Part One. The Shift: Scholarly Views of Family Structure Effects on Children, 1977-2002
Introduction
What are the effects of family structure on children? Over the past few decades,
that question has been a major source of controversy within family scholarship.
Two opposed camps have squared off in debate. One perspective—call it
“concerned” or pro-marriage—holds that the decline in marriage
has been a troubling trend, especially for children. Its adherents argue that
father absence and divorce tend to have important negative consequences for
child well-being. The opposing perspective—call it “sanguine”
or pro-family diversity—holds that families haven’t been weakened
by divorce and unwed childbearing but have just changed in form. Advocates of
the sanguine view argue that the supposed effects of family structure for children
are exaggerated, if they exist at all. There are of course views intermediate
to these two, including that negative family structure effects on children exist
but only because other institutions have not adapted to changes in the family.
In the 1970s, while the divorce rate was skyrocketing, the sanguine view seemed
ascendant among family scholars. By the mid-1980s, however, some observers noticed
an apparent shift underway. In 1987, the Journal of Family Issues published
18 essays by prominent scholars in a special edition on “The State of
the American Family.” The majority of commentaries expressed more concern
than optimism about recent increases in divorce and single-parenting.
In the issue’s introductory essay, then-editor Norval Glenn (1987) speculated
that the emphasis on concern “reflects a change in recent years in the
modal views of family social scientists, [though] one cannot be sure”
(p. 349). In any event, a number of influential scholars remained optimistic
about diverse family structures, and no clear consensus emerged. Research and
debate on family structure effects continued in social scientific journals through
the 1980s and into the 1990s.
Recently, the so-called “family wars” seem to have subsided. The
debates over divorce and unwed childbearing have not disappeared, but they are
less heated than in years past. Partly it is because same-sex marriage is the
new flash point that dominates discussion of marriage and family. Yet partly
it is because the debate about family structure has, in many ways, played itself
out. Since the dramatic shifts in children’s living arrangements began
four decades ago, scholars have had time to study the effects of divorce and
single parenting. Most family scholars apparently now agree that the preponderance
of the evidence indicates that children tend to do best when they grow up with
their own two married parents, so long as the marriage is not marred by violence
or serious conflict. The debate now centers on the exact nature and size of
family structure effects and on whether or not societal adjustment to new family
forms can substantially reduce their negative effects. Some pro-marriage scholars
and advocates have trumpeted this development as a “new consensus”
in the research literature about the positive effects of marriage for children.
However, enough disagreement remains that “consensus” may be too
strong a word, and evidence for the increase in agreement—the alleged
shift in views—is impressionistic and anecdotal. It is true that some
prominent family scholars have changed their minds about the consequences of
divorce and single-parenting (Glenn, 2001) and that statements of negative effects
of divorce and single parenting have been issued by respected nonpartisan research
organizations (e. g., Moore et al., 2002), but there has been no systematic
study of the purported shift.
The purpose of the study reported here is to provide better evidence of the
trend in family scholars’ views of family structure effects on children
by examining all relevant articles published in the Journal of Marriage
and Family over a recent 26-year period. We set out to estimate whether
anything approaching a “new consensus” has really emerged among
mainstream family scholars—the kinds of persons who publish in the premier
family social science journal—and the extent to which empirical research
findings have contributed to any trends found. Incidental purposes of the research
include assessing the trend in amount of attention given to family structure
effects and variations in views of these effects among different kinds of family
scholars. The ultimate purpose, of course, is to provide information to inform
policy discussions about family issues.
The main hypothesis that guided our research was that there has been a steady
increase in concerned views about family structure effects and a steady decrease
in sanguine views. We also expected that the increase in concern would lead
to a steady increase in attention paid to the topic.
Methods
To achieve our ends, we examined all articles (excluding book reviews) published
in the Journal of Marriage and Family (JMF)* from 1977 through
2002 to identify those that deal with family structure effects on children.
We selected the JMF because it was, and remains, the most prominent
and influential journal in family social science in the United States (and in
the world). Any substantial trends in the field, we reasoned, were likely to
be reflected in the pages of the JMF.
We define family structure in terms of number of adults in a family and their
relationship to any children present. The three main family structures we look
at are (1) a child or children living with married biological or adoptive parents,
(2) a child or children living with only one parent, and (3) stepfamilies. All
of our comparisons are between intact married families and one or both of the
other types (it is likely that unmarried cohabiting adults with children often
fall into the single-parent category in the studies we examined). We do not
include other family structures, such as children living with grandparents or
with gay and lesbian parents, due to a dearth of relevant articles in JMF.
