“The Inside Story Of The Harvard Dissertation That Became Too Racist For Heritage”

Mark Palko points me to a news article by Zack Beauchamp on Jason Richwine, the recent Ph.D. graduate from Harvard’s policy school who left the conservative Heritage Foundation after it came out that his Ph.D. thesis was said to be all about the low IQ’s of Hispanic immigrants. Heritage and others apparently thought this association could discredit their anti-immigration-reform position. Richwine’s mentor Charles Murray was unhappy about the whole episode.

Beauchamp’s article is worth reading in that it provides some interesting background, in particular by getting into the details of the Ph.D. review process. In a sense, Beauchamp is too harsh. Flawed Ph.D. theses get published all the time. I’d say that most Ph.D. theses I’ve seen are flawed: usually the plan is to get the papers into shape later, when submitting them to journals. If a student doesn’t go into academia, the thesis typically just sits there and is rarely followed up on. I don’t know the statistics on this, but I’m guessing it’s a typical pattern for a policy school Ph.D. to go into the policy world, not academia, and so then the details of the thesis won’t be taken so seriously. At some point, the goal is for the student to graduate, it’s not required that the thesis have all its holes plugged. That can be done in the submission-for-publication stage, if that ever happens.

We last heard from Richwine on this blog a few years ago, in the context of an article where he tied himself into knots on the topic of intelligence and politics, ultimately arguing that Saul Bellow’s aunt (?) was more politically astute than Bellow, even though, in Kristol’s words, “Saul’s aunt may not have been a brilliant intellectual.” Huh? We’re taking Richwine’s testimony on Saul Bellow’s aunt’s intelligence?

The point is, Richwine has had conflicted views on IQ and politics, seemingly undecided about whether to take the line that intelligent Americans mostly have conservative views (“[George W.] Bush’s IQ is at least as high as John Kerry’s” and “Even among the nation’s smartest people, liberal elites could easily be in the minority politically”) or the fallback position that, yes, maybe liberals are more intelligent than conservatives, but intelligence isn’t such a good thing anyway.

P.S. On a related issue, here is my discussion of Murray’s recent book on the upper class.

85 thoughts on ““The Inside Story Of The Harvard Dissertation That Became Too Racist For Heritage”

  1. The irony is it requires almost no intelligence – (it’s remarkably obvious) – to suggest “The bottom line is that a political debate will never be resolved by measuring the IQs of groups on each side of the issue.” And yet that’s his key takeaway.

    Anyway, it doesn’t seem extraordinary at all to think that GWB is as smart (if not smarter) than John Kerry. I would also guess on first approximation “elite” circles tend to seem more liberal than they are. Because it’s not the done thing to be an elite educated guy who denies evolution, or whatever. (Or, to put it another way, it isn’t politically correct to contradict certain liberal stylized facts). This is especially true of the asymmetry at the extremes. It might be okay to be your proverbial “fiscal conservative social liberal” [which, nowadays, is what the signal for a conservative running in educated, liberal circles?] who basically supports all the premises of the liberal agenda, but definitely not someone who’d dare support the Ryan plan.

    I went to a debate party (the one with the bayonet joke) at Harvard last year. It was “co” sponsored by the young democrats and republicans or whatever. I counted 4 people – in hundreds – that actually stood up and cheered when Romney walked. I would also have been quite socially ostracized, I imagine, from supporting such a ridiculous campaign.

    Now that’s all fair (because the Romney campaign did run on certain policies that should be ostracized), but it has nothing to do with IQ. And it does not represent the Harvard student body.

    Which, incidentally, is why Niall Ferguson is (or at least was) so important to the GOP donor agenda.

    • Oh and another part of the irony is that Charles Murray (who Richwine loves in the TP article) in Coming Apart basically uses “elite” as a place holder for “liberal, Ivy League, Democrat elite”. Or that’s my take. And that’s very contra-Richwine.

      Of course it’s important to note that in Murray’s world an accomplished writer with a snazzy blog following living on $23,000 a year is above a plumber earning $55,000.

      Anyway me and Murray were born on the east coast. And you teach at Columbia. Whatever we consider to be “elite” is probably a far, far cry from what most of America considers elite. For example I would expect very, very little support for the liberal cause in a highly exclusive country club in the Midwest. And that’s “elite” in every single way. Just more southern.

      Also, I don’t think Murray himself has a coherent definition of what he considers to be “upper” and “lower”. Sometimes it’s the “mass affluent” that most definitely includes the solid conservative heartland of nuclear, white, families earning a solid amount in boring jobs. But without much alert it can change to the New Englandy, Harvard-esque elite that’s very, very different. Opinions on religion, for example which are endorsed strongly by the former but probably ridiculed by the latter.

      But there’s a big part of the elite that is conservative. Economically and professionally.

    • There seems to be a lot of field to field variation on these requirements, as well as country to country variations. But I have seen a lot of programs where having the papers published is considered a serious problem for the dissertation. But I have also seen places where being published is required to graduate.

      It is a mess.

      But my modal experience is close to Dr. Gelman’s experience — most PhD theses have some issues that need to be addressed but there is a lot of pressure to get the student out of the program and into a post-doc where become a lot more cost effective (a 50% time RA is often as expensive as a 100% time post-doc due to tuition).

      • >>>having the papers published is considered a serious problem for the dissertation<<<

        Interesting. I'd never heard of this. So early publication becomes a liability?

        Which programs are these? I wonder what their rationale is behind deterring publication.

        • I have seen this approach in very book thesis heavy programs that seemed to think that the work should be completed (i.e. defended) before it went out for review. It’s definitely in decline in the modern publishing paradigm.

