| |
Abortion
and Mental Health: Studies based on the National Longitudinal
Study of Youth (NLSY)
Editor's Note: The NLSY is one of the surveys of the National
Longitudinal Surveys of Labor Market Experience (NLS) conducted
by the Center for Human Resources Research (CHRR) at Ohio State
University in collaboration with the U. S. Bureau of the Census.
It conducts annual interviews with stratified, multi-stage random
samples, including a national probability sample of noninstitutionalized
civilian men and women age 14-21 years as of 1979, with oversampling
of Blacks, Hispanics, and poor Whites. A more complete description
of the method can be found in the NLS Handbook (CHRR, 1988). This
section reports on three studies that have been based on NLSY
data.
Citation: Russo, N. &
Zierk, K. [Department of Psychology, Arizona State University,
Box 871104, Tempe, AZ 87287-1104], (1992). Abortion, Childbearing,
and Women's Well-Being. Professional Psychology, Research and
Practice, 23, 269-280.
Introduction: These longitudinal
analyses reflect a stress and coping perspective in which reproductive
events such as unwanted pregnancy and abortion are considered
to present both challenges and difficulties. Whether or not such
stressful life events have negative psychological outcomes depends
on a woman's resources - psychological, social, and economic -
for coping with such events. The primary goal of the study was
to determine the extent to which the relationship of abortion
to well being can be explained by level of well being that existed
before having an abortion or by childbearing and coping resources
(i.e., education, income, employment, & marriage).
Method: This study is based
on a secondary analysis of NLSY interview data from 5,295 women
who were interviewed annually from 1979 to 1987. Among this group
773 women were identified in 1987 as having at least one abortion,
with 233 of them reporting repeat abortions. Well-being was assessed
in 1980 and 1987 by the Rosenberg Self-Esteem Scale. The researchers
used analysis of variance (ANOVA) and multiple regression to examine
the combined and separate contributions of preabortion self-esteem,
contextual variables (education, employment, income, and marital
status), childbearing (being a parent, numbers of wanted and unwanted
children) and abortion (having one abortion, having repeat abortions,
number of abortions, time since last abortion) to women's post
abortion self-esteem.
Results: In 1987, women
who had one abortion had higher self-esteem than women who reported
no abortion or women who reported repeat abortions. Number of
abortions was positively correlated with number of children and
number of unwanted births. One out of three women having abortions
had more than 2 children underscoring the importance of not assuming
women who have abortions reject motherhood in general. Women who
had unwanted births had the lowest levels of well-being, and were
also more likely to have repeat abortions than other women. Multiple
regression analyses revealed that when previous levels of self-esteem,
contextual variables, childbearing, and abortion variables (one
abortion, number of repeat abortions) were all included in the
regression equation, the most important predictor of well-being
in 1987 was well-being in 1980. In addition, employment, higher
income, more years of education, and having fewer children all
independently contributed to increased well being in 1987. Being
married did not make a difference in women's well being, possibly
because happy marriages could not be identified. Neither having
one abortion nor having repeat abortions had an independent relationship
to well being when the other variables were controlled, suggesting
that the relationship of abortion to well-being reflects abortion's
role in controlling fertility and its relationship to coping resources.
Evaluation: This study
is an important contribution to the literature because it is longitudinal,
based on a large national sample, and examines effects of having
one abortion separately from having repeat abortions. It is of
sufficient size and variability in the critical variables that
if the claim that the experience of abortion has a widespread
and substantial impact on women's well being were true, it should
have been reflected in the findings. However the data collection
occurred between 1979 and 1987 before the systematic attempt in
the United States to stigmatize and shame women having abortions
got underway. The negative mental health effects that would be
expected from these efforts would not be reflected in these results.
The study would be stronger if
there were multiple measures of mental health variables (e.g.,
depression, anxiety). However, the measure of well-being used,
the Rosenberg Self-esteem Scale, has been shown to be a highly
reliable and valid predictor of a variety of mood and anxiety
disorders and is arguably one of the best all-around measures
of global well-being. As the authors themselves point out (p.
278), in addition to the standard methodological limitations associated
with any survey research and analyses of secondary data, there
is the additional possibility that women who were most distressed
would conceal having an abortion because they perceive it as having
a stigma. Underreporting has been a problem with the NLSY (Jones
& Forrest, 1992), and indeed, the staff at CHRR have later
taken steps to overcome it. It is important to recognize that
the study cannot be used to estimate numbers of women having abortions
- in that case underreporting would be a concern and weighted
data would have to be used to construct a profile of the population.
