Science

Violence: Are Percentages or Raw Numbers a Better Measure?

|

Steven Pinker has been getting much attention lately for arguing in his new book The Better Angels of Our Nature, against a lot of people's surface assumptions based on a pretty geopolitically violent past century, that modernity and the modern state have in fact made the world a far less violent place. 

David Bentley Hart at First Things wonders if Pinker's measures make the most sense, given that we are still, after all, sitting on a pretty big pile of murdered bodies here in the aftermath of the 20th century:

Pinker's method for assessing the relative ferocity of different centuries is to calculate the total of violent deaths not as an absolute quantity, but as a percentage of global population. But statistical comparisons like that are notoriously vacuous. Population sample sizes can vary by billions, but a single life remains a static sum, so the smaller the sample the larger the percentage each life represents. Obviously, though, a remote Inuit village of one hundred souls where someone gets killed in a fistfight is not twice as violent as a nation of 200 million that exterminates one million of its citizens. And even where the orders of magnitude are not quite so divergent, comparison on a global scale is useless, especially since over the past century modern medicine has reduced infant mortality and radically extended life spans nearly everywhere (meaning, for one thing, there are now far more persons too young or too old to fight). So Pinker's assertion that a person would be thirty-five times more likely to be murdered in the Middle Ages than now is empirically meaningless.

In the end, what Pinker calls a "decline of violence" in modernity actually has been, in real body counts, a continual and extravagant increase in violence that has been outstripped by an even more exorbitant demographic explosion. 

A Reason.tv inteview with Pinker:

NEXT: The First Climate Change Trade War Breaks Out?

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. OK, look you guys, you just have to stop posting that photo of Pinker! It is just too creepy.

    1. Wouldn’t take much makeup to put him as an extra on The Walking Dead, huh? Get that ol’ boy some time in the Sun!

      1. Or, with no makeup, to put him here.

      2. He looks like Danny Devito in Frank’s Little Beauties.

    2. you just have to stop posting that photo

      No! Peek some Pink!

  2. So Pinker’s assertion that a person would be thirty-five times more likely to be murdered in the Middle Ages than now is empirically meaningless.

    I love it: The fact that you are far less likely now to die young of innumerable pathogens than in the Middle Ages makes our lower homicide rates meaningless?

    No, they would mean more!

    1. Last I checked, pathogens don’t commit murder.

    2. His point was that the extension of life, and reduction of infant mortality, swell the demographic cohorts least likely to commit or suffer violence: the very young and the very old. Thus, a comparison of raw percentages does not take account of changes in the demographic shape of society.

  3. Personal antedotes are the most useful because you can completely bypass logic and reason with them.

    1. Stop getting my name wrong!

      1. Really?

        1. *slurp* Really!

          1. Does your mommy know you’re not at school today?

    2. Can I give you a C in spelling and remove your T? It’s anecdote:) Try Firefox. It has spell check.

  4. Well, his orders of magnitude argument works both ways. We wouldn’t say that a country of a million people that had one murder was half as violent as a village of a hundred people that had two murders, would we?

    No, anytime you have big changes in the denominator, the best comparison is rates and percentages.

    What, perhaps, is most interesting is that the huge body counts we’ve been running up have been the product of state action, in a way that was not true until pretty recently on a historical scale.

    1. the huge body counts we’ve been running up have been the product of state action

      The more power we give to those whose job it is to protect people and ensure peace, the more people they kill.

      Go figure.

      1. A+ for brevity and relevence. A perfect comment.

        1. For that reason, I find this all too bizarre to be worthy of much argument. You’re much less likely to be killed in a street fight with Caravaggio than 500 years ago. You’re much more likely to die when your entire neighborhood is ethnically cleansed by the government than 500 years ago. Didn’t we all know that?

          1. Pinker’s actually arguing that you are currently less likely to die when your neighborhood is ethnically cleansed by some group of armed men claiming some kind of authority.

            The only reason we think of genocide and “total war” as recent developments in human history is that no one in ye goode olde days thought they were a big enough deal to comment on.

      2. we had to destroy the village in order to save the village….

    2. Not disputing that meta-point RC but the methodological individualist in me would say that if the whole number or murders is one or two, that the relevant explanations are best looked at in something other than the systematic incentives–there are probably observable and discoverable individual reasons for the murders for which resort to systemic social science explanations seems less necessary/

      1. Oh, sure. I was just that Hart’s rhetorical tactic is pretty null when trying to come to grips with the issues.

      2. I think you have to average over a longer time span.

        If there is one murder in the Inuit village and then none for the next 20 years, that one might be an outlier and can be ignored.

