Essentialist beliefs about essentialist beliefs

« previous post | next post »

Most people don't understand statistical ideas like distribution, correlation, and regression. And even if they understand these concepts, they often find them too complicated for everyday thinking, or impractical for everyday communication. So instead they fall back on essentialist beliefs and generic statements.

For example, "The lower classes are not merely unfortunate, according to the upper classes; they are genetically inferior". This statement attributes a belief about  one generic  category ("the lower classes") to  another generic category ("the upper classes"). It might be someone's subjective assessment based on personal experience, but in this case it's a journalist's description of the results of a series of psychological experiments, describing the fact there was a significant correlation (and a significant multiple-regression coefficient) between a "social class rank" variable and an "essentialist beliefs about class" variable. And as generic statements often do, it risks leaving readers with misleading ideas about the experiment and about the attitudes that it tested.

Here's the context for that quotation — Matthew Hutson, "Social Dawinism Isn't Dead", Slate 1/3/2014:

In 2012 the top 0.01 percent of households earned an average of $10.25 million, while the mean household income for the country overall was $51,000. Are top earners 200 times as smart as the rest of the field? Doubtful. Do they have the capacity to work 200 times more hours in the week? Even more doubtful. Many forces out of their control, including sheer luck, are at play.

But say you’re in that top 0.01 percent—or even the top 50 percent. Would you want to admit happenstance as a benefactor? Wouldn’t you rather believe that you earned your wealth, that you truly deserve it? Wouldn’t you like to think that any resources you inherited are rightfully yours, as the descendant of fundamentally exceptional people? Of course you would. New research indicates that in order to justify your lifestyle, you might even adjust your ideas about the power of genes. The lower classes are not merely unfortunate, according to the upper classes; they are genetically inferior.  

 The research in question was reported in a paper by Michael Kraus and Dacher Keltner, "Social Class Rank, Essentialism, and Punitive Judgment", Journal of Personality and Social Psychology, 2013. Their abstract:

Recent evidence suggests that perceptions of social class rank influence a variety of social cognitive tendencies, from patterns of causal attribution to moral judgment. In the present studies we tested the hypotheses that upper-class rank individuals would be more likely to endorse essentialist lay theories of social class categories (i.e., that social class is founded in genetically based, biological differences) than would lower-class rank individuals and that these beliefs would decrease support for restorative justice—which seeks to rehabilitate offenders, rather than punish unlawful action. Across studies, higher social class rank was associated with increased essentialism of social class categories (Studies 1, 2, and 4) and decreased support for restorative justice (Study 4). Moreover, manipulated essentialist beliefs decreased preferences for restorative justice (Study 3), and the association between social class rank and class-based essentialist theories was explained by the tendency to endorse beliefs in a just world (Study 2). Implications for how class-based essentialist beliefs potentially constrain social opportunity and mobility are discussed.

That all makes such perfect sense that some might see this research as analogous to experimental proof of the wetness of water. But that reaction would be inappropriate, since the most striking thing about the experimental results, in my opinion, is how weak they were: the tested fluid is not really very wet at all.

I'm afraid that the facts of the case are complicated, and so it's going to take a while to go through (even some of) them.  If you're pressed for time, or not especially interested in the hard life of science writers (and Matt Hutson is a good one, in my opinion), you might want to turn your attention to some of our shorter posts.

Let's start with some of the numbers, and then see what they mean. Here's a table of correlations from that Kraus & Keltner paper:

In order to understand this table, we need to know what the authors mean by the ten variable names that label the rows and columns.

Variable (1), "Social class rank", is the result of a guided self-assessment. Participants "[were] asked to rate their social class based on a measure of subjective socioeconomic status used in previous research (Adler et al., 2000; Kraus et al., 2009)". The cited reference is N.E. Adler et al., "Relationship of subjective and objective social class with psychological functioning: Preliminary data in healthy white women", Health Psychology 2000, which explains that

Participants were given a drawing of a ladder with 10 rungs that was described as follows: “Think of this ladder as representing where people stand in our society. At the top of the ladder are the people who are the best off, those who have the most money, most education, and best jobs. At the bottom are the people who are the worst off, those who have the least money, least education, and worst jobs or no job.” They were then asked to place an X on the rung that best represents where they think they stand on the ladder.