Therefore, family structure issues other than divorce, out-of-wedlock childbearing,
and remarriage are beyond the scope of this study.
An article was included in our analysis if it contains a viewpoint or finding
on the relationship between family structure and child well-being or a global
assessment of the effects of trends in family composition. Family structure
does not have to be a major focus of the article for it to be included. An article
that treats child abuse, for example, was included if it discusses whether or
not family structure is a significant risk factor. We included all articles
reporting research in which at least one examined outcome is typically considered
to be a measure of well-being, such as academic performance, adolescent pregnancy,
drug use, delinquency, and emotional problems.
We judged that 266 articles,† or 12.2 percent of the 2,189 articles published
during the 26-year-period, met our criteria for inclusion. Since the inclusion
criteria are somewhat subjective, it is possible that if other persons did the
reviewing, a few articles currently included would be excluded, and vice versa.
Nevertheless, it is unlikely that a few studies at the margins would substantially
alter the results of our analysis.
In order to track scholarly conclusions over time, we assigned each article
that meets the inclusion criteria a “family structure effects rating,”
on a scale from 1 to 5, based on how much the authors seem to think family structure
has an effect on child well-being or on outcomes that persist into adulthood
(see Table 1). A rating of “1” means that the author appears to
believe that family structure, divorce, and unwed childbearing are generally
unimportant when it comes to child well-being. In other words, a “1”
represents a very sanguine view of family structure diversity. (For the purposes
of this article, we use the term “family diversity” exclusively
in terms of family structure, i.e., intact two-parent families, single-parent
families, and stepfamilies.) A rating of “5” means that the author
takes a strong stance that intact marriages substantially promote child well-being,
overshadowing many other important influences. An article with a “5”
rating, then, reports high concern about increases in divorce and unwed childbearing.
A rating of “2” indicates that the author is more sanguine than
concerned about the decline of marriage (although the quantitative articles
we rated “2” found at least some evidence of family structure effects
or conceded that it is likely that such effects exist) and a rating of “4”
means that the author is more concerned than sanguine. A rating of “3”
represents an intermediate stance between the extremes. The family structure
effects scale, then, represents the spectrum from sanguineness to concern. Although
authors of articles with high ratings did not always use explicit language of
concern, they reported conclusions about effects that almost anyone would consider
reason for concern.
Table 1. Number of Articles
by Family Structure Effects Rating and Article Type |
Article Type |
Number of Articles by Family Structure
Effects Rating |
Number of Articles by Type |
Mean Family Structure Effects Rating by Article
Type |
| |
1 |
2 |
3 |
4 |
5 |
|
|
| Quantitative |
1 |
28 |
40 |
131 |
1 |
201 |
3.51 |
| Qualitative |
0 |
2 |
1 |
1 |
0 |
4 |
2.75 |
| Theoretical |
9 |
7 |
3 |
7 |
5 |
31 |
2.74 |
| Synthetic |
1 |
8 |
11 |
12 |
0 |
32 |
3.06 |
| Total |
11 |
45 |
54* |
150* |
6 |
266* |
3.36 |
*Two articles each of which fit into two category types (i.e., quantitative/qualitative
and qualitative/theoretical). |
We divided the articles into four categories: quantitative, qualitative, theoretical,
and synthetic. Quantitative and qualitative articles report the results of empirical
research; synthetic articles summarize and integrate the results of research,
as in literature reviews and meta-analyses; while theoretical articles include
“think pieces” and opinion essays as well as discussions and formulations
of theory. We traced the trend in the ratings within these categories as well
as among all articles, and we compared ratings among the categories.
The coding, done by one person (Thomas Sylvester), is necessarily somewhat
subjective, though we took measures to minimize systematic bias. For instance,
the articles were not rated in the order in which they were published, and all
articles in each category were not rated at the same time—a procedure
that should have minimized bias due to our expectation of trends or of differences
among the categories. Some scholars will undoubtedly disagree with the coding
of certain articles, but we doubt that the ratings of a handful of articles
at the margins would substantially alter any observed trends or differences
among categories. In any event, skeptical readers can judge the accuracy of
our ratings by examining them at www.familyscholarslibrary.org.
An important point is that the ratings are based on the conclusions of the
authors of the examined articles, not on the conclusions we would have drawn
from the findings reported. For instance, the same findings could lead to a
rating of a “3” or a “4” depending on the interpretations
given by the authors. Thus, an article reporting evidence that we thought indicates
strong effects was given a fairly low rating if the author concluded that the
effects were modest.
