Blinded by Gender: A Comment on
The Dangerous Experiment on Teen Girls
This post is related to (but not a part of) the Youth Suicide Rise project
Jonathan Haidt has recently published an article in The Atlantic titled The Dangerous Experiment on Teen Girls that provides important facts about recent changes in the mental health of girls while simultaneously providing severe misinformation about similar developments affecting boys.
Haidt is right in pointing out that the mental health of adolescent girls has deteriorated over the last 10 years and that this coincides (partly) with the rise of social media. He is also right in pointing out that the use of social media by girls is associated with substantial risk elevations of psychological disorders.
Unfortunately Haidt also severely misinforms and misleads readers about the deterioration of mental health among boys as well as about the associations between digital technology and psychological disorders for boys.
Girls not Boys
Haidt mentions "girls" more than twenty times in his essay, always in a context that emphasizes harm to girls over harm to boys. On the other hand, the seven times that Haidt mentions "boys" is each and every time within the context of asserting a comparatively lesser harm experienced by boys.
The relentless focus on how declines in mental health supposedly affected girls far more than boys is not justified and misrepresents a fundamentally more complex reality.
And Haidt does not merely mislead -- he also outright misinforms readers.
Much Smaller Increases for Boys?
Haidt writes "Girls in the U.K. also experienced very large increases in anxiety, depression, and self-harm (with much smaller increases for boys)" and links this assertion to a study published in the BMC Psychiatry journal.
This study, however, found that percentage increases were often much stronger for boys, not girls: e. g. 240% vs. 110% increase of depression for boys vs. girls ages 13-16 and 190% vs. 90% increase of self-harm for boys vs. girls ages 13-16.
Why would Haidt characterize such results as much smaller increases for boys?
For Girls Only?
Puzzling is also Haidt's report of an ER study when he writes that there were increases in ER admissions due to self-harm "for girls only" -- and yet the relevant graph indicates that admissions for boys age 15-19 increased by about 40% from 2009 to 2015 (the study does not offer year-to-year data tables).
Haidt then adds that "From 2010 to 2014, rates of hospital admission for self-harm did not increase at all for women in their early 20s, or for boys or young men, but they doubled for girls ages 10 to 14" -- but rates for boys age 10-14 might have actually doubled from 2009 to 2013 (it is hard to tell precisely from the graph).
Why are several statements by Haidt on trends for boys so misleading?
Visual Misinterpretation of Graphs
The hint to answering that question is in an earlier part of the essay, where Haidt precedes a graph of depression trends with the assertion that "when we look at what happened to American teens in the early 2010s, we see many such turning points, usually sharper for girls."
Yet when we do look at the graph carefully, we realize that the prevalence of depression has roughly doubled for both girls and boys. The reason this increase looks 'sharper' for girls is because it is a much steeper slope if we go from 12 to 24 than if we go from 4 to 8 over the same time period on a graph.
As Haidt confirmed to me, when looking at the graphs in the U.K. study he missed the fact that relative changes there were often similar or greater for boys.
In other words Haidt did not intend to mislead his readers -- rather he misled himself by relying solely on visual impressions from graphs and then inadvertently misled his readers as well.
Elevated Risks
There is yet another area of major misinformation regarding boys in The Atlantic essay.
Haidt tells readers that the reason some studies found (supposedly) 'small' or 'tiny' correlations is that they "lump all screen-based activities together" and because they "lump boys and girls together". Haidt then points out that "Girls who use social media heavily are about two or three times more likely to say that they are depressed than girls who use it lightly or not at all."
This passage makes no sense unless Haidt is implying that there are no elevated risks for boys comparable to the doubling and tripling of risk that he mentioned for girls. In fact Haidt implies that elevated risks for boys are so minuscule that merely lumping boys with girls dooms the overall associations to be small at best.
In reality this is completely false.
Digital Technology
For example, on the 2017 High School Youth Risk Behavior Surveillance System (YRBS) survey the top 20% of digital tech users among teenage boys were nearly five times as likely to report a suicide attempt in the past 12 months as students with only about 1 hour daily use. More robust measures, such the risk ratio comparing the heavy users to all students with lesser use, also revealed strong risk increases (well over double the majority risk).
