The GiveWell Blog

Has violence declined, when large-scale atrocities are systematically included?

Note: I wrote the following on my personal time, then cleaned it up slightly for public consumption. This post is not directly related to GiveWell’s work, but we thought readers might find it interesting anyway. It provides a simple supplementary analysis to the argument presented in The Better Angels Of Our Nature that violence has declined over time. I conclude that the book’s big-picture point stands overall, but my analysis complicates the picture, implying that declines in deaths from everyday violence have been significantly (though probably not fully) offset by higher risks of large-scale, extreme sources of violence such as world wars and oppressive regimes.

Thanks to Steven Pinker for reviewing a draft of this post.

One of my favorite nonfiction books is The Better Angels of our Nature by Steven Pinker. It argues that “violence has declined over long stretches of time, and today we may be living in the most peaceable era in our species’ existence … it is an unmistakable development, visible on scales from millennia to years, from the waging of wars to the spanking of children.” For the most part, I think the book is quite convincing on this point.

This post focuses on what I see as the biggest missing piece of its analysis. The major large-scale atrocities of the 20th century – particularly the two World Wars and the regimes of Josef Stalin and Mao Zedong – stand as an obvious challenge to the book’s theme of declining violence over time. Better Angels does address these events, arguing that they are not as historically anomalous as they may seem. However, the book does not give a comprehensive, quantified picture of how recent centuries compare to older ones in terms of total deaths from such large-scale atrocities. It also does not compare the relative death toll of large-scale atrocities to that of other sources of violent deaths it discusses (homicide, witch hunts, executions, etc.) to determine whether the atrocities of the 20th century were violent enough to offset other kinds of improvements. While many critics have highlighted the atrocities of the 20th century, I don’t believe any of them have done this sort of analysis either, with the exception of a partial analysis on the Uncommon Descent blog.

Using some of the data cited in Better Angels, I’ve done a simple analysis to lay out estimated “deaths from major atrocities” for each century, going back to the 5th century BC. I’ve also looked a bit into how these figures would look if we included deaths from everyday violence as well. Having done this, four points stand out:

  • There are two other centuries (13th and 17th) that look to have been at least as bloody as the 20th, though this observation is very sensitive to very imprecise death toll estimates of a very small number of atrocities. (“Bloody” here refers to high violent deaths per capita per year; “atrocity” means an enormous large-scale mass killing, like a war or conquest or democide.) The 13th century death toll comes almost entirely from estimates of the damage done by Genghis Khan, while the 17th century death toll comes mostly from estimates about the fall of the Ming Dynasty. I don’t see a clear trend overall on “death risk from large-scale atrocities” from the 13th through 20th centuries.
  • Prior to the 13th century, it looks like per-century death tolls from the largest atrocities were consistently lower, and I doubt that this is an artifact of the data.
  • Around the 15th century, a documented fall in the homicide rate seems to have started. The homicide rate decline and the rise in deaths from very large-scale atrocities that took place between the 13th and 15th centuries seem to be in the same ballpark as each other, consistent with the idea that violence shifted from individuals to regimes. I would guess that the net effect was a decline in violent deaths, especially when bearing certain issues with the data in mind, but it isn’t clear.
  • Large-scale atrocities account for enormous numbers of violent deaths. While Better Angels describes multiple trends, it does not compare them to each other in an apples-to-apples way. My sense is that large-scale atrocities account for far more violent deaths than most of the other sources of violence the book discusses – so the lack of a positive trend means that the overall global risk of dying from violence may not have improved greatly over time (though it probably has improved). To make this point vivid, the global rate of violent deaths from the “big four” atrocities alone (two World Wars, regimes of Josef Stalin and Mao Zedong) – spread out over the entire 20th century – is ~50 violent deaths per 100,000 people per year; that’s comparable to the very worst national homicide rates seen today, whereas the homicide rate for high-income countries such as the U.S. tends to be less than 1/10 as high. In other words, the two World Wars + Stalin and Mao alone were enough to make the 20th century as a whole more dangerous than homicide makes today’s homicide-heaviest countries, and they were enough to offset the benefit of the European homicide rate decline that Better Angels describes from Medieval times through the Enlightenment.