        But then again, by that standard, can we ignore Hiroshima and Nagasaki?

        1. Since they are part of war, no. How can they be outliers, just because that specific method of destruction hadn’t been used before or since? There is nothing unique about killing large numbers of arguably innocent people that belong to an enemy state. It’s actually quite common during war even if nukes are not.

      3. Brian: I am not sure why you apparently think that methodological individualism provides some kind of objection to Pinker’s findings. Of course, percentages are derived by comparing one subgroup of “individuals” out of all the “individuals” who make up the whole group. Rates of murder, rape, and assaults can only be calculated for populations.

        With regard to war deaths and democides (death by government), Pinker’s data are pretty persuasive that those rates are also lower than in the past.

        My reading of the Hart piece is that it’s just one more (failed) attempt to bolster the common belief that humanity is irremediably depraved and that we live in the worst of times.

        1. The part of it I thought was interesting is the part I quoted. And regardless of whether you find Hart’s metaphysics valuable or tenable or not, and if it is true that our vast increase in population that allows a far greater number of violent deaths to result in a far lower rate of violent deaths happened for reasons unconnected with that increase in population, then the conclusions that I’ve seen many draw from Pinker’s observation about the value of the modern state seem questionable. And this is an interesting point to me.

          1. Also, the “methological individualism” part is specifically about R.C.’s example of communities where there were merely one or two murders. I am questioning the need for any larger theory to explain what system is comparatively less violent in such a world; we can merely investigate why those one or two individuals were violent.

            1. I believe the example of a community where there is only one killing is Hart’s not mine. As long as you apply your methodological objection to his version of a comparison involving vastly different denominators, I’m cool.

          2. And regardless of whether you find Hart’s metaphysics valuable or tenable or not, and if it is true that our vast increase in population that allows a far greater number of violent deaths to result in a far lower rate of violent deaths happened for reasons unconnected with that increase in population, then the conclusions that I’ve seen many draw from Pinker’s observation about the value of the modern state seem questionable.

            I read this Greenwaldian behemoth 10 times and I still have no clue what it means.

        2. With regard to war deaths and democides (death by government), Pinker’s data are pretty persuasive that those rates are also lower than in the past.

          Not having read the book, I am very surprised to hear this.

          I would have thought that the industrialization of warfare would have raised the killing potential of armies by an order of magnitude. I assumed killing people retail by swinging sharpened metal bars at them would have limited the numbers compared to the wholesale killing of industrialized war.

          1. I would have thought that the industrialization of warfare would have raised the killing potential of armies by an order of magnitude.

            This is weird to me. It seems libertarians never miss a chance to point out that trade results in more prosperity, technological in this case, and less war.

        3. I would see humanity not as ‘irremediably depraved’, just ‘utterly imperfectible’. The point being that there is an indivisible quantum of evil in human nature, and it is imprudent if one would counter it to think it soluble.

    3. “What, perhaps, is most interesting is that the huge body counts we’ve been running up have been the product of state action, in a way that was not true until pretty recently on a historical scale.”

      True. Pinker’s analysis assumes that death by violence is purely a random event, and not the result of intentional government policy. He overlooks the critical distinction between death by pathogen in the Middle Ages versus death by war, attempted mass extermination and ethnic cleansing, intentionally-imposed famine, etc.

      1. … What book did you read?

        1. Pinker is the new Obama. Everyone talks about the books but nobody bothers to read them.

    4. I too think it’s unreasonable to use percentages alone when dealing with really low numbers.

      If that village of 100 had one murder last year and three this year, it’s murder rate went from 1% to 3%, a 200% increase in the overall murder rate.

      I think it’s more statistically significant if 100 people are murdered in a town of 10,000 then I do 1 in a town of 100.

      When one person can cause a crime rate statistic to jump more then a fraction of a percent, it’s probably unreasonable to treat that increase as anything but a fluke or freak occurrence. They did this when dealing with effects of 9/11 on various statistics in New York.

      1. There are well understood means of dealing with the problem of statistical measures of small numbers.

        This is a hobby horse of mine, you understand: but I think that it is one of the most neglected easy subjects in analysis.

        We can make comparisons that include freak events just fine.

        * You can average over enough time to get bigger numbers.

        * You can apply Poisson statistics to the rates (usually the right thing to do).

        * You can often construct a power law to model the rates, and compare the exponents (with error bars, of course).

        Just because newspapers are afraid to deal with this stuff doesn’t mean we don’t understand it.

