National Estimates of Missing Children: Selected Trends, 1988-1999
National Estimates of Missing Children: Selected Trends, 1988-1999
National Estimates of Missing Children: Selected Trends, 1988-1999
Department of Justice
Office of Justice Programs
Office of Juvenile Justice and Delinquency Prevention
December 2004
www.ojp.usdoj.gov
The words “missing child” call to mind tragic and frightening kidnap
pings reported in the national news. But a child can be missing for many
reasons, and the problem of missing children is far more complex than
the headlines suggest. Getting a clear picture of how many children
become missing—and why—is an important step in addressing the
problem. This series of Bulletins provides that clear picture by summariz
ing findings from the Second National Incidence Studies of Missing,
Abducted, Runaway, and Thrownaway Children (NISMART–2). The series
offers national estimates of missing children based on surveys of house
holds, juvenile residential facilities, and law enforcement agencies. It also
presents statistical profiles of these children, including their demographic
characteristics and the circumstances of their disappearance.
1 For methodological reasons, the analysis does not cover nonfamily abductions (including
stereotypical kidnappings) and children categorized as “thrownaway.” See methodology sidebar
on pages 4–7.
J. Robert Flores, Administrator
Access OJJDP publications online at www.ojp.usdoj.gov/ojjdp
NISMART
Key Findings Comparison of Incidence Rates for Missing
Children, 1988 (NISMART–1) and 1999
In the three categories considered (family abductions,
(NISMART–2)
runaways, and lost, injured, or otherwise missing chil
dren), the analysis finds: Rate per 1,000 children ages 0–17
8
■ No evidence of any increase in the incidence of miss
ing children between 1988 and 1999. 7
Background 4
Results
This section presents findings for three categories of
end of a legal or agreed-upon visit and the child
children: victims of family abductions, runaways, and
was kept at least overnight. Policy focal family
children categorized as lost, injured, or otherwise miss
abductions meet at least one of three addi
ing. The subsections begin with definitions of the cases
tional conditions: (1) the abductor attempted to
included in the analysis. In each category, definitions
conceal the taking or whereabouts of the child
distinguish between “broad scope” and “policy focal”
or to prevent contact with the child, (2) the
cases—i.e., all cases and those considered to be more
abductor transported the child out of state, or
serious—as explained below. The findings are summa
(3) evidence existed that the abductor intended
rized in the figure on this page.
to keep the child indefinitely or to affect custo
Family Abductions dial privileges permanently.
Definitions. An episode qualifies as a broad ■ The incidence rate for children who experienced
scope family abduction if, in violation of a broad scope family abductions declined significantly,
custody agreement or decree, a family member from 5.62 per 1,000 children in 1988 to 4.18 in 1999.
took a child or failed to return a child at the
2
NISMART
■ The slight increase in the incidence rate for children decline in the incidence rate for children who experi
who were victims of policy focal family abductions, enced broad scope family abductions. Stepparents may
from 2.59 per 1,000 children in 1988 to 3.15 in 1999, be less aware of, or less likely to report, less serious fam
is not large enough to be statistically significant—i.e., ily abductions, which tend to involve conflicts between
the difference in rates may simply be due to chance biological parents.
and may not reflect an actual increase in the inci
dence rate. Runaways
Definitions. An episode qualifies as a broad
One might expect an increase, not a decrease, in the
scope runaway if it meets one of the following
incidence of family abduction victims, given that the
criteria: (1) a child left home without permis
population has become more geographically mobile,
sion and was away at least one night; (2) a
divorce rates remain high, and parents (particularly
child made a statement or left a note indicat
fathers) have rising expectations for more equitable cus
ing intent to run away and then stayed away
tody arrangements. However, countervailing trends may
at least overnight; (3) a child age 15 or older
be at work: greater public awareness and avoidance of
was away, chose not to come home when
risks associated with catastrophic conflicts over custody,
expected, and stayed away at least two nights;
improved public access to custody dispute resolution
or (4) a child age 14 or younger was away,
services, and court system reforms that make it harder
chose not to come home when expected, and
for disgruntled parents to turn to other states for more
stayed away at least one night. Policy focal
favorable treatment.
runaway episodes meet the additional condi
Such factors may explain the decline in the incidence tion that the child was without a familiar and
rate for children who experienced broad scope family secure place to stay for at least one night.
abductions. The lack of decline in the incidence rate for
■ The incidence rate for broad scope runaways declined
children who experienced more serious policy focal fam
from 7.09 per 1,000 children in 1988 to 5.28 in 1999.
ily abductions may mean that these improvements have
This decrease approaches statistical significance,
had less impact on the most acrimonious, combative
meaning that it probably, but not conclusively, indi
custody situations in which noncustodial parents resort
cates an actual decline.
to extreme unilateral actions.
