consanguinity and islam and democracy

i said last week that the week would be devoted to the woodley & bell consanguinity and democracy paper … and then i got distracted. typical. so, now, back on track…

aside from looking for any straight up connection/s between consanguinity and democracy (see previous post), woodley & bell also looked at consanguinity and democracy and several other possible factors that might affect the success of democracy in the nations included in the study: economic freedom, inequality, exports of fossil fuels (the “resource curse”), pathogen load (i’ll come back to that one!), and islam.

using path analysis, they found that islam seems to have a direct impact on democracy in muslim nations and ALSO that islam has an indirect impact on democracy via consanguinity.

recall that woodley & bell used two different indices of democracy: data from the polity iv project and the eiu democracy index. so they worked up two path analyses (click on charts for LARGER view). percent muslim for each country came from pew:

both analyses indicate: “that Islam has both direct effects on democracy and effects that are mediated by consanguinity, although the direct path from percentage Muslim to democracy [in the first model] only approached the conventional cutoff for significance (p = .096).”

from the paper (pg. 12):

“The largest impacts on consanguinity in the path models were produced by pathogen load and the effect of the percentage of Muslims within a nation. In the first path model the latter variable did not have a significant direct path to democracy, which suggests that its effects on democracy are largely mediated by consanguinity. Both pathogen prevalence and the influence of Islam have been described in the literature as having an inhibitory effect on democracy (e.g., Fincher et al., 2008; Fish, 2002; Fukuyama, 2001; Huntington, 1984; Thornhill et al., 2009). Here we indicate that these variables, which had previously been posited to have independent effects on democracy, are actually mediated by consanguinity.”

so, if a nation is islamic, that will affect how democratic it is (or not!), but what seems to be more important is if the population practices cousin marriage. it’s islam+consanguinity that is the key here, not just islam.

i think it makes sense that the effects islam has on democracy are “mediated” by how much cousin marriage there is in a society. cousin marriage directly affects the genetic relatedness between the individual members of a population, making individuals more related to their family members than would happen in an outbred society, while making those same individuals less related to non-family members, again unlike in an outbred society. i think this pretty clearly leads to clannish or tribal behavioral patterns which, as woodley and bell point out, are not conducive to liberal democracy at all.

islam doesn’t demand cousin marriage, but it doesn’t prohibit it either. since muslims are supposed to emulate mohammed (who married a cousin – see below), it probably rather encourages it. and anyway — which came first, cousin marriage or islam? yup. cousin marriage. one of mohammed’s wives was a cousin of his (his fzd) — and ali (yes that ali), who was mohammed’s cousin, married mohammed’s daughter, ali’s first cousin once removed. cousin marriage was very much the norm amongst the arabs in mohammed’s day. and, unlike roman catholic church policy makers, neither mohammed nor any imam since him (at least none that count) seem to have come down against cousin marriage afaik.

furthermore, good ol’ father’s brother’s daughter (fbd) marriage, the form of cousin marriage that leads to the most inbreeding, and that is still the preferred form amongst many muslims, was probably already well established amongst the arabs in mohammed’s day. fbd marriage was probably introduced to the arabs by jewish tribes from the levant who migrated into the arab peninsula starting in the second century b.c. so not only is cousin marriage amongst the arabs old, it’s really old — and it’s fbd marriage to boot. the arabs went on to introduce fbd marriage to the peoples of north africa, the mashriq and south asia (like the pakistanis and the afghanis).

my guess is that it’s not just the amount of consanguinity in a nation that negatively affects the success of democracy in that country, but the length of time the people have been practicing cousin marriage AND how close that cousin marriage is. like i said in the previous post, i think the evolution of “genes for altruism” comes into play here, not just the immediate genetic relatedness between the individuals in these societies, although it’s important, too.

so, i would bet that democracy would fare the worst in the levant, where fbd marriage originated, and the arab peninsula, where fbd marriage has been present for so very long, and that distance from that core region would predict better odds of democracy working at all.

