risk for a new state-led or non-state-led mass killing. Figure 2 displays the estimated risk in 2020 or 2021 for the 30 highest-ranked countries. For every country in the top 30, we recommend that policy makers consider whether they are devoting sufficient attention to addressing the risks of mass atrocities occurring within that country. Strategies and tools to address atrocity risks should, of course, be tailored to each country’s context.
Further qualitative analysis is needed to understand the specific drivers of risk in a given situation, the mass atrocity scenarios that could be deemed plausible, and the resiliencies that could potentially be bolstered to help prevent future atrocities. This kind of deeper qualitative assessment is exemplified in Early Warning Project reports on Côte d’Ivoire (2019), Mali (2018), Bangladesh (2017), and Zimbabwe (2016). Concerned governments and international organizations should consider conducting their own assessments of countries at risk, which should suggest where adjusting plans, budgets, programs, and diplomatic strategies might help prevent mass killings in high-risk countries. Because these qualitative assessments are resource intensive, policy makers should prioritize that type of analysis on countries whose risk estimate is relatively high according to this Statistical Risk Assessment, and where opportunities for prevention exist.
Note: * indicates ongoing state-led mass killings; ° indicates ongoing non-state-led mass killings. Some countries have multiple ongoing episodes of one or both type (e.g., Burma/Myanmar has two ongoing state-led mass killings; Nigeria has an ongoing state-led and an ongoing non-state-led mass killing). Risk-based ranking is in parenthesis. The probabilities displayed here are associated with the onset of an additional mass killing episode. See the full list of ongoing mass killings on our website.
~ For more information on crimes against humanity in China, see later section on "Unexpected Results."
In the paragraphs below, we discuss each country’s risk according to our statistical model, and note any instances of ongoing violent conflict, group-targeted human rights abuses, and significant events that pose risk for major political instability. These brief summaries include information that goes beyond the data in our statistical model, but they are not intended to provide a comprehensive analysis of factors contributing to atrocity risk. Rather, they are intended to serve as starting points for those who are interested in deeper qualitative analysis. For each country, we also identify the specific factors that account for the risk estimates from our model (see “Methods” below for more detail on the risk factors in the model) and note whether the country is experiencing an ongoing mass killing.
The results of this risk assessment should be a starting point for discussion and further analysis of opportunities for preventive action. For countries in each of the following categories, we recommend asking certain key questions to gain a fuller understanding of the risks, adequacy of policy response, and to identify additional useful lines of inquiry.
The remaining seven countries in the top ten are Yemen, India, Nigeria, Somalia, Turkey, Ethiopia, and Burma/Myanmar. Yemen, India, and Nigeria are discussed in different sections below. To learn more about the factors that contributed to the high-risk estimate of any of these countries, visit the country pages on our website.
In addition to Pakistan, Afghanistan, and the DRC, a few other countries have appeared near the top of our rankings for several years.
Two countries that may be conspicuously absent from our highest-risk rankings are Burma/Myanmar and Syria. In Burma, the United States Holocaust Memorial Museum determined in 2018 that the Burmese military had committed genocide against the Rohingya population. The scale and intensity of the war crimes and crimes against humanity in Syria is well-known, with devastating effects on civilians.
The percentage risk and ranking for each country represents the probability that a new onset of mass killing begins in that country—that either a new perpetrator group emerges and kills more than 1,000 civilians of a specific group, or an existing perpetrator group begins targeting a new group of civilians—not that an existing mass killing continues. In the cases of Burma and Syria we already count two ongoing mass killings in each.
In Burma there is the genocide against the Rohingya that culminated in 2017, as well as a long-simmering conflict in the country’s east in which the state has been perpetrating mass killing against other minority groups (i.e., the Karens, Shan, and Mon) since 1948. Note that we consider a mass killing to be “ongoing” until we see three consecutive years with fewer than 100 civilians killed as part of the campaign. Burma’s risk and ranking (six percent risk and tenth rank) represents the likelihood that a new perpetrator group emerges or that the state begins a campaign of violence against a new target group in 2020 or 2021.
In Syria there is an ongoing, state-led mass killing against perceived political opposition since 2011, as well as a non-state-led mass killing perpetrated by the self-proclaimed Islamic State (IS) and its affiliates against perceived opposition since 2012. In the case of Syria, it is difficult to imagine the state or IS targeting a new group of civilians, as the current parameters of the target groups are so broad. That means that Syria’s risk and ranking (5.9 percent risk and 12th rank) is the likelihood that a new perpetrator group emerges in 2020 or 2021.
We highlight three countries that moved up in our rankings substantially between the 2019–2020 and 2020–21 assessments.