■ The decrease in the incidence rate for policy focal
At least one demographic change between 1988 and 1999
runaways, from 2.06 per 1,000 children in 1988 to
may have affected family abduction estimates: a decline
1.26 in 1999, is not large enough to be statistically
in the number of children who live with both biological
significant. The difference in rates may simply be due
parents. During that period, the U.S. Census Bureau’s
to chance and may not reflect an actual decrease in
Current Population Surveys (conducted in March of each
these cases.
year) show a 4.6-percent decline in the number of chil
dren living with both parents, and NISMART caretaker The likely decline in the incidence rate for broad scope
respondents who were biological parents of the child in runaways and the possible decline in the incidence rate
question declined 4 percent between NISMART–1 and for policy focal runaways could have a variety of expla
NISMART–2. In NISMART–1, 9 out of 10 respondents nations. As context, it is important to note that most
were parents, and the rest were primarily grandparents, runaway episodes (as defined in NISMART) are brief,
aunts, and uncles (Finkelhor, Hotaling, and Sedlak, lasting no longer than a day or two. These episodes often
1992); in NISMART–2, fewer (86 percent) were biological involve children doing things they think their parents
parents, more (4 percent) were stepparents, and 6 percent may disapprove of, such as traveling to a distant party or
(about the same as in NISMART–1) were grandparents, event, or spending time with a boyfriend or girlfriend.
aunts, and uncles. Between 1988 and 1999, caretakers may have begun to
The increase in the percentage of NISMART caretakers give teenagers greater independence or may have be
who are stepparents may be especially relevant to the come less able to establish limits in light of competing
3
NISMART
demands on caretakers’ time. This could mean that abuse and neglect, including sexual abuse (Jones and
fewer children who are away from home meet the run Finkelhor, 2001; Finkelhor and Jones, 2004).
away definition of being away without permission.
Lost, Injured, or Otherwise Missing Children
Another possible factor is the increasing availability of
Definitions. An episode qualifies as broad
cell phones and other modes of communication. Chil
scope lost, injured, or otherwise missing if it
dren who have taken liberties may be more likely to
meets one of the following criteria: (1) a child
negotiate an agreement with their caretakers, preclud
disappeared from home or from parental super
ing a sense of alarm or violation of permission.
vision and could not be located for varying
In addition, running away may hold less allure as chil amounts of time depending on age—any
dren become more aware of the dangers involved. The amount of time (ages 0–2), 2 hours (3–4), 3
Internet may now serve as a substitute for hanging out hours (5–6), 4 hours (7–10), 8 hours (11–13),
with friends, taking risks, and having adventures. The overnight (14–17)—or, for a child of any age
decline in running away is also consistent with improve with a serious or permanent physical or men
ments in child and youth well-being during the 1990s tal disability or impairment or life threatening
(Foundation for Child Development, 2004). These medical condition, for 1 hour; (2) a child who
improvements include declines in child poverty (Federal was out with parental permission failed to
Interagency Forum, 2000), teenage drinking (Johnston, return, could not be located, and was gone at
O’Malley, and Bachman, 2000) and pregnancy (Federal least overnight; or (3) a child who was out
Interagency Forum, 2000), youth violence and victimiza with parental permission failed to return or
tion (U.S. Bureau of Justice Statistics, 2000), and child make contact with the parent for at least an
4
NISMART
format and content changed. For example, not all ques NISMART–1 and NISMART–2, reports the estimates as inci
tions were replicated verbatim and, in some instances, dence rates of missing children per 1,000 children ages
the designers changed question format from open- 0–17 in the population, and measures change in these
ended to close-ended, changed question order, col rates over time.