kinda looks that way, don’t it? (eui democracy index 2011 – click on map for LARGER view):

syria, saudi arabia, yeman and oman have the worst scores for democracy in the muslim world (in the world!). iran, turkemenistan and uzbekistan have similar scores and all three of those countries were “arabized” in the early- to mid- seventh century a.d. pakistan was not brought under the arab sphere of influence until later (the early eighth century) and conversion to islam and arabization (and, presumably, the adoption of fbd marriage) took some time. this, i think, might partially explain why, even though pakistan today has similar consanguinity rates to saudi arabia, it does better as far as having a democratic state goes — the pakistani populations haven’t been marrying their fbd for as long as arabs.

similarly, at the other end of the “arab” world, north africans are relatively better at democracy than the saudis since they, too, were arabized — and adopted fbd marriage — comparatively late. the far flung islamic nation, indonesia, manages democracy fairly ok since they’ve hardly adopted fbd marriage at all, although they’ve probably been marrying their mother’s brother’s daughters for a while like other east asian populations.

previously: consanguinity and democracy

(note: comments do not require an email. albatrosses!)

Advertisements

32 Comments

  1. I’ve been thinking more about consanguinity and cruelty. But I’m still a little confused about China. Won’t polygamy and exogamous patriarchal clans lead to inbred villages even in the absence of (a lot of) cousin marriage? What is the limit there?

    Reply

  2. A terribly poetic translation. The water margin is a much more accurate title, and outlaws of the marsh a much more descriptive one.

    By the way Luke Lea, do you have any more links to primary data about asian under-performance vis-a-vis test scores? All I’ve managed to find are monomoniacal postings by Sineruse Steve Hsu’s who has the irritating habit of dropping names without ever actually giving the exact paper title. I’ve actually read two papers from an author and year mentioned, finding the only two relevant papers on the topic published that year and my reading of it is that the he is mis-interpreting the data, possibly deliberately.

    http://public.econ.duke.edu/~hf14/ERID/Arcidiacono_et_al.pdf

    http://public.econ.duke.edu/~psarcidi/grades_4.0.pdf

    In the first paper the data does point to a slight asian over-prediction of first year GPA (table 2 page 16). However this was not predicated on test scores but relied on instead of asking the students themselves what they thought their first year cumulative GPA would be. Secondly the actual first year cumulative GPA for Asians was higher than that for whites according to the raw CLL (table 1 page 13). Keep in mind that the white and asian scores were from a randomized sample and not field of study specific.

    The second paper primarily addresses black/white discrepancies and the only mention of Asian is in the beginning with a nice colorful graph. Asian GPA’s are higher than whites for the Freshman and Junior years, lower for Sophmore and Senior. More critically the author explicitly states in the paragraph below graph does not account for either variance (major switching, droping out) and course selection. The author also explicitly states that natural science GPA are on average 8% lower than humanities GPA’s (A 0.25 GPA difference when using 3.5 as the benchmark). This seems to be a rather deliberate lie as the blog posts were claiming that the GPA’s were being weighted for course selection.

    There is a third paper I found that was referenced incorrectly.

    http://www.princeton.edu/~tje/files/webAdmission%20Preferences%20Espenshade%20Chung%20Walling%20Dec%202004.pdf

    This was by Espenshade, Chung, and Walling (not Radford) that discusses minority, athlete, and legacy admissions at three elite universities. This probably was the most egregiously erroneous reference of them all. Not only does the paper contain nothing about asian college GPA’s vis-a-vis whites. It was singularly focused on the admissions process itself and not the results. All reference to GPA’s and class rank were about the high school level and how they factored into admissions. The conclusions drawn by the authors themselves are exactly the opposite of what Sineruse claims.

    Reply

  3. @luke – “Won’t polygamy and exogamous patriarchal clans lead to inbred villages even in the absence of (a lot of) cousin marriage?”

    polygamy, yes, ought to lead to a population being more inbred — there will be more half-siblings overall than in an outbred populations.

    exogamous patriarchial clans is just mbd marriage by another name, i think — see the new post here. (^_^)

    Reply

  4. @ hbd – exogamous patriarchial clans is just mbd marriage by another name?

    Can you explain that in one sentence in plain English? If a patriarchal clan is importing wives from a variety of other nearby clans doesn’t the inbreeding quotient depend on how many and how unrelated those other clans are? I would imagine this would lead to a spectrum, the extreme case being female exchange between just two clans. What the situation is empirically both in the past and today I don’t know but currently there are more than 500,000 villages 50,000 townships (order of magnitude) in China so the variation must be considerable.