Another important point is that the ratings do not imply an author’s
support for any particular public policy or political ideology. In this study,
“sanguine” is not code for “liberal,” and “concern”
is not code for “conservative.” (A large majority of family scholars
are politically liberal [Klein & Stern, 2004/2005; Brookings Institution,
2001].) Authors whose articles have identical ratings may disagree about the
policy implications of their findings.
In order to track trends without giving undue weight to year-to-year fluctuations,
we divided the 26 years covered by the study into four five-year periods and
one six-year period (1977-1982), assigning an extra year to the first period
because, as expected, there were fewer relevant articles per year in the earlier
years.
Results
Has There Been a Shift in Views?
Overall, we found strong evidence for increased concern about negative family
structure effects on children, but our main hypothesis, that there would be
a steady shift from sanguine to concerned views during the 26 years, was
not supported (see Figure 1). Rather, there was a substantial, and statistically
significant, change in the expected direction from 1977-1982 to 1983-1987 but
essentially no change after that. The mean rating for 1977-1982 is 2.81, very
near the center of the scale, whereas the later means range from 3.4 to 3.47.
The later means are well within the “concerned” range, although
we have greater confidence in the indicated change than in the exact meaning
of the ratings for any particular period. The articles reporting quantitative
research evinced a pattern of change parallel to that for all articles, with
a mean rating of 3.11 for 1977-1982 and mean ratings above 3.5 for each of the
five-year periods. Thus, concerned views clearly became more prevalent, though
there was not a steady change, and according to our judgment they outweighed
sanguine ones throughout the last 20 years covered by the study.
Figure 1. Number and Average Rating of Articles on
Family Structure Effects by Time Period |
 |
Our expectation that we would find an increase in attention to family structure
effects on children was strongly supported (see Figure 1). Both the number of
articles and the percentage of all articles in the JMF dealing with
the topic increased steeply during the 26 years covered, the proportion almost
quadrupling within the quantitative research category.
By the mid-1990s, 15 percent of all quantitative research articles in the JMF
dealt with family structure effects on children, and that proportion held steady
through 1998-2002. We believe this change very likely reflects increased concern
about possible negative effects, or at least increased openness to the possibility
of such effects. The accumulation of research findings indicating or suggesting
negative effects probably spurred an increase in research and writing on the
topic.
The strongest evidence of a change in views of family structure effects on
children comes from our examination of the 32 “synthetic” articles
on the topic published in the JMF from 1977 through 2002. For the purpose
of assessing the trend in the views reflected in these articles, we divided
the 26 years into three shorter periods so that each would include at least
10 articles (see Figure 2). In 1977-1987, the conclusions of these articles
were generally sanguine (mean rating of 2.4), whereas by 1999-2002, the mean
was 3.64—well within the concerned range—and the indicated change
is statistically significant. We view this trend as particularly important,
because it reflects the contents of family social science literature in general
and not just that in the JMF.
Figure 2. Average Rating of Synthetic Articles by Time
Period |
 |
Further evidence of the shift comes from the observations and judgments of
the authors of articles in the JMF. For instance, in 1988 Sara McLanahan
and Karen Booth wrote, “Whereas a decade ago the prevailing view was that
single motherhood had no harmful effects on children, recent research is less
optimistic with respect to the long-term outlook” (p. 557). A widely cited
1991 meta-analysis by Paul Amato and Bruce Keith also reflected a change in
scholarly thinking on the effects of parental divorce. After pointing out that
a host of negative outcomes are associated with parental divorce, the authors
stated, “The results lead to a pessimistic conclusion: the argument that
parental divorce presents few problems for children’s long-term development
[which, as the authors imply, had been prevalent] is simply inconsistent with
the literature on the topic” (p. 55).
Other authors expressed differing views in articles published in the JMF
early and late in the period covered by our study, a conspicuous example being
in papers by Marilyn Coleman and Lawrence Ganong, who are among the foremost
expects on stepfamilies. In 1986, they wrote, “[In a] comprehensive review
of empirical studies of stepchildren, we concluded that there were few differences
between stepchildren and children from intact nuclear families. If, in fact,
stepchildren are not ‘different’ in any major way from children
in other families, the negative perceptions [of stepfamilies] need to be corrected”
(p. 309). In 1990, however, these same authors reported research suggesting
a link between stepfamilies and low levels of child well-being, and in 2000,
they, with colleague Mark Fine, wrote, “Since at least the middle of the
decade, it has been safe to conclude that stepchildren are at somewhat greater
risk for educational difficulties and internalizing and for externalizing behavior
problems than children living with two parents…” (p. 1300). The
authors appropriately pointed out that most stepchildren do not exhibit the
problems mentioned, but their take on the effects of children’s living
in stepfamilies obviously had become less sanguine.