        1. So, one man’s median is another man’s Poisson?

          1. Nice.

  5. I don’t see either side of this debate using personal anecdotes. The question at issue isn’t about the rates or the numbers, which neither side seems to be disputing. It’s about what conclusions about the value of modernity and the modern state one should be led to by the rates and the numbers. (Particularly for me as a libertarian, the value of the modern state as it relates to those rates and numbers.)

    1. This sounds anecdotal, or at least thought-experimental:

      a remote Inuit village of one hundred souls where someone gets killed in a fistfight

      At least part of what makes this example misleading is the small sample size for the village. At least for me, my brain discounts the significance of that single death accordingly.

      As others have pointed out, history shouldn’t get extra credit for having fewer people to die of violence in the first place. People were busy dying of other things that don’t kill them now. If current-day Earth were replicated 100 times on a 100 different planets, with the same rate of violent death, who would call that a tragedy?

  6. Would Hart rather live in a society of 100 million with a 1% violence rate or a society of 1 million with with a 25% violence rate? Ceteris paribus, pretty much any sane person would choose the former society even though there are a larger absolute number of violent incidents.

    1. But the question isn’t which one we want to live in, but which has more or less violence. If it’s a question of absolute morals, then clearly one would prefer the one which had fewer *specific instances* of violence (in your example, the society of 1 million). But that’s not what Pinker is getting at. The problem is that Hart and Pinker are really approaching the question from different perspectives. I can sympathize with Hart’s perspective, but it’s not really what Pinker is getting at.

      1. If it’s a question of absolute morals, then clearly one would prefer the one which had fewer *specific instances* of violence (in your example, the society of 1 million).

        Would we? Even if it also had fewer specific instances of human flourishing, well-being, cooperation, and warm fuzzies?

        I don’t think it’s at all clear that absolute numbers are the yardstick for absolute morals.

  7. The argument here seems to mainly be that if someone dies from a fist fight (preferably in a small population area, I guess) then that form of violent death just isn’t even close equal to the death of someone who died in a mass genocide. So, what… violence isn’t violence… unless it’s violence?

  8. percentages or raw numbers

    My friends and I usually use the “quality kills” metric. It works for movies, anyway.

  9. Sounds to me like they might jsut be onto something dude.

    http://www.Privacy-Pros.tk

  10. Lenny–No, not to put words in Hart’s mouth, but what made his point resonate with me is again about what we learn about the value of modern institutions from those numbers: would one rather like in a world where some human being might beat another in fist fight and that’s the only violent death, or one where an institution systematically leads to wars and genocides? I’m not saying the answer is obvious; Pinker’s conclusions and data are interesting, but I’m not sure they fully support the conclusions made of them about the overall value of the modern state. (especially since I’m pretty sure that purely social or intellectual changes not necessarily connected with the modern state also feed into the lower violence rates, though that’s of course a highly arguable point that we can highly argue over.)

    1. I think I would agree with you that Pinker’s overall conclusions that the modern state is responsible for the reduction rate is not necessarily the real factor. But seems to me that Hart’s response has little to do with that part of the conclusion, rather he seems to be arguing that “quality” of the death is some overriding factor in if we can really consider a violent death rate only in sheer numbers over percentages. In that sense I’m not sure I can agree with Hart’s assessment.

      1. I think the mode of the violence is important if the goal is assess the value of the modern state. Where most violence is one-on-one, people will have and feel like they have more control over whether they become victims. Where the violence is state directed (or protected), individuals have no control- it’s like being subject to the forces of nature.

        1. State sanctioned violence is still often one-on-one. It’s just that the person perpetrating the violence is the person you’re supposed to call for help when someone is perpetrating violence.

    2. Lenny–No, not to put words in Hart’s mouth, but what made his point resonate with me is again about what we learn about the value of modern institutions from those numbers: would one rather like in a world where some human being might beat another in fist fight and that’s the only violent death, or one where an institution systematically leads to wars and genocides?

      If I’m in imminent danger of being violently murdered, I’m not gonna care greatly whether the person wielding the gun is a free lance thug or someone wearing a badge or military uniform that somehow is supposed to make that imminent violent death somehow more palatable.

      Now, if your point is that a state might have the same murder rate as an anarchist organization, but the state will perpetrate additional violations of my rights, then the state would be worse.

    3. Brian: I am no fan of the state qua state, but the anthropological literature on homicide rates in pre-state societies pretty clearly shows that violence is considerably higher than in state organized societies. It turns out that Hobbes was on to something when he declared that life in the state of nature was nasty, brutish, and short.

