In this appendix, we summarize the 15 longitudinal studies on crime and delinquency included in Table 1. We have grouped these longitudinal studies into three categories, offender-based samples, cohort samples, and general population samples as well as described the different characteristics inherent in each type of sample. We also include the specific descriptive results from each sample with regard to age of last criminal justice contact by age of last follow-up.
Offender-based samples are ones that purposefully select individuals based on their involvement with the criminal justice system, typically as a juvenile. Given that offender-based samples, by design, have had a criminal justice interaction (e.g., arrest, adjudication) as part of the sampling criteria, individuals in these samples tend to have extensive criminal careers that span later into the life course. We draw on five offender-based samples in the current study, which are detailed below.
California Youth Authority
While there are multiple samples of cohorts of parolees released from the California Youth Authority, the sample we use is the publically available longitudinal data set of 524 racially diverse (51.5% non-white) serious male offenders released from California Youth Authority (CYA) institutions when they were in their late teens to early 20s (ages 16 to 22) . The offenders were released from the CYA between 1965 and 1984 and were followed up for 7 years post-release with ages ranging from 16 to 22 at baseline and 22 to 28 at the final wave with the majority of the sample followed from ages 18–20 to 24–26 (85.7%). Data were collected on violent and nonviolent criminal arrests for each individual in each year of follow-up as well as the number of months they were not serving time in a jail, prison, or a CYA detention center.
As stated in the “Data and Analysis
” section, the criteria for inclusion requires at least two criminal justice contacts and for offender samples the criteria is one adult contact (age 17+); four men were not arrested for the entire 7-year follow-up period leaving 99.2% of the sample with an arrest post-release. The average age of last arrest for the sample of 520 male offenders is 23.75 and an average age of last follow-up of 24.63. As the age of follow-up increases for these men, the age of last arrest increases in a linear fashion with an age of last arrest of 22 for those who were last followed-up at age 23 and an age of last arrest of 26 for those followed to age 27. This pattern was similar for white and non-white males. Also, when incarceration is taken into account (i.e., the individual had to be “on the street” for at least half of the last year of follow-up (n
= 313)), the pattern of age of last arrest is consistent with the pattern displayed when incarceration is not taken into account.
Cambridge-Somerville Youth Study
The Cambridge-Somerville Youth Study (CSYS) is a delinquency prevention experiment embedded in a prospective longitudinal survey of offending over the life course. Begun in 1939, the study enrolled 650 predominantly white (91%) underprivileged boys, ages 5 to 13 years (mean birth year = 1928), from Cambridge and Somerville, Massachusetts. The boys were placed in matched pairs and one member of each pair was chosen at random (on a coin toss) to be in the treatment group, which received individual counseling and home visits. While the boys were not selected based on delinquency status, this study is included as an offending sample for the purposes of the current research to reflect the fact that the potential for offending was part of the selection criteria for inclusion. The CSYS boys were recommended by police, teachers, church personnel, and social workers, and represent either “difficult” or “average” temperament . The boys were matched on a number of background factors including “delinquency prognosis” by teachers and social workers . Thus, this sample comprises individuals who had not necessarily displayed delinquent behavior at the time of its inception, but includes individuals who had been identified early in life as likely to be delinquent. The follow-up sample includes 506 boys from the original sample, comprising 253 matched pairs (for a historical overview, see ).
Estimates on age of last conviction for the 506 boys were made available by Brandon Welsh and Steven Zane from Northeastern University, who are currently working with the Cambridge-Somerville data. Of the total sample, 255 had an age of last conviction for a non-traffic offense (50.4% of the follow-up sample). Among those alive at the last follow-up (n
= 219), which was in 1979, 168 had two or more convictions with at least one of those convictions occurring in adulthood (age 17 or older). The average age at last follow-up was 50 years and the average age of last conviction for this sample is 35.51.
