Tale of Two Studies: What Causes Teen Pregnancy, Violence, and Drug Use?

The headlines proclaimed the controversial news: race, poverty, and single-parents were NOT the irrevocable harbingers of drug abuse, teen pregnancy, and violence. Instead, researchers were claiming that behaviors that parents and teens could influence — such as problems at school and the amount of time spent “hanging out with friends” and the type of friends they chose — could predict trouble.

Some cynics speculated that this was a right-wing conspiracy to justify abolishing federal programs for the poor. Some assumed that the media, always keen for a “dog bites man” story, had oversimplified research results to get a newsworthy sound bite. On the other hand, some youth advocates welcomed the idea that impoverished minority youth with single parents didn’t necessarily have three strikes against them.

This study has important implications for all parents and adults who work with youth. To get the whole picture, I’ll describe two versions of the study: one that was published in a public health journal, and another that was aimed at the media and the public.

“Study #1: The Journal Article”

“The Effects of Race/Ethnicity, Income, and Family Structure on Adolescent Risk Behaviors” is by Robert Blum, MD and five colleagues, all from the University of Minnesota Adolescent Health Program. It was published in the American Journal of Public Health in December 2000.

The article examines the extent to which race and ethnicity (white, black, or Hispanics), income (6 categories, ranging from $10,000 or less to $61,000 or more), and single-parent homes (i.e., family structure) predict several specific problem behaviors among middle school students and among high school students:

  • Smoking
  • Alcohol consumption
  • Suicidal thoughts or attempts
  • Weapon-related violence
  • Ever had sexual intercourse (yes/no)

These were measured as a continuous range of behaviors, except for sexual intercourse.

The results show that race, income, and family structure each influence many of these problem behaviors. It was a well-designed analysis (using multivariate regression analyses) that can look at each of these influences separately. For example:

Smoking: teens that live in a single-parent home are more likely to smoke, regardless of age (middle school or high school), income, or race/ethnicity. Whites are more likely to smoke than Blacks or Hispanics, regardless of income, family structure, or age, although the differences were greater among the older teens. Teens from higher income homes were somewhat less likely to smoke, but among the high school students, neither income nor gender was related to smoking.

Alcohol: Students living in single-parent homes were more likely to drink, regardless of age, income, or race/ethnicity. Among high school students, those from more affluent families tended to drink more, regardless of family structure or race/ethnicity.

Weapon-related violence: Teens from more affluent homes were less involved with weapon-related violence regardless of family structure or race/ethnicity. Teens who lived with one parent and who were Black or Hispanic were more likely to be involved with weapons regardless of income. Girls were less likely to be involved in violence than boys, in middle school or high school.

Suicide Risk: More affluent students, Black students, and those in two-parent homes were slightly less likely to consider or attempt suicide. Girls were more likely to have suicidal thoughts or attempts, in middle school and in high school.

Sex: Teens from lower income families were more likely to report having sexual intercourse, regardless of family structure or race/ethnicity. Black teens from single family homes were more likely to report having sexual intercourse, regardless of income. These patterns were stronger in middle school. Middle school girls were less likely to report having intercourse than middle school boys, but there were no gender differences among high school students.

Getting an article published in a research journal depends on having statistically significant differences, regardless of how large those differences are or what the implications are for real life. However, Dr. Blum and his colleagues asked a question that is rarely asked in social science research: how important are these differences?

The answer was shocking: not much. The conventional wisdom is that minority race, poverty, and single-parent families make it almost impossible for kids to succeed, but this study found that their influence was weak, accounting for less than 10 percent of the variance.

What does “variance” mean? Here’s an example: these students vary a lot in their smoking habits, ranging from never smoked to smoking several packs of cigarettes every day. That variety of responses — how much they vary –creates the “variance.” If you knew a student’s race, income, and whether or not they lived in a single parent home, you could only account for 4 percent of that variation among middle school kids.

That means that 96 percent of that variation is caused by other things – perhaps how religious they are, or whether their parents smoke. (This article doesn’t say what those other things are.) Knowing the student’s race, income, and whether they live with one or two parents does not enable you to predict how much they smoke. Predicting alcohol use or suicide is even harder: race, income, and family structure account for 2 percent of the variance or less. Race, income, and family structure account for only 3 percent of the variance for weapon-related violence in either age group, and for sexual intercourse among high school students. The best prediction can be made for sexual intercourse for middle school students, but even that only accounts for 10 percent of the variance. So, if you know that Tanya is a white middle school student with a family income of $25,000/year and lives with her mother, you have a slightly better chance of correctly predicting whether she has had sex than predicting how much she drinks or smokes — but you’ll probably be wrong in predicting any of those behaviors.

