Genetics and the Historical Decline of Violence?

There is an insightful and data-rich blog called hbd* chick that is well worth our reading. A recent post discusses Steven Pinker’s new book about the decline of violence. I haven’t read the book, but the discussion on hbd* chick’s blog suggests that Pinker does not give much weight to the possibility of genetic change in European history causing the long term declines that we see, for example from Eisner’s careful tabulation of homicide rates in Europe from Medieval times.

In the Western European countries for which Eisner found data on homicide rates there was a bumpy but steady decline from Medieval times through the twentieth century. Here, for example, is his chart showing homicide rates in England: note the log scale on the Y axis. The rates are homicides per 100,000 population.

Eisner england

The data from other European countries are similar but some, especially Scandinavia, seem to have undergone the change later by a few centuries.

Homicide rates (these are local rates and do not include war casualties) declined dramatically or so it seems. There is a near hundred-fold difference between English homicide rates in 1300 and 2000. What if, as Gregory Clark would suggest, these declines reflect genetic, i.e. evolutionary changes, in the populations. How much selection would be required to cause such a decline in violence? An interesting recent development in human genetics is the resurgence of classical quantitative genetics as an older faith that we would quickly find “genes for” one thing and another mostly failed. It seems that RA Fisher got it essentially right in 1918. A good entree to current literature is this remarkable paper showing that cryptic distant genetic kinship predicts IQ correlations just fine.

In this tradition our model is that there is some underlying partially heritable trait that we might call “propensity to violence”. It doesn’t matter what we call it since it is just some direction in a high dimensional space of characters. This trait has a Normal distribution in the population, and individuals with a high enough value of the trait commit homicide. In other words homicide is in our model a classical threshold trait. We assume that selection on this trait is mild so that directional selection acts to shift the whole distribution up or down.

In 1300 the homicide rate was about 50 per 100,000 people, or 0.5 per thousand. Homicide must have caused on the order of 1 to 2 percent of all deaths and a much higher proportion of deaths of young adult males. Our assumption of a Normal distribution of the underlying trait immediately implies that the threshold was 3.3 standard deviations greater than the mean (from any table of the Normal distribution). Natural selection, social selection we would say in this case, disfavors homicide and the distribution is shifted so that the homicide threshold is surpassed by only 1 in 100,000 people rather than 1 in 2,000. By the year 2000 the homicide threshold is at 4.3 standard deviation from the population mean. In other words selection has moved the distribution 1 standard deviation in 700 years or 28 generations.

This amount of change, 1 standard deviation, would correspond to a change in IQ of 15 points or a change in mean male stature of about 2.5 to 3 inches. What strength of natural selection would have been required to cause this amount of genetic change?

The workhorse of quantitative genetics is the “breeder’s equation” which says that the response to selection is the product of the additive heritability and the selective differential, which is the difference between the population mean of a trait and the mean among parents. It is usually written as


In the present case we need a response of 1/28 of a standard deviation per generation. Assuming an additive heritability of 0.5 (the true value is probably 0.8 or so from literature on the heritability of aggressive behavior in children) the selective differential must be about 1/14 or .07 standard deviations per generation. In terms of IQ this would correspond to a one point IQ advantage of parents over the population average and in terms of stature parents with a mean stature 0.2 inches greater than the population average. This would occur if the most homicidal 1.5% of the population were to fail to reproduce each generation.

Justice was famously brutal and harsh in Medieval and Renaissance England so this may not be an entirely meaningless exercise. In this excellent essay Peter Frost suggests that the nearly the same selection against violence occurred in the several centuries before the fall of the Roman Empire, and he provides grisly details of Roman treatment of criminals.

About these ads
This entry was posted in Uncategorized. Bookmark the permalink.

48 Responses to Genetics and the Historical Decline of Violence?

  1. FredR says:

    Beautiful! Plz post more on this issue. Clark’s model doesn’t seem very distinguishable from the class-based differential reproduction many early eugenicists proposed as the basic mechanism for the progress of civilization.

    • Kiwiguy says:

      @ FredR,

      see also, Steve Hsu had a post discussing a basic model where you could get a 1 SD shift in heritable trait in less than 1000 years.