How Strong is the Evidence of Family Structure Effects on Children?
The study reported here was designed primarily to assess trends in views rather
than the accuracy of any particular position, but the findings do have relevance
for the debates about family structure effects. The general shift toward more
concerned views as more attention has been devoted to the topic in itself suggests
that the concern is warranted, and other findings from the study also support
that conclusion.
We think it is important, for instance, that on average the authors of the
articles reporting quantitative research evinced more concerned views than the
other authors, the difference in mean ratings between the quantitative and “theoretical”
pieces being statistically significant (see Table 1). (We place no importance
on the low average rating—2.75—of articles reporting qualitative
studies, because only four such articles were published.) A major reason for
this difference is probably, though not certainly, that the views of the authors
of the quantitative pieces were more constrained by “hard data”
than those of the other authors and thus were less affected by preconceptions
and ideological biases. All of the quantitative articles were presumably selected
for publication by the normal review process, while several of the “think
pieces” were invited and thus their authors were less constrained by having
to satisfy reviewers—a difference that may also account for some of the
disparity in mean ratings. It is interesting that only one percent of the quantitative
articles received the extreme ratings of “1” or “2”
whereas 44 percent of the “theoretical” articles did so—again,
we suspect, partly because of the greater constraints placed on the authors
of the former by the empirical evidence and the reviewing process. However,
it is likely that some of the authors were invited to write opinion pieces because
they were known to hold rather extreme views.
We did not try to assess the overall adequacy of the methods used for the quantitative
research reported in the JMF, but it is generally agreed that, everything
else being equal, studies with large representative national samples are better
than those with small and local convenience samples. It is relevant, therefore,
that the average rating for the 119 studies using nationally representative
samples is 3.66 compared with 3.28 for the 82 other studies. Furthermore, the
87 longitudinal studies have a mean rating of 3.68, compared with 3.37 for the
114 cross-sectional ones. The 60 longitudinal studies with nationally representative
samples have a mean rating of 3.7 while the 55 cross-sectional studies with
non-representative samples have a mean of 1.11. Only the last difference is
statistically significant, but the others approach significance. These differences
suggest, but by themselves cannot prove, that the better the research design,
the more likely the researchers were to conclude that there were important family
structure effects on children.
Finally, we think it is important to point out that of the 201 relevant quantitative
studies reported in the JMF over the 26-year period, only one, or a
half of one percent, failed to find any evidence whatsoever of family structure
effects and failed to concede that such effects are likely, although the authors
of another 28, or 14 percent, either failed to find evidence for the effects
but conceded that they were likely, concluded that the evidence for effects
was not strong, or concluded that the indicated effects were too weak to be
important. Among quantitative family researchers, then, disagreements about
family structure effects on children seem to be very largely about their magnitude
and importance and not about whether or not they exist.
Discussion and Conclusion
Although the evidence for the hypothesized shift among family scholars toward
concerned views about family structure effects on children is not exactly what
we expected, it does strongly indicate that a shift in the expected direction
did occur. We expected a steady increase in concerned views, but at least in
the pages of the JMF, most of the change occurred between the 1970s
and the late 1980s.
Whether or not there has been an emergence of a “new consensus”
among family scholars about family structure effects on children depends on
how one defines the consensus. Consensus about what? There does seem to be widespread—almost
universal—agreement that some negative family structure effects exist,
or have existed in the recent past. Beyond that, however, there is still considerable
disagreement. In a related paper, we (Glenn and Sylvester, 2006) show that many
authors of papers published in the JMF during 1977-2002 tended to downplay
evidence of negative family structure effects on children. Other scholars have
legitimate questions about the evidence for family structure effects. Pro-marriage
activists may disagree with the views of these scholars, but we think it would
be a mistake to deny or disregard them.
Furthermore, it is incorrect to claim that the evidence for negative family
structure effects on children is conclusive, because it is not. Anything approaching
conclusive evidence would have to come from randomized experiments, which cannot
be used to assess family structure effects. For instance, we cannot randomly
divide children into an experimental group, to be subjected to a parental divorce,
and a control group, from which the stimuli associated with a parental divorce
are withheld. The non-experimental and quasi-experimental methods we must use
instead are all fallible. As the family diversity advocates like to point out,
correlation does not equal causation. Although children who grow up in non-traditional
families clearly do not fare as well on the average as children raised with
both biological or adoptive parents, that fact alone does not prove family structure
effects. Something (or some things) that commonly affect family structure and
child outcomes could account for the correlation. Quasi-experimental research
designs statistically “hold constant” some of those things likely
to commonly affect family structure and child outcomes, but one can never be
sure that everything that needs to be “held constant” has been measured
and incorporated into the research design. In recent years, there have been
attempts to “hold constant” even unmeasured variables, but these
procedures remain controversial.
