As was in the news everywhere, in the penultimate round of the AFL, Crows star Izak Rankine made a homophobic sledge of some Collingwood player. Rankine has been suspended and has publicly apologised, and the incident has inspired about a thousand articles. Because homophobia in the AFL is a big problem. Maybe.
I am not convinced that the many tons of media moralising over Rankine’s “highly offensive homophobic slur” were warranted. For a sport where, even these days, players can knock the stuffing out of each other, the reaction to Rankine’s sledge struck me as over the top. I’m not going to take that on here, however. I am just going to take on one aspect of one article.
A couple weeks after Rankine’s sledge, there appeared in The Age an article by Erik Denison, a behavioural science researcher at Monash University. Denison does not mention Rankine, instead framing his article on Mitch Brown’s post-AFL coming out as bisexual, with a focus on “the AFL’s harmful hypermasculine and homophobic culture” and what to do about it. Fine. It is perhaps a little odd to expect the AFL to be other than hypermasculine, but Brown argued from his personal experience, which he clearly found painful. There is reasonable stuff to discuss.
Much less reasonable, however were the headline and opening of Denison’s article. The dramatic title to Denison’s article was,
More than 100 current AFL players are likely gay. One thing would make it easier for them to come out
Really? Out of the 700-ish AFL players, 100+ of them “are likely gay”? The “likely” is poorly placed but, stepping over the resulting ambiguity, Denison’s (presumed) claim seems a hell of a stretch. But Denison backs it up in his second paragraph:
Based on recent high-quality studies, about 15 per cent of young people identify as something other than heterosexual. In the AFL alone, this should mean more than 100 openly gay or bisexual players. Instead, there is only one player, who retired in 2016.
So, 15% of 700 is 105, and there you go. Or not.
One cannot translate “15 per cent of young people” to 15% of AFL players, as Denison has done. If for no other reason, and given AFL homophobia is as prevalent and as damaging as is being claimed, one would imagine that this homophobia would significantly lower the participation percentage. But, more fundamentally, what is the origin of that “15 per cent of young people”?
Denison’s “high quality studies” link provides scant evidence of the associated claim. Denison’s link is to a paper by Denison and his Monash mates, which is on the exclusion and discrimination of gay-et-al people from sports. The paper may be valuable in this regard but it is simply not concerned with the prevalence of homosexuality and the like. The paper appears to contain just one relevant sentence (p 391):
For example, the American Centers for Disease Control and Prevention’s (CDC) biennial youth surveillance surveys of high-school students now provide reliable data across a wide range of areas relevant to sport managers, including showing 14.6% of high-school aged young people identify as LGBQ (Kann et al., 2018), and 1.8% identify as transgender (Johns et al., 2019).
So, these “high quality studies” are not studies but, rather, surveys. This matters; a survey is not reality but only a reflection of reality, which may be less or more distorted. What also matters is that we have gone from “young people” to “school aged young people”. American school aged young people.
Is how a bunch of American high school kids choose to describe themselves likely to be an accurate indicator of AFL footballers. Even ignoring the circuitous route taken, why did Denison choose to go there?
Denison could have chosen to go with the Australian Bureau of Statistics. This is still a survey rather than reality, but it is at least a survey closer to home and to the population of interest, and more recent. Why did Denison not do so? Maybe because of what the survey indicates:
Taking the ABS survey as reality, and an AFL reality, and there are all manner of reasons to doubt both of these, this suggests a ballpark of around 8% of AFL footballers are gay: fifty or so. I very much doubt the reality is close to that. What there is little doubt about is the accuracy of Denison’s headline; there is not a snowflake’s chance in hell that a hundred AFL footballers are gay.
Does this matter? To the salient point, of whether or not homophobic sledges are to be considered off limits, and how egregious, I cannot see that it does. Which makes Denison’s gilding the lily all the sillier. It would be a lot easier to buy Denison’s genuine points if the article hadn’t led off with such fantastic trumpeting.
Let me offer two comments.
Perhaps Denison did not write the headline for his article.
I gather that ABS will ask questions about ‘sexual orientation and gender’ in the Census to be held next year.
Perhaps. But he presumably wrote the second paragraph.
Let me offer an alternative hypothesis – the university system is broken – ideology, budget cuts, publish or perish and corporatisation have fundamentally weakened academic behaviour.
An alternative to what? Of course the university system is broken, but what are you suggesting is the effect of that here?
That some academics will pursue a headline whatever it takes
Nah. The guy is much more a true believer than that.
Oh I see – I looked at his LinkedIn page (about the main use for LinkedIn to see what someone might actually be like) and it’s all about combating homophobia. In which case he really should be familiar with the generally estimated of proportion of people identifying as gay. So it would appear he’s driven more by ideology than rigour in this case. Indeed he self-describes as “using science to drive meaningful change to negative behaviours in workplaces, sports clubs, and schools”.
Typical Simpson Paradox fallacy. Any person who claims to be a researcher must know it.
How is this Simpson?
Simpson’s Paradox is a statistical phenomenon where a trend that appears in different groups of data reverses or disappears when the groups are combined. “Trend” in this case doesn’t necessarily refer to a time series phenomenon—it can be a correlation or association. What matters is that a hidden variable (confounding factor) isn’t properly accounted for when groups are combined or analysed separately.
The author is incorrectly applying percentages measured across the general population to a particular segment without considering confounding variables. This is a fundamental statistical error. It’s well-documented that certain occupations have higher-than-average percentages of LGBTQ+ individuals. According to a 2021 UK census analysis, these include air travel attendants, actors and entertainers, leisure and theme park attendants, and coffee shop workers. The same principle applies to sports.
Women’s sports, particularly soccer, basketball (WNBA), and rugby, have significant numbers of openly lesbian and bisexual athletes, with figures reaching 33% of WNBA players and 12% at the 2023 Women’s World Cup. For men’s sports, there are notably more openly gay Olympians in diving, equestrian events, and swimming, though comprehensive data remains limited for non-Olympic sports. The point is that some sports might have a higher percentage, some lower and some might not have at all.
I know what Simpson is. I know the guy is improperly pretending that a stat for Group A can be presumed for group B. I don’t know why this is a case of Simpson?
But it’s even worse than that as he uses the 15% being gay for his study, which is well outside the range of figures used almost everywhere else of more like 10%.
I think it is all well explained in my post
Well done, then.