    • In many journals, particular old-school journals, you give up all copyright in your work when you publish it with them, and you would actually be breaking the law by publishing your paper in your dissertation. Sometimes the journal is nice enough to license it back to you for use in a dissertation, but not always. Sometimes journals will consider publication in a dissertation to be “publication” and wouldn’t accept the “paper version” as novel anymore.

      It seems to me perfectly reasonable that people publish their work in the form of a dissertation before publishing it in peer reviewed journals. There are all kinds of reasons, from slowness on the part of collaborators or advisors to actually prepare the work for publication, to a certain number of people having no real interest in academia, people may be scooped in the last few months of their PhD and simply need to finish up what they’ve done so they can move on to a new topic, in some fields (econ apparently for example) publication of a paper might be a 2 or 3 year long process, there’s no point in making people stay in their PhD long after they’ve done the work just waiting for a stupidly slow bureaucratic process to occur.

      • Do you have examples of journal-policies which are restrictive in the ways you described?

        (a) not accepting papers which were included as sections of an already defended thesis OR

        (b) not allowing an author to include a published paper in his dissertation.

        I’ve never encountered either (in Engineering) so am curious to know more.

        “It seems to me perfectly reasonable that people publish their work in the form of a dissertation before publishing it in peer reviewed journals.”

        I agree. What I don’t agree is allowing a grossly reduced standard of scrutiny for the Thesis versus peer review. This isn’t about flawed theses, but about the effort PhD committee members invest in a thesis assessment. I think faculty ought to at least attempt more. I’ve often encountered a very cursory effort. A part of this is the absolute lack of accountability when bad theses go through (because no one reads them after wards).

  2. (1) Was he fired for his views or his competency? i.e. If I wrote a misogynist or antisemetic pamphlet I might still get fired from my accounting job. (no one likes a PR disaster on its roster).

    (2) Regarding, Andrew’s “PhDs-are-often-flawed” point, are we even so sure this same work may not have been accepted by a (reputable) journal? I’m not so sure. Bad work often makes it to journals too simply because so very often the standards of what’s “bad” are so subjective.

    (3) Sometimes faculty members must face a very tough dilemma: Whether to destroy a students career (and 5-6 long years of effort) by outright rejecting a marginally flawed thesis or to countenance it in the light of the knowledge that mostly no one ever reads your theses again, anyways. Sadly, this was an exception.

    (4) Is there a overwhelming consensus among Professors (in the relevant area) that the work was flawed? Even if, say 10% of (reputable) academics think the work is OK, there’s no reason, in hindsight to question the thesis. A PhD thesis is intended to be on the frontier of an area so a certain degree of uncertainty and disagreement is natural. The standard of evidence should not be “beyond all reasonable doubt”

    PS. Harvard didn’t retract his degree, right? I say rightly so (even though I don’t agree with his work and conclusions)

  3. Andrew, the dissertation was not flawed, nor is it fair to say that it was “all about the low IQs of Hispanic immigrants.” Please see my response to the Beauchamp piece here:

    http://thinkprogress.org/justice/2013/05/24/2059311/jason-richwine-responds-on-race-iq-and-his-dissertation/

    On the “Are liberals smarter than conservatives?” article: I made two simple points about why comparing the IQs of liberals and conservatives is not likely to help resolve any political disputes. The first point was philosophical: A high IQ does not necessarily lead to better political choices. The second point was statistical: A positive correlation between IQ and liberalism does not necessarily imply that liberalism is the preference of most smart people. Surely you agree with both points? I wasn’t trying to reach a unified field theory of intelligence and politics, only to throw cold water on a trope that is especially (but not exclusively) popular on the Left, i.e., “Our side of the aisle is correct because we’re smarter.”

    • What does the IQ of Hispanic immigrants have to do with the (supposedly) “popular trope on the Left, i.e., ‘Our side of the aisle is correct because we’re smarter.’”???

    • Dr. Richwine,

      Would you be so kind as to publish your dataset and code that would reproduce the tables in your dissertation?

      Thank you.

    • Jason:

      Thanks for the note. As I have not read your dissertation, it is inappropriate for me to write that it is “all about low IQs of Hispanic immigrants.” I rephrased above to write that your thesis was “was said to be” on that topic, which is all I can really say.

      Regarding your earlier article, I agree that you were not attempting to give any all-encompassing theory. But I think your story of Saul Bellows’s aunt indicates that there are some convolutions in your argument.

    • In my experience, the “liberals are stupid; conservatives are evil” trope is much older and more widespread than this “we’re smarter” trope, especially among conservatives. It sounds like you’re dealing with a straw man.

    • Here’s a link to the full thesis’ pdf:

      http://www.scribd.com/doc/140239668/IQ-and-Immigration-Policy-Jason-Richwine

      I felt a lot of discussion here was hearsay (e.g. Andrew’s “As-I-have-not-read-your-dissertation….”), so maybe it will help people read (or at least skim through) the thesis to get a better feel for it.

      I’ve no credentials to judge, but personally it came across as informal and superficial. But maybe that’s just my bias.

    • Dr. Richwine got Richwined not for being wrong but for being right. You don’t get in trouble for saying something wrong but for saying something that seems accurate but nobody wants to admit. There is a huge amount of testing data — IQ, SAT, GRE, AFQT, school achievements, etc — and real world accomplishment levels that show that Mexican-Americans start out on average below average and, while the second generation gets a boost from growing up with English, rates of closing the gap over the generations are poor.

      This is certainly the experience in California, where there have been sizable Mexican-American populations for the last century, yet Mexican-American achievements in huge local elite industries like Silicon Valley and Hollywood are remarkably lacking.