The bottom line is that we do not know for sure what the nature and magnitude
of family structure effects on children are and that we are unlikely to know
for sure in the foreseeable future. The evidence for important effects is very
convincing, especially since there are strong theoretical reasons for expecting
such effects, but convincing and conclusive are not the same.
Does that mean that no personal or policy decisions concerning possible family
structure effects should be made until “all of the evidence is in”?
In our opinion, of course not. Individuals make many decisions every day on
the basis of probabilistic knowledge, and to say that policy decisions should
await certain knowledge is to say that such decisions are not going to be made.
These decisions should be, must be, made on the basis of the “preponderance
of evidence” available at the time.
At the present time, the preponderance of the evidence indicates that family
structure matters, and matters to an important degree, for children. Accepting
that conclusion leaves many questions unanswered, such as, what, if anything,
should be done to try to change the distribution of family structures in the
society. Trying to answer those questions is beyond the scope of this paper.
Notes
*Prior to 2001, the Journal of Marriage and Family was known as the
Journal of Marriage and the Family. For ease, the new journal title
is used throughout this article and in the online bibliography at www.familyscholarslibrary.org.
†The bibliography of articles analyzed for this study, along with their
family structure effects ratings, are available online at www.familyscholarslibrary.org.
Works Cited
Amato, P.R., & Keith, B. (1991). Parental divorce and adult well-being:
A meta-analysis. Journal of Marriage and Family, 53 (1), 43-58.
Brookings Institution (2001). National survey on government endeavors.
Prepared by Princeton Survey Research Associates.
Coleman, M., & Ganong, L.H. (1990). Remarriage and stepfamily research
in the 1980s: Increased interest in an old family form. Journal of Marriage
and Family, 52(4), 925-940.
Coleman, M., Ganong, L., & Fine, M. (2000). Reinvestigating remarriage:
Another decade of progress. Journal of Marriage and Family, 62 (4),
1288-1307.
Cooney, T.M. (1994). Young adults’ relations with parents: The influence
of recent parental divorce. Journal of Marriage and Family, 56 (1),
45-56.
Ganong, L.H., & Coleman, M. (1986). A comparison of clinical and empirical
literature on children in stepfamilies. Journal of Marriage and Family,
48 (2), 309-318.
Glenn, N.D. (1987). Continuity versus change, sanguineness versus concern:
Views of the American family in the late 1980s. Journal of Family Issues,
8 (4), 348-354.
Glenn, N.D. (2001). Social science findings and the ‘family wars’.
Society, 38 (4), 13-19.
Glenn, N. D., & Sylvester, T. (2006). The Denial: Downplaying the Consequences
of Family Structure for Children. New York: Institute for American Values.
Available at http://www.familyscholarslibrary.org.
Klein, D.B., & Stern, C. (2004/2005). Political diversity in six disciplines.
Academic Questions, 18 (1), 40-52.
McLanahan, S., & Booth, K. (1989). Mother-only families: Problems, prospects,
and politics. Journal of Marriage and Family, 51 (3), 557-580.
Moore, K.A., et al. (2002). Marriage from a child’s perspective:
How does family structure affect children and what can we do about it?
Child Trends Research Brief. Washington, DC: Child Trends. Available at http://www.childtrends.org/PDF/MarriageRB602.pdf.
About this Report
This report was commissioned by the National Fatherhood Initiative, in partnership
with the Institute for American Values, and written by Professor Norval Glenn
of the University of Texas at Austin and Thomas K. Sylvester of the Institute
for American Values. The sponsors are grateful to their academic advisors for
their review and suggestions, and to their financial contributors for their
generous support.
This project was supported by Grant No. 2001-DD-BX-0079 awarded by the Bureau
of Justice Assistance, U.S. Department of Justice. The Bureau of Justice Assistance
is a component of the Office of Justice Programs, which also includes the Bureau
of Justice Statistics, National Institute of Justice, Office of Juvenile Justice
and Delinquency Prevention, and the Office for Victims of Crimes. Points of
view or opinions in this document are those of the author and do not represent
the views of the Department of Justice.
© 2006, Institute for American Values.
Institute for American Values
1841 Broadway, Suite 211
New York, NY 10023
Tel: (212) 246-3942
Fax: (212) 541-6665
Website: www.americanvalues.org
Email: info@americanvalues.org