    • harpend says:

      Yes, I agree. Although Greg Clark seems almost to dodge the topic the beginning of the industrial revolution was also (more or less) the beginning of the demographic transition. The old correlation between wealth and fitness turned upside down, there was a flood poor people who could now survive, and this caused alarm among the talking classes and was I think the proximate cause of the eugenics movement.

      @Kiwiguy: thanks for the reference to Steve’s post. I had forgotten it, shame on me for not citing it.

  2. Candide III says:

    Doesn’t selection apply for non-genetic (broadly speaking, cultural) heritable traits as well?

    • harpend says:

      Certainly, yes. The problem is that quantitative genetics provides a well worked out and well tested model. Despite years of effort we still have nothing equivalent about cultural transmission.

  3. 1.5% seems high, considering that, according to the chart, an average Englishman’s lifetime odds of committing a homicide never exceeded about 0.3% at any point after 1600. (His/her odds of being executed in his/her lifetime were also around 0.3%, though most executions were not of murderers – murders were rarely solved – but of highway robbers and horse thieves.) You need additional mechanisms to select against individuals with homicidal tendencies.

    An additonal data point to consider is that, if you translate annual murder rates into odds of committing a murder due to the changes in life expectancy and age structure, and correct for increased survival rates due to greatly improved EMT response in the second half of the 20th century, you’ll see a significant increase in murder odds (on the order of 5x) starting at least in 1950 and possibly as early as 1930, and peaking around 1990.

    • harpend says:

      The high frequency changes in rates of violence like the increase to 1990 and the drop afterward likely have little to do with genetics. IMHO we are more or less stopped in our tracks because we have no general coherent theory of cultural transmission.

      As Steve points out below we think (most of us) that homicide is an indicator of something else—proneness to violence, high time preference, fast life history, and so on. We cannot treat homicide as a discrete real trait. It is instead a shadow on the wall of the cave.

      • Maybe. All I’m saying is that, when you have a 1930-1990 swing that is almost half the size the purported 1300-1930 genetic effect, your theory better be able to explain it.

      • observer says:

        In response to Eugene, I think it will be necessary to dig up and sample a lot of old cemeteries in order to complete this kind of argument. The math from Harpending and Cochran is persuasive but doesn’t resolve the question of just how important the genetic vs cultural component is in inducing the change in murder rates. One could imagine several interesting possible scenarios.

        E.g. if the rise of Roman or English law enforcement reduces the fitness value of a ‘disorderly’ phenotype, this in time could give rise to a still more orderly society in which murder rates decline faster than the phenotype itself, due to more effective policing or the like.

        Of course in a thoroughly peaceful society this process could halt or even reverse if the pacified majority lost the ability to stomach the sterilization or execution of the disorderly. Such a scenario could conceivably result in an interesting cycling effect rather than a stable state, depending on how much genetic changes emphasized compassion over simple orderliness.

        That leads to another interesting question: was pacification achieved through the same genetic means over the history of different civilizations, e.g. Rome and Imperial China, or England and Japan? Possibly not.

      • JL says:

        Maybe. All I’m saying is that, when you have a 1930-1990 swing that is almost half the size the purported 1300-1930 genetic effect, your theory better be able to explain it.

        Like Henry said, it’s a log scale. In 1300, the homicide rate was about 23 per 100,000 while the 1930-1990 swing was from about 0.8 to about 1.2 per 100,000.

  4. Steve Sailer says:

    “there is some underlying partially heritable trait that we might call “propensity to violence”.”

    I suspect “propensity toward disorder” might be a little closer to something real. The main problem I had with Pinker’s book is that by talking about “violence,” he seemed to be lumping together a number of quite different things. He seemed to be drawing an overly sharp distinction violent and nonviolent disorder and not a sharp enough distinction between organized and disorganized violence.

    Here’s an example. Here are three groups of people. Which two are most similar?

    A gang of car thieves
    A gang of carjackers
    The physicists of the Manhattan Project

    In Pinker’s approach, the carjackers and the Los Alamos physicists would be the similar ones because they are both engaged in violence, while the car thieves are the odd man out because they are nonviolent.

    In my approach, the carjackers are more like the car thieves than they are like the Manhattan Project.

