Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 31
Participation in Criminal Careers
Efforts to develop effective crime con-
tro! policies can be enhancer! by an un-
derstanding of the ways in which inclivict-
ual criminal behavior contributes to total
crime. The panel's approach to the study
of criminal behavior rests fundamentally
on a criminal career paradigm of indivict-
ual offending, which disaggregates the
various elements of indiviclual offending
into four key dimensions. This chapter
summarizes the literature on one of those
dimensions, the level of participation in
crime and some demographic and social
covariates of participation; Appendix A
(Visher and Roth) ofthis report provides a
detailed critical review of incliviclual
stuclies.i
iGordon and Gleser (1974) formally defined
"prevalence of delinquency" (referred to here as
participation), and resolved some of the confusion
surrounding its precise definition, computation, and
terminology. In a later paper, Gordon (1976) re-
viewed several important studies of"prevalence"
based on official-record data through the juvenile
period (to age 18~. The Visher and Roth review
(Appendix A) of participation in offending, on which
this chapter is based, updates the review by Gordon
and also includes studies of adult participation and
self-reported participation. While Gordon used ra
31
Participation in offending is reflected
in the fraction of a population ever com-
mitting (or "doing") at least one crime
before some age a (Da) or currently active
during some observation period (ct). Esti-
mates of participation have been derives!
from self-report surveys of ofiencling in
cross-section samples. Commonly, these
studies present respondents with clescrip-
tions of specific criminal acts that corre-
spond to legal definitions of crimes. Cu-
mu~ative participation (Da) is computed
from responses to questions of the form
"Have you ever. .. ?" Questions of
the form "Within the past year, have
you . . . ?" provide the basis for estimates
of current participation all. Because the
fraction of people who have committed
any crime during the prece(ling year is
smaller than the fraction who have ever
committee] a crime, estimates of D obvi
cial differences as a defining variable and adjusted
published participation rates to reconcile differ-
ences across studies, race is treated here as just one
source of variation, and participation rates are re-
ported as they appear in the published literature.
OCR for page 32
32
ously cannot be smaller than estimates of
for the same sample and crime type.
Estimates of the fraction of a popula-
tion that is ever arrested, B. and the per-
cent of a population arrested within some
observation period, b, measure participa-
tion as reflected in arrest (or other official)
records. Since many people who commit
crimes are never arrested, estimates of
arrest participation are obviously less
than those of crime participation, that is, b
is less than d. As careers continue, how-
ever, the chance of at least one arrest
increases, so that the difference narrows
between B and D.
Although studies based on official rec-
ords vary widely in design details na-
ture of sample, criterion for identifying
individuals who have participated, and
length of observation period- b and B
can be computed in only a limited num-
ber of ways. Agency statistics on the age
distribution of arrestees within a year can
be used to compare b for any age by
dividing the number of arrestees of that
age by the entire population of that age.
Because researchers usually want to
study influences on participation beyond
the demographic and criminal history
variables contained in agency records
and because only a fraction of crimes
result in arrest, current participation is
usually stuclied in terms of self-reported ct
rather than officially recorded b.
To estimate officially recorcled cumula-
tive participation rates for example, BE
(the fraction of a population arrested by
age 18), three approaches are available.
First, in prospective longitudinal studies,
a cohort of individuals is selected and
followed over time; notations of the ages
at first arrest are used to compute B by
each age. Second, in cross-section stud-
ies, agency files are searched to ascertain
the fraction of any sample of 18-year-olds
for whom official records exist. Third, if
arrest or court referral records for first
offenders are easily distinguishable, a
CRIMINAL CAREERS AND CAREER CRIMINALS
life-table approach similar to that used in
mortality studies can be used: the clistri-
bution of age-specific initiation rates, Ia
(i.e., the fraction of the population whose
first arrest occurs at each age) can be
computed and cumulated to age 18 to
estimate Bit. This rate can be calculates]
from the age (listribution of first arrestees
in one year (T. Monahan, 1960, as cor-
rected by Gordon and Gleser, 1974), or
from arrest history data on members of a
bird cohort (Wolfgang, Figlio, and Sellin,
1972~. However, to understand the rela-
tionship between age and initiation rates,
the conditional initiation rate, or hazard
rate (ha)' may be more useful: this is the
fraction that age-a initiators represent of
the total a-year-old~s who have not yet
initiated their criminal activity.2 Each of
these three research designs has different
implications for the measurement of B.
Designs of self-report and official-
record participation studies vary in other
cletaiTs, such as the time period in which
the data were collected; the seriousness
of the behavior considered; and the
threshold of contact with the criminal
justice system, ranging from none (i.e., an
uncletected crime) to a police contact, a
court referral, or conviction. Self-report
studies of delinquent activity usually in-
clude a broader range of deviant behav-
iors and less serious crime types than
studies based on official police or court
records. The source of official records also
affects estimates: juvenile court recorcls,
which record only incidents referred for
frontal adjudication, are likely to produce
2Thus, for example, in a cohort of 1,000 boys, if
200 have initiated their careers by age 16 and 100
more begin in their 17th year, then In = 100/1,000
= 0.10, and he = 100/~l,000 - 200) = 0.125. At
later ages, more of the offenders in the sample have
already begun offending and they are therefore no
longer at risk of becoming offenders for the first
time, so Ja is likely to decrease but ha could well
increase.
OCR for page 33
PARTICIPATION IN CRIMINAL CAREERS
lower estimates of participation than are
records of police contacts.
As inventories of all crimes committee!
by an individual, both official records and
self-reports are subject to error. The prob-
ability that an individual's offense will be
recorded by police depends on such
factors as patrol allocations, victims' re-
porting behavior, offenders' success in
avoiding police detection, and police dis-
cretion in recording offender contacts,
which is especially wide for juveniles.
Errors in self-reports of criminal involve-
ment arise primarily from under- ant!
overreporting of criminal behavior and
differences between individual and legal
definitions of crime. These variations in
measurement and their effects on esti-
mated participation rates are a central
concern in the following discussion of
empirical research on participation.
STUDY DESIGNS AND
PARTICIPATION ESTIMATES
Criminal participation is an incliviclu-
ally basecI phenomenon that typically
cannot be derived from aggregate arrest
statistics. However, early estimates of B
were made using record systems that dis-
tinguish arrests or court referrals of first-
time offenders from those of other of-
fenclers. Although such systems are rare
in American agencies, one exception is
the Philadelphia Juvenile Court. In one
of the first studies of criminal participa-
tion, T. Monahan (1960) combined one
year's age distribution of persons arrested
or referred to court for the first time with
data for the same year on the age clistri-
bution of the general population (see also
Ball, Ross, and Simpson, 1964~. British
agencies routinely keep records of first
offenders and this technique has also
been used to estimate B in Britain (e.g.,
Little, 1965; Farrington, 19811. These
data petit generalizations to contempo-
raneous subpopulations, but they only
33
approximate participation rates in any
single birth cohort represented in the
subpopulation. Recorc! systems in most
jurisdictions, moreover, do not collect the
data needec! to study nondemographic
correlates of participation; these data
must be collected through interviews or
searches of other record systems for incli-
viduals being stuclied in cross-section
samples or being tracker] longitudinally.
Studies baser! on indivicluals' self-re-
ports do permit study of the relationship
between criminal participation (D and d)
ant! a broacler range of characteristics.
Data are gathered from cross-section or
prospective longituclinal studies of sam-
ples of persons representing some popu-
lation of interest. Respondents are asked
if they have committed specific criminal
or delinquent acts, either ever (D) or
during some definer] prececling interval
(ct), and many other data are also col-
lectecI. This method can generate cle-
tailec! information on crime-specific par-
ticination and its relationship to family
structure, group membership, other be-
haviors, ant] a wicle variety of socioeco-
nomic characteristics. But measuring
criminal involvement through self-re-
ports can introduce several sources of
error. For example, underrepresentation
of offenders is a problem in self-report
surveys because youths who are at a
higher risk of criminal involvement are
often school dropouts or truants or other-
wise difficult to locate. Some evidence
also suggests that underreporting and
poor recall of crimes is common in self-
reports of criminal activity. However,
some respondents may exaggerate their
criminal involvement. Differences in
question wording ant! method of admin-
istration also affect self-reporte(1 partici-
pation rates in ways that make it difficult
to compare results across studies.
