Further survey results

By Razib Khan | September 21, 2012 1:36 am

I have now reformatted the responses into a csv file. So you can do something boring like create a scatterplot with Excel, as above. Or, you can import into R and dig for more interesting patterns. Here is an updated to the PDF which shows you the simple non-crossed results.

Also, out of curiosity I separated respondents into geneticists vs. non-biologists (so I exclude biologists who are not geneticists from this analysis). First, here are the non-biologists on race:

And now the geneticists:


For what it’s worth, I agree with the 66% of geneticists. The sample size is small and biased, but still interesting nonetheless.


  • Eurologist

    Regardless of the low-R fit, what is interesting in your first figure is how many social liberals are not economic liberals. In other words, the lower right quadrant seems to be about equally populated as (if not more highly than) the upper right one.

  • marcel

    Elaborating on Eurologist’s point, it appears that the line in your scatterplot has a non-zero slope only because 3 quadrants are roughly equally densely populated while one quadrant is almost empty. It’s not so much that as people tend more liberal on one dimension they also do so on the other; rather it’s that roughly equal numbers of people are

    a) both socially liberal and economically liberal
    b) socially liberal and economically not liberal
    c) neither socially liberal nor economically liberal

    and very few are socially not liberal and economically liberal.

    Drawing a line in this situation, I think, misrepresents what is going on. it suggests that if you know that someone is socially liberal you have some insight into their level of economic liberalism. Rather, it appears to me that if you know that someone is socially liberal, you have very little information about the likelihood that they are economically liberal: conditional on someone’s being socially liberal, there appears to be a roughly 50% chance that they are economically liberal.* If you know that they are socially conservative, then you have some information about them that is useful for predicting their level of economic liberalism – they are pretty likely not economically liberal.

    We can also turn this around. If they are economically liberal, then they are likely socially liberal, but if you know only that they are not economically liberal, then you cannot say, with much confidence, whether they are likely to be socially liberal or not.

    Attaching labels that are appropriate to the US political scene to each quadrant, as a marketer might: the upper right quadrant is Libertarians, the lower right quadrant is the liberal wing of the Democratic party; the upper left is the religious right; and the lower left would include (this is the label I’m least confident about) a hodgepodge of political centrists, communitarians and probably others.

    *I’m eyeballing the graph here, not checking the original numbers in the csv file – I have to get ready for work and really should not be doing this now!

  • https://plus.google.com/109962494182694679780/posts Razib Khan

    Drawing a line in this situation, I think, misrepresents what is going on.

    huh. why do you think i displayed the r-squared?

  • marcel

    huh. why do you think i displayed the r-squared?

    I assumed (yeah, yeah, “When you assume, you make an ass of you and me”) that it was there because that is a standard statistic to display with a graph, but the comment doesn’t really get at my point, which is almost Tufte-esque (if I may be a bit grandiose and self-aggrandizing). When a line is drawn in a scatterplot, I think that the author of the graph is implying a relationship/correlation of some degree of reliability between the variables in the graph (and generalizing from my own response, which may or may not be legitimate, I suspect many others do as well). Something about the relative efficiency of conveying information graphically vs. with words (whether or not that information is accurate).

  • https://plus.google.com/109962494182694679780/posts Razib Khan

    #4, the r-squared leaves nothing implicit. if i drew a flat line with an r-squared of 0.01 would you think i’m implying a relationship? 😉

  • marcel

    Except that having looked at the numbers in the csv file, my intuition has been defeated or I am more confused than normal. You get a reasonable, upward sloping line for the right side of the plot (coefficient = 0.59 SE=0.12), but not the left side (coefficient=0.17, SE=0.17). So in the upper half of social-liberalism, the relationship between the 2 variables is reasonably good, not so much in the lower half. I’ll have to think about this a bit, not the relationship itself, but why I cannot see it in the plot.

    In re: R-sq: with 350+ observations, an R-sq of 0.2 for 2 variables is not bad. Not great, but it tells me that there is some information there. For an R-sq of 0.01, I’d probably want an N of several thousands, maybe 10k (I’d have to look it up).

  • https://plus.google.com/109962494182694679780/posts Razib Khan

    with 350+ observations, an R-sq of 0.2 for 2 variables is not bad. Not great, but it tells me that there is some information there.

    which was my implied point.