    Reply

  5. Duke of Qin – cool name! Right, Marsh Men is the standard title and besides “All Men Are Brothers” does not imply that these bandit groups were all of the same clan, so me bad. Do you happen to know how closely related bandit groups typically were?

    As for the under performance issue, all I remember is that one commenter on Steve Sailer’s blog. He seemed to be well informed with supporting data. Did you find him?

    Reply

  6. One more point on under performance: people who do a lot of test prep would be expected to under perform. Their scores do not reflect their natural endowment to the extent they would for people who walk in cold and make the same scores. Wouldn’t you agree?

    Reply

  7. I should point out that a township in China is an administrative unit containing a number of villages, ten on average (roughly speaking), each village with 1500 people in a hundred or more households depending on number of male children in last three or four generations, typically around 5 today. This from memory and recent reading.

    Reply

  8. I meant the average household has approximately 5 members on average. Obviously there will be variation. The one child policy has been unevenly enforced. Less enforced in rural areas than cities, and among cadres and party members than common people. I could be wrong about that, but have seen these things mentioned in my reading. I look out for stuff like that.

    Reply

  9. No I don’t have any idea how closely related Chinese bandit groups are to one another but I suspect highly. A criminal enterprise is generally not conducive to trust and what is there more trustworthy than blood?

    As for the commenter on Steve Sailer, they were also referencing Sineruse if not Sineruse themselves.

    Any fairly intelligent slacker used to flying by the seat of their pants and bullshiting their way through the abridged cliff notes version easily becomes adapt at spotting other people’s bullshit. That someone kept name dropping authors without giving the exact paper name was a yellow flag.

    Here is one claim made regarding Asian underperformance

    http://infoproc.blogspot.com/2011/12/differential-validity-of-sat.html

    2nd comment

    “This is consistent with the Espenshade and Duke University (Arcidiacono et al 2011) studies of grades, where white was used as the baseline, and being Asian (American) was a negative predictor for college grades, after accounting in some fashion for choice of major or course difficulty, HS academic variables, and socioeconomic characteristics.”

    This is contradicted by both Arcidiacono papers (the only 2 papers in 2011 which cover anything close to a related topic)i linked to in my earlier post which makes no such claim.

    http://infoproc.blogspot.com/2011/12/asian-admissions-statistical-prediction.html#disqus_thread

    A comment 20 or so down

    “Espenshade-Chung-Radford’s study, among others, found substantial overprediction of Asian college grades. Their most interesting finding was that for prediction given SAT, HS GPA, some socioeconomic indicators, field of college major, athelete or legacy status, *and first year college class rank* there was still a negative Asian effect on class rank in the later years (i.e., in the more advanced courses). The effect was larger than that of being an athlete or legacy and equivalent to more than the famous “140 SAT points” when quantified in the manner that so impressed US News, of holding all other factors equal.”

    There is no study by Espenshade, Chung, and Raford. There is one by Espenshade,Chung, and Walling that studies the impact that minority status, athlete status, and legacies have on admissions. There is no data in the study at all about college GPA’s. Rather the study contains only high school GPA data and class rank and its impact on admissions to colleges. It does not show any Asian under-performance at all and the conclusion is in fact the opposite.

    All the comments I’ve found on the internet about Asian underperformance vis-a-vis whites have pointed to pretty much this one individual who it seems, after taking the time to read the authors he name drops and the studies obliquely mentioned, is simply a big fat dirty liar.

    Reply

  10. Duke of Qin: This is the comment on Steve Sailer’s blog I referred to:

    Anonymous said…

    I’m copying and pasting a very insightful comment by Sineruse on Steve Hsu’s blog that nobody has rebutted yet.

