Child Abuse & Neglect 42 (2015) 30–39

Contents lists available at ScienceDirect

Child Abuse & Neglect

Research article

Adolescent exposure to violence and adult illicit drug use Scott Menard a,∗ , Herbert C. Covey b , Robert J. Franzese c,1 a b c

Department of Criminal Justice and Criminology, Sam Houston State University, Campus Box 2296, Huntsville, TX 77341, USA Adams County Human Services Department, USA Department of Sociology, University of Oklahoma, USA

a r t i c l e

i n f o

Article history: Received 12 September 2014 Received in revised form 30 December 2014 Accepted 5 January 2015 Available online 24 January 2015 Keywords: Marijuana use Illicit drug use Adolescent exposure to violence Life course National Youth Survey Family Study Logistic regression

a b s t r a c t Informed by a strain theory perspective, this study utilizes data on adolescent exposure to violence (AEV) from a prospective, longitudinal, national household probability sample that originally consisted of 1,725 respondents, first interviewed as adolescents in 1977 and last interviewed in middle adulthood in 2003. Findings from bivariate correlations and logistic regression models indicate that AEV is associated with both adolescent and adult illicit drug use, but some of the association between AEV and adult illicit drug use becomes nonsignificant when controlling for adolescent illicit drug use. Specific types of AEV associated with adult illicit drug use differ by gender. Implications, limitations, and future research directions are discussed. © 2015 Elsevier Ltd. All rights reserved.

Introduction Are exposure to (witnessing or awareness of) neighborhood violence, witnessing violence between parents, physical abuse by parents, and more general violent victimization (excluding physical abuse by parents) associated with adolescent illicit drug use and, net of adolescent drug use, predictive of adult illicit drug use? Both exposure to violence and illicit drug use are matters of concern, each in its own right. Each affects millions of Americans (Finkelhor, Turner, Ormrod, & Hamby, 2009; Substance Abuse and Mental Health Services Administration, 2011) and is associated with physical and mental health problems (Chen & Lin, 2009; Springer, Sheridan, Kuo, & Carnes, 2007; Tajima, 2004), economic disadvantage (Covey, Menard, & Franzese, 2013; Huang, Evans, Hara, Weiss, & Hser, 2011), criminal victimization and crime perpetration (Barroso et al., 2008; Elliott, Huizinga, & Menard, 1989; Menard, 2012; Rebellon & Van Gundy, 2005). Studies of the relationship of childhood and adolescent exposure to violence to illicit drug use have been marked by methodological limitations and inconsistency of results (Kendall-Tackett, 2013; Lynch, 2003; Widom, 2014; Wolfe, Crooks, Lee, McIntyre-Smith, & Jaffe, 2003), particularly with respect to the relationship of witnessing parental violence to illicit substance use; compare, for example, Fergusson and Lynskey (1997) and Herrenkohl, Sousa, Tajima, Herrenkohl, & Moylan (2008) with Zinzow et al. (2009). Widom (2014) summarizes the state of this research by noting that “Hundreds of papers have been published describing a relationship between child maltreatment and substance abuse, primarily based on crosssectional designs. Few longitudinal studies have followed abused and/or neglected children into adulthood, and, based on these few studies . . . the evidence linking child abuse and substance abuse is mixed” (p. 231, italics in original). While there

∗ Corresponding author. 1 Retired. http://dx.doi.org/10.1016/j.chiabu.2015.01.006 0145-2134/© 2015 Elsevier Ltd. All rights reserved.

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

31

may be hundreds of studies of maltreatment (a term including but not limited to physical abuse) in childhood (sometimes including adolescence, under age 18, but often limited to younger ages), and short term associations with illicit drug use, there are far fewer studies of specifically adolescent exposure to violence and its long term relationship to illicit drug use. In their meta-analysis of 118 studies of child exposure to domestic violence, Kitzmann, Gaylord, Holt, and Kenny (2003) found that only 10 utilized adolescent samples. Yet evidence from Thornberry, Ireland, and Smith (2001; see also Ireland, Smith, & Thornberry, 2002) suggests that when adolescents and younger children are compared, it is exposure to violence in adolescence, rather than earlier childhood, that poses the greater risk of later illicit drug use. This may occur for three reasons. First, childhood is more removed in time than adolescence is from adulthood, and it may be that more proximate experiences of exposure to violence have greater impact. Second, if exposure to violence occurs in childhood but not adolescence, that may indicate that the problem has been solved; but exposure to violence in adolescence indicates either continuity of problems from childhood, or the emergence of new problems in adolescence, either of which may have a greater impact on substance use than exposure to violence that has been successfully alleviated or mitigated. Third, in addition to the general issues of identity formation and transition to adult statuses characteristic of adolescence, adolescence is the stage of the life course in which exposure to, access to, and onset of illicit drug use is typically initiated (Elliott et al., 1989). It is thus more likely that adolescents, rather than children, might respond to exposure to violence in the short term with illicit drug use which, once initiated, may become a stable behavioral pattern that continues into adulthood. In view of the relative neglect and suggested importance of exposure to violence in adolescence as a risk factor for illicit drug use, in the present study we focus on adolescent exposure to violence (hereafter AEV) as a predictor of adult illicit drug use. AEV, as described by Covey et al. (2013), Eitle and Turner (2002), Finkelhor et al. (2009), and Menard, Weiss, Franzese, and Covey (2014), is a general term encompassing direct physical abuse, witnessing parental violence, and perceptions of neighborhood, violence, as different and specific forms of the broader concept of exposure to violence. We also consider more general adolescent violent victimization, separate from physical abuse by parents, as a predictor of illicit drug use. We examine the relationship of AEV to two forms of illicit drug use, hard drug use and marijuana use. Consistent with much of the literature, we distinguish between the use of marijuana, considered a “softer” drug whose use is comparatively widespread, and the less prevalent “harder” drugs including cocaine, heroin, and hallucinogen use plus nonprescription amphetamine and barbiturate use. We consider the association between AEV and illicit drug use in adolescence, but focus more on the much less studied, longer term, relationship of AEV to illicit drug use in middle adulthood.