O~cial-record estimates of participa-
tion (B ant! b) are computed by using a
sample of inclivicluals selected on some
OCR for page 34
34
basis other than status as an offender,
such as a random sample of a population
or of participants in a prior study, perhaps
unrelatecI to crime. Searches of official
records have been carried out in conjunc-
tion with prospective longitudinal or
cross-section studies of demographic and
social characteristics hypothesized to be
relate c] to participation. Some researchers
have gathered both offlcial-record and
self-report data on the same individuals
because of interest in comparing D with
B (e.g., Gold, 1966; Hindelang, Hirschi,
and Weis, 1981; Elliott et al., 19831.
Because juvenile and adult records are
generally maintained in separate reposi-
tories, researchers concerned with cumu-
lative participation estimates for adults-
in particular, the lifetime risk of participa-
tion, BL. have generally had to combine
age distributions of participation rates
from multiple sources and even multiple
jurisdictions (Christensen, 1967; Belkin,
Blumstein, and Glass, 1973; Blumstein
and Graddy, 19821. To the extent that
juvenile and adult participation behavior
is similar across jurisdictions, this ap-
proach can yield! reliable estimates of
lifetime participation for different sub-
populations. But since subpopulation
composition may vary across jurisdic-
tions, it is important that these subpop-
ulations, which may differ systematically
in terms of their participation (e.g., sex or
race categories), be analyzed separately.
Another concern in combining jurisdic-
tions is that arrest recording thresholds
and definitions may differ across the juris-
dictions.
Official records are becoming more
widely accessible as a source of sequen-
tial information about involvement in
crime. However, since only a fraction (q)
of all criminal acts lead to arrests that are
recorded by the criminal justice system,
official-record estimates of participation
wit] always be Tower than actual levels of
CRIMINAL CAREERS AND CAREER CRIMINALS
involvement. In aciclition, crimes re-
ported in an individuaT's official record
may not be a representative sample of all
crimes committed by that offender.
Crimes of specific types (e.g., robbery
relative to burglary) or of specific subpop-
ulations (e.g., intoxicated offenders and
inexperienced offenders) may be more
likely to lead to arrest than others,
thereby distorting comparisons of partici-
pation across different groups. In addi-
tion, police patrolling or investigation pri-
orities defined in terms of location or
crime type may also lead to over- or un-
clerrepresentation of particular subpop-
ulations in official arrest records (see Wil-
son, 1968; Black and Reiss, 1970; Black,
1971; Smith and Visher, 19811. Informa-
tion in arrest records, however, can be
combined with other data to infer esti-
mates about the likelihood of arrest for a
given criminal event and for different
subpopulations of offenders.
A crime victim, in reporting an event to
the police, has an important role in gen-
erating an official arrest record. Reporting
probabilities are known to differ by char-
acteristics of the crime, the offender, and
the victim (Hindelang, 1978b; Bureau of
Justice Statistics, 1984a). Also, some re-
ported crimes are "unfounded" which
usually means that the crime is not re-
corded- because the victim refuses to
follow through with a complaint or be-
cause police conclude that the reported
event does not constitute a crime.
The choice of data source also affects
the type of criminal behaviors that are
represented in any estimate of participa-
tion. Most self-report studies do not have
large enough samples to reliably measure
D and, especially, ct for serious but less
common offenses, such as burglary, large
thefts, and robbery. Thus, self-report
studies usually cover a much broa(ler set
of criminal or deviant acts, frequently
including school-relate(1 infractions and
OCR for page 35
PARTICIPATION IN CRIMINAL CAREERS
minor delinquencies (e.g., fighting, tres-
passing, vandalism, alcohol use) in which
a large proportion of all youth are in-
volved. Official records generally include
only behaviors that are serious enough for
a crime to be reported to police ant!
thereby brought to the attention of the
juvenile or criminal justice system. Even
when official records ant! self-reports
capture the identical set of events, dis-
crepancies are likely to exist because of
differences between legal and public def
. . r
Unctions or crimes.
PARTICIPATION AMONG MALES
Participation in offending is more often
estimated for males than females. Female
participation receives less research atten-
tion because female offenders account for
a much smaller share of crime, as mea-
surec! by official statistics on total arrests.
Because of the large amount of research
on mates, the pane] used estimates of D
and B for mates as a "baseline" for com-
parison with other subpopulations. Crim-
~na~ paruc~panon among males has been
estimated using the methods described
above for vastly different samples and
across several decades. Although few
studies have attempted to produce na-
tional estimates of participation, consis-
tent findings across jurisdictions increase
the confidence with which generaliza-
tions can be made from local studies to
subgroups regardless of their location.
Our discussion focuses primarily on mate
participation before the 18th birthday,
Bit, and lifetime participation, BI, as de-
rived from official-record estimates in
published English-language research.
Other measures of participation-current
participation, c! (e.g., within the year pre-
ceding the interview) and cumulative
participation by high school age, DHS
which are generally derived from self-
reported data are also considered (for
35
more detail, see Visher and Roth, Appen-
dix A).3
O~cial-Record Estimates
Wolfgang, Figlio, and Sellin (1972) re-
ported that 35 percent of a sample of
mates born in 1945 and residing in the
city of Philaclelphia from ages 10 through
18 had at least one recorded police con-
tact for a nontragic offense by their lSth
birthday (see Table 2-11.4 This estimate
has become the most widely cited esti-
mate of criminal participation among U.S.
males. Since not all juvenile police con-
tacts result in court referral, it is not sur-
prising that analyses of Philadelphia juve-
nile court statistics collected at about the
same time (mid 1950s to mid 1960s) yield
slightly smaller estimates of Bit. Those
analyses found that 25 percent of mates
had a juvenile court record by age 18
(reanalysis of T. Monahan, 1960, by Gor-
don ant! Gleser, 1974~. Findings in Ken-
tucky and, more recently, in Oregon are
consistent with the Philadelphia court
statistics: by age 18, 21 percent (Ball,
Ross, and Simpson, 1964) and 25 percent
(Polk et al., 1981) of males had juvenile
court records.
Table 2-1 presents estimates of partici-
pation by 18 (Bit) for all ofEcial-record
studies reviewed by the panel. While
some of the estimates differ substantially
3Not all types of data exist for these four measures
of criminal participation. In fact, estimates of partic-
ipation by age 18 and in a lifetime are exclusively
based on studies with official-record data, with one
exception (Porterfield, 1946~. In contrast, active par-
ticipation and participation by high school age are
usually estimated from self-reported data (a few
studies report BHS) Estimates of Dis, Do, and b do
not generally appear in the published literature.
4"Nontraffic offenses" in the Philadelphia study
encompass a very broad range of reasons for police
contact, including many nonserious charges such as
disorderly conduct, liquor violations, drunkenness,
and investigation.
OCR for page 36
Crime Type and Level of Involvement
.
U.S. Studies
Known to police or juvenile court
Nontragic arrest or police contact
36
CRIMINAL CAREERS AND CAREER CRIMINALS
TABLE 2-1 Estimates of Male Participation Rates by Age 18 (official record data)
Participation Rates (percent)
White Black
All Males
52
23, 29
Juvenile arrest or court referral
Juvenile court conviction
Arrest for index offense
British Studies
Conviction or police caution, indict-
able offenses
Conviction for indictable offense
64
42, 50
13, 17, 18
8, 8, 9
31, 38, 38,
43, 50
26, 27, 27
56
25, 33, 34, 35,
41, 44, 47
17, 21, 25, 26
26, 28
12, 14, 18
15, 28, 29
12, 15, 16, 17,
17, 26
NOTE: See Table 1 and discussion in Visher and Roth (Appendix A) for sources and analysis of these
alternative estimates.
from Wolfgang et al.'s estimates, much of
the variation can be explained in straight-
forward ways. Differences in the set of
offenses included, in the threshoIc3 of in-
volvement (e.g., police contact, arrest, or
conviction), or in the composition of the
base population lead to considerable vari-
ation across studies. A broad definition of
criminal activity that includes clisorclerly
conduct or suspicious behavior generates
higher estimates of Bit, which may ex-
plain the high average estimates of 56
percent reported by Palmore and Ham-
mond (1964) for a high-risk sample in
New Haven, Connecticut. Similarly, the
difference between Shannon's (1982a) es-
timates of BE for nontragic contacts (41,
44, and 47 percent) and those from the
Philadelphia cohorts (33 and 35 percent)
is likely due to the fact that in the Shan-
non stucly consiclerable emphasis was
placed on recording police contacts for
minor infractions of the law as well as for
status offenses (L. W. Shannon, 1985, per-
sonal communication).
When the measure of participation is
restricted to arrests or contacts for index
offenses, BE drops to 14 percent in the
1945 Philadelphia cohort and to 17 per-
cent in Shannon's 1942 cohort (L. W.