    “You have that backwards. East Asians underperform at the higher level selections relative to credentials, for the same reason that their absolute performance is higher on the earlier and lower-level metrics (such as SAT, grades, HS graduation rate, college admission results). This is unavoidable no matter whether E.Asians have more or less ability than whites, as long as they tend to “work harder” at any given level of ability. The Asian discrimination complaints would have us ignore this phenomenon, as though all Asian high performance is a result of pure ability with no group differences in effort or investment biasing the results when compared to whites. It is a popular gambit to claim that “ETS studies show” that test prep rarely improves scores by much, even though every East Asian immigrant enclave has a thriving multitude of test prep centers based on the opposite experience. Examples of how this works out in reality on objective measures harder than the SAT:

    – declining East Asian-American to white ratio in US math competitions as you follow a cohort from grade 8 to high school to college. e.g., proportionally fewer E.Asian than white US IMO team selectees get gold medals; the number and ranking of US E.Asian high scorers relative to whites (even within the same set of students, so as to control for dropouts from the pipeline) takes a sharp decline from USAMO to the Putnam; no E.Asian winners of the Morgan Prize (top national undergraduate math research award) even though most white and Indian winners are former competition winners.

    – the same for grad school fellowships (compare the tiny proportion of Hertz fellowship East Asian American recipients to the 60+ percent share of Siemens AP scholars or high school science contest winners); Ivy League valedictorians and Phi Beta Kappa numbers relative to the Asian share of the selection pool (e.g., the top 200 students per year by SAT+ high school GPA, at each university); or professional prizes and tenure appointments in theoretical math/physics/CS.

    – a mountain of Law School Admissions Committee studies (see their website) on LSAT, law school GPA, and bar exam passage (and failure…) rates. Combined LSAT + GPA regression overpredicts bar exam pass rates for Asians. Asian GPA gets lower year by year in law school compared to whites. High verbal SAT scores are 2x more common for Asians than whites in high school, but this advantage disappears on the LSAT, and the Asian LSAT distribution as a whole is well below that of whites, both for the 1990’s and much more recent data. The distribution looks like a mixture of two subgroups, gen-1 immigrants + native English speakers, but this does not explain why the latter category that has the same upper LSAT score distribution as whites, lost its huge advantage from the high school verbal SAT results. That is, Asian verbal SAT scores “overpredict” LSAT performance, and Asians underperform their college-level credentials in law school and beyond……….

    12/19/11 12:25 PM
    Anonymous said…

    continued ….
    – As with the LSAT, immigration and socioeconomics does not account for the bar passage rates. Whites in the lowest income level considered passed the bar exam at rates higher than the wealthiest category of Asians. Control for difficulty of bar exam in different regions did not make any of the negative Asian effects go away. The Asian disadvantage in LSAT and LGPA and bar passage has been a robust pattern for as long as LSAC has asked the questions. The implication is that a meritocratic admissions policy would apply an Espenshade-style penalty to Asian LSAT scores. I assume this is illegal, but it is a politically incorrect academic reality if the past 20 years of studies, all singing the same tune, are correct.

    – medical licensing exam pass rates (USMLE step 1) are overpredicted for Asians based on GPA and MCAT score. Authors of one of the studies that pointed this out suggested to use a different medical school admissions formula for Asians, so as to avoid the disastrous scenario of several years’ med school plus loans followed by a failure on USMLE. Those authors were mostly East Asian, and their proposal amounts to applying yet another Espenshade-style penalty of some number of virtual MCAT points for Asians to more truthfully assess their prospects after medical school. Whites not being able to compete under a meritocracy was not a concern of the study.

    In addition to objective underperformance, there has been no subjective sign of any CCNY-style overperformance from East Asian-Americans at the UCs where, if discrimination existed, many highly qualifed Asians locked out from Stanford and the Ivy Leagues would find refuge. From the data I looked at (admittedly not complete), it does not appear that Asians at Berkeley are even reaching their population share of NSF fellowships and academic prizes compared to whites. Note that to assess population share you need not the W:A ratio of enrolled students but the one for students with high average SAT score, where the Asian proportion is much higher.”
    12/19/11 12:25 PM
    Anonymous said…

    My own comment on Siserune’s analysis:

    I do believe Indians have low mean IQs, but I also suspect Indians perform (in terms of productive research) about the same as Whites of a similar IQ. As you go from elite contests to elite research awards, Indians and Chinese in the US are about equal in numbers (Hertz winners, IEEE Fellows, ACM Fellows, ACM Doctoral dissertation winners, Morgan Prize winners, Cole Prize, etc) and are dwarfed by Jews/Whites. Yan Shen will continue to claim that Asians are uber geniuses, but there is little evidence for such overconfidence. Here’s what I think is true: for a Chinese to perform as well at research at a White/Jew, he/she needs significantly more cognitive horsepower as measures by g. To quote Christopher Chang, Asians are good at optimizing, not theory building. Something RKU once worried about on a Mangan’s thread, but refused to repeat on Steve Hsu’s blog :)
    12/19/11 12:26 PM

    Reply

  11. @the duke – “No I don’t have any idea how closely related Chinese bandit groups are to one another but I suspect highly.”

    something that needs finding out.