Theory Although no theoretical perspective has been specifically developed to explain the relationship of experiencing physical abuse, witnessing parental violence, and exposure to neighborhood violence in adolescence with illicit drug use in adulthood, the broad category denoted as strain theories is straightforwardly applicable to this relationship. According to the anomie theory of Merton (1938), at the individual level, the experience of strain may lead to different modes of adaptation, one of which, retreatism, is particularly associated with substance use problems. Retreatism involves the abandonment of both success goals and of normative constraints defining legitimate means of achieving goals. Research by Menard (1995, 1997) supports the relationship of strain via retreatism to illicit drug use. Menard, like Merton, however, focused on economic strains rather than experiences of victimization and exposure to violence. More directly applicable to the present issue in that respect is the general strain theory (GST) of Agnew (1985). GST proposes that there are three types of strains: (1) strains that result from the inability to achieve positively valued goals; (2) strain that occurs when positively valued stimuli are removed; and, most pertinent to the present study, (3) strains that are present with the introduction of negative or noxious stimuli, including but not limited to witnessing or being a victim of violence. GST has been tested extensively in relationship to juvenile delinquency in general (e.g., Broidy, 2001), but largely absent in previous tests of GST is a clear, direct examination of the relationship of exposure to violence (strain type 3) to illicit drug use. Some evidence for the applicability of strain theory to the explanation of illicit drug use comes from Harrison, Fulkerson, and Beebe (1997), who found that physical abuse was associated with increased likelihood of marijuana and other drug use and that victims indicated that they were using substances as a form of self-medication, a pattern consistent with the retreatist adaptation in anomie theory. Strain theories vary in how they suggest a linkage between exposure to violence, including AEV, and illicit substance use. As noted above, Merton’s (1938) anomie theory allows for different modes of adaptation, not all of which lead to substance use. Agnew’s (1985) GST suggests that strain is mediated by negative emotionality, including feelings of depression and anger, but does not specify how negative emotionality might lead to illicit substance use, as opposed to violence (externalizing) or mental health problems (internalizing). Anda et al. (2006) and Perry (2001) suggest a neurobiological approach in which trauma affects brain functioning in ways that lead to adverse behavioral and mental health outcomes, but are not specific in describing how AEV would result in illicit substance use. Other theoretical perspectives might be suggested here, but to or knowledge none besides strain have been used to date to explain or predict the relationship of AEV (or exposure to violence more generally) to illicit drug use. Moreover, of the theories linking victimization and offending discussed in Menard (2012), some would clearly not apply here: learning theories would expect AEV to result in further violence, not substance use, and self-control theories, which emphasize voluntarily placing oneself in risky situations, would not appear to be applicable to the (presumably involuntary) situation of witnessing parental violence (although an argument might be made for self-control provoking physical abuse or resulting in greater exposure to neighborhood violence).

32

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

Based on both strain theory and past empirical results, then, we hypothesize that AEV in the family and neighborhood, in at least one of its forms (physical abuse, witnessing parental violence, exposure to neighborhood violence) will be positively associated with illicit drug use in middle adulthood. While this hypothesis is based on anomie and general strain theories, as well as prior research, the present study is not intended as a formal test of those theories, but rather is informed by those theories in its concern with the relationship between AEV and illicit drug use. The hypothesis states that at least one form of AEV will be associated with illicit drug use because existing research does not provide an adequate basis to specify which types of AEV will be associated with illicit drug use in adulthood (or in adolescence); that specification is one of the contributions of the present study to the study of AEV and illicit drug use. While support for such an association would clearly be evidence consistent with the strain theories from which it is derived, the present research does not constitute a formal test of any strain theory, and does not preclude the possibility of developing other theories to explain that association. Methodological Limitations of Prior Research Addressed in the Present Research As suggested above, there are several weaknesses in the existing research on the relationship of AEV (and also childhood exposure to violence) with subsequent illicit drug use. First, this research often relies on samples that are small in size and/or limited to clinical samples of individuals who have experienced certain types of exposure to violence, or to predominantly urban, minority ethnicity, lower socioeconomic status samples, without comparison samples (Gewirtz & Edleson, 2007; Heyman & Slep, 2002; Lynch, 2003; Margolin & Gordis, 2000). In particular, studies involving national probability samples representative of the general population are rare (Rebellon & Van Gundy, 2005); an exception is the National Study of Children’s Exposure to Violence or NATSCEV (Finkelhor et al., 2009). Second, some studies have failed to distinguish between directly experiencing violence (as a perpetrator or a victim) from broader exposure to violence such as witnessing violence in the family or awareness of violence in the neighborhood (Acosta, Albus, Reynolds, Spriggs, & Weist, 2001; Gewirtz & Edleson, 2007). Third, several reviews of the literature have indicated the rarity of and the need for longitudinal studies (e.g., Gewirtz & Edleson, 2007; Widom, 2014). Kitzmann et al. (2003) found that only 7 of the 118 studies they reviewed utilized longitudinal data sets. Even in existing longitudinal studies, individuals are often followed only into adolescence (e.g., Ehrensaft et al., 2003) or young adulthood (e.g., Yates, Dodds, Sroufe, & Egeland, 2003), neglecting longer term consequences of adolescent and childhood exposure to violence (Gewirtz & Edleson, 2007; Margolin & Gordis, 2000). In the present study, weaknesses in past research are addressed, first by examining a national probability sample (as opposed to a local or clinical sample) of over 1,000 total respondents. Second, we distinguish among direct victimization (physical abuse) and more general exposure to violence (witnessing parental violence and neighborhood violence). Third, we measure exposure to violence primarily in adolescence, and measure past-year prevalence of hard drug and marijuana use in middle adulthood, adding to our knowledge of both of these otherwise underrepresented age ranges. Fourth, we not only examine bivariate correlations but also include the different measures of AEV in the same analyses and consider their separate and combined relationships to hard drug use and marijuana use. Fifth, we include adolescent violent victimization other than physical abuse, in light of the possibility that violent victimization in general may be a risk factor for illicit drug use (Menard, 2002; Ruback & Thompson, 2001). Sixth, we control for prior substance use, in light of evidence that the when controlling for prior substance use, the relationship of AEV to adult illicit drug use may not be direct (Menard, 2002), but may reflect a short term impact of AEV on adolescent illicit drug use, with subsequent continuity in illicit drug use explaining the bivariate association between AEV and adult substance use. Seventh, we control as well for other potential confounding influences, including gender (by analyzing the data separately for females and males), ethnicity, urban–suburban–rural residence, and socioeconomic status. The separate analysis by gender is necessary because past research suggests that females and males may respond differently to exposure to violence (Gewirtz & Edleson, 2007; Herrenkohl et al., 2008; Kendall-Tackett, 2013; Widom, 1989). Methods Sample Data for this study are taken from the National Youth Survey Family Study (NYSFS). The NYSFS study design, described briefly here and in more detail in Covey et al. (2013); see also Elliott et al. (1989), involves twelve waves of data (eleven for the focal respondents in the present study) over a 27-year period, based on a probability sample of households in the continental United States selected in 1976 using a multistage, cluster sampling design. The sample consisted of an estimated 2,360 eligible youth respondents aged 11–17 at the time of the initial interview, of whom 1,725 (73%) agreed to participate in the study, signed informed consents and completed interviews in the first year. Completion rates over waves 2–4 (collected 1977–1980) were above 94% of the original sample; 87% for waves 5 (1981) and 6 (1984); 80% for wave 7 (1987); 83% for wave 8 (1990); 78% for Wave 9 (1993); and, of the surviving respondents, 75% for wave 10 (2002) and 70% for wave 11 (2003). There appears to be no systematic initial nonparticipation or sample loss over time based on demographic characteristics (Covey et al., 2013; Elliott et al., 1989), and analyses of the effects of sample attrition in the NYSFS suggest that it has little or no impact on substantive findings (Brame & Paternoster, 2003; Elliott et al., 1989; Jang, 1999; Lackey, 2003; Menard & Elliott, 1993).

