A number oJ surue ys haue by now provided information about the Gelatine risk of

persons from various back grounds to experience sexual abuse during childhood.
Interest ingl'f, they are Haiti y uni Joim in failing to find differences in rates according
Io social class or race. How cues, s‹nieral other factors haue emerged from communit y
studies as being consistently associated uiitfi higher risk/or a6use. (a) when a child
lives u›ithout one oJ the biological parents; (b) when the mother is unauailab le to the
child either as a result of employment outside the home or disabilit y and illness; (c)
u'hen the chtld reports that the parents’ marriage is un happy or conJliclual; (d) when
the chiId reports hauing a poor relationshi p ui ith the parents or being sub)ect to
extremely punitive discipline or child abuse; (e) ui hen the child reports having a
stepfather. The article drams some implications Jiom these 'findings and makes
recotnmetidations for uiays to improve subsequent studies of risk factors.

Hiafi Factors for

Child Sexual Abuse
DAVID FINXELHOR
Universit y oJ New Hampshice
LARRY BARON
Yale Universit y

In resf›onding to the problem of child sexual abuse, one of our most

pressing tasks is to identify any group of children who may be at high
risk. Not only will such identification allow us to focus our preven-
tion efforts where they are most needed, it will also give us valuable
new clues about the causes of sexual abuse.
The task of identifying high risk children is not a simple one,
however. Because much sexual abuse is hidden, so are the risk
factors. Any characteristic that appears to be common to victims
who come to public attention may not apply to the vast number of
victims who do

Authors’ Note: This research has been made possible by a grant from the
National Center on Child Abuse and Neglect (90CA 0936/01). We would also like
to thank Linda Gott and Ruth Miller for help in preparing the manuscript. This is
one of a series of papers on sexual abuse and family violence available from the
Family Violence Research Program at the University of New Hampshire.

JOURNAL OF INTERPERSONAL VIOLENCE, Vol. 1 No. 1, March 1986 43-71

0 1986 Sage Publications, Inc.

43

Downloaded from jiv.sagepub.com at GEORGE MASON UNIV on December 21, 2014

from the SAGE Social Science Collections. All Rights Reserved.

44 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

not. If most reported sexual abuse victims are from impoverished,

disorganized families, for example, is it because these children are
at higher risk or simply because these victims are more readily
detected? Recently, however, a number of studies have appeared
that have gathered data on relatively large samples of unreported
cases of sexual abuse. These cases are mostly adults looking back
on their childhood experiences, not children currently being
abused. But they do tell about risk factors among a sample of
victims, much more representa- tive than cases coming to the
attention of social agencies and police. In the following review of
findings concerning risk factors,' we will rely exclusively on those
studies that employed samples of the general population—either
whole communities or students. One crucial feature of these studies
is that they also contain built-in comparison groups. The whole
notion of a “risk factor” requires such a comparison between
characteristics of abuse victims and char- acteristics of nonvictims
from the same samples. We will from time to time make reference
to a few studies without comparison samples, particularly when
these studies are large in scope; but it will be
primarily for supporting evidence.

THE SURVEYS

The most notable findings of the large-scale surveys do not

concern who is at high risk but rather how large and widely
distributed the risk appears to be. At one time it may have been
thought that sexual abuse was confined to a small number of
children in certain unusual family and social circumstances that
might be readily identifi- able. However, the findings from the
surveys establish conclusively that this is not the case; sexual abuse
is prevalent in remarkable large quantity in individuals from
virtually all social and family circumstances.
Of the surveys, ten were actual communitywide surveys of adults.
One of the best was a sample of 933 adult women in San Francisco,
identified through household listing and screening, and interviewed
about lifetime history of sexual victimization (Russell, 1983). In that
study 289» of the women reported sexual abuse involving contact
before the age o1 14, 38% reported such abuse before age 18, and l6$o
reported such abuse at the hands of a family member. In another

Downloaded from jiv.sagepub.com at GEORGE MASON UNIV on December 21, 2014

 Finkelhor, Baron / CHILD SEXUAL ABUSE 45

well-designed community study, Wyatt (1984; Peters, 1984) recruited

a matched quota sample of 122 white and 126 Afro-American women
in Los Angeles through random digit phone dialing, and then inter-
viewed each in person. Child sexual victimization was reported by
429o of that sample. A national sample survey has also been done.
The Nos Angeles Times Poll (Lewis, 1985; Timnick 1985a, 1985b)
con- ducted phone interviews with 1252 males and 1374 females, and
found that 279» of the women and l69o of the men reported childhood
sexual victimization.
In addition to the community surveys, there have been a number of
studies that used student samples. For example, Finkelhor (1979)
surveyed 796 college students with self-administered questionnaires
and found that 199» of the women and 99» of the men had
experienced abuse by age 16. Fromuth (1983) replicated this on a
sample of 482 women students, finding a comparable rate of 229».
Other student surveys include Fritz, Stoll and Wagner (1981), Landis
(1956), and Seidner and Calhoun (1984).
There are striking variations in the rates uncovered in these sur-
veys, variations that are due to differences in method and definition of
abuse, and perhaps actual differences in true prevalence in different
groups (for analysis of these differences, see Peters, Wyatt, & Finkel-
hor, 1986). However, it should be noted that the higher rates tend to
come from the more meticulous studies with the more carefully
designed questionnaires, the better trained interviewers, and more
intensive efforts to achieve candor. This suggests that underreporting
is a serious problem, especially in the less meticulous studies.
Nonetheless, all the studies leave the impression that sexual abuse
is a relatively common event for children. In fact, the high rates across
all social classes, ethnic backgrounds, and family types is more nota-
ble than the differences within these subgroups. However, most of the
studies also provide some opportunity to look at relative risk accord-
ing to these various factors.