One way global statistical risk assessments are helpful is in identifying countries whose relatively high (or low) risk estimates may surprise regional experts. In cases where our statistical results differ substantially from expectations, we recommend conducting deeper analysis and revisiting assumptions. The purpose of this analysis is not to pit qualitative analysts and statistical models against one another but rather to deepen our understanding of risk in the country in question. We highlight three countries that, in our informal judgment, fall into this category.
The data used to produce this assessment is from 2019 (published by most sources in early- to mid-2020). This means that changes due to the COVID-19 global pandemic and various other economic and social crises that occurred in 2020 are not captured in this risk assessment. To enable users to explore how such changes might affect a country’s risk estimate and ranking, we are releasing a new interactive data tool, which allows users to:
For example, in 2020–21, Mali ranks 38th, with a 2.3 percent estimated risk. This assessment is based on 2019 data. However, someone following events in Mali may suspect that events over the course of 2020 may have an impact on that risk, specifically the coup d’état that occurred on August 18. Soldiers from the Malian military detained several government officials including President Ibrahim Boubacar Keïta, who resigned and dissolved the government later that night. Using the tool to change “any coup attempts in the last 5 years” from “no” to “yes”, we see that Mali’s updated risk assessment is 3%.
Another example is Venezuela. On the risk factor “freedom from political killings,” the Varieties of Democracy dataset codes Venezuela as, “Weakly respected by public authorities. Political killings are practiced frequently and top leaders of government are not actively working to prevent them.” The next interval in their coding scheme is “Not respected by public authorities. Political killings are practiced systematically and they are typically incited and approved by top leaders of government.”
Some analysts may argue that Venezuela merits the latter coding – that in 2019 political killings were practiced systematically and approved by top leaders of the government. Changing this variable in in the interactive tool shifts Venezuela’s estimated risk from 3% to 5%, moving its ranking from 88th to 19th.
To produce this assessment, we employ data and statistical methods designed to maximize the accuracy and practical utility of the results. Our model assesses the risk for onset of both state-led and non-state-led mass killings over a two-year period.
The data that inform our model come from a variety of sources. On the basis of prior empirical work and theory, we selected more than 30 variables, or risk factors, as input for our statistical model (see the discussion of our modeling approach, below). All data used in our model are publicly available, regularly updated, and available without excessive delay. They also have, in our estimation, minimal risk of being retrospectively coded in ways that could depend on observed mass killings or their absence, cover all or almost all countries in the world, and go back at least to 1980 (but ideally to 1945). We include variables reflecting countries’ basic characteristics (e.g., geographic region, population); socioeconomic measures (e.g., changes in gross domestic product per capita); measures of governance (e.g., restrictions on political candidates and parties); levels of human rights (e.g., freedom of movement); and records of violent conflict (e.g., battle-related deaths, ongoing mass killings). Alongside the model, we publish a data dictionary and make the model and all data available on our GitHub repository. The only dataset the Early Warning Project maintains is that of ongoing mass killing.
In 2020, the Center for Systemic Peace stopped producing its annual Polity dataset, which included measures of regime type and duration, formerly used by our model. The 2020-21 assessment does not include measures of these risk factors.
Our modeling approach is described in detail on our website. We use a logistic regression model with “elastic-net” regularization. In summary, based on a set of about 30 variables and data on mass killing going back to 1945, the algorithm identifies predictive relationships in the data, resulting in an estimated model. We then apply this model to recent data (from 2019 for the 2020–21 assessment) to generate forecasts. While the exact number of countries varies by year, the project includes all internationally recognized countries with populations of more than 500,000. The model automatically selects variables that are useful predictors; see our methodology page for a list of variables selected by the model. We emphasize that these risk factors should not be interpreted as causes or “drivers” of risk but simply as correlates of risk that have proven useful in forecasting.
We assessed the accuracy of this model in ways that mimicked how we use its results: We built our model on data from a period of years and then tested its accuracy on data for later years (i.e., we conducted out-of-sample testing). Our results indicate that about two out of every three countries that later experienced a new onset of mass killing ranked among the top-30 countries in a given year. See the accuracy page on our website for more details. We are preparing a technical paper in which we assess our model and others according to multiple performance measures.
Early warning is a crucial element of effective atrocity prevention. The purpose of our statistical risk assessment is to provide one practical tool to the public for assessing risk in countries worldwide. This tool should enable policy makers, civil society, and other analysts to focus attention and resources on countries at highest risk, especially those not currently receiving sufficient attention.
This quantitative assessment is designed to serve as a starting point for additional analysis. States and international organizations have developed and implemented tools for qualitative atrocity risk assessments—we see the application of such tools as a complementary next step after our statistical analysis. These in-depth assessments should in turn spur necessary adjustments in strategic plans, budgets, programs, and diplomatic strategies toward high-risk countries. By combining these approaches—global risk assessment, in-depth country analysis, and preventive policy planning—we have the best chance of preventing future mass atrocities.
from the Early Warning Project and the Simon-Skjodt Center for the Prevention of Genocide