Friday, April 11, 2014

Does Steven Pinker Distort the Data for Declining Violence?

Steven Pinker's Better Angels of Our Nature has over 115 figures--an average of one for every 6 pages of text.  Many of these figures are visual presentations of data to support his argument for a historical trend towards declining violence from the Stone Age to the present.  These figures are based on data found in thousands of cited sources.  This is one of his most impressive rhetorical techniques for persuading his readers that his reasoning is based on a meticulous statistical analysis of data.

Most readers will not take the trouble to read the sources for each figure to see whether Pinker is being accurate in his presentation of the data.  But some of his critics have done this for some of the figures, and they are accusing Pinker of manipulating the data to make it look more supportive of his argument that it really is.  Having looked into this myself, I think this is a fair criticism, although it's not fatal to his argument.  If Pinker had been totally honest about the gaps and uncertainties in the data, he could still have made a plausible argument for his conclusions.

Here I'll point to two examples: the table on page 195 that ranks the greatest atrocities in human history and Figure 2-2 on page 49 that shows the "percentage of deaths in warfare in nonstate and state societies."

Pinker identifies his table of the greatest atrocities as taken from Matthew White's list of "(Possibly) The Twenty (or so) Worst Things People Have Done to Each Other."  White identifies himself as an "atrocitologist" who for many years has maintained a website where he compiles records of the greatest atrocities in human history based on his estimates of violent deaths drawn from historical sources.  This work has been published as a book--The Great Big Book of Horrible Things: The Definitive Chronicle of History's 100 Worst Atrocities (Norton, 2012)--with a Foreword by Pinker.

In Pinker's table, he says that he's following White in ranking the 21 worst atrocities.  Number 1 is the Second World War with a death toll of 55 million.  Number 2 is Mao Zedong who was responsible for a death toll of 40 million (mostly through a government caused famine).  Number 3 is the Mongol Conquests of the 13th century with a death toll of 40 million.  Number 4 is the An Lushan Revolt in China in the 8th century with a death toll of 36 million.

This seems to confirm the common belief that the 20th century was the bloodiest in human history, especially when one notices that 5 of the top 21 atrocities were in the 20th century; and this would seem to refute Pinker's theory of a historical trend of declining violence.  In fact, White concludes his book by identifying the bloody events of the first half of the 20th century as the "Hemoclysm" (Greek for "blood flood"), which he sees as a series of interconnected events stretching from the First World War to the deaths of Hitler, Stalin, and Mao.  The collective death toll here would be 150 million, which would make it the other Number 1 atrocity of human history.

If Pinker is to save his theory of declining violence, he must reinterpret White's account of the historical record of violence culminating in the Hemoclysm of the 20th century.  Pinker does this with three arguments.

His first argument is that we must adjust White's numbers to overcome the illusion that the 20th century was much bloodier than past centuries.  Pinker adjusts the absolute numbers of violent deaths, and he also asks us to look at the relative numbers, calculated as a proportion of the populations.  Once these adjustments are made, Pinker can conclude that "the worst atrocity of all time was the An Lushan Revolt and Civil War, an eight-year rebellion during China's Tang Dynasty that, according to censuses, resulted in the loss of two-thirds of the empire's population, a sixth of the world's population at the time" (194).  In an endnote to this sentence, Pinker writes: "An Lushan Revolt.  White notes that the figure is controversial.  Some historians attribute it to migration or the breakdown of the census; others treat it as credible, because subsistence farmers would have been highly vulnerable to a disruption of irrigation infrastructure" (707, n. 13).

A reader who notices this endnote might become curious about what White has said about these controversial calculations concerning the An Lushan Rebellion.  A reader who looks at White's book will notice that he revises the estimates of violent deaths--moving from 36 million to 26 million to a final estimate of 13 million.  With the lower estimate, the An Lushan Rebellion ranks Number 13 on the list of atrocities, not Number 4 as Pinker has it, because Pinker accepts the highest estimate of 36 million (White 88-93, 529).