Variables (2) and (3), "Income" and "Education" were measured through simple multiple-choice questions:

Educational attainment was assessed with four categories: (a) less than high school education, (b) high school education, (c) college graduation, and (d) postgraduate degree. Annual income was assessed with eight categories: (a) less than $15,000, (b) $15,001–$25,000, (c) $25,001–$35,000, (d) $35,000–$50,000, (e) $50,001–$75,000, (f) $75,001–$100,000, (g) $100,001–$150,000, and (h) greater than $150,000.

Note in passing that the highest-income category (h) corresponded to about the top 10% of Americans in 2011. And also note that although Social class rank is significantly correlated with Income and Education, the correlations are not very strong: r=0.36 and r=0.21 respectively. We can square these correlations to get "percent of variance accounted for", which tells us that the self-reported Social class rank measure accounted for about 13% of the variation in Income, and about 4% of the variation in Education. To get a sense of what a correlation of r=0.36 means, here's a plot using fake data (unfortunately Kraus & Keltner don't provide the underlying data tables from their experiment):

Variables (4), (5), and (6) ("Biological beliefs", "Discreteness beliefs", and "Informativeness beliefs") were determined by averaging Likert-scale responses to a set of questions about each topic, e.g. "the kind of person someone is can be largely attributed to their genetic inheritance", "everyone is either a certain type of person or they are not", "it is possible to know about many aspects of a person once you become familiar with a few of their basic traits". There were eight questions for the first two belief scales and seven for the last one.

How about the measure of  Variable (7), "Social class essentialism"? The authors used standard psychometric techniques to find ten questions whose answers can be combined to yield one overall measure and two sub-scales that correspond roughly to their preconceptions about such attitudes:

On the basis of previously developed self-report measures of essentialism, we created 22 items for assessing people’s essentialist beliefs about social class categories. Essentialist conceptions of social class include beliefs that it is easy to judge social class through brief interaction, that children can be separated into different social class categories at birth, that social class can be determined without clothing cues, and that social class is at least partially based in biological temperament or genetic tendency. We administered these 22 items to a large sample (n = 703) of undergraduate students at a public West Coast university. The sample was collected as part of a larger battery of psychological questions administered to psychology majors online, at the start of the academic semester. 

We then subjected participant ratings to a principal-components factor analysis with a varimax rotation. Examination of the scree plot revealed a first factor with an eigenvalue of 5.30 and a second factor with an eigenvalue of 2.05. […] When we inspected the items, the first factor appeared to tap into beliefs that social class is a discrete social category (e.g., “Other people’s social class is easy to figure out”). The second factor appeared to tap into beliefs that social class is biological (e.g., “Social class is partly biological”). We selected items for the two factors of the essentialist beliefs about social class categories scale if the item loaded at .40 or higher on one factor and .30 or below on the other factor. This procedure yielded a total of 10 items: six on the discreteness subscale (M = 3.93, SD = 0.96) and four on the biological subscale (M = 2.64, SD = 1.00). The two scales were significantly positively correlated (r = .35, p < .05), and so a total score for essentialist beliefs about social class categories was computed (M = 3.42, SD = 0.81, α = .74).

The variable they call "Social class essentialism" then corresponds to that "total score for essentialist beliefs about social class categories"; and the "Discreteness subscale" and the "Biological basis subscale" (variables (9) and (10)) are the two PCA dimensions that they extracted from the pattern of answers to the ten questions.

Variable (8), "Conservatism", was based on direct self-assessment:

To assess political orientation, we collected a single item assessing the extent to which people self-identify as liberal or conservative on a 7-point Likert scale (1 = very liberal, 7 = very conservative; M = 3.63, SD = 1.67).

Finally, who were the subjects?

