The new Sustainable Development Goals (SDGs) include a target to “Significantly reduce all forms of violence and related deaths everywhere.” Given the vast decline in violence since the Middle Ages, particularly since the end of the Cold War, this ambitious target is achievable. But policymakers know the least about the countries receiving the most aid. To ensure that aid and policy are effective, current data gaps and deficiencies must be fully understood and improved. Equally important, the target must include indicators that capture all the main types of violence, not just homicide.

The Data Problem

  • Current statistics are marred by problems that make them incomparable across countries. Policymaking that ignores flawed data may focus on less effective goals or assume programs are working when, in fact, violence is being hidden through statistical manipulation.
  • Policymakers know the least about the countries receiving the most aid. Among the top ten British aid recipients, four have reported no homicide statistics or have had only one data point in twenty-seven years. Eight of the top ten U.S. aid recipients have no reported homicide statistics for the past four years. Egypt, Iraq, and Jordan have no reported homicide statistics since the Arab Awakening.
  • Failure to accurately count different types of violence obscures possible relationships among them. For instance, these include connections between the end of civil war and rising homicide, between state brutality and increased insurgency, and possible connections between state repression and homicide.
Rachel Kleinfeld
Rachel Kleinfeld is a senior fellow in Carnegie’s Democracy, Conflict, and Governance Program, where she focuses on issues of rule of law, security, and governance in post-conflict countries, fragile states, and states in transition.
More >

The Way Forward

  • A global violence dataset that accounts for “all violent deaths everywhere” should include four disaggregated types of data: homicides, deaths among armed groups in conflict, deaths of unarmed civilians perpetrated by state or nonstate actors, and deaths caused by on-duty government security forces.
  • The international community needs accurate data across these categories to know which programs and policies actually reduce violence, rather than simply alter the form violence takes.
  • If the international community does not explicitly include state repression and terrorist killings in the SDG 16.1 target, it opens a loophole to politicizing numbers through reclassification and the use of state violence to try to reduce homicide and rebellion.
  • International actors should press for a comprehensive set of indicators for SDG 16.1, which currently only include homicides.
  • Data reporting and collection could be improved by investing in independent observatories, standardization of definitions and methodologies, and other crucial steps.

These decisions are not technical, but political. Statistical manipulation is inevitable and occurs in countries from the United States to Russia. Impartial, trained, and internationally funded violence observatories can assist in gaining accurate statistics so resources can target the most effective places and programs.


The decision to include a target to “Significantly reduce all forms of violence and related deaths everywhere” (SDG 16.1) represents an enormous opportunity. Since no low-income country facing violent conflict achieved a single one of the Millennium Development Goals, SDG 16.1 offers the international community a chance to make a difference in an area that has held back development and human empowerment in myriad countries.1 Yet it is not enough to have a goal—the worth of the SDG endeavor will depend on the quality and scope of the indicators. They will determine whether progress is actually being made or whether problems are simply being hidden under semantic cloaks. Further, the international community—whose advocacy was essential to putting the landmark SDG 16 and its focus on good government, functional institutions, and peace on the agenda—must pay equal attention to the data used for these indicators.

The goal of reducing violence is not a pipe dream. Violent death, from conflict to homicide, has been falling since the Middle Ages and declined even more precipitously after the Cold War.2 Despite a recent uptick in deaths from warfare, most conflict is now concentrating in just a few places, so that peace is becoming more widespread.3 Meanwhile, homicides continue to decline globally.4 Focused funds and policies could truly reduce violent deaths—if they are well targeted.

Before policies and programs can be judged as successful, however, policymakers and donors must determine whether they actually reduce violence. Unfortunately, the statistics on global violence are currently so poor and so susceptible to manipulation that using them to judge program success risks misdirecting funds.

The statistics on global violence are currently so poor and so susceptible to manipulation that using them to judge program success risks misdirecting funds.

Many policymakers who rely on violence statistics may not understand just how incomplete, incomparable, and problematic the underlying data are for most forms of violence. The United Kingdom has pledged to give half of all its foreign aid to countries that are fragile or in conflict. But among the top ten countries receiving British assistance, Ethiopia, Nigeria, and South Sudan have each reported only one homicide data point in the last twenty-seven years. No data points exist for Tanzania. Sierra Leone—highly at risk for the typical rise in murders following the end of a civil war—stopped reporting four years ago. Only two of the top ten recipients of U.S. economic assistance have reported their homicide statistics in the past four years. Egypt, Iraq, and Jordan, among the top ten U.S. security assistance recipients, have not reported international data for the years following the Arab Awakening, despite being among the largest recipients of U.S. security aid.5 Given that homicide is far and away the greatest global cause of violent death, these facts should be worrying. Yet, the problem is not confined to homicide. Conflict data are even more problematic. The fact is—we know the least about the countries to which we are giving the most money and that are the most beset by violence (see Table 1).

Better data will make it easier to discern where violence is occurring, how different forms of violence interact and change over time, and whether efforts to reduce violence are indeed working. This paper focuses on the need to count different types of violence to avoid political manipulation and enhance understanding of interaction effects among the forms of violence. It also delineates the problems of incomparable and incomplete data and the importance of funding statistical efforts to improve data reporting and collection, so that policy is targeted toward the most effective goals.

What Should Be Counted

The process of counting violence has run into challenges—both political and practical. A group of leading statisticians, national statistics offices, multilateral institutions, and nongovernmental organizations have come together in the United Nation’s (UN) Praia Group on Governance Statistics to craft the indicators that will demonstrate whether a given country is making progress toward SDG 16.1. Since no government is keen to look bad internationally and many are concerned about privacy and sovereignty, the determination of what should be counted has been contentious. 

After a flurry of global activity to get SDG 16 on the agenda, advocacy groups have paid far less attention to how the targets will be counted. To make Target 16.1 meaningful, more global scrutiny of the indicators under development is needed.

