The Germans say “yes”, although their study leaves much to be desired.
This article was brought to my attention by a friend, so let the record show I take requests.
“Stigmatization of obese individuals by human resource professionals: an experimental study”
You can skip the part under “Abstract”. That’s just the teaser.
The paper itself starts with the “Background” (i.e. the “Introduction”.) This is the part where you talk about how important your topic is so that people will give you funding. They stretch out “weight stigmatization is mean” into 2 paragraphs. Why? Because that is what they want to talk about. Their experiment was a justification to say this, not the other way around. The data are all just rhetorical ammo for convincing you of this point. But I’m getting ahead of myself…
“Using a computer-based paradigm…”
You ever use a thesaurus when you were writing a paper for school to make it sound like you knew more than you did? That’s what this is. “We put a survey online” just doesn’t have the same ring….
“…we asked HR professionals from a broad range of industries and employers to evaluate standardized photographs of individuals differing in gender, ethnicity, and Body Mass Index (BMI) in relation to employment access, work-related prestige and career achievement.” This is the actual introduction.
This is their hypothesis: “Based on previous research, we hypothesized that a) HR professionals would evaluate obese individuals lowest and normal-weight individuals highest on work-related prestige and achievement, b) obese individuals would be the least likely to be hired by HR professionals and c) that this potential weight bias would be more pronounced for obese women than men.”
To an extent, I can accuse them of being blinded by preconceived biases. But that’s how science is done. And culturally, you’re not allowed saying “we shotgunned a bunch of different psychical characteristics for this and looked for anything that jumped out”, which would be the honest way to do this experiment. This isn’t a knock at these guys, that’s just how it works.
The researchers got a dozen pictures of people 40-50 years old: “two male and two female individuals were obese (average BMI of 37.9 kg/m²) and the other individuals were normal-weight (average BMI of 22.4 kg/m²). Of the normal-weight individuals, two males and two females had immigrant backgrounds and were identifiable as ethnic minorities. All individuals were photographed frontally displaying the face and the upper torso, and wore a white T-shirt. Depending on the task, photographs were either presented separately or six simultaneously on a computer screen. The computer program randomized photograph selection and display position.” They got a group of HR people they recruited from an HR trade show–sounds exciting–in Cologne to judge these pictures by the following criteria:
- “Allocating one out of six professions to a photographed individual. ‘Please allocate the following presented individuals to one of the designated professions. It is possible to allot multiple individuals to the same profession.’ We predefined two professions as high prestige (medical doctor and architect), two as medium prestige (optician and retailer) and two as low prestige (usher and cleaner) according to the Standard International Occupational Prestige Scale (SIOPS).”
- “Disqualifying one out of six individuals from being hired.‘ Very often, the first impression a hiring manager receives from a job application photo influences his or her choice of job applicant. Which of the following presented individuals would you by no means hire?'”
- “Nominating three out of six job candidates to a supervisory position. ‘Generally, all applicants were well suited for the job. Which three applicants do you think were short listed for the position?'”
This is the entirety of the experiment. Any intent to discover something other than what you will find from these prompts is 100% irrelevant. Anything of value they can possibly find will be in these results. Remember that as you read the results, discussion, and conclusion.
“We defined stigmatization using two different approaches: The first approach is based on the idea that in the absence of stigmatization, there should be an equipartition of occupational achievement and prestige…. the second approach, which is based on the idea that in the absence of stigmatization, occupational achievement and prestige should be distributed as it is within German society when taking into account gender, ethnicity, and BMI. Here, we considered the divergence in the observed allocation and nomination data from the actual data within German society as a stigmatization tendency.”
“127 HR professionals meeting the inclusion criteria participated in the study. The largest participant subgroup worked in trade and industry (23.4%), followed by health and welfare (16.4%). Companies of all sizes were equally represented.”
Their data are actually pretty nicely interpretable from their tables, which is nice. Table 1 shows us some stats about the people doing the judging. Pretty much entirely middle-aged, borderline-overweight college-educated folks with 1-3 decades in the field. Which is fine, but that’s a specific demographic. If you want to look at younger, skinnier women and how judgmental they are, this study won’t help you.