Participants were 169 adults from a national online sample participating for the opportunity to win small gift certificates to an online retailer. The participants were recruited through advertisements on http://Craigslist.org, and the website where data were collected was maintained by a public West Coast university. Of the 169 participants (age M = 35.04 years), the majority (n = 112) were female. Participants self-identified primarily as European American (n = 112), Asian American (n = 27), other or multiple ethnic groups (n = 17), African American (n = 9), Native American (n = 3), or Latino (n = 1).1 The social class backgrounds of participants in the sample were also quite diverse: 42.9% of participants reported annual incomes of $50,000 or less, and 39.3% of participants reported graduated from high school as their highest level of education completed.

So how about the rest of the results?

From their Table 2 (as reproduced above), we can see that there was indeed a positive and statistically significant correlation (of r=0.24) between their Social class rank and Social class essentialism measures. The correlation between Conservatism and Social class essentialism (r=0.17) was also statistically significant, though smaller.  But "statistically significant" here just means that the observed numbers are probably not simply due to sampling error in comparing two populations whose distributions are in fact identical. If we ask how useful e.g. self-reported "Social class rank" is in predicting their measure of "Social class essentialism", the answer is that it accounts for 0.24^2 = .0576 or about 6% of the variance.

Again in graphical form, here's what a correlation of r=0.24 looks like (and again this is fake data — at some point in the marvelously improved future, authors of scientific papers will routinely be required to supply their tables of raw data in digital form):

 

In fairness to Kraus & Keltner, a correlation of r=0.24 is quite good by the standards of social psychology. Here's a plot of "results from a century of social psychological research, more than 25,000 studies of 8 million people", reported in F.D. Richard, C.F. Bond, and J.J. Stokes-Zoota, "One hundred years of social psychology quantitatively described", Review of General Psychology, 2003, and previously discussed in "The and a sex: a replication", 8/16/2009.

And small but real effects can sometimes be practically important: in elections, for example, a percent or two of the vote is a big deal.

Returning to Kraus & Keltner, the correlation between Income and Social class essentialism (0.12) was not statistically significant, undermining even the statistical version of Matt Hutson's framing of social class essentialism as an attitude naturally characteristic of "that top 0.01 percent [of the income distribution]—or even the top 50 percent".

This all naturally leads us to ask why K&K to be having such a hard time showing that water is wet. Perhaps the assessment methodology is inherently noisy. But it's possible that the fluid in question just isn't in fact very watery, apparent common sense to the contrary. After all, Americans are raised on a diet of Horatio Alger stories, which presuppose both meritocratic possibilities and non-meritocratic realities, and we're often taught that hereditary aristocracy is a bad idea that our ancestors emigrated to the America to escape. There are other factors that tend in the opposite direction, especially racial and ethnic stereotypes, but it shouldn't be a shock to discover that Americans, whatever their income and self-assessed class standing, might be resistant to baldly-stated "social class essentialism".

It's also possible that the relevant space of social attitudes involves some complexities that K&K glossed over. In this connection I'd like to point to an interesting fact about their Table 2: the observed correlation between Income and Conservatism is 0.00.

Compare this to the work of Andrew Gelman, Red State, Blue State, Rich State, Poor State: Why Americans Vote the Way They Do, 2009 (powerpoint summary here). Stealing a couple of his slides:

He resolves the apparent paradox this way: "Context matters […]: The poor are similar in Red and Blue America, but the rich are different."

The ideas underlying "Social class essentialism" are (I think) a mixture of those generally described as "social issues" and those generally described as "economic issues"; and Kraus & Keltner's sample of 169 subjects is some mixture of the groups underlying Gelman's distinction between "Republican States" and "Democratic States". So a larger and better-described sample might show us a clearer picture with stronger relationships between social categories and attitudes.

Whatever the reasons for the low correlations in K&K's study, we're left with a typical difficulty for the interpretation of generic statements as a summary of weak statistical tendencies. As I wrote in "Generic comparisons", 11/7/2011:

[T]he relationship between mere prevalence and … generic statements is a very weak one.  Everyone knows that "Ticks carry Lyme Disease", although only a minority of ticks do so (14% in one study). Everyone knows that "Mosquitoes carry West Nile Virus", though the highest infection rate found in the epicenter of a recent epidemic was estimated at 3.55 per thousand (and the rate was essentially zero outside of the epicenter).

