Study 1
Miller and colleagues (2020) used a randomized controlled trial design to investigate the effectiveness of the Coaching Boys Into Men program, implemented in middle schools in Pennsylvania. Initially, 26 middle schools were randomized to the treatment condition, and 26 middle schools were randomized to the control condition. After randomization, five schools withdrew from the treatment condition and four schools withdrew from the control condition. Later, 14 teams from the treatment group and 24 teams from the control group declined to participate. Athletes in sixth through eighth grade sports who completed the baseline survey and had signed assent and consent forms, or who a had a waiver for consent through their school, were eligible to participate. Two schools, one treatment and one control, were removed due to lack of parental consent. This resulted in a final sample of 973 male student–athletes from 41 private and public middle schools (20 treatment schools and 21 control schools). Sports included in the sample were basketball, wrestling, baseball, golf, lacrosse, swimming, track and field, volleyball, cross country, football, soccer.
Students were enrolled in the study from spring 2015 through fall 2016. Athletes completed 15-minute online surveys at schools at the start and end of each sports season, about 4 to 6 weeks after baseline, and again 1 year after baseline. In the treatment group, prevention advocates from local rape crisis centers and domestic violence agencies provided an introduction to the program through 1-hour training sessions and continued to provide support to coaches throughout the sport season. In the control group, coaches at control schools were asked to interact with their male athletes as usual.
There were no statistically significant demographic differences between male athletes in the treatment group and male athletes in the control group. The treatment group (n = 518) consisted of fifth graders (0.6 percent), sixth graders (12.0 percent), seventh graders (34.0 percent), eighth graders (49.4 percent), and ninth graders (0.6 percent). Of athletes in the treatment group, 49.8 percent were White, 31.9 percent were non-Hispanic Black, 3.1 percent were Asian, 1.5 percent were Hispanic, 5.2 percent were multiracial, and 4.4 were “other race.” Most of the athletes (87.3 percent) were born in the United States. They reported that their parents completed some high school (2.5 percent), were high school graduates (12.9 percent), completed some college or technical school (9.1 percent), graduated college (28.4 percent), or completed graduate school (20.1 percent); 22.2 percent of student–athletes did not report parent education.
The control group (n = 455) consisted of sixth graders (7.7 percent), seventh graders (38.0 percent), eighth graders (51.9 percent) and ninth graders (0.7 percent). Of the athletes in the control group, 59.8 percent were White, 25.7 percent were non-Hispanic Black, 1.3 percent were Hispanic, 0.9 percent were Asian, 6.4 percent were multiracial, and 3.7 percent were “other race.” Most of the athletes were born in the United States. (93.9 percent). They reported that their parents completed some high school (1.8 percent), were high school graduates (11.7 percent), completed some college or technical school (7.9 percent), graduated college (28.6 percent), or completed graduate school (27.7 percent); 18.9 percent of student–athletes did not report parent education.
There were seven outcomes of interest: a) positive bystander behavior; b) negative bystander behavior; c) adolescent relationship abuse, d) cybersexual abuse, or sexual harassment; e) recognition of abuse; f) sex-equitable attitudes; and g) intention to intervene with peers. Positive and negative bystander behaviors were measured using investigator-developed items. Participants were asked whether they had witnessed nine various abusive behaviors among their friends in the last 3 months. For each of the nine they had witnessed, they next indicated their responses from a list that included two negative intervention behaviors (“I didn’t say anything” and “I laughed or went along with it”) and four positive intervention behaviors (“I told the person in public that acting like that was not okay”; “I told the person in private that acting like that was not okay”; “I talked to our coach about it privately”; “I talked to another adult [not coach]”). Positive intervention scores were a sum of the number of positive intervention behaviors checked across the nine items, while negative intervention behaviors were a sum of the negative intervention behaviors indicated across the same nine items.
Sex-equitable attitudes were measured with mean scores from an 11-item scale with questions modified from Barker’s Gender-Equitable Norms Scale (Pulerwitz and Barker 2007). An example item is “If a girl is raped it is often because she did not say no clearly enough.” Responses range from “strongly agree” to “strongly disagree” on a five-point scale. Recognition of abuse was measured with mean scores from a 12-item scale with response options for each item (i.e., “threatening to hit them”) also on a five-point scale, ranging from “not abusive” to “extremely abusive.” Intentions to intervene with peers were measured using eight investigator-developed items. Each item indicated an abusive behavior (i.e., telling sexual jokes that disrespect women and girls), and participants report how likely they would be to do something to stop that behavior. Responses range from “very unlikely” to “very likely.”
Athletes who reported having had a heterosexual dating relationship were asked about perpetrating specific adolescent relationship behaviors toward a female partner in the past 3 months. A dichotomous composite score was calculated by combining three scales: a 10-item measure of adolescent relationship abuse (revised from the Conflict Tactics Scales, Straus et al. 1996), a 5-item measure of sexual harassment, and an 8-item measure of cybersexual abuse. Participants were assigned a score of yes or no based on whether they endorsed any of the 23 items across the three measures—for example, “called them names, like ‘ugly’ or ‘stupid.’”
Treatment effects were tested using mixed effects models that accounted for individual and school-level clustering, baseline scores on the outcomes being predicted, time, and treatment group membership. The analysis predicting any relationship abuse, sexual harassment, or cybersexual abuse was also adjusted for having ever dated at baseline. Exploratory analyses were conducted to test program effects on self-reported homophonic teasing. Subgroup analysis was conducted to test whether intervention intensity affected the strength of the program effects. Intensity-adjusted intervention effects were assessed by assigning athletes whose coaches delivered the full intensity program (meaning they covered at least nine cards over a minimum of 4 weeks) an intensity score of 1.0 and assigning athletes whose coaches did not meet these implementation benchmarks a score between 0.5 and 1.0 based on the proportion of the program they administered.
Study 2
Jones and colleagues (2021) used the sample from the randomized controlled trial described in Study 1 (Miller et al. 2020) to investigate program effects on student reports of any dating abuse, any sexual harassment, and any sexual assault 1 year after baseline.
A measure of dating abuse revised from the Conflict Tactics Scales (Straus et al. 1996) asked participants whether they had engaged in 10 behaviors toward someone with whom they were in a relationship—for example, “yelled at them or destroyed something that belonged to them.” A yes to any of these questions was coded as a yes for dating abuse. Sexual harassment and sexual assault were measured using subsets of items from the dating abuse measure. Endorsing any of three items (for example, “showed friends or posted pictures of them naked or doing something sexual”) was coded as a yes for any sexual harassment, and endorsing either of two items (for example, “convinced them to have sex, after they had said no a few times”) was coded as a yes for any sexual assault.
At baseline, 8.9 percent of treatment group athletes and 5.1 percent of control group athletes reported any dating abuse, 5.0 percent of treatment group athletes and 3.1 percent of control group athletes reported any sexual harassment, and 2.9 percent of treatment group athletes and 1.3 percent of control group athletes reported any sexual assault. Program effects were tested using maximum likelihood estimates between the treatment and control groups that accounted for all available data, and student- and school-level clustering of observations. No subgroup analysis was conducted.