Criteria 1: This is where most of their information is. Looking over Table 2 is worth the time. Let’s see what we find out of the ordinary. What we want to look at is the “Expected N (%) in German Society”, and compare it to the “observed”. If you’re looking for discrimination, comparing it to the actual situation is going to be more illuminating than comparing it to a hypothetical system of parity. The “residual” column does the math for you. The greater the absolute value of the residual, the more the observe deviates from the expected. So a high value would be something that should catch your eye. Remember, they are sorting by job “prestige” categorized thus: “high prestige (medical doctor and architect), medium prestige (optician and retailer), and low prestige (usher and cleaner)”. Those 6 jobs are actually what the people surveyed are sorting for. “Prestige” is a word the authors are using that has no bearing on the choices. A residual over ~15 seems to be flagged as significant. It’s arbitrary, but this isn’t unreasonable. The corollary is, though, that the insignificant deviations are still data. That is, what did they get right? Where didn’t they discriminate?
- Obese women were heavily over-placed in usher/cleaning jobs and under-placed by an even amount in optician/retailer and doctor/architect jobs.
- Obese men were over-placed in usher/cleaning jobs and under-placed in doctor/architect jobs. They were realistically assessed in the optician/retailer jobs.
- Non-ethnic Normal-weight Women were over-placed in doctor/architect jobs and under-placed in optician/retailer jobs. They were realistically assessed in usher/cleaning jobs.
- Non-ethnic Normal-weight Men were over-placed in doctor/architect jobs and under-placed in optician/retailer jobs. They were realistically assessed in usher/cleaning jobs.
- Ethnic Normal-weight Women were under-placed in usher/cleaning jobs. They were realistically assessed in doctor/architect jobs and optician/retailer jobs.
- Ethnic Normal-weight Men were over-placed in doctor/architect jobs and under-placed in usher/cleaning jobs. They were realistically assessed in optician/retailer jobs.
There were 18 total slots for assignation. They only realistically assessed 6 of those slots. So right there, when you grab a bunch of people in HR and show them pictures, they don’t judge people in a way that reflects the work-world. So do they suck at their jobs? Or is there more at play here? Either way, these results do not match expectations in most categories.
Criteria 2: “42% of HR professionals disqualified the obese female when asked whom of the six displayed individuals they would absolutely not hire.” Figure 1 illustrates it, but some asshole put it in grayscale, so it becomes pretty useless for a quick visual. This is used as corroboration that obese people are discriminated against. But go back to the prompt. They had to disqualify one of the people with only the picture to go off of for 6 different jobs. In what universe do HR people pick whether or not they hire somebody based only on a picture? And if that is a system these people are using, I think the headline should read “HR people are so lazy that they’ll disqualify you for a job based only on your picture”. It doesn’t prove that HR discriminates against the obese in their job, it proves that, if forced to discriminate, they’ll discriminate against the obese. And again, only in a set of 6 jobs. The conclusion is meaningless; the prompt is a farce.
Criteria 3: “the selection frequency of obese and non-ethnic normal-weight candidates to a supervisory position differed significantly from equipartition, with the obese candidates nominated significantly less often and the non-ethnic normal-weight candidates nominated significantly more often.” This is a great example of where data visualization can help you get a better sense of the weight of the data. Take a look at figure 2. Non-ethnic normal-weight people are slightly more likely to get picked, the ethnic normal-weights a bit less likely, the obese males a bit less likely, and the obese females much less likely. So claiming discrimination against the obese is fair, but it’s disingenuous insofar that excepting females would put it on par with discrimination against ethnicities, which is there, but not super substantial. What this feature of the experiment can arguably show is discrimination against overweight females. BUT AGAIN, this is based on picture only. This is (hopefully) an unrealistic scenario. We aren’t getting information on what actually happens. There is no evidence of actual discrimination, only hypothetical.
With this, we have enough data to make our own conclusions. What do we know, and how well do we know it? Are there any methodological issues that could weaken these results?