As a result, I once suggested banning such statements altogether ("Mandatory treatment for generic plurals", 9/13/2009):

I propose a voluntary ban on the use of generic plurals to express statistical differences, especially in talking to the general public about scientific results in areas with public policy implications.  In other words, when we're looking at some property P of two groups X and Y, and a study shows that the distribution of P in X is different from the distribution of P in Y to an extent that is unlikely to be entirely the result of chance, we should avoid explaining this to the general public by saying "X's have more P than Y's", or "X's and Y's differ in P", or any other form of expression that uses generic plurals to describe a generic difference.

 But this is completely impractical. Like all other natural languages, English offers us the irresistable temptation to substitute simple generalizations for complex ideas about the complex relationships implicit in tables of observations. You might as well try to ban alcohol or marijuana.

So instead, let me suggest a policy of responsible indulgence. Let's focus on preventing people from driving public policy debates when their judgment is impaired by generic intoxication, and offer effective statistico-medical intervention to those whose addiction to essentialist thinking blights their consciousness in personally as well as socially destructive ways.

Update — It's not just science writers who are tempted to make careless generic statements. Researchers themselves are often equally guilty. Thus Joshua Holland, "Why the Wealthy Favor Harsh Punishment — for Criminals and Errant Schoolchildren", Moyers and Company 1/13/2014, quotes Dacher Keltner as saying:

We asked people from different class backgrounds to think about the rich and the poor. And then we asked them, “To what extent do you think that these categories, rich and poor, are about people who have different genes, or different temperaments, or different biological makeups?” And again we found this similar pattern, which is that upper class individuals think of class as being based in biology and genes, and you don’t see that belief in people who are less wealthy.

He says that "upper class individuals think of class as being based in biology", not "upper class individuals tend to think of class as being based in biology". Fair enough — so far this is just a typical generic statement like "Ticks carry Lyme Disease". But then he tells us that "you don’t see that belief in people who are less wealthy". This statement must be technically false, given the correlation of merely r=0.20 between  the Income variable and the Biological beliefs and Biological beliefs subscale variables in their study.



17 Comments

  1. Lorenzo Vegro said,

    January 19, 2014 @ 4:42 pm

    Very interesting, thank you. There's a small typo: "…the observed numbers are probably not simply do to sampling error in comparing two populations whose distributions are in fact identical". Also, unless I missed something, shouldn't it be "incorrelated" instead of "identical"?

    >at some point in the marvelously improved future, authors of scientific >papers will routinely be required to supply their tables of raw data in >digital form

    I do look forward to that day, too.

  2. AntC said,

    January 19, 2014 @ 4:54 pm

    Thank you Mark, again, for a deep analysis.
    Speaking as a Brit (and steeped in European history), I find it puzzling to equate class with wealth/income, rather than include background and upbringing/relationship to the means of production. [Upbringing is only loosely related to educational level.]
    Does the U.S. not have any class of impoverished nobility looking down their noses at the nouveau riche? Is there not the equivalent of 'Sloane Rangers'? Are there not self-made men/women who adopt elitist attitudes out of all proportion to their de facto position on the income scale? Is Europe-envy a la Henry James and T.S. Eliot no more? Where do the 'chattering classes' fit in? It has been well said in Britain that it's the lower middle-class who hold strongest support for the status quo, because they have most to lose.
    I see class as much more being a bundle of beliefs/attitudes, in which case corrrelating class with, err, beliefs/attitudes would be much more like proving the wetness of water. OK, U.S. culture is known for having no appreciation of history (but forgetting is not "escaping" to use your word, and the "American dream" could only survive in defiance of history — this is your non-meritocratic reality), but do class-based attitudes not have persistence in the psyche?
    To make a linguistic point: given the long, long study of social 'class', I find it somehow dishonest and illegitimate for that word to appear in Kraus and Keltner's title.