Sex
Virtually all studies that have included men as well as women have
found higher abuse rates for women. The differences range from more
than 4 women for each man in a Minnesota phone survey (Murphy,
1985) to 1.5 women for each man in one of the more obsc:ure student
48 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

Age of Child at Onset

Almost all studies report figures showing that children are more
vulnerable to sexual abuse starting in the preadolescent period
between ages 8 and 12. The mean or median age of onset for several
of the survey studies is as follows: Finkelhor (1984) girls 10.2; boys l
1.2; Russell (1983) girls 11.2; Wyatt (1985b) girls 11.2.
However, mean or median figures are not the best for conveying
the true distribution of risk because they do not show the relative
vulnerability at different ages. Moreover, they are affected by the
choice of an upper limit (Russell, 1983, with an upper limit of 18
will show an inflated mean or median compared to Finkelhor’s
1979 study, which used age 16). Thus we have constructed Table l
to illustrate risk on a year-by-year basis. In Table 1, studies are
listed so as to show the risk per 100 for each year of age. When
studies reported the exact number of cases that occurred for each
year, these figures were simply divided by the total N and
multiplied by 100 to obtain the risk per 100 for that year. When
studies only had data grouped for a range of years, the number of
cases in each group was divided by the total N and by the number
of years in each group to obtain an estimate of the risk for that
year. Then a summated average was created for all six studies
(shown in the last row of Table 1) by simply averaging the rates for
all studies that reported cases for that age.
The pattern in Table 1 shows an increase in vulnerability at age
6 to 7 and another very dramatic increase at age 10. Ages 10
through 12 appear to be years of particularly acute risk, when
children are victim- ized at more than double the average rate.
However, looking at studies individually, there are certain dis-
crepancies. Some studies (Finkelhor, 1979, 1984; Fromuth, 1983;
Lewis, 1985) show particularly high vulnerability in the preadoles-
cent period (10 to 12) with a decline in the adolescent period.
Russell’s (1983) study, however, shows continuing or higher
vulnerability in the adolescent period. Some of this difference is due
to definitions. Russell included certain types of peer aggressions in her
tallies, whereas Finkelhor (1979, 1984) and Fromuth (1983) only
included experiences with older partners. Moreover, both Finkelhor
(1979) and Fromuth (1983) made criteria for abuse in adolescence
more restricted. This tells us that the distribution for age of onset is
affected by what types of experiences are being counted.
 TABLE 1
Rate of Female Childhood Sexual Abuse by Age of Onset

Age af Onse t
Do
wn Swdies 1 2 3 4 5 6 Z 8 9 10 11 12 I3 l4 I5 l6 l7 IB
lo
ad
ed Finkelhor (1979) (17 (271 (19)
fro
m (N — 530) .06 1.06 1.06 1.69 1.69 1.69 3.52 3.52 3.52 .89 .89 .89 .89
jiv.
sa
Finkelhor (1984)
ge (1 ) (6) (5) (9) (2) (4) (12) (9) (6) (4) (5) (2)
pu (N = 334)
b. .29 1.79 1.49 2.69 .59 1.19 3.59 2.69 1.79 1.19 1.49 .59
co
m Fromuth (198 3)
at
(1 ) (15) (2 7) 72 (24)
(N = 482)
G .06 .06 .06 1.03 1.03 1.03 *1.86 1.86 1.86 4.97 4.97 4.97 1.24 1.24 1.24 1.24
E
O Keckley (1983)
R (31 ) (9) (2)
G (N — 320)
E .43 .43 .43 .43 .43 1.93 1 93 1 9 1 3 1.93 .70 .70 .70 .70 .1B .15 .15 US
M Lewis (1985)
A 53 (34) (41 ) (26) (56) (30) (53) (39) (15) (8) (2 3)
S (N = 1374)
O .64 .64 .64 .64 .64 ,64 2.47 2.98 1.89 4.07 2.18 3.85 2.18 1.09 .58 .55 .55 .55
N
Russell (1983)
U (2) (3) (1 4) (15) (26) (22 T (23) (23) (42) (31 ) (7 6) (105) (42) (65) (81 ) (70)
NI (N — 930)
V .2 1 .32 1.50 1.61 2.79 2.36 2.47 2.47 4.51 3.33 8.17 11.29 4.5 1 6.98 8,70 7.52
on
De Summated
ce
average
m
be
.37 .33 .36 .82 1.09 1.49 2.16 1.92 1.83 3.76 2.89 3.83 2.91 1.68 1.88 2.02 2.74 .35
50 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

All of the studies do agree about lower rates of vulnerability for

children under 6 and 7. Here again, however, methodological
consid- erations urge some caution. Clinicians have pointed out
how easy it is for the memories of early sexual abuse experiences
to be forgotten and repressed, especially when they occur to
children without a cognitive framework for interpreting their
experience. It is possible that the lower rates for younger children
may simply reflect a greater loss of memory of these experiences.
In discussing the age distributions for victims of sexual abuse, it is
also important to include some comment about figures that come
from studies of reported cases. In general the figures from such studies
show much higher mean or median ages for abused children than do
the studies using survey methodology. For example, in the National
Incidence Study of Child Abuse and Neglect, 609» of the sexually
abused children were 12 or older. It must be remembered, though, that
most studies using reported cases record the age of the child at the
time of the refiorting, which is in a great many cases some years after
the onset of the abuse. Moreover, it is probable that older children are
more likely to reveal abuse, thus biasing upward the ages of children
in studies of reported cases. Thus the discrepancies between the
clinical and survey studies are less contradictory than they might
seem.

