Statistics

Researchers: Calling Girls Fat Makes Them Fat

My brother when I was 10 said I was fat. So now I am.

My brother when I was 10 said I was fat. So now I am.

Whatever you do, don’t call somebody a “researcher”. It could condemn them to a nasty, brutish, and short-tempered life. The kind where most of their time is spent scrambling for grants, position, and anti-commonsense headlines.

Like this one, from the Boston Globe: Stop Calling Young Girls ‘Fat’. Before we move on, please to note the scare quotes.

Paper sez, “Researchers concluded that 10-year-old girls who had been told they were ‘too fat’ were more likely to find themselves in the obese range of the body mass index by age 19.” More scare quotes. Can somebody not be fat but only “fat”? Never mind.

The news was culled from the peer-reviewed letter in JAMA Pediatrics, “Weight Labeling and Obesity: A Longitudinal Study of Girls Aged 10 to 19 Years” by A. Janet Tomiyama and the inaptly named Jeffrey Hunger.

The authors worry that some kids don’t like to be called fat and take it badly when they are. The authors say that dreaded “stigma processes can begin when an individual experiences weight labeling.” Weight labeling is scientifically defined: “Recent research suggests that the negative psychological effects of weight stigma can begin when one is simply labeled as ‘too fat’ by others.”

So Tomiyama followed a couple of thousand girls, some of whom at age 10 were “weight labeled” and some not. She then looked at the same girls (those she could find, anyway) at age 19 and put them on the scale. Some of these girls turned out fat and some did not.

The older fat girls were, Tomiyama discovered, more likely to be called fat when they were younger. Conclusion? “Weight stigma may contribute to weight gain by increasing obesogenic stress processes and triggering weight-promoting coping behaviors like overeating”.

In other words, calling fat 10-year-old girls fat made them fat at 19. How? By activating their obesogenic stress processes, a super-sophisticated-sounding term which I have memorized and will drop into conversation every chance I get.

The results were provided by a statistical model (logistic regression) which “controlled” for the girls’ weight when 10, so therefore the 19-year-old girls’ fatness couldn’t have been caused by their earlier fatness. Wee p-values (actually, narrow “confidence intervals”) confirmed the statistics. And this is all the proof we need.

The authors’ final word: “Researchers, public health officials, and clinicians should consider nonstigmatizing approaches to improving the health and well-being of overweight children.” In other words, don’t tell the kiddies they’re fat because it’ll cause them to stay fat.

The theory that some people are just fatter than others, and that if a kid is fat at 10 she is likely to be fat at 19, is to be discounted. Especially when it could be society’s fault.

————————————————–

Thanks to our friend Willis Eschenbach for alerting us to this study.

Categories: Statistics

36 replies »

  1. My sister, the tiny little person that she is, always told me I was fat. I ignored the little twerp and still do.
    (I’m in the normal range for weight and BMI today and always have been.)

  2. Did they take another set of girls that were overweight and call them “thin” or “bean pole” for a few years and see if their obesogenic indifference processes kicked in and they were then thin at age 19? Seems like a logical step to take.

  3. Paul: Very good catch! For the study to be actual “science”, you would need girls who had been called “fat”, ones never labelled and ones called “thin”. Otherwise, how would you know the label was the problem?

  4. Paul, Sheri;

    For a proper scientific experiment you would need 9 groups. Groups of overweight kids, underweight kids and normal kids; some from each group called ‘thin’, some from each group called ‘fat’ and some unlabeled.

    For a truly strong study it should be done double blind, with those doing the labeling not knowing the subjects starting weight category.

  5. Maybe they should first demonstrate that overeating does indeed cause obesity rather than just assuming it. Now that would be a research project.

    Is this science, YOS? Maybe the researchers were influenced by a consideration of one or more of the deadly sins.