Shannon, 1985, personal communica-
tion). Thus, a narrower seriousness defi-
nition of criminal involvement substan-
tially lowers Bit: according to four
estimates, about 15 percent of all males
participate in serious criminal activity,
defined as an arrest for an index offense.
Estimates of participation for a specific
time period and location are also vuIner-
able to changes over time in the scope of
crimes and in population composition.
For example, analysis of delinquency in a
second Philadelphia cohort, born in 1958,
was initially interpreted as implying that
BE was almost identical with BE for
mates born 13 years earlier (for the same
offense definition and threshold of in-
volvement). In fact, participation rates ac-
tually decreased for both whites (from 29
to 23 percent) and blacks (from 50 to 41
percent); however, because of the large
increase in the black population between
1945 and 1958, the overall rate appeared
relatively unchanged. One possible ex-
planation contributing to the race-specific
declines in participation is that the range
OCR for page 37
PARTICIPATION IN CRIMINAL CAREERS
of crimes that prompted police to stop
suspects or make arrests may have nar-
rowec] cluring this interval. If police in
Philadelphia began to focus on more
serious crime types in the 1960s, the
participation estimates could have cle-
cIined.
Other Philadelphia data and another
cohort study do show an increase in par-
ticipation rates for serious crimes over the
period (Shannon, 1982a; Tracy, Wolf-
gang, and Figlio, 19851. For all Philadel-
phia mates, Bit for violent index offenses
(murder, rape, aggravated assault, rob-
bery) increased from 4 to 10 percent. A
similar pattern emerges from Shannon's
cohort data: cumulative participation es-
timates, based on at least one felony ar-
rest by age 18, were 9 percent, lo percent,
and 15 percent for males in the 1942,
1949, and 1955 cohorts, respectively.
However, additional analyses are needed
to partition this apparent increase in par-
ticipation among effects of changes in
population composition, of changes in
participation behavior across successive
cohorts in any subpopulation, and of
changes in police arrest and recor~keep-
ing practices.
Estimates of lifetime participation, Be,
are considerably more scarce-16 esti-
mates, 10 of which are from two studies-
and subject to some methodological con
37
cems rooted in the imprecision of mortal-
ity adjustments and in the nature of the
samples (see Table 2-21. In general, about
half of all mates have been projected to
have at least one police arrest for nontraf-
fic offenses by ages 40 to 50 (Robins,
1966; Christensen, 1967; Belkin, Blum-
stein, and Glass, 1973) and perhaps as
many as 25 to 35 percent of U.S. mates
will have arrests for an index offense
(Robins; 1966; Blumstein and GracIdy,
1982~.
As the threshold of criminal involve-
ment narrows to a conviction record, esti-
mates of BE would be expected to drop.
McCord (1979) reports that about 30 per-
cent of a Massachusetts sample of mates
treated for childhood problems in the
1930s had a conviction record for any
offense by age 50. A British projection
based on aggregate statistics (Farrington,
1981) also concludes that 44 percent of
British mates wit! have a conviction
record] in their lifetime. These surpris-
ingly high participation estimates reflect
particular aspects of the studies' designs:
broad definitions of crime were used,
including serious misdemeanors (Far-
rington) and disorclerly conduct (Mc-
Cord); McCord's sample is generally ac-
knowledged to have been at high risk of
criminal participation; and the very high
probability of conviction following arrest
TABLE 2-2 Estimates of Male Lifetime Participation Rates (official record data)
Participation Rates (percent)
White Black
Crime Type and Level of Involvement
Nontragic arrest
Arrest for index offense
Conviction for serious offenses (unspecified)
Conviction for property offense
Conviction for violent offense
All Males
47 68
15 52
50, 60, 60
23, 36
6, 29, 30, 35, 44
26
17
NOTE: See Table 2 and discussion in Visher and Roth (Appendix A) for sources and analysis of these
alternative estimates.
aThis anomalous 6 percent applies to the adoptive fathers in Mednick, Gabrielli, and Hutchings (1984),
who were presumably screened for absence of criminal records before being permitted to adopt.
OCR for page 38
38
in Britain makes the arrest and conviction
thresholds more comparable there than in
the United States.
Self-Report Estimates
Altemative estimates of cumulative
participation in crime (D) are available
from self-report studies of specific crimi-
nal activities (e.g., theft over $50, bur-
glary, robbery) by cross-section samples
of youth, usually those enrolled in high
school. Since several grade levels are typ-
ically sampled, the reported measure is
interpreted as participation by an age
range (e.g., 13 to 18 years), or more sim-
ply, by "high school age," which we de-
note as DHS. Considerable imprecision
surrounds this measure since both D and
vary strongly with age during the teen-
age years. Nevertheless, the literature on
these estimates provides much of the
available information on relative crime-
specific participation rates, and general
consistency emerges from 10 studies con-
ducted over the last 25 years.5
From these studies, a systematic pat-
tem emerges in criminal participation
across crime types. The two studies based
on self-reports of high-school-age mates
during the 1950s and 1960s report that 13
percent have stolen autos and 5 to 6
percent admit having stolen items other
than autos worm more than $50 (Short
and Nye, 1958; Jensen and Eve, 19761. In
more recent studies conducted in the
1970s ant! 1980s, DHS for non-auto thefts
over $50 has increased to around 15 per-
cent, and similar recent estimates for auto
sMany other studies have obtained self-reports of
delinquent involvement among samples of high
school youths, but the panel chose to focus on the
subset of studies that reported involvement in seri-
ous crime types for which an adult could be ar-
rested. In addition, many self-report studies of
delinquency only report aggregate frequency mea-
sures or composite scales and not crime-specific
individual participation rates.
CRIMINAL CAREERS AND CAREER CRIMINALS
theft have decreased to 10 percent. These
changes in crime-specific participation
likely reflect Me effects of inflation and
enhancer] security features on newer au-
tos. Burglary appears to be a relatively
common crime among high school mates:
in three different studies, 16 to 20 percent
reported ever committing at least one
burglary. For assault with injury, cumula-
tive participation averages 13 percent in
both the oIcler and more recent studies.
Last, participation rates for robbery are
generally consistent across several stud-
ies, with about 6 percent of high-school-
age mates reporting commission of a rob-
bery, although one study, using a
constructed single cohort, reports DHS for
robbery of 15 percent.6
A few stuclies of high school males also
measure participation rates through both
self-reported and officially recorded in-
volvement with the police or juvenile
court. These estimates vary widely, prob-
ably because of differences in definitions
of crime, in thoroughness of record
searches, in year of survey, and in geo-
~ranhic area. One recent study (Hinde-
lang, Hirschi, and Weis, 1981) permits a
direct comparison of official-record and
self-report estimates of the fraction of a
single sample with police contacts. In
that study, involving a sample of high
school students in Seattle selected in
1978, one-third to one-half of the teenag-
ers did not report their police contacts to
interviewers, if the official records are
accurate; the underreporting was highest
among white females and black mates.
However, the findings from earlier stud-
ies comparing police contacts as mea
6Estimates of cumulative participation by age 17
were calculated from data reported by Elliott and
Huizinga (1984:62) for a synthetic cohort that was
aggregated from three birth cohorts (1959, 1962, and
1965) to get cumulative participation rates at each
age from 11 to 21 for the entire sample. The offense
types used in their analysis are actually broad clus-
ters of crimes, but some comparisons are possible.
OCR for page 39
PARTICIPATION IN CRIMINAL CAREERS
sured by self-report and offlcial-record
data are inconsistent: some show higher
rates on self-report than official-record
data, and some show other differences by
subgroup (see Visher and Roth, Appendix
A:Table 31.
For white mates in the Seattle study,
about 27 percent had an official record of
police contact, which is consistent with
official-record estimates for white males
of 26 percent for BHS in a 1964 Califomia
sample and 29 percent for BIS, in Phila-
clelphia. The self-report estimate of po-
lice contact of 17 percent for the white
males in the Seattle cohort (Hindelang,
Hirschi, and Weis, 1981) is also similar to
that reported in two other studies (Gold,
1966; Williams and Gold, 19721. In con-
trast, only 5 percent of one 1967 national
sample of youths had official police rec-
ords (Williams and Gold, 19721. This lat-
ter estimate may reflect both the difficulty
of obtaining complete official records on a
nationwide sample and Tower participa-
tion rates among the rural youth that were
more heavily represented in that sample.
