This study is not sufficiently controlled for authors to reach the
conclusion that “television may have a substantial role in the high rates
of teenage pregnancy in the United States” and that “reducing the exposure
of US teens to sexual content might substantially reduce teen pregnancy
rates.” Authors are correct to worry that “there is the possibility that
we did not account for all the factors that may possibly explain the
relationship we uncovered.”
This study includes no control for the most important variable: the
pregnancy behaviors of adults of the same families, locales, and
population groups corresponding to the teenagers whose pregnancy rates are
being analyzed. Indeed, using tabulations by the Centers for Disease
Control, Alan Guttmacher Institute, and Census to create a matrix of rates
of birth, unwed birth, abortion, and fetal loss rates per 1,000 female
teens age 10-19 and 1,000 female adults ages 20-44 for each race/ethnicity
(white non-Hispanic, black non-Hispanic, and Hispanic) by state, the
correlations between adult and teenage outcomes are strikingly high. For
example, a bivariate regression shows that the rate of pregnancies and the
proportion of births to unwed mothers among adults ages 20-44 is
associated with 87% of the variation in teenage pregnancy rates (R=0.933,
adjusted R2=0.869). A similar adult-teen correspondence exists for rates
of birth (R=0.929), unwed birth (R=0.771), abortion (R=0.958), and fetal
loss(R=0.773) (p<.0001 in all cases).
This paper also fails to control directly for poverty levels, another
covariate strongly associated with rates of teenage birth (R=0.725),
abortion (R=0.468), fetal loss (R=0.573) and overall pregnancy (R=0.731).
When these adult-behavior and socioeconomic variables are combined in
multivariate regression, they are found to predict 88% of teen pregnancy
rates (multiple R=0.940, adjusted R2=0.882) (p<.0001, all cases). These
staggeringly high correlations suggest that what we call “teenage
pregnancy” is not a distinct “teenage” behavior at all, but simply a
function of adult sexual norms, the fact that most pregnant teenagers have
adult partners and are part of the adult sexual world, and the tendency of
poorer populations to reproduce earlier in life.
With regard to these factors, the authors’ regression values for the
most relevant variables show that black race and absence of two parents
are significantly more predictive of higher teen pregnancy rates than is
television viewing. (This study’s more accurate conclusion should be:
“Socioeconomic conditions and family structure are much more important
factors in teenage pregnancy than are television habits”). Even so,
controlling for race is not a sufficient substitute for controlling for
poverty, since teenage pregnancy rates are correlated with poverty within
each race. Nor does controlling for two-parent presence adequately control
for the range of adult reproductive behaviors.
When researchers fail to control for most important covariates of
teenage pregnancy, less important, even trivial, covariates of these
larger variables may appear to be more important than they are. Television
viewing may well covary with poverty and with adult pregnancy, birth,
unwed birth, and abortion levels (just as do dozens of other behaviors).
That poorer, single parents with numerous children and frequent
pregnancies may let their children watch somewhat more television (or,
say, eat more fast food) than more affluent, two-parent households with
few kids does not mean television viewing (or fast-food consumption)
itself is a significant factor explaining why poorer teens get pregnant at
rates many times higher than do richer ones. Without inclusion and precise
specification of these larger variables, it is not valid to argue that a
lesser covariate is an important causal factor, or even a true correlate,
of the teen pregnancy outcome being analyzed.
Finally, self-reporting surveys are relatively weak tools afflicted
by many confounds. That survey findings often do not reconcile well with
outcome statistics (which themselves are imperfect) is a serious problem
that needs more attention. For example, this and other papers in this
issue argue that increasingly explicit and violent television, video game,
and internet content are strongly related to higher rates of pregnancy and
violence among adolescents. Yet, in reality, the rise in objectionable
media content has accompanied plummeting rates of pregnancy and violence
(especially serious violence) among teens over the last 10 to 15 years.
Claims that high rates of US teen pregnancy are related to television
programming do not explain why Western European and Japanese teenagers
whose cultures include explicit media sexuality have low pregnancy rates.
Behavior surveys should be subjected to much more rigorous specification
of key variables and validation by outcome statistics before they can be
used to recommend policy changes.
Mike Males, Ph.D.,
Center on Juvenile and Criminal Justice,
http://www.YouthFacts.org
Conflict of Interest:
None declared
eLetters: ![[Read eLetters]](/icons/misc/sectionDOWN.gif){kind=icon}