      This suggests that the ruling class’s policy over the last generation of betting the country on Mexican immigration was a mistake. Saying that important people made a mistake is not a pathway to popularity.

  4. The article didn’t, couldn’t or wouldn’t get into the racism within Hispanic or Latino society. It skates above it but I find it difficult to read about the “research” without seeing how it directly reflects the views of the smallish crust within Central and South American society that views itself as “white” and “Spanish” versus the [fill in the blank with curse words]. I have a number of anecdotes that are too disgusting to detail. I also witnessed how consultants to various parties in these countries had to grapple with the blunt racism which dominated internal policy, and which nearly always provided the underlying rationale for why they had the money and power and those other people did not. Stupid, lazy and uneducable – or only within certain limits. It’s rather easy to see how this kind of research would fit the conscious and unconscious agenda.

    I’m waiting for research that says the reason US manufacturing fell is the lower IQs of the people versus those in finance. How can you argue with the shrinking world share of the former versus the explosive growth of the latter. It must be rooted in intelligence. Right?

  5. I was puzzled by Beauchamp’s and Professor von Whatever’s focus on the definitional issues of “Hispanic”. I remember many many surveys and forms asking “Are you Hispanic?”, some quite important forms and surveys too.

    If all these institutions consider Hispanic-ity as self-evident to the respondent, (or at least not controversial enough to footnote and clarify what they mean by Hispanic) why is defining what exactly Hispanic means such an issue?

    So long as one uses a clear and consistent definition, what’s the big deal here? Richwine seems to have stuck to “born in either Mexico, Central America and the Caribbean, or South America” which seems eminently reasonable (and even better, non-vague) to my non-expert eye. Seems as good as any other definition one could come up with. Why the need to belabor that point beats me (unless the ulterior goal here is to never be able to agree about a definition thereby precluding any questions based on the concept of Hispanic-ity itself).

    Why classifying someone in a survey as Hispanic or not might necessitate Prof. van Whatever’s “substantive component of analysis from the qualitative, historical, cultural, normative, and theoretical perspectives “ is beyond me.

    For all I know Richwine’s thesis may be totally flawed, but Beauchamp’s critique on definitional grounds of “Hispanic” seemed quite weak.

    • +1. I rarely +1, but this is exactly right. Heterogeneity among those Richwine calls Hispanic should, if anything, attenuate his hypothesis. Argentinians aren’t Mexicans. I think we all get that. And I’m not my neighbor, but we both count in the income statistics of my town.

      • “Heterogeneity among those Richwine calls Hispanic should, if anything, attenuate his hypothesis.”

        Exactly right. The von Vacano-style objections to Richwine’s dissertation are 180 degrees backward in terms of statistical impact.

        • The fact that social scientists have criticized Richwine for using the category of Hispanic is surreal. If social scientists were not allowed to use categories like Hispanic, white, or black, lots of them would be out of work. Ethno-racial comparisons are the bread and butter of American social science.

        • Yes, but in the 21st Century intellectual discourse, logical consistency has taken a backseat to Lenin’s dictum that central question is always “Who? Whom?”

          Dr. von Vacano is good and Dr. Richwine is bad, and that’s all you need to know.

          Judging form photos online, Dr. von Vacano and Dr. Richwine are similar in coloration, but Dr. von Vacano’s name (despite the ironic “von”) ends in a vowel, so he is good.

  6. Dear Rahul,
    My point is simple: the concept of “Hispanic” is deeply and widely contested. So there is no one single, clear definition. In fact, there may be none. And Dr Richwine does not delve deeply on any definition (after all that’s part of dissertations, so yes this is a flaw). By the way your own definition is deeply flawed: US-born “Hispanics” or Spaniards are still often considered “Hispanic.” There is also a debate with the term “Latino” (See Jorge Gracia’s book).
    Perhaps it helps if you think in terms of the word “Asian”. Does it mean from India? or also from China? or also from Japan, Iran, parts of Russia, etc or also those born in the US of “Asian” or Indian parents? At the very least, you must try to THINK critically about what those ethno-racial terms mean. That is what a dissertation is for. It is not a simple policy paper or Policy Analysis Exercise. It is one of those things one must understand to do advanced work in places like IAS or CASBS. (one cannot just simply use ANY definition; I can’t define “Hispanics” as a kind of fruit or vegetable for example).
    Cheers,
    Professor von VACANO

    • Yes you can, as long as you’re clear about what you’re doing.

      Deeply and widely contested the conept of Hispanic may be, but surely no scholar will be confused by an operational definition.

      • I didn’t read anything, but I think that an operational definition is necessary, but no sufficient for a dissertation in social sciences. I would expect some discussion of the several ways they define a concept, because the way you define has implications. You may follow some standard ans say that explicitly and refer the reader to the literature that discusses the concept in more depth. But you need to address it at least a little bit.
        See, for instance, the discussion about democracy.

        • John Nash’s dissertation was 32 pages with handwritten equations and two citations, one to his own work. Blathering on about other people’s definitions of things might be interesting or not, but it is in no way required for anything.

        • 1. Maybe because it was shorter!
          2. So if you make a minor point, you have to talk longer about it? Sounds like academia to me.
          3. What’s the point in a dissertation talking about things that you aren’t going to directly address? (In this case, definition of Hispanic.) To convince the reader you’ve read widely in the subject? That might be good for getting an academic job, but has little subject matter information for the already informed reader. (I refer again to Nas’s dissertation, which indicated that von Neumann and Morgenstern had written a book that had some stuff he wanted to talk about… that was roughly the literature review section.) To hint to the reader of your ideological leanings? Now we’re getting somewhere… but it ain’t somewhere good.