When participation is measured by the
presence of a juvenile court record that
includes adjudication, the estimates
range from 2 percent for index offenses in
a 1967 national sample to about 11 per-
cent for a broader range of offenses in the
recent Seattle study.
The three cumulative measures dis-
cussed thus far are the most common
statistics on participation reported in the
literature either from official records or
self-reports. But for many purposes, an
annual measure of current participation
by active offenders' 3, is useful. (Esti-
mates of b are not commonly reported in
the literature.) As discussed earlier, c! is
computed as the fraction of some popula-
tion committing at least one crime within
a specified interval, usually one year. Es-
timates of ct are less common in the liter-
ature than estimates of D or B. cumulative
participation.
39
All estimates of ct reported here were
obtained by the self-report method
through questions about involvement in
specific offense types during a specific
preceding reference period. In one study,
reports of ~ were based on a 12-month
reference period; for the other studies,
12-month estimates were interpolates]
from values basest on reference periods of
18 months and 3 years. The panel re-
viewed four studies (Weis, 1976; Bach-
man, O'Malley, and Johnston, 1978; Tit-
tle, 1980; Elliott et al., 1983) that used
similar methods ancl reported ~ for five
serious crime types: grand theft (over
$50), breaking and entering, assault with
a weapon or injury, robbery, and auto
theft. Published statistics were annual-
ized to provide comparable estimates
across all studies.
In any given year, about 3 to 6 percent
ofteenage males steal items worth at least
$50. Slightly more youths, 4 to 7 percent,
report breaking en cl entering a building.
Estimates of ~ for assault with a weapon
(or causing injury) are less consistent
across studies, with a range of 3 percent
for younger adolescent mates (12-14
years) to 13 percent for oIcler males in
high school (16-17 years). The age di~er-
ences and personal definitions of"as-
sault," even with injury or a weapon, are
likely explanations of this variation. Last,
~ for robbery and auto theft among teen-
agers averages 5 percent and 2 percent,
respectively, although some variation ex-
ists between studies and age categories.
Few studies report `1 for aclults, but some
limited data suggest that current partici-
pation declines rapidly in the early 20s.
In summary, estimates of cumulative
participation rates among mates by high
school age, by age 18, and by age 40-50
(lifetime) and current participation rates
are influenced by: (1) the type of data
used (aggregate statistics, official records,
or self-reports); (2) the crimes in which
participation is being measured; (3) level
OCR for page 40
40
or threshold of involvement (police con-
tact, arrest, court referral, or conviction);
and (4) the characteristics and representa-
tiveness of the sample. These factors ac-
count for most of the variation in esti-
mates of participation (see Visher ant!
Roth, Appendix A, Tables 1-4, for more
details).
CRIMINAL CAREERS AND CAREER CRIMINALS
crime type, level of involvement, or mea-
sure of participation.
With few exceptions, the ratio of male-
to-female participation in U.S. samples
ranges from less than 3:1 for broadly de-
fined offense categories to 5:1 for index
offenses and specific crime types. In the
1958 Philadelphia cohort study, which
included females, 14 percent of females
had a recorder] nontragic police contact
by age 18, as compared with 31 percent of
males. In the three Racine, Wisconsin,
cohorts, the average BE for males (with
crime types broadly cleaned) was about
45 percent; for females, BE averaged 19
percent (Shannon, 1982a). Juvenile court
referral estimates of B. which average 25
percent for males, range from 5 to 14
percent for females in four studies. Life-
time participation estimates for females
converge at 15 percent, compared with
BE for males of 50 percent based on any
nontragic offense. For specific crime
types, corresponding male/female ratios
of DHS and c! vary more across studies as a
result of low reported participation values
based on self-reports.
PARTICIPATION ESTIMATES BY SEX,
RACE, AND AGE
Differences in participation across de-
mographic subgroups~efined in terms
of sex, race, and age, are discussec! be-
cause much available data on participa-
tion are reported in these terms. Further-
more, the often substantial bivariate
relationships are robust in multivariate
analyses ant! stable across a variety of
temporal and geographic settings and
data collection methods. The policy im-
plications of these empirical associations
ant] those of other covariates are dis-
cussed in later chapters.
Sex
Some of the research that reports par-
ticipation rates for males also reports
comparable measures for females. For fe-
males, 16 estimates of Bit, 5 estimates of
By, and 11 estimates of DHS were located
in the literature reviewed by the panel;
current participation, 4, among females in
specific crime types was also reported in
three studies. In general, patterns of par-
ticipation among females parallel those
among males: higher estimates for broad
crime domains (e.g., all nontragic com-
pared with index offenses) and low
thresholds of involvement (e.g., police
contacts or arrests compared with convic-
tions). The most consistent pattern with
respect to gender is the extent to which
mate criminal participation in serious
crimes at any age greatly exceeds that of
females, regardless of source of data,
Race
Racial differences in criminal participa-
tion have been examined with both self-
report and official-record data for cumula-
tive participation measures, Big, BL., and
DHS, and for current participation, ct.
These differences reflect a number of
social and economic factors that are cor-
related with both ethnicity and participa-
tion. Self-report studies that report partic-
ipation rates separately for white and
black male subgroups (e.g., Gold, 1966;
Hirschi, 1969; Hindelang, Hirschi, and
Weis, 1981; Elliott et al., 1983) generally
find negligible differences for primarily
minor delinquent acts. The few self-
report studies that permit race compari-
sons for serious offenses show that the
estimated ratio of black/white participa
OCR for page 41
PARTICIPATION IN CRIMINAL CAREERS
tion is only slightly above 1:1 for self-
reported serious property offenses such
as grand theft, auto theft, and burglary
(see Hindelang, Hirschi, and Weis, 1981;
Visher and Roth, Appendix A:Tables 3
and 41. For self-reports of robbery and
assault ant! for scales limited to index
offenses, this ratio increases, especially at
younger ages. For example, in the Na-
tional Youth Survey, annual current par-
ticipation rates for robbery, ct. for blacks
ant! whites aged 11-17 were 9 and 4
percent, respectively, for a black/white
ratio of 2.25:1. But in the fib wave of
interviews, when the age range was
1~21, ct for black and white youth in the
same sample was 3 and 2 percent, respec-
tively, for a lesser ratio of only 1.5:1 (El-
liott et al., 1983; see discussion in Visher
and Roth, Appendix A:Table 4~.
Studies using official data generally re-
port greater black;/white ratios and stron-
ger associations between race and partic-
ipation in crime as the seriousness of
criminal behavior increases from nontraf-
fic offenses to all inclex offenses to violent
index offenses. Combining data from sev-
eral studies with criminal participation
broaclly defined as nontragic offenses, the
black/white ratio averages 1.8:1; for index
offenses, the ratio averages 3.2:1 (Visher
and Roth, Appendix A:Table 81.
The Philadelphia cohort data are the
only source of offlcial-record participation
estimates by racial group across several
levels of offense seriousness (Tracy,
Wolfgang, and Figlio, 1985:Tables 4a and
5a). For the cohort born in 1945, the
black/white ratios for Bit were 1.8:1 for
nontragic offenses, 3.3:1 for UCR index
crimes, and 4.5:1 for offenses with injury.
For the cohort born in 195S, the ratios
were very similar: 1.8:1, 2.9:1, and 3.2:1
for nontragic offenses, UCR index crimes,
and offenses with injury, respectively.
Analysis of all the data shows that racial
differences in aggregate measures of
criminal behavior appear to be largely a
41
function of differences in participation
rather than offending frequencies. For
example, the 1980 black/white ratios for
arrest rates of males uncler age 18 were
1.6:1, 2.4:1, and 3.4:1 for nontragic of-
fenses, all inclex offenses, and violent in-
clex offenses, respectively.7 These ratios
are very consistent with the participation
ratios just presenter! for the Philaclelphia
cohort born in 1958. Since aggregate ar-
rest rates are a function of both participa-
tion rates and indiviclual frequency rates
(A= ,ub), the similarity in the black/white
ratios suggests that racial differences in
criminal behavior are largely a function of
differences in participation. More black
youth are arrested at least once, but black
and white offenders tend to be rearrested!
at similar rates (see also Blumstein ant]
Bradley, 1982; Tracy, Wolfgang, and
Figlio, 1985).
Age
The relationship between age and par-
ticipation in criminal behavior can be
examined using several measures. Com-
mon measures are age-specific values of cl
and of the hazard rate for initiation (h), as
discussed above. Often, hazard rates can
be calculated from official records (e.g.,
Wolfgang, Figlio, and Sellin, 1972; Far-
rington, 1983a). Age distributions of cur-
rent participation rates, data, are usually
generated by studies that gather multiple
waves of self-reportecl data on criminal
involvement during recent intervals (e.g.,
Elliott et al., 1983~.