    [(myl) Note that their notion of "social class rank" is a self-reported placement that is only "moderately" (in their description) correlated with income (r=0.36). Presumably this is because there are people with relatively high incomes who like to think of themselves as salt-of-the-earth types, and people from an elevated class background (at least in their own estimation) who happen not to be making much money. For example, maybe some of the 112 females among the 169 subjects were single mothers living on a limited income after divorce or death of their spouse. Without more depth in the analysis of the subject pool, it's hard to know what to make of the numbers.]

  3. Bob Ladd said,

    January 19, 2014 @ 6:04 pm

    @AntC: Speaking as a long-time UK resident brought up in North America, I can tell you that yes, there are categories of people in the US like the ones you ask about, but they are far less obvious than in the UK, I think because there are far fewer such people. I don't mean to suggest that American society doesn't have a class structure (as many Americans would like to believe, or used to believe), but the effects of class membership on one's attitudes and beliefs – and one's chances in life – seem to me to be far deeper and more pervasive in the UK than in the US. Your objection to Kraus and Keltner's research could well be valid in the UK, but I think it's much less so in the American context.

  4. AntC said,

    January 19, 2014 @ 7:04 pm

    @myl Without more depth in the analysis of the subject pool, it's hard to know what to make of the numbers.
    Indeed. What social class [by any characterisation] is defined by those "…participating for the opportunity to win small gift certificates to an online retailer."? Would anyone take part who is genuinely wealthy, or genuinely go-getting? Is this the same demographic as watch daytime TV?
    (From your critique, I think K&K fail to throw much light even on how people "… like to think of themselves …", let alone correlate it to any social construct.)

  5. Lane said,

    January 20, 2014 @ 5:36 am

    It's interesting to see so few people in the bottom-right quadrant of the last slide. One way in which America is supposed to differ from Europe is in having a strongish "leave us alone" contingent on both social and economic matters. But that is only strong-"ish", apparently. It sometimes seems like most of the people I know describe themselves as "fiscally conservative, socially liberal." But I guess that just tells you more about certain small social circles than it does about America generally.

  6. P said,

    January 20, 2014 @ 7:44 am

    If those of higher social class are more likely to think that socioeconomic differences reflect genetic differences (at least to some degree), it may be simply because they are better informed. Here's a recent study showing that the lower bound for the heritability of children's socioeconomic status in the UK is about 20%. Note that because children cannot cause their family SES, this heritability reflects "the causal genotypic factors responsible for their parents' education and occupation" and that the heritability of the parents' SES would, of course, be higher. They also show that the correlation between children's IQ and SES appears to be mainly (if not entirely) mediated genetically by alleles influencing both IQ and SES. All these analyses are based the GCTA method which uses DNA data to investigate if genetic similarity among unrelated individuals is related to phenotypic similarity among them.

    The main shortcoming of the Kraus and Keltner study is that their sample is not representative of the US population.

    [(myl) I haven't read the Trzaskowski et al. paper carefully, but you need to worry about the fact that "heritability" often doesn't mean what you probably think it does: read e.g. Cosma Shalizi on the heritability of zip codes and number of fingers.]

  7. P said,

    January 20, 2014 @ 8:27 am

    Please don't make assumptions about what I do and don't know. I am perfectly aware of what heritability means. Note that the new GCTA methodology circumvents all the traditional (and false) criticisms of behavioral genetics.

    [(myl) Without making any assumptions about your knowledge and beliefs, I need to point out that not everyone is convinced that "the new GCTA methodology circumvents" the problems that Cosma Shalizi (among many others) pointed out. For some discussion, see Evan Charney, "Still Chasing Ghosts: A new Genetic Methodology Will Not Find the 'Missing Heritability'", Independent Science News 9/13/2013, which argues that "GCTA studies are highly vulnerable to confounding by population stratification":