Social Class
Child abuse in general has been thought to be more prevalent in
lower social classes. But child welfare advocates have generally found
this a politically unappealing stereotype and have tried to emphasize
the fact that child abuse occurs in all social strata. For some writers,
this emphasis has turned into a claim that child abuse is unrelated to
social class. Empirical studies have repeatedly undercut these claims,
however, finding strong associations between class and physical
abuse (Pelton, 1981; Straus, Gelles, & Steinmetz, 1980). Although
abuse is certainly not limited to the lower class as the stereotype might
suggest, to most researchers it makes sense that the frustrations of
poverty, joblessness, lack of education, and inadequate housing
would contribute to the conditions that increase violence toward
children.
Child sexual abuse, however, may be an entirely different story.
The most representative surveys of child sexual abuse in the com-

Downloaded from jiv.sagepub.com at GEORGE MASON UNIV on December 21, 2014

52 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

(1956) also reported higher rates among students whose fathers

were employed in public service, skilled, semiskilled, and unskilled
jobs compared to those with business-class fathers (although no
mention is given of the magnitude of the difference). Fromuth
(1983), however, found no social class association in another
student study. It is likely that the social class structure in student
populations is complicated by the question of who gets selected
into college. Students are not the best samples to test for social
class relationships.
A positive relationship exists between sexual abuse and social class
among reported cases, too, but here again the data are not well suited
for determining the class linkages. It is generally acknowledged that
the child welfare system is heavily biased toward identifying abuse in
lower social strata. The data, of course, show this bias. The National
Incidence Study of Child Abuse and Neglect (NCCAN, 1981) found
that 389» of all of sexual abuse cases known to professionals came
from families with incomes less than $7,000, whereas only 179c of
American families had incomes in this category. This is about the
same lower- class bias as for physical abuse (35$» of cases from
incomes less than
$7,000), but not as much as for child neglect (53$• of cases are
from the poorest bracket). Similarly, the American Humane
Association (Trainor, 1984) reports that although only 11.9$o of
U.S. families were receiving public assistance in 1980, 29.39• of
officially reported sexual abuse cases in 1982 came from such
families. Interestingly, they note that there has been a marked
decline since 1976 (from 39.8$•) in the percentage of sexual abuse
cases that involve welfare recipients. Un- fortunately, these data
probably tell us more about the system for identifying child abuse
than they do about the true social class distri- bution of sexual
abuse.
The currently available data on sexual abuse suggest that sexual
abuse and social class are unrelated. The most representative com-
munity surveys suggest little connection (Miller, 1976; Peters,
1984; Russell, 1983). The strong overrepresentation of the poor
among reported cases is readily accounted for as a reporting bias in
that system.
An argument could be made that the surveys are masking a social
class connection. It is possible that women from higher social classes
are more candid with middle-class interviewers and report more of
their abuse, thus creating an artificial parity with less candid lower-
class women. But there is no evidence for this specific effect, and not
 Finkelhor, Baron / CHILD SEXUAL ABUSE 58

much inferential support for it from the methodological studies of

survey research.

Ethnicity
Across the board, studies have consistently failed to find any black-
white differences in rates of sexual abuse. Even among reported cases,
where it is thought that blacks suffer from a labeling bias (O’Toole,
Turbott, & Nalepka, 1983), the percentage o1 black cases is no more
than the percentage of blacks in the population as a whole (NCCAN,
1981). In four community surveys (Keckley Market Research, 1983;
Kercher, & McShane, 1984; Russell, 1983; Wyatt, 1984), rates among
whites and blacks have been very similar. In the only survey showing
any black preponderance (Miller, 1976)—l8Po for blacks and 139c for
whites—the difference disappeared when controlling for city size and
other demographic factors.
Among reported cases, sexual abuse has consistently been the
type of abuse for which blacks have the lowest representation. In
the American Humane Association national records (Trainor,
1984), for example, black families accounted for 15.39» of sexual
abuse cases, almost exactly their proportion in the American
population. In the National Incidence Study (NCCAN, 1981), the
percentage of blacks was ll%, somewhat below their representation
in the population. This is especially significant considering that
discrimination, pover- ty, and stereotyping usually influence
professionals to more readily label black families as abusive.
Wyatt’s (1984) study, based on a community survey, and the
related follow-up by Peters (1984) were specifically designed to
look at possi- ble black-white differences, and were remarkable in
the similarities they found. Wyatt (1984) showed no overall
difference in rates or types of abuse, and Peters (1984) found few
differences either in risk factors or effects. The studies did show
that because of different family patterns in the black community,
certain factors associated with vulnerability in the white
community—such as separation from father, changes in caretakers,
and mother’s labor force participation— did not operate in the
black community (Peters, 1984). However, overall the studies were
remarkable in the similarities that they showed between sexual
abuse in the black and white communities.
Although studies have found no black-white differences, there
are hints of possible differences for other ethnic group. Kercher
and
54 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

McShane (1984), for example, in a mail survey of Texas driver’s

license holders, found victimization rates of 21.7 • for Hispanic
women compared to 9.89o for whites and 10.49o for blacks.
Russell’s (1983) findings concerning Hispanic women give some
possible sup- port to the Texas study. Although not statistically
significant, Russell’s overall rates for incestuous victimization for
Hispanic women are somewhat higher than for the sample as a
whole (20Po versus l6Po for the whole sample), and in the case of
father-daughter incest, Hispanic women have a rate of 7.5%
compared to 4.She. Russell’s data also show low rates (this time
statistically significant) for Asian and Jewish women, only 89o and
109» of whom were inces- tuously victimized. Ethnic differences
need to be the subject of further analysis.