Strong risk elevations with various psychological disorders appear regularly on large surveys of adolescent boys, and they do so without the need to examine any specific subset of digital technology, such as playing video games (see Links below).
Social Media
It turns out that even if we look specifically at social media, the notion that boys have far lower risk elevations than girls is also false.
In the study relied upon by Haidt, the heavy users of social media (5 or more hours a day) among girls have roughly double the depression risk of all the other girls -- and yet the same is true for boys. The risk elevation associated with going from moderate to heavy and very heavy use is also nearly identical for boys and girls.
Summary
The Dangerous Experiment on Teen Girls contains misinformation that could easily cause readers to mistakenly conclude that there has been no severe deterioration of the mental health of boys. It also contains misinformation that could cause readers to disregard any concerns about the role that digital technology may play in the increases of psychological disorders among boys.
The single-minded focus on girls is not justified by current evidence -- and neither is the dismissive attitude toward potential risks for heavy users of digital tech among boys.
Ironically, the more Haidt might be right about risks for girls, the more dangerous his current message might be for boys.
Appendix:
Scope of the Critique
This critique is concerned with one aspect of The Atlantic essay, namely its mistreatment of gender issues. My lack of comment on other aspects of the essay does not indicate agreement.
Self-reports
The elevated risks mentioned above are typically based on self-reports of both digital tech use and psychological indicators. On one hand, the elements of approximation and faulty memories in self-reports could severely decrease the actual risk ratios present in reality. On the other hand, there could be some hidden biases, such as depressed teens tending to over-estimate their digital tech use time.
That said, it should also be noted that objective does not mean accurate. We can objectively measure how long a TV is turned on, but this can be a highly inaccurate measure of time spent actually watching the TV.
Suicide
Haidt mentions suicide fatalities only once in passing, despite deaths being the most tragic outcome of mental health problems among kids.
Had U.S. suicide rates remained roughly at 2007 levels, there would have been about 5 thousand fewer teen suicides among teenage boys versus about 3 thousand fewer suicide deaths among teenage girls between 2008 and 2019 (see Data section below).
The excess deaths are considerably higher for boys (absolute change) even though suicide rates rose faster for girls (relative change). This is the reverse of depression trends in U.K. described in the study relied upon by Haidt.
Recent developments in adolescent mental health are a lot more complicated than they seem to be from The Atlantic essay -- and not just regarding gender but also regarding age and other factors (the details are beyond the scope of this critique).
Neurodevelopmental Disorders
The U.K. study linked to by Haidt shows that neurodevelopmental disorders increased the fastest, yet such trends seem to be largely ignored in discussions dealing with the recent deterioration of mental health in adolescents.
Is it a mere coincidence that such disorders have also been increasing rapidly? If not, could it be a hint there might be, say, some new environmental poisoning factor like micro-plastics affecting the brains of kids?
Mental Health and Gender
The prevalence of psychological disorders typically considered in mental health research is the greatest for anxiety and depression, and these in turn are far more common in females than in males.
Yet does this really mean that females are much more likely than males to experience mental health problems?
I doubt this (see the Note at the end) and I am unaware of any professional consensus on this matter.
Social Media and Depression
The difference between girls and boys is that girls with no SM use at all have substantially lower risk than those with moderate use (2 hours a day), while for boys there is no difference. If we let 1.0 be the (separate) risks for girls and boys in the moderate use category, which is quite robust as it contains about one third of both girls and boys, then the relative risks look thus:
Girls: 0.6 0.8 1.0 1.4 2.1 (here 1.0 is about 18% risk of depression among girls)
Boys: 1.1 1.1 1.0 1.7 2.1 (here 1.0 is about 7% risk of depression among boys)
As we can see, the relative risk elevation is similar for girls and boys going from moderate to heavy use.
Statistical concerns aside (the zero use category has only 4.4% of all girls), the results indicate that moderate use of social media is harmless for boys but may be harmful for girls. That could indeed be due to the body image issues and the relational conflicts that Haidt mentions in his essay.