      BTW, Pinker makes it clear that in early states, rulers reduced violence among their subjects not out of benevolence, but for the same reason that farmers reduce violence among their cows, pigs, and chickens – it gives rulers more surplus to enjoy.

      1. But White Injun tells me that the gamboling and frolicking people of North America and prehistory never committed homicide and it’s only farming and the state that turns us into beasts. Who am I supposed to believe? I can’t make up my mind: the person who cites a few crackpot books or someone who knows something about anthropological data?

  11. Obviously, though, a remote Inuit village of one hundred souls where someone gets killed in a fistfight is not twice as violent as a nation of 200 million that exterminates one million of its citizens.

    Just as obviously, an Inuit village of 100 with 20 murders is not preferable to a city of 1 million with 21 murders.

    1. I was thinking the same thing. If I recall, Pinker’s data showed >20% chance of violent death in the most primitive societies he studied. The 1% rate Inuit village is a straw man.

      1. I wonder about cause and effect. Pinker seemed to draw his data about past primitive societies by studying present primitive societies. But what if the extreme levels of violence is the reason that those societies are still primitive and not a result of their primitiveness?

        1. Croesus: Actually, Pinker also cites paleo-anthropological studies that, among other things, looked for signs of violent death in graves of early peoples. Researchers found evidence for high percentages of violent deaths.

          1. White Injun says this evidence does not exist, that these people were free to frolic.

  12. I wonder how much of a lower murder rate could the attributed not to the state, but to economic freedom and the ability to create wealth.
    If wealth that you create will be taken from you by agents of the state, what’s the point in trying?
    May as well murder and steal.

    1. In his book Pinker listed trade as one of the factors contributing to the decline in violence.

      1. and hookers. Lots of hookers.

      2. affenkopf: The modern slogan is: Make Money, Not War!

  13. Politics surpasses all other causes of violent death combined. Wars, political executions, colonization, starvation (at least in the 20th century) – all attributable to politics.

    Seems like the thing to do is lessen the importance of politics. If only there was an ideology that stressed that.

    1. But, but, but… ROADZ!11!!1! SOMALIA!11!!111

    2. Politics surpasses all other causes of violent death combined.

      Politics is almost synonymous with violence, so…yeah. Tautologies are tautological.

  14. Dig this:

    http://www.huffingtonpost.com/…..89145.html

    Wonder what Feministing would say…

    1. They would say “Hooray! Now we can point to an alarming increase in rapes, without mentioning the ‘increase’ is due to a change in the definition!”

      1. Removing the “forcible” part seems really dumb. Saying that men can be raped too seems really obvious. So, sort of a mixed bag overall.

        1. Here’s something you never hear a guy say: “Stop sucking my dick, or I’ll call the police!”
          -George Carlin

        2. Yes, I’m not seeing how any person would submit to being raped without putting up any physical resistance whatsoever and without any threat of violence if they did resist.

          Now, if they were lowering the threshold for physical resistence, then maybe, depends on what level was chosen, but none whatsoever?

      2. I’d figure the Jezebel types would say “good riddance to men getting raped, but they don’t deserve any justice over it”.

    2. Wow. Not a single major news outlet actually prints the new definition though.

      1. It wasn’t hard to find. Even HuffPo had it:

        The new terminology says rape is “penetration, no matter how slight, of the vagina or anus with any body part or object, or oral penetration by a sex organ of another person, without the consent of the victim.”

        So what is it with the consent of the victim? The use of victim here is a little odd.

  15. i havent read the book so maybe someone can correct me on this but it seems the even bigger flaw with pinker’s argument is that he doesnt seem to account for the fact that our emergency lifesaving technology has advanced so much that it has downgraded a lot of homicides into assaults simply by being able to save the victim.

    1. If you haven’t read the book, how do you know that it’s a flaw in his argument?

      1. i’ve read a bunch of reviews of the book and interviews and i havent seen it addressed so i’m curious.

        1. Most of the reviews are ideological bullshit. This will give you a pretty good sense of the book.

    2. I’ve definately heard that about war casualties. Many of the soldiers who end up amputees or paralized today would have died on the battlefield in earlier wars.

  16. By the way, could you blow the White Indian dog whistle any harder, Brian? We’re past 30 comments and he hasn’t even reposted the Psychology Today link yet.

  17. In the inuit village you just have to NOT be that one asshole who is constantly starting fights.

    In a large society (say Iran) you have to BOTH not be an asshole who starts fights AND not be part of a group that is threatened by the government.

    It seems like the expansion of freedom and global interdependence is driving down violence more than the proliference of large “modern” states.