As the authors themselves note,
however, underreporting problems do not pose major issues given
the purpose of the study, particularly given that such problems
cannot easily explain the pattern of results. The correlations
among the abortion, demographic, and childbearing variables parallel
the typical findings in the scientific literature, buttressing
confidence in them. If it were true that the most distressed women
in the sample did not report their abortions, a finding of no
difference between women having an abortion and other women might
indeed be explained by underreporting. Further, underreporting
on the part of the most distressed women would blur the difference
between women having one abortion and women having repeat abortions.
But the expected difference between women having one abortion
and repeat abortions was indeed found. Further, the main points
of the study are based on regression analyses that examined the
relative combined and independent contributions of abortion, contextual,
and childbearing variables to subsequent well being.
Thus, the key findings are not
challenged by the underreporting issue - indeed, the fact that
they were obtained despite underreporting places more confidence
in them. Consequently, in developing a list of significant and
independent predictors of women's well being, well being before
ever having an abortion is first on the list followed by contextual
and childbearing variables. Whether or not a woman has had an
abortion, as an independent predictor, does not make the list.
As the authors point out, that is not to say that abortion is
unimportant or has no effect. Repeated unwanted pregnancy, whether
ending in unwanted birth or repeat abortion, was correlated with
lower education, lower income, and greater likelihood of being
in poverty. Insofar as having an abortion enables women to delay
childbearing, obtain an education, get a job, and have a good
income, even though it has no independent effect it is still contributing
to positive mental health.
It has also been suggested that
African-American and Catholic women would be both most distressed
and more likely to underreport having an abortion, and these groups
would be the primary source of underreporting bias in this study.
Consequently, a second study based on NLSY data but focusing on
race and religion was conducted.
Jones, E. F., & Forrest, J.
(1992). Underreporting of abortion in surveys of U. S. women:
1976 to 1988. Demography, 29¸ 113-126.
Citation: Russo, N. [Department
of Psychology, Arizona State University, Box 871104, Tempe, AZ
87287-1104], Dabul, A. (1997). The Relationship of Abortion to
Well-being: Do Race and Religion Make a Difference? Professional
Psychology, Research and Practice, 28, 23-31
Introduction: In response
to criticisms of Russo & Zierk (1992, above), these analyses
were based on data from 1,189 Black and 3,147 White women who
participated in the National Longitudinal Study of Youth.
Method: Analyses of variance
and regression analyses parallel to those conducted by Russo &
Zierk (1992) were used to examine whether the relationship of
abortion, contextual, and childbearing variables to well being
separately by race (Black, White), religion's support for abortion
rights (anti-, neutral- pro-) and whether or not the respondent
was Catholic (Catholic, NonCatholic).
Results: The results were
similar to the previous study in that education and income were
positively and independently related to well being for all women,
regardless of race or religion. Abortion was not independently
related to well being when preexisting well being and the other
variables were controlled, regardless of race or religion. This
study used regression to examine the factors predicting well being
among the women who had at least one abortion. Again, level of
pre-existing well being was the most important predictor of postabortion
well being.
Evaluation: This study
is an effective response to the assertion that the findings of
Russo & Zierk (1992) reflect underreporting on the part of
Black women. It also provides evidence that underreporting due
to having a religious background or being Catholic did not significantly
influence the results. However, in addition to the cautions identified
above, the authors point out that the Catholic Church's Operation
Rescue had not begun to receive national attention at the time
that religion and religious attendance was assessed. Thus, while
the findings apply to women who identified and attended church
when they were initially interviewed, 1987 church attendance was
not measured. Further, insufficient sample size and the way that
the NLSY asked about religion did not enable separation of conservative
fundamentalist Christian groups from other Protestants. The lack
of relationship of abortion to well being regardless of race,
even when individual religions were examined, suggests that the
result would remain unchanged even with a more precise classification,
however.
Citation: Reardon, D. &
Cougle, J. Depression and unintended pregnancy in the National
Longitudinal Survey of Youth: a Cohort Study. British Medical
Journal, 324, (12 Jan 2002), pp. 151-52.
Introduction: The authors
seek to examine the question of whether "prior psychological
state is equally predictive of subsequent depression among women
with unintended pregnancies regardless of whether they abort or
carry to term" (p. p. 151).
Method: The analyses are
described as based on "Women scoring in 'high risk' range
for clinical depression (CES-D score >15) who had their first
abortion or first unintended childbirth between 1980 and 1992.
Logistic regression stratified by marital status was used to compare
the percentage with scores above the CES-D cut-off for the 128
women "with unintended births" and the 293 "aborting
women." Because the findings are based on miscoded data they
will not be repeated here.
Evaluation: Flawed conceptualization combined with miscoding
of data make the findings of the study inaccurate and meaningless
except as a source of prime examples of how research can go wrong.