Some patterns emerge from the studies
that the pane] reviewed. Even though
only a small fraction of youth at risk begin
criminal careers at any given age, a con-
centration of initiations among youth is
7These ratios were computed from 1980 race-
specific arrest data (Federal Bureau of Investigation,
1981) and 1980 population figures (Bureau of the
Census, 1983:Table 33~.
OCR for page 44
OCR for page 46
OCR for page 47
OCR for page 48
OCR for page 49
OCR for page 50
OCR for page 51
OCR for page 52
OCR for page 53
OCR for page 54
Representative terms from entire chapter:
criminal behavior
44
One reason for this result may be that a
single teacher's or parent's rating of child-
rearing practices may be influenced by
various subjective factors that affect the
generalizabiTity of the findings. Another
explanation might be that the presence of
multiple risk factors within a family (such
as poor supervision, poor cliscipline) pre-
dicts delinquency better than single risk
factors.
Recent studies of the effects of parent-
ing techniques on pre-aclolescent con-
duct disturbances as precursors to later
criminal behavior have also pointer! to
the adverse effects of parents' lax super-
vision, passive attitudes, ant! ineffective
cliscipline.~° In adclition, cross-section
research has also founcl correlations be-
tween some family management vari-
ables principally discipline and super-
vision ant! criminal involvement, but
many of these studies do not report crim-
inal participation rates (e.g., Nye, 1958;
Hirschi, 1969; Patterson and Stouthamer-
Loeber, 1984~.
Parental Criminality
A factor consistently associated with
serious delinquency and aclult criminal
participation in many different studies is
criminal behavior of parents. In one
study, for example, twice as many chil-
ciren at age 10 with a convicted parent
had a criminal conviction by age 25 as
those without a convicted] parent (Far-
rington, 1933a; see also Osborn and West,
19791. A multivariate analysis of these
same clata shower! that having a convicted
parent was a significant predictor of juve-
nile convictions and self-reportec! delin-
quency. This relationship persists in
i°Recent discussions of this topic can be found in
Hirschi (1983), Patterson (1982), and Butter and
Giller (1984:18~188~. For reviews of studies relat-
ing nondelinquent but disturbing antisocial conduct
to family functioning, see Butter (1977), Hinde
(1980), or Patterson (1980~.
CRIMINAL CAREERS AND CAREER CRIMINALS
studies of children whose parents, or
even grandparents, have a history of juve-
nile delinquency, arrest, or other antiso-
cial behavior, such as excessive drinking
or a poor work record (see Robins and
Lewis, 1966; Robins, West, and Herjanic,
1975).
rental criminality and participation
their children in delinquency is sup-
norted bv studies that suggest that some
offender characteristics (e.g., low arousal
levels of the autonomic nervous system,
certain EEG patterns, low intelligence)
may arise from a mix of inherited ancl
environmental factors. The strongest evi-
dence for a partially genetic origin
emerges from several studies concluctec!
in Scandinavian countries, showing that
adopted chil
PARTIClPaTION IN CRIMINAL CAREERS
Likely environmental explanations for
the observed empirical association in un-
broken families are that children who are
raised in homes with criminal parents are
exposed to aggressive behavior and anti-
social attitudes (Rutter and Giller, 1984),
poor child-rearing practices (Wilson,
1975), and adverse family conditions as-
sociatec! with lower-cIass upbringing
(Van Dusen et al., 1983~. Because the
relationship exists even with respect to
parental delinquency occurring before
the child's bird, it is apparently not a
consequence of direct involvement in the
parents' criminal activity (Farrington,
Gundry, and West, 1975~. Last, parental
antisocial behavior, including unstable
work patterns, alcohol ant] drug use, and a
record of delinquency, also appear to be
significant in the genesis of serious delin-
quent behavior in their children.
Family Disruption
The impact of family disruption, espe-
cially divorce and separation, on the par-
ticipation of youth in criminal behavior
has been a popular topic in studies of
juvenile delinquency (for reviews, see
Hennessey, Richards, and Berk, 1978;
across studies in the base rate of delinquents and
the selection ratio used by the researchers (for
further discussion of RIOC, see Chapter 4 in this
volume and Copas and Tarling in Volume II). Eight
different predictors were examined for their associ-
ation with male delinquency: composite measures
of parenting; child problem behavior; stealing, ly-
ing, or truancy; criminality or antisocial behavior of
family members; poor educational achievement;
single measures of parenting; separation from par-
ents; and socioeconomic status. Loeber and Dishion
found that the median RIOC for these predictors
(averaged over studies with data on a specific pre-
dictor) followed the above order. For parental crim-
inality, the median RIOC was .24 with a range of 1.0
to .08. For example, the Hutchings and Mednick
(1975) study of adoptees was used to calculate three
RIOC measures: .11, .13, and .20; in addition, the
false-positive rates for this study exceeded 70 per-
cent.
45
Wilkinson, 19801. Family disruptions are
consistently related to a record of convic-
tion (Wadsworth, 1979; Farrington,
1983a) and to self-reported commission of
index offenses or serious offense types
(Farrington, 1979b; Elliott et al., 1983),
with participation rates almost twice as
high among youth from disrupted fami-
lies as those from intact families.
However, self-report studies that use
nonserious offenses or other measures of
criminal behavior besides participation
report small and inconsistent associations
between family disruption and delin-
quency, perhaps because such minor acts
are so widespreacI. For example, one
study found a slight effect of family dis-
ruption on a six-item scale of self-report-
ed criminal activity that included prop-
erty damage and theft uncler $2 (Hirsch),
1969; see also Nye, 1958; DentIer and
Monroe, 1961; Berger and Simon, 1974~.
It appears that "broken homes" may dif-
ferentially affect mates and females, ur-
ban and rural youth, and different ethnic
groups (see Datesman and Scarpitti,
1975; Austin, 1978; Wilkinson, 1980), ant!
these variables conic] account for some of
the inconsistent results in research on
family disruption and criminal participa-
tion. There is also some evidence that
family disruption has a selective impact
on delinquency, primarily affecting rebel-
lious behavior such as running away and
truancy.
In any case, it does not appear that
family disruption per se is associated with
clelinquency. Rather, parental discord,
which usually precedes a break-up, or
other difficulties that follow a break-up
apparently impair normal family func-
tioning and increase the risk of clelin-
quency. Marital conflict within intact
homes is associated with participation in
delinquency, perhaps to a greater extent
than broken homes (Rutter, 1981; Mc-
Corcl, 19821. Evidence that parental death
is unrelated to criminal involvement
46
while parental divorce increases delin-
quency suggests that conflict between
parents may account for the association
between family disruption and delin-
quent behavior (see Butter, 1981; see
also, West and Farrington, 1973; Wads-
wor~, 1979; but see Rankin, 19831. When
father absence was broadly defined to
include death and other separations unre-
lated to divorce in two other studies, it
was not a significant predictor of adult
criminal behavior when multivariate
analysis was used to sort out the impor-
tance of several variables (Farrington,
1983a). And, a recent empirical review of
several studies found that separation from
parents at an early age was only weakly
predictive of delinquency as an adoles-
cent (Loeber and Dishion, 19831.
A composite measure of parental be-
havior that includes parental conflict,
however, does predict adult criminal par-
ticipation (Farrington, 1983a) and juve-
nile convictions (McCord, 19791. As the
operative factor, parental discord proba-
bly reduces the effectiveness of parental
supervision and discipline, interferes
with supportive parent-child relation-
ships, and offers negative, antisocial mod-
els for a child. Moreover, the association
between family conflict and criminal par-
ticipation among children is probably a
complex relationship affecting some
youth but not others, depending on other
family variables and the youth's charac-
teristics.
Family Size ant] Structure
Family size appears related to delin-
quency, but explanations for this finding
are not yet established. Large families-
defined as those with at least four chil-
dren may handicap effective parenting
and result in increased delinquency par-
ticipation by children. However, the fam-
ily size/delinquency relationship does
not appear in a few studies of children
CRIMINAL CAREERS AND CAREER CRIMINALS
from large middle-class families, suggest-
ing that the pertinent explanatory vari-
able is the "disadvantages which tend to
accompany family size in poorer sections
of the community" (Rutter and Giller,
1984: 186) and not simply family size. For
example, the decreased control of chil-
dren because of lack of resources for child
care and overcrowding in homes are plau-
sible intervening explanatory factors (see
Hirschi, 1983; Butter and Giller, 19841. In
Farrington's (1983a) multivariate analy-
sis, family size was not an independent
predictor of criminal involvement in any
of the four age groups examined; how-
ever, associated variables, such as low
family income and poor housing, were
significant predictors of delinquency by
ages ~10 and 1~13.