        • Actually, judging by its widespread use – including here – hispanics seems to be more well-defined than most terms in social science. So that’s not the point.

      • Well not a scholar in the area I have to admitt I always thought of Hispanic as implying Spanish-speaking so I don’t particularly understand why Portugese speaking Brazilians apparently are included yet people from Spain and Portugal are not.

        The operational definition seems clear enough but I dont’ see any rationale as to why one would accept it.

        Note I’m from Canada so what may seem obvious to an American audience may not be so to me.

        • Brazilians other latin americans are mostly mixed race people with disadvantage identities. Portuguese and Spanish are the white man burden of latin america. Mixed europeans and latin american hides this fundamental fact. It like mixing dutch and african in south africa. Obviously some latin americans won’t fit the stereotype but most will do. It is not that complicated. The exceptions hopefully go to the error terms.

    • So you don’t like his definition. All dissertations use definitions. I assume that you do not have any substantive criticisms, or you would not be posting silly quibbles about definitions.

      • Actually IIRC Richwine did not offer a definition, instead he only counted people as “Hispanic” if they defined themselves as such, if nothing else this is problematic because that term itself is not used in Spanish-speaking countries AFAIK (it’s an Americanism); that is a serious methodological flaw which undercuts the entire dissertation itself (because it introduces a huge bias into the sample), not a “non-substantive criticism” as you so assert.

    • Prof. van Vacano:

      Essentially, I respectfully still disagree with your post.

      The fact that a term does not have a single universally accepted definition does not make it undefinable. e.g. Take your example of “Asisan”: Do the variants of usage you listed (rightly too) preclude usage of the term “Asian” in serious research? Can I not talk about Asian markets, or Asian parenting or Asian work-ethics or one of any number of different attributes?

      I see literally thousands of references to the term “Asian” in scholarly work from the social sciences and I doubt you’d ridicule those too for just sweeping aside your definitional concerns by a mere “…for purposes of this work we take Asian to mean….” And to be fair to Richwine, whatever definition he was using was no analog to your cute “Hispanic as a fruit” witticism.

      I’d be more convinced if you offered a better definition of “Hispanic”, but no: you insist that the term’s undefinable. That’s a little hard to digest seeing the wide usage and prevalence of this term in various contexts.

      Yes, it may mean different things to different people but that does not mean it means nothing!

      • I think Prof. Vacano’s point was that Richwine does not offer *his* own definition of what Hispanic means, thus the entire dissertation is wildly open to divergent interpretations depending on the reader’s concepts. (And unlike a journal article, a doctoral dissertation is not limited by space or page limitations so that’s not an excuse.)

        Also just because people use a certain term doesn’t mean it actually corresponds to anything in the real world, think about for how long Western thought divided people into clear racial categories (based on “DNA”), only in the past couple decades have they realized that only a tiny percentage of human genes relate to physiognomy and that much (most?) of the time people have more in common genetically with individuals from other ethnic groups than they do with members of their own “race”.

        • It’s very silly to denounce Richwine for the conceptual vagueness of the term “Hispanic” when nobody else gets denounced for using it. For example, according to Google News, the term has been used by the professional news media over the last few weeks 99,800 times. The term “Hispanic” appears 75,600 times on census.gov, the official site of the Census Bureau.

        • much (most?) of the time people have more in common genetically with individuals from other ethnic groups than they do with members of their own “race”.

          That’s nonsense. Genetic similarity decreases with (pre)historical geographic distance between populations, and widely separated populations are completely genetically divergent in the sense that, say, every single European is more similar to every single European than to any East Asian or sub-Saharan African, and vice versa. See this article for details. They write:

          Thus the answer to the question “How often is a pair of individuals from one population genetically more dissimilar than two individuals chosen from two different populations?” depends on the number of polymorphisms used to define that dissimilarity and the populations being compared… [I]f genetic similarity is measured over many thousands of loci, the answer becomes “never” when individuals are sampled from geographically separated populations.

    • This term is widely used including by the “hispanics” themselves. For example, every paper on vote choice will use it. These are mostly mixed raced people with a “disadvantage” and “brown” identities. Obviously not everyone from latin american fits these definitions but lack of fit is also true for other definitions as well. How many “whites” fits all stereotypes of “being whites” ? How many people with white names are actually mixed? Also mixing Spanish with with hispanics is kind of joke, such as mixing germans or english with black people from their colonies. It hides the fundamental historical fact that the iberians were the true “white man burden” in the Latin American.

      You are also exagerating things when you talk about Russians and Asians. Russians are not asians simply because they came from Europe to populate Asia later. Russians are people of norther european descend and christian creed so there is no way to call them “asian” unless one wants to create confusion.It is like calling white americans amerindians just because they move to the americas. Thankfully today we have DNA testing so that we can trace the real origins of the human people as oppose to being endlessly digressing about culture.

    • Dr. von Vocano:

      An FAQ for you:

      Q. But how can test-givers tell who is Hispanic?
      A. They just ask them. It’s called “self-identification.”

      Q. But is that scientific?
      A. It’s good enough for government work. The government spends billions to count Hispanics, and it’s all done just by letting anybody check whether or not they are Hispanic.

      Q. But how do we know that Hispanic test scores won’t suddenly change?
      A. Well, I’ve been following this question for over 40 years. Jencks has been following it longer. So far, they haven’t. Maybe you are right, or maybe Professor Jencks is right. Personally, I’d rather bet the country on Jencks’s assessment of the evidence.