A count of violent deaths that is most accurate and least susceptible to manipulation would include four types of violence:

  • Homicide. While war grabs headlines, homicides account for, by far, the greatest number of violent deaths per year. At between 370,000 and 450,000 homicides annually over the past decade according to WHO and UNODC numbers, about three-quarters of violent deaths globally are homicides.6 To put it in perspective, more people are killed by homicide each year than the higher estimates for five years of the brutal war in Syria.7 

    Homicide is the only category of violent deaths certain to be included in the indicators under SDG 16.1. Yet the most frequently cited datasets, such as those of the WHO and UNODC, are based on reporting from only around 50 percent of all countries.8 Both datasets have missing or incomplete data for populous countries, including China, Brazil, and Russia. Data for sub-Saharan Africa are modeled from only a handful of data points, and the numbers that are reported are not currently comparable across countries.
  • Battle-related deaths. In recent years, deaths among armed actors in battle have constituted the second largest number of deaths after homicides. While interstate wars are at an all-time low, when they occur, they can be the most deadly category of all. Meanwhile, civil war deaths are on the rise after a decline.9 While about 70,000 people died from warfare from 2007 to 2012, the number rose to 90,000 from 2010 to 2015, with 80 percent of those deaths emanating from just three countries: Afghanistan, Iraq, and Syria.10

    Despite the importance of including these large and growing numbers in a count of violent deaths, a number of countries are arguing against including battle-related deaths in SDG 16.1’s indicators. One line of argument is that weaknesses in data collection undermine the value of these statistics. There are certainly problems with this data, but there are equally significant problems with homicide data, and the United Nations has chosen to treat those hurdles as a challenge to tackle, not a reason to abandon a crucial statistic. Battle-related deaths should be treated similarly.
  • One-sided violence. One-sided violence is the name Uppsala University’s Conflict Data Program (UCDP) has given to a dataset of deaths caused when organized groups such as terrorists or governments attack unarmed civilians. While in most years, these numbers are smaller than for battle-related deaths and homicides, acts of genocide (for example, in Rwanda in 1994) can kill in numbers akin to interstate war.11

    Meanwhile, major terrorist activities, such as in Nigeria today, can cause significant spikes in violent deaths and are omitted from homicide statistics in some countries. For instance, the largest recent terrorist attack in the United States occurred in June 2016, when a man who had pledged allegiance to the self-proclaimed Islamic State fired into a Florida nightclub frequented by gays, killing forty-nine people. Since the United States does not include terrorist attacks in homicide statistics, those forty-nine deaths would risk being lost, statistically, depending on whether those murders were classified as a terrorist attack or homicidal hate crime.

    One-sided violence is not even being considered as an SDG 16.1 indicator. It should be, but since the likelihood is low due to political realities, it should be included in an independent count. Failing to include one-sided violence makes repressive countries look more peaceful than they are. This skews policymakers’ thinking about the relationship between violence and various political regimes, and it prevents the possible identification of causal relationships between repression and citizen violence that could be important to preventing violent deaths.

    It also opens a gaping hole for politicizing statistics. Unless terrorism is clearly included, countries could choose to reclassify some portion of violent deaths as terrorist attacks or could legally declare groups such as gangs to be terrorists to lower their reported violence numbers. Statistical manipulation of violence numbers has occurred from Chicago and New York to Russia to make administrations look more effective at fighting crime than they actually are, and should be anticipated as violence data begins to be used to determine aid and international programs.12 
  • Legal intervention deaths. Killings perpetrated by government security personnel in the course of domestic duties, such as when a police officer shoots a suspected criminal, are known as legal intervention deaths.

    Legal intervention deaths are also not under consideration for inclusion in the SDG 16.1 indicators. This is understandable, given countries’ reluctance to report such numbers, but it is a mistake. While in most places these numbers are fairly low, in some cases, they are high enough to affect overall violence rates in a country. For instance, Nigeria in 2008 reported nearly 2,000 homicides, but had an additional 857 deaths as a result of legal interventions that did not appear in homicide data.13 These deaths would have raised the homicide rate by more than 40 percent. Even where numbers are low, however, they can be important due to their potential outsized political effects—for example, the current unrest in the United States over police killings of African Americans.

    Moreover, when legal intervention deaths represent a systemic problem, such as in places featuring government death squads or broad-scale brutality, this category of violence can cause parts of the population to lose confidence in state security services, leading to vigilantism and other forms of citizen violence that grow when a state loses legitimacy.14 In extreme cases, rising legal intervention deaths could be a warning sign of other forms of violence to come.

These four types of violence, together, would give an accurate rendering of direct violent deaths. Allowing countries to draw artificial semantic divisions among types of violence undermines the international community’s ability to understand the factors fueling violence and how to reduce it. Meanwhile, during conflict, many types of violence occur simultaneously and motives are difficult to disentangle. Accounting for multiple types of violence is thus the best way to gain an accurate sense of the number of dead, without falling into the impossible trap of trying to uncover the motives of perpetrators and without opening easy loopholes for manipulation.

Some argue that indirect deaths from conflict should also be counted—for example, deaths resulting from the destruction of health systems and food delivery. It is true that these deaths number far higher than direct deaths.15 Moreover, forms of violence short of death, such as rape and domestic violence, clearly deserve counting as ends in their own right and because of the well-documented role of female empowerment in promoting development. However, a reduction in indirect deaths depends less on violence itself than on a multitude of other variables, including the state of a country’s medical infrastructure and roads, which should be captured in other SDG targets.16 The challenges to collecting serious data on sexual and family violence are so severe that they merit a separate assessment and an indicator unto themselves. For Target 16.1, the four indicators above would provide the clearest understanding of direct violence in a country.

Policy Gains from Counting All Types of Violence

Criminologists who study criminal violence such as homicide are distinct from scholars who look at conflict, insurgency, and terrorism, who are themselves separate from those who study genocide and the killing of civilians from legal interventions. Scholars and practitioners from these different disciplines rarely talk to one another or read one another’s work. Datasets that look at these different forms of violence are currently incompatible and rarely contain enough data on the same country to draw causal relationships. Yet studying only one type of violence can obscure interrelationships between them. That prevents policymakers from understanding how a policy intervention or political change may alter, rather than simply increase or decrease, violence within a country.

In El Salvador, one of the few places where data are comparable and available, the end of civil war violence heralded a decline in battle-related deaths but an uptick in homicide. The homicide rate jumped from 1,464 people killed in the year before the end of its civil war (1991) to 2,480 the year after the war ended (1993).17 Only by looking at both homicide and battle-related deaths together can policymakers see that the end of war transformed, rather than reduced, violent death in the country.

Criminal and political violence might be exacerbated by similar root causes— particularly governments that lose the trust of their people and exacerbate divisions among their populations, often through selective state repression.

