The Better Angels of Our Nature: Why Violence Has Declined
THE TRAJECTORY OF WAR IN THE REST OF THE WORLD
What was the rest of the world doing during the six hundred years when the great powers and European states went through their Ages of Dynasties, Religions, Sovereignty, Nationalism, and Ideology; were racked by two world wars; and then fell into a long peace? Unfortunately the Eurocentric bias of the historical record makes it impossible to trace out curves with any confidence. Before the advent of colonialism, large swaths of Africa, the Americas, and Asia were host to predation, feuding, and slave-raiding that slunk beneath the military horizon or fell in the forest without any historian hearing them. Colonialism itself was implemented in many imperial wars that the great powers waged to acquire their colonies, suppress revolts, and fend off rivals. Throughout this era there were plenty of wars. For the period from 1400 through 1938, Brecke’s Conflict Catalog lists 276 violent conflicts in the Americas, 283 in North Africa and the Middle East, 586 in sub-Saharan Africa, 313 in Central and South Asia, and 657 in East and Southeast Asia.11 Historical myopia prevents us from plotting trustworthy trends in the frequency or deadliness of the wars, but we saw in the preceding chapter that many were devastating. They included civil and interstate wars that were proportionally (and in some cases absolutely) more lethal than anything taking place in Europe, such as the American Civil War, the Taiping Rebellion in China, the War of the Triple Alliance in South America, and the conquests of Shaka Zulu in southern Africa.
In 1946, just when Europe, the great powers, and the developed world started racking up their peaceful zeroes, the historical record for the world as a whole snaps into focus. That is the first year covered in a meticulous dataset compiled by Bethany Lacina, Nils Petter Gleditsch, and their colleagues at the Peace Research Institute of Oslo called the PRIO Battle Deaths Dataset.12 The dataset includes every known armed conflict that killed as few as twenty-five people in a year. The conflicts that rise to the level of a thousand deaths a year are promoted to “wars,” matching the definition used in the Correlates of War Project, but they are otherwise given no special treatment. (I will continue to use the word war in its nontechnical sense to refer to armed conflicts of all sizes.)
The PRIO researchers aim for criteria that are as reliable as possible, so that analysts can compare regions of the world and plot trends over time using a fixed yardstick. Without strict criteria—when analysts use direct battlefield deaths for some wars but include indirect deaths from epidemics and famines in others, or when they count army-against-army wars in some regions but throw in genocides in others—comparisons are meaningless and are too easily used as propaganda for one cause or another. The PRIO analysts comb through histories, media stories, and reports from government and human rights organizations to tally deaths from war as objectively as possible. The counts are conservative; indeed, they are certainly underestimates, because they omit all deaths that are merely conjectured or whose causes cannot be ascertained with confidence. Similar criteria, and overlapping data, are used in other conflict datasets, including those of the Uppsala Conflict Data Project (UCDP), whose data begin in 1989; the Stockholm International Peace Research Institute (SIPRI), which uses adjusted UCDP data; and the Human Security Report Project (HSRP), which draws on both the PRIO and UCDP datasets.13
Like Lewis Richardson, the new conflict-counters have to deal with failures of thinginess, and so they divide the conflicts into categories using obsessivecompulsive criteria.14 The first cut distinguishes three kinds of mass violence that vary in their causes and, just as importantly, in their countability. The concept of “war” (and its milder version, “armed conflict”) applies most naturally to multiple killing that is organized and socially legitimated. That invites a definition in which a “war” must have a government on at least one side, and the two sides must be contesting some identifiable resource, usually a territory or the machinery of government. To make this clear, the datasets call wars in this narrow sense “state-based armed conflicts,” and they are the only conflicts for which data go all the way back to 1946.
The second category embraces “nonstate” or “intercommunal” conflict, and it pits warlords, militias, or paramilitaries (often aligned with ethnic or religious groups) against each other.
The third category has the clinical name “one-sided violence” and embraces genocides, politicides, and other massacres of unarmed civilians, whether perpetrated by governments or by militias. The exclusion of one-sided violence from the PRIO dataset is in part a tactical choice to divide violence into categories with different causes, but it is also a legacy of historians’ long-standing fascination with war at the expense of genocide, which only recently has been recognized as more destructive of human life.15 Rudolph Rummel, the political scientist Barbara Harff, and the UCDP have collected datasets of genocides, which we will examine in the next section.16
The first of the three categories, state-based conflicts, is then subdivided according to whom the government is fighting. The prototypical war is the interstate war, which pits two states against each other, such as the Iran-Iraq War of 1980–88. Then there are extrastate or extrasystemic wars, in which a government wages war on an entity outside its borders that is not a recognized state. These are generally imperial wars, in which a state fights indigenous forces to acquire a colony, or colonial wars, in which it fights to retain one, such as France in Algeria from 1954 to 1962.
Finally there are civil or intrastate wars, in which the government fights an insurrection, rebellion, or secessionist movement. These are further subdivided into civil wars that are completely internal (such as the recently concluded war in Sri Lanka between the government and the Tamil Tigers) and the internationalized intrastate wars in which a foreign army intervenes, usually to help a government defend itself against the rebels. The wars in Afghanistan and Iraq both began as interstate conflicts (the United States and its allies against Taliban-controlled Afghanistan, and the United States and its allies against Baathist-controlled Iraq), but as soon as the governments were toppled and the invading armies remained in the country to support the new governments against insurgencies, the conflicts were reclassified as internationalized intrastate conflicts.
Now there’s the question of which deaths to count. The PRIO and UCDP datasets tally direct or battle-related deaths—the people who are shot, stabbed, clubbed, gassed, blown up, drowned, or deliberately starved as part of a contest in which the perpetrators themselves have to worry about getting hurt.17 The victims may be soldiers, or they may be civilians who were caught in the crossfire or killed in “collateral damage.” The battle-related death statistics exclude indirect deaths arising from disease, starvation, stress, and the breakdown of infrastructure. When indirect deaths are added to direct deaths to yield the entire toll attributable to the war, the sum may be called excess deaths.
Why do the datasets exclude indirect deaths? It’s not to write these kinds of suffering out of the history books, but because direct deaths are the only ones that can be counted with confidence. Direct deaths also conform to our basic intuition of what it means for an agent to be responsible for an effect that it causes, namely that the agent foresees the effect, intends for it to happen, and makes it happen via a chain of events that does not have too many uncontrollable intervening links.18 The problem with estimating indirect deaths is that itit requires us to undertake the philosophical exercise of simulating in our imagination the possible world in which the war didn’t occur and estimating the number of deaths that took place in that world, which then is used as a baseline. And that requires something close to omniscience. Would a postwar famine have taken place even if the war had not broken out because of the ineptitude of the overthrown government? What if there was a drought that year—should the famine deaths be blamed on the war or on the weather? If the rate of death from hunger was going down in the years before a war, should we assume that it would have declined even further if the war hadn’t occurred, or should we freeze it at its level in the last year befor
e the war? If Saddam Hussein had not been deposed, would he have gone on to kill more political enemies than the number of people who died in the intercommunal violence following his defeat? Should we add the 40 to 50 million victims of the 1918 influenza pandemic to the 15 million who were killed in World War I, because the flu virus would not have evolved its virulence if the war hadn’t packed so many troops into trenches?19 Estimating indirect deaths requires answering these sorts of questions in a consistent way for hundreds of conflicts, an impossible undertaking.
Wars, in general, tend to be destructive in many ways at once, and the ones that kill more people on the battlefield also generally lead to more deaths from famine, disease, and the disruption of services. To the extent that they do, trends in battle deaths can serve as a proxy for trends in overall destructiveness. But they don’t in every case, and later in the chapter we will ask whether developing nations, with their fragile infrastructure, are more vulnerable to knock-on effects than advanced nations, and whether this ratio has changed over time, making battle deaths a misleading index of trends in the human toll of conflict.
Now that we have the precision instrument of conflict datasets, what do they tell us about the recent trajectory of war in the entire world? Let’s begin with the bird’s-eye view of the 20th century in figure 6–1. The viewing was arranged by Lacina, Gleditsch, and Russett, who retrofitted numbers from the Correlates of War Project from 1900 to 1945 to the PRIO dataset from 1946 to 2005, and divided the numbers by the size of the world’s population, to yield an individual’s risk of dying in battle over the century.