Looking purely at quantified violent death risk by century, the picture that emerges from these figures is one of falling everyday violence that is significantly (though probably not fully) offset by higher risks of large-scale, extreme sources of violence such as world wars and oppressive regimes. The net impact is probably lower levels of violence, but it’s not entirely clear. The key transition looks like it was around the 13th-15th centuries; I don’t see much reason to think that the Scientific Revolution should bear much blame for rising atrocity tolls (the timing doesn’t work), but the “rule of law and rising power of governments” dynamic that Better Angels credits for much of the decline in everyday violence could be argued to have had a significant cost in terms of rare mass atrocities.

The dynamics of violent deaths discussed above are consistent with a picture of modernization as improvement in everyday conditions, accompanied by larger rare catastrophic events. This picture can be applied to more recent times as well, even as death tolls from atrocities have fallen: everyday peacefulness has continued to improve, but the potential maximum damage of global catastrophic risks (such as power grid failures, natural and engineered pandemics, climate change and artificial intelligence) seems to be on the rise as well. Today, the potential bad news is even more unlikely and infrequent, but potentially even more enormous, than ever before.

Taking a more holistic view – looking at ways in which non-fatal violence has declined, the phenomenon of the “long peace” since the mid-20th century, and other improvements over time – I think it remains the case that the modern world has become greatly less violent, as well as a better place to live in other ways. I do think that the overall point of Better Angels stands with my analysis in mind, though there is some added complexity to it.

Finally, I note that the literature on this topic appears extremely thin. Steven Pinker is not a historian, yet I believe his systematic examination of historical trends in violence is the first of its kind. Many critics of Better Angels highlight the question of how 20th century atrocities compared to past atrocities. However, I’ve seen only one critic who did either of the following: (a) spelled out a more systematic comparison Pinker could have done; (b) performed a rough version of this comparison. This critic was Uncommon Descent, a blog whose main purpose appears to be arguing for Intelligent Design.

Details follow. From this point on I abbreviate Better Angels as BA.

  • I go through BA’s discussion of the major atrocities of the 20th century, and discuss why I believe more analysis is called for. BA’s argument and the need for more analysis
  • I discuss my own rough attempt to make these comparisons, and what it shows: a lack of clear trend in deaths from large-scale atrocities from the 13th through 20th centuries, a smaller death toll from large atrocities but a higher toll from homicides prior to the 15th century, and the relative importance of large-scale atrocities vs. other sources of violence. More: My analysis
  • I reflect on how one should think about long-term historical trends in violence and quality of life with these corrections in mind. Reflections

BA’s argument and the need for more analysis
In Chapter 5, BA confronts the challenge that 20th century atrocities pose to its argument that violence has declined:

The 20th century would seem to be an insult to the very suggestion that violence has declined over the course of history. Commonly labeled the most violent century in history, its first half saw a cascade of world wars, civil wars, and genocides …

BA argues that these atrocities aren’t as much cause for pessimism as they might seem, primarily because:

  1. There were comparably bad large-scale atrocities in the past, and we tend to overlook these because recent times are better-documented and more vivid to us.
  2. The two World Wars could simply have been flukes; the odds that this is true are higher than we’re intuitively inclined to think.
  3. To the extent there were real forces pushing the world toward more damaging wars, they came primarily from (a) improvements in states’ abilities to assemble large armies; (b) the rise “counter-Enlightenment ideologies” (Communism, Fascism, romantic militarism) that have since fallen out of favor.
  4. The second half of the 20th century saw unprecedented peace and stability.

My main concern is with #1: BA doesn’t systematically analyze whether past atrocities were, in fact, as bad in aggregate (in terms of “annual risk of violent death per person”) as 20th-century atrocities. That’s what I’ll mostly discuss here. At the bottom of this post (“Addendum”) I briefly discuss the others.

“Was the 20th century really the worst?”

BA: “When one corrects for the availability bias and the 20th-century population explosion by rooting around in history books and scaling the death tolls by the world population at the time, one comes across many wars and massacres that could hold their head high among 20th-century atrocities.” The main evidence for this statement comes from a list called (Possibly) The Twenty (or so) Worst Things People Have Done to Each Other, which BA has adjusted to consider the “percentage of world population killed” rather than just the “number of people killed” (an adjustment I approve of; more on that below).