Other family structure characteristics
that may increase participation through
their influence on parenting skills in-
clude short intervals between first and
second birds, continued growth in family
size after a child reaches age 6, younger
mothers (under age 20), and short inter-
vals between marriage and the first
child's birth, especially less than 1 year
(see Wadsworth, 1979~. But these types of
variables, including family size, have not
been extensively studied with U.S. data
(for an exception, see Hirschi, 1969~.
Early Antisocial Behavior
Using various measures of antisocial
behaviors in the elementary school or
preadolescent years, many investigators
have examined the association of early
antisocial behavior with later serious de-
linquency and adult criminal activity.
With few exceptions, the evidence indi-
cates that such conduct as aggressive be-
havior, "troublesomeness" in school, dis-
honesty, and stealing are all related to a
record of arrest or court conviction in later
years (see Loeber and Dishion, 1983~.
For example, in one study, 62 percent of
PARTlCIPaTION IN CRIMINAL CAREERS
children who were rated by teachers,
peers, or parents as troublesome at ages
~10 had court convictions by age 25,
compared with 26 percent of other chil-
dren (Farrington, 1983a; see also
Feldhusen, Thurston, and Benning,
19731. In adclition, the pane} found that
more general measures of early antisocial
behavior-referral to a chiTct guidance
clinic, teacher rating as "potentially de
. ,, `` . . ,,
. .~nquent or ant~soc~a , anc . excessive
elementary school absences were simi-
larly associated with criminal activity at
oIcler ages (Reckless, Dinitz, and Kay,
1957; Robins, 1966; Robins and Wish,
1977; Ouston, 19841.~2
Further evidence of a strong associa-
tion between preadolescent antisocial be-
havior and delinquency appears in some
research that has examined the impor-
tance of various categories of variables in
predicting delinquency. In a review of
more than 20 studies, child problem be-
havior including stealing, Tying, and tru-
ancy- was second only to poor family
management practices as a predictor of
later juvenile and adult criminal involve-
ment (Loeber and Dishion, 19831. Using
several different techniques, Robins
(1966; Robins and Wish, 1977) concluded
that childhood deviance and adult antiso-
cial behavior represented different stages
of a single developmental process and
that childhood behaviors were better pre-
clictors than family variables. Last, Far-
rington (1983a) also supports the view
that criminal behavior is part of a devel-
opmental sequence usually initiated by
troublesome, clefiant, or aggressive be-
havior in children before the age of 12. In
i2A large literature outside the scope of the pan-
el's focus on criminal careers addresses the preva-
lence of antisocial behaviors (e.g., fighting, lying,
stealing) in elementary school children and the
evidence of continuities between these behaviors
and various psychosocial problems in adulthood
(Olweus, 1979; Loeber, 1982; Loeber and Dishion,
1983; Butter and Giller, 1984:Chapter 2~.
47
his data, family influences and the child's
antisocial behavior were equally impor-
tant in predicting conviction by age 25.
Despite the apparent consistency of Me
relationship between early antisocial be-
havior and subsequent criminal participa-
tion, data are lacking on several signifi-
cant points. First, while continuities
between antisocial behavior and partici-
pation in crime have been fount! in stud-
ies of older children (ages ~14) followed
into adolescence or early adulthood, vir-
tually no studies have followed children
from ages 4 to 7, when antisocial behavior
tends to begin, into young adulthood,
when the risk of criminal career initiation
has largely passed. Therefore, under-
standing of the relationship between
early behavior and criminal participation
is incomplete. Second, most children
who exhibit antisocial behaviors do not
become involves! in criminal activity as
teenagers or adults. There is little knowI-
edge of the factors that reliably identify
antisocial preadolescents who do not
progress to offending patterns involving
serious crimes. Furthermore, there is ev-
idence that suggests it is even more diffi-
cult to preclict eventual serious criminal
behavior among persons who first be-
come offenders in young adulthood.
Social Class
Social class is often discussed as an
important correlate of criminal behavior,
although there is substantial debate over
the consistency of the relationship be-
tween Tow social class anc! involvement
in criminal activity and about its underly-
ing meaning. Most of the recent U.S.
studies of this cIass/crime association
have relied on self-report data, which tap
nonserious criminal activity that is rela-
tively common and does not necessarily
attract justice system intervention.
In the panel's review of the research
relating participation rates to individual
48
level measures of social class, no signifi-
cant association emerges in self-report
studies when the measure of criminal
activity is primarily nonserious behavior,
such as vandalism, fighting, or small
chews. This result probably reflects He
fact Hat a large fraction of youths engage
in that behavior regardless of Heir social
cIass.~3 Also, in studies of the class/
participation relationship in which mea-
sures of official involvement (e.g., police
contact, arrest record) encompass a wide
range of delinquent behaviors, no associ-
ation is usually found (e.g., Hathaway and
Monachesi, 1963; Polk, Frease, and Rich-
mond, 1974; Hindelang, Hirschi, and
Weis, 1981~.~4
When criminal participation is re-
stricted to serious offenses, social class
differences emerge in one major study
Hat user! self-report measures (Elliott
and Huizinga, 1983), and this association
persists after convolving for race. Win
official data (e.g., juvenile court or police
recorcis), He cIass/serious crime relation-
ship is fairly s~ong, especially in He
pre-1970 studies (e.g., Reiss and Rhodes,
i3Many investigators have examined the class/
crime issue with self-report data (e.g., Nye, 1958;
Akers, 1964; Hirschi, 1969; Williams and Gold,
1972; Johnson, 1980; Krohn et al., 1980~. For a
detailed review of studies considered by the panel,
see Visher and Roth (Appendix A). However, some
studies using other measures of criminal behavior,
principally aggregate "incidence" rates, have re-
ported differences by class group (e.g., Elliott and
Voss, 1974; Hindelang, Hirschi, and Weis,
1981:194; Elliott and Huizinga, 19831. Other impor-
tant reviews on this topic appear in Tittle, Villemez,
and Smith (1978), Hindelang, Hirschi, and Weis
(1979, 1981), and Clelland and Carter (1980~.
i4In the Racine, Wisconsin, cohort study, which
had official data on broadly defined police contacts
(including investigations, suspicion, and informa-
tion-gathering), Shannon (1982a) found social status
to be unrelated to number of police contacts; how-
ever, participation rates by social status could not be
computed from published data.
CRIMINAL CAREERS AND CAREER CRIMINALS
1961; Gold, 1966; for another review, see
Tittle, Villemez, and Smith, 19781.~5 Re-
cent evidence, however, is limited be-
cause individual official records usually
do not include individual-level social
class measures, and arrest statistics are
not clisaggregated by this variable. Some
existing data sets presumably contain of-
ficial-record data that would provide fur-
ther information about the relationship
between participation in serious offenses
and social class (e.g., McCorc3, 1979; Hin-
delang, Hirschi, and Weis, 1981; Shan-
non, 1982a; Thornberry and Farnworth,
1982), but the appropriate analyses have
not been reported in the literature.
The strongest evidence for a negative
association between social class and crim-
inal participation appears in research us-
ing ecological (area) measures of social
class (e.g., Lander, 1954; ChiTton, 1964;
Shaw and McKay, 1969; Wolfgang,
Figlio, and Sellin, 1972~. Indeed, until
1950, ecological correlations were the
only basis for conclusions about social
class and criminal behavior (Hinclelang,
Hirschi, and Weis, 1981: 1841. However,
comparisons between these studies and
those with individual social class indica-
tors must be made cautiously. For one
~5Hindelang, Hirschi, and Weis (1981) refer to
Hirschi (1969) as another study with a moderate
class/official-record relationship (correlation of
-.21), but participation rates by social class are not
reported in the original source. In three British
studies (Wadsworth, 1979; Farrington, 1983a;
Ouston, 1984) and one study with data from Den-
mark (Van Dusen et al., 1983) the class/crime rela-
tionship is particularly strong.
i6Using regression analysis and follow-up data
from the 1945 Philadelphia cohort initially collected
by Wolfgang, Figlio, and Sellin (1972), Thornberry
and Farnworth (1982) found a strong association
between individual social status and both the fre-
quency of officially recorded serious (e.g., index or
violent arrests) and nonserious (e.g., total arrests)
adult criminal behavior, but participation rates by
social class were not reported.
