Thursday, June 07, 2012

Consanguinity and corruption

In the wake of MG's essay on the nature and nurture of corruption, I wondered if a hard correlation between consanguinity rates and graft at the national level had been discovered. Searching for as much, the top returns I received were from MG and HBD Chick. Apparently, it hasn't been an area of academic interest, though HBD Chick is deserving of an academic spot for her intellectual curiosity about and indefatigable efforts researching and relaying consanguinity through history and up to the present to any who happen to be interested in as much.

Why should academics and policy makers take note, though, when they've already identified the culprits? They are, of course, bad laws, bad leaders, and bad institutions! Fix these things and any country is capable of resembling Norway. Any day now we'll get the right laws and enforcement mechanisms in place and use them to throw out the crooks and set things straight in Zimbabwe, Zaire, Syria, Sudan, the Congo, Nigeria, Sierra Leone, Somalia, Burma, Iraq, Afghanistan, Papua New Guinea...

As is often the impetus here, not finding what I was looking for meant needing to figure it out. The data aren't perfect by any stretch, but something is better than nothing. Computing simple, unweighted averages for each country for which studies and surveys have been conducted and subsequently recorded on and then comparing them to Transparency International's 2011 Corruptions Perception Index yields a correlation of .44 (p = 0). In places where extended families are important and family members are more closely related to one another than they are in the West, outsiders are treated with much less even handedness than kin are and nepotism is, if not the rule, at least perfectly acceptable. In these places, if you're not blood, you're going to have to pay to play.

A correlation of .44 is considered fairly strong in the infinitely varied world of the social sciences, but the true relationship between corruption and consanguinity is almost certainly even more vigorous than that. I'm using imperfect and sporadic data. There is nothing available on inbreeding for about half the countries in the world while for India there are 45 studies for which I must, by necessity, compute a simple average from, because even if I wanted to try and weight the sources for geographic and demographic representativeness within India, I'd be utterly unable to do so competently since I know so little about that extremely complicated country of over 1 billion people.

Further, even to the extent that the data are representative, they leave something to be desired, as the chickadee explains:
what we are talking about here when we discuss inbreeding vs. outbreeding and nepotism and/or corruption are types of altruistic behaviors -- and these behaviors/attitudes have evolved differently in different populations, of course, over time. so you can't just take a population that has been inbreeding for scores of generations, and likely evolved certain altruistic behaviors, and change their behavior patterns via just one or two generations of outbreeding. there is going to be some lag-time.

 why do i say this? because the problem with using the numbers for the kind of analysis you describe is that there is no time depth to them. if you look at the data, it appears as though the chinese have similar inbreeding/outbreeding rates to western europe or canada, but that's only in the last generation or so (and even that is debatable). as i've blogged about, the chinese have been inbreeding for literally millennia. any effects that's had on altruistic behaviors are NOT going to be overturned in one or two generations.
what needs to be done is that the histories of inbreeding/outbreeding in different populations need to be quantified (part of my ongoing, neverending project @hbd chick (~_^) ), and then those numbers need to be compared to transparency international's and/or other figures.
Yet despite this, we still see a rigorous, statistically significant correlation between corruption and consanguinity. Randomly generated numbers don't correlate with one another. If (when?) much of the remaining randomness in the consanguinity numbers is removed and the appropriate adjustments for time depth are made, the observed correlation will prove to be stronger still.


Anonymous said...

"Fix these things and any country is capable of resembling Norway."

Well, that plus 10k years of selection.

sykes.1 said...

I'm not sure nepotism is corruption, especially as applied to an extended family. Favoring one's family is different than stealing for one's self.

There is also the issue of long-standing social norms. A society in which winner-takes-all (the norm throughout history everywhere) has a different standard of corruption than Norway. They might regard Norway as corrupt for not caring about their families.

Anyway, the ruling class in Norway is virulently anti-semitic, and they teach their children to hate Jews in organized camps. So why is Karzai corrupt and not Jens Stoltenberg? Norway is plainly a corrupt society.

Anonymous said...

1. Correlation isn't causation
2. Show your work

Audacious Epigone said...


1. Right. I'm not claiming causation here, just speculations on what the data show, per usual.

2. I provided the data sources. You want a graph? I didn't make one because I don't have a program labels individual points. There's a tedious way of doing it using Excel and Paint if necessary.

Anonymous said...

"Well, that plus 10k years of selection."

If you know what the biological mechanism is then you might not need 10K years of accidental selection.

"I'm not sure nepotism is corruption, especially as applied to an extended family."

It's a question of scale. You can see the effect of large-scale nepotism and corruption throughout the poorest countries of the world.

If it's true that the more inbred a group is the more corrupt and nepotistic they'll be then you have a practical answer to a practical problem.

As mentioned, a lot of countries e.g. Japan, Korea, have seen a dramatic drop in consanquinity since WWII. If the effect of consanqunity is cumulative over generations then this drop would have a distorting effect.

for example

Country A has default human consanquinity of 100 and a corruption index of 100

Country B has a low level of consanquinity, 50, and a corruption index of 50

Country C is transitioning from A to B but has a corruption index higher than would be expected for its *current* level of consanquinity when plotted on the same graph as A and B if the effect was linear. This is because the influence of that country's *past* levels of consanquinity are still part of the equation.

This is because measurements of consanquinity generally measure *current marriages* between close relatives as a proxy for consanquinity itself.

So if for example it was possible to get the data for current marriages *and* the rate for the previous few generations and make an average the correlation would be stronger imo.

e.g. the data for the last four generations of country A might be100, 100, 100, 100, averaging 100

for country B 50, 50, 50, 50 averaging 50

but for country C 100, 90, 80, 70 averaging 85

so if it worked in a direct and linear way (unlikely to be as simple as that obviously) you'd expect country C to have a corruption index of 85 despite having a *current* consanquinity index of 70.

M.G. said...

Thank you Audacious Epigone for running these numbers. I've added a link to this in my original post, so people can see that some quantitative work's been done on the subject.

So if for example it was possible to get the data for current marriages *and* the rate for the previous few generations and make an average the correlation would be stronger imo.

Definitely. It's a pity the data's so spotty. HBD Chick's out there like a lone adventurer trying to put the pieces together without (I presume) funding or any university backing. This is a question that IMO has clear and immediate interest for international policy-makers.

Audacious Epigone said...


Congrats on the mention by Steve Sailer. He's doing us a service by pointing out the service you're doing.

Anonymous said...

"This is a question that IMO has clear and immediate interest for international policy-makers."

Absolutely. The potential is mind-blowing.

hbd chick said...

@anonymous - "Well, that plus 10k years of selection."

i don't think 10,000+ years is required here. i think we're more in 10,000 Year Explosion territory, iow less then 10,000 years -- possibly way less.

if these guys are right, just 50 generations of inbreeding plus some strong selection can result in some pretty high frequencies of "genes for altruism" (and, i'm guessing, by extension similar amounts of outbreeding result in the opposite). fifty generations is just something like 1000 years....

hbd chick said...

@the awesome epigone - "...a correlation of .44 (p = 0)."

wow. higher than i would've guessed! thanks for running the numbers. (^_^)

james wilson said...

I doubt 50 generations are required to significantly alter heritable behaviors--or physical characteristics--depending on the degree of randomness invovlved. A geneticist specializing in canines claimed that any breed of dog could be selected within that breed to produce any other breed in a relatively short time.
And bad behaviors cannot be bred out of humans, because they are as likely to cling to them as they are to good ones. Since bad behaviors are no longer rewarded with starvation, incurable disease, or annhilation (they may be rewarded for it), it's a moot point anyway.