My concern with this list is that it lists the 20 worst “things,” where a “thing” can be as specific as “Second World War” (1939-1945) or as broad as “Mideast Slave Trade” (which went on for ~1200 years). Furthermore, 6 of the 21 listed atrocities were entirely within the 20th century, which seems like a lot for a single century. This table can tell us that there were some events arguably as bloody as World War 2, but it doesn’t really undermine the idea that the 20th century was the bloodiest, all things considered.

Multiple critics, including the New Yorker’s Elizabeth Kolbert, have made similar observations about this table, but of the critiques I’ve found, Uncommon Descent gives the most detailed critique:

Numbers of victims aside, the question we really need to answer is: if you were living in Central Asia during the time of the Mongol Empire, what was the chance that you’d eventually be killed in some act of violence? Knowing the percentage of people that were killed in the worst massacre during that period is of little use unless you also know how long the massacre lasted, and whether there were any other major massacres during that century.

…“Imagine,” [Pinker] writes in his response to Kolbert on his FAQ page, “that that slave trade was abolished after a year, or that Genghis Khan was defeated after a month, or that the Holocaust was called off after a week. Would we not judge those events as vastly less violent?” Yes, certainly we would. But we’re not judging events; we’re comparing the overall level of violence during various time periods. For instance, would a typical person living in the 8th century be at a higher or lower risk of being killed in wars and atrocities than an individual living in the 13th? To answer that question, you’d need to look at the annual death rate from wars and atrocities, per 1,000 people, and because the annual death rate might fluctuate markedly from year to year as atrocities come and go, you’d need to average the annual death rate out over the course of a century, in order to get a typical figure that represented your annual risk of dying as a result of an atrocity … I would opt for the annual death rate from all atrocities occurring during a fixed time period (say, the 13th century) as the best measure of how violent that period was.

Uncommon Descent goes on to compare the 20th century to others. It mostly uses the metric of “percentage of all deaths that were caused by violence from atrocities,” which I don’t think is the best metric to use: more recent centuries had fewer deaths from disease, so the percentage of deaths due to violence would have gone up even if violence had remained constant. At one point Uncommon Descent does discuss the “death rate from acts of violence per 100,000 people per year,” which I think is the better metric. However, here it looks only at four centuries (1st, 13th, 17th, 20th). Accordingly, I decided to do similar analysis (using the same data sources as Uncommon Descent) to get the numbers for every century as far back as the data can go.

My analysis
I pulled the figures from the “one hundred deadliest multicides” ranking on page 529 of Atrocities: The 100 Deadliest Episodes in Human History by Matthew White. This is the same source behind the figures that Uncommon Descent primarily uses, and it is essentially the same as the source behind BA’s own list of the worst atrocities over time (same author, but published a year after the table BA cites).

I don’t have the context to say how good this source is, but it seems appropriate to use it because: (a) it is the main source of choice for both BA and the criticism of BA by Uncommon Descent; (b) it doesn’t seem that the author (Matthew White) has taken much of a position one way or another on the hypothesis under debate (whether modern times are more peaceful than the past). The Uncommon Descent post cites figures by others in various places, but for simplicity and consistency I’ve stuck to one source. I also used the same population source as the Uncommon Descent post (it looks reasonable) as a basis for estimating the median population in each century.

My full calculations are here. Below is a summary table – for each century, I give the overall “deaths from atrocities” rate and the 3 worst atrocities that contributed to it.