    • Matt says:

      Manhattan Project physicists were very intelligent. Not sure if they were particularly orderly, net of intelligence helps you plan and organise. Unleashing nuclear chaos and the arms race does not particularly seem like the action of one too much bothered with order. “Propensity to disorder” and “propensity to violence” may well be real traits, but I am convinced your example is particularly good evidence for either (as behavioural differences in these groups are explained well largely by IQ and curiosity and differences in other traits are more speculative).

      • Toddy Cat says:

        Actually, the “Arms Race”, insofar as it existed, was not particularly chaotic, and one could make the argument that the development of nukes was actually anti-chaotic, with regard to international relations. There can be little doubt that the (relatively) peaceful world we have seen since the end of WWII is a result of the invention and possession of nuclear weapons by the major powers. At least some of the Los Alamos physicists hoped, throught the development of nuclear weapons, to make major wars unthinkable, and they succeeded, up to a point. However, we seem to be entering a new era of nuclear proliferation, so the “nuclear chaos” may be yet to come, as actors less responsible than the Pentagon, the Politburo, and #10 Downing get their hands on the bomb. Interesting times…

      • brahms says:

        “You can’t be serious!”
        John MacEnroe

  5. gcochran9 says:

    Anytime you drop a population into a new but survivable environment, they change – genetically. The breeder’s equation tells you how fast this happens, at least in the short to medium run. Few biologists are familiar with it, particularly people working in human biology. Certainly very few social scientists are. Indeed, they’re not just unfamiliar with it, they actively deny it.

    Our Ashkenazi paper discussed a simple case: the main features were an unusual degree of reproductive isolation, combined with equally unusual job specialization. Selective pressures that hit the population as a whole are more common: there you mainly need reproductive isolation from past populations, which happens automatically, absent time machines.

  6. dearieme says:

    “Justice was famously brutal and harsh in Medieval and Renaissance England”: famously, perhaps, but accurately? How often were the brutal laws enforced? I ask because the leading historian of landscape history, Oliver Rackham, likes to point out that everyone believes in terribly severe laws against poaching deer but nobody can ever cite even one name of anyone executed under them.

    • Toad says:

      Also called Jonny Scot, and Johnny of Braidsley

      And he’s away to the gay green wood,
      For to hunt the red tail doe.

      Then by there coom a silly old man,

      And he’s away to the king’s foresters,
      For to tell on young Johnny.

      Seven men they heard him out,
      And then swiftly they did ride,
      And they coom on Johnny all alain,
      And they shot him in the side.

      Johnny shot six of them,
      And the seventh he wounded sore.

  7. j says:

    Greg’s elegant equation fails to convince me not because it is wrong (it is not) but because all the eugenic and pre-eugenic literature emphasizes that the criminal class outreproduces the lawabiding class. For example, “The Fertility of the Unfit” by WA Chapple refers to one criminal having 800 descendants. The book goes back to the 11th century in Italy to illustrate this phenomenon. The Roma, sometimes considered a criminally inclined population, did exceedingly well in Medieval Europe. My point is that it has not been demonstrated that criminality and violence lowers reproductive outcomes today or in historical times.

    • Anthony says:

      Henry says that one need not assume the death penalty was the mechanism of selection. There are many other plausible mechanisms of selection:

      More violent people are generally poorer, and thus less marriageable. Even if this meant only a later marriage, it might mean fewer offspring.
      More violent people are generally poorer, and thus less able to prevent their children from dying during famines or plagues.
      More violent people are generally poorer, and thus less able to prevent their children from being stunted or ill or otherwise less fertile as adults.
      More violent people might be more likely to kill their children.
      More violent people might be more likely to abuse their wives while pregnant, causing miscarriages and lowered fertility.

      Also, while the death penalty wasn’t often applied to murders (because of the difficulties of solving them), it was likely applied primarily to the more violent fraction of the population – if horse thieves are the most likely to be executed, the state is still thinning out the 10% most violent, if not the 1.5% most violent.

      • Toad says:

        “More violent people are generally poorer”

        And the poor have higher fertility.

      • harpend says:


        The poor have higher fertility these days but it is only after the Industrial Revolution that we see that. Clark has good data on wealth and fitness in Medieval England. Wealthy gentry and merchants had twice the surviving offspring as the poor.