      Q. But if Hispanics are an ethnicity, not a race, how can we know that the next generation of Hispanic immigrants won’t be very different?
      A. I could imagine one event that would drive up new Hispanic immigrants’ children’s test scores substantially: another revolution in Mexico. If rich white Mexicans, like the world’s richest man Carlos Slim, had to flee for their lives from Mexico, the next generation of Mexican newcomers might be a lot like the prosperous Cubans who arrived in Miami after Fidel.

      But the way immigration from Mexico has been working since the end of the last revolution almost a century ago is via family chain migration. New immigrants tend to belong to the extended families of old immigrants.

      Q. But that’s genetic determinism!
      A. Actually, it’s both nature and nurture. If, say, a young fellow from Sinaloa moves in with his uncle in East LA, the newcomer shares a lot, genetically and culturally, with the old-timer.

      http://takimag.com/article/frequently_asked_questions_about_the_jason_richwine_brouhaha_steve_sailer/print#ixzz2axzS67G0

  7. It is a good thing that most dissertations are flawed. I know of a certain professor who did not permit flawed dissertations to be approved, and, by golly, after a few revisions, all the flaws disappeared.
    We want the flaws in research to be evident, along with the reasons for them. The apologies come at the end.
    Don Campbell was eloquent on such matters.
    Lee

  8. It’s trivial, but I thought this part was a little funny:

    ‘Richwine finished his dissertation in his fourth year at Harvard, while he was working out of AEI. That’s not unheard of, but it’s faster than the Kennedy School average. “It can be done with some sacrifice to quality or depth,” one Kennedy School alum told me.’

    From that snide remark, it sounds like “one Kennedy School alum” is just a teeny bit jealous.

  9. I find ad hominem statements like “jealous” absurd and puerile. I have no idea what I should be jealous of. In any case, Dr Richwine wrote a dissertation (ie, doctoral work, not MA) at the Kennedy School of GOVERNMENT. Not the Kennedy School of Statistics.
    Politics and policy are complex, not mere numbers-crunching exercises. If one is going to write a doctoral dissertation at a government school, one needs to address all relevant issues. Not merely those that are of significance to statistical methods. If a dissertation is about “race”, one must critically examine the relevant concepts, not simply use the “It was a methods dissertation” screen to shield from gaping holes in social science. Race involves politics, culture, and history. Again, the mususe of “Hispanic” is similar to a broad, uncritical use of the term “Asian,” which covers almost half of the world, without a substantive defense of the use of the term.
    And Rahul, my last name is VON Vacano, not “van Vacano.” What is your last name?

    • The terms Hispanic and Asian include people of different races. The terms white and black include multiple ethnic groups. Everybody knows that. Your complaint seems to be that you would have preferred for Richwine to write a dissertation on a different subject.

    • Was the quote I was referring to yours or did you just not read the whole comment? The article attributes it to “One Kennedy School alum.” If anything is absurd and puerile it is making the claim that completing a degree in 4 years is impossible without producing a sub-par dissertation.

      b.t.w. Statisticians involved in the social sciences (as many of us are) are acutely aware of definitional issues of many classes and categories. All categorizations can be broken down though post-modern discursive techniques into meaninglessness, but we shouldn’t let that paralyze us. If the only problem with the research was that he used self-identification as the definition of hispanic, our grounds for outrage would be very weak indeed.

      b.t.b.t.w. It is poor form to demand the full name of those you are having a discussion with on the internet.

      • Outrage may indeed be too strong a word but for the fact that work(s) such as Mr. Richwine’s are often used by some individuals and/or groups to promote institutionally racist policies, these can and do have very powerful real world effects on many people’s lives (this wasn’t just some recondite grad paper about a hitherto-undiscovered tribe on a small island near Borneo).

        Also while I think your description of the necessity of defending against excessive deconstructivism is well-taken, the methodology behind the operative term in this work sounds so flawed to me that it does call into question any conclusions which are drawn from the data in it.

        • OK I looked a little deeper and apparently Mr. Richwine did use a definition of his own, essentially someone born in South or Central America (“Latin America” in general); also he makes concessions to the concept of mestizos.

          While I think this is valuable because at least rules out the Iberian Peninsula, it’s still problematic because it negates the tremendous number of black Africans who make up the mestizo concept (Richwine defines it strictly in terms of mixes between European-descent and Amerindian.)

          Also I think the question of Brazil, by far the largest and most populated country in the grouping, needs to be teased out a little better (at the very least differentiating Spanish from [Brazilian] Portuguese, as close as they may be]. (Certainly I think grad level work need not be perfectly flawless, but as Prof. von Vacano points out this was not just a master’s level work from a “run of the mill” school, it was a doctoral dissertation from Harvard, presumably the standards should have been more rigorous if anything.

        • I agree but I think it’s worse than that: Chapter 3 pg 60 of dissertation:
          Over 56% of immigrants living in the US in 200 were Hispanic—that is, born eiter in Mexico(32 of total immigrants), Central American and the Caribean (17%) or South America (7%).
          and then
          “… Most [Hispanics] come from Spanish-speaking nations with cultures heavily influenced by Catholicism.”

          As far as I can see this implies that Haitians and Jamacans are Hispanic, not forgetting the Brasilians. Obviously not what was intended but the definition is just not clear enough.

      • I agree with Ian Fellows: for example, self report identity as hispanic strongly predicts vote choice, so it might mean something.

      • “on the internet” covers a lot; I know a forum in which full real names are expected and anonymous identifiers deprecated.

        But in this case I read Diego von Vacano as just _asking_ for Rahul’s name; _demand_ is a strong word here. The context was that Rahul twice misspelled his name, once as “von Whatever” and once as “van Vacano”.

        On this blog anything seems to go as far as names go from fully anonymous to fully explicit, and a little respect both ways seems in order. Those who want to hide or obscure their identity can do so; anyone is entitled to expect correct spelling.