Century Deaths from atrocities per 100k people per year #1 worst atrocity (% of deaths this century) #2 worst atrocity (% of deaths this century) #3 worst atrocity (% of deaths this century)
5th BC 3.1 Age of Warring States (60%) Second Persian War (40%) N/A (0%)
4th BC 4.3 Age of Warring States (54%) Alexander the Great (46%) N/A (0%)
3rd BC 11.2 Qin Shi Huang Di (34%) Second Punic War (26%) Age of Warring States (16%)
2nd BC 3.7 Roman Slave Wars (52%) Gladiatorial Games (48%) N/A (0%)
1st BC 8.1 Gallic War (30%) Gladiatorial Games (21%) Roman Slave Wars (20%)
1st 35.0 Xin Dynasty (94%) Gladiatorial Games (5%) Roman-Jewish Wars (2%)
2nd 3.7 Gladiatorial Games (43%) The Three Kingdoms of China (42%) Roman-Jewish Wars (15%)
3rd 12.6 The Three Kingdoms of China (88%) Gladiatorial Games (12%) N/A (0%)
4th 3.2 Fall of the Western Roman Empire (53%) Gladiatorial Games (47%) N/A (0%)
5th 18.9 Fall of the Western Roman Empire (97%) Gladiatorial Games (3%) N/A (0%)
6th 2.3 Justinian (90%) Goguryeo-Sui Wars (10%) N/A (0%)
7th 5.2 Mideast Slave Trade (73%) Goguryeo-Sui Wars (27%) N/A (0%)
8th 37.6 An Lushan Rebellion (90%) Mideast Slave Trade (10%) Mayan Collapse (1%)
9th 5.6 Mideast Slave Trade (63%) Mayan Collapse (37%) N/A (0%)
10th 3.6 Mideast Slave Trade (94%) Mayan Collapse (6%) N/A (0%)
11th 3.5 Mideast Slave Trade (95%) Crusades (5%) N/A (0%)
12th 11.2 Fang La Rebellion (40%) Crusades (31%) Mideast Slave Trade (29%)
13th 98.0 Chinggis Khan (90%) Mideast Slave Trade (3%) Crusades (3%)
14th 54.7 Timur (56%) Fall of the Yuan Dynasty (29%) Hundred Years War (7%)
15th 20.8 Timur (29%) Atlantic Slave Trade (22%) Hundred Years War (16%)
16th 30.4 Atlantic Slave Trade (30%) Conquest of the Americas (25%) French Wars of Religion (20%)
17th 104.9 Fall of the Ming Dynasty (48%) Thirty Years War (14%) Atlantic Slave Trade (9%)
18th 33.3 Famines in British India (34%) Atlantic Slave Trade (17%) Conquest of the Americas (14%)
19th 44.6 Taiping Rebellion (36%) Famines in British India (16%) Congo Free State (11%)
20th 81.1 Second World War (33%) Mao Zedong (20%) Joseph Stalin (10%)

My comments:

1. There’s no clear trend from the 13th through 20th centuries in terms of total death rate. The jumpiness of the totals makes it very hard to see any sort of trend, even when aggregating 100-year periods; as discussed below, one every-few-centuries giant atrocity tends to account for a huge chunk of the bad centuries’ death tolls.

2. The death rates from these atrocities (the 100 worst of all time in absolute death rates) are much lower before the 13th century.

3. The 20th century ranks as the 3rd bloodiest, but could easily become the bloodiest if a couple of very uncertain estimates were changed. The 13th century estimate is basically all from Genghis Khan (referred to above as “Chinggis”); Uncommon Descent notes that this figure is heavily disputed by others, and Matthew White (the source of the above figures) concedes a lot of doubt here (see page 123 of Atrocities). The 17th century death toll is more evenly spread out, and probably better documented, but half of it comes from an estimate based on census records of the death toll for the collapse of the Ming Dynasty.

4. BA used a figure of 36 million for the An Lushan rebellion in the 8th century, but Matthew White has since revised his estimate to 13 million. With that revision, the 8th century doesn’t look especially violent compared to later centuries; without it, it would look like the bloodiest century of them all. In my view, this highlights how fragile these figures are, especially for earlier centuries. For context, 36 million is the figure for “missing” people given at the beginning of Matthew White’s chapter on the rebellion; in the version of the book I used (which has the more recent figure), White later states: “What happened to 36 million people? Is a loss of two-thirds in one decade even possible? Perhaps … On the other hand, these numbers could also represent a decline in the central government’s ability to find every taxpayer rather than an actual population collapse … the actual population collapse may have been closer to one-half, or 26 million. For the sake of ranking, however, I’m being conservative and cutting this in half, counting only 13 million dead in the An Lushan Rebellion.”