The graph reminds us of the freakish destructiveness of the two world wars. They were not steps on a staircase, or swings of a pendulum, but massive spikes poking through a bumpy lowland. The drop-off in the rate of battle deaths after the early 1940s (peaking at 300 per 100,000 people per year) has been precipitous; the world has seen nothing close to that level since.
Eagle-eyed readers will spot a decline within the decline, from some small peaks in the immediate postwar decade to the low-lying flats of today. Let’s zoom in on this trend in figure 6–2, while also subdividing the battle deaths according to the type of war that caused them.
FIGURE 6–1. Rate of battle deaths in state-based armed conflicts, 1900–2005
Source: Graph from Russett, 2008, based on Lacina, Gleditsch, & Russett, 2006.
FIGURE 6–2. Rate of battle deaths in state-based armed conflicts, 1946–2008
Civilian and military battle deaths in state-based armed conflicts, divided by world population. Sources: UCDP/PRIO Armed Conflict Dataset; see Human Security Report Project, 2007, based on data from Lacina & Gleditsch, 2005, updated in 2010 by Tara Cooper. “Best” estimate used when available; otherwise the geometric mean of the “High” and “Low” estimates is used. World population figures from U.S. Census Bureau, 2010c. Population data for 1946–49 were taken from McEvedy & Jones, 1978, and multiplied by 1.01 to make them commensurable with the rest.
This is an area graph, in which the thickness of each layer represents the rate of battle deaths for a particular kind of state-based conflict, and the height of the stack of layers represents the rate for all the conflicts combined. First take a moment to behold the overall shape of the trajectory. Even after we have lopped off the massive ski-jump from World War II, no one could miss another steep falloff in the rate of getting killed in battle that has taken place over the past sixty years, with a paper-thin laminate for the first decade of the 21st century at the end. This period, even with thirty-one ongoing conflicts in that mid-decade (including Iraq, Afghanistan, Chad, Sri Lanka, and Sudan), enjoyed an astoundingly low rate of battle deaths: around 0.5 per 100,000 per year, falling below the homicide rate of even the world’s most peaceable societies.20 The figures, granted, are lowballs, since they include only reported battle deaths, but that is true for the entire time series. And even if we were to multiply the recent figures by five, they would sit well below the world’s overall homicide rate of 8.8 per 100,000 per year.21 In absolute numbers, annual battle deaths have fallen by more than 90 percent, from around half a million per year in the late 1940s to around thirty thousand a year in the early 2000s. So believe it or not, from a global, historical, and quantitative perspective, the dream of the 1960s folk songs has come true: the world has (almost) put an end to war.
Let’s take our jaws off the table and look more closely at what happened category by category. We can start with the pale patch at the bottom left, which represents a kind of war that has vanished off the face of the earth: the extrastate or colonial war. Wars in which a great power tried to hang on to a colony could be extremely destructive, such as France’s attempts to retain Vietnam between 1946 and 1954 (375,000 battle deaths) and Algeria between 1954 and 1962 (182,500 battle deaths).22 After what has been called “the greatest transfer of power in world history,” this kind of war no longer exists.
Now look at the black layer, for wars between states. It is bunched up in three large patches, each thinner than its predecessor: one which includes the Korean War from 1950 to 1953 (a million battle deaths spread over four years), one which includes the Vietnam War from 1962 to 1975 (1.6 million battle deaths spread over fourteen years), and one which includes the Iran-Iraq War (645,000 battle deaths spread over nine years).23 Since the end of the Cold War, there have been only two significant interstate wars: the first Gulf War, with 23,000 battle deaths, and the 1998–2000 war between Eritrea and Ethiopia, with 50,000. By the first decade of the new millennium, interstate wars had become few in number, mostly brief, and relatively low in battle deaths (India-Pakistan and Eritrea-Djibouti, neither of which counts as a “war” in the technical sense of having a thousand deaths a year, and the quick overthrow of the regimes in Afghanistan and Iraq). In 2004, 2005, 2006, 2007, and 2009, there were no interstate conflicts at all.