  • marcel

    Beating a dead horse at this point, but i just realized what is interesting (to me) about this.

    If you tell me that someone’s social liberalism index is high (i.e. > 50), I will have very little information about their economic liberalism, but if you tell me that their soc-index is low, a reasonable guess is that their economic liberalism is as well.

    OTOH, within the low group, there is pretty nearly no correlation between the 2 indices, but there is a reasonable correlation in the higher group. In the soclib < 50 group, most of the people also have econlib 50 group, although overall people are reasonably evenly distributed into the 2 groups, econlib >50 and econlib<50, as soclib rises within this group, econlib does as well. Looking at the graph, I missed these 2nd order effects (i.e., those within each of the soclib groups).


    Also, having trained in stem and not having the kind of literary crit background of many people I
    know, I'm not good recognizing implications: the only inferential skills that I trust in myself are the statistical ones, and we can see from the above comments how reliable that is!

  • T

    When I answered those questions I looked very carefully and saw that you mentioned “leftism” in both cases. Leftism consistently means = anti-traditionalism + economic socialism. (Rightist generally doesn’t have a real meaning, since leftists use the term to refer to the other three corners of the plot).

    However the term liberalism (which the question also use) has different meanings in different countries. By pairing it with leftism you seemed to be using it in the American sense, but at least some of your respondents (including Marcel) interpreted it differently; they were using the classic definitions. Evidence of this is him referring to the upper right corner (both economically and socially liberal) as libertarian.

    Even if I am wrong about how to interpret the questions my point is valid: we weren’t all answering the same way.

  • albatross

    There’s a wrinkle in the socially liberal/conservative question that might be worth asking another question or two to untangle in some future survey: How many people are socially liberal in the political sense, but are conservative in their personal life? I’m thinking of Charles Murray’s recent work here–it seems like a lot of educated, relatively successful people are politically social liberals, but are personally social conservatives. Further, I imagine most people would benefit from a more socially conservative lifestyle than is the norm (like, you really ought not to have kids till you’ve been married for awhile and you’re pretty stable), even though I don’t think it’s a good idea to have the police and courts involved in forcing people to do that stuff.

  • marcel


    I’ve been married nearly 30 years and actually thought the main reason to get married was that it simplified a lot of issues around child rearing (so my son was born 9mos, 5 days after the wedding). However, my understanding is that it is increasingly the norm in both W. Europe and Quebec for the middle class not to marry before starting a family, at least not in a formal, legal sense. The current president of France (as well as the losing candidate in the previous election) both — well, actually, together — are a prime example. Any thoughts on implications of this for your parenthetical statement?

  • https://plus.google.com/109962494182694679780/posts Razib Khan

    #11, marriage is just a stand-in for a monogamous committed relationship. the high cohabitation rates in much of the western world is qualitatively different from the patterns you are seeing among the working and lower classes in the USA today, where the relationships which produce children aren’t very firm in the first place.

  • DK

    Re: races. As a biologist who answered with the last option, I would like to hear your opinion on what is phylogenetic reality then. I find it difficult to believe that the racial differences between Japanese and Yoruba is equal parts social construct and genetics. (Obviously, I had geographic races in mind, not things like “Arian race”).

  • https://plus.google.com/109962494182694679780/posts Razib Khan

    #13, i was thinking the racial difference between blacks and whites, or middle easterners and whites, as social construction. these classes are not as clear and distinct as japanese or yoruba.

  • Syon

    Razib:”#13, i was thinking the racial difference between blacks and whites, or middle easterners and whites, as social construction. these classes are not as clear and distinct as japanese or yoruba.”

    The racial difference between Middle Easterners and Whites is quite clearly a social construction (as evidenced by the fact that Christian and Jewish Middle Easterners who immigrate to the West are folded into the White category), but how is the racial difference between Whites and Blacks a social construction? Are you talking about things like the one drop rule (people with fractional Black ancestry counting as fully Black)?


  • https://plus.google.com/109962494182694679780/posts Razib Khan

    #15, yes.