5. Both BA and White point out that the farther back in time one looks, the more likely it is that there are lots of undocumented atrocities, and thus that the numbers above are underestimates. I think this is probably true. On the other hand, I doubt that any of the undocumented atrocities are big enough to really stack up with the biggest atrocities in this table: as shown above, for any given century there is a very steep dropoff from the 1-3 most damaging currently known atrocities to the rest. So if totals from the past are understated, I’d guess it’s due to very large numbers of relatively small atrocities rather than missing giant atrocities. In other words, if the question is which centuries had the worst truly massive atrocities, it looks pretty robustly true that they were all 13th century or later.

5. The picture gets more complex when one accounts for homicide rates in addition to death rates from large-scale atrocities. For convenience, I’m just going to use the term “rate” instead of saying “per 100k people per year.”

  • Modern homicide rates in the developed world tend to be very small compared to the above figures. The rate is ~5 for the U.S., and much lower for many other rich countries. (Canada 1.6, France 1, Japan 0.3.) Even when looking at poor countries, there are only 24 countries where the rate is over 20, and the very worst countries are around 50 (except for Honduras at 90). The overall global rate is reported as 6.2. (Source: Wikipedia.) For comparison, just the two world wars + Stalin + Mao would be enough for a 50-rate for the entire 20th century worldwide. In other words, deaths from major atrocities are very large compared to deaths from homicide in the modern world, even when one spreads atrocity deaths across the whole world and the whole century. A random person in the 20th century had a very high chance of being killed by one of the two world wars, Josef Stalin’s regime, or Mao Zedong’s regime – much higher than the odds of being killed in an everyday homicide.
  • Homicide rates used to be much higher, high enough to compete with these atrocities. Table 2 of this paper (which I found by tracing BA’s references in another chapter) estimates that Europe had an average homicide rate of 32 in the thirteenth-fourteenth centuries and 41 in the fifteenth century, rapidly declining to 19 in the 16th century and all the way down to 3.2 in the 18th. Imagine adding the homicide rate to the “deaths from giant atrocities” rate in the table above: if one assumed the 13th-15th century European figures are representative of the early world as a whole, earlier centuries would get a substantial bump (in the range of 25-30 points) relative to later ones, which would erase the current apparent peacefulness of these earlier centuries. As argued elsewhere in BA, much earlier civilizations (for example, hunter-gatherers) generally had far higher rates of violent death, and look robustly more violent.
  • In reality, there are good reasons to think a 25-30 point bump isn’t nearly a large enough adjustment.
    • The homicide rates from the above bullet points appear to come primarily from court records. I would guess that there were many unreported/unregistered homicides compared to modern times. Consistent with this, it looks like the Oxford and London figures for homicides are higher than other English figures for homicides (this is from Figure 3-1 in BA); I’d guess this is because the data was better in Oxford and London than elsewhere.
    • The homicide rate increases from the 13th to 14th century, and the 14th century Oxford data point is all the way up at 100. While I don’t know this, my guess is that (similar to what I’ve seen for malaria data) the 14th century showed a rise in homicides because of better data collection; the improved data collection (and thus apparent rise in homicides) directly preceded the actual fall in homicides, as rule of law improved. If I’m right about that, it would imply that centuries before the 14th were even worse, and we just don’t have data for them at all (consistent with the idea that rule of law was weaker then).
    • This idea is supported by data on 1600s America and the 1800s American West, both of which show homicide rates in excess of 100 early on and then a decline (figures 3-15 and 3-16 in BA).
    • Rule of law at that time may have been weaker outside Europe, consistent with the fact that we only have data for Europe.
    • These figures also don’t include other sources of violent deaths, especially smaller-scale wars that are theorized to be underreported in the above table as well as raids, skirmishes, etc.

It’s highly uncertain, but I’d guess that small-scale war and homicide before the 13th century were at least a 100 rate if not much more. That would more than make up for the lower death tolls from giant atrocities.