Bottom line: If you focus on the group differences, you would conclude that Black and Hispanic teenagers from poor, single-family homes are likely to get in trouble. If you focus on the low predictive power of the statistics, you will realize that a white teen from a middle-class two-parent family is almost as likely to get into trouble (or stay out of trouble) as a Black or Hispanic teen from a poor single-parent home.

Is the research credible? It’s based on the National Longitudinal Study of Adolescent Health, which includes very detailed information about more than 12,000 students attending 134 high schools and middle schools. The data represent 7th to 12th graders in schools across the country. It is a relatively new, amazingly well-designed dataset, and Dr. Blum and his colleagues are just of few of the dozens of well-respected researchers who are analyzing pieces of information from this enormous study.

“Study #2: The Monograph

The second study, using the same dataset but with different analyses, was published in December as a monograph entitled “Protecting Teens: Beyond Race, Income and Family Structure” by Blum, Beuhring, and Rinehart. It has photographs, glossy paper, an attractive cover, and many more details.

This report can be read and understood by most adults and many youth. Before it gets into the complicated analyses, it describes the teenagers in the study, and how many reported problem behaviors. For example, most of the teens reported never having smoked an entire cigarette, but one quarter said they had smoked in the past 30 days. It also reports some of the findings of the journal article about race, income, and family structure.

Most important, it analyses data that were not included in the journal article: the impact of what the researchers call “individual factors, friends, and family,” such as:

  • Frequent problems with school work; ever repeated a grade
  • Frequently “just hangs out with friends”
  • Frequent religious activities; religious beliefs
  • A job during the school year
  • Physical maturity
  • History of rape or sexual abuse
  • Bad temper
  • Self-esteem
  • Wants and expects to attend college
  • Number of best friends who smoke; drink
  • Friend attempted or committed suicide
  • Prejudice among students at school
  • Smoker in the home, frequency of parent drinking; parent suicide attempt
  • Teen sets curfew
  • Positive relationship with parent/family
  • Parent is frequently present at dinner
  • Number of siblings, extended family in the home
  • Access to guns in the home
  • Ever dated; ever kissed; romantic relationships in past 18 months; perceptions about sex

Not surprisingly, some of these factors predicted some problem behaviors, and others predicted other problem behaviors — parents who smoke tend to have teens who smoke, parents who have a drinking problem tend to have kids with a drinking problem, and so on. What is surprising is that the power of these predictions were much, much greater than race, income, or whether a teen lived with one or two parents. Instead of 3 percent or even ten percent of the variance (referred to in this report as “individual differences”), a few personal, friends, and family variables accounted for 24 to 49 percent of the variance for most of the problems studied. (The exceptions were suicide and girls’ violent behaviors, which were harder to predict). The traits that tended to be most strongly associated with substance abuse, violence, sex, and suicide were how much time kids spent “hanging out” without supervision, whether they were having problems in school, and the kinds of kids they spent time with.

Dr. Blum cautions us that “I’m not saying poverty or ethnicity is irrelevant. They are tremendously important. I’m just saying you shouldn’t base predictions of problems on them, because there is so much variation within these families. We need to get beyond “racial profiling” to understand what is really going on.”

Blum also cautioned that a study like this can’t prove what causes what — only what correlates with specific problems. “But, the strong message to parents is that you need to be in your kid’s life: know their friends, what their friends do, and who their friends’ parents are.” And the message for all adults are to “set clear expectations regarding school performance. Skipping school or doing poorly is not just an educational threat, it’s a health threat. We need to provide resources to help schools capture the interest of kids who are disenfranchised.”

Blum also stresses that the study supports “the importance of adolescents having after-school and evening programs and places that provide structured time that is supervised by adults.” He concluded: “Our society will pay at the front end, or pay at the back end — and the price at the back end is much higher.”

This report should encourage parents of disadvantaged minority students, and the teachers and other adults who work with them, because it indicates some important influences that can be modified to help these children. On the other hand, it provides a real wake-up call for white, suburban families who think their children are immune from serious problems. And, it helps explain why substance abuse and tragic violence can occur even in seemingly safe, middle class communities.