  6. I didn’t read the study, but this Boston Globe headline and the UCLA article make it seem that simply calling someone “fat” alone will increase the probability that they are fat later in life. The UCLA article quotes “Even after we statistically removed the effects of their actual weight, their income, their race and when they reached puberty, the effect remained.” Is income a proxy for greater access to food and more sedentary lifestyles?

    If so, um, okay…

    If not, and if I take at face value that the population generally weighs more now than previous decades (I think this is right – and would make sense given greater access to food and more sedentary lifestyles, which, remember, don’t matter here), THEN if I went back in time to any other decade long period (let’s say pre-1980), I would find that people were less likely to call young girls “fat”? Right? Is that a logical conclusion I can draw from this study?

    Of course, during those earlier periods, the young girls were less likely to be fat, therefore less likely to be referred to as such, therefore they would not be fat later in life.

    See, it’s fixed. This works.

  7. I wonder if they used an obesomatic to measure the obesogenic stress processes? The use of a manual obesometer would certainly not be appropriate for such a careful study.

  8. Hmmm, in context one would think a Longitudinal Study is a study over many longitudes, IOW, girth. Who says science papers are humorless?

  9. The theory that some people are just fatter than others, and that if a kid is fat at 10 she is likely to be fat at 19, is to be discounted.

    I clicked through several links from the original story to see if I could find the actual paper itself. No dice. I was specifically looking for information about whether they controlled for that or not becaue it’s a good point. I didn’t come up with much — journalists typically garble, fumble, or completely ignore the things that can help us discern whether a particular study might actually be good science or not. Closest I found was this statement:

    That means it’s not just that heavier girls are called too fat and are still heavy years later; being labeled as too fat is creating an additional likelihood of being obese.

    Especially when it could be society’s fault.

    I don’t need “good” science to know that insulting people is damaging and wrong, especially when repeated often and with cruel intent. Conversely when I see “bad” science telling me something that I already know from my own experiences, it would be the epitome of ridiculous to say, “Aha! What I already know to be true must now be false becaue a bunch of shrieking liberal care bear ‘academics’ did a study that I think is crappy.”

  10. PS: apologies for the ALL CAPS , I used the wrong html tag. I was looking for blockquote.

  11. You never know! Just like there can be long lasting effects from calling young children stupid!

    Anyway, I am surprised that parental weight is not included in the analysis.

    “Wee p-values (actually, narrow “confidence intervals”) confirmed the statistics. “

    Mor appropriately, the intervals for odds ratio do not include 1. Not “narrow CIs. ”

    P.S. I like the new design. Clean and simple. The two movie stars behave like RATS in the picture though.

  12. The social “sciences” discover things that are novel and true. Alas, their novel findings are seldom true, and their true findings are seldom new.

  13. Well put YOS. I have seen what I consider “good” social science being done — as far as a soft science can be good. I believe that a more reasonable commentary on this particular study is, “Why spend valauble resources of money and human effort to tell us that which we already knew to begin with?”

  14. Thanks, JH. Here’s the moneyquote:

    Baseline BMI and weight labeling status were moderately correlated (r= 0.41,P < .001). Logistic regression analy-ses (Table) evaluated the association between baseline labeling and obesity 10 years later. Adjusting for baseline BMI,household income, parental education, race, and age at menarche, being labeled “too fat” at age 10 years remained a significant predictor of obesity at age 19 years(oddsratio=1.66). The odds ratio was 1.62 when family members were the source of labeling and 1.40 when non family members were the source. These effects were not modulated by race.

    So they controlled for several factors. r=.41 is rather high for a social sciences study if I recall correctly, and look, it’s got a wee pee value!

    As this is a reacearch note only, and the data were gathered from already existing sources this kind of looks like turning over rocks to find a publishable statistical exercise for an MA candidate. Nothing to see here more than that.

  15. JH, thanks for the paper link.

    From the paper:

    Baseline BMI and weight labeling status were moderately correlated (r= 0.41,P< .001).

