Nonheterosexual adolescents are vulnerable to health risks including addiction, bullying, and familial abuse. We examined whether they also suffer disproportionate school and criminal-justice sanctions.
The National Longitudinal Study of Adolescent Health followed a nationally representative sample of adolescents who were in grades 7 through 12 in 1994–1995. Data from the 1994–1995 survey and the 2001–2002 follow-up were analyzed. Three measures were used to assess nonheterosexuality: same-sex attraction, same-sex romantic relationships, and lesbian, gay, or bisexual (LGB) self-identification. Six outcomes were assessed: school expulsion; police stops; juvenile arrest; juvenile conviction; adult arrest; and adult conviction. Multivariate analyses controlled for adolescents' sociodemographics and behaviors, including illegal conduct.
Nonheterosexuality consistently predicted a higher risk for sanctions. For example, in multivariate analyses, nonheterosexual adolescents had greater odds of being stopped by the police (odds ratio: 1.38 [P < .0001] for same-sex attraction and 1.53 [P < .0001] for LGB self-identification). Similar trends were observed for school expulsion, juvenile arrest and conviction, and adult conviction. Nonheterosexual girls were at particularly high risk.
Nonheterosexual youth suffer disproportionate educational and criminal-justice punishments that are not explained by greater engagement in illegal or transgressive behaviors. Understanding and addressing these disparities might reduce school expulsions, arrests, and incarceration and their dire social and health consequences.
Comments
Author's response
We appreciate the comment regarding the incorrect RR in the abstract of our article. This was a typographic error. The text of the article contains the correct RR regarding the rate of serious bacterial infection in neonates with fever and RSV infection.
Criminal-Justice Sanctions Minor and Mostly Against Bisexual Youth
Himmelstein and Bruckner (2011) have presented an interesting analysis of Add Health data over waves 1 and 3. They concluded that “nonheterosexual youth, particularly girls, have greater odds than their peers of experiencing school and criminal-justice sanctions” (p. 49). They did report that “Of Add Health respondents who indicated same-sex attraction in wave 1, only 27% reported such attraction in wave 3” (p. 51), a result which concurs with several other research reports they did not cite [1, 2].
I recreated much of their data based on the percentages and sample sizes reported in their tables and reanalyzed it. Consequently, I believe that their conclusion, while technically accurate, is misleading in at least two ways.
First, the effect sizes associated with their outcomes are relatively small, even when significant statistically. For example, their largest odds ratio in Table 3 (p. 53) was 4.34 (p = .007), which in Table 2 (p. 53) featured 1.5% of female nonheterosexuals arrested after age 18 compared to 0.3% of heterosexual females (p = .0001). Though significant, the effect size (ES, Cohen’s d) associated with the bivariate result was only 0.13, below the 0.20 magnitude deemed “small” [3]. Male expulsion from school was significant (p < .05) in Table 2 – 11.2% for heterosexuals and 6.7% for nonheterosexuals – with an effect size of only 0.16, again below the small level. The largest effect size I was able to locate was 0.29, well below the “medium” level of 0.50, for females stopped by police, even though the results were very significant (p = .0001), and appeared to involve substantial percentage differences - 19.5% for nonheterosexuals and 9.5% for heterosexuals. Thus, even some of their strongest outcomes involve relatively small effects.
Second, their results minimize the differences between bisexuals and homosexuals. If one compares only those adolescents who were 100% homosexual or mostly homosexual (N = 227) versus those who were 100% heterosexual (N = 13,490) as shown in Appendix 2 (p. 56), the differences for expulsion from school (ES = 0.06), juvenile convictions (ES = 0.07), and adult convictions (ES = 0.10) were not significant statistically, though the outcomes were worse for the heterosexuals. The results for under-18 arrests were significant (p < .05, two-sided Fisher’s Exact Test, ES = 0.16) but were worse for heterosexuals. The results for over-18 arrests were significant by a one-sided Fisher’s Exact Test (p < .05, ES = 0.19), again worse for heterosexuals. Though the effect sizes were below “small”, all five outcome variables favored the homosexual group of adolescents. To conclude that homosexuals are being sanctioned in the criminal justice system is not accurate; what would be accurate would be to say that bisexuals were being sanctioned more often than either homosexuals or heterosexuals. An important question for future research would be why and how the criminal justice system would be “focusing” on bisexuals rather than homosexuals if indeed it was biased against nonheterosexuals. Is it possible that bisexuals are actually one sexual orientation or the other but are rebelling against society in numerous ways, one being to practice more sexual diversity? I don’t claim to know the answer, but the question would appear to invite further research.
References
1. Dickson N, Paul C, Herbison P. Same-sex attraction in a birth cohort: prevalence and persistence in early adulthood. Social Science & Medicine 2003;56(8):1607-1615. 2. Kinnish KK, Strassberg DS, & Turner CW. Sex differences in the flexibility of sexual orientation: a multidimensional retrospective assessment. Archives of Sexual Behavior 2005;34(2):173-183. 3. Cohen J. A power primer. Psychological Bulletin 1992; 112(1): 155- 159.
Conflict of Interest:
Dr. Schumm served as an expert witness in a trial in Florida in 2008 concerning gay adoption rights.
Overstated Conclusions
There is a sharp disconnect between the statistical findings presented in this report and the authors' conclusion that “nonheterosexual youth suffer disproportionate educational and criminal-justice punishments that are not explained by greater engagement in illegal or transgressive behaviors.” With the exception of the outcome variable “ever stopped by police,” all of the bivariate correlations presented in Table 2 were found to be non-significant in the multivariate analysis (Table 3) after adjusting for race, behavior, and socioeconomic status, and correcting for multiple testing (“only findings with a P value of <.001 were deemed statistically significant”).
With regard to the “ever stopped by police” outcome variable, the results of the multivariate analysis were inconsistent across the three measures of sexual orientation: of the nine tests, just three were statistically significant. In the stratified analysis by sex, one in six tests met the criterion of statistical significance. In sum, the categorical language of the authors' conclusion is based almost entirely on a pattern of findings that did not reach the stated criterion of statistical significance. Under the circumstances, the conclusions should have been stated in more tentative terms.
Conflict of Interest:
None declared