Social Isolation
Social isolation is a risk factor of interest to sexual abuse
researchers for two reasons. First, there are anecdotal reports of
concentrations of sexually abusive families in rural areas (Summit &
Kryso, 1978). Second, social isolation has proven to be correlated in
some research with other forms of child abuse and neglect (Garbarino
& Stocking, 1980).
However, rural residence has been associated with more sexual
abuse in only one study, and in two it has actually been associated
with less. Finkelhor (1984) found a significantly high rate (449o)
among the college student women who had grown up on a farm.
Russell (1983) also found rates of incestuous abuse somewhat higher
among women from both rural (l7$o) and town (209») backgrounds,
compared to city-raised women (149c), but the relationship did not
reach statistical significance. On the other hand, Miller’s (1976)
results were completely opposite, with adolescents from farm com-
munities reporting the lowest rates (109c) compared to urban (l59o)
and suburban (l99o) areas. Wyatt (1985a) also found higher rates
among women raised in urban areas. So the idea that sexual abuse is
more common in rural areas seems doubtful.
However, another aspect of social isolation has been empirically
confirmed by three studies: sexual abuse victims appear to be isolated
among their peers. Finkelhor (1984) found that women with two or
fewer friends at age 12 had more experiences of sexual abuse.
Fromuth
 Finkelhor, Baron / CHILD SEXUAL ABUSE 65

(1983) confirmed this in her student sample with the exact same
question. Peters (1984) found higher rates associated with women
who reported either lack of closeness with peers or lack of
closeness with siblings. (The peers were a better predictor with
white women and the siblings with black women). In fact, lack of
closeness with siblings was one of three variables that best
predicted sexual abuse in Peters’s multivariate analysis of all risk
factors.
A plausible hypothesis is that social isolation is related to abuse. If
children have few friends, this may create a need for contact and
friendship that sexual abusers can capitalize on. Friendless children
may be easy marks. However, the friendlessness found in three studies
may, unfortunately, be the result of having been victimized rather
than a risk factor. Children who are being abused by family members
are often prohibited from having friends. Children who are feeling
shame and stigma as a result of having suffered victimization often
isolate themselves from others. At age 12, the year asked about in both
the Finkelhor and Fromuth studies, more than half of the respondents
had already suffered abuse. This casts doubt on whether friendlessness
was truly a risk factor. Social isolation may indeed be a risk factor for
sexual victimization, but we need to find other, more sensitive vari-
ables with which to measure it.

Parental Absence and Unavailability

In general, the background factors that have shown the strongest
connection to sexual abuse, both across and within studies, have been
those relating to parents and family. We have subdivided these into
four categories: (1) parental absence and unavailability; (2) poor
relationship with parents; (3) parental conflict; and (4) presence of a
stepfather. However, it should be kept in mind that these may all be
measuring aspects of the same issue.
In regard to parental absence, seven studies have found higher
vulnerability to sexual abuse among women who lived without
their natural mother or father at some time during childhood. In
Finkel- hor’s (1984) student study, having lived apart from a
natural mother was the most powerful risk factor, resulting in an
almost threefold higher risk. Herman and Hirschman (1981) found
separation from mother to be a serious risk for father-daughter
incest. However, these findings were not replicated by Fromuth
(1983) or Peters (1984).
56 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

Finkelhor, Fromuth, and Peters all found that girls who had ever
lived without their natural father were at higher risk, although the
Peters finding applied only to white women. Russell (198$) also
found higher rates among girls living with their biological mother
and no father, but the finding applied only to extrafamily abuse.
Similarly, Miller (1976) found elevated risk for adolescents who
were currently living without their father or without both natural
parents. Bagley and Ramsay (in press) found that any separation
from either parent for a period of six months or more before the
age of 16 was correlated with the likelihood of being a victim. This
is an impressive number of studies with positive findings on the
question of parental absence.
Another variable that in some of its forms might be related to
parental absence is the mother’s employment outside the home. At
least four studies have found this related to sexual abuse (Fromuth,
1983; Landis, 1956; Peters, 1984; Russell, 1983). Peters found that
the relationship held only for white women, and Russell, whose
findings were short of statistical significance for the whole sample,
found that mother’s employment increased risk significantly in
households with stepfathers. Finkelhor (1984) did not find any risk
for mother’s employment.
Parental unavailability may also be indicated by a disabled or ill
parent. Finkelhor (1984) found higher rates of sexual victimization
among girls who reported that their mothers were often ill. Peters
(1984) reported somewhat higher rates (only at p = .10) for girls
with one parent who was disabled by substance addiction,
emotional problems, or physical ailments. Looking specifically at
father- daughter incest families, Herman and Hirschman (1981)
found signif- icantly more mothers who were seriously ill (most
commonly due to alcoholism, depression, or psychosis) and also
mothers who were burdened with many children and pregnancies.
However, Fromuth (1983), using the same variables as Finkelhor,
did not confirm any association between sexual abuse and mother’s
illness.