PARTICIPATION IN CRIMINAL CAREERS
thing, aggregate relationships do not nec-
essariTy permit valid inferences to indi-
viduals (see Robinson, 1950; Hannan,
19711. Urban areas that are cleaned as
lower class may have higher crime rates
because of a small group of very active
criminals or because outsiders come into
the area to commit crime. In addition,
differential police surveillance and arrest
practices in lower-cIass areas may con-
tribute to ecological correlations between
social status and crime. In discussing this
issue, TittIe, Villemez, and Smith (1978)
argued that census tracts a common unit
of analysis in ecological studies are typ-
ically quite diverse with regard to family
income, and especially so during the
1950s and 1960s when most ofthe ecolog-
ical research was carried out.
In summary, individual social class
may be empirically related to some types
of serious clelinquency and adult criminal
behavior when participation is measured
either by self-reports or official records,
but relevant research is limited, espe-
cially that based on recent samples. In a
review of socioeconomic class as a preclic-
tor of a chiefs later clelinquency in seven
studies, Loeber and Dishion (1983) con-
cluded that social status was a poor pre-
dictor compared with early antisocial be-
havior or measures of family functioning.
But low social class and low family in-
come at age 14 significantly predicted
convictions at ages 17-20 and 21-24 in a
multivariate analysis that controller] for
some family influences (Farrington,
1983a). The meaning ofthis association is
far from clear, but it may be due in part to
social class differentials in police detec-
tion, official recording of criminal behav-
ior, or victim reporting. Indicators of
social status may also overlap with
unmeasured! aspects of parental behavior
and family structure that are consistently
related to delinquency.
49
School Performance and Intelligence
The relationship between participation
in delinquency or adult criminal activity
and various measures of school perform-
ance or intelligence measured by school
achievement test scores, standard IQ
tests, vocabulary skills, and "school fail-
ure" has been examined in many studies.
In the panel's review of eight studies (see
Visher and Roth, Appendix A:Table 15),
the empirical association between Tow
intelligence or school achievement and
criminal involvement was consistent in a
wide variety of samples and with both
self-report and official measures of delin-
quency.
In general, participation decreases
with higher IQ scores, grade point aver-
ages, reading test scores, and other mea-
sures of academic achievement. Polk,
Frease, and Richmond (1974) reported
that 42 percent of a general high school
sample with low grades had a juvenile
court record compared with 22 and 9
percent of students in the sample with
average and high grapples, respectively. In
some studies, the relationship between
participation and intelligence/achieve-
ment is independent of race (e.g.,
Wolfgang, Figlio, and Sellin, 1972) and
social class (e.g., Reiss and Rhodes, 1961;
Polk, Frease, and Richmond, 19741; it
also exists in research that relates school
achievement or intelligence to other mea-
sures of criminal behavior, such as scales
clerived from factor analysis or aggregate
frequency measures (Weis, 1973; data
from Hirschi, 1969, cited in Hirschi and
Hinclelang, 1977; Menard and Morse,
1984~. In addition, empirical links
emerge between various non(lelinquent
conduct disorders, which may precede
clelinquency, and measures of intelli-
gence (for a review, see Butter and Giller,
19841.
50
While empirical results are consistent
the search for possible explanations ofthe
association between delinquency and
Tow intelligence/achievement continues
(see discussions in Gordon, 1976:256-
270; Hirschi and Hindelang, 1977; Wil-
son and Herrnstein, 1985~. Hirschi and
Hindelang argue that IQ influences de-
linquent behavior primarily through its
correlation with school performance, al-
though the empirical evidence for their
hypothesis is weak ancI is contradicted by
at least one recent study (Menard and
Morse, 19841. Alternatively, low scores on
IQ tests may reflect a predelinquent's
resistance to authority and unwillingness
to make the effort needed to do well on
such tests. The association between
school conduct disorders and IQ is fairly
well establishecI (see Butter and Giller,
1984:163-168), but no study has success-
fully resolvecI the issue of how school
performance, IQ, and early antisocial be-
havior might interact in predicting delin-
quency.
In two multivariate analyses, Tow IQ
emerged as a strong predictor of juvenile
convictions along with poor parenting
and antisocial behavior (Farrington,
1983a), but was not significant in another
study once school variables and social
class were controlled (Wolfgang, Figlio,
and Sellin, 1972~. Thus, what we know at
present is that Tow intelligence and weak
school performance appear to be closely
intertwined, and several studies have
shown them to be strong predictors of
serious delinquency and adult criminal
behavior (Loeber and Dishion, 1983), but
any causal structure among these factors
is not yet established.
Substance Abuse
The relationship between substance
abuse and criminal activity is widely
thought to be firmly established, sup-
portecl by empirical research as well as
CRIMINAL CAREERS AND CAREER CRIMINALS
informal observations of criminal justice
operations. But this relationship is much
more complex than it initially appears,
especially when "criminal activity" is de-
fined in terms of criminal career dimen-
sions. The distinction between participa-
tion in criminal behavior (D) and the
frequency of that behavior (A) is particu-
larly important when discussing sub-
stance abuse. We examine here the rela-
tionship of criminal participation and
substance abuse, that is, whether drug
users are more likely to be involved in
crime than non-users. In the next chapter,
on active offenders, we review the re-
search on whether drug-using criminals
commit crimes more frequently than non-
using criminals (see also Wish and
Johnson, Volume II).
The panel did not consider much of the
research on substance abuse and criminal
participation, which has focused on juve-
niles, their use of marijuana or alcohol,
and their involvement in relatively minor
criminal behavior (e.g., small thefts,
school crime, vandalism) (see Burkett and
lensen, 1975; Wechsler and McFadden,
1976; Kellam, Ensminger, and Simon,
19801. But the available evidence on par-
ticipation in serious criminal activity sug-
gests that drug users, especially multiple
drug users, are much more likely to be
involved than non-users (e.g., Johnston,
O'Malley, and Eveland, 1978; Elliott and
Huizinga, 19841. Table 2-3 presents data
from a national sample of youths aged
11-17 in 1976 and 1~21 in 1980 (Elliott
and Huizinga, 1984~. The self-reportecl
participation rates for felony assault, fel-
ony theft, and robbery increase dramati-
cally as drug use becomes more serious,
from no (lrug use to alcohol only, alcohol
and marijuana, and multiple drugs.
Data of this type, however, cannot an-
swer the frequent question whether sub-
stance abuse leads persons into crime. A
longitudinal study of both criminal in-
volvement and drug use is needed to sort
PARTICIPATION IN CRIMINAL CAREERS
TABLE 2-3 Current Criminal Participation Rates, 4, by Drug User Types
(percent)
51
Crime Typea Drug User Typeb 11-17 yearsC 15-21 yearsC
Felony assault No drugs 12.7 4.1
Alcohol 18.3 5.2
Alcohol/marijuana 33.8 13.5
Multiple drugs 51.7 24.2
Felony theft No drugs 6.3 2.3
Alcohol 18.6 4.4
Alcohol/marijuana 32.4 13.5
Multiple drugs 55.2 27.3
Robbery No drugs 4.0 0.8
Alcohol 5.7 0.4
Alcohol/marijuana 6.8 2.8
Multiple drugs 22.4 6.4
aThese crime types are actually clusters of related offenses. Felony assault is composed of aggravated
assault, sexual assault, and gang fights; felony then includes auto theft, theft over $50, breaking and
entering, and bought stolen goods; robbery encompasses strong-a~ming students, teachers, and others.
bBased on self-reports of drug use in the preceding year. Since the researchers wished to focus on non-
experimental drug use, use of the specific drug at least four times was a definitional requirement.
CActive participation rates for the preceding year, d.
SOURCE: Data from Elliott and Huizinga (1984:Tables 1 and 2), for first and finch waves of
longitudinal study.
Out the causal relationship between sub-
stance use and criminal activity. Table
24 presents such data on youths aged
11-19 from the National Youth Survey
(Elliott and Huizinga, 19841. Although
the measure of criminal activity is not
strictly based on participation, the ciata
are relevant here. In this particular sam-
ple, the predominant pattern among drug
users who are also delinquent was for
initial drug use to follow delinquency or
to occur simultaneously, rather than for
drug use to prececle delinquency. This
study concluded that both substance
abuse and criminal behavior may be a
result of similar social and individual fac-
tors, principally ineffective socialization
in the home, involvement with clelin-
quent peers, and school-related clifficul-
ties (see also Elliott, Huizinga, ant]
Ageton, 19851. In another study (Robins
and Wish, 1977), the median age of initi-
ation for serious alcohol problems ant!
barbiturate ant! amphetamine use (16.3
years) was later than the median age at
initiation for juvenile arrests (15.2 years).