Gluecks’ Unraveling Juvenile Delinquency Follow-up Study
The original Unraveling Juvenile Delinquency study was conducted by the Gluecks , and consisted of 500 juvenile delinquent males and 500 matched non-delinquent males in Boston, Massachusetts born as early as 1925. The 500 delinquent boys were selected from two reform schools in Massachusetts. Beginning in 1987, Laub and Sampson reconstructed the Gluecks’ original Unraveling Juvenile Delinquency data to age 32; in 1994, they began a follow-up data collection effort for the Glueck men. In part, this follow-up effort included collecting annual criminal histories and death information for 480 men from the delinquent sample (for more information on the full follow-up effort, see : Chapter 4). These criminal history data, which include all violent, property, and drug or alcohol offenses, are used to estimate the age of last arrest for all types of crimes. Traffic offenses along with other more minor types of offenses, such as vagrancy and conspiracy, are excluded. Unlike the other offender sample estimates, the estimates for age of last arrest can not account for incarceration information due to a lack of incarceration data from ages 32 to 69.7
Given the presence of a criminal justice contact inherent in the selection for this sample, we restricted the sample to include those men who had at least one arrest as an adult by age 69 (n = 408, 85.0%). However, death information is particularly important to consider due to the age of the sample. Thus, to more accurately estimate the age of last follow-up, we further restricted the sample to those alive in 1994 (50.5% of the 408 men). While estimating the age of last arrest for each of the ages of last follow-up or death would be ideal, there are too few men in each age to give reliable estimates for the mean age of last arrest for the younger ages. Therefore, we report the mean age of last arrest for those men who remained alive until 1994 to ensure comparability with other samples (n = 206). However, it should be noted that when we plot the age of last arrest and age of death over the life course, there is a linear trend for those who die in early adulthood, which begins to diverge by the mid-20s to early 30s, similar to the patterns shown in Fig. 1.
The pattern for the Glueck data shows that for those who remained alive, the average age of last arrest is 39.18, which is over 25 years before the average age of last follow-up, which is 65.36 years. Moreover, when disaggregated by crime type, the mean age of last property, violent, and drug/alcohol offenses for all of the men (age 60 to 69) was 32.25 (n
= 175), 35.06 (n
= 127), and 38.80 (n
= 133), respectively.
Ohio Life Course Study
The Ohio Life Course Study (OLS), led by Giordano [33
], is a longitudinal study comprised of three interviews that spans a period of 22 years. The original sample consisted of 254 delinquent and institutionalized males (50%) and females (50%) in Ohio in 1982. Participants were racially diverse (62% white, 32% black) and averaged 16 years of age at the first interview. The first follow-up interview was completed in 1995, when the respondents averaged 29 years of age. Although there was a second follow-up study, which was completed in 2003 when the respondents were 37 years of age, on average, this second follow-up did not include a criminal records search or self-report questions on arrest. Thus, the official arrest data are gleaned from the 1995 data collection effort. Applying the sample selection criteria of at least two non-traffic offenses, one of which occurred in adulthood, resulted in 86 men and women with an age of last arrest. Overall, the mean age of last arrest is 25.29 with the now familiar trend of lower ages of last arrest for those followed to younger ages and older ages of last arrest for those followed into their 30s.
Pathways to Desistance Study
The Pathways to Desistance study is a multi-site, longitudinal study of 1354 serious adolescent offenders who were adjudicated of a serious offense in the juvenile and adult court systems of Maricopa County, Arizona, and Philadelphia County, Pennsylvania, between 2000 and 2003 . The age ranges of the sample at the start of the study were 14 to 19 years. The majority of the baseline sample is male (86.4%) and racially/ethnically diverse (20.2% white, 41.4% black, 33.5% Hispanic, and 4.8% other). Data were collected for 11 waves over a span of 7 years, with the first six waves of follow-up interviews corresponding to 6-month recalls and the latter five waves corresponding to 12-month recalls.
For this study, age of last police contact was derived from the answer to whether the police picked up the respondent and accused him/her of something in the recall period. This question was asked at each wave. However, the nature of the crime or reason for being picked up is not available. As a result, estimates may be biased upward and may include minor crimes. Once we selected on only those whose age of last arrest was in adulthood, the sample was reduced from the original 1354 to 947 individuals ranging from age 17 to age 26 as the last age of follow-up. Overall, this offender-based sample that has been followed into the mid-20s has an age of last police contact 1 to 2 years before the age of last follow-up. This pattern is consistent by gender, race, and when accounting for incarceration time (i.e., selecting only on those who were “on the street” for more the 50% of the recall period) (data available upon request).
Cohort studies are a subset of longitudinal panel studies where individuals who share a common characteristic are followed over time. Seven of our studies are categorized as cohort studies: three birth cohort studies, three grade cohort studies, and an age cohort study. When calculating the mean age of last criminal justice contact for non-offender-based samples, we selected on those cohort members with at least two criminal justice contacts to ensure that there was a distinct age of first and age of last contact (although these could be the same age). We also restricted each sample to include only those who have at least one contact in adulthood (age 17 or older). With respect to consideration of incarceration time, most of these studies do not have that information available; however, this lack of data is less problematic than for offender-based samples.
Marion County Youth Study
The Marion County Youth Study is a grade cohort study that originally comprised 1227 sophomore males in Marion County (i.e., Class of “1967”), who were 98% White [77
]. While several subsamples were later assessed from this original sample (i.e., high school dropouts, juvenile delinquents, a 25% random sample of the base population), this study focuses on the 379 males (30.9% of sophomore sample) who had an adult criminal record by 1980 (by age 32). Official police and court records are used to assess the subject’s law violations including non-minor traffic offenses. From this sample, 217 males (57.3% of the adult criminal group) had at least two arrests by age 32. The average age of last arrest for this grade cohort was 24.32 years, nearly 8 years prior to the age of last follow-up.
Philadelphia 1958 Birth Cohort
The Philadelphia 1958 Birth Cohort study includes 27,160 racially diverse males and females from both high and low socioeconomic statuses who were born in 1958 and who remained Philadelphia residents from ages 10 to 18. Of the full sample, 52.7% are non-white, 47.3% are white; 51.5% are female and 48.5% are male [32
]. This study looks at police contacts through age 26 using official juvenile and adult records. All types of index and non-index crimes are included for juveniles and adults. While 28.8% of the cohort had at least one police contact (n
= 7821), 16.4% had two or more offenses (n
= 4463). Once the sample is further restricted to those with at least one adult contact, the sample is reduced to 3319 men and women (12.2% of the original birth cohort, 42.4% of the arrested sample). The mean age of last arrest for this subsample is 21.31 years of age. Similar to the patterns seen with the other cohort studies, this pattern holds across gender and race; men (n
= 2856, mean = 21.45), women (n
= 463, mean = 20.45), whites (n
= 984, mean = 20.88), and non-whites (n
= 2335, mean = 21.49).
Pittsburg Youth Survey
The Pittsburgh Youth Study is a community-based sample of boys in three grade cohorts first, fourth, and seventh from inner-city Pittsburgh, Pennsylvania, who have been followed since 1987. Each cohort represents a random selection of the grade cohort at baseline; a subgroup of approximately 250 “high-risk” boys, selected based on a risk screening tool, and a subgroup of approximately 250 boys randomly selected from the original sample were followed in subsequent waves. The oldest (seventh grade) cohort (used in this study) was followed from ages 13 to 36; 41.7% are white, 55.6% black, .2% Hispanic, .4% Asian, and 3.2% mixed ethnicity .
Although these data are not publicly available, we leverage published estimates that allow us to calculate an age of last arrest for the 506 boys from the seventh grade cohort. Jolliffe and colleagues  conducted a systematic review of life-course-persistent, adolescent-limited, and late-onset offenders for seven prospective longitudinal studies, one of which is the Pittsburgh Youth Study’s seventh grade cohort followed to an average age of 36. These published data provide the prevalence, average age of onset, and criminal career duration of moderate to serious theft and violence for each offending group. From these published data, we estimate the average age of last arrest for the life-course-persister and late-onset groups, combined. To ensure that the most similar inclusion criteria possible is applied to this sample (i.e., at least one offense occurs in adulthood), we do not include the adolescent-limited group since the definitional criteria does not require an adult arrest. The exclusion of the adolescent-limited group introduces the likelihood that the age of last contact is biased upward.
To calculate the average age of last arrest, we first calculated the average age of onset for the life-course-persister (n
= 68) and late-onset offenders (n
= 73) and estimated a weighted average that is calculated by summing the products of the average age of onset and number in each group and dividing this product by the total sample. We then, similarly, created a pooled estimate of the average duration and added this to the age of onset estimate. This procedure resulted in an average age of last arrest occurring at age 30.40 for this cohort.
Project on Human Development in Chicago Neighborhoods (PHDCN)
The Project on Human Development in Chicago Neighborhoods (PHDCN) includes an accelerated longitudinal cohort study designed to examine delinquency and criminal behavior among multiple cohorts as well as their family and community contexts. Between 1994 and 2001, data were collected at three time points for the over 6000 individuals from the original seven age cohorts (ages 0, 3, 6, 9, 12, 15, and 18) . Most recently, a fourth wave of data collection was completed that includes the criminal histories from official records of a subsample of the original sample (1057 individuals from cohort 0, 9, 12, and 15).