Poor Relationship with Parents

About as impressive as the findings concerning parental
absences are the findings concerning poor relationships with
parents, particu- larly with the mother. This has also been
substantiated by a large
 Finkelhor, Baron / CHILD SEXUAL ABUSE 57

number of studies (six) and is one of the most consistent findings to

date. Landis (1956) was one of the first to observe that molested
women reported a more distant relationship with their mothers.
They also were less likely to mention their mothers as their primary
source of sex information. Finkelhor (1984) found in his sample of
college women that those who were at higher risk were the ones
who said they were not close to their mother or received little
affection from their mother or father. The same was true if they
rated mothers high on a scale of sexual punitiveness. All of these
factors except maternal affection made an independent contribution
in a multivariate anal- ysis of risk factors. In her multivariate
analysis, Peters (1984) found that not being close to mother was the
variable that was most predic- tive of sexual abuse. Miller (1976)
also used multivariate analysis and found that poor relationships
with either parent were potent corre- lates of abuse. Bagley and
Ramsay (in press) showed an impressive correlation of .56 between
sexual abuse and the EMBU, a well-vali- dated instrument for
measuring a child’s recollections of harsh, puni- tive, and
emotionally cold parenting.
One partially dissenting finding was Fromuth (1983), who found
that only 4 o of the victims and 2 o of the nonvictims reported
having a distant relationship with their mother. The wording of the
question was perhaps not sensitive enough to discriminate
problematic relation- ships from satisfactory ones in her sample.
However, even Fromuth found a significant association on the
question of closeness with father. She also found higher risk if
father had ever been physically abusive toward the child. In another
dissenting finding, Gruber and Jones (1983) actually found poor
relations with mother to be a p7otective factor in a multivariate
analysis of risk; but their sample consisted of 20 victimized and 21
nonvictimized girls in a delinquency intervention program, most of
whom came from extremely deprived family environments.
Altogether, the collection of positive findings suggests that having
a poor relationship with one’s parents is one of the most common
correlates of sexual abuse.

Conflict Between Parents

Besides reporting a poor relationship with their parents,
sexual abuse victims are also more likely to report a poor
relationship between their parents. Five out of five studies
confirm such a risk
 Finkelhor, Baron / CHILD SEXUAL ABUSE 59

familial abuse (Finkelhor, 1984; Miller, 1976). It is unfortunate

that more of these studies have not made sure that the family
condition under question existed prior to the time that the abuse
occurred.
If these clusters of variables are risk factors, there are at least two
major dynamics that explain the relationship. One concerns super-
vision. When a child is missing a parent, has a poor relationship
with a parent, or has parents who are hostile to one another, the
child may be less well supervised and thus less well protected from
predatory adults. The finding that abused girls received less sex
education from their mothers (Finkelhor, 1984) is related here,
because giving sex education may be one of the ways concerned
mothers help protect their children. Although the omission may be
unintentional or even well-intended, not receiving sex education is
a form of not being supervised and not being protected.
A second possibility is that the poor parental relationships are
connected to sexual abuse through emotional disturbances in the
child. If a child is unhappy, emotionally deprived, or needy, then
she may be more conspicuous and more vulnerable to an adult
interested in molesting her. She may be more amenable to the
offers of friend- ship, appreciation, and the material rewards that
the offender makes, and she may be less able to stand up for
herself. Moreover, she may be more afraid to tell her parents about
the abuse because she has reason to believe she will not be
supported. This second dynamic does not necessarily preclude the
first. In fact, it seems likely that emotional disturbance and lack of
supervision are both at work.

Stepfather Families
Besides parental absence, conflict, and lack of support, another
parental factor—the presence of a nonbiologically related father—has
been widely discussed as a risk factor for sexual abuse. Out of six
studies examining this factor, four have confirmed its significance.
Of these studies, three—Miller (1976), Finkelhor (1980), and Gruber
and Jones (1983)—found that having a stepfather increased a girl’s
risk for all types of sexual abuse. Russell (1983) found an increased
risk for father-daughter abuse only. Fromuth (1983) and Peters (1984)
did not find any increased risk for any type of abuse. Although most of
these studies, as mentioned earlier, indicate that simple father absence
is a risk factor, both Finkelhor (1984) and Gruber and Jones ( 1983)
 Finkelhor, Baron / CHILD SEXUAL ABUSE 61

to confirm and even specify the risk. One caution about some of
the findings reported here is that they may be historically outdated.
They are based mostly on studies of women who grew up before
the 1970s, when fewer children had stepfathers. It may be very
different to live in a stepfather family in an era when there are few
such families, com- pared to the current era when such families are
much more common- place. Assuming that this vulnerability does
show up even in studies of current cohorts of children, it would
support an important new direction for social policy. A case might
be made that families con- taining stepfathers should be the target
of sexual abuse prevention efforts.

SUMMARY OF RISK FACTORS

This review of findings concerning risk factors has been fruitful in

identifying some emerging consistencies across studies. A number of
background factors seem to be clearly associated with higher risk in
several studies (see Table 2).
Not surprisingly, girls are at higher risk than boys. Also, preadoles-
cents appear to be more at risk than either younger or older children.
Girls with few friends in childhood report more abuse, but this may
be an effect of abuse and not a risk factor. The strongest and most
consistent associations across the studies concerned the parents of
abused children. Girls who are victimized are (l) more likely to have
lived without their natural father, (2) more likely to have a mother
who was employed outside the home, (3) more likely to have a mother
who was disabled or ill, (4) more likely to witness conflict between
their parents, and (5) more likely to report a poor relationship with
one of their parents. Girls who lived with stepfathers were also at
increased risk for abuse.
The studies also seem to agree that two factors do riot increase
risk for abuse. Blacks have no higher rates of abuse than whites;
and girls from lower social strata are at no more risk than others. In
fact, sexual abuse appears very democratic in its social class
distribution. The studies have demonstrated strong agreement on
these points.
62 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

TABLE 2
Risk Factors for Child Sexual Abuse: Summary of
Empirical Findings from Retrospective Surveys