    Reply

  12. @luke – “‘exogamous patriarchial clans is just mbd marriage by another name?’

    Can you explain that in one sentence in plain English?

    well, i just meant that if a patriarchial clan is not marrying within patrilineages (fbd marriage), then they must be marrying their kids off to the mothers’ relatives (mbd marriage). but, now that i’ve thought about it some more, that’s not exactly right for china.

    if you have a custom saying you can’t marry anyone with the same surname (china), but you still want to marry a cousin, you’ve got to marry a maternal cousin or you could marry your father’s sister’s daughter as she’ll have her father’s surname. so there is another option there.

    however, there does seem to be a preference for the maternal side in china, though, at least in modern times.

    @luke – “If a patriarchal clan is importing wives from a variety of other nearby clans doesn’t the inbreeding quotient depend on how many and how unrelated those other clans are?”

    absolutely. if you always get your brides from just one other clan (which you would think mbd marriage would push towards), the two clans would be more closely related than if you got your brides from ten different clans.

    if you always follow the mbd pattern, though (which was hardly the case), then you should always be drawing your brides from just one other clan. there’s no way around that that i can see.

    Reply

  13. @luke – “What the situation is empirically both in the past and today I don’t know but currently there are more than 500,000 villages 50,000 townships (order of magnitude) in China so the variation must be considerable.”

    yes, i agree. one broad pattern that i think exists — and this is only an impression i have based on the little that i’ve read so far — is that historically there’s been more cousin marriage amongst the southern han than the northern. not certain that that’s 100% correct, but that’s my impression so far.

    Reply

  14. Let’s pose the China inbreeding issue this way: there are on the order of 50,000 townships, each township typically consisting of around ten villiages. A typical village may have one or two (or three?) clans with a total population of around 1500 people living in two or three hundred households. These villages clans have typically lived where they do now for centuries. Assuming they usually marry withing the township, that means a breeding population of around 15 or 20 thousand people (obviously this has increased with general population increase over last 300 years). The total number of clans in a township would be on the order of 15 or 20 (one or two per village). If everything were random except for rule of clan exogamy, what would be the inbreedin quotient after x generations? Would there be a tendency to choose mates from villages that are closest in proximity? How about another clan in the same village. Marriages were arranged, keep in mind, and were often conceived of as family alliances if I am not mistaken. There were dowries involved, right? I get the idea that families didn’t completely lose track of their daughters in case the marriage didn’t work out.

    Maybe somebody can do the math. I can’t. But my intuition tells me that there will be a high inbreeding quotient given enough generations and assuming clans stay in place for many centuries.

    Reply

  15. No I don’t have any idea how closely related Chinese bandit groups are to one another but I suspect highly.

    Given the name “water margin” i’d guess they were a clannish group living in a swamp, like the Cajun in the US, reinforced by outlaws on the run from other areas.

    .
    if you always follow the mbd pattern, though (which was hardly the case), then you should always be drawing your brides from just one other clan. there’s no way around that that i can see.

    http://home.cc.umanitoba.ca/~anthrop/tutor/marriage/matxcuz.html

    If brides always go from A to B from B to C from C to D and lastly from D to A then each patrilocal clan always keeps the name and always gives and recieves brides in only one direction but over generations the genes get mixed up because after one generation
    – the A children are now AD children
    – the B children are now BA children
    – the C children are now CB children
    – the D children are now DC children

    so in the next iteration
    – AD brides marry BA husbands
    – BA brides marry CB husbands
    – CB brides marry DC husbands
    – DC brides marry AD husbands

    so their children
    – A clan’s children are now AABD
    – B clan’s children are now BBAC
    – C clan’s children are now CCBD
    – D clan’s children are now DDAC

    so in the next iteration
    – AABD brides marry BBAC husbands
    – BBAC brides marry CCBD husbands
    – CCBD brides marry DDAC husbands
    – DDAC brides marry AABD husbands

    so their children
    – A clan’s children are now AAABBBCD
    – B clan’s children are now ABBBCCCD
    – C clan’s children are now ABCCCDDD
    – D clan’s children are now AAABCDDD
    etc

