Falling everyday violence, bigger wars and atrocities: how do they net out?

Adapted from an old GiveWell blog post, which has more detail.

The Better Angels of our Nature 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.”

The book gives many examples of ways in which violence has declined. It looks at trends in homicides, in executions, in bans on slavery and torture, in corporal punishment at school, in hunting, even in apologies by political and religious leaders.

What it doesn't provide is a systematic, long-run examination of trends in violence: a single, consistent set of key indicators reported in the same way and tracked over long time periods (in particular, going back further than a few hundred years). Without this, it's hard to know whether "violence has declined" is a consistent, across-the-board phenomenon, or just a description of the particular measures and time periods that the book chooses to focus on.

And my best attempt to look at a single systematic indicator (the violent death rate) presents a muddier picture about long-run trends in violence, because it's plausible (though not clear) that the increasing severity, over time, of the largest wars and atrocities has been big enough to mostly (if not entirely) offset a lot of the improvements on everyday (and other) dimensions of violence.

Below, I will:

• Outline what I see as the "missing piece" of the existing analysis on trends in violence over time: the question of what happened to the violent death rate, accounting both for everyday violence (e.g., homicides) and large-scale wars and atrocities.
• Present some rough analysis that I did to address the "missing piece" and thus examine trends in overall violent death rates over the long run. In brief:
• Deaths from wars and atrocities appear to have gone up significantly starting around the 13th century, which is pretty close in time to the first documented signs of falling homicide rates.
• When accounting for deaths from wars and atrocities, it's not clear whether the violent death rate declined between 500 BCE and the mid-20th century. (And that's as far back as the wars/atrocities data goes.)
• The last 50 years or so have had low death tolls from both homicides and wars/atrocities.
• One spin on these observations - a bit on the provocative side, but a simple take that has a number of implications I agree with - might be: "Over time, increasing state power and order has led to falling everyday violence, but offsetting (at least partially, maybe fully) risk of increasingly infrequent, extreme violence. You could think of these twin trends as continuing even now, as observed violence levels are low across the board, but global catastrophic risks may be at all-time highs."
• To be clear, today's violent death rates look solidly lower than those of very early humans. So it seems like there must have been a decline in violence at some point - it just remains very unclear when that decline was and how steady it was. (This is similar to my general take quality of life over the very long run: we know the most recent period has seen improvement, but the picture is murkier before that.)
• Estimates of death tolls from wars and atrocities seem highly debatable. Some of the discrepancies and revisions are really huge, so all of the claims here are quite uncertain.

The "missing piece": total violent death rates

In my view, the single best indicator for long-run trends in violence is the violent death rate: how many total deaths from violence there were, per person (or per 100,000 people), per year. Deaths tend to be easier to verify and measure than most other relevant indicators, and to the extent that Better Angels discusses very long time frames, it is mostly discussing some measure of violent deaths. But the way it presents these is, in my opinion, easy to get confused by:

• It presents many charts of declining homicide rates over the past several centuries.
• It does not present comparable charts for deaths from wars and atrocities (famines, genocides, etc.) It does have an extensive discussion of what we should make of the fact that some of history's bloodiest wars and atrocities were in the 20th century - but it does not look at the numbers for these wars and atrocities in the same terms as homicides, and plot the overall trend of all violent deaths combined. (Some more discussion of the analysis that is presented for atrocities is summarized at my 2015 blog post; see "BA's argument and the need for more analysis.")

In my view, this means there's a "missing piece" of the story of trends in violence: we see that one kind of violent death has become less common over the last several hundred years, but we don't see the trend in all kinds of violent deaths. And I consider this missing piece significant, because my sense is that wars and atrocities account for far more violent deaths than most of the other sources of violence the book discusses.

For example - and this surprised me - the global rate of violent deaths from the 20th century's “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.

Deaths from wars and atrocities, by century

I did this analysis a few years ago, so it's possible that it doesn't incorporate some corrections to the data source I'm using.

I believe the main source that Better Angels uses to tabulate the death tolls of the biggest wars and atrocities is Atrocities: The 100 Deadliest Episodes in Human History by Matthew White. (A bit more discussion of why I used this source in this post.) So I pulled the numbers from its "one hundred deadliest multicides [wars and atrocities]" table (pg 529).

I used these numbers to create estimates of the death toll from wars and atrocities each century, per 100k people per year.¹ Here's the result (calculations here):

My observations:

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

The figures are noticeably lower before the 13th century (which is strikingly close to when we have our first documented evidence of declining homicides).

This could be misleading: as both Better Angels and Atrocities point out, 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.

On the other hand, I somewhat 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 earlier centuries had comparable war/atrocity death tolls, they probably came from much larger numbers of smaller wars and atrocities.

There is a lot of uncertainty in these figures, and revisions could change the picture quite a bit.

• 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 nearly all from Genghis Khan (referred to above as “Chinggis”), and there is a lot of room for doubt here - alternative estimates² imply less than half the death toll. 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.
• For the 8th century, Atrocities estimates 8 million deaths from the An Lushan rebellion. But this is a downward revision from an earlier figure (by the same author) of 36 million, cited in Better Angels. The older figure would make the 8th century appear as one of the most violent. Some context on why the revision happened is in this footnote.³

What happens when we account for both atrocity+war deaths and homicides? It's very hard to say, because the only homicide rate data I've found (see here) is from (a) a few European countries starting in the 1300s; (b) the USA starting in the 1600s.

But I wanted to get a sense of the rough ballpark, so I threw together an estimate of the global homicide rate (assumptions I made are in this footnote⁴). Here's the result (spreadsheet here):

In this chart, I'm assuming that only Europe saw falling homicide rates as early as 1300 or so,⁵ and I'm assuming that homicide rates started falling in the rest of the world starting around the Industrial Revolution. That is my best guess.

Here's an alternative version, in which I assume that homicides fell worldwide the same way they fell in Europe:

This is all very imprecise, because global homicide rates are a mystery before very recently. But the big picture is that the rise in death tolls from huge wars and atrocities plausibly offset the decline in homicide rates, leading to a flat or unclear trend in overall violent death rates between ~500 BCE until the Industrial Revolution (~1700s), or even later.

Since the mid-20th century, deaths from wars and atrocities have been much lower - though it's hard to make too much of this, because in general, wars and atrocities seem very volatile, with massive death tolls sometimes occurring after centuries of relative peace. Furthermore, as argued in Toby Ord's book The Precipice, there's a case to be made that today represents an all-time high in terms of risks that could decimate a significant fraction (or even all) of humanity. So maybe rather than declining violence, what we're seeing is a continued trend of more and more infrequent risks of larger and larger catastrophes.

Footnotes