By Steven Pinker
When I began to examine trends in warfare while writing The Better Angels of Our Nature, I quickly realized that without a fixed yardstick you can demonstrate any trend you want. If you wish to paint a historical period as violent, just include killings of all kinds, lump together deaths on the battlefield with indirect deaths linked to famines and epidemics, and accept the highest estimates that have been bruited about. Conversely, if you want to portray a period as peaceful, restrict yourself to declared wars between governments, count only the battlefield deaths, and be stringent about which estimates you allow in.
There is only one way to elevate a discussion of war trends above the level of a barroom argument, and that is to consult quantitative datasets assembled by disinterested scholars who define what they count as a "war," stick to one criterion for which deaths to tally, and exhaustively list all wars known to have taken place during a set interval.
Several of these datasets are available, such as those of the Human Security Report Project (HSRP). Most of the scholars who have examined them agree that the decades since 1945 have seen a decline in wars among great powers and developed states (what I call the Long Peace), and the decades since the end of the Cold War have seen a decline in deaths from wars of all kinds (the New Peace). Disagreement persists, to be sure, about the causes of the declines and how long they will last.
John Arquilla's Dec. 2 article "The Big Kill," which claims that war is on the increase, is a good illustration of the pitfalls of cherry-picking conflicts, mashing up categories, and credulously selecting extreme estimates when they help an argument along.
To explain away the decline in deaths on the battlefield, Arquilla recycles the urban legend that the proportion of war deaths suffered by noncombatants has risen from 10 percent during World War I to 90 percent in the wars of today. But this factoid has been debunked by three political scientists (Andrew Mack, Joshua Goldstein, and Adam Roberts), who each discovered that it compares battle deaths in one era with battle deaths, indirect deaths, injuries, and refugees in another.
Arquilla then tries to make the decline go away with a different tactic: he claims it is "skewed" by population growth since 1940. But this defies the mathematical principle that a tendency can be estimated only by dividing the number of occurrences of an event by the number of opportunities for it to occur; it's like choosing to have surgery at a small hospital with a high complication rate rather than at a large hospital with a low rate because the larger hospital has a greater absolute number of complications. A tripling of the world population since 1940 means that there are three times as many people who can start wars and three times as many people who can be killed in them. Even if the number of war deaths had been constant over that interval, it would indicate a sharp decline in the likelihood that a person will be killed in a war. In any case the absolute number has not remained constant. It plunged from 224,000 a year in the 1950s to 85,000 a year in the 1990s and 31,000 a year in the 2000s, so it doesn't even matter whether you look at rates or absolute numbers.
A decline is apparent even if one totals up the sheer number of armed conflicts in the HSRP dataset, ignoring their annual death tolls (which can be as low as 25): the count fell from 53 in 1991 to 30 in 2010, the most recent year available. Arquilla points out that the 1991 peak represents an increase from the 1950s, but that is consistent with the New Peace, which refers to a worldwide decline in all kinds of war after the Cold War ended (as opposed to the post-World War II decline of wars among great powers and developed states).
Arquilla turns to what he calls "big-kill" wars, those that led to a million deaths, and claims they have doubled every half-century since 1800. This trend, too, is being measured with a rubber ruler. For the 20th century, Arquilla includes both wars and genocides (such as those of Stalin and Pol Pot); for the 19th, he includes only wars, omitting the massive death tolls associated with colonial wars, the Atlantic and Mideast slave trades, the man-made famines in British India, and the depredations in the Congo Free State. For that matter, his list of pure wars is selective, omitting Shaka Zulu, the Panthay Rebellion, and the Mahdi revolt, all with death tolls which exceed a million if one includes indirect deaths, as Arquilla does for his 20th century. Indeed, for the second half of the 20th century, Arquilla cites nonstandard highball estimates for the Rwandan genocide and the Mozambique civil war, waffles between absolute and relative death counts, and tries to sneak in near-misses (such as the Iran-Iraq war), none of which he does for the earlier periods.
The casual attitude toward data extends to quotations. Arquilla has me writing that the past 70 years have seen "no spikes, just a couple of ‘blips." In fact, I do refer to "spikes" during this period, while the word "blip" appears nowhere in the book.
Steven Pinker, the author of The Better Angels of Our Nature, is Harvard College Professor and Johnstone Family Professor in the Department of Psychology at Harvard University.
Next: Andrew Mack and Sebastian Merz of the Human Security Research Project also respond.