Yet is this a valid reason to characterize social media as potentially harmful primarily to girls?
To put such a message in perspective, consider obesity.
No doubt that mild obesity has far greater social penalties for girls than it does for boys, and this likely affects mental health as well physical health of girls more than of boys.
And yet hardly any doctor would characterize obesity as mainly a female health issue, not to mention neglect or outright belittle the risks for males. The reason few doctors would do so is because the greatest health risks are due to the more severe forms of obesity -- and here the consequences for men are in general as grave as they are for women.
Any doctor who frames obesity as primarily a female health issue would likely face a backlash from colleagues concerned that such a misleading message might set back efforts to stem obesity among males.
Similarly, characterizing social media as primarily a threat to girls could easily lead the public to under-estimate the potential risks for boys.
Haidt vs. Twenge on Girls vs. Boys
Although Jean M. Twenge emphasized harms and risks to girls in the past, I've never seen her imply that harms and risks to boys are not substantial -- and never seen anything from her remotely as dismissive of harms and risks to boys as is The Atlantic essay by Haidt.
Perhaps the strongest emphasis on girls by Twenge can be found in her paper Increases in Depression, Self‐Harm, and Suicide Among U.S. Adolescents After 2012 and Links to Technology Use: Possible Mechanisms -- yet as we can guess already from its title, even there she made it clear that she considers harms and risks to boys to be substantial.
As recently as this summer, Haidt and Twenge published an opinion in The NY Times, titled This Is Our Chance to Pull Teenagers Out of the Smartphone Trap, where they warned about worldwide increases of teen students feeling lonely in school -- all teens, not just girls or mainly girls. Their proposed solutions also made no distinction between girls and boys.
Meta-Studies
Haidt is extremely misleading when he follows a passage about supposedly small correlations revealed in three meta-studies by arguing that girls and social media are the exception due to their use of social media being associated with greatly elevated risks of depression.
That argument makes sense only if the three meta-studies mentioned by Haidt looked at elevated risks.
None of them did.
In fact none of them allowed any example of elevated risks to appear within their texts.
In other words the elevated risks revealed in numerous papers by Twenge, starting back in 2017, or in papers by Kelly and others since, were simply ignored by the authors of the three meta-studies.
If you deliberately omit inconvenient evidence, you can always say there is none. It's that simple.
Needless to say, censorship is not a valid scientific methodology.
Haidt was fooled by authors on the very border of bad science and outright pseudo science who are being empowered by journal editors who should know better. And Haidt is now unfortunately helping to fool the public on this matter.
A Public Intellectual
I have no reason to think that Haidt tried to intentionally mislead his readers -- in fact I am convinced he tried to be as truthful as one can be about a complex matter in a short essay.
Haidt is a public intellectual in the best sense of the term -- someone who tries to use his intellect to serve the public, not just his career. He has entered the perilous waters of digital tech debates out of a genuine concern over its effects on society, especially its adolescent members.
Haidt is also exceptional in welcoming and encouraging criticism, practicing what he believes about the necessity of public dialogues. He admits errors without needless excuses and qualifications and he tries to fully understand the arguments of his opponents. All this can be seen on sites that Haidt has set up to address such criticism -- he even lets other researchers to add content directly (see Links below).
The reasons the best of us can falter can be mundane. I suspect Haidt may be dealing with too many disparate issues at once, not always finding enough time to sit back and review evidence and arguments with sufficient care and prior consultation with colleagues. Haidt has also waded into an area of psychology rampant with misuse of statistics, a subject that is not one of his strengths.
There is no shame in having only rudimentary understanding of statistics. One must be careful, however, not to fool oneself and not to be fooled by others when adopting statistical arguments without full comprehension.