    #

    Alternatively hubchik posted some stats from (i believe) aristocratic families in medieval Italy. If i recall right the pattern seemed to suggest four allied families with two main pairs i.e. four families A, B, C and D with A&B and C&D having each other as main ally so
    – family A married B c50% of the time and families C and D c25% of the time each
    – family B married A c50% of the time and families C and D c25% of the time each
    – family C and D married each other 50% of the time and family A and B 25% of the time each

    This would make sequential sense because if

    http://home.cc.umanitoba.ca/~anthrop/tutor/marriage/bixcuz.html

    is the first stage i.e. two familes forming a long term marriage alliance, then the next logical step if they needed to build a bigger alliance would be for two of those pair-clans to form a doublepair-clan.

    .
    But my intuition tells me that there will be a high inbreeding quotient given enough generations and assuming clans stay in place for many centuries.

    MBD is still inbreeding. It’s just aiming at a larger group of related allies so the total breeding pool is a bit larger.

    Reply

  16. Grey wanderer, no that is not really correct. Outlaws of the Marsh is a work of fiction based on actual historical figures akin to the Robin Hood legend. The bandits in the novel were, apart from some brothers, not related at all. In fact, it valorizes friendship over kin duty. One character literally states that his wife and children are replaceable, of less worth than the sworn brotherhood of men.

    I suspect in reality actual outlaws in medieval China were quite a bit different than the romanticized depictions.

    Reply

  17. Had dinner with my new anthropologist Chinese friend last night. He grew up in China. I asked him about cousin marriage. He said yes, there is a lot of it. He described it as “cross cousin” marriage but said the phrase means something different than usual. Women keep their clan last names when they marry (I didn’t know that). In a village if one woman marries into a clan it is common for her sisters to do likewise. Into the same family or same village? At this point the children interrupted so I didn’t get an answer to that last question. We got along so I expect to pursue this further. He’s supposed to recommend me some books on the subject.

    Reply

  18. BTW, the Chinese anthropologist’s wife is 5′ 9″ tall. I asked her if that was unusual. She said not at all for the current generation. Who knew? (She was quite beautiful too — her skin was whiter than white.)

    Reply

  19. @luke – “my intuition tells me that there will be a high inbreeding quotient given enough generations and assuming clans stay in place for many centuries.”

    yes — i, too, think that must be right. (i have a quote related to this from a phd thesis that i just got through reading — i’ll post it tomorrow.) as for all your mathematical questions (which are all good ones!) — i’ll have to sit down and think (hard) about those one day. (^_^)

    @luke – “Oh, and I forgot: throw in some polygamy.”

    yup! that, too.

    Reply

  20. @g.w. – “MBD is still inbreeding. It’s just aiming at a larger group of related allies so the total breeding pool is a bit larger.”

    exactly.

    there seems to be a range of inbreeding from 1) just general geo-endogamous marriages with little or no focus on marrying family, to 2) slightly more inbreeding like the scots-irish/greeks do where you do marry in the clan, but it’s more likely to be a third- or fourth-cousin, to 3) seriously focusing on mbd marriage, to 4) seriously focusing on fbd marriage.

    there should be more variations on my spectrum there — those are just the main ones that spring to mind. plus, polygamy, too.

    Reply

  21. @luke – “He described it as ‘cross cousin’ marriage but said the phrase means something different than usual.”

    huh. wonder what he meant by that? mbd marriage is certainly cross cousin. there is some mzd (mother’s sister’s daughter) marriage and fzd (father’s sister’s daughter) marriage in china, too, although less of both of those. fzd is also cross cousin.

    @luke – “Women keep their clan last names when they marry (I didn’t know that).”

    oh, i didn’t know that, either.

    @luke – “In a village if one woman marries into a clan it is common for her sisters to do likewise. Into the same family or same village?”

    into the same family would make sense if folks are really trying to abide by the mbd rule — but that’s not always possible, of course, because there are not always enough available sons/daughters.