El Salvador is not alone. Scholars have long recognized that homicides often increase following civil wars, due to reprisal political killings that appear as regular homicides, difficulty disarming and reintegrating combatants, disorganized law enforcement, delegitimized governments, the availability of arms, and the decivilizing process of war.18 Yet as civil wars ended following the end of the Cold War, homicide declined globally. Does that global decline hide rising numbers in the countries where civil wars ended, suggesting that the reduction in other countries was even greater than previously believed? It is impossible to know, because the WHO was the only entity keeping public, globally comparable homicide statistics during the 1990s, but WHO data capture less than a third of the global population and are particularly sparse in countries facing conflict.19

Scholars have long assumed that political and criminal violence stem from different motivations and follow different trajectories. However, little is actually known about the relationships among these forms of violence because the numbers simply do not exist. Case studies from Iraq, Latin America, and the United States provide tantalizing glimpses that criminal and political violence might be exacerbated by similar root causes—particularly governments that lose the trust of their people and exacerbate divisions among their populations, often through selective state repression.20 Criminal and political violence may also, at times, move in tandem. It is well-known that groups engaged in conflict often use criminal means, including drug or human trafficking, to finance themselves. For instance, a specialist on the Irish Republican Army estimated that its top two sources of income in the 1970s were robbery and racketeering, while Colombia’s illegal logging, mining, and multiple other illicit businesses are tied to financing the paramilitaries and guerrilla groups.21 In Colombia, where exemplary countrywide statistics exist, Colombian violentologists found that when paramilitary groups were negotiating an amnesty and they and the Revolutionary Armed Forces of Colombia (FARC) agreed to a ceasefire, homicide and violent deaths from organized crime also plummeted in the country.22 This is not to say that revolutionary groups are the same as criminal groups. But in such cases, looking at homicide and battle-related deaths side by side provides a more accurate understanding of how people are experiencing violence within a country and how forms of violence are related.

Looking at multiple forms of violence could provide important information that would enhance understanding of how to address some of the most violent places in the world, where multiple forms of violence exist simultaneously. Many important questions could be answered if these different forms of violent deaths were counted and disaggregated. To highlight just a few:

  • In countries where multiple types of violent groups exist, are the methods for negotiating peace different than in those countries that must address only one violent entity? For example, the Colombian government’s prior agreement with paramilitary groups formed a baseline for its negotiations with the FARC, while negotiators were aware that remaining insurgent groups were watching the deal offered to the FARC. In 1991, Colombia’s drug cartels attempted to negotiate a truce based on an earlier peace agreement the government had made with the guerrilla group M-19. The iterative nature of such peace deals seems intuitively different than those in countries seeking to address just one violent faction.
  • Many case studies suggest that government repression yields an increase in violent insurgency.23 In her studies of civil war, for example, Barbara Walter finds that significant reductions in the number of political prisoners and extrajudicial killings make the renewal of civil war between two and three times less likely than in countries with higher levels of human rights abuses.24 Could looking at one-sided violence and legal intervention violence help policymakers predict and perhaps intervene to prevent future battle-related deaths?
  • Might repression also create a lack of trust in the government that increases homicide? Criminal violence is generally treated separately from politics, and yet the historian Randolph Roth found that trust in government was highly correlated with the homicide rate in the United States; and the decivilization theory of Nobert Elias suggests that government repression might reduce inhibitions and lead to greater societal violence.25 Collecting one-sided violence and legal intervention deaths alongside homicide statistics could enable tests of such a theory, which would have serious implications for policing policy.
  • To what extent are violent governments blaming criminals or rebels for what is, in actuality, violence connected to the state or elite economic interests? In El Salvador, for instance, which has had among the top homicide rates in the world in recent years, gang members are commonly blamed for the stunning rate of violence. But in 2007, the United Nations found that gangs were responsible for only about one-quarter of the killings.26 Instead, the United Nations pointed a finger at organized crime, which is almost always in league with parts of a government. It also identified a growing number of police-run “social cleansing” units that tortured and killed both political targets and gang members. The units came to light after three Salvadoran politicians were killed in Guatemala and an ensuing investigation uncovered high-level links among the Ministry of Internal Affairs, the National Civilian Police, and drug cartels.27 If state, rather than gang, violence was a cause for some significant portion of the violence in a country, that would be important information for measuring the effectiveness of, say, gang-violence reduction programs versus more political or diplomatic interventions.

Much of the world is becoming more peaceful, while violence is concentrating in a few places.28 Counting and disaggregating multiple types of violent deaths will allow policymakers, development practitioners, and scholars to better understand the dynamics of places exhibiting “compound violence” and to gain a deeper sense of how to fight multiple forms of violent activity.29

Counting and disaggregating multiple types of violent deaths will allow policymakers, development practitioners, and scholars to better understand the dynamics of places exhibiting “compound violence.”

Conversely, choosing to count homicide alone, as some are arguing in the Target 16.1 debates, would misallocate resources among regions and lead to policy solutions that could harm the very people the international community is trying to help. For instance, homicide appears to be high in Latin America and surprisingly low in Africa, given that homicides usually rise during conflict and the African continent has the greatest number of conflicts. However, low homicide numbers likely have to do more with the failure of many African countries to report homicide statistics than with the reality of violent deaths. Choosing not to count both homicide and conflict-related deaths under Target 16.1 would further obscure the number of dead in Africa and would likely lead to theories on violence that (being based on the overrepresentation of data from Latin America) are actually comments on Latin America’s particular problems, rather than an assessment of global reality. 

Practical Difficulties With Data Collection

There are problems with collecting all forms of data on violence. Most of these challenges, while difficult, can be tackled through standardizing methodologies and definitions, building on and amending existing datasets, and investing in government statistics offices. All of these choices are political, however, and data issues involving politicization are not easily fixed technically. They are best addressed by maintaining a skeptical and cautious international public and by investing in an independent count of violence globally and in independent, impartial, and well-trained violence observatories (similar to those funded in many Latin American countries).

Many countries keep individual data, and a number of universities and nongovernmental organizations such as the Human Rights Data Analysis Group or the Armed Conflict Location and Event Data (ACLED) group maintain data on specific countries, regions, and conflicts. The data from these sources are often more comprehensive than any other statistics, but they are limited in scope. Only a handful of datasets are regularly maintained and updated, are suitable for cross-country comparisons, and disaggregate violence to the level of individual death rather than by conflict or event. The most useful and comparable datasets are generated by the WHO, UNODC, Uppsala University/the Peace Research Institute, Oslo, and the Small Arms Survey. But even these datasets have large gaps in data and must contend with different country definitions of homicides and different country choices regarding death reporting requirements. The Small Arms Survey dataset is by far the most comprehensive source of data—including homicides, battle-related deaths, legal-intervention killings, and one-sided violence—but it does not go far enough back in time and suffers from the same problems of underreporting that other data sources do.