Links:
The Dangerous Experiment on Teen Girls
Temporal trends in annual incidence rates for psychiatric disorders and self-harm among children and adolescents in the UK, 2003–2018
Trends in Emergency Department Visits for Nonfatal Self-inflicted Injuries Among Youth Aged 10 to 24 Years in the United States, 2001-2015
Increases in Depressive Symptoms, Suicide-Related Outcomes, and Suicide Rates Among U.S. Adolescents After 2010 and Links to Increased New Media Screen Time
Social Media Use and Adolescent Mental Health: Findings From the UK Millennium Cohort Study
Increases in Depression, Self‐Harm, and Suicide Among U.S. Adolescents After 2012 and Links to Technology Use: Possible Mechanisms
This Is Our Chance to Pull Teenagers Out of the Smartphone Trap
Note: for my critiques of some aspects of the Smartphone Trap essay by Haidt & Twenge, see:
The Perils of Improper Terminology: A Comment on The Smartphone Trap
The Crime of Parsimony: A comment on The Smartphone Trap
Besides the YRBS data analyzed by Twenge back in 2017 and the Millennium Cohort Study analyzed by Kelly (both see links above), other massive survey data that show substantial risk elevations for boys (either alone or together with girls) include National Survey of Children's Health (NSCH) and Monitoring the Future (MtF):
Associations between screen time and lower psychological well-being among children and adolescents: Evidence from a population-based study
Why increases in adolescent depression may be linked to the technological environment
The three meta-studies that Haidt cited as evidence that there are no substantial associations between digital tech use and psychological disorders:
The association between adolescent well-being and digital technology use
Annual Research Review: Adolescent mental health in the digital age: facts, fears, and future directions
Like this meta-analysis: Screen media and mental health.
A site set up by Haidt where he collects and respond to criticism:
Social Media Use and Mental Health: A Review
Data:
The 2017 YRBS data can be downloaded from https://www.cdc.gov/healthyyouth/data/yrbs/data.htm in a raw ASCII format or in Access format (plus there are scripts to convert the ASCII data into SAS or SPSS). As I query the 2017 ASCII file directly with bash commands like
total_students=$(echo_data $1 | awk -v i=$i -v j=$j '{FS=""}{print $i $j}' |grep -c [1-7][12] )
it would be of little use for most readers if I post any code. Instead, I plan to write R and Python scripts to query the combined data set for all the past YRBS surveys and post these once completed.
CDC suicide data can be queried easily with their WISQARS tool at https://wisqars.cdc.gov/fatal-reports -- raw ASCII data can be downloaded from their CDC WONDER site (https://wonder.cdc.gov/). Below is a screenshot of suicide fatalities for boys and girls age 10-19 -- these allow approximations of excess deaths due to the rises starting at 2008.
Child Suicides Double in a Decade
The Rise: Boys and Girls
The Rise: Excess Deaths by Child Age
Note:
Regarding women being more psychologically troubled than men, my doubts (presented here with the caveat that I am not a psychologist) are partly motivated by the following considerations:
Could it be that we a) under-count externalizing disorders such as physical and verbal aggression and b) over-count internalizing disorders like anxiety and depression?
Depression per surveys is routinely defined as mere two weeks of incapacitating sadness, which is certainly an important indicator of mental health, but should it be counted an outright disorder?
Is not such passivity often a protective reaction to trauma, such as the loss of a parent? Inability to concentrate during certain activities must have been a huge risk during human evolution, where it could have lead to death, especially when combined (in youth) with inexperience.
Adolescent boys take more risks and, to put it bluntly, seem to have been more expandable during human evolution than girls, so it makes sense that mild types of 'depression' -- really a form of passivity -- would be more common among girls as a protective mechanism.
So I am not convinced the high prevalence figures for female depression reflect true disorders -- it may actually indicate that females are better at coping with trauma, at least as long as safer equals better. For true disorders, perhaps we should count only chronic or severe forms of depression.
Admittedly I am on a thin ice here as I am not a psychologist.
I never expected to sound like some gender studies feminist damning a whole field outside of my expertise for sex bias, but does it really make any bloody sense to judge females as far more likely to be mentally troubled than males? And is it a coincidence that such judgments are based on measures defined in an era when psychology was completely dominated by men?
No comments:
Post a Comment