    @luke – “He’s supposed to recommend me some books on the subject.”

    excellent!

    thanks for sharing! (^_^)

    Reply

  22. Duke
    “Outlaws of the Marsh is a work of fiction based on actual historical figures akin to the Robin Hood legend.”

    Ah, my clever deduction wasn’t quite as clever as i hoped.

    Reply

  23. I found this thread while searching for some data I had posted at Steve Hsu’s blog.

    Flailing rebuttals from China Pride warriors (in this case, Han supremacist “Duke of Qin”) are a dime a dozen in any discussion of US East Asian underperformance, the goal being to shout down unpleasant data. This has always ended badly for the China Pride trolls when the data analyses and references to literature turn out to be correct. But since DoQ went to the trouble of looking at some of the sources, and has made many specific (and false) statements about those papers, I’ll answer his remarks here.

    1. Qin: “There is no study by Espenshade, Chung, and Raford. There is one by Espenshade,Chung, and Walling that studies the impact that minority status, athlete status, and legacies have on admissions.”

    Reply: if you search for “Espenshade Chung Radford” you will discover as one of the top hits the 2009 book by Espenshade & Radford (Chung did the statistical work). This book is the source of the entirely bogus but much reported “140 SAT point” figure for a negative Asian effect in admissions, and it also contains regression analyses finding Asian underperformance relative to whites AFTER the admission. If you then search within the book for the words “OLS Regression of Percentile Class Rank” you can see that, with some controls for field of major, SAT, HS GPA and other student characteristics, the study found is a strongly negative Asian effect on 4-year college GPA, and there remains a negative effect when including first-year college GPA. (i.e., there is a decline during the college years, as also found in the Duke CLL studies — more below).

    2. Qin: “There is no data in the [E,C & W] study at all about college GPA’s. Rather the study contains only high school GPA data and class rank and its impact on admissions to colleges. It does not show any Asian under-performance at all and the conclusion is in fact the opposite.”

    Reply: I never cited the Espenshade-Chung-Walling study, but your comment on underperformance is incorrect. The paper has some data on the higher proportion of Asians in the applicant pool with high credentials, but that is neither “over” nor “under” performance. The over/underperformance question is how do (US East- ) Asians perform *relative to those credentials* after the selection, in comparison to whites. In the jargon of the statistical studies, Asian credentials tend to “overpredict” later performance. It would be surprising to see any study in the modern era of intensive Asian test prep that does NOT find overprediction relative to whites. Either you believe that all that test prep (and additional studying, and an earlier start, and weekend schools, and summer schools, etc etc) accomplishes nothing, or the credentials are less predictive for Asians in ways that can be detected statistically.

    3. Qin: ” http://public.econ.duke.edu/~hf14/ERID/Arcidiacono_et_al.pdf

    http://public.econ.duke.edu/~psarcidi/grades_4.0.pdf

    In the first paper the data does point to a slight asian over-prediction of first year GPA (table 2 page 16). ”

    Reply: I referred to the second paper only. The first paper is about whether applicants know anything the admission office does not when evaluating their abilities, and the answer was generally that Asians over-rate themselves (i.e, their predicted GPA) relative to whites. This is not directly relevant to the underperformance discussion except as yet another indicator that Asian families are not unbiased determiners of which students in their area were most qualified for admission. The study also indicates that the admission office does not under-rate Asian applicants; if they did, it would manifest as an Asian overperformance effect. So in two respects the first study does contradict some sacred cows of the US E.Asian community, though on a medium-size sample at one school. Continuing to the second paper, which does deal with underperformance…

    4. Qin: “The second paper primarily addresses black/white discrepancies and the only mention of Asian is in the beginning with a nice colorful graph. Asian GPA’s are higher than whites for the Freshman and Junior years, lower for Sophmore and Senior. More critically the author explicitly states in the paragraph below graph does not account for either variance (major switching, droping out) and course selection. ”

    Reply: The second paper (title is “Time Path of …”) is about changes in GPA over time and does contains findings on Asians whether or not that was the goal of the study. Indeed the Asian raw GPA starts out higher and ends up lower than whites’, which is one crude population-level indicator of underperformance, but the study is much more precise than that. It uses students’ pre-admission credentials and admission office ratings, and then tries to account for the difficulty of every single course at Duke, and even for the difficulty of competition in each class (the ability levels of the other students enrolled). The authors create a variable for every student’s estimated ability at every semester, and every single course grade awarded is modelled as a function of overall class difficulty (average GPA awarded in that class) plus individual student’s ability plus ability of the other enrolled students in that class that semester.