      • I seem to recall that John von Neumann was really just plain old John Neumann (or however you say it in Hungarian) and he added the von to sound more important. But maybe that’s just something I’d like to believe…

        • Neumann was born Neumann János. The “von” was added along with the reversed order when translated into English to accentuate his noble status of margittai in the Austro-Hungarian Empire.

  10. ps always be wary of statements like “The dissertation was not flawed”, as Richwine writes above. All dissertations are flawed in some sense or another. There’s nothing wrong with that. What matters is their systematicity, depth and contribution to a field.

    • I’ll be wary of that statement if you’ll admit that all concepts of Hispanic are flawed, thereby allowing free rein for any researcher to study any part of the field in depth without the least defensiveness against your absurd request for nonexistent generality. Deal?

  11. Everybody has been using “hispanic” as a term in the USA. Variations of it such as “non-hispanics white” are even more strange but quite common. These terms have mostly not been contested, even by the “hispanics” themselves, who use these to advance their own agendas. Lots of papers and books have used it without been criticized. But now, for some reason, there is a diatribe against the author of a politically uncorrected study. So that assumption of the study – the very existence of the hispanics as a group – is been now challenged just because of the study’ conclusions? Interesting.

  12. Is it already time to trot out cosma shalizi’s writeup about IQ again? I know a lot of people have a hard time understanding it, but if you don’t understand it, you really have no business contributing to the pile of pseudo-scientific baloney in this field. I’m trying to be polite out of respect for andrew, but this racist crap has been getting old for 30 years now. They all have the same damn background section on “the science” of IQ (hint – real science doesn’t have to go out of its way to remind people that it’s science). They all appeal to academic free speech and whine about getting attacked for political correctness, when the truth is the research is just crap.

    http://vserver1.cscs.lsa.umich.edu/~crshalizi/weblog/520.html
    http://vserver1.cscs.lsa.umich.edu/~crshalizi/weblog/523.html

    • Unfortunately, the racist ignorant rubes who fail to understand Shalizi all have low g, so it’s next to impossible to rid them of their errors. Even a high g guy like Cosma couldn’t do it.

      • The Asian physicist (strange that you felt compelled to refer to my ethnicity) happens to read Andrew’s blog.

        It’s obvious you’re in the group that doesn’t understand Shalizi’s essay. He’s not refuting the claim that g has some predictive power, or is a decent way to compress the information from multiple cognitive tests. He’s pointing out that the positive correlations between those cognitive tests does not, by itself, imply that there is an *underlying causative factor* called g. This particular point is not central to the IQ debates that plague society, although opponents of IQ cling to Cosma’s essay the way a cripple clings to God.

        As long as g is relatively stable (i.e., in adulthood) and predictive of outcomes, it’s a useful construct for psychology and social science. Those properties could be true even if g is merely a good summary statistic for a more complicated description of cognitive abilities.

        It’s also true that g is quite heritable, but that’s another story …

        • “asian physicist” was meant to parallel “racist pseudo science”. You may have missed the sarcasm which was directed at the original poster above. I did get Shalizi’s essay and I also get that to a casual observer, it reads as a devastating critique of IQ, which is presumably why “As an online discussion about IQ or general intelligence grows longer, the probability of someone linking to statistician Cosma Shalizi’s essay g, a Statistical Myth approaches 1”

        • Give me a break, that’s the other common response – “these PC idiots are just like fundamentalist christians!”, as if GWAS statistics was too sophisticated for anyone to understand. The problem with research in this area is that they may pay lip service to acknowledging that all they have is a low dimensional representation of things that determine these iq test scores (including environmental effects) is that their conclusions and implications always rely on a stronger assumption regarding IQ as an intrinsic, heritable property of individuals.

          For example, take the recent interview with Zhao Bowen in wired – http://www.wired.com/wiredscience/2013/07/genetics-of-iq/ ““You cannot ask a kid with low IQ to just work hard and then become a really talented mathematician,” he says. “It’s impossible.” And yet, Zhao says, that’s what is currently expected in China. He wants to stop the vast majority of Chinese students from wasting their time.

          Take also Richwine’s own thesis – “The optimistic argument says that if today’s immigrants gradually get better educations and move up the socioeconomic ladder, then they could assimilate culturally and economically just as Europeans did. However, this optimism is unwarranted if the average immigrant lacks the raw cognitive ability, or intelligence to pursue higher education and take on skilled labor. Just as low intelligence will limit an individual’s career choices, low average intelligence in a group will inhibit its overall success”

          You can’t wave your hands about “well IQ is meant to be predictive and it’s okay if it’s just a low dimensional projection of environmental factors” and then make a policy statement which is completely inconsistent with allowing for that interpretation. It’s like saying, “well yes, the ice cream sales could be correlated with swimming pool drownings because they’re dependent on the time of year and weather, the importance of ice cream sales is that it’s a strong predictor” and then concluding with “My research suggests that we should be banning ice cream sales to save children.”

        • As someone quite naive in these IQ debates, one assertion in that long Cosma Shaizi piece puzzled me:

          “Do I really believe that the heritability of IQ is zero? Well, I hope by this point I’ve persuaded you that’s not a well-posed question. “

          Why is it an ill-posed question? I tried reading his long, quite technical logic (and maybe grokked 20% of it!) but I still am confused.

          Is asking “Is heritability of height zero?” also an ill posed question?