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

33

Measures Two measures of self-reported illicit drug use are considered in the present study. Self-report data on illicit drug use are generally recognized as having good reliability and validity, especially compared to official data on illicit drug use; for a more extended discussion on the reliability and validity of self-reported illicit drug use data, see for example Elliott et al. (1989) and Harrison (1995). (1) The scale for hard drug use, used in previous research with this data set (Elliott et al., 1989; Menard, 2002) asks how many times in the past year the respondent has engaged in amphetamine (nonprescription), barbiturate (nonprescription), cocaine, hallucinogen, and heroin use. (2) Marijuana use is represented by a single item asking how many times in the past year the respondent has engaged in marijuana use. For both types of illicit drug use as dependent variables in the present study, we use the prevalence of illicit drug use coded as 0 = no use, 1 = use one or more times. For the focal predictors in this study, described briefly here (for further detail, see Covey et al., 2013) we use multiple waves of a single measure for each of being physically abused by parents, witnessing parental violence, and exposure to neighborhood violence. (3) For being physically abused by parents, we use a single item asking whether (yes or no) the respondent was beaten up by a parent in the previous year, measured at waves 1–5 (ages 11–17 to 15–21). From this we obtain a measure of cumulative prevalence (0 = no, 1 = yes) of being beaten up by a parent at least once for waves 1–5. (4) Witnessing parental violence consists of retrospective questions asked at waves 8–11 (ages 24–30 through 37–43) about whether the respondent witnessed either of their parents physically hurting each other in a fight. Some respondents reported witnessing no parental violence; others gave inconsistent reports (answering yes in one wave and no in a subsequent wave); and others gave consistent responses (once they answered yes, they continued to answer yes in any subsequent waves). Setting not witnessing parental violence as the reference category, we have two dummy variables, one for inconsistent and one for consistent reports of witnessing parental violence. As indicated in Covey et al. (2013), this distinction appears to be important with respect to later reported outcomes. (5) For exposure to neighborhood violence, asking whether assaults and muggings were a problem in the neighborhood, we have only parental reports at wave 1, and only respondent reports at waves 5 and subsequently. For individuals with either only parental reports or only self-reports of whether they had witnessed assaults and muggings, the available scale was used. For individuals with both parental and self-reports (the vast majority of the cases), if either the parent or the focal respondents reported neighborhood violence, this variable was coded as indicating the presence of violence in the neighborhood. In addition to the predictors and outcomes with which we are most concerned, we also include controls for (6) violent victimization other than physical abuse in adolescence; (7) sex/gender (coded 0 = male, 1 = female), controlled by performing separate analysis for females and males, in light of evidence cited earlier that the impacts of AEV may be different for females and males; (8) race/ethnicity (coded 0 = white/majority, 1 = nonwhite/minority, the latter including individuals who identify themselves as Latino or Hispanic (consistent with U.S. Census Bureau classifications when the study began in 1976, Latino/Hispanic is not treated as a category separate from race); (9) place of residence (dummy variables for urban and rural, with suburban as the reference category); (10) family structure at wave 1 (0 = other than a two parent family, 1 = two parent family); (11) socioeconomic status of parents at wave 1 based on the Hollingshead index of social position (an index that combines parental education and occupational prestige; see Bonjean, Hill, & McLemore, 1967) coded into two dummy variables, upper/middle class and lower class, with working class as the reference category; and (12) prior prevalence of the dependent variable (hard drug or marijuana use) in waves 1–5. Analytical Strategy We begin with a presentation of descriptive statistics and bivariate correlations. Our main interest in the bivariate correlations is to examine the relationship on the outcomes (hard drug and marijuana use) of each of the predictors of interest (physical abuse, witnessing parental violence, exposure to neighborhood violence) on each without controlling for one another or for other variables in the model. These bivariate correlations serve two purposes: first, to place the present study in the context of other studies in which only such bivariate relationships, and not multivariate analyses, are used to examine the relationship of AEV with illicit drug use and other adult outcomes, and second, when expected relationships are not observed in the multivariate analysis, to offer clues to whether such relationships are altogether absent in the data, or whether they may be mediated by other variables in the analysis. We then examine the relationship of adult illicit drug use to the predictors controlling for each other and for the other variables in the model. Because all of the dependent variables in the analysis are dichotomous, logistic regression is used. Since all of the predictors in the analysis are also measured on the same dichotomous (0, 1) scale, we focus on the odds ratio as a measure of the strength of the relationship. While standardized logistic regression coefficients are often appropriate, especially when the predictors are measured on different scales, the use of the odds ratio is an acceptable and intuitively meaningful option when all of the predictors are measured on the same dichotomous scale (Menard, 2010). Results Descriptive statistics for the variables used in the present study are presented in Table 1. Only those cases which have valid values for all of the variables in the analysis (N = 1,119, 65 percent of the original sample) are included in Table 1. All of the variables are dichotomous with minimum = 0 and maximum = 1, and the mean for each variable is equal to the

