Social scientists claim to have demonstrated that bullying in early adolescence causes behavioral problems in late adolescence and adulthood, like depression, anxiety, anti-social behavior (see Juvonen & Graham, 2014 and Arseneault, 2017) and, according to a recent paper (Earnshaw, 2018), substance abuse. In response, “anti-bullying” campaigns have sprung up across the nation. I contend that “bullying research” is fatally flawed and “anti-bullying” campaigns are excessive, are hurting our children by creating unreasonable expectations and feelings of entitlement. I’ll focus on Earnshaw and colleagues’ paper because examples are usually the best way to make a general point.
Earnshaw measures one variable, “peer victimization in early adolescence,” and determines its relationship to two other variables, depression and substance abuse. “Structural equation modeling” is used to evaluate the relationship between variables to support a hypothesis that an explanatory (independent) variable, bullying, is causal of the response (dependent) variables, depression and substance abuse. Of course, “causation” cannot be assumed because that requires manipulation of variables, which is obviously impossible here. However, a strong statistical correlation between an explanatory and response variable is usually a good indicator of causation. The overarching purpose is laudable, as all parents and behavioral scientists want to determine why some young adults but not all young adults are depressed substance abusers. Is it early adolescent peer victimization, or bullying? That is, if we prevent adolescent bullying, will young and older adults be less likely to suffer depression, anxiety, substance abuse, etc.?
Most definitions of bullying paraphrase the CDC: “any unwanted aggressive behavior(s) by another youth or group of youths, who are not siblings or current dating partners, that involves an observed or perceived power imbalance. These behaviors are repeated multiple times or are highly likely to be repeated” (CDC, 2018).
With that background, I jump into the fatal flaw in bullying research: the collected data has low external reliability but results usually suggest high confidence. Earnshaw concludes, “our results from 3 US metropolitan areas show that youth who experience more frequent peer victimization in the fifth grade are more likely to engage in alcohol, marijuana, and tobacco use in the tenth grade.” Most bullying research suffers these problems: 1) self-reports, especially when querying youngsters, have very low external reliability, but data gleaned from self-reports is entered as either cardinal or ordinal values; 2) beginning behavioral states are seldom measured, such as a preexisting conditions affecting the response variable; 3) sample populations consist of volunteers, are convenience samples; and 4) attrition in longitudinal studies are very high (16.5% of Earnshaw’s sample dropped out).
The third and fourth problems should be self-evident, but bullying research usually downplays or ignores them. As to the first problem, it boils down to the adage, “garbage in, garbage out.” Take another look at the CDC’s definition of bullying. Questions posed to fifth graders by Earnshaw is typical: “How often do kids kick or push you in a mean way?” and “How often do kids tell nasty things about you to others?” The authors claim, “adequate reliability,” by which they mean internal consistency. That may be, but what about external reliability? A reasonable person would expect almost every kid in Western cultures to answer, “Yes, I’ve experienced unwanted and repeated aggressive behavior from other youths.” Converting answers to an ordinal scale (as Earnshaw does) does not solve the problem, as one kid’s definition of “unwanted” and “aggressive behavior” will often be different than another kid’s.
The second problem is even more critical. All humans vary in their reaction to the same social encounters. Earnshaw controlled for a host of variables, but not for predispositions. Some kids are born to be more anxious than others, which is why their “unwanted” or “aggressive” experience with others is different than another kid’s. More importantly, errors during development or inherited genotype predispose many humans to anxiety related disorders. If you don’t even measure a participant’s preexisting behavioral states, correlating later factors (e.g., bullying) with an ending behavioral state (e.g., depressed, substance abuse) is meaningless. Arguments that we cannot eliminate these problems in human studies are sound, but that necessarily means high confidence cannot be achieved.
Finally, the adolescent period is one of learning to “navigate” a complex social network. Learning only takes place if the child has “good” and “bad” social encounters, from which to compare reward values for use in future social decisions. Creating an environment in which “bad” social encounters are eliminated (e.g., “entitlement” to only pleasant social encounters) denies children a valuable component to learning. Many argue that this perpetuates “violence” (nonsense; that’s a straw man argument) or refer to a report from the National Academies of Sciences, Engineering, and Medicine (NASEM, 2016) that opens, “Bullying has long been tolerated by many as a rite of passage among children and adolescents. There is an implication that individuals who are bullied must have ‘asked for’ this type of treatment or deserve it. Sometimes, even the child who is bullied begins to internalize this idea.” That rejects out-of-hand legitimate, contrary evidence, which preconceived rejection is contrary to good scientific methods.
For these reasons, anti-bullying research is not scientific research.
Earnshaw, V. A., Elliott, M. N., Reisner, S. L., Mrug, S., Windle, M., Emery, S. T., … & Schuster, M. A. (2017). Peer victimization, depressive symptoms, and substance use: a longitudinal analysis. Pediatrics, e20163426. http://pediatrics.aappublications.org/content/early/2017/05/04/peds.2016-3426?utm_source=TrendMD&utm_medium=TrendMD&utm_campaign=Pediatrics_TrendMD_0