The study was based on a two-page survey instrument sent to high school students around the United States. Commercial lists of high school students nationally were identified. A composite list of more than eight million names and home addresses from among approximately 15 million high school students in the country was used to draw a random national sample of 20,000, stratified by juniors and seniors. In April 2000, surveys were mailed to 20,000 students at their home addresses. A follow-up, first-class postcard to 5,000 students for whom we still maintained access to their names was mailed one-week after the survey mailing encouraged students to respond. The mailing lists were destroyed following this one-time use for research purposes only. The post office returned as undeliverable 5.46 percent of the original surveys. We therefore assumed that at most 18,600 arrived at their destination. We received in the mail 1,541 completed, useable surveys. The sample, therefore, reflects an 8.28 percent response rate.Survey Instrument
One survey instrument was developed for high school students. A cover letter announced the survey, provided enough information about the survey to constitute informed consent, and stated the due date for the survey's return. A postage-paid return envelope was enclosed. The instrument covered the following areas:
Background: Sex, race, state of residence, location (urban, suburban, rural), grade in school, school type (public, private, home), and membership in any type of group.
Definition and Experience with Hazing: Students were asked, for each type of group, if they: were hazed to join, didn't join because they were afraid of hazing, or left the group because of it. Students were asked at what age they were first hazed, if ever; if they would report hazing; and what they thought were the most effective strategies for preventing hazing.
Initiation Activities: Finally, the survey asked what activities were expected of them to join a high school group or team. Four categories of initiation behaviors were used from last year's study of NCAA athletes: community building, humiliating, substance abuse, and endangering. The first group of behaviors had the most changes since we'd learned the most about it from last year's survey. The last groups of behaviors remained fairly stable with a few items edited and a few items added. The final behavior list had 36 behaviors as opposed to 27 behaviors on the 1999 survey form.
Open-ended Questions: Students were consistently asked for their input on such items as: other prevention strategies, reasons for participation, experiences as a consequence, feelings afterwards, and activities expected of them.
New Items developed for this survey were categories of high school groups, motivational factors, consequences of hazing.
Most of the students who responded attended public school (90%), whereas five percent attended church schools, five percent attended other private schools, and one percent was home schooled. All geographic types were represented: 46 percent suburban, 30 percent rural and 25 percent urban, as measured on a self-perception scale. Successful students were disproportionately represented in the sample: 84 percent reported average grades of A or B, 14 percent C, and one percent D or F.
Synthetic Estimation of the Population
The sample of respondents was regionally proportional to the Federal Current Population Survey data for sixteen and seventeen year olds in the United States in September 1999. The response from mid-western students was slightly higher than to be expected. The other regions are in proportion when the mid-west is discounted.
|Gender||Sample %||n||National %|
|Ethnic Origin||Sample %||n||National %|
Findings are therefore reported based on the sample weighted by gender and region to best approximate anticipated national percentages as closely as possible given these data. For weighting, the high school student population was estimated by taking the federal Current Population Survey data for sixteen and seventeen year-olds in September 1999 and multiplying by 71 percent, the federal estimate of school participation. To estimate the entire high school-aged population, the entire number of fourteen and fifteen year-olds was added to that amount. The entire high school-aged population was therefore estimated at 13,628,460.
Based on the survey, 91 percent of the students joined high school groups and of those 48.4 percent were subjected to some form of hazing. Based on this finding, an estimated six million high school students in the United States were subjected to hazing in the past four years, or more than one and a half million students per year. Given the limitations of this study, more studies are needed to verify and refine this estimate.
There were approximately 8,096,967 16 and 17 year olds in the United States in September 1999. The U.S. Department of Education estimates that 71 percent of 17-year-olds are enrolled in high school. Discounting by 71 percent, we estimated 5,748,847 juniors and seniors in high schools nationally.
Stratifying the sample by gender and region, we divided the estimated national population by the number respondents in the sample to weight the sample for national synthetic estimation. The sample was taken from juniors and seniors in high school. It was thought that given the content of the survey, it would be better to address it to older students rather than the younger students and that they could reflect on their entire high school experience.
Data Collection and Handling
Surveys were mailed directly to students' home addresses. Surveys were returned to Alfred University in self-addressed, stamped business reply envelopes. Returns were due postmarked no later than May 15, 2000. One week after the deadline, on May 27, 2000, all surveys were boxed and mailed to National Computer Systems, Inc (NCS) for optical scanning. To ensure anonymity, no unique identifier was printed on the survey. Qualitative responses were compiled by hand. The final database including all respondents was transferred to a secure server at Alfred University.
Initially, frequency tables were analyzed for each respondent group of students and staff by the entire group and by each demographic variable. A cross-tab of gender and geographic region was used to develop sample weights. Each of the six groups of males and females by region were weighted by the sample frequency divided by the population to produce synthetic estimations of prevalence and to produce percentages expected for the population based on this currently limited set of knowledge.
Factor analyses were run to test our theoretical categories of community-building, humiliating, substance abuse, and dangerous initiations. These categories were confirmed by the factor analyses. A few behaviors were better understood by these analyses:
- "Tattoo, pierce, or shave" is more about isolation than about being hurtful to self.
- "Deprive oneself of food, sleep, or cleanliness" clusters with substance abuse behaviors, which are about being hurtful to self rather than about being humiliating, although for face validity we left this variable in the humiliating hazing category.
- "Drink or exercise until you pass out" clusters with substance abuse behaviors, which are about being hurtful to self rather than being destructive or disruptive.
To further investigate significant differences among groups, cross-tabular analyses using chi-square statistics were conducted. Demographic groups were crossed with activities to identify at-risk groups and hot spots, while the perceived hazing variables were crossed with variables to validate perceived versus actual hazing. Group variables were crossed with demographic variables to identify group differences. Gender was crossed with all variables to identify male and female differences.
Limitations of the Study
Because of the secrecy surrounding hazing activities, we guaranteed complete anonymity by conducting a direct-mail survey. High school students have a consistently low rate of response on mail surveys. The low response rate begs for further studies to confirm or refute, and further refine these findings. The sample size was large enough, however, to use inferential statistical analyses with high levels of confidence (95-99%).