lapsed sequences of questions into a single question,
Standardizing the survey data involved weighting the data
or partitioned a single question into a sequence.
to reflect the census-based population ages 0–17 at the
■ Other analyses in this series of Bulletins focus entirely time of each survey. All comparisons between the two
on NISMART–2 and use that survey’s definitions to surveys take into account the fact that the estimates are
calculate estimates. However, to permit comparisons based on samples and, as a result, sampling error affects
with NISMART–1 data in this Bulletin, researchers cal their precision. This sampling error is reflected in the con
culated NISMART–2 estimates according to the origi fidence interval around each estimated rate—i.e., a range
nal NISMART–1 definitions. Because the extent to within which the true number should fall 95 percent of the
which changes in questionnaire format and content time when a study like this one is conducted. Details of
may have influenced responses is not clear, these esti the NISMART–1 weighting procedures and variance esti
mates are close approximations to (but not strict repli mation are available in NISMART Household Survey
cations of) the original NISMART–1 definitions. Methodology (Sedlak, Mohadjer, and Hudock, 1990). Similar
details for NISMART–2 will be available in OJJDP’s NIS-
■ NISMART–2 included a household survey of youth.
MART–2 Household Survey Methodology Technical Report
NISMART–1 did not. Therefore, comparisons between
(Hammer and Barr, forthcoming).
the two years must be limited to the adult caretaker
household surveys conducted in both years. Comparing the NISMART–1 and NISMART–2
Household Samples
Technical Notes: Rates, Weighting, and
Sampling Errors A primary challenge in designing the NISMART–2 house
hold survey was to have a large enough sample to support
According to the U.S. Census Bureau, the population ages
an estimate of the number of children who were victims of
0–17 increased from just over 63 million in 1988 to almost
nonfamily abductions. Thus, NISMART–2 attempted to reach
72 million in 1999. Because of this change, the analysis in
a much larger household sample of adult primary caretak
this Bulletin standardizes the incidence estimates for
ers than was reached in NISMART–1. The table on page 7
hour after return, or contact was expected triggered a call to police to help locate the missing child.
because the child suffered harm or an injury Most commonly, lost, injured, or otherwise missing
that required medical attention. Policy focal episodes involve children who are delayed coming home
lost, injured, or otherwise missing episodes or contacting their parents because of weather, travel
meet the additional condition that the police delays (such as car breakdowns), and other extenuating
were contacted to help locate the child. circumstances, and children who are perceived as miss
ing because of a miscommunication about their plans or
■ The incidence rate for children who experienced
intentions.
broad scope lost, injured, otherwise missing episodes
decreased significantly, from 6.95 per 1,000 children Once again, a factor in the observed decline could be
in 1988 to 3.40 in 1999. improvements in communication technology, such as
■ The incidence rate for children who experienced the proliferation of cell phones. However, it is also
policy focal lost, injured, otherwise missing episodes possible that the change could arise from methodological
declined from 2.21 per 1,000 children in 1988 to 0.51 differences between NISMART–1 and NISMART–2.
in 1999. This decrease approaches statistical signifi
cance, meaning that it probably, but not conclusively, Summary
indicates an actual decline. This analysis examines trends in the incidence of chil
These findings suggest a clear decline in the incidence dren who are family abduction victims, runaways, and
rate for children who experienced broad scope lost, lost, injured, or otherwise missing, based on NISMART
injured, or otherwise missing episodes and a possible data for 1988 and 1999. It considers children who experi
decline in the incidence rate for children who experi enced a broad scope (i.e., any) incident and those who
enced episodes classified as policy focal because they experienced a policy focal (more serious) incident.