    In other words, if you have higher BMI, then you’re more likely to be called fat. Stop the presses, apparently we have proof that people call it like they see it!

    The odds ratios for baseline BMI and baseline labeling are similar, with BMI having a slight edge. The CIs for baseline labeling have a range of about 1, so there’s a lot of uncertainty there. I was under the impression that correlated predictors can cause problems with coefficient estimates. They didn’t remove the control for baseline labeling to see if their model fit just as well (or predicted any unknown data just as well).

    That, along with the lack of control groups, pretty much makes this study bunk.

  16. James, BMI and other factors WERE the control groups. Among other factors, they looked at “thin” girls (lower BMI) and “fat” girls (higher BMI). The results are not astounding, or particularly revelatory. This was a homework assignment novel enough to warrant being published as short note. Briggs is making much ado about nothing and most of the commentary appears to be uncritically following along with it. That’s the real issue here.

  17. Looked at the letter.
    Weight labeling was assessed by asking participants, “Have any of these people told you that you were too fat?” [followed by a list] Participants reporting “yes” to any item were considered “labeled.”

    Thus, not distinguishing the intensity of the labeling. If the participant had been told this once by her mother or teased constantly by bullies, it’s still a “yes.” It also does not seem to take into account when between 10 and 19 they experienced the stigmatizing. It does not seem to say that they were called “fat” at age 10 specifically.

    IIRC, logistic regression is useful when X is continuous and Y is binary. But if the binary is the yes/no answer to the stigma question, shouldn’t this be treated as the X and weight gain as the Y? Seems a paired t-test would make more sense if X=labelled yes/no and Y=change in BMI.

    this relationship was independent of initial BMI and thus not attributable simply to participants’ objective weight at baseline.

    BMI ≠ weight. But if so, what was the basis for calling them fat if they were not? Possibility: low initial weight at 10; gain weight by 13 for other reasons; getting called fat at 13-18; high BMI at 19.

    future research should examine these potential mechanisms

    We were all waiting for this one, admit it.

  18. BMI ≠ weight. But if so, what was the basis for calling them fat if they were not?

    Of course BMI does not equal weight. “Fat” in this context does not mean massive. BMI is body mass index, not body mass. It’s calculated as the ratio of mass in kg over height in meters squared. The resulting unit then becomes kg/m^2, not kg.

    A 19 year old woman is presumably taller than a 10 year old. One needs to factor in that change in height because obviously a taller person is going to tend to be more massive than a shorter one.

    BMI has limitations. Athletes have a higher muscle to fat mass ratio, therefore BMI overestimates “obesity” for athletes. There are issues with seemingly arbitrary standards such as where to draw the line between “normal” and “obese”, or “underweight” and “normal”. There’s the problem of “what is normal”? If average BMI is rising in a population as a whole, what was once “overweight” becomes the new “normal” over time.

    These are reasonable questions to ask. But for a homework assignment that tells us what we already know: it’s awfully rude and hurtful to call someone of any age “fat”, and those of us who are men know that it’s a death sentence to make so much as a hint that a lithe athletic woman might be a tad overweight.

    So I again ask, what exactly is the big deal here? What hay is there to be made from this entire puff piece?

    future research should examine these potential mechanisms

    We were all waiting for this one, admit it.

    Ah, I see.

  19. Brandon,

    “Controlling for” and a “control group” are two different things. I realize that the latter is practically impossible in sociology, but it means that “controlling for” will always lump in any correlated or proxy variables that aren’t examined or explicitly addressed.

  20. Basically, a control group is something you arrange ahead of time; “controlling for” means you take what you get (wrt e.g. race or BMI) and try to factor them out of the calculations. What I would have tried first cut in this case would be to pair the pre- and post-BMIs for each girl and look at the deltas, sorting them into those who were stigmatized in youth and those who were not. If race might be a factor, I would do this separately for each racial group in the data. But then I’d have to say “paired t-test,” which doesn’t sound as kewl as “logistic regression.”