34

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

Table 1 Descriptive statistics (female and male).a Variable Prevalence of adult hard drug use Prevalence of adult marijuana use Prevalence of adolescent hard drug use Prevalence of adolescent marijuana use Prevalence of parental physical abuse (beaten up by parents) Prevalence of witnessing parental violence (inconsistent) Prevalence of witnessing parental violence (consistent) Prevalence of neighborhood violence Prevalence of adolescent violent victimization other than parental physical abuse (general adolescent violent victimization) Nonwhite Urban Suburban (reference category) Rural Two-parent family structure Upper/middle class Working class (reference category) Lower class a

Mean (proportion) Female/male

Standard deviation Female/male

.126/.144 .140/.234 .224/.251 .532/.583 .105/.077 .115/.117 .168/.126 .149/.148 .579/.810

.332/.352 .347/.424 .417/.434 .499/.493 .307/.266 .320/.322 .374/.332 .356/.356 .494/.393

.163/.185 .201/.194 .633/.614 .166/.192 .818/.850 .241/.241 .338/.298 .421/.461

.369/.388 .401/.396 .480/.491 .372/.394 .386/.357 .428/.428 .467/.462 .494/.499

Listwise n = 572 for females, 547 for males.

Table 2 Bivariate correlations: females.a Variable; Pearson’s r (p)

Physical abuse Inconsistent witnessing parental violence Consistent witnessing parental violence Neighborhood violence Adult hard drug use Adult marijuana use Adolescent hard drug use Adolescent marijuana use Adolescent general violent victimization Nonwhite Urban Rural Two-parent family Upper/middle class Lower class a

Neighbor-hood violence

Parental physical abuse (beaten up by parents)

Inconsistent witnessing parental violence

1 .001 (.974)

Redundant 1

Redundant Redundant

Redundant Redundant

.228 (.000)

−.162 (.000)

1

Redundant

.082 (.051) .111 (.008) .158 (.000) .282 (.000) .196 (.000) .292 (.000)

.049 (.241) −.038 (.363) .028 (.505) .095 (.023) .032 (.444) .087 (.039)

−.030 (.477) .154 (.000) .129 (.002) .073 (.080) .047 (.265) .194 (.000)

1 −.010 (.805) .030 (.475) .012 (.783) .038 (.368) .078 (.063)

.050 (.231) .013 (.750) −.045 (.278) −.194 (.000) −.100 (.017) .101 (.016)

.063 (.130) .037 (.373) .001 (.989) −.028 (.498) −.050 (.231) .080 (.057)

.018 (.674) −.015 (.717) .076 (.069) −.286 (.000) −.078 (.061) .100 (.017)

.162 (.000) .158 (.000) −.107 (.010) −.134 (.001) .017 (.682) .012 (.778)

Consistent witnessing parental violence

Listwise n = 572.

proportion of cases coded 1 for that variable. As indicated in Table 1, prevalence of hard drug use is similar (no more than five percentage points different) for females and males, 22–25% in adolescence and about half that in adulthood; but prevalence of marijuana use, while similar in adolescence at 53–58%, is different in adulthood, with females reporting 14% and males 23% prevalence of marijuana use in adulthood. Adolescent physical abuse (11% for females and 8% for males) and consistent witnessing of parental violence (17% for females and 13% for males) are slightly higher for females than for males, but prevalence of more general adolescent violent victimization is considerably higher for males (81%) than for females (58%); and rates of inconsistent witnessing of parental violence (12% for both) and exposure to neighborhood violence (15% for both) are practically identical. Also highly similar, as we would expect, are percentages who are nonwhite, urban rural, in two-parent family structures, upper/middle class, and lower class (all within 0–4 percentage points). Tables 2 and 3 present bivariate correlations between the four focal predictors and the other variables in the analysis for females and males, respectively. In Table 2 for females, parental physical abuse is significantly correlated with adult hard drug and marijuana use, and it is also positively related to adolescent hard drug and marijuana use, to general violent victimization, and to lower socioeconomic status. It is significantly negatively related to two-parent family status and to upper/middle socioeconomic status. Inconsistent witnessing of parental violence is not related to either of the measures of adult illicit drug use, and only to prevalence of hard drug use in adolescence. It is also significantly related to adolescent general violent victimization. In contrast, consistent witnessing of parental violence is positively related to adult hard drug use (but only marginally, .05 ≤ p ≤ .10, to adolescent hard drug use) and to adult marijuana use, but not to marijuana use in adolescence. It

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

35

Table 3 Bivariate correlations: males.a Variable; Pearson’s r (p)

Physical abuse Inconsistent witnessing parental violence Consistent witnessing parental violence Neighborhood violence Adult hard drug use Adult marijuana use Adolescent hard drug use Adolescent marijuana use Adolescent general violent victimization Nonwhite Urban Rural Two-parent family Upper/middle class Lower class a

Parental physical abuse (beaten up by parents)

Inconsistent witnessing parental violence

Consistent witnessing parental violence

Neighbor-hood violence

1 .002 (.966) .097 (.023) .170 (.000) .077 (.073) .100 (.019) .134 (.002) .049 (.254) .140 (.001) .022 (.607) .084 (.048) −.053 (.213) −.033 (.444) −.098 (.021) .023 (.596)

Redundant 1 −.138 (.001) .024 (.569) .028 (.507) .027 (.526) .052 (.224) .008 (.856) .089 (.037) .120 (.005) −.035 (.420) −.062 (.148) −.070 (.101) −.046 (.285) .029 (.503)

Redundant Redundant 1 .043 (.314) .079 (.065) .050 (.242) −.004 (.933) .031 (.472) .072 (.093) .103 (.016) .023 (.596) −.003 (.936) −.164 (.000) −.098 (.021) .069 (.109)

Redundant Redundant Redundant 1 .122 (.004) −.024 (.579) .104 (.015) .050 (.246) .097 (.023) .253 (.000) .199 (.000) −.125 (.003) −.185 (.000) .005 (.899) −.024 (.577)

Listwise n = 547.

is also positively related to adolescent general violent victimization, negatively to two-parent family status, and positively to lower socioeconomic status. Neighborhood violence is not statistically significantly related to adult or adolescent hard drug or marijuana use, but is positively related to minority ethnicity and urban residence, marginally positively related to adolescent general violent victimization, and negatively related to rural residence and two-parent family status. In Table 3 for males, parental physical abuse is positively related to adult marijuana use and adolescent hard drug use (and marginally to adult hard drug use), positively to adolescent general violent victimization and urban residence, and negatively to upper/middle socioeconomic status. Inconsistent witnessing of parental violence is not statistically significantly related to either of the adult or adolescent illicit drug use measures, but is positively related to adolescent general violent victimization and minority ethnicity. Consistent witnessing of parental violence has only a marginally significant relationship with adult hard drug use and adolescent general violent victimization, a positive relationship with minority ethnicity, and negative relationships with two-parent family status and upper/middle socioeconomic status. Logistic Regression Results In the logistic regression analysis, we followed the protocol for logistic regression diagnostics suggested in Menard (2010). Based on tolerance statistics all above .69, there is no evidence of collinearity. Because all of the predictors were dichotomous, nonlinearity in the logit is not an issue (it is only relevant for predictors that have more than two values and are at least ordinal). As would expected for a good model fit, approximately 5 percent of the deviance residuals were greater than 2, and none was greater than 3, indicating that none of the cases was an outlier with respect to predicted and observed values. Dfbetas, which are standardized measures of the impact of individual cases on the logistic regression coefficients, were all less than .02, indicating that no cases stood out as being particularly influential. Table 4 presents the logistic regression results for the prevalence of adult hard drug use for females and males. For females, the strongest predictor (based on the odds ratio) of adult hard drug use is consistent witnessing of parental violence, followed by prevalence of adolescent general violent victimization, then upper/middle socioeconomic status, all in the positive direction. The positive relationship with upper/middle class parental socioeconomic status may reflect the inclusion of “pills” (nonprescription amphetamine and barbiturate use) in the drug use measure, and the greater access to such drugs for individuals with better financial resources. The other two influences are in the direction expected based on strain theory. For males, the positive relationship of consistent witnessing of parental violence is only marginally significant (p = .068), and the strongest influences are prior hard drug use in adolescence and exposure to neighborhood violence, both in the positive direction. In Table 5, for females, adult marijuana use is most strongly influenced by the prevalence of adolescent general violent victimization, prevalence of adolescent marijuana use, and consistent witnessing of parental violence, all in the positive direction (and there is a marginally significant positive relationship between urban residence and adult marijuana use). For males, statistically significant relationships are not all in the same direction, but if we follow the suggestion of Menard (2010) and convert the odds ratios to a canonical form (if the odds ratio, OR, is less than one, compute 1/OR = the odds ratio we would get if the coding of the predictor were reversed to produce a positive relationship), then the strongest influences on adult marijuana use, in descending order, are the positive relationship with prevalence of adolescent marijuana use, followed by the negative relationships with minority ethnicity, upper/middle class socioeconomic background, and two-parent family structure, respectively. If we consider only relationships that are significant at p ≤ .05, none of the AEV variables, including prevalence of adolescent general violent victimization, is significantly related to adult marijuana use.

36

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

Table 4 Logistic regression results for prevalence of adult hard drug use. Predictors

Odds ratio Femalea /maleb

Physical abuse Inconsistent witnessing parental violence Consistent witnessing parental violence Neighborhood violence Prevalence of adolescent hard drug use Prevalence of adolescent general violent victimization Nonwhite Urban Rural Two-parent family structure Upper/middle class Lower class Constant a b

1.719/1.041 .749/1.325 2.493/1.904 .699/1.972 .725/2.748 2.214/1.230 1.409/.688 1.644/1.381 .465/.654 1.213/1.156 2.091/.650 1.387/.932 .042/.083

Unstandardized coefficient (b) Female/male .542/.041 −.288/.281 .913/.644 −.358/.679 −.322/1.011 .795/.207 .343/−.374 .497/.323 −.766/−.425 .193/.145 .738/−.431 .327/−.070 −3.180/−2.485

Statistical significance (p) Female/male .166/.925 .543/.464 .005/.068 .357/.045 .349/.000 .012/.582 .361/.322 .119/.299 .101/.281 .582/.688 .042/.236 .327/.812 .000/.000

For females, RL 2 = .086, p = .000; and p = 0, p = 1.000 (two-tailed). For males, RL 2 = .077, p = .001; and p = −.013, p = .912 (two-tailed).

Table 5 Logistic regression results for prevalence of adult marijuana use. Predictors

Physical abuse Inconsistent witnessing parental violence Consistent witnessing parental violence Neighborhood violence Prevalence of adolescent marijuana use Prevalence of adolescent general violent victimization Nonwhite Urban Rural Two-parent family structure Upper/middle class Lower class Constant a b

Odds ratio Femalea /maleb

Unstandardized coefficient (b) Female/male

Statistical significance (p) Female/male

1.634/1.834 1.333/1.180 2.086/1.153 1.081/.639 2.211/2.910 2.269/1.478

.491/.607 .287/.165 .735/.143 .078/−.448 .793/1.068 .819/.391

.161/.103 .449/.621 .023/.654 .821/.189 .005/.000 .008/.218

.830/.469 1.812/1.220 .926/.637 1.583/.514 1.037/.508 1.427/.887 .023/.261

−.187/−.756 .594/.199 −.077/−.452 .459/−.665 .037/−.678 .356/−.120 −3.774/−1.345

.627/.023 .064/.472 .837/.154 .181/.025 .919/.029 .237/.624 .000/.004

For females, RL 2 = .090, p = .000; and p = .012, p = .912 (two-tailed). For males, RL 2 = .095, p = .000; and p = .016, p = .841 (two-tailed).

In addition to the analyses presented here, we performed additional analyses, not shown in detail here, but which may give some idea of the robustness of these findings to alternative models. First, we considered models that excluded prevalence of adolescent general violent victimization as a predictor. Without adolescent general violent victimization in the model, for females but not males, it is replaced by parental physical abuse as a statistically significant predictor, with the expected positive relationship between parental physical abuse and adult hard drug and marijuana use. None of the other substantive conclusions was affected by this change. Second, we examined models with frequency instead of prevalence of illicit drug use as both outcomes and predictors, using ordinary least squares regression analysis. At the .05 level of significance, none of the substantive conclusions was changed. Third, we examined the joint impact of the three different types of AEV by adding the prevalence of physical abuse, consistent witnessing of parental violence (given the evidence here and in Covey et al. (2013), and Menard et al. (2014), that consistent and inconsistent witnessing of parental violence produce different results, and that it is consistent and not inconsistent witnessing of parental violence that produces results most congruent with both theoretical expectations and prior research), and exposure to neighborhood violence into a single AEV index. The resulting index is a count of the number of types of AEV experienced, ranging from 0 to 3. Consistent with prior research using the same index (Menard et al., 2014) and with our focus specifically on the measures of exposure to violence specific to the family and neighborhood, adolescent general violent victimization was not included in this index, but was included as a separate variable in one analysis and excluded altogether in another (parallel to its inclusion in the analysis presented in Tables 4 and 5 and its exclusion in the first of the alternative analyses described above). The substantive results using the index were similar to the results presented here, in the sense that if any of the separate measures of AEV was statistically significant in Tables 4 and 5, the AEV index was likewise significant in the alternative analysis. Fourth, we considered a continuation ratio logit model (Menard, 2010) with a sequence from nonuse to illicit drug use to problem use, defined as in Elliott et al. (1989) and Menard (2002) as having problems with family, work, police, health,

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

37

accidents, or involvement in fights, as a result of either drug or marijuana use, respectively. The substantive results for illicit drug use, the first stage of the model, are the same as in Tables 4 and 5. However, the small number of problem users in middle adulthood (5 females and 10 males with problem drug use, 8 females and 23 males with problem marijuana use) mean that except for problem marijuana use for males, there are fewer cases with problem use than there are variables in the second stage of the sequential logit model, suggesting that the data are not adequate for that analysis, so those results are not detailed here. Discussion There are several limitations to the present study. The present research focuses on adolescent physical abuse, witnessing parental violence, and perceptions of neighborhood violence. Other forms of adolescent maltreatment such as sexual abuse, physical neglect, and emotional abuse, were not available from the NYSFS data set. It would be useful in future research (as in past research; e.g., Herrera & McCloskey, 2001; Widom, 1989) to use data that access a broader range of child maltreatment measures. The single-item dichotomous indicators used in the present research are different from the multiple item measures used in other research, but this concern may be mitigated somewhat by the similarity in rates of AEV presented here and in the NATSCEV study (Finkelhor et al., 2009). For future research, it would be interesting to examine frequencies of physical abuse, witnessing parental violence, and exposure to neighborhood violence, to compare those results with the present results for prevalence, but this goes beyond the data available in the NYSFS. Reliance on self-report data reduces the risk of false negatives (concluding that there was no AEV when in fact there was no officially recorded AEV but there was unrecorded AEV), but also risks some false negatives resulting from recall failure or concealment. It would be useful in future research (as has been done in past research) to supplement self-report data with official data, to provide a more comprehensive picture of adolescents’ experiences with exposure to violence and their subsequent behaviors, including but not limited to illicit drug use. It would also be interesting to examine the extent to which different family configurations, not only at one point in time but also stability and instability in those configurations, might moderate the relationship between AEV and illicit drug use, but we lack the information in the present data set to perform that analysis. Also not possible in the present study but desirable in future studies would be the examination of both adolescent and earlier childhood exposure to violence, to replicate the research of Ireland et al. (2002) and Thornberry et al. (2001) on the relative impact of adolescent as opposed to earlier childhood exposure to violence. As noted earlier, this study does not constitute a formal test of strain theory, but the present results clearly provide evidence consistent with the hypothesis derived from strain theory. Most of the significant relationships in Tables 2 and 3 are consistent with the research hypothesis, with strain theory, and with past research, including past findings that the impact of exposure to violence is different for males and females. Physical abuse is more strongly and consistently associated with substance use for females, mirroring past findings (e.g., Widom & White, 1997), and consistent reports of witnessing parental violence are significantly associated with adult illicit drug use for females but not males, while exposure to neighborhood violence is significantly associated with adolescent and adult hard drug use for males but not females. In the logistic regression analysis for females, if adolescent general violent victimization were excluded from the analysis, physical abuse would continue to be a significant predictor, but with both in the equation, it is the more general adolescent violent victimization rather than the more specific parental physical abuse that is the significant predictor of adult hard drug and marijuana use. This suggests that it is the broader experience of violent physical victimization, not just specific to the family context, that matters most for females (but, again, appears to be nonsignificant for males) in predicting female adult illicit drug use. Besides physical abuse and general violent victimization, in the logistic regression analysis, consistent reports of witnessing parental violence continue to be associated with adult hard drug and marijuana use for females and not (except for the marginal relationship with hard drug use) for males, while adolescent exposure to neighborhood violence continues to be significantly associated with adult hard drug use but not marijuana use for males and not females. It seems readily apparent that programs successful in reducing AEV (even without the side effect of reducing the risk of illicit drug use), and programs successful in reducing the continuity from adolescent to adult illicit drug use, are beneficial in their own right. Perhaps less readily apparent, but suggested by the results presented here, is that if programs are successful in identifying individuals likely to be at risk of illicit drug use in response to AEV (particularly, based on the present results, females exposed to violence in the home and males exposed to violence in the community), and are able to offer an effective alternative to self-medication during adolescence as a response to AEV, they should also reduce the risk of adult illicit drug use. Finally, these results indicate that the general hypothesis, that AEV will be positively related to adult illicit drug use is supported in the bivariate analysis for both female and male hard drug and marijuana use, and in the logistic regression analysis, controlling for prior drug use and adolescent general violent victimization, for hard drug use for both females (witnessing) and males (neighborhood), and for marijuana use for females (witnessing) but not for males. Given the adolescent trauma and stress associated with violence within the family, physical abuse, and later substance abuse, treatment approaches that focus on family dynamics are indicated. While this study did not assess different treatment approaches to adolescent substance abuse treatment, given evidence for strain theory, it does suggest that family-based approaches that focus on conflict and violence within the family warrant consideration. While interventions to reduce family violence do not typically focus on substance abuse, family-based approaches to reducing adolescent substance abuse and other delinquency in the adolescent, such as functional family therapy (Slesnick & Prestopnik, 2009) and multisystemic therapy (Henggeler, Clingempeel, Brondino, & Pickerel, 2002) not only have been designed to address substance abuse but

38

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

also are designed to engage family members and sometimes peers in the adolescents’ treatment. In addition to substance abuse, these approaches typically address a wide variety of issues, such as domestic violence, conflict, and other potential sources of adolescent stress. Research has shown that such family-based treatments can be highly effective, often more so than individual and group treatment approaches (Hogue & Liddle, 2009; National Institute on Drug Abuse, 2014). In this same vein, interventions that reduce the incidence of the problem before it occurs through prevention, reduction, or minimization of the harm of domestic violence, given the findings of this study, may reduce the risk of later long-term adolescent and adult substance abuse.

References Acosta, O. M., Albus, K. E., Reynolds, M. W., Spriggs, D., & Weist, M. (2001). Assessing the status of research on violence-related problems among youth. Journal of Clinical and Adolescent Psychology, 30, 152–160. Anda, R. F., Felitti, V. J., Bremner, J. D., Walker, J. D., Whitfield, C., Perry, B. D., Dube, S. R., & Giles, W. H. (2006). The enduring effects of abuse and related adverse experiences in childhood: A convergence of evidence from neurobiology and epidemiology. European Archives of General Psychiatry and Clinical Neuroscience, 256, 174–186. Agnew, R. (1985). A revised strain theory of delinquency. Social Forces, 64(1), 151–167. Barroso, C. S., Peters, R. J., Kelder, S., Conroy, J., Murray, N., & Orpinas, P. (2008). Youth exposure to community violence: Association with aggression, victimization, and risk behaviors. Journal of Aggression, Maltreatment and Trauma, 17, 141–155. Bonjean, C. M., Hill, R. J., & McLemore, S. D. (1967). Sociological measurement: An inventory of scales and indices. San Francisco, CA: Chandler. Brame, R., & Paternoster, R. (2003). Missing data problems in criminological research: Two case studies. Journal of Quantitative Criminology, 19, 55–78. Broidy, L. M. (2001). A test of general strain theory. Criminology, 39(1), 9–35. Chen, C., & Lin, K. (2009). Health consequences of illegal drug use. Current Opinion in Psychiatry, 22, 287–292. Covey, H. C., Menard, S., & Franzese, R. J. (2013). Effects of adolescent physical abuse, exposure to neighborhood violence, and witnessing parental violence on adult socioeconomic status. Child Maltreatment, 18, 85–97. Ehrensaft, M. K., Cohen, P., Brown, J., Smailes, E., Chen, H., & Johnson, J. G. (2003). Intergenerational transmission of partner violence: A 20-year prospective study. Journal of Consulting and Clinical Psychology, 71, 741–752. Eitle, D., & Turner, R. J. (2002). Exposure to community violence and young adult crime: The effects of witnessing violence, traumatic victimization, and other stressful life events. Journal of Research in Crime and Delinquency, 39, 214–237. Elliott, D. S., Huizinga, D., & Menard, S. (1989). Multiple problem youth: Delinquency, substance use and mental health problems. New York: Springer-Verlag. Fergusson, D. M., & Lynskey, M. T. (1997). Physical punishment/maltreatment during childhood and adjustment in young adulthood. Child Abuse and Neglect, 21, 617–630. Finkelhor, D., Turner, H., Ormrod, R., & Hamby, S. L. (2009). Violence, abuse, and crime exposure in a national sample of children and youth. Pediatrics, 124(5), 1411–1423. Gewirtz, A. H., & Edleson, J. L. (2007). Young children’s exposure to intimate partner violence: Towards a development risk and resilience framework for research and intervention. Journal of Family Violence, 22, 151–163. Harrison, L. (1995). The validity of self-reported data on drug use. Journal of Drug Issues, 35, 91–111. Harrison, P. A., Fulkerson, J. A., & Beebe, T. J. (1997). Multiple substance use among adolescent physical and sexual abuse victims. Child Abuse and Neglect, 21, 529–539. Henggeler, S. W., Clingempeel, W. G., Brondino, M. J., & Pickerel, S. G. (2002). Four-year follow-up of multisystemic therapy with substance abusing and dependent juvenile offenders. Journal of the American Academy of Child and Adolescent Psychiatry, 41, 868–874. Herrera, V. M., & McCloskey, L. A. (2001). Gender differences in the risk for delinquency among youth exposed to family violence. Child Abuse and Neglect, 25, 1037–1051. Herrenkohl, T. I., Sousa, C., Tajima, E. A., Herrenkohl, R. C., & Moylan, C. A. (2008). Intersection of child abuse and children’s exposure to domestic violence. Trauma, Violence and Abuse, 9, 84–89. Heyman, R. E., & Slep, A. M. (2002). Developmental issues in the impact of child maltreatment on later delinquency and drug use. Criminology, 40, 864–870. Hogue, A., & Liddle, H. A. (2009). Family-based treatment for adolescent substance abuse: Controlled trials and new horizons in services research. Journal of Family Therapy, 31, 126–154. Huang, D. Y. C., Evans, E., Hara, M., Weiss, R. E., & Hser, Y. (2011). Employment trajectories: Exploring gender differences and impacts of drug use. Journal of Vocational Behavior, 79, 277–289. Ireland, T. O., Smith, C. A., & Thornberry, T. P. (2002). Developmental issues in the impact of child maltreatment on later delinquency and drug use. Criminology, 40, 359–399. Jang, S. J. (1999). Different definitions, different modeling decisions, and different interpretations: A rejoinder to Lauritsen. Criminology, 37, 695–703. Kendall-Tackett, K. (2013). Treating the Lifetime Health Effects of Childhood Victimization (2nd ed.). Kingston, NJ: Civic Research Institute. Kitzmann, K. M., Gaylord, N. K., Holt, A. R., & Kenny, E. D. (2003). Child witness to domestic violence: a meta-analytic review. Journal of Consulting and Clinical Psychology, 71(2), 339–352. Lackey, C. (2003). Violent family heritage, the transition to adulthood, and later partner violence. Journal of Family Issues, 24, 74–98. Lynch, M. (2003). Consequences of children’s exposure to community violence. Clinical Child and Family Psychology Review, 6, 265–274. Margolin, G., & Gordis, E. B. (2000). The effects of family and community violence on children. Annual Reviews of Psychology, 51, 455–479. Menard, S. (1995). A developmental test of Mertonian anomie theory. Journal of Research in Crime and Delinquency, 32, 136–174. Menard, S. (1997). A developmental test of Cloward’s differential opportunity theory. In N. Passas, & R. Agnew (Eds.), The future of anomie theory (pp. 142–186). Boston: Northeastern University Press. Menard, S. (2002). Short- and long-term consequences of adolescent victimization. Washington, DC: Youth Violence Research Bulletin, Office of Juvenile Justice and Delinquency Prevention. Menard, S. (2010). Logistic regression: from introductory to advanced concepts and applications. Los Angeles, CA: Sage. Menard, S. (2012). Age, criminal victimization, and offending: Changing relationships from adolescence to middle adulthood. Victims & Offenders, 7, 227–254. Menard, S., & Elliott, D. S. (1993). Data set comparability and short-term trends in crime and delinquency. Journal of Criminal Justice, 21, 433–445. Menard, S., Weiss, A. J., Franzese, R. J., & Covey, H. C. (2014). Types of adolescent exposure to violence as predictors of adult intimate partner violence. Child Abuse and Neglect. Advance online publication, http://dx.doi.org/10.1016/j.chiabu.2014.02.001 Merton, R. K. (1938). Social structure and anomie. American Sociological Review, 3, 672–682. National Institute on Drug Abuse. (2014). Principles of adolescent substance use disorder treatment: A research-based guide. Washington, DC: National Institute on Drug Abuse. Perry, B. D. (2001). The neurodevelopmental impact of violence in childhood. In D. Schetky, & E. P. Benedict (Eds.), Textbook of child and adolescent forensic psychiatry (pp. 221–238). Washington, DC: American Psychiatric Press. Rebellon, C. J., & Van Gundy, K. (2005). Can control theory explain the link between parental physical abuse and delinquency? A longitudinal analysis. Journal of Research in Crime and Delinquency, 42, 247–274. Ruback, R. B., & Thompson, M. P. (2001). Social and psychological consequences of violent victimization. Thousand Oaks, CA: Sage.

S. Menard et al. / Child Abuse & Neglect 42 (2015) 30–39

39

Slesnick, N., & Prestopnik, J. L. (2009). Comparison of family therapy outcome with alcohol-abusing runaway adolescents. Journal of Marital and Family Therapy, 35, 255–277. Springer, K. K., Sheridan, J., Kuo, D., & Carnes, M. (2007). Long-term physical and mental health consequences of childhood physical abuse: Results from a large population-based sample of men and women. Child Abuse & Neglect, 31, 517–530. Substance Abuse and Mental Health Services Administration. (2011). Results from the 2010 National Survey on Drug Use and Health: Summary of national findings. Rockville, MD: SAMHSA. Tajima, E. (2004). Correlates of the co-occurrence of wife abuse and child abuse among a representative sample. Journal of Family Violence, 19, 399–410. Thornberry, T. P., Ireland, T. O., & Smith, C. A. (2001). The importance of timing: The varying impact of childhood and adolescent maltreatment on multiple problem outcomes. Development and Psychopathology, 13, 957–979. Widom, C. S. (1989). Child abuse, neglect, and adult behavior: research design and findings on criminality, violence, and child abuse. American Journal of Orthopsychiatry, 59, 355–367. Widom, C. S. (2014). Longterm consequences of child maltreatment. In J. E. Korbin, & R. D. Krugman (Eds.), Handbook of child maltreatment (vol. 2) (pp. 225–247). New York: Springer. Widom, C. S., & White, R. H. (1997). Problem behaviors in abused and neglected children grown up: prevalence and co-occurrence of substance abuse, crime, and violence. Criminal Behavior and Mental Health, 7, 287–310. Wolfe, D. A., Crooks, C. V., Lee, V., McIntyre-Smith, S., & Jaffe, P. G. (2003). The effects of children’s exposure to domestic violence: a meta-analysis and critique. Clinical Child and Family Psychology Review, 6(3), 171–187. Yates, T. M., Dodds, M. F., Sroufe, L. A., & Egeland, B. (2003). Exposure to partner violence and child behavior problems: A prospective study controlling for child physical abuse and neglect, child cognitive ability, socioeconomic status, and life stress. Developmental Psychopathology, 15, 199–218. Zinzow, H. M., Ruggiero, K. J., Hanson, R. F., Smith, D. W., Saunders, B. E., & Kilpatrick, D. G. (2009). Witnessed community and parental violence in relation to substance use and delinquency in a national sample of adolescents. Journal of Traumatic Stress, 22(6), 525–533.