    Genetic studies (by whatever method) that have so far purported to identify SNPs associated with one or another trait have more often than not been false positives [18-20]. A prime cause of this has been the failure of researchers to take adequately into account population stratification. Population stratification refers to the fact that frequencies of polymorphisms can differ in different populations and subpopulations (ethnic or geographical) due to unique ancestral patterns of migration, mating practices, and reproductive expansions and contractions. Nearly all outbred (i.e., nonfamilial) populations exhibit population stratification, including populations deemed relatively homogenous (e.g., among Icelanders). One well-known example of a false association between a polymorphism and a trait was the link between the dopamine receptor gene DRD2 and alcoholism. Initial studies suggested a strong association, but subsequent investigations found none when more effective controls for population stratification were imposed. In retrospect, it is clear why this initial result was vulnerable to confounding due to population stratification: DRD2 alleles vary widely by ethnic ancestry, and ethnic differences in alcoholism rates are pronounced.

    Surely socio-economic status, however quantified, is towards the high end of variables subject to population stratification; so I would reserve judgment on Trzaskowski et al. until I've read it carefully. They claim to have avoided the problem using the "standard procedure in genome-wide association analyses to avoid artificial associations due to ethnic or other types of population stratification", but whether this "standard procedure" is effective is not clear to me — apparently Evan Charney isn't convinced.]

  8. P said,

    January 20, 2014 @ 9:25 am

    As it happens, I have critiqued Charney's silly article in the comments here. You can read the specifics there, but here's a general point I made about the peculiar style of argumentation that the likes of Charney prefer when (and only when) talking about behavioral genetics:

    The critics of behavioral genetics have this huge catalogue of possible confounds that, while no empirical support for any of them exists, somehow, as if magically, always bias the results of completely different study designs in an identical manner. The critics actually never empirically test their claims, preferring to pose far-fetched theoretical challenges. In contrast, behavioral/quantitative geneticists themselves have been eager to put their assumptions to test, developing ever more advanced methods of heritability estimation, and new methods, most recently GCTA, have repeatedly vindicated the older methods.

    [(myl) I get it that arguing about heritability in comments threads is your thing, but that's not the topic of this post or the theme of this blog, so let's leave it at that.]

  9. Matthew Hutson said,

    January 20, 2014 @ 12:59 pm

    Hi Mark,

    I appreciate the close attention you paid to my article.

    I did slip up here: Rather than attributing social class essentialism to people in the top 50% of the income distribution, I should have attributed it to people who *believe themselves to be* in the top 50%. (As for the top 0.01%, I imagine that such a high income would affect their subjective social class rank enough that actual income and not just subjective rank would be correlated with essentialism.)

    More important, I take your point about generic comparisons. Where I said "The lower classes are not merely unfortunate, according to the upper classes; they are genetically inferior,” I could have said something like, “People in the upper classes disproportionately believe that the lower classes are not merely unfortunate; they are genetically inferior.” And/or I could have mentioned the size of the correlation in the article.

    I do feel it’s worthwhile to report on even small and moderate effects found in social psychology studies. It’s rare to find a large effect, and when you do it's usually not interesting. (Gender strongly correlates with attraction to females. Surprise!) Understanding complex social phenomena requires compiling many moderately sized effects. And in select real-world circumstances, effects that are on average small in a study can largely define a situation, so it’s good to understand them.

    Also, when I talked to Michael Kraus, the paper’s first author, about his effect sizes, one thing he said was, "Some small effects have amazingly large influences. The aspirin effect on mortality from heart disease is smaller than the effects we observe here, but it amounts to an extra five years of life.”

    Whether or not science journalists go cold turkey on generic comparisons, many of us—and some scientists as well—could do a better job of highlighting the difference between statistical significance and correlation strength.

    Matt

  10. D.O. said,

    January 20, 2014 @ 3:20 pm

    Mr. Hutson.
    Your line about 0.01% is not based on any observation, but is a pure projection. How many of 169 participants of the experiment belong to 0.01% interval near any percentile? Yes, probably 0. (BTW, what is it about "a public West Coast university"? Prof. Kraus teaches at UCSF and Prof. Keltner at UCB, why be so coy? And even if we couldn't guess the university in question with any certainty, why hide it?)