Estimates of age of last arrest by age of last follow-up are provided by Robert Sampson and Roland Neil at Harvard University. By 2017, 165 of the subsample had been arrested two or more times with at least one non-traffic arrest in adulthood (15.6% of the 1057 subsample) with the age of last follow-up ranging from 21 to 37 years. Due to data challenges, arrest histories for the age 12 and 15 cohorts were calculated using only adult arrests (17 years of age or older). This introduces the possibility that the age of last arrest estimates are biased upward. The analytic sample is majority male (n
= 123; 75%) and mostly non-white (n
= 145; 88%).
Racine 1942 and 1949 Birth Cohorts
Shannon  conducted a longitudinal study on three birth cohorts in Racine, Wisconsin (for birth years 1942, 1949, and 1955). The data from the cohort with information up to age 25 (1949 cohort) and up to age 32 (1942 cohort) are included in this study. The original subjects were identified from police, and private and parochial school records. Official records of police contacts were collected for both traffic and non-traffic offenses; the estimation of the timing of last police contact is based on reports on the non-traffic police contact data to ensure comparability with the other longitudinal samples reviewed.8
The original sample for the 1942 birth cohort consists of 1352 predominantly White (94.5%) men and women with data from age 6 to age 32. For this sample, 435 men and women (32.2% of the original sample) had at least one non-traffic police contact. Once we restricted the sample to those with at least two or more contacts and at least one of them as an adult, the sample size reduces to 213 with a mean last arrest at age of 24.31 years. This mean age varies little when estimated for men (n
= 175, mean = 24.31), women (n
= 38, mean = 24.29), and whites (n
= 179, mean = 23.80) with a higher mean for non-whites (n
= 34, mean = 27.00).
The original sample for the 1949 birth cohort consists of 2099 predominantly White (90.8%) men and women born in 1949 with data from age 6 to age 25; 38.2% of the sample had at least one non-traffic offense. Applying the selection criteria for this study resulted in 375 men and women with two or more police contacts with at least one adult (age 17 or older) police contact (297 men and 78 women; mean last arrest at age 21.02 years). The pattern across gender and race was similar; men (n
= 297, mean = 21.02), women (n
= 78, mean = 21.19), whites (n
= 288, mean = 20.86), and non-whites (n
= 87, mean = 21.57).
The Woodlawn Study comprises 1242 African-American children who were in first grade in 1966 (636 females and 606 males) in the nine public and three parochial schools in Woodlawn, a neighborhood community on the South Side of Chicago . While the data include multiple components from a variety of data sources, this study utilized the official criminal history data collected from the Chicago Police Department and the Federal Bureau of Investigation (FBI) in 1993, which spanned ages 17 to age 32, and the more recently updated criminal histories from the Illinois Criminal Justice Information Authority (ILCJA) to age 52. Mortality information was also assessed through searches of the National Death Index (NDI) as well as by corroborating reports of deaths from families and friends (n = 132 as of 2009, 11% of the original cohort) (for more information on the full data collection effort, see ). Although data are not available for days incarcerated, as a proxy, we incorporate the sentencing data from the arrest records into the criminal histories such that a person is considered incarcerated in any year where he or she has zero offenses and is known to have been sentenced to more than 1 year in prison at that age. Using this strategy, we reduce the chances of presuming someone has desisted from offending who in fact was incarcerated.9
For this sample, 565 men and women (46.4% of the original cohort, 384 men and 181 women) had at least one arrest. Given that there is no information on juvenile arrests, we restricted the sample to those with at least two or more adult arrests (to err on the conservative side, which potentially biases the estimates upward). This selection criterion reduces the sample size to 441 (329 men and 112 women). Similar to the Glueck sample, there are only a few cases of death at each age, which reduces the reliability of any aggregate estimate for those ages; yet, it should be noted that the general pattern of increasing differences between age of last arrest and age of death as this latter age increases is apparent with this cohort.
Here, we report the mean age of last arrest for those men and women who remained alive until age 52 (n
= 381), which is 38.88 years. This mean age of last arrest varies little when estimated for men (n
= 288) and women (n
= 93) separately (means of 39.26 and 37.71, respectively). Moreover, when the Woodlawn offenses are disaggregated by crime type, the average age of last violent and property offense are similar (33.17, n
= 358 and 34.46, n
= 296, respectively) followed by drug and alcohol offenses (37.46, n
US General Population Samples
General population samples utilize probability sampling strategies, such as random sampling techniques, in order to produce a sample that is representative of the US population as a whole. Unlike offender-based samples that are selected for their offender status, or cohort samples that are selected based on a shared characteristic, general population samples are selected at random from a larger population. The following three samples are ones that represent the US population’s youth and young adults.