SFudies Finding Association/

Factor Studies Looking for Association COI C) fSlO/g

Female gender 11/ 12 Gir Is higher r isk than

boys
Preadolescence (I 0-13) 5/7 May depend on
pea k definition of abuse
Lower SES 2/10 No associatio n with
social class in
community studies.
Only student studies
found relationsh ip
Blacks 0/5 Black -white risk
Rural residence 1/4 equal Uncertain
Urban residence 2/4 Uncertai n
Few friends in childhood 3/3 Possible r isk
factor; but
possibly effect of
abuse
Ever I ived without natural mother 3/6
Mixed
Ever Iived without natural father 6/6
Strong association
Mother’s employment outside home 4/5
Strong association
Mother disabled or ill 2/4
Possible association
Poor relationship with parent 6/7
Strong association.
Three studies
conf irm in
multivariate
analysis
Parents in conf|ict 5/5
Strong association
Stepfather fam i| ies 4/6
Strong association

RECOMMENDATIONS FOR FUTURE STUDIES

Considering that the study of child sexual abuse is still in its early
stages, it is impressive that there are so many convergent findings
 Finkelhor, Baron / CHILD SEXUAL ABUSE 63

concerning risk factors. But the agreement on certain points should

not obscure the fact that these findings are subject to many caveats.
It also should not obscure the fact that to confirm fully the
implications from current research, we need more research of much
greater depth and sophistication. In this section, we would like to
make a variety of suggestions about the direction future research
should take.

Better Specification of Risk Factors

The research reviewed here has concentrated on confirming the
importance of certain risk factors such as stepfamilies, social isola-
tion, and poor parental relationships. But as was pointed out earlier,
we must be careful to acknowledge all that is still uncertain about
these findings. First, it needs to be substantiated that these are
actually risk factors—that they precede the sexual abuse and are
not simply the result of abuse. Second, why and how these risk
factors create risk must be ascertained. This can be done in part by
making the variables more precise. What is meant, for example, by
“poor relationship with one’s parents”? This variable needs to be
broken down into such specific categories as “parent did not spend
enough time with the child,” “parent had negative feelings toward
the child,” “parent used restrictive or abusive parenting practices,”
and so forth. In addition, what is meant by “parental conflict?”
Does this mean “violence between parents,” “conflict over money,”
“conflict over child care?” This also needs to be specified.
In some cases, the risk factor identified may be a proxy for
another variable that is the actual risk factor. Is “rural residence,”
for exarri- plc, a proxy for “less supervision?” Is “few friends in
childhood” a proxy for “feeling lonely”? Because some of the
identified risk factors, such as having a stepfather, may create risk
through various causal sequences (“more predatory men in
household” and “alienation from mother”), some additional
variables that would specify these sequences need to be added to
the analysis. All of this will improve our understanding of how
these risk factors operate.

New Classes of Risk Variables

In addition to better specification of already-identified risk
factors, whole sets of new risk factors need to be considered. For
example, no one has examined whether physical characteristics of
children—their
64 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

size, weight, strength, or attractiveness—make a difference in their

risk for abuse. Similarly children’s temperamental characteristics
— whether they are passive, aggressive, introverted, or extroverted—
may play a role. There are also many psychological variables that
have not yet been examined, such qualities as the child’s locus of
control and self-esteem. As a comparison, it should be noted that
the literature on rape avoidance found that women who were
stronger and more athletic were less likely to be raped (Bart, 1981).
Such factors might have pertinence to child sexual abuse and
development. Various cognitive factors also need to be considered:
Does the level of a child’s sex education influence vulnerabilityi•
What about their understand- ing of child molestation? What about
their beliefs about adults, obedience, and their own rights?
Intelligence may also play a role; children with more intelligence
may have better coping skills and might be able to avoid sexual
victimization. Before the field gets too narrow in its concerns,
researchers need to brainstorm about a wide range of other possible
risk factors to test.

Evaluation of Smaller Demographic Subgroups

In the analysis of possible sociodemographic risk factors, such
large categories as black, white, rural, urban, lower class, and
middle class have been considered. But a much more detailed
analysis of subgroups is also required. Perhaps risk is particularly
high or low, as some preliminary findings suggest, in various
ethnic populations, such as Asians, Hispanics, and Native
Americans. Moreover, rural populations are not uniform; perhaps
risk is particularly high among rural populations in particular parts
of the country. Certain specialized family structures need to be
assessed, such as children who live in single-father families (a
rather small subgroup of all children), as well as children who live
with adopted parents and children who live in institutions.
Certainly given recent public concern we need to assess whether
there is higher risk for children who participate in day-care
programs. It is also important to look at the risk to children who
have emotional difficulties or are handicapped in some way.

Better Specification of Risk by Developmental Stage

Most of the risk factor studies done so far have considered child-
hood as a single stage. However, it is also plausible that different risk
 Finkelhor, Baron / CHILD SEXUAL ABUSE 66

factors create vulnerability at different stages of childhood. For

example, it may be that a working mother creates risk for her
child’s molestation only from age two to eight, but not afterward. It
may be that a stepfather creates risk for molestation from ages 10 to
16, but not before. Children’s needs and their environments change
with different developmental stages, so it is plausible that risk
factors for each stage would also change. Researchers need to
isolate risk factors by developmental stage.

Multivariate Analysis
One implication of the more complex questions that we have
proposed here is that researchers need to use more complex data
analysis techniques. Much of the research to date has limited itself
to simple bivariate relationships. As a result, it is still unknown to
what extent some of these observed risk factors are simply
expressions of other antecedent risk factors, and would disappear in
multivariate analysis. For example, it is possible that the increased
risk in step- parent families may be explained by parental
unavailability or low marital satisfaction. It is imperative that in
any future analysis of risk factors researchers take advantage of
multivariate techniques.