Only a handful of datasets are regularly maintained and updated, are suitable for cross-country comparisons, and disaggregate violence to the level of individual death rather than by conflict or event.

Below are some of the specific constraints impeding data collection under the four types of violence essential for a comprehensive understanding of Target 16.1.


No country is arguing that homicide data be excluded from the SDG targets. Yet the problems inherent in homicide statistics are well-known among researchers. These issues make the numbers far less comparable across countries and useful for policymaking than is generally acknowledged.

Underreporting, Particularly in Sub-Saharan Africa

The UNODC is the most comprehensive multinational source of homicide statistics. However, in 2012, the UNODC received reports from only 50 percent of all countries.30 Responses are particularly low in sub-Saharan Africa, while only four countries in Southeast Asia and one country in South Asia reported their homicide numbers in 2010–11. These response rates are similar to previous years.

Far fewer countries report health statistics than police statistics. Africa is again particularly underrepresented, forcing the WHO to model more than two-thirds of the country data rather than report empirical counts.31 The paucity of empirical data forces modelers to make broad generalizations from few data points and potentially undermines findings by building assumptions into the data itself.32

Finally, where both law enforcement and morgue statistics are provided, there are sometimes significant discrepancies. Africa had just three countries in 2012 that provided both sets of statistics, and there was almost a full standard deviation of difference between them.33 In Asia, there was also a significant gap.


For both law enforcement and morgue purposes, most countries require a body to register a homicide. Thus, in Colombia, criminal organizations have turned to “chop houses,” where victims are dismembered to avoid leaving a statistic.34 In Sicily, the mafia prefers to drop some victims into vats of acid, a method known as lupara bianca, to avoid reports of a homicide. In many countries, numerous individuals who have disappeared—some percentage of whom are likely to be victims of homicide—are not counted. In Mexico, for instance, the disappeared appear to total at least 27,000 as of 2015, and none are counted as victims of homicide.35 Paradoxically, countries beset by criminal groups strong enough to have developed means to cover their tracks may end up looking less violent than those with less organized crime.

Lack of Comparable Data

Finally, although the definition of homicide is more homogenous than that of more culturally defined crimes such as rape, it is not uniform. For example, no two European countries have precisely the same definition. Worldwide, some countries such as Mexico include accidental deaths (for instance, vehicular accidents) in their homicide statistics, but most countries omit these numbers or included them for a time and then stopped.

Altering how and when police report a death as a homicide rather than as a death caused by unknown or other causes is a favorite method of manipulating statistics.

Countries also differ in how and when they report law enforcement numbers. For example, some countries report homicides only after successful prosecutions or once a case is closed. Given low clearance rates, this can cause significant undercounting. Altering how and when police report a death as a homicide rather than as a death caused by unknown or other causes is a favorite method of manipulating statistics.36 Because of the potential to politicize law enforcement numbers and country differences on definitions and when a homicide is counted, most academic researchers prefer to rely on the WHO’s homicide counts, which come from mortuaries counting bodies. Yet these counts are far less comprehensive and are particularly weak in countries in conflict. Thus, homicide statistics are not truly comparable across countries, though they are frequently used that way.

These examples are not intended to single out countries or agencies; no entity is immune from the regular bureaucratic difficulties of counting homicides that tend to result in data discrepancies. Instead, they show the need for caution in drawing cross-country or global comparisons based on current homicide statistics—at least until broader consensus is achieved on a more standardized definition and methodology.

The UNODC is working toward improving these statistics. In 2013, the UN Statistical Commission and the UN Commission on Crime Prevention and Criminal Justice developed a two-part program to improve statistics. They plan to bolster collection through country-level focal points, national violence observatories, and regional groups. They also wish to enhance comparability through a new protocol known as the International Classification of Crime for Statistical Purposes (ICCS). The ICCS would standardize how homicide data are collected and would include terrorist murders, killings by police officers using excessive armed force, and extrajudicial killings. If implemented, and if these categories were disaggregated, these changes would result in significant progress toward gaining better homicide statistics. Currently, the data have been tested in pilot countries, and a technical advisory group is working on creating data collection manuals and other tools (translated into a number of languages), which are expected in 2017. However, resources currently lag far behind ambition. Without money to help countries overcome problems with collection, and pressure to overcome issues of political will, implementation is unlikely to be comprehensive, particularly in some of the world’s most violent countries.

Legal Intervention Killings

Many of the same issues that trouble homicide statistics also plague statistics on legal intervention killings. Again, Africa’s numbers barely exist, country data are not comparable, and perhaps most crucially, undercounting is endemic and politicized.

The global thinking on violence might fundamentally alter if legal intervention killings were counted. As noted earlier, if Nigeria counted its legal intervention deaths, its homicide rate would rise by more than 40 percent. The problem occurs globally. In the Dominican Republic in the mid-2000s, the total killings may have been undercounted by as many as fifty-eight violent deaths per month.37 In the United States, legal intervention killings in 2015 may have been as high as eighty per month.38 It is hard to know for certain, because reporting from each of the hundreds of police precincts across the United States has been voluntary. A new law, if it holds, will make reporting mandatory in the future.

The paucity of statistics on homicide and the lack of legal intervention death numbers are so great that they render any attempt to draw global conclusions from current numbers suspect. For instance, many policymakers, journalists, and institutions such as the World Bank claim that Latin America, home to just 9 percent of the world’s population, contains more than 30 percent of its homicides.39 However, there is little evidence to support this statement. It is clear that homicides are high in Latin America, but to create an accurate global comparison, more numbers from other continents would be needed.

The paucity of statistics on homicide and the lack of legal intervention death numbers are so great that they render any attempt to draw global conclusions from current numbers suspect.

Modeling is, of course, an accepted method of addressing gaps in data. Yet the gaps in data are so large and so complete in some regions that statistical models are based on little empirical reality—a problem of which the modelers themselves are keenly conscious.40 Not only is sub-Saharan Africa a dark spot of inferred reality, but so are other important areas where trends may be unique. For many years, the Soviet Union refused to report statistics to the United Nations, claiming that there was no crime to report; today, researchers believe Russia’s real homicide rate may again be a third higher than it is reporting.41 China, home to one-seventh of the world’s population, has extremely patchy homicide statistics; and deaths caused by forced labor camps or extrajudicial execution are unknown, though it is evident that state execution rates are high. Other regions with known high levels of legal intervention killings include populous South Asia, where these killings are known as encounter killings. They are believed to have risen in the 1990s in India, but between underreporting and false encounters that actually entail police killings in custody, it is difficult to know the true magnitude of the violence.