  8. Alice Finkel says:

    The death penalty was very common in England as was the punishment of transport to the colonies. Such methods may have curtailed the growth of the criminal classes temporarily.

    Modern immigration policies in Britain are rapidly reversing any gains accrued via the former historical methods of population control.

    The US confines large numbers in prisons, but still allows conjugal visits — which defeats the purpose, in population terms.

    • Death penalty was applied in early modern England at the average rate of 100 executions per ~10 million residents per year. A lot compared to modern Western countries, but almost negligible as a method to curtail the growth of the criminal classes. Transport to colonies was more common. The primary destination for these convicts prior to 1790 was North America. (Makes you wonder why the US ended up doing as well if not better than England.)

      • harpend says:

        Nothing to disagree with in the points that you make. But we need to step back and realize that the model I put up is nothing more than a model, a cartoon, that might help us understand the history and perhaps predict things that were not obvious. The numbers that came out are not to be taken too seriously.

        There are lots of ways that selection could have happened: no need to assume that the death penalty played a big part. For example I might suggest that both murderers and murderees came disproportionately from the violence prone end of the spectrum.

        Re the recent decline in violence, “high frequency” changes, these almost certainly have cultural causes rather than genetic causes. No argument at all with that. Giuliani showed us how that can work.

    • Toad says:

      “punishment of transport to the colonies.”

      Fertility rates of colonists in the low population density colonies was much higher than those that stayed behind in the built-up mother countries.

  9. Priceeqn says:

    J. Seger has a manuscript from way back where he shows that you could, due to sexual selection, get a 1SD shift in 100-200 years.

  10. anon says:

    Interestingly, the decline of violence in Mexico happened much faster than in Europe

  11. agnostic says:

    It’s just a nitpick, but strictly speaking these are victimization rates rather than offender rates. However, making reasonable assumptions about how many offenders there were, given the number of victims, does not seem to change the basic conclusion. For example, if there were half as many murderers as murder victims in both periods, if there were half as many in 1300 but one-tenth as many in 2000, etc. That doesn’t affect the results.

  12. agnostic says:

    An overlooked mechanism of selection here is female choice that would affect male reproductive success.

    Perhaps as part of the broader shift toward stable and nose-to-the-grindstone lifestyles from 1300 to 2000, women began moving away from having kids by the kind of guy who winds up sticking another guy with a knife in a public tavern, just because he drunkenly stepped on his shoes without apologizing. And choosing in his place the kind of husband who would fit better into the “industrious revolution” / Greg Clark story.

    In fact, this genetic mechanism could feed back into and accelerate the cultural mechanism of more powerful police forces, death penalty, or whatever. When cultural changes make it less worthwhile to be a violent and disorderly guy, women will find him less worth choosing — his way of life seems to bode poorly for the future, what with all the greater policing forces roaming around. Then once violent males become rarer due to lower success with females, the state / police have an even easier time at isolating and weeding them out, since the state’s policing resources now do not have to be spread as thinly as they would in a population with more, rather than fewer, violent males.

  13. Rob King says:

    Just in case no-one has shared this yet
    More recent evolving occuring

    • harpend says:

      Greg has some very interesting and very strong things to say about this paper. (What a surprise!) I am urging him to put a post about it.

  14. Maule Driver says:

    Consider this:

    “The result of this breeding program conducted over more than 40 generations of silver foxes was a group of friendly, domesticated foxes. These domesticated foxes, which were bred on the basis of a single selection criteria, displayed behavioral, physiological, and anatomical characteristics that were not found in the wild population, or were found in wild foxes but with much lower frequency. One of the reasons that these findings were so compelling was that the criterion used to determine whether an individual fox would be allowed to breed was simply how they reacted upon the approach of a human. Would they back away, hissing and snarling, and try to bite the experimenter? Or would they approach the human and attempt to interact?

    It doesn’t take much.

  15. nooffensebut says:

    “How much selection would be required to cause such a decline in violence? An interesting recent development in human genetics is the resurgence of classical quantitative genetics as an older faith that we would quickly find ‘genes for’ one thing and another mostly failed.”