Studies of drug addicts, which show that a
history of criminal behavior preceded ad-
cliction, also appear to support this view
(see Wish and Johnson, Volume II; Rob-
ins, 1979; Friedman ant] Friedman, cited
in Kaplan, 1983~.
Employment
The relationship between unemploy-
ment and participation in criminal activ-
ity has been a source of considerable
controversy. Presumably, a disorganized
life-style is associated with both unstable
employment and involvement in crime,
but the specific causal relationships have
not been fully explored (for one recent
attempt, see Thornberry and Christen-
sen, 19841. In a comprehensive review of
25 studies using aggregate (lata, Freeman
(1982) concludecl that there is no solid
evidence Mat unemployment affects
52
TABLE 2-4 Temporal Order of Drug
Use and Delinquency
Involvement in Drug Use and
Delinquency
Percent of
Total Sample
(1976-1978)
.
No drug usea and no delinquencyb 46.0
Drug use and no delinquency 26.5
Alcohol
Alcohol and marijuana
Alcohol, marijuana, and other
drugs
No drug use and delinquency
Initial drug use before delin
quency involvement 4.4
Alcohol
Alcohol and marijuana
Alcohol, marijuana, and other
drugs
Initial drug use after delinquen
cy involvement
Alcohol
Alcohol and marijuana
Alcohol, marijuana, and other
drugs
Initial drug use and delinquency
involvement occur in same
year
Alcohol
Alcohol and marijuana
Alcohol, marijuana,~and other
drugs
Other, not classifiable
8.9
8.0
3.7
4.3
t0.1'
aDrug use is defined as use of the specific drug
at least four times in the preceding year, or use of
each drug at least four times for multiple drug use
types.
bDelinquency involvement is defined as engag-
ing in 12 or more self-reported delinquent of-
fenses and at least 2 index offenses in the preced-
ing year.
SOURCE: Elliott and Huizinga (1984:Table 9).
criminal behavior (see also Cantor and
Land, 19851. However, the panel's re-
view of research based on individual ciata
did find some consistent patterns (see
Visher and Roth, Appendix A).
In particular, the association of unem-
ployment with criminal participation ap-
pears to be different for adults than for
school-age youths, possibly because un-
employment may have different mean
CRIMINAL CAREERS AND CAREER CRIMINALS
ings for the two age groups. Unemployed
or erratically employed adults are more
likely to be involved in criminal activity
than those with stable employment, and
in one study this relationship persisted
even when controls were introcluced for
criminal history and drug use (Viscusi,
`~8.6) 19831. An unstable work record at age 18
(7 i) was also a significant predictor of a con
viction record at age 21-24, even control
-ling for antisocial and delinquent behav
ior at earlier ages (Farrington, 1983a). In
contrast, full-time employment appears to
influence school-age youths adversely,
`0.6)
t3.9'
(a 5) with higher fractions of employed youths,
LO ~ ~ especially at ages 1~17, than others cur-
rently participating in serious offenses
(Elliott et al., 1983, 19851. Unfortunately,
(4 7) research on the effects of school-age em
(2 i) ployment is scarce, and some studies only
report mean arrest data or seriousness
measures, rather than participation
rates.~7
Taken together, analyses of unemploy-
(2 6) ment and crime suggest that employment
may Inhibit criminal participation In
adult years, while employment during
adolescence may be associated with crim-
inal activity, perhaps because it separates
teenagers from parental supervision and
gives them funds to buy drugs and alco-
hol. The research that shows different
17The studies reviewed by the panel vary in their
measures of criminal behavior (e.g., active partici-
pation, cumulative participation to age 25), and the
respondents in these studies range from a represen-
tative sample in the National Youth Survey (Elliott
et al., 1983), to 2,000 young black men in three
highly urban settings (Viscusi, 1983), to a study of
400 British youngsters followed for more than 15
years (Farrington, 1983a). The study by Viscusi used
a broader definition of criminal activity than the
others: respondents were asked whether they had
participated in any of 10 specific criminal acts in-
cluding the category "any other illegal activities"
(1983:16). Three studies do not report participation
rates by employment status (Bachman, O'Malley,
and Johnston, 1978; Shannon, 1982a; Thornberry
and Christensen, 1984), but their results are consis-
tent with the findings discussed in the text.
PARTICIPATION IN CRIMINAL CAREERS
directions of effects at different ages high-
lights the necessity of partitioning the
population uncler study into separate
groups within which observable factors
have consistent theoretical meanings.
Unless such partitioning is done, the un-
derlying relationships may be masker! by
confounding effects. This possibility is
particularly relevant to the relationship
between employment and criminal be-
havior, and it may also have an impact on
other factors associates! with participa-
tion.
Peer Group Influences
Involvement with delinquent friends is
widely believed to be positively related
to juvenile criminal behavior. Indeed,
this concept lies at the heart of one major
theory of delinquency, differential associ-
ation theory (SutherIancI and Cressey,
19781. Several longitudinal studies report
that association with delinquent friends is
clearly related to participation in serious
criminal behavior at later ages (Far-
rington, 1983a; Polk et al., 19811. In Far-
rington's British study, 59 percent of
males who had extensive involvement
with delinquent friends at age 14 had a
conviction record at age 25, in compari-
son with only 25 percent of other chfl-
dren.
Unfortunately, few empirical tests of
differential association theory present
data on participation in serious criminal
activity and negative peer influences.
Many studies examine marijuana and
drug use because these activities are of-
ten tier! to peer contacts (e.g., Burkett and
Jensen, 1975; Kandel, 1978; Akers et al.,
1979), or focus on a delinquent popula-
tion and their peer contacts (see espe-
cially, Reiss, Volume II; Hinclelang,
1976; Morash, 1984~. Other research has
used multivariate mocleling techniques.
While some of these studies report signif-
icant direct effects of negative peer influ
53
ences on delinquency (e.g., Jensen, 1972;
Matsueda, 1982; Elliott, Huizinga, and
Ageton, 1985), the delinquency measure
is usually a composite index of minor and
moderately serious behaviors (but see
lohnstone, 1978~.
Analyses of peer influences and crimi-
nal behavior indicate that involvement
with many delinquent friends is a signifi-
cant risk factor for participation in delin-
quency. These results are consistent across
a wide variety of samples, measures of
delinquency and peer influences, and esti-
mation techniques. Some recent research
(e.g., Matsueda, 1982; Elliott, Huizinga,
and Ageton, 1985) is attempting to sort out
the underlying causal relationships, includ-
ing the possible mediating effects of paren-
tal supervision and attachment, involve-
ment in conventional activities, and
exposure to conventional attitudes.
SUMMARY
Perhaps the most striking observation
about participation is the high rate at
which males ever become involved in
crime and even in arrest. Typically, about
15 percent of urban mates are arrested for
an index offense by age 18 ant! about 25 to
35 percent will be arrester] for such an
offense sometime in their lifetime. De-
mographic differences in participation, as
measured by arrest, are large, most strik-
ingly between the sexes and less so be-
tween the races.
The differences in participation among
demographic groups vary consiclerably
with offense seriousness. When the clefi-
nition of criminal behavior is broadly de-
fined, participation is widespread in all
demographic groups and so relative dif-
ferences in participation are small. For
serious offenses, for which the base rates
of participation are low, the demographic
differences are considerably larger. This
interaction of demographic variables with
changes in the scope of criminal partici
54
pation resolves only part of the discrep-
ancy between some self-report studies
(which indicate small and inconsistent
black/white differences in "offending"
when the offense threshold is extremely
low) and official-record data for serious
offenses, which show large differences
between blacks and whites.
Of more theoretical and policy signifi-
cance is the fact that research on other
factors associated with participation in
serious offending consistently points to
the same variables that have long been
associated with crime: ineffective parent-
ing, poor school performance, Tow mea-
sured IQ, drug use, and parental criminal
CRIMINAL CAREERS AND CAREER CRIMINALS
ity. It is extremely difficult, however, to
develop reliable measures of the relative
influence of each of these variables,
largely because of the complexity of the
underlying relationships among them
and because different studies highlight
only one or a few of the variables rather
than all of them.
The factors that distinguish participants
from nonparticipants couIcl well be dif-
ferent from the factors that distinguish
among participants, in terms of their
offending frequency. The next chapter
considers estimates of frequency and the
factors associated with variation in fre-
quency.