        • (same anon)

          Also “He’s pointing out that the positive correlations between those cognitive tests does not, by itself, imply that there is an *underlying causative factor* called g. This particular point is not central to the IQ debates that plague society”

          What, then, _is_ central to this field of research? If IQ/g is not an intrinsic property of individuals, but is just a convenient low dimensional representation of environmental effects, why are we spending all this money studying it? What exactly is the causal hypothesis that is being examined when one looks for a correlation between hispanic immigrants and IQ? Because without the existence of an underlying causative factor, the entire enterprise is rather circular. We project down environmental factors into a number, then lo and behold, individuals with different environments are associated with that number.

        • No one believes that IQ is “just a convenient low dimensional representation of environmental effects”. Perhaps in the 1970s, but not anymore.

        • @P I’m specifically responding to Steve’s assertion that:

          “He’s pointing out that the positive correlations between those cognitive tests does not, by itself, imply that there is an *underlying causative factor* called g. This particular point is not central to the IQ debates that plague society”

          which sounds to me that he’s not all that concerned with whether heritability estimates are environmental or not. My point was then, if one is not willing to defend g as an underlying causative factor, then it’s rather disingenious for papers such as Richwine’s to discuss implications assuming their measurement is not just an intermediate outcome.

          It sounds like you and the rest of the field might differ from Steve in that you do care about this particular point. If so, show me a study that compellingly measures a direct effect without environmental confounders. Yes, I’m familiar with twin studies, no, I haven’t found any of them to make a convincing case that they’ve obtained a generalizable estimate, without questionable causal identification, which eliminates _all_ environmental contributions and gene-environment interactions, but I’m open to being convinced if there’s a compelling study that exists.

        • which sounds to me that he’s not all that concerned with whether heritability estimates are environmental or not.

          Whether g is real has no bearing on the fact that IQ is heritable. If g is not a real causal entity, then it’s a composite of real causal entities and its heritability reflects the combined heritability of its constituent elements.

          Look at it this way. A linear combination of individuals’ height, eye color, and handedness is surely heritable, because each three are heritable. A combination of randomly chosen heritable traits is itself heritable, even though there is no common set of genes that causes variation in those traits. In principle, it’s possible that g is simply a sum or average of independent mental abilities, its heritability reflecting the heritability of many specific abilities. However, there are many reasons to believe that this is not the case, e.g., high genetic correlations between different mental abilities.

          If so, show me a study that compellingly measures a direct effect without environmental confounders.

          http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3182557/

        • @P Thanks for the polite response. I had a quick look at the paper and frankly if this is the state of the art in the field, it’s time to take a step back because both the study design and the model are quite inadequate to back up the causal claims. I should get back to my real work, so these here are a few quick thoughts.

          The model does not include any environmental factors as direct effects or genotypic interactions, so the model is entirely dependent not only on the study design completely de-correlating gene and environment. It’s unconvincing in this respect – “a correlation between environmental and genetic similarity might occur if similarity due to environmental factors between relatives segregates with the degree of separation” – really, this is really the only way they could imagine gene-environment correlation? Principal components capture large-scale variation, but it’s not a compelling solution when genotype-environment correlations are entrenched into every last crevice of social network in society. More generally, if you’re going to take “subtractive” approach to confounding (addressing only confounders that come to mind) in an observational setting with a low-dimensional outcome, you’ve got to be humble about the possibility of missed backdoor paths. “Establish that” is not an appropriate title phrase for this kind of study.

          There’s a secondary issue with the conflation of “explained variance” and “explained causality” in genomic associations. Take this ‘explained by’ logic to it’s genome-wide extreme – what if they could fit their “idealized” model of every SNP? You would ‘explain’ 100% of the IQ variance since the pattern of SNPs uniquely identifies an individual, but it says nothing about causality. To go from the other direction, if I measured every single environmental factor in a persons life, it would ‘explain’ 100% of the variance as well because a person’s life history uniquely identifies the individual as well, but again, says nothing about causality. Since they’re using a low dimensional projection of genotype, this is not the main issue in this study, but the fact is their “idealized” model is not ideal for answering the question at hand.

          Regarding your point about g, I see what your saying, but it’s not what is tested in either the idealized model (for the above reason) or their applied model because they have not eliminated the correlation between genotype and environment in their data.

          Regarding your point about if X is heritable, then X + anything is heritable, yes this is true. And some well-defined aspects of intelligence are heritable (eg in specific diseases), which were mentioned in Cosma’s original blog post. The problem is the jump from those well-defined diseases with well-defined hypotheses to these underdetermined observational correlations. As estimates of causal effect, the estimates presented here are unconvincing and they overreach in their conclusions. Overreaching in science is common, but in this literature you have a bad combination of self-serving overreach and harmful policy recommendations.

        • The model used in that study is based on the assumption that the correlation between genetic and phenotypic similarity of unrelated people from various parts of Great Britain represents a genetic causal effect. You say that potential genotype-environment correlations render such an interpretation unfounded. Could you explain your model explicitly? What kind of g-e correlations could explain the findings? The only ones I can think of are of the sort that can be thought of as genetic effects (e.g., having certain genes causes one to seek intellectual stimulation, which boosts IQ).

          Take this ‘explained by’ logic to it’s genome-wide extreme – what if they could fit their “idealized” model of every SNP? You would ‘explain’ 100% of the IQ variance since the pattern of SNPs uniquely identifies an individual, but it says nothing about causality.

          I don’t see how that follows. Using their method, the upper bound of IQ variation that can be explained is the actual extent of additive heritability, no?

        • >I don’t see how that follows. Using their method, the upper bound of IQ variation that can be explained is the actual extent of additive heritability, no?

          You’re right that the extreme case applies more to a case where all genetic information is taken into account, both additive and epistatic.