The Long Peace—an avoidance of major war among great powers and developed states—is spreading to the rest of the world. Aspiring great powers no longer feel the need to establish their greatness by acquiring an empire or picking on weaker countries: China boasts of its “peaceful rise” and Turkey of a policy it calls “zero problems with neighbors”; Brazil’s foreign minister recently crowed, “I don’t think there are many countries that can boast that they have 10 neighbors and haven’t had a war in the last 140 years.”24 And East Asia seems to be catching Europe’s distaste for war. Though in the decades after World War II it was the world’s bloodiest region, with ruinous wars in China, Korea, and Indochina, from 1980 to 1993 the number of conflicts and their toll in battle deaths plummeted, and they have remained at historically unprecedented lows ever since.25
As interstate war was being snuffed out, though, civil wars began to flare up. We see this in the enormous dark gray wedge at the left of figure 6–2, mainly representing the 1.2 million battle deaths in the 1946–50 Chinese Civil War, and a fat lighter gray bulge at the top of the stack in the 1980s, which contains the 435,000 battle deaths in the Soviet Union–bolstered civil war in Afghanistan. And snaking its way through the 1980s and 1990s, we find a continuation of the dark gray layer with a mass of smaller civil wars in countries such as Angola, Bosnia, Chechnya, Croatia, El Salvador, Ethiopia, Guatemala, Iraq, Liberia, Mozambique, Somalia, Sudan, Tajikistan, and Uganda. But even this slice tapers down in the 2000s to a slender layer.
To get a clearer picture of what the numbers here are telling us, it helps to disaggregate the death tolls into the two main dimensions of war: how many there were, and how lethal each kind was. Figure 6–3 shows the raw totals of the conflicts of each kind, disregarding their death tolls, which, recall, can be as low as twenty-five. As colonial wars disappeared and interstate wars were petering out, internationalized civil wars vanished for a brief instant at the end of the Cold War, when the Soviet Union and the United States stopped supporting their client states, and then reappeared with the policing wars in Yugoslavia, Afghanistan, Iraq, and elsewhere. But the big news was an expl
osion in the number of purely internal civil wars that began around 1960, peaked in the early 1990s, and then declined through 2003, followed by a slight bounce.
FIGURE 6–3. Number of state-based armed conflicts, 1946–2009
Sources: UCDP/PRIO Armed Conflict Dataset; see Human Security Report Project, 2007, based on data from Lacina & Gleditsch, 2005, updated in 2010 by Tara Cooper.
Why do the sizes of the patches look so different in the two graphs? It’s because of the power-law distribution for wars, in which a small number of wars in the tail of the L-shaped distribution are responsible for a large percentage of the deaths. More than half of the 9.4 million battle deaths in the 260 conflicts between 1946 and 2008 come from just five wars, three of them between states (Korea, Vietnam, Iran-Iraq) and two within states (China and Afghanistan). Most of the downward trend in the death toll came from reeling in that thick tail, leaving fewer of the really destructive wars.
In addition to the differences in the contributions of wars of different sizes to the overall death tolls, there are substantial differences in the contributions of the wars of different kinds. Figure 6–4 shows the second dimension of war, how many people an average war kills.
FIGURE 6–4. Deadliness of interstate and civil wars, 1950–2005
Sources: UCDP/PRIO Armed Conflict Dataset, Lacina & Gleditsch, 2005; adapted by the Human Security Report Project; Human Security Centre, 2006.
Until recently the most lethal kind of war by far was the interstate war. There is nothing like a pair of Leviathans amassing cannon fodder, lobbing artillery shells, and pulverizing each other’s cities to rack up truly impressive body counts. A distant second and third are the wars in which a Leviathan projects its might in some other part of the world to prop up a beleaguered government or keep a grip on its colonies. Pulling up the rear are the internal civil wars, which, at least since the Chinese slaughterhouse in the late 1940s, have been far less deadly. When a gang of Kalashnikov-toting rebels harasses the government in a small country that the great powers don’t care about, the damage they do is more limited. And even these fatality rates have decreased over the past quarter-century.26 In 1950 the average armed conflict (of any kind) killed thirty-three thousand people; in 2007 it killed less than a thousand. 27