Even if that’s true, though, it’s interesting that large-scale warfare became more deadly fairly close to the time that smaller-scale violence was on the decline. It arguably implies a transfer of violent power from the individual to the state. And it’s extremely interesting that these large-scale atrocities, even with how infrequent they are, can end up being as significant – for a particular century – as the huge decline from Medieval to modern-day homicide rates. The number of people killed under the world wars, Stalin and Mao is really stunningly large. The total deaths from those 4 wars/democides was around 140 million; the population of the world around 1950 was ~2.5 billion.

6. Much of the rest of BA discusses other positive trends, such as the end of Medieval torture, the decline of capital punishment, the end of burning accused witches at the stake, the decline of lynchings and race riots. I think these are all important trends, but it’s worth noting that the death figures for all of them are comparatively tiny. If you’re looking at broad trends in attitudes and everyday life, BA makes an extremely compelling case (which I haven’t discussed in detail) that the trend has been overwhelmingly positive. However, if you’re just looking at violent deaths, the situation is more ambiguous, because all of these improvements combined are significantly (though probably not fully) offset just by the biggest few atrocities of the 20th century.

BA also discusses trends that don’t relate to violent deaths, but do relate to violence more broadly; more on this immediately below.

I think the point that the world has gotten much better still stands, even if overall violent death rates basically haven’t changed (and my guess is that they have in fact improved, when including homicide and smaller-scale atrocities).

  • BA describes a lot of trends that haven’t primarily been about death at all, including declining child abuse, declining slavery, and the rise of feminism. Arguably any of these alone could be as important as all sources of violent deaths put together, speaking broadly. A bad violent death rate (100) means ~1 in 1000 people per year; issues like feminism profoundly affect a far greater proportion of the population. I can’t think of negative trends that are good candidates for offsetting these positive developments.
  • BA doesn’t even bother to discuss disease, which has been amply covered elsewhere. Disease is a much bigger killer than human-on-human violence and the trends on this front are huge and unambiguously positive. For example, see this chart: at the start of the 20th century, infectious diseases were killing people at a rate of ~800 per 100k people per year, which is nearly 10x the worst of the violent-death rates discussed above. That death rate has collapsed.
  • I think there is genuine reason for optimism that the world has made progress in reducing the likelihood of giant atrocities. The second half of BA’s Chapter 5 (“The Long Peace”) lays this out well. More broadly, the picture one gets from the book is one of progress on a small number of problems at a time; just because periodic large-scale atrocities have been ~steady for 800 years doesn’t mean we won’t see great progress in the future on this front as well.

The dynamics of violent deaths discussed above are consistent with a picture of modernization as improvement in everyday conditions, accompanied by larger rare catastrophic events. This picture can be applied to more recent times as well, even as death tolls from atrocities have fallen: living standards and everyday peacefulness have continued to improve, but the potential maximum damage of global catastrophic risks (such as power grid failures, natural and engineered pandemics, climate change and artificial intelligence) seems to be on the rise as well. Today, the potential bad news is even more unlikely and infrequent, but potentially even more enormous, than ever before.

Finally, I note how thin the relevant literature appears to be. The question of what the overall historical trend in violence looks like – which seems to me like an extremely interesting and important question – has very little discussion and debate around it, and doesn’t seem to be associated with any academic field at all. Broadly speaking, I’ve felt similarly about discussions of other broad historical trends relevant to aggregate well-being.

Addendum: some notes on BA’s other arguments re: 20th century atrocities

In addition to claiming that the 20th century wasn’t necessarily the most violent, BA makes some other claims about 20th century atrocities:

1. BA argues at length that the two World Wars could be statistical flukes, and that such a phenomenon wouldn’t be as unlikely as it intuitively seems. For example:

In making sense of the 20th century, our desire for a good story arc is amplified by two statistical illusions. One is the tendency to see meaningful clusters in randomly spaced events . Another is the bell-curve mindset that makes extreme values seem astronomically unlikely, so when we come across an extreme event, we reason there must have been extraordinary design behind it. That mindset makes it difficult to accept that the worst two events in recent history, though unlikely, were not astronomically unlikely . Even if the odds had been increased by the tensions of the times, the wars did not have to start. And once they did, they had a constant chance of escalating to greater deadliness, no matter how deadly they already were. The two world wars were, in a sense, horrifically unlucky samples from a statistical distribution that stretches across a vast range of destruction.”