          I took a closer look at the model in their previous paper. The fundamental issue comes up in their previous paper on height (Yang 2010): “Heritability is the proportion of phenotypic variation due to additive genetic factors; we therefore fitted a model in which SNPs have additive effects. Non-additive genetic variation and variation due to gene-environment interactions may exist, but they are not part of the missing heritability because they do not contribute to the heritability. ”

          This is circular causal inference logic. If environment-gene interactions exist, then the correlation that they’re estimating isn’t the causal effect of additive heritability. If they’re unable to credibly estimate or disprove environment-gene interactions, then they have absolutely no business claiming that they’re estimating additive heritability.

          > You say that potential genotype-environment correlations render such an interpretation unfounded. Could you explain your model explicitly? What kind of g-e correlations could explain the findings? The only ones I can think of are of the sort that can be thought of as genetic effects (e.g., having certain genes causes one to seek intellectual stimulation, which boosts IQ).

          This is one of those things where the scientific community has the burden of proof backwards. Causality in nature is both sparse (due to locality of direct causal phenomena) and connected, so associations really should be assumed to be confounded unless proven otherwise. Economists get a lot of things wrong, but on the issue of causal identifiability, at least some of the applied practitioners may be onto something – if you can’t argue that you have the conditions for a natural experiment, it’s confounded by default.

          That being said, it’s easy to suggest backdoor causal paths. The starting point (which should be self evident) is that genotypes are not randomly assigned to environments. Rather, the correlation between gene and environment is the consequence of a migratory process which is going to introduce correlations at pretty much every conceivable scale. There’s lots of room for such correlation in between the bookend scales of my family and large-scale PCA variation (the two things they try to eliminate).

          For example, I not only tend to know more people from the country I’m from, but within the country I’m from, I will tend to know people from particular region of that country where my family comes from. Within my community (which can also be defined at multiple scales) I share my resources as well as information on social connections, parenting, education etc. This is just one small example, but such correlations are entrenched at every scale of the social network. Removing pairwise family relationships and PCAs is not sufficient to convincingly estimate the direct causal effect of additive heritability.

        • You suggest that g-e correlations in the form of “social connections, parenting, education etc” could invalidate the results. This is very unlikely. Firstly, we know from traditional family studies that the effect of the shared family environment (which includes parenting, schools, etc.) on IQ is weak if not non-existent (adoption studies, for example, indicate zero correlations between non-biological family members by adulthood). Secondly, even if such factors strongly influenced IQ in the case of a particular individual, the same factors would not be able to influence unrelated individuals who happen to be genetically somewhat similar to that individual. Yet it is the similarities, genetic and phenotypic, between these unrelated individuals that they use to estimate heritability.

          This replication study helps clarify why the method is reliable. They had a sample of thousands of Swedish male twins who had taken an IQ test upon induction into mandatory military service. They genotyped them, excluded one twin from each pair, and dropped individuals who were related. Then they estimated pairwise genetic relationships between all pairs in the sample, and correlated them with IQ similarity. The estimated narrow-sense heritability was 47 percent, similar to the earlier study.

          How could any g-e correlations cause IQ similarity between unrelated individuals in the Swedish sample? Twinning happens pretty much randomly across the country, and any effect of clustering of twins in families is prevented by excluding related individuals. Thus we have a sample of unrelated people who grew up all over Sweden, most of them never meeting each other. Yet when their IQs were tested in late teens, these people resembled each other to the degree they shared genes. The sort of g-e correlations you suggest are completely unable to explain these results. Your case against these methods is very weak if you cannot present even hypothetical environmental mechanisms that could account for the results.

        • “adoption studies, for example, indicate zero correlations between non-biological family members by adulthood” I’d be interested in seeing citations (and sample sizes) for this 0 correlation estimate.

          “Secondly, even if such factors strongly influenced IQ in the case of a particular individual, the same factors would not be able to influence unrelated individuals who happen to be genetically somewhat similar to that individual.”

          Unrelatedness is not a binary proposition. You haven’t subtracted out genetic similarity by excluding direct (or even extended) family members. I’ll put it another away, if you took environmental variables and applied a similar regression (eliminating family members and subtracting out PCAs), I’m willing to bet there is going to be substantial “explained variation”.

          I don’t see how the swedish data as used in the replication study addresses this issue. The problem is _not_ that the standard errors are too small because of correlated observations (which removing family members tries to address), it’s that removing direct family members is insufficient to block a causal path through shared environments.

          Anyway, I appreciate the discussion, but this thread could go on forever and I don’t have time to pick apart a flawed literature one paper at a time.

        • I know of three studies of unrelated siblings reared together who were tested after childhood (there are more child studies, but they are not very informative because with age, heritability increases and shared environmental influences decrease).

          Correlations for the three samples of unrelated siblings reared together were as follows:

          +0.02 (N=24) (Teasdale & Owen, 1984)
          -0.03 (N=84) (Scarr & Weinberg, 1978)
          -0.01 (N=181) (Loehlin et al., 1997, in the book Intelligence, Heredity and Environment)

          The weighted average is -0.01.

          Unrelatedness is not a binary proposition. You haven’t subtracted out genetic similarity by excluding direct (or even extended) family members.

          In these studies, related individuals were excluded based on genomic measures of relatedness. Pairs of individuals whose coefficient of relatedness was greater than 0.025 were excluded. This means that the most closely related pairs in the data were fourth cousins or so. Such distant “family” relationships are socially 100 percent meaningless — certainly in a nuclear family culture like Sweden’s or Britain’s. If you want to claim that such relationships, which are completely unknown to individuals (I have little idea who my second cousins are, never mind fourth cousins), are associated with shared environments, you have your work cut out for you.

Comments are closed.