Sample Size
Many of the foregoing suggestions for improving research also
dictate research designs using large samples. Large samples are
needed for studies that (1) look at the risk in small subpopulations;
(2) use multivariate techniques; (3) break up the dependent variable
into subcategories of extra-familial and intra-familial abuse; and (4)
look at abuse in the various stages of childhood. The majority of
sexual abuse studies to date have had under 1,000 respondents.
New studies are needed that use samples of 2,000 or more
respondents. Such studies, of course, will be expensive, but there
are techniques for reducing survey costs and these need to be
reviewed for their applica- bility to sexual abuse research. It may be
possible, for example, to do surveys using the telephone, at least for
the identification of subjects (Wyatt, 1984). It may be possible to
combine sexual abuse research with other kinds of studies, such as
health status and crime victimi- zation surveys.
66 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

Other Survey Designs

large sample surveys are not the only way in which risk studies can
be conducted, however. If there are small subgroups in which, as we
suspect, risk may be particularly high, special studies can be done
within these subgroups. For example, researchers may wish to do a
study with a cohort of children who live in stepfamilies, or a cohort of
children whose mothers are disabled by psychiatric illness. These can
be done both prospectively or retrospectively. By gathering a large
number of subjects from one subgroup, it may be possible to tell
under what conditions this particular risk factor seems to operate. For
example, it may be that children in stepfamilies that are formed when
the child is very young are not at the same degree of risk as step-
families formed when the child is already a preadolescent.

Studies of Child Populations

Virtually all of the risk factor studies thus far have been of
adults or young adults looking back on their childhood. These
studies have some important limitations. For example, it is very
hard to get detailed and accurate information about parent-child
relationships and other childhood variables from somebody who
has been out of their family for 20 years. The validity of the data
on the relationship between the parent and the child may be very
poor, and there may be a loss of memory concerning other
important events that might be crucial for understanding the
antecedents or the sexual abuse.
There is an obvious need for more contemporaneous studies that
gather information on families immediately after the abuse has
occurred to assess how these families and children may differ from
others. Questions about supervision, parental relationships, and
parental conflict may be much more reliably evaluated in these
immediate studies. Of course, one problem with such studies is that
only a small portion of abuse comes to public attention. As a result,
they may not tell us about risk factors for sexual abuse so much as
risk factors for sexual abuse that gets detected.
The optimal kind of design to take care of both problems is a
longitudinal study that would gather information early in the family
life cycle on families with young children, and then follow them and
the children until the children become adults. Then it might be
possible to ascertain from self-reports in interviews whether sexual
 Finkelhor, Baron / CHILD SEXUAL ABUSE 67

abuse had occurred, even if it had never been reported. The data
gathered early in the study would give fairly unbiased information on
possible predictors of subsequent abuse.
A modified version of this design, which has advantages
because it does not take as much time, would be for investigators
to find groups of children and families of a previous generation
who had already been the subjects of study early in the family life
cycle in some other unrelated research. It might then be possible to
follow up those children (now adults) and find out whether they
had been abused during their childhood. The data gathered in the
earlier study on their families and childhoods, if it were still
available, could be used to test risk factors. It would provide a
particularly good opportunity to see if it would have been possible
to predict which children would be subsequently abused.

A FINAL CAVEAT: BLAMING THE VICTIM

Risk factors always require cautious handling, but this is partic-

ularly so when dealing with child sexual abuse, which has been
fraught with myth and misunderstanding. In the past there have
been those who have taken findings such as the fact that children
without friends are at higher risk and used these to hold victims
responsible for being abused. Others have taken findings such as
the fact that children with working mothers are at higher risk and
used these to hold mothers responsible for abuse.
Conclusions such as these are particularly likely when victims
are studied without simultaneously studying offenders. All the data
in this review are based on victim surveys that, because they gather
detailed information on victims and their families, and correspond-
ingly little about offenders and their backgrounds, tend to over-
emphasize the role of victims and obscure the contribution of the
perpetrators.
It is important to emphasize that true causal responsibility for abuse
lies with offenders. All the research suggests that it is offenders
who initiate the sexual activity. In an effort to put this issue in
perspective, we have developed what we call the “Four
Preconditions Model of Sexual Abuse” (Finkelhor, 1984). This
model postulates that four
68 JOURNAL OF INTERPERSONAL VIOLENCE / March 1986

preconditions must exist for sexual abuse to occur: (1) There must
be an offender with the motivation to sexually abuse, (2) the
offender must overcome internal inhibitions against abusing, (3) the
offender must overcome external obstacles against abusing; and (4)
the offender must overcome resistance by the child. To “explain”
sexual abuse fully, one must account for the presence of all four of
these preconditions.
Most of the risk factors that we have been discussing in this
article refer to considerations of preconditions 3 and 4. They are
situations that may reduce the supervision of the child and thus
make it easier to overcome external obstacles to abuse or they are
conditions that under- mine a child’s emotional health and thus
their ability to resist abuse. But one of the values of the model is
that it makes clear that preconditions 3 and 4 only make a
difference after preconditions l and 2 have already been met. That
is, they are only relevant once there is an offender who is
predisposed to molest and who has overcome internal inhibitions
against doing so. Thus many children who may suffer from risk
factors at preconditions 3 and 4 will never experi-
ence abuse.
It is important that we know about risk factors at preconditions 3
and 4. Through knowledge of these factors we may be able to
design programs to help protect children and families at higher risk.
For exmaple, if children with few friends are more vulnerable,
these might be children who school personnel want to target for
special attention. But the search for these risk factors should not
obscure the fact that these are not the ones that cause abuse. Our
ult•rriate goal should be not simply to protect high risk children
from abuse but to eliminate abusive behavior itself.