I think BA is right to point out how our intuitions about randomness can be flawed, but this is still a long way from establishing that the 20th century’s atrocities were actually flukes. Most of BA’s analysis concerns the frequency and duration of wars – not their total death tolls. BA concedes that wars seem to have become more destructive even as they’ve become less common. In addition, the two world wars weren’t the only remarkable atrocities of the 20th century: BA’s main source attributes similar death tolls to the regimes of Josef Stalin and Mao Zedong. The models BA proposes encourage the reader to think of wars as primarily driven by randomness, but these models aren’t empirically defended (when it comes to death tolls as opposed to duration/frequency, and when it comes to arguing that death tolls are random and not merely power-law-distributed), and they don’t address non-war atrocities.

2. BA offers two possible explanations for why wars might have become systematically more damaging over time.
One is improvements in states’ abilities to build militaries: “The greater lethality of the wars that did take place was the result of a development called the military revolution. States got serious about war. This was partly a matter of improved weaponry, especially cannons and guns, but it was more a matter of recruiting greater numbers of people to kill and be killed.”

I find this explanation plausible, but as stated it only explains war, not other atrocities such as Stalin’s and Mao’s. One might extend the explanation more broadly to say that states became more powerful and more adept at controlling their populations.

The other offered explanation is the rise of “counter-Enlightenment ideologies” – “conservatism, nationalism and utopian ideologies” including Communism and Fascism. BA observes that these ideologies have since been on the decline, and I think that’s true, though limited consolation: as shown in the table above, there can be several relatively peaceful centuries followed by a really catastrophic one, so we’re pretty far from being able to say that harmful ideologies are in permanent as opposed to temporary decline.

3. BA gives an extended discussion of the relatively low death toll from large-scale atrocities since the mid-20th century, a time period it calls the “Long Peace.” I see no problems with BA’s claims here and think that the level of peacefulness over the last ~70 years is incredible and historically special. But again, it’s a fairly short time period in the scheme of things when examining very infrequent, extremely damaging atrocities.


  • David Roodman on July 8, 2015 at 3:21 pm said:

    Holden, what confuses me in all this is the semantic relationship between “violence” and “famine” (which I guess are both examples of “atrocity”). Pinker’s thesis is the decline of violence. I would think that the everyday interpretation of “violence” would exclude the great famines under Stalin and Mao, which both arose from sudden changes in the economic rules. I recognize that there are gradations between “violence” and “famine”—from war deaths, to wars that destroy crops and cause hunger as a byproduct, to gun-point enforcement of economic policies clearly intended to cause starvation, to massive famines that are unintended yet nevertheless ignored by an all-powerful, violence-monopolizing state. How much of your difference from Pinker is in the semantic expansion of the term “violence,” or perhaps equivalently, the shift from “violence” to “atrocity.”

  • 27chaos on July 8, 2015 at 3:41 pm said:

    Excellent. Now all I need is for someone to articulate some kind of justification for the annoyingly vague intuition haunting my thoughts that to measure violence in terms of the percentage of the overall population is rather misleading.

  • Colin Rust on July 21, 2015 at 10:44 am said:

    I haven’t read Better Angels, though it’s one (of a long list) I’ve been “meaning” too. My sense without having read the book was that Pinker was arguing not only that non-state violence has declined dramatically, but that paradoxically despite the horrors of the twentieth century (WWI, WWII, etc.), total violence is declining. Maybe that’s idiosyncratic to me, but I suspect many people who were casually aware of the thesis had the same overstated understanding. So this post is a helpful corrective: we simply don’t really know if in the long arc of history total violence is declining.

    (Also: what David said. I share the confusion.)

  • Colin Rust on July 21, 2015 at 12:45 pm said:


    intuition haunting my thoughts that to measure violence in terms of the percentage of the overall population is rather misleading

    For combatant deaths, you might have a point: Most combatants are relatively young men. So if a man dies at 80 rather than 60 from natural causes (in both cases), that shouldn’t really count as progress on combatant deaths. But I’d argue it should count as progress on civilian deaths: if he died at 80, that’s an extra 20 years he wasn’t the victim of violence.

    And of course, if you measure death rates as a percentage of deaths, high infant and childhood mortality from natural causes makes violent death rates look lower, which makes no sense. So even for combatant deaths alone, using the percentage of deaths will tend to understate progress.

    Overall, if you have to pick a single, simple measure, I’d go with the violent death rate as a percentage of population, as Holden has done.

  • Holden on July 25, 2015 at 9:39 am said:

    Colin and David, I agree that the distinction between e.g. violence and famine can get fuzzy. Dr. Pinker and I have both attempted to outsource the question of how to draw those distinctions to the same source (Matthew White’s book). I only gave White’s book a cursory review and can’t vouch for the way it did this, but the key point seems to me to be that Dr. Pinker and I are using the same figures for the same atrocities.

    27chaos, I agree with Colin’s response.

  • Filipe on July 29, 2015 at 7:37 pm said:

    I haven’t fully developed this thought, but it seems to me it might make sense, so I’ll mention it here and you can see if it seems to lead somewhere.

    It seems to me that this type of analysis (both yours and Dr. Pinker’s) can be hindered by the fact that we are separating the counts into arbitrary discrete time units, that is centuries. This might not have been a problem for the earlier types of catastrophes and conflicts, because they used to take very long, but it might become a problem as we move closer to the present.
    For example, if we had began counting centuries on what is currently called the year 50, we would have the previous century spanning the years 1850-1950, and the current one spanning 1950-2050. In this case, we would have split the wars to one side and Mao to the other, which would probably create two “not as bad” centuries. Or maybe if we changed the size of our measurement unit to, say, 200 or 500 years. This would “dilute” some catastrophes that are more concentrated in time, but would not affect those that span many centuries. The opposite is also true.

    Therefore, more obviously, this implies that it does not make much sense to talk about “best” or “worst” centuries, since we could be just getting lucky doing some sort of “gerrymandering” with the conflicts.

    But I also believe it shows that we should be careful when analyzing these trends only with arbitrarily discretized data. One solution would be to have a continuous count of the death tolls. This is obviously practically (if not also theoretically) impossible. However, one could argue that some time steps would be short enough to be negligible; the question is what should this time length be. I doubt a century would do it, maybe a year or a decade? It is unlikely, however, that we have data with this level of precision as well. Another solution that comes to mind is taking a fixed time-span (a century for example) and trying to position the “year zero” in different places (as I did in the example in the beginning of this post). One could then maybe measure how much certain measures vary only as a function of changing the alignment of the time intervals.

    At last, a reanalysis that would seem simple to do is to just get the century data and try to pack it into different time spans, such as 200, 300, 400 years, and see how this affects the trends in how catastrophes are progressing, and which time-period seems “worst” or “best”.

    As I said in the beginning, these are all ideas fresh out of reading the post, so they are not fully developed and might have issues. But I believe they are sufficient for at least initiating a discussion. Feel free to respond to and criticize them.

    (Also, has there been any interesting response by Dr. Pinker or anyone else to this blog post?)

  • Holden on July 30, 2015 at 2:50 pm said:

    Thanks for the comment, Filipe. I tried to examine this idea by creating a “rolling chart” that graphs the 100 years starting 1500, then the 100 years starting in 1501, etc. all as one continuous line – see

    My takeaway was that it was basically the same picture you get from the table.

    I’m not aware of any responses to this post other than here.

  • Filipe on August 13, 2015 at 5:51 pm said:

    Thanks for your reply, Holden. I played a bit with your spreadsheet as well and the results do seem to be similar to the ones in the post, regardless of how you look at the data, as you said.

    I find your criticisms very interesting, as much of the rebuttals to BA I had read so far were pretty poorly constructed. This seems like a sound statistical analysis, and that’s why I mentioned I would be interested in seeing what Pinker himself would have to say about it (for the sake of good dialectics haha). You mentioned he reviewed a draft of this, do you know what were his reactions/thoughts? I would really like to see more debate on your observations from people knowledgeable on the subject (such as yourself, Mr. Pinker, etc), as they do seem to raise some interesting questions.

Comments are closed.