To enhance the usefulness of the article as a training exercise,
the problems are listed in detail. Other comments can be found
on the journal's website at http://bmj.com/cgi/eletters/324/7330/151#19026
First, a total of 795 women had
at least one abortion by 1992. If the analyses are based on 293
"aborting women," what happened to rest of them? It
appears that the researchers did not identify women who had an
abortion on their first pregnancy - they identified women who
had an abortion on their first pregnancy since their last NLSY
interview, which was most likely in 1990, but could have been
earlier depending on when the woman's last interview took place.
In addition, although the findings
are generalized to "first pregnancies" in order to "control"
for previous psychiatric state measured in 1979, 251 women who
had their first pregnancies and abortions before 1979 are excluded
from the analyses. These are the individuals who had first pregnancies
at a younger age and who arguably are the most distressed.
But suppose that the data were
not miscoded and the findings were not based on a selective sample
of first pregnancies - the study still has several additional
fatal methodological flaws:
- It purports to control for
previous psychiatric state by controlling for score on the Rotter
Internal-External Locus of Control Scale. This is neither a
measure of psychiatric state nor a measure of well being.
- The authors suggest that marital
status and first pregnancy outcome interact in their effects
on depression, with differences found for married but not unmarried
women. They argue that their results may be due to underreporting
of abortions on the part of unmarried women, suggesting that
their findings may reflect the stress of raising a child without
support. However their data analyses are based on marital status
in 1992, and not marital status at time of first pregnancy.
Further, as Russo & Zierk
(1992) pointed out:
- Women who "abort" and those
who don't cannot be classified into two groups solely based
on their first pregnancy outcome. Thus, many of the individuals
classified in the 'not aborting" group based on first pregnancy
(since last interview) have indeed had abortions, some multiple
times.
- Women who have one abortion
differ from women who have repeat abortions. Grouping these
women together on the one hand overestimates the relationship
of abortion to mental health outcomes for women who have one
abortion while on the other hand underestimates the relationship
of abortion to mental health outcomes for women who have a pattern
of repeated unwanted pregnancy.
- A CES-D cutoff score is used
to construct the depression variable. Although the scale is
widely used as a screening tool, it was designed to determine
the extent to which people differ along a continuum of psychological
distress and not to detect depressive disorder (Radloff, 1977).
As Santor and Coyne (1997) point out, a number of studies have
documented overdiagnosis of depression resulting from the use
of the CESD. They point out that in some cases, "as few as 11%
of individuals classified as depressed using standard cutoff
scores were actually diagnosed as depressed in a diagnostic
interview (Roberts et al, 1991)" (p. 233). It would have been
more appropriate also report the results of analyses that were
based on a continuous CES-D score.
- Claims for the implications
of the findings made in the article as well as in the authors'
press release are based on flawed logic and inappropriate generalization.
Even were their findings not based on miscoded data, the results
of this study could not be generalized to the mental health
outcomes of unintended pregnancy in the context of illegal abortion.
There is no random assignment to groups here. The fact that
abortion is a legal alternative to unintended pregnancy means
that women who are most severely distressed at the idea of having
a child can choose to have an abortion. Taking them out of the
childbearing population thus would lower the mean level of depression
found among women bearing unplanned or unwanted children
Making abortion illegal, however,
strengthens the link between unintended pregnancy and unwanted
childbearing. Under illegal conditions the profile of women who
give birth (and who place children for adoption) would be expected
to perform more poorly on mental health indicators. For this reason,
cross-national studies of pregnancy outcomes in contexts that
vary in access to abortion are needed if the goal of a study is
to generalize findings to those contexts.
Although not cited in the published
article, the authors' press release, which can be found on the
website of the Elliot Institute, suggests that the findings of
this study contradict those of Russo & Zierk (1992). A comparison
of the aims and methods of the two studies, however, reveals the
fallaciousness of this claim. Russo & Zierk (1992) did not
focus on first pregnancy - they examined whether having zero,
one or more abortions made a difference in predicting a women's
well being. The same measure was used to assess well being pre-
and post- abortion. They compared their findings for women having
abortions before and after 1980 to identify bias that might occur
as a result of restricting the data to women having abortions
after 1980.
In summary, the findings in this
article are fatally flawed, and claims that it refutes the conclusions
of Russo & Zierk (1992) are unfounded.
Radloff, L. (1977). The CES-D
Scale: A self report depression scale for research in the general
population. Applied Psychological Measurement, 1, 385-401.
Roberts, R. E., Lewinsohn, P.
M. & Seeley, J. R. (1991). Screening for adolescent depression:
A comparison of depression scales. Journal of the American
Academy of Child and Adolescent Psychiatry, 30, 58-66.
Santor, D. A. & Coyne, J.
C. (1997). Shortening the CES-D to improve its ability to detect
cases of depression. Psychological Assessment, 9, 233-243.
|