5
NISMART
presents detailed sample comparisons for NISMART–1 and statistics. NISMART-2 had a significantly lower contact rate
NISMART–2 and reveals significant differences between the (the proportion of all cases in which the survey reached
two surveys. Although NISMART–2 contacted nearly 2.5 some responsible member of the household) and coopera
times as many households as NISMART–1, the numbers of tion rate (the proportion of completed interviews among
interviews completed and children represented were not eligible households) and a somewhat higher refusal rate
proportionately larger. (the proportion of all cases in which a household or
respondent refuses to do an interview or breaks off an
Of the 85,522 households contacted in NISMART–2, it was
interview among all potentially eligible households). These
possible to screen 85 percent for the presence of children,
outcomes yielded a lower response rate (the proportion of
yielding 20,170 eligible households with children. Of these,
completed interviews among all eligible households in the
16,111 households (representing 31,787 children) yielded
sample) for NISMART–2. This is consistent with the increas
completed adult interviews, for a response rate of 61 per
ing prevalence of nonresponse (including noncontact and
cent. In NISMART–1, 11,617 eligible households yielded
refusals) noted by other researchers in general population
10,367 completed interviews (representing 20,138 children),
surveys, particularly in random-digit dial surveys like NIS
for a response rate of 78 percent.*
MART (Groves and Couper, 1998; Hox and De Leeuw, 1994;
The recruitment statistics in the table show that the Harris-Kojetin and Tucker, 1999; Steeh et al., 2001).
NISMART–2 sample yielded proportionately fewer con
The recruitment and outcome differences between
tacts with households, fewer eligible households with chil
NISMART–1 and NISMART–2 may stem from their use of
dren, fewer completed interviews among eligible house
different methodologies to sample and screen phone num
holds, and more telephone numbers with unknown
bers. The differences may also reflect an increased reluc
eligibility. Only the percent of ineligible telephone numbers
tance of respondents to report the presence of children
is similar between the two surveys.
living in the household, or possibly a larger proportion
The outcome rates in the table, which are based on of eligible households (i.e., households with children)
American Association for Public Opinion Research stan “hidden” in phone numbers with unknown eligibility.
dard definitions (AAPOR, 2004), mirror the recruitment
A discussion of these differences and their potential impli
cations will be available in OJJDP’s NISMART–2 House
*Response rates for NISMART–1 and NISMART–2 have been computed
hold Survey Methodology Technical Report (Hammer and
with standard definition RR4 (AAPOR, 2004) to facilitate this comparison.
Barr, forthcoming).
The analysis offers evidence of statistically significant their care, the quality of family life and parental atten
declines in incidence rates for children who experienced tion, availability of social and medical services, and
broad scope family abduction episodes and lost, injured, community safety and cohesion—has not worsened and
or otherwise missing episodes, plus some evidence of de may have improved. It is also possible that the allure of
clines in incidence rates for children who experienced running away may have dissipated as parents have given
broad scope runaway episodes and policy focal lost, children greater freedom and independence. The prolifer
injured, otherwise missing episodes. Observed changes ation of cell phones and other modes of communication
in incidence rates for children who experienced policy has made it easier to locate children and clear up misun
focal family abduction episodes and policy focal runaway derstandings, which may help to explain the decline in
episodes are not statistically significant. lost, injured, or otherwise missing children.
The most important finding is the absence of increases Although the findings reported in this Bulletin are
in any of these problems. This finding is consistent with encouraging, they are no cause for complacency. The
growing evidence of improvements in child and youth NISMART estimates for 1999 also reveal large numbers
well-being during the 1990s (Foundation for Child Devel of children and youth still caught up in circumstances
opment, 2004; Federal Interagency Forum, 2000; John of crisis and vulnerability. The family and community
ston, O’Malley, and Bachman, 2000; U.S. Bureau of problems these statistics reflect are unlikely to disappear
Justice Statistics, 2000; Jones and Finkelhor, 2001; anytime soon.
Finkelhor and Jones, 2004).
Finally, methodological differences between the 1988
Such evidence suggests that the environment affecting and 1999 NISMART surveys (see sidebar on pages 4–7)
children—possibly the economic resources available for may explain some of the change (and lack of change)
reported in this Bulletin. Further analysis is required to
6
NISMART
Note: All percents are rounded to the nearest whole number. c Percent computed from screened households with children (equivalent to
a Percent computed from total households contacted. cooperation rate, COOP2).
d Outcome rates are based on American Association for Public Opinion Research
b Percent computed from households screened for children.
standard definitions CON2, COOP2, REF2, and RR4 (AAPOR, 2004).
reach definitive conclusions about the impact of these Studies. Washington, DC: U.S. Department of Justice, Office of
differences. Justice Programs, Office of Juvenile Justice and Delinquency
Prevention.