    1. Wanna rub noses? No? You’re not Inuit?

      1. I smell what you did there.

    2. In the inuit village you just have to NOT be that one asshole who is constantly starting fights.

      He’s fighting with someone, so you also have to not be the guy he decides to attack.

      1. You also have to not be related to either of them, in case the loser gets pissy and decides to retaliate by, say, raping the winner’s womenfolk and sowing his fields with salt.

  18. Obviously, though, a remote Inuit village of one hundred souls where someone gets killed in a fistfight is not twice as violent as a nation of 200 million that exterminates one million of its citizens.

    Oh, yes, that is really obvious.

    So, by this reasoning, if you take a bunch of these little villages that add up to one million population, and they form a geopolitical unit, and each year 10,000 people in these villages get violently murdered (i.e., 1% per year), and compare them to one of the villages by itself, where on average 1 out of 100 residents gets violently murdered each year, then logically the geopolitical unit is 10,000 times as violent as any given village in the unit, and so * logically * it is 10,000 times safer to reside in a little village that secedes from the geopolitical unit.

    * facepalm *

  19. Shorter — Pinker doesn’t understand math, at least where it relates to rates versus totals.

    1. Pinker doesn’t? You sure about that?

      1. prolefeed: I think you mean Hart doesn’t.

    2. Is this really prolefeed? Since the handle says protefeed, I’m wondering if it’s a typo or someone being sly about prolefeed.

  20. What’s wrong with that sort of analysis, whole numbers or percentages, is that it’s averaged over a large and diverse geographical area – rendering it essentially meaningless for any specific location (say, a neighborhood).

    Even in a godforsaken place like Richmond, CA, there are neighborhoods with low crime.

  21. According to dunphy, anecdotes will do in a pinch.

  22. Maybe we should divide by the logarithm of the population instead of the population.

  23. Pinker’s proxy for violence, imposed death, skews the state’s contributions- states executed criminals a lot more back then, but we forcefully confine a lot more. It’s still violence.

    1. I’m sure he doesn’t consider that nearly all acts of government are violence. But I think it is still an interesting question whether one is more or less likely to die violently than in the past.

  24. unfortunately the argument glosses over the point if one man dies in a population of 200 due to a fistfight and that is a representative sample of fistfight deaths then the society of 200 millione that slaughters one million of its citizens is still going to be accountable for the 1 million fistfight deaths and is thus substantially more violent. furthermore the example does not cover a duration of tyhe data set if both death rates occurred over 60 years that would be one thing but if that was the death rate per year than both would be equally violent speaking about number out of context is not factually relevant

    1. Punctuation is your friend.

  25. The fact that nobody died laughing at Ron Paul’s fake eyebrows must mean something.

  26. I would have expected that a Middle Ages culture of violence, with modern weaponry, would result in a LOT more death than we see now, even with modern medical technology.

    Example: I don’t know how much medical technology really helps the people we hit with Predators.

    So, I think his point remains. The rate of mayhem is going down. The global media makes it look worse than it is, and than it *would have been* if Genghis Khan had nukes.

  27. Did anybody else get a little lightheaded during Pinker’s chapter with all the detailed descriptions of crucifixion, stretching on the rack, breaking on the wheel, Judas cradles, drawing and quartering, etc?

    The wheel thing is haunting my nightmares.

  28. The more important value to a given person is their own likelihood of safety from disease and violence and their ability to move and live without the risk of these things. Not the total number of victims, but the level of concern for a given person.

  29. I agree that the quoted part is the most interesting part of Hart’s article– the rest is mostly sarcasm.

    But, though he doesn’t follow it up by actually doing the math (apparently preferring to simply assume it works out as he thinks?), he does raise at least one interesting methodological criticism: if (implicitly assumed) violence is always mostly committed on young adults, then improvements in modern medicine that make more children and old folks survive nonviolent dangers will skew the age distribution of the population and make the same rate of violence per young adult look like a lower rate of violence per living human.

    It’s not clear to me that violence is inherently a danger only to young adults. But it may be a disproportionate one: wars mostly kill soldiers, gang violence mostly kills gang members, etc.

    Wouldn’t it be straightforward to check for, though? Strictly comparing violent deaths within a limited age band, say? (12-25?) Maybe the data isn’t available for that, though.

    But that quibble and ron’s comment above about modern technology turning what otherwise would be killings into maimings do seem like worthwhile criticisms, to me. But pointing out *possible* flaws is not the same as running the numbers…

Please to post comments

Comments are closed.