  • RedZenGenoist

    From http://en.wikipedia.org/wiki/Social_constructionism:

    “A social construction (also called a social construct) is a concept or practice that is the construct (or artifact) of a particular group. When we say that something is socially constructed, we are focusing on its dependence on contingent variables of our social selves rather than any inherent quality that it possesses in itself. ”

    “Social constructs are generally understood to be the by-products of countless human choices rather than laws resulting from divine will or nature. Social constructionism is usually opposed to essentialism, which instead defines specific phenomena in terms of inherent and transhistorical essences independent of conscious beings that determine the categorical structure of reality.”

    I think everyone understands what Razib is asking with the question.

    But taken by literal definition, we’re looking at EITHER a phylogenetic reality OR a social construct. The Ven diagram between the two is non-overlapping, so to speak. The categories are discrete. Social construct literally means “NOT a phylogenetic reality, has NO inherence”. That, in a different society, racial groupings could be malleable and interchangeable, because they have no objective substance.

    Of course, what we mean when we answer “equally” is that there is *some* superficial degree of interchangeability (one drop rule, etc, is all quite arbitrary), but that the deep phylogenetic realities are highly objective, and not interchangeable. Answering that it’s a “phylogenetic reality” implicitly includes all of the subjective squishiness of “one drop rule”, semitics being caucasian or not, etc, because that’s always the case in all phylogeny.

    (I’m a genomicist who answered “equally”, cognizant that I was being literally incorrect, and a bit stupid)

  • Douglas Knight

    15 people claimed to be exactly centrist. Since that was the default setting, they should probably be interpreted as not answering and be excluded. For each axis, another 20 changed that axis, but left the other at 50. I’m uncertain as to whether to exclude them. Anyhow, it doesn’t make a difference to the regression line, R^2, or lowess to exclude them.

    One danger of scatterplots is that you can’t see people with identical responses, which happen with default values, extreme values, and, to a lesser extent, round values. One answer is to add noise. Another is transparency, here done to your data.

    I’m surprised by the large number of people who put themselves exactly on the diagonal.

  • https://plus.google.com/109962494182694679780/posts Razib Khan

    #18, i removed the ’50 – 50′ people under the assumption that they were just default settings.

  • Miguel Madeira

    “How many people are socially liberal in the political sense, but are conservative in their personal life? I’m thinking of Charles Murray’s recent work here–it seems like a lot of educated, relatively successful people are politically social liberals, but are personally social conservatives. ”

    I am not much sure that the concept of “personally social conservative” makes much sense – at least, if we associate “social conservatism” with the idea that social rules, traditions, tabus, etc. are needed to repress the doomed moral nature of man and stop us from descending in barbarism, there is nothing intrinsically “conservative” in stable and hard-working families with A-student kids. If anything, I suspect that are exactly this kind of people that are more prone to believe in the “rosy” view of human nature, that humans are “naturally” good, diligent, reasonable, friendly, etc. and don’t need an external moral force.

    A concrete example – if you live in a social environment where divorce is rare, I think you will be more prone to believe that divorced people probably had very strong reasons to do that (after all, in your environment, people only divorce in very strong cases) and then they should not be morally judged nor legally restricted in their ability to divorce. In contrast, if you live in a neighborhood full of divorced or single mothers with semi-delinquent kids, you will more prone to believe that social (and eventually legal) rules against divorce, sex-out-the-wedlock, etc. are needed to sustain civilization.

  • Karl Zimmerman

    I made this chart, which shows the breakdown of readership in the HBD (Steve Sailer) and non-HBD categories. On the first chart I forgot to exclude the 50/50 results, but they are excised on the “quadrant breakdown”

    As expected, the non-HBD readership is mostly left-liberal, with a fair salting of libertarians as well. HBDers tend to be either right wing economically or socially, but given there are more non-HBD libertarians, I’d say their social conservativism stands out more. Interestingly, all but two of the more extreme “populist” cluster are HBDers.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Gene Expression

This blog is about evolution, genetics, genomics and their interstices. Please beware that comments are aggressively moderated. Uncivil or churlish comments will likely get you banned immediately, so make any contribution count!

About Razib Khan

I have degrees in biology and biochemistry, a passion for genetics, history, and philosophy, and shrimp is my favorite food. In relation to nationality I'm a American Northwesterner, in politics I'm a reactionary, and as for religion I have none (I'm an atheist). If you want to know more, see the links at http://www.razib.com


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar