Methodology of the Current Study

Objectives

This Bulletin examines how adolescent victimization affects victims' lives during adolescence and adulthood. The study is an "inventory of effects" (see Blalock, 1969:41) in that it emphasizes a single cause and traces its multiple effects. The research conducted for this Bulletin investigates four questions related to adolescent victimization:

  • What are the immediate effects on the victim, in terms of physical injuries and financial losses?

  • How is adolescent victimization related to certain voluntary problem behaviors and involuntary problems in both adolescence and adulthood? (Voluntary problem behaviors include illicit drug use, perpetration of violent and property crimes, and perpetration of domestic violence. Involuntary problems include mental health problems and further victimization, including domestic violence victimization.)2

  • Is adolescent victimization related to certain specific problems in adulthood, regardless of whether the same problems were present in adolescence?

  • How does adolescent victimization affect adult life as measured by a general index of success?

The research goes beyond simply correlating adolescent victimization with subsequent problems. It introduces controls for prior problems, so the true impact of victimization can be determined, and satisfies the three usual criteria for inferring a causal relationship between one variable and another (Menard, 1991): association, time ordering (by relating adolescent victimization to adult problem outcomes), and control for spuriousness (by accounting for other adolescent problems as possible predictors of the corresponding adult problem outcomes). The analysis may omit some variables that could conceivably affect the relationship between adolescent victimization and subsequent problems, but such variables are likely to be captured in the study's measure of prior problems. At most, such an omission would raise the question of whether the influence of adolescent victimization on adult problems is direct or indirect—it would not call into question the existence of the influence.

Sample

Data for the research presented in this Bulletin are taken from all nine waves (interview cycles) of the National Youth Survey (NYS), a prospective, longitudinal study of a probability sample of Americans who were 11–17 years old in 1976 (the first year for which data were collected) and 27–33 years old in 1992 (the most recent year for which data are available).3 From 1976 to 1980 (waves 1–5 of the survey), data were collected annually. After 1980, data were collected at 3-year intervals, beginning with 1983 (wave 6) and continuing to 1992 (wave 9). The NYS sample allowed researchers to follow the same individuals, who collectively are representative of the total U.S. population in their age group, from their adolescence in the 1970s well into their adult years in the 1990s. (For additional details on the NYS sample, see the sidebar below.)

The National Youth Survey Sample
The National Youth Survey (NYS) used a probability sample of households in the continental United States, based on a self-weighting, multistage, cluster sampling design. The sample was drawn in late 1976 and contained an estimated 2,360 eligible youth born between 1959 and 1965, of whom 1,725 (73 percent) agreed to participate in the study, signed informed consents, and completed the interviews in the initial wave of the survey. Overall completion rates were more than 94 percent of the original respondents for waves 2 and 3, 87 percent for waves 5 and 6, 80 percent for wave 7, 83 percent for wave 8, and 78 percent for wave 9. A comparison of the age, gender, and race/ethnicity of individuals who participated in the survey and of those who were eligible but did not participate indicated that the loss rate from any particular age, gender, or racial/ethnic group was proportional to that group's representation in the population. Differences in the sample at wave 1 and subsequent waves with respect to social and demographic characteristics, illegal behavior, and substance use were small and not statistically significant. With respect to sociodemographic characteristics, NYS respondents at each wave appear to be representative of the total U.S. population born between 1959 and 1965, as established by the U.S. Census Bureau. In particular, the NYS sample is representative of the U.S. population in this age range with respect to gender, race/ethnicity, and rural/suburban/urban residence.

Questions about violent victimization were asked in all nine waves of NYS interviews (for the years 1976-80, 1983, 1986, 1989, 1992); however, in the second year (1977), for administrative reasons, 57 percent of the respondents were not asked about property victimization. This essentially divides the NYS sample into two subsets, one with and one without potentially complete data on property victimization for the full nine waves. In addition, for some individuals, data for one or more years are missing. This combination of systematically missing data for property victimization at wave 2 plus sporadically missing data for some respondents allows two possible approaches to the analysis of the data. One approach, known as "pairwise elimination," uses all available cases in each year, so that calculations of the prevalence and frequency of victimization and correlations between victimization and other problems are based on a variable number of cases. The second approach, known as "listwise elimination," uses only complete cases, i.e., cases with data for all variables in all years used in the analysis. The two approaches produce similar results. Consistent with general practice, the present analysis uses pairwise elimination in the calculation of prevalence, frequency, and correlation coefficients (tables 1, 2, and 3) but listwise elimination in the logistic regression analysis of the influence of adolescent victimization on adult outcomes (tables 4, 5,and 6).

Measurement

Types of victimization. The analysis includes two types of victimization: property and violent. Property victimization includes having a vehicle (car, motorcycle, or bicycle) stolen; having items stolen from a vehicle, including packages, vehicle parts, and bicycle locks; having items such as clothing or other possessions stolen from a public place; experiencing vandalism (deliberate damage of one's property, such as slashing tires or ripping up books); and having a pocket picked or purse or wallet snatched (or experiencing an attempt to do so).4 Violent victimization includes having something taken directly by force or threat (or experiencing an attempt to do so); being beaten up or threatened with a beating;5 being attacked with a weapon such as a knife, gun, bottle, or chair;6 and being raped or otherwise sexually attacked (or experiencing an attempt to do so).

Prevalence and frequency. The analysis considers both the prevalence and the frequency of victimization. Prevalence refers to whether an individual has been a victim of crime during a given measurement period, and frequency refers to how many times the individual has been a victim during that period. Although for some purposes the analysis considers annual prevalence and frequency of victimization, it focuses on the cumulative prevalence and frequency over the full 17 years covered by the NYS.

Measures of problem outcomes. The analysis uses the following measures of problem outcomes (in addition to subsequent criminal victimization):

  • Nondomestic violent offending (felony assault).7
  • Serious property offending (felony theft).
  • Serious domestic violence offending.8
  • Serious domestic violence victimization.9
  • Marijuana use.
  • Polydrug use.
  • Problem drug use.10
  • Mental health problems.11

The analysis considers measures of problem outcomes for both adults and adolescents. The measures as applied to adults are described in greater detail in the sidebar below. In some cases, it was not possible to use the same problem outcome measures for adolescents as for adults; endnotes 9–11 explain how the analysis handles such cases.

Adult Problem Measures
  • Nondomestic violent offending: Attacking someone with intent to seriously injure or kill the person, participating in gang fights, having or attempting to have sex with someone against his or her will, or taking or attempting to take something from someone by force or threat of force. In wave 9, battery (hitting or beating someone so badly that the person probably will need a doctor) replaced gang fighting. This is the NYS felony assault scale; the nondomestic violent victimization scale contains similar measures.

  • Serious property offending: Stealing or attempting to steal something worth more than $50, stealing or attempting to steal a motor vehicle or a bicycle, breaking into a house or car to steal something or to look around, or buying stolen goods. This is the NYS felony theft scale; the serious property victimization scale contains similar measures.

  • Serious domestic violence offending: Kicking, biting, or hitting one's spouse or partner with one's fist; hitting or trying to hit one's spouse or partner with something; beating up one's spouse or partner; threatening one's spouse or partner with a knife or gun; or using a knife on or firing a gun at one's spouse or partner. Measures for serious domestic violence victimization parallel the measures for serious domestic violence offending.

  • Marijuana use: Use of marijuana or hashish.

  • Polydrug use: Use of amphetamines, barbiturates, cocaine, hallucinogens, or heroin. (Use of any of these drugs almost always occurs in combination with marijuana and alcohol use.)

  • Problem drug use: Presence of problems resulting from illicit drug use, including interpersonal relationships, fights, absence from work, or trouble with the police.

  • Mental health problems: Indications of clinical anxiety, depression, obsessive-compulsive behavior, or schizophrenia, based on the Diagnostic Interview Schedule (Robins et al., 1981).

Additional descriptions of these measures may be found in Elliott, Huizinga, and Menard (1989).

Sociodemographic factors. In addition to examining the problem outcomes noted above, the study considers four sociodemographic factors: age in 1976 (11, 12, or 13)—essentially a control for any effects related to age cohort; gender; race/ethnicity (majority or minority); and socioeconomic background (parents' socioeconomic status in 1976), as measured by the Hollingshead two-factor index (Hollingshead and Redlich, 1958).

Analysis

Age cohorts. The analysis is limited to the three youngest NYS age cohorts. It measures adolescent problems in respondents who were ages 11–17 during 1976–80 and adult problems in the same respondents when they were ages 21–29 during 1986–92. The measurement of adolescent problems encompasses respondents who were both young enough not to have experienced certain problems (especially illicit substance use and mental health problems) and old enough to have had the opportunity to experience all of the problems. At the point of analysis for adult problems, most respondents had made the transition from school to the labor market. An interval of at least 4 years of age (from 17 to 21) and 6 calendar years (from 1980 to 1986) thus separates the measurements of adolescent and adult problems.

Mental health problems. Adult mental health problems are analyzed only for 1992, because that is the only year for which the NYS measured specific mental health problems other than depression (see endnote 11). In particular, 1992 is the only year for which the NYS measured PTSD. The 1992 measures, however, include both past-year prevalence and ever-prevalence of anxiety, depression, and PTSD.

Logistic regression models. To examine the relationship of victimization to other problems as the sample develops from adolescence to adulthood, the analysis uses logistic regression models (Hosmer and Lemeshow, 1989; Menard, 1995) to explore the prevalence of specific problems in adulthood. In each model, the dependent variable is the prevalence of an adult problem outcome, and the independent variables (i.e., the possible predictors of that problem) are the prevalence of the same (or comparable) problem during adolescence, the prevalence of violent victimization and property victimization during adolescence,12 and the sociodemographic variables mentioned earlier. The analysis focuses on the prevalence of problem behaviors and conditions in adolescence and adulthood rather than their frequency (see explanation of prevalence and frequency). This focus ensures consistency across the different problem measures. It also reflects the author's primary interest in exploring the continuity of problem behaviors and conditions from adolescence to adulthood (i.e., whether a problem persists rather than how many times it occurs).



Previous Contents Next

Line
Short- and Long-Term Consequences of
Adolescent Victimization		 Youth Violence Research Bulletin • February 2002