    That is the finest-grained control for course difficulty ever attempted in a GPA study. In addition, their method favors Asians by treating the socioeconomic variables, such as parents having highschool, college, or graduate education, as having the same meaning for US whites and immigrant Asians when (due to the lesser availability of college education in East Asia, and the immigrant parents tending to be scientists) the Asian parents’ educational strata mean more. Despite this, the Duke researchers still found Asian underperformance, which in their methodology literally means that the “estimated ability” parameter for individual Asians tends to decline over the four years. A diagnosis of inflated credentials doesn’t get any more specific than that.

    5. Qin: “The author also explicitly states that natural science GPA are on average 8% lower than humanities GPA’s (A 0.25 GPA difference when using 3.5 as the benchmark). This seems to be a rather deliberate lie as the blog posts were claiming that the GPA’s were being weighted for course selection.”

    Reply: The Duke study had extremely detailed controls for course selection, as I just explained in some detail. The Espenshade-Radford-Chung study had controls for type of major (Engineering, Science, Humanities, …) and the Asian effect they found was strong enough to not be explainable by differences in selection of majors within the type. The Asian underperformance coefficient in one the E-R-C regressions was more than the effect of being an athlete or legacy admission, and more than the famous 140 SAT points. That is, if you accept the bogus methodology they used for constructing the 140 point estimate, and apply it to their class rank results, you get a result larger than 140 for the number of points Asian SATs should be discounted when predicting future class rank.

    Reply

  24. @sineruse – “But since DoQ went to the trouble of looking at some of the sources, and has made many specific (and false) statements about those papers, I’ll answer his remarks here.”

    thanks! dunno if the duke is still checking out this thread, but the resurrection of old comment threads — and the general sharing of interesting info — is always welcome here on hbd chick. (^_^)

    Reply

  25. Very interesting but surely there must be some acknowledgement of the industrial development of a country. For example Oman – a country barely a generation out of the desert, indeed barely a country. It would be very unusual for ‘democracy’ to have developed there in such a short time. Sure there are a lot of cousins there as in the more developed UAE where the economy is far more developed – yet not much sign of democracy not least as there is arguably too much wealth. Everyone (migrants aside) is very well looked after, no taxation – who needs to bother about a vote? Why rock that very steady boat?

    Reply

  26. I asked Duke Of Qin about his reply (if any) to Sineruse, in a comments discussion ongoing at Steve Sailer’s website.

    Q http://www.unz.com/isteve/act-score-gaps-we-didnt-know-it-was-this-bad/#comment-1998175
    A http://www.unz.com/isteve/act-score-gaps-we-didnt-know-it-was-this-bad/#comment-1998534
    A2A http://www.unz.com/isteve/act-score-gaps-we-didnt-know-it-was-this-bad/#comment-2003377

    To keep the information in one place, I extract here the part of his answer that addresses Sineruse’s comment above.

    ~~~~~~~~~~~~~~~~~~~~~~~~~~
    Duke of Qin:
    (Sineruse wrote) “The Duke study [http://public.econ.duke.edu/~psarcidi/grades_4.0.pdf ] had extremely detailed controls for course selection, as I just explained in some detail”

    Actual paper authors wrote page 3, paragraph 2: “There are, however, at least two reasons to be skeptical of Figure 1: variance and course selection.” and ” Note that these averages do not take into account selection into courses”

    …(also, ) these categories are for a broad “Asian” and not specifically “Chinese”.

    ~~~~~~~~~~~~~~~~~~~~~~~~~~~

    My comment: that is not a serious answer to Sineruse, whose quoted comment seems to be about the regression analysis (Table 6, page 15) of a class rank measure that is controlled for course selection. Qin’s reference point, Figure 1 from the introduction, graphs the group average GPAs based on uncontrolled grades, but as the authors (and DukeOfQin) explain, it is necessary to add controls to get a more meaningful comparison, which is why the rest of the paper is done with the controls.

    Reply

Leave a Reply

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

WordPress.com Logo

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

Google photo

You are commenting using your Google 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 )

Connecting to %s