Conclusion Finkelhor, D., Hotaling, G., and Sedlak, A. 1992. The abduction
The period between 1988 and 1999 saw significant mobi of children by strangers and nonfamily members: Estimating
lization on behalf of missing children. Law enforcement the incidence using multiple methods. Journal of Interpersonal
Violence 7(2):226–243.
officers received special training, and public awareness
grew as a result of media coverage and educational pro Finkelhor, D., and Jones, L.M. 2004. Explanations for the
grams disseminated to schools and families. Have these Decline in Child Sexual Abuse Cases. Bulletin. Washington,
efforts made a difference? What else should be done? DC: U.S. Department of Justice, Office of Justice Programs,
The kinds of data collected thus far are of limited use Office of Juvenile Justice and Delinquency Prevention.
in answering these questions. Data gathered more fre Foundation for Child Development. 2004. The Foundation for
quently and more locally could provide a tool for public Child Development Index of Child Well-Being (CWI),
policy analysis that might lead to even greater effective 1975–2002. Durham, NC: Duke University, Foundation for
ness in combating the problem of missing children. Child Development.
7
U.S. Department of Justice
PRESORTED STANDARD
Office of Justice Programs POSTAGE & FEES PAID
*NCJ~206179*
Washington, DC 20531
Official Business
Penalty for Private Use $300
Harris-Kojetin, B., and Tucker, C. 1999. Exploring the relation This Bulletin was prepared under grant number 95–MC–CX–K004 from
of economic and political conditions with refusal rates to a the Office of Juvenile Justice and Delinquency Prevention, U.S. Depart
ment of Justice, to Temple University.
government survey. Journal of Official Statistics 15(2):67–184.
Points of view or opinions expressed in this document are those of the
Hox, J.J., and De Leeuw, E.D. 1994. A comparison of non- authors and do not necessarily represent the official position or policies
response in mail, telephone, and face-to-face surveys. Applying of OJJDP or the U.S. Department of Justice.
multilevel modeling to meta-analysis. Quality and Quantity
28:329–344. The Office of Juvenile Justice and Delinquency Prevention is a com
ponent of the Office of Justice Programs, which also includes the
Johnston, L.D., O’Malley, P.M., and Bachman, J.G. 2000. Moni Bureau of Justice Assistance, the Bureau of Justice Statistics, the
toring the Future. National Results on Adolescent Drug Use: National Institute of Justice, and the Office for Victims of Crime.
Overview of Key Findings 1999. Washington, DC: U.S. Depart
ment of Health and Human Services.
Jones, L., and Finkelhor, D. 2001. The Decline in Sexual Abuse Acknowledgments
Cases. Bulletin. Washington, DC: U.S. Department of Justice, Heather Hammer, Ph.D., is a Senior Study Director at the
Office of Justice Programs, Office of Juvenile Justice and Delin Temple University Institute for Survey Research, Philadelphia,
PA, and Principal Investigator of NISMART–2. David Finkelhor,
quency Prevention. Ph.D., is Professor of Sociology and Director of the Crimes
against Children Research Center, University of New
Sedlak, A.J., Mohadjer, L., and Hudock, V. 1990. NISMART
Hampshire; and Advisor to NISMART–2. Andrea J. Sedlak,
Household Survey Methodology. Washington, DC: U.S. Gov Ph.D., is Associate Director of Human Services Research at
ernment Printing Office. Westat, Inc.; Project Director of the NISMART–2 Unified
Estimate, Juvenile Facilities Study, and Law Enforcement
Steeh, C., Kirgis, N., Cannon, B., and DeWitt, J. 2001. Are they Study; and Advisor to the NISMART–2 Household Survey.
Lorraine E. Porcellini, B.A., is Study Direction/Sampling
really as bad as they seem? Nonresponse rates at the end of the
Coordinator at the Temple University Institute for Survey
twentieth century. Journal of Official Statistics 17(2):227–247. Research.
U.S. Bureau of Justice Statistics. 2000. National Crime Victim The authors extend their appreciation to the many individuals
who contributed to this Bulletin, to OJJDP NISMART–2
ization Survey trends, 1973–98: Adjusted victimization rates. Program Manager Barbara Allen-Hagen, and to the NISMART
Retrieved August 13, 2000, from the Web: survey respondents.
www.ojp.usdoj.gov/bjs.