Sample size is something I see a lot of people jump on. And to an extent, that’s good. It shows that people do understand the value of a wide sample, and honestly that really is a recent innovation in science historically. Really wasn’t cemented until the 20th century even among scientists, so if laymen at least pay it lip-service, it’s a good sign. But it’s often an unnecessary criticism. Yeah, a bigger sample size is always nice. But they got around 120, which is going to give you some useful info. My concern size-wise is the number of pictures these folks had to choose from. To what extent are they representative of the society these people are interacting in? If 4 in 12 Germans are obese, then only having 2 obese people in the set is already showing a form of bias. You’re showing people an abnormal situation and asking them to judge. But this is a trifle. Wouldn’t ruin the data. Another trifle: they should have run an ANOVA comparing the difference in the HR peoples’ characteristics with their sorting–they only compared sex. Do obese ethnic women discriminate differently than normal-weight non-ethnic males? This is double the data for very, very little effort. Oh well.
What massively undermines the data is the fact that the picture is all they have to go on. Presumably, HR doesn’t function with just pictures. This experiment does not even approximately model an actual hiring process. This whole questionnaire is essentially aimed at constructing a straw man. I’d have passed on publishing this on those grounds. However, someone less cynical than me could reasonably argue that if they were being upfront about this massive limitation that there is still data that can be published. “Using only a picture, here’s how HR sorts people into 6 jobs”. That’d be an accurate title. The usefulness of the data may be questionable, but me not knowing its use shouldn’t be grounds for censorship, should it? Ideally no. The problem is that the media isn’t going to keep that headline, and it will turn into “HR hates fat people” anyway. Is it fair for science to pay for the sins of the media? I suppose not.
So if the discussion brings up these limitations and shows all deviations and non-deviations from the expected, this would be a fair paper, albeit ones with results with limited value. Place your bets now, folks.
Some completely non-scientific gripes, though, if I may have a moment. The 1st paragraph in the Discussion is unnecessary and should be removed for brevity’s sake. Paragraph 6 should go to the introduction.
In a non-gripe, but to bring it to your attention, paragraph 8 is the obligatory “what we’ll do next if we get more funding” paragraph. Same with the Conclusion.
So did they explain away their massive limitation to this study? ” Limitations are that we do not have data on HR professionals who were reluctant to participate in this study and that we did not include résumés of the respective individuals and HR professionals. Thus, participants had to base their decision solely on the applicant’s picture.” Ugh, again, stylistic gripe, but the first sentence is two entirely different points. The second clause should go with the second sentence. Sorry. Anyway, yeah, so, they admitted the limitation, but didn’t justify their data’s value in light of this substantial limitation. You’re basing hiring decisions on only a picture. Not even a damn resume, let alone an interview. How in God’s name is that anywhere near enough to give you decent info on how HR discriminates (if at all)? Mentioning limitations does not exculpate them! And that’s in paragraph 7.
“Overall, HR professionals showed an overestimation of occupational prestige in normal-weight individuals and an underestimation in obese individuals….
We interpret our study’s data as strong evidence of stigmatization against obese individuals by HR professionals….
Our results suggest that obese individuals, especially women, are at a significant disadvantage for occupational advancement and prestige….”
What they do not discuss are any of the other deviations from the norm that HR made in these studies. Because this wasn’t about finding out the reality of the situation. They already had their mind made up. This was ammo for their arguments about discrimination of over-weight individuals. This is what they actually wanted to write: “Our data strongly suggest that interventions targeting this type of stigmatization tendency in HR professionals should be a high priority due to its significant and complex impact on the individual and society.”
They data do not suggest that in the slightest. This is a political policy prescription. Or rather, a call for such. Data do not make policy calls. These people want political action, so they half-assed science and weaponized it as a rhetorical device. You have to do what we say; science says so. Why do you hate science? People who disagree with us are just anti-science troglodytes! And the media will run with this, because methodology checking is hard, and they don’t care about that anyway.
The references are all more rounds in the magazine. Nobody’s going to check those 30 references. They aren’t there to be checked, they’re there to look intimidating. It’s Gish-galloping. You don’t need 30 references to justify seeing if HR is biased towards psychical attributes. They’re a rhetorical device to persuade you that this is just icing on the cake of settled science.
The reality is that the model is weak. Any evidence gathered is in an extreme, unrealistic set of circumstances. Does this give you probable cause for a warrant to keep digging? I’d grant that. But does their data “strongly suggest that interventions targeting this type of stigmatization tendency in HR professionals should be a high priority due to its significant and complex impact on the individual and society”? Hardly.