National Longitudinal Survey of Youth 1997 (NLSY97)
The National Longitudinal Survey of Youth 1997 (NLSY97) consists of a nationally representative sample of approximately 9000 boys (51.2%) and girls (48.8%) who were 12 to 16 years old in 1996 . The sample was racially and ethnically diverse (51.9% non-black/non-Hispanic; 26.0% black; 21.2% Hispanic). The first survey took place in 1997, where surveys were administered through an in-person interview. From 1997 through 2011, surveys were completed annually; subsequent surveys are completed on a biennial basis. The study has completed 17 waves of data starting in 1997 until the most recent data collection effort in 2015–2016. Although the purpose of the survey is to document the transition from school to work and into adulthood, information on self-reported arrest and the timing of that arrest is included in each wave. Specifically, NLSY97 youth respondents were asked whether they had ever been arrested by the police or taken into custody for an illegal or delinquent offense (not including arrests for minor traffic violations).
In order to retain as much of the sample as possible, we calculated the age of last follow-up for those who dropped out of the study as well as those who remained throughout all of the waves.10
This allowed us to look at a wide range of ages of follow-up (ages 17 to 35) although the majority of the sample was followed until ages 31 to 35. Restricting the sample to those with two or more arrests, at least one of which is in adulthood, results in an analytic sample of 1747 men and women (74.9% men).11
A slight majority are non-white (black, Hispanic, and mixed race, n
= 779; 53.5%); 676 (46.5%) are white (non-black, non-Hispanic). For those followed in the teens and early 20s, the average age of last arrest is also in the teens and early 20s. In contrast, for those followed into the early to mid-30s, the average age of last arrest is in the mid-20s. This pattern is consistent when disaggregated by sex and race.
National Youth Survey
The National Youth Survey, initiated by Elliott [27
], focuses on a representative sample of 1725 males (53.2%) and females (46.8%) who were originally surveyed in 1976, when they were ages 11 to 16. The sample is racially mixed with the majority white (78.9%) followed by black (15.1%) and Hispanic (4.4%). The sample was followed for seven waves, with the last data collection effort in 1987. During waves 5 (1980), 6 (1983), and 7 (1987), men and women were asked to self-report whether they had been arrested for anything other than a minor traffic offense ever (for waves 5 and 6) or since January 1984 (for wave 7) and if so, how many and when the most recent arrest occurred. Based on these questions, we calculated the age of last arrest for the 352 arrested men and women (22% of the 1598 with data available). Of these 352 arrested men and women, 161 had two or more arrests (45.7%) with 144 men and women having at least one adult arrest (age 17 or older) (122 men and 22 women). While only 8% of the original sample met the criteria for this study, those in the analytic sample display a similar pattern of 1 to 2 years between age of last follow-up and age of last arrest compared with close to 5 years between these two among those followed into the mid-20s.
National Longitudinal Study of Adolescent to Adult Health (Add Health)
The National Longitudinal Study of Adolescent to Adult Health (Add Health) was initiated in 1994 to study adolescent health and health risk behaviors. Funded over the years through the National Institutes of Health as well as a number of other funding agencies ,12 the Add Health study has followed over 15,000 randomly selected US adolescents from 1994, when the students were in seventh through 12th grade, through 2008, when these young adults were aged 24 to 32. In-home interviews collect data on a range of factors including social, behavioral, psychological, biological, and economic domains as well as contextual data for neighborhoods, communities, and schools. Data available for public use include a random sample of one-half of the generally representative core sample, and one-half of the oversample of African-American adolescents with a parent who has a college degree.
At the time of this study, four waves of data are available for analysis (wave 5 data collection is in progress). In the final wave (wave 4), sample respondents were asked to self-report their history of arrest by the police. Most pertinent to this research are questions tapping into arrest (i.e., have you ever been arrested or taken into custody by the police?) and the age when the arrest(s) occurred. Of the wave 4 sample (n
= 5114), an age of last arrest could be calculated for 726 (14%) men and women who were arrested two or more times with at least one of these arrests occurring in adulthood. Roughly three-quarters of those with two or more arrests were men (78%), two-thirds were white (66%). Again a similar pattern emerges between age of last arrest and age of last follow-up among this sample, with a consistent picture by gender and race.