    For violence? I beg to differ. MAOA-2R doubles violent delinquency in men and quadruples risk of arrest in black men. An index of MAOA-3R, ANKK1, DAT1, DRD4, and 5-HTTLPR is a significant predictor of adult black male violent delinquency, unlike an index of the mother-son childhood relationship.

  16. Slow Learner says:

    r = h^2 * s …

    In the present case we need a response of 1/28 of a standard deviation per generation. Assuming an additive heritability of 0.5 (the true value is probably 0.8 or so from literature on the heritability of aggressive behavior in children) the selective differential must be about 1/14 or .07 standard deviations per generation.

    Perhaps I am thick but I am having problems making the numbers add up.

    1/28 = 1/2 * 1/2 * 1/14?

    Could it be that the selective differential needs to be 1/7?

    • harpend says:

      I do have a nearly perfect record on this blog of getting the algebra wrong with my short examples. My logic apparently was that to get 1/28 of a standard deviation per generation the selective differential must be twice that or 1/14 per generation in which case r=1/2 * 1/14 = 1/28. Where did you get the second 1/2 ?

      • Slow Learner says:

        I thought that since h was 0.5 and that it was squared in the equation, I had to square it when plugging the numbers in. I guess I will just have to go back and read up on that stuff again.

      • Slow Learner says:

        OK, I see that h^2 is 0.5 or 1/2, and thus my mistake.

  17. Matt says:

    This may be of interest -

    “Dr. Jacob Bock Axelsen of the Biomathematics Unit at Tel Aviv University’s Department of Zoology says that individuals are markedly more peaceful now than throughout human history—and now he’s got a mathematical model to show it. Originally designed to explain population densities of language groups, it’s the first mathematical model to map the intensity of conflicts through the global demographic history of the past millennium.”

    Finding that hunter-gatherer societies were the most violent per individual, the model revealed that historically the most violent areas of the world have been Papua New Guinea, Africa, Europe, Asia, and South America, as determined by the strength of conflicts in the region.

  18. Slow Learner says:

    This would occur if the most homicidal 1.5% of the population were to fail to reproduce each generation.

    Or if the dumbest 1.5% of the population fail to reproduce …

  19. The fourth doorman of the apocalypse says:

    Q: What will it take to close the achievement gap?

    A; If you are asking about the black-white achievement gap, about 700 years.

  20. athEIst says:

    he provides grisly details of Roman treatment of criminals

    I saw three methods mentioned but no grisly details.

  21. bubble says:

    “This would occur if the most homicidal 1.5% of the population were to fail to reproduce each generation.”

    I think there’s a second factor that makes the numbers come out even easier. In terms of trait bundles i don’t think it’s just propensity for violence i think it’s a combination of propensity for violence and capacity for violence where propensity is likelihood to be violent and capacity is the likelihood of successfully kiling.

    Individuals would then fit into four categories:
    1) high propensity and high capacity
    2) high propensity and low capacity
    3) low propensity but high capacity
    4) low propensity and low capacity

    The majority of murder would come from group (1). Group (2) would probably come second in homicides but more manslaughter than murder. Group (3) generally only with extreme provocation like they come home and find their wife with another man.

    Who knows what the proportions would have been but if for the sake of argument they were equal between the four categories then the number would be closer to ~ 0.4% rather than 1.5%?

    The reason i say this is a lot of cops and soldiers have exactly the same *capacity* for violence genes as violent criminals (e.g. MAOA imo) but where they differ is they have restraint genes as well which reduces the *propensity*. If you talk to cops in a non-PC context they’ll always say the problem with extremely violent gangbanger types is they lack self-control. They don’t even notice the irony.

    “though most executions were not of murderers – murders were rarely solved – but of highway robbers and horse thieves”

    Same difference. If you executed muggers the rape and murder rates would dissolve.

    “all the eugenic and pre-eugenic literature emphasizes that the criminal class outreproduces the lawabiding class”

    There are lots of decent poor people. The people writing those books disproportionately noticed the scary ones.

    nb Latvia, Lithuania etc lagged Scandinavia i think hence their higher rates now.

  22. brahms says:

    There seems to be a sustained uptick in homicide rate right around 1965 – any major changes in Western Civ at that time:)?

  23. pablo brandt says:

    This happens cause the Police Force grew up in strength, and the femalization of White males has its impact too.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s