CONCLUSION

The study of risk factors for sexual abuse is a very promising

research area. In a few short years, a number of studies have
produced quite convergent findings. We are now at the point where
we can readily build on these findings and obtain some very useful
answers. The tools for doing this research are readily available as
well. We have good experience in surveying nonclinical
populations to obtain reports on abuse. Other epidemiological
techniques for risk factor
 Finkelhor, Baron / CHILD SEXUAL ABUSE 69

analysis are also well developed. Such studies become even more
important as we develop techniques of prevention education. They
may mean that we can better target our education at the child popula-
tion at greater risk. It is an area where our research should soon bear
results in terms of achieving concrete reductions in the number of
sexually abused children.

NOTES

1. “Risk factor” is a concept used widely by epidemiologists and defined by

Mausner and Bahn (1974) as “factors whose presence is associated with an
increased likelihood that disease will develop at a later time.” Thus although the risk
factors are presumed to precede the disease temporally, they are not necessarily
causal factors— simply markers of higher susceptibility. Thus sex, age, or ethnicity
may be risk factors without having any direct connection to etiology.
2. Peters’s (1984) study is a companion study to Wyatt’s (1984) Los Angeles
survey. However, they have somewhat different samples. Wyatt’s (1984) study
recruited a quota sample of 250 black and white women in Los Angeles for extensive
interviews about many aspects of sexual history including sexual abuse. Peters (1984)
recontacted this sample and completed follow-up interviews with 119 of the original
women, using an interview much more exclusively focused on the risk factors for and
effects of sexual abuse.

REFERENCES

American Humane Association. (1981). National stud y on child neglect arid abuse
reporting. Denver: American Humane Association.
Badgley, R., Allard, H., McCormick, N., Proudfoot, P., For tin, D., Ogilvie, D., Rae-
C rant, Q., Gelinas, P., Pepin, L., & Sutherland, S. [Committee on Sexual Offences
Against Children and Youth l ( 1984). Sexttof offenses against children (Vol. 1).
Ottawa: Canadian Government Publishing Centre.
Bagley, C., & Ramsay, R. (in press). Disrupted childhood and vulnerability to sexual
assault: Long-term sequels with implications for counselling. Social lYor/t and
Human Sexuality.
Bart, P. (1981). A study of women who both were raped and avoided rape. Journal oJ
Sociaf Issues, 37 (4), 123-137.
Burnam, A. (1985). Personal communication concerning the Los Angeles
Epidemio- logical Catchment Area Study.
De Francis, V. (1969). Protecting the child victim o/ sex crimes committed by adults.
Denver: American Humane Association.
Finkelhor, D. (1979). Sexual 1 y -victimized childien. New York: Free Press.
 Finkelhor, Baron / CHILD SEXUAL ABUSE 71

Pelton, L. H. (Ed.) (1981). The social context oJ child abuse and neglect. New York:
Human Sciences Press.
Peters, S. D. (1984). The relationshi p between c/it/d/iood sexual victimization and
adult depression among APro-American and inhile uiomen. Unpublished Ph.D.
dissertation, University of California, Los Angeles.
Peters, S. D., Wyatt, G., & Finkelhor, D. ( 1986). Prevalence: A review of the research. In
D. Finkelhor, A sourcebook on child sexual abuse. Beverly Hills, CA: Sage.
Plummer, C. (August, 1984). Prwenting sexual abuse: What in-school programs teach
children. Paper presented at the Second National Conference for Family Violence
Researchers, Durham, NH.
Russell, D. (1983). IntiaJamil iaI child sexual abuse: A San Francisco suruey. Final
report to the National Center on Child Abuse and Neglect.
Seidner, A. L., & Calhoun, K. S. (1984, August). Childhood sexual abuse: Factors
related to differential adult ad justment. Presented at the 2nd National Conference
on Family Violence Research, Durham, NH.
Straus, M. A., Gelles, R., & Steinmetz, S. {1980). Behind closed doors: P'iolence in the
American Family. New York: Doubleday.
Summit, R., & Kryso, J. (1978). Sexual abuse of children: A clinical spectrum. Amexi-
can Journal oJ Oithopsychiatry, 48, 287-251.
Timnick, L. (1985a, August 25). 22$ in survey were child abuse victims. Los Angeles
Times, p. 1.
Timnick, L. ( 1985b, August 26). Children’s abuse reports reliable, most believe. Los
Angeles Times, p. I.
Tobias, J. L., & Gordon, R. (1977). Operation Lore. Michigan State Police. (mimeo)
Trainor, C. (1984). Sexual maltieatment in the United Slates: A fire-year perspective.
Paper presented at the International Congress on Child Abuse and Neglect,
Montreal.
Wood, S. C., & Dean, K. S. (1984). Final report: !S exual abuse of males research project.
Washington, DC: National Center on Child Abuse and Neglect (90 CA/812).
Wyatt, G. (1984). Sociocttfturaf issues in the prevention of child sexual abuse. Unpub-
lished manuscript.
Wyatt, G. (1985a). Personal communication.
Wyatt, G. E. (1985b). The sexual abuse of Afro-American and American White women
in childhood. Child Abuse and Neglect, 9, 507-519.

David Finkelhor is Assotiate Chair of the Family Research Laboratory and Associate
Director of the Pamily Violence Research Program. He has published three books on
the subject of thild sexual abuse. the most recent of which, Sourcebook on Child Sexual
Abuse, reviews current knowledge concerning prevalence, offenders, risk yactors, and
long-term effects.
Larry Baron received his Ph. D. in sociolog y from Rutgers Universit y and is a
Lecturer irt the Department of Sociolog at Yale Univeisit J. He is cuiientl y completing
a hook on tape artd its relalionsh ip to pornography, sexual inequalit , and cultural
support Jot violence in the United S tates.