  11. Jonathan Mayhew said,

    January 20, 2014 @ 3:27 pm

    Depending on how much one expects two variables to be connected, the headline might be: "Water not as wet as expected." In other words, one wouldn't think that people would think that they themselves are inferior, so a study finding only weak correlation where you would expect a strong one would be significant, but not to confirm that correlation.

  12. peterv said,

    January 20, 2014 @ 8:18 pm

    "Most people don't understand statistical ideas like distribution, correlation, and regression. And even if they understand these concepts, they often find them too complicated for everyday thinking, or impractical for everyday communication."

    Well, one plausible explanation for this phenomenon is that these ideas are genuinely and intrinsically difficult. After all, the formal mathematical representation of uncertainy by means of probabilities has been around since the 1660s, and we STILL don't have expert consensus on the semantics of probability statements. See Donald Gillies' 2000 book, Philosophical Theories of Probability, for an exposition of the four main competing semantics.

    Likewise, a decade ago, the expert members of the Uncertainty in AI (UAI) academic email list sent scores of messages and spent hundreds of expert hours trying to rigorously define the term "random variable", without, as I recall, reaching agreement.

    [(myl) On one hand, yes. See "The Pirahã and us", 10/6/2007, for some discussion of the history and progress of the relevant ideas in our culture. On the other hand, let's not exaggerate the degree of genuine intellectual uncertainty here. The fact that there might still be some controversy about the appropriateness of various versions of the axiom of choice doesn't get in the way of the use of arithmetic in engineering or financial applications, and similarly, the differences of opinion about the philosophical foundations of probability theory don't get in the way of widespread applications of statistical exploration and reasoning.]

  13. Jon said,

    January 21, 2014 @ 4:50 am

    I would suggest that the reason the correlations are weak is that the method of data collection is flawed. This type of survey (online, for a small reward) has been discussed on the BBC Radio 4 statistical programme, More or Less. There is a group of people who search for such surveys and fill them in as quickly as possible to maximise their rewards. They have no interest in maintaining accuracy, some may click boxes at random.

  14. Michael said,

    January 21, 2014 @ 5:25 am

    You've referred to the study in question (along with some of those commenting) as an experiment. That's a wrong, even misleading, description: It was a correlational study.

  15. P said,

    January 21, 2014 @ 7:21 am

    (myl) I get it that arguing about heritability in comments threads is your thing, but that's not the topic of this post or the theme of this blog, so let's leave it at that.

    Well, excuse me for thinking that a post dealing with, among other things, "social class categories […] founded in genetically based, biological differences" had something to do with heritability. I have recently found myself arguing about heritability in comment threads precisely because people are so misinformed about it. For example, Kraus and Keltner never consider the possibility that better-educated people are more likely to think that genetic differences influence class differences simply because they are better informed about scientific facts.

  16. J. W. Brewer said,

    January 21, 2014 @ 11:49 am

    I am trying to understand what makes the view "that social class [in the modern U.S.] is at least partially based in biological temperament or genetic tendency" an "essentialist" view, other than that these researchers so designated it. For example, the view that adult height (in the modern U.S.) is "at least partially based in . . . genetic tendency" is not typically a belief that that are Platonic forms of tallness and shortness instantiated, perhaps imperfectly, by various individuals, but is perfectly consistent with the statistical notion that the height of the various adults in a given population is distributed along a continuum in a bell-curvish sort of shape. It is also consistent with the notion that assessing whether a particular individual is "tall" or "short" (or neither) can often be a comparative/contextual judgment rather than an absolute one. Labeling the entire "nature" side of a nature/nurture type of dispute as "essentialism" seems to me to unfortunately obscure the rather significant possibility that "nature" is often best understood in statistical/distributional ways rather than naive-essentialist ways.

  17. J. W. Brewer said,

    January 21, 2014 @ 11:52 am

    Note FWIW that according to the late Robert Fogel adult height (at least for males – there apparently isn't a good enough dataset for females far enough back in time) was in prior centuries correlated with social class, but no longer is.

RSS feed for comments on this post