Meanwhile, the Middle East is particularly difficult to parse. Consider Egypt’s murder rate, which has risen threefold since the Arab Awakening but is still relatively low. In 2011, 990 deaths were reported as homicides and thirty-one deaths were reported as conflict-related deaths. Yet according to the media, in January and February alone, at least 841 people were killed in unrest.42 Unless almost no one died the rest of the year, somewhere, the numbers are going awry. Given what we know about Egypt’s 2011 crackdown, many of these deaths may be legal encounters, but as they are not accounted for, it is impossible to know what the actual death rate is or what is causing it. Even if the numbers are relatively low, a rise this great in a single country matters to the trajectory of that country. The international policy choices should differ depending on whether the deaths are related to civil wars, interpersonal killings, or intentional government targeting of its civilians through its security services or allied militias.

Battle-Related Deaths and One-Sided Violence (Together Known as Direct Conflict Deaths)

Annual Versus Event Reporting

To gain the most useful and comparable statistics, conflict deaths, like homicides, would be assessed annually. Yet traditional counts of conflict deaths are determined by “event”—in other words, deaths are assessed per conflict, not per year. Sometimes a single conflict spans many years, while in other cases, multiple conflicts occur in a single year in a given country and are counted separately. This problem is frequently cited in the protestations against collecting conflict data for Target 16.1, yet it is hardly insurmountable. Uppsala University’s dataset makes its conflict data accessible in a country-year format. Other datasets could also be broken down from event into annual data, depending on how the underlying information has been collected and coded.

Linking Deaths to Countries

Once conflict or one-sided violence data are broken down by year, data analysts will need to assign country locations to various episodes of violence. This is easy to do for violence within a given country, which, given the current paucity of interstate war, accounts for the vast majority of most of these types of violence. It is more difficult for a subset of these data: transnational violence that crosses borders. Consider Boko Haram terrorists in the Lake Chad region. If terrorists from one country inflict deaths in another, should deaths be assigned based on the location of death or the country of the perpetrator? If deaths result from an interstate war, or a conflict using interstate proxies, should the count be based on the country of citizenship of the dead or the country of origin of the perpetrator(s)? Should the number of deaths simply be divided among the number of countries in conflict? A decision must be made, and it will not be perfect in every case. Yet as long as it is made consistently, it will not hamper statistics over the long run. Uppsala University made such a decision and its newest data are now georeferenced, so researchers can reclassify incident data by individual countries.


Finally, the raw data must capture as many violent deaths as possible. Most conflict datasets, such as the University of Michigan’s Correlates of War dataset, begin to count at 1,000 deaths, which has long been the scholarly definition of war. However, an increasing number of violent deaths are being caused by small organized groups, such as guerrillas and paramilitaries. These deaths are typically not captured in homicide datasets, but do not meet the 1,000 deaths level to be included in conflict datasets. This problem could be solved by the use of Uppsala University’s twenty-five-death threshold, which would allow far fewer deaths to escape through the net.

Meanwhile, one-sided violence perpetrated by terrorists, rebels, or governments is captured in a variety of datasets. Some governments include these data in their homicide statistics, but others do not, and though they are required in the new ICCS, implementation is likely to be slow and patchy at best. The more than 3,000 victims of the September 11, 2001, terrorist attacks were not counted in U.S. homicide statistics, and the nearly 200 killed in Madrid in 2004 were omitted from its counts.43 If one-sided violence is to be counted separately, which is essential, these deaths will need to be disaggregated so they are not double-counted among the countries that do currently include these numbers in their homicide statistics.

Uppsala University’s UCDP dataset is the most comprehensive and useful for longitudinal purposes, but it is particularly vulnerable to undercounting.44 It requires a known perpetrator to record a death, meaning that it undercounts deaths in messier circumstances, such as Mexico’s drug war. The undercounting can be severe. For instance, in Iraq in 2006, the data provided by the Iraq Body Count Project record 27,000 civilians killed, while the combined UCDP datasets—covering state-based conflict, nonstate conflict, and one-sided violence—register only 4,261 deaths in total for the same period, a six-fold difference.45 The undercounting problem could be somewhat ameliorated if the UCDP would relax its perpetrator requirement. Given that it already collects the underlying data and simply does not report those deaths for which a perpetrator cannot be assigned, this would not be a major hurdle.

The more than 3,000 victims of the September 11, 2001, terrorist attacks were not counted in U.S. homicide statistics, and the nearly 200 killed in Madrid in 2004 were omitted from its counts.

A larger but equally solvable problem is overcoming UCDP’s reliance on English-language media to attain its conflict counts. Relying on media reports generally results in undercounting, particularly in countries where media are repressed or in rural regions where media are less prevalent. Relying on English-language sources greatly exacerbates the problem. For instance, when a Bogota-based think tank, CERAC, compared its figures of Colombian battle-related deaths with those from the UCDP, they found that Uppsala counted less than half the deaths most years.46

Some of this variation can be attributed to definitional issues. However, much of the gap can be attributed to a large number of clashes that each resulted in a small number of deaths that appeared in local Colombian press, but did not receive coverage in English-language press.47 Uppsala has begun addressing this problem in a limited fashion by including some local media news translated by the BBC monitoring service. In the short term, the problem of undercounting could be offset by modeling with a proxy variable for conflicts in non-English-speaking countries or regions. In the long term, more accurate statistics could be gained with greater funding to enable multiple methods of data checking and by piecing together indices that are less comprehensive globally but more thorough within a particular country or region, such as those employed by the Human Rights Data Analysis Group or ACLED. Meanwhile, more user-friendly formats, similar to those generated by the ACLED, enable simpler mapping and infographics, thus improving the ability of users to spot and help correct errors.

Who Is Authorized to Count?

Many countries protesting the inclusion of conflict statistics in SDG 16 aim to protect their sovereignty over data. In other words, they argue that data should be subject to their own laws and fear that data held in servers abroad may not be safe from the intelligence agencies and defense establishments of other countries. This concern makes it difficult for Target 16.1 indicators to draw on accurate and comprehensive counts from groups like the Center for Systemic Peace, which has connections to the U.S. government. However, there are multiple means to gain globally accessible statistics while maintaining the privacy and sovereignty of country data. The African Union has collaborated with the UN Economic Commission for Africa, the African Development Bank, and members of the African Statistical System to develop the Strategy for the Harmonization of Statistics in Africa. Similarly, Eurostat led the process of collecting, validating, and harmonizing the statistics of European Union member states. These regional mechanisms for data protection can be extended to cover conflict deaths or could inform a separate solution.

If countries succeed in blocking a UN-approved statistic, it does not mean these deaths go uncounted. They will simply continue to be counted by nongovernmental organizations, academics, intelligence agencies, and potentially, global big data firms such as Google.

Of course, if countries succeed in blocking a UN-approved statistic, it does not mean these deaths go uncounted. They will simply continue to be counted by nongovernmental organizations, academics, intelligence agencies, and potentially, global big data firms such as Google. Some of these entities, particularly the Small Arms Survey, do an estimable, difficult job, but others may have inherent biases. Private groups may use data to further business interests. Humanitarian organizations may have an incentive to portray greater numbers of deaths to garner donations. For instance, the International Rescue Committee, which does valuable and excellent humanitarian work, has been accused of augmenting death figures in the Democratic Republic of the Congo based on shaky statistical methods, and one of the researchers was later cited for falsifying mortality statistics in the Iraq War.48

The media also shapes the public perception of violence. The incentive structure of the press tempts journalists to dramatize the intensity of war. To make headlines, journalists are better off using the highest estimates of casualties. The yellow journalists of the late 1800s, for instance, famously sensationalized deaths in Cuba to trigger a U.S. intervention there over a century ago.

These numbers become “true” by repetition in the public mind, but can lead to serious flaws in policymaking and public opinion. The world has made great strides in learning how to facilitate and maintain peace. But exaggerated death tolls shape how the public views the tractability of violence, which thus shapes the nature and feasibility of the response. Sensationalized reporting contributes to the problem of violence by creating a sense of helplessness among countries that can, in fact, help. It can make violence seem inevitable when it is, in fact, caused by and can be solved by human agency. Comprehensive, comparable data from unbiased entities will allow policymakers and development practitioners to understand how violence actually functions, where it is most severe, and what policies and strategies might improve the situation.


Failure to create comparable, accurate, reliable statistics on homicides, battle-related deaths, one-sided violence, and legal intervention deaths will maintain or worsen the international community’s skewed understanding of violence. It may also have serious practical consequences for fragile states by undermining a statistical lever that could galvanize the global community into providing more consistent funding to better prevent, contain, and end violence worldwide. It could also exacerbate the misallocation of funds as money is directed to countries that keep more accurate statistics, while failing to reach countries that face high levels of violence but have difficulty reporting or are deliberately failing their citizens by choosing not to report.

Failure to create comparable, accurate, reliable statistics on homicides, battle-related deaths, one-sided violence, and legal intervention deaths will maintain or worsen the international community’s skewed understanding of violence.

Many statistical challenges have practical solutions. Standardizing definitions, as the ICCS has already done for homicide and as is equally possible for conflict-related deaths; relaxing perpetrator requirements for reporting deaths; and ensuring that local language sources are used for counting conflict-related deaths would each be significant steps forward in ensuring that every life matters and every violent death is counted. Choosing to count and disaggregate the four most important sources of direct violent death would reduce the chances for data manipulation and allow for the crucial research that policymakers need to reduce violence.

None of these interventions are technical, however simple they seem. Each has significant political ramifications. Thus, creating an accurate, reliable, and comparable data source held by a respected nonpartisan international organization—but outside the SDG system—may be the best solution. In tandem, investing in independent violence observatories in areas undergoing conflict or with potentially high levels of violence will be crucial for gaining accurate data; this approach already has a successful track record in multiple countries.

While it will be hard to gain buy-in, making a sustained effort will be worthwhile. The Millennium Development Goal experience has demonstrated how data collection can be nurtured and improved significantly when political will and expertise come together. SDG 16 provides a unique opportunity to get at the root of a problem that has inhibited development for many decades. For the half million people who die violently each year, and the more than 1.5 billion people living under the threat of violence, the need to invest in a comprehensive assessment of global violence worldwide, whether inside the UN system or independently, is clear.49


1. “UN System Task Team on the Post-2015 UN Development Agenda: Peace and Security,” United Nations Peacebuilding Commission, May 2012, 5

2 For a full collection of the evidence of decline, see Steven Pinker, The Better Angels of Our Nature: Why Violence Has Declined (New York: Penguin, 2011). For graphics depicting the stunning nature of this fall in violence, see Max Roser, “Homicides,” Our World in Data, accessed January 3, 2016,; and Max Roser, “War and Peace,” Our World in Data, accessed January 3, 2016, The underlying data are drawn from  “Data on Armed Conflict,” Peace Research Institute Oslo (PRIO), accessed November 16, 2016,; and Manuel Eisner, “Modernization, Self-Control and Lethal Violence: The Long-Term Dynamics of European Homicide Rates in Theoretical Perspective,” British Journal of Criminology 41, no. 4 (September 2001): 618-–36.

3 Violence is increasingly concentrating in a small number of countries that are growing more violent. Just eighteen countries with only 4 percent of the world’s population account for nearly one-quarter of all violent death worldwide. Three countries account for 80 percent of all direct conflict deaths. See “Monitoring Trends in Violent Deaths,” Research Notes 59, Small Arms Survey, September, 2016, 4; and “Every Body Counts,” Research Notes 49, Small Arms Survey 2015, 51.

4 See the United Nations Office on Drugs and Crime (UNODC) data in Commission on Crime Prevention and Criminal Justice, “World Crime Trends and Emerging Issues and Responses in the Field of Crime Prevention and Social Justice,” UN Economic and Social Council, February 12, 2014; and UNODC, Global Study on Homicide 2013: Trends, Contexts, Data (Vienna: UNODC, 2013).

5 In this paper, internationally comparable data are treated as emanating from the UNODC or WHO. Some countries have locally available data that differ from or are not reported to these organizations, but it is difficult for most researchers or policymakers to navigate through raw police statistics.

6 Approximately 83 percent of violent deaths occur outside of conflict and are mostly homicides. See “Monitoring Trends in Violent Deaths,” 2.

7 Syrian war death statistics vary, and most counts include what the UCDP calls one-sided violence, in which the government or terrorist groups murder unarmed civilians rather than combatants. Thus, being most conservative, the Syrian Center for Policy Research estimated that around 470,000 Syrians had been killed in the war as of December 2015, while the Syrian Observatory for Human Rights, which does not disclose its methodology, estimated 370,000 deaths in the first five years of the war, and the United Nations estimated around 250,000 but noted that the numbers were indicative, not accurate. Homicide statistics are drawn from Small Arms Survey datasets and the UNODC.

8 The UNODC’s global studies on homicide each have a methodology section at the end with reported numbers, which hover around 50 percent each time. See, for instance, the UNODC Global Study on Homicide 2013, 100.

9 See the Civil War Datasets kept by Uppsala University at “UCDP Datasets,” Department of Peace and Conflict Research, Uppsala University, accessed November 17, 2016,; and the Centre for the Study of Civil War and the Uppsala Conflict Data Program, “Data on Armed Conflict,” PRIO, accessed November 16, 2016,

10 These three countries accounted for 80 percent of battle-related deaths. See “Monitoring Trends in Violent Deaths.”

11 Estimates for the Rwandan genocide range from 500,000 to 2 million. Uppsala University puts the death toll at around 800,000, and scholarship suggests that between 600,000 and 800,000 Tutsis and perhaps an additional 60,000 Hutus lost their lives in 1994. See Marijke Verpoorten “The Death Toll of the Rwandan Genocide: A Detailed Analysis for Gikongoro Province,” Population 60, no. 4 (2005): 401–39.

12 For more on statistical manipulation in New York, see John A. Eterno and Eli B. Silverman, The Crime Numbers Game: Management by Manipulation (New York: CRC Press, 2012). For Chicago, see the three-part investigative report by David Bernstein and Noah Isackson: “The Truth About Chicago’s Crime Rates,” Chicago Magazine, April 7, 2014, and May 19, 2014; and their follow-on article, “New Tricks,” Chicago Magazine, May 11, 2015. For Russia, see Alexandra Lysova and Nikolay Shchitov, “What Is Russia’s Real Homicide Rate? Statistical Reconstruction and the ‘Decivilizing Process,’” Theoretical Criminology 19, no. 2 (2015): 257–77.

13 Numbers are drawn from the CLEEN Foundation, a Nigerian justice sector reform organization, cited in The Global Burden of Armed Violence 2015: Every Body Counts, eds. Anna Alvazzi del Frate, Keith Krause, and Matthias Nowak (Cambridge, UK: Cambridge University Press, 2015), 37.

14 Vanda Felbab-Brown, David Kilcullen, and the author of this paper have all documented how a state’s loss of legitimacy can lead to vigilantism and citizens turning to violent groups such as criminal gangs and insurgents for succor. See David Kilcullen, Out of the Mountains: The Coming Age of the Urban Guerilla (Oxford: Oxford University Press, 2013), 116–68; Vanda Felbab-Brown, Aspiration and Ambivalence (Washington, DC:Brookings Institution, 2012); and Rachel Kleinfeld (New York: Knopf, forthcoming).

15 Robert Muggah, “Counting Conflict Deaths: Options for SDG 16.1” (briefing note to the members of the Inter-Agency and Expert Group on SDG Indicators), Saferworld and Igarapé Institute, October 2015.

16 To be fair, homicide and conflict deaths also depend on medical infrastructure—advances in medicine are responsible for some reduction in direct violent death from both these causes, particularly in the developed world.

17 The data on homicides are drawn from WHO annual statistics. See the World Health Organization Global Health Observatory Data Repository, “Homicide Estimates by Country,” WHO, accessed November 16, 2016,

18 Edith Abbott, “The Civil War and the Crime Wave of 1865–70,” Social Service Review 1, no. 2 (June 1927): 212–34; Dane Archer and Rosemary Gartner, “Violent Acts and Violent Times: A Comparative Approach to Postwar Homicide Rates,” American Sociological Review 41, no. 6 (December 1976): 937–63; and Ted Robert Gurr, “Historical Trends in Violent Crime: Europe and the United States,” in Violence in America: The History of Crime (New York: Sage Publications, 1989), 21–54, 47.

19 The WHO includes 140 countries in its datasets. Yet many of these countries have sparse data points; for instance, Bosnia has only one data point between 1989 and 2014. Bangladesh, Cambodia, Ethiopia, Kenya, Libya, and Nigeria, among others, do not appear at all. Neither do China and India—rather large omissions. During the 1990s, the WHO collected data on no more than 2 billion people in the world.

20 Randolph Roth, American Homicide (Cambridge, MA: Harvard University Press, 2009); Michael Stohl and George Lopez, Government Violence and Repression: An Agenda for Research (New York: Praeger, 1986); and Keith Krause, Robert Muggah, and Elisabeth Gilgen, eds., The Global Burden of Armed Violence 2011: Lethal Encounters (Cambridge, UK: Cambridge University Press, 2011).

21 Brendan O’Leary and Andrew Silke, “Understanding and Ending Persistent Conflicts: Bridging Research and Policy,” in Terror, Insurgency, and the State: Ending Protracted Conflicts, eds. Marianne Heiberg, Brendan O’Leary, and John Tirman (Philadelphia: University of Pennsylvania Press, 2007), 205; author interview with Jorge Restrepo (executive director of CERAC), Colombia, January 27, 2015.

22 Author interviews with Jorge Restrepo, Colombia, January 27, 2015; author interview with Jorge Giraldo Ramirez (EAFIT), February 4, 2015; statistics from CERAC and Ramirez.

23 Indiscriminate state repression tends to help violent insurgencies recruit, as people begin to see such violent organizations as more able to offer protection than more moderate groups. See Jeff Goodwin, No Other Way Out: States and Revolutionary Movements, 1945–1991 (Cambridge, UK: Cambridge University Press, 2001), 48, 143, 162, 178; John Walton, Reluctant Rebels: Comparative Studies of Revolution and Underdevelopment (New York: Columbia University Press, 1984); Benedict J. Kerkvliet, The Huk Rebellion: A Study of Peasant Revolt in the Philippines (Berkeley: University of California Press, 1977), 192–93, 240–41, 227, 238; T. David Mason and Dale A. Krane, “The Political Economy of Death Squads: Toward a Theory of the Impact of State-Sanctioned Terror,” International Studies Quarterly 33, no. 2 (June 1989): 175–198; and Ted Robert Gurr, “The Political Origins of State Violence and Terror: A Theoretical Analysis,” in Government Violence and Repression, eds. Stohl and Lopez, 45–72.

