Making It Go Away

Let’s call it Plan B. (Elsewhere I discuss Plan A and Plan C.) It has more than a few adherents. The basic strategy is first to concede the correlation between family structure and child outcomes, then perform various academic acts intended to make that correlation seem insignificant and even misleading, like a mirage in the desert. Although these re-visionings typically involve mundane mathematical procedures, academics often describe them in highly evocative terms. For example, when the "it" is how family breakdown affects children, a chosen methodological move is frequently said to "make it go away" or "make it disappear." Like something inconvenient that can be gotten rid of. Or like a killer talking about his victim. (From the first Godfather movie: "Oh, Pauly. You won't see him no more.")

In the U.S. family debate, Dr. Robert W. Blum, a professor of pediatrics and the director of the Center for Adolescent Health at the University of Minnesota, has recently emerged as a prominent practitioner of Plan B. In an article published late last year in the American Journal of Public Health, and in a companion monograph presented at the National Press Club and widely reported in the national media, Dr. Blum and his colleagues analyzed data from the federally funded National Longitudinal Study of Adolescent Health, usually called the Add Health Survey. Their main goal was to evaluate the relative importance of various risk factors for U.S. adolescents linked to five harmful activities: involvement in weapon-related violence, early sexual intercourse, drinking, smoking, and thoughts of or attempts at suicide.

One of their main conclusions is that, in assessing risk factors, scholars and policy makers must look "beyond" family structure. Why? Blum freely concedes that living in a one-parent home measurably increases the likelihood that teenagers will experience these problems. But his statistical analyses also show that a broad demographic variable such as family structure is less accurate as a predictor of outcomes than any of several individual and social variables, such as having trouble with schoolwork, not getting along with parents, and hanging out with friends who drink, smoke, and are sexually active.

Here is how Blum explains this finding to the Washington Post: "Let's say I'm a physician counseling a teenager, and that teen tells me he lives in a single-parent home. What is that going to tell me about whether he smokes cigarettes or is having sex? Not much. But if that kid tells me he spends his Friday and Saturday nights with kids who drink, that's going to tell me a lot." Not only is family structure a comparatively "weak predictor" of adolescent behavior, it is also, in Blum's estimation, "not especially amenable to change." We need, Blum insists, to focus on those teen risk factors that we can actually do something about - which is why scholars and policy makers should "abandon" the issue of family structure as a useful way of understanding adolescent behavior, while simultaneously, for example, treating the issue of school failure as a major "public health problem."

Let us count the ways in which Blum's basic proposition is either trivial or silly. First, he trumpets the greater predictive power of "individual and social" variables without acknowledging that these variables themselves are partly the effects of family structure. For example, it's fine for Blum to report that lots of unsupervised "hanging out" is a risk factor for teens, but ridiculous to offer up this fact as evidence that we must move "beyond" family structure, since countless studies have found that children in one-parent homes are more likely to engage in unsupervised "hanging out." In formal terms, Blum is arguing that evidence of how a demographic variable (one-parent homes) can express itself (more teenage hanging out) constitutes a compelling reason to abandon scholarly focus on the variable itself. Very embarrassing. Blum is committing an elementary error in logic.

Here's another example of the same error. In the monograph's "Summary," just prior to demanding that we "move away from" concern with family structure, he reports this overall finding: "No protective factor cut across all health-risk behaviors. However, the one most consistently protective factor found was the presence of a positive parent-family relationship." Let's see if I've got this straight. If I am a teenager in trouble, my "family structure" - the fact that my father after the divorce moved to another state to marry another woman - is not an important factor. What's important instead is whether I have a "positive parent-family relationship." Apparently, not only do scholars interpret how I get along with my family as unconnected to my "family structure," but the vital importance of getting along with my family means that society should "move away from" any concern about family structure. Is something wrong with this picture?

Blum's second violation of basic logic is even more embarrassing than his first. He fails to distinguish a predictor of a behavior from the behavior itself. For example, Blum makes much of the fact that living in a one-parent home is less predictive of teenage drinking than "the number of best friends who drink at least monthly." Similarly, living in a one-parent home is less predictive of teenage sexual activity than "whether they have an on-going relationship." See how this works? Blum is increasing the accuracy of the predictor by making sure that the predictor is integrally connected to what is being predicted. He is therefore clearly perpetrating a shell game. Drinking stems from…spending lots of time with friends who are drinking. Sex comes from...having a steady girlfriend. (A Blum-like insight from the first Rocky movie: "You hang out with yo-yo people, you get yo-yo friends. It's simple mathematics.")

Consider an analogy. I am a scholar who studies car driving, and here is the press release on my latest research: "Previous scholars have suggested that being rich is linked to the likelihood of driving a Mercedes-Benz automobile. But new statistical analyses reveal that being rich is a comparatively weak predictor of this behavior. Instead, a number of individual and social variables - such as owning a booklet called Your Mercedes-Benz, test-driving Mercedes-Benz cars, and regularly "hanging out" where Mercedes-Benz products are sold - are significantly more predictive of whether a person drives a Mercedes-Benz. What do these findings mean? Clearly, we should move away from a focus on comparative affluence in understanding these behaviors, especially since the distribution of wealth is not especially amenable to change."

Which brings us to the third and final reason why Blum's study is ultimately unserious. Everything - the contradictory categories, the near-tautological assertions, the heavy-handed rhetoric about going "beyond" family structure - ultimately stems from Blum's one big idea, which is that family structure is not "amenable to change," while nearly everything else is. Name any reform (except one), and Blum is ready and willing. We can fix the schools. We can raise children's self-esteem. We can reduce racial prejudice. We can foster healthier peer groups. We can even improve individual parenting skills. Indeed, we must, as a matter of "public health," do all of these things for our young people. But reduce divorce and unwed childbearing? Increase the proportion of children growing up in two-parent homes? Sorry, that issue is untouchable. Can't do a thing about it.

I am aware of no evidence - certainly Blum does not cite any - that would support this thesis. The proposition that positive change is possible in virtually all areas of social life, except the area of family structure, is little more than a naked ideological assertion - one that tells us very little about society, but a great deal about the philosophical orientation of the person making the assertion.

Recent data from the Census Bureau are relevant to this issue. Divorce rates in the U.S. have been declining modestly for some time. Rates of unwed childbearing, after three decades of steady increases, remained essentially unchanged during the last half of the 1990s. The proportion of all U.S. children living in a two-parent home stopped declining in 1995, and may, as of 2000, be modestly increasing. Importantly, the proportion of African American children living in two-parent homes has increased significantly since 1995.

These data tell us that positive change in the area of family structure is more than just possible; it may already be occurring. Insofar as Blum's guiding premise about the impossibility of strengthening family structure is not only unsupported, but also on the verge of being empirically disproved, his entire exercise becomes not just flawed, but pointless.

Sources: Robert W. Blum, et al., "The Effects of Race/Ethnicity, Income, and Family Structure on Adolescent Risk Behaviors," American Journal of Public Health 90, no. 12 (December 2000): 1879-1884. Blum, et al., Protecting Teens: Beyond Race, Income, and Family Structure (Minneapolis: Center for Adolescent Health, University of Minnesota, 2000). Laura Stepp Sessions, "New Study Questions Teen Risk Factors," Washington Post, November 30, 2000. "Household and Family Characteristics: March 1998 (Update)," U.S. Bureau of the Census, Current Population Reports, P20-515 (Internet release date: December 11, 1998). Unpublished Census data from 1999 and 2000 obtained from the Bureau's Fertility and Family Branch.

First published Spring/Summer 2001.