    Monetary inflation sets in when more money is printed than there is value to be purchased. Each individual dollar becomes worth less. Historian John Lukacs has pointed out that this applies to the multiplication of historical documents, the multiplication of degree-granting institutions, and (I would add) the multiplication of professional sports teams. We may be seeing inflation affecting Peer-Reviewedâ„¢ Scientifical Papers.

    This instance is actually not too shabby (IMHO) compared to some others we have seen, such as this one:
    http://tofspot.blogspot.com/2014/05/somebody-paid-money-for-this.html

  21. This drivel is a really example of the logical fallacy called “post hoc, ergo propter hoc”.

    Because event or condition “A” preceded condition “B” does NOT mean “A” caused “B”.

  22. I wonder if this only works for the word “Fat?” “Cause when my sister was 11 years old, I called her a “Big jerk!”, and she was certainly bigger at age 19. Come to think of it most of the 10-11 year old people I called “You big ####” (fill in what ever you like) were all bigger at age 19! I might be on to something here. Perhaps I could get a publication in JAMA Pediatrics. “Bigogenic Stress Causes Bigness.” Yeah,I know, fat chance.

  23. From Mark J Perry’s website, Carpe Diem:

    Spurious Correlations Website, Updated Daily: Recent example: Per capita consumption of cheese correlates with the number of deaths by becoming tangled in bedsheets from 2000-2009, correlation coefficient = 0.947.

    Spurious Correlations Website is at http://www.tylervigen.com/.

  24. My first reaction when I read about some new study is to ask “how’d they measure that?” If the answer clearly can only be “they asked people some questions”, then I tend to dismiss the study out of hand. I believe that among the most unreliable ways to find out something about somebody is to ask them.

    I love the Spurious Correlations web site, Terrence; thanks! Which reminds me – was that statement “Correlation is a necessary but not sufficient condition for causation” ever resolved here? I’m still not buying the first part.

  25. James, my apologies for missing the distinction between control groups and controlling for. I see that I was using them sloppily. My main problem with your post was the statement, “that pretty much makes this study bunk.” I’m glad to see that you recognize that control groups are pretty much impossible in social sciences. Sometimes we must use the data at hand. Whenever we study humans themselves, in the back of all our minds should be the thought that understanding such complex animals is fraught with uncertainty no matter how well designed the “experiment”. That doesn’t necessarily make a given study completely bunk. And it doesn’t make the whole of “science” bunk either, as some comments here seem to suggest.

  26. I know of some situations where two causally-linked variables failed to correlate because the X was not measured over a sufficiently wide range.

  27. I’m sure it happens all the time. Such are the limits of finite human understanding.

  28. Brandon,

    I was in the snarky mood of the posting when I said “bunk” 🙂 I was also not super clear in my first post on the meaning of control, so no worries!

    I think there are valuable things to be learned from the social sciences. The article discussed here, in my view, is not one of them (hence, bunk). I’m also bringing my own bias in, which is that we should spend less time controlling external behavior, and rather train ourselves (and our children) to be able to control their responses to it.

    In other words, if can only say either “don’t call people fat” or “sticks and stones”, the latter is worth more. The first can certainly be true, but I don’t think it’s the most valuable thing to do for children.

  29. James: Yeah, I missed the light-hearted nature of the snark when I read it. I had my own biases that I bring to the table when I read Briggs, so you definitely caught some unintended friendly-fire.

    “I’m also bringing my own bias in, which is that we should spend less time controlling external behavior, and rather train ourselves (and our children) to be able to control their responses to it.”

    That is worth the price of admission right there. I think one of the main messages of the original piece and I just missed it because I’d already gone into “kill the silly theist” mode. One of the reason I read this blog is to stop having that reaction so much.

    Unfortunately, I haven’t the energy to do either those two conversations justice, so I will leave you now with my thanks.

Leave a Reply

Your email address will not be published. Required fields are marked *