24 Barbara F. Walter, “Conflict Relapse and the Sustainability of Post-Conflict Peace” (background paper for the World Development Report, 2011), cited in World Development Report 2011: Conflict, Security, and Development (Washington DC: World Bank, 2011), 82.

25 Roth, American Homicide, 17–18, 450–51; and Norbert Elias, The Civilizing Process: Sociogenetic and Psychogenetic Investigations, revised edition (Cambridge, MA: Blackwell, 2000); these theories also are explored in Amy E. Nivette and Manuel Eisner, “Do Legitimate Polities Have Fewer Homicides? A Cross-National Analysis,” Homicide Studies 17, no. 1 (February 2013): 3–26.

26 United Nations Security Council, Report of the Joint Group for the Investigation of Politically Motivated Illegal Armed Groups in El Salvador (New York: United Nations, 1994), transmitted with letter dated August 11, 1994, from the UN Secretary-General addressed to the president of the UN Security Council S/1994/989.

27 Ibid.; and Procuraduria para la Defensa de los Derechose Humanos (PDDH), Violaciones a los derechos humanos por responsabilidad de la Policia Nacional Civil de El Salvador [The responsibility of the national civil police of El Salvador for human rights violations], (San Salvador, El Salvador: PDDH, 2007), 60–61.

28 See endnote 3.

29 These linkages are explored in the author’s forthcoming book (New York: Knopf, forthcoming).

30 The UNODC’s global studies on homicide each have a methodology section at the end with reporting numbers, which hover around 50 percent each time. See, for instance, UNODC, Global Study on Homicide 2013, 100;  see also, United Nations Development Account, “Strengthening Statistical Capacity for Crime Prevention in Asia,” United Nations, accessed January 3, 2016, The countries that reported in Asia in 2010–11 were India, Japan, the Philippines, South Korea, and Thailand. Notable nonreporters such as Afghanistan and Pakistan would likely have significant violence statistics were they to be counted; the trajectory of a country in transition such as Myanmar requires a baseline and yet will not have accurate counts to allow an understanding of change over time.

31 UNODC, Global Study on Homicide 2013, 49.

32 Smalls Arms Survey, Graduate Institute of International and Development Studies, Small Arms Survey 2011: States of Security (Cambridge, UK: Cambridge University Press, 2011).

33 UNODC, Global Study on Homicide 2013, 101.

34 Author interview with a member of Colombia’s National Security Council, January 26, 2015.

35 Between 1980 and 2010, the UN Working Group on Enforced or Involuntary Disappearances asked governments to take a look at many cases of the disappeared. The numbers could be quite large—for instance, Iraq had 16,545 cases, Sri Lanka had 12,230, and Guatemala had 3,155. Not all disappearances are murders, but many are. Between 2009 and 2011, Colombia found that 35 percent of the reported disappearances it cleared were homicides, and in Mexico, nearly 40 percent were found to be homicides. See Krause, Muggah, and Gilgen, eds., The Global Burden of Violence 2011.

36 See endnote 12, particularly the articles on Chicago.

37 Todd Foglesong and Christopher Stone, “Measuring the Contribution of Criminal Justice Systems to the Control of Crime and Violence: Lessons From Jamaica and the Dominican Republic,” KSG Working Paper no. RWP07-019, Harvard University, April 2007, 17–18.

38 In 2015, the Washington Post began an attempt to count police homicides and found 965 deaths at the hands of police (Kimberly Kindy et al., “A Year of Reckoning: Police Fatally Shoot Nearly 1,000,” Washington Post, December 26, 2015), while the Center for Disease Control counted 16,121 homicides, which exclude police killings. In 2014, in response to public turmoil, Congress passed the Death in Custody Reporting Act, requiring local and state police to report data on civilian deaths in interactions with law enforcement. As of this writing, the Federal Bureau of Investigation is just beginning to set up a system to enable systemic knowledge of civilians killed nationally by law enforcement in the line of duty—though it is still not clear whether the numbers will be counted in homicide statistics.

39 See, for instance, Victoria Ojea, “Latin America Accounts for More Than 30% of the World’s Homicides,” World Bank, March 5, 2014, accessed December 15, 2016,

40 Author and research assistants conversations with data analysts at the OECD, Small Arms Survey, SIPRI, and UCDP, January 2015–December 2016.

41 Lysova and Shchitov, “What Is Russia’s Real Homicide Rate?”

42 Smalls Arms Survey, Graduate Institute of International and Development Studies, Small Arms Survey 2013: Everyday Dangers (Cambridge, UK: Cambridge University Press, 2013), 36,

43 UNODC, Global Study on Homicide 2013, 60.

44 The UCDP database is also the most useful for longitudinal research, as it has been collected since 1989. It could be combined with the PRIO dataset, which covers 1946–2008, to create longitudinal projections.

45 Krause et al., The Global Burden of Armed Violence 2011, 15.

46 Jorge Restrepo, Michael Spagat, and Juan F. Vargas, “Special Data Feature; The Severity of the Colombian Conflict: Cross-Country Datasets Versus New Micro-Data,” Journal of Peace Research 43, no. 1 (January 2006): 99–115.

47 Particularly problematic is the gap between UCDP’s and CERAC’s death figures in 2002, ostensibly because multiple events—including a pivotal presidential election, a series of local and national elections, and a presidential inauguration—crowded out the space in the international press for issues of violence.

48 Like in many African civil wars, most victims of the protracted civil conflict in the Democratic Republic of the Congo (DRC) (1998–2003) did not die because of the fighting, but because of the higher vulnerability to diseases during a conflict that obstructed access to food and medicine. The International Rescue Commission’s (IRC) early study in November 2002 corroborated this thesis. Yet, a few years later, IRC researchers thought that they had underestimated mortality because of empty households and security conditions, which impeded access to the worst areas. Consequently, they extrapolated findings from war zones to the whole country, speaking about the First African World War—a conflict that, according to IRC, claimed 5.4 million victims in 10 years (1998–2008), totaling 8 percent of the total population of the DRC. These figures were never validated by experts, and the IRC never provided confidence estimates. Moreover, the IRC found child mortality rates that were twice as high as those of different credible estimates. See Marc-Antoine Pérouse de Montclos, Elizabeth Minor, and Samrat Sinha, eds., Violence, Statistics, and the Politics of Accounting for the Dead (Cham, Switzerland: Springer International, 2016).

49 The World Development Report 2011: Conflict, Security, and Development found that 1.5 billion people lived in countries affected by conflict. The homicide statistics are drawn from the UNODC.