Historians know that the Chinese census recorded a population of 52,880,488 in the year 754, and then after ten years of civil war, the census of 764 recorded a population of 16,900,000.  This would suggest that 36 million people died in the war, which would be two-thirds of the entire population of China.  Pinker accepts these numbers, which allows him to rank the An Lushan Revolt as Number 4 on the list of atrocities.

But White indicates that most historians doubt the accuracy of these numbers, because they suspect that the chaos created by the war had impeded the ability of the Chinese census takers to find every taxpayer.  He cites five historians who commented on the census numbers.  He reports that two of them express "major doubt" about the census numbers, one expresses "slight doubt," one expresses "apparent acceptance," and one expresses "acceptance."  But a reader who checks these sources will see that the doubt is even greater than is reported by White.  The historian whom White identifies as expressing "slight doubt"--Peter Stearns--actually says that the population census of 16,900,000 was "certainly too low," which surely shows "major doubt."  And the historian whom White identifies as expressing "acceptance"--Peter Turchin--actually says there is "a certain degree of controversy among the experts" about the numbers, which surely indicates "slight doubt."

Not only does Pinker depart from White in accepting the 36 million estimate of violent deaths, Pinker also insists that death tolls should be adjusted as a proportion of the populations, because this allows us to judge the relative risk of being killed at different points in history.  The 55 million deaths in World War Two is higher than the 36 million in the An Lushan Revolt, but then the world population at the middle of the 20th century was much larger than that in the 8th century.  So if 36 million violent deaths was a sixth of the world's population in the 8th century, this would be the equivalent of 429 million violent deaths in the middle of the 20th century, which would raise the An Lushan Revolt to Number 1 on the list of atrocities; and World War Two would drop to Number 9 on the list.  White does not adjust the ranking in this way.

Pinker's second argument for why the Hemoclysm of the 20th century does not refute his theory of declining violence is that the causes of war can be so contingent that we can have something like World War Two erupt by chance without altering the otherwise declining trend of violence.  We can thus see World War Two as "an isolated peak in a declining sawtooth--the last gasp in a long slide of major war into historical obsolescence" (192).  If wars start and stop at random, then the accidents of history and the peculiarities of particular individuals can result in cataclysmic spasms of violence (200-222). 

In 1999, there was a lot of discussion about who should be considered the Most Important Person of the 20th Century.  White's answer was Gavrilo Princip.  And who was he?  He was the 19-year-old Serbian terrorist who assassinated Archduke Franz Ferdinand of Austria-Hungary.  This was a lucky accident for Princip.  If the archduke's driver had not made a wrong turn in Sarajevo, this would not have happened, and it's likely that World War One would not have happened, and this would not have set off the series of events leading to Lenin, Stalin, Mao, Hitler, World War Two, and the Cold War (White, 344-58; Pinker, 207-10, 262-63).

In the 80-year-long Hemoclysm sparked by Princip's bullets, three individuals--Stalin, Hitler, and Mao--were responsible for most of the violent deaths.  The communist regimes were responsible for 70 million deaths, which would justify ranking communism as the Number 1 atrocity--even greater than World War Two--except that it's hard to think of the whole communist movement as one event (White, 453-57).  Notice that what we see here is that most of the violence of the 20th century has been caused by illiberal ideology--Nazism and communism.

This supports Pinker's third argument for why the violence of the 20th century does not deny his theory of history.  The historical trend towards decreasing violence and increasing liberty depends on the spreading influence of classical liberal culture based on the principle that violence is never justified except in defense against violence.  That illiberal regimes have been the primary sources of violence in the 20th century confirms Pinker's argument. 

Because of the contingency of history, we can never be sure that illiberal leaders will not arise and cause great disasters.  Some day, we might see another Stalin, or Mao, or Pol Pot.  And that's why Pinker is clear in stating that there is no inevitability in the historical trend towards declining violence, because it could be reversed by illiberal turns (xxi, 361-77, 480).  But insofar as classical liberal ideas and norms spread around the world, they can increase the odds in favor of declining violence, which is what has happened since World War Two.

In my next post, I'll turn to Figure 2-2.

My first long series of posts on Pinker's Better Angels was written from October, 2011, to January, 2012.

No comments: