General and specific predictors of behavioral

and emotional problems among adolescents

Based on a sample of 1,218 students in the 10th and 11th grades, 14 variables measuring behavioral and emotional problems were modeled as four factors via confirmatory factor analysis. The factors were labeled Polydrug Use, Delinquency, Negative Affect, and Academic Orientation. A similar four-factor structure was supported I year later, and a cross-lagged autoregressive structural equation model indicated high stability but also some significant cross-lagged predictions. For instance, higher levels of Delinquency at Wave I predicted higher levels of Wave 2 Polydrug Use and Negative Affect and lower levels of Academic Orientation. Significant predictors common to the four factors included stressful life events and family social support. Specific predictive relations were indicated for particular temperament attributes and specific problem factors (e.g., high Activity Level predicted Delinquency; Behavioral Inflexibility predicted Negative Affect), and for peer substance use variables and specific problem factors (e.g., Peer Drug Use predicted Polydrug Use).The findings are discussed with regard to the need to use research designs and statistical models that facilitate the evaluation of the generality and specificity of influences on behavioral and emotional problems during adolescence.

Adolescence is a phase in the life span characterized by transitions across numerous developmental domains, including challenges in biological (e.g., puberty), cognitive (e.g., issues of personal identity), and social (e.g., shift toward peers) spheres. Intertwined with these numerous developmental changes and associated increased levels of stressful life events (Newcomb, Huba, & Bentler, 1981; Windle, 1992a) is the expression of a range of behavioral and emotional problems (Dryfoos, 1990; Jessor & Jessor, 1977; Petersen et al., 1993). Some of these problems are age-normative and involve experimentation (e.g., alcohol use, minor delinquency), whereas the more severe expressions by others (e.g., precocious or promiscuous sexual activity and teen pregnancy, criminal arrest, depressive disorders) may impose constraints on both current functioning and probable future developmental trajectories (Harrington, Rutter, & Fombonne, 1996).

The research literature is replete with examples of high rates of internalizing and externalizing problems among children and adolescents (Achenbach, 1991; Loeber, Farrington, Stouthamer-Loeber, & Van Kammen, 1998). Contingent on the research goals of particular studies, some investigators have focused on narrowband factors (e.g., withdrawal, depressive symptoms, anxiety symptoms, attention deficits), whereas others have focused on broadband factors such as internalizing and externalizing problems that consist of composite scores derived from narrowband factors. In addition to variation in studies with regard to focusing on narrow- or broadband factors, there is also considerable heterogeneity across studies with regard to which domains or areas have been most critical to measure for which age groups and for which research purposes (Achenbach, 1991; Jessor & Jessor, 1977; Loeber et al, 1998: White & Labouvie, 1994). In this study, we focused on the four domains of polydrug use, delinquency, negative affect, and academic orientation. These domains are not intended to provide an exhaustive mapping of all relevant domains of psychopathology and misbehavior among adolescents but rather to provide a reliable assessment of several critical domains that have been reported consistently to be of relevance to youth functioning in adolescent school populations (Dryfoos, 1990; Jessor & Jessor, 1977: Loeber et al., 1998). Moreover, we focused on general (e.g., family support is significantly associated with all four outcomes) and specific (e.g., alcohol-using peers are significantly associated only with polysubstance use) predictors of these four outcome variables. Findings regarding common and specific predictors of multiple problem behaviors among youth may be used to understand the etiology, or causes, of these problems, as well as to facilitate the design of interventions (Weiss, Susser, & Catron, 1998).

For example, findings by White and Labouvie (1994) indicated that adolescents who rated high on both delinquency and drug use factors could be distinguished from adolescents who rated high on only one of these two problems with regard to heightened levels of temperamental characteristics related to impulsivity, low sell-control, and emotional instability or neuroticism. On the basis of data from the National Youth Survey, Elliott, Huizinga, and Menard (1989) reached the following conclusion with regard to delinquency, drug use, and mental health:

Loeber et al. (1998) also reported that a unitary model of externalizing problems does not adequately account for types of delinquent offenders (e.g., overt, violent offenders vs. covert, minor delinquents) or differences in terms of time of onset and differential life-course trajectories among offender subtypes.

In the current study, our objectives were threefold. First, data on 1.218 "middle" adolescents (11th and 12th graders) were used to evaluate a four-factor model of 14 manifest variables representing the behavioral and emotional problem domains of polysubstance use, general delinquency, academic orientation (i.e., educational performance and expectations for educational achievement), and negative affect (e.g., depressive and anxiety symptoms). The second objective was to test the stability and cross-lagged prediction of derived factors across a 1-year interval. This facilitated both the evaluation of the stability of these problem behaviors across a 1-year period and the prospective evaluation of the possible impact of high levels of one problem behavior (e.g., delinquency) on a second problem behavior (e.g., polysubstance use), and vice versa. The third objective was to test the specificity versus generality of prediction of different behavioral and emotional problems (e.g., polysubstance use and negative affect) by several prominent risk factors that have been identified in the literature (Hawkins, Catalano, & Miller, 1992: Windle, 2003). Although not exhaustive of all possible risk factors, predictors were selected from the domains of temperament, family support, substance-using peers, stressful life events, and family income. The specificity-generality of prediction has implications for both conceptual and analytic models of adolescent dysfunction, and for prevention and treatment. For example, if the same variables are equally predictive of all behavioral and emotional problems, more universally targeted interventions may be appropriate and even optimal. However, more tailored interventions may be optimal if different variables predict different outcome variables.

METHOD

Participants

The sample consisted of 1,218 high school juniors (55%) and seniors (45%) recruited from three homogeneous, nondenominational, suburban high schools in western New York . Fifty-one percent (616) of the sample were girls and 49% (602) were boys. The average age of the respondents at the first measurement occasion was 15.51 years (SD = 0.64), and 98% were White. Sixty-seven percent of the sample was Catholic, 21% Protestant, and 11% "other." Ninety-two percent of the fathers and 43% of the mothers were employed full-time outside the home (31% of mothers were employed part-time outside the home). Fathers completed an average of 13.90 years of education (SD = 2.32), and mothers completed an average of 13.25 years (SD = 2.18). The average number of children per family was three. The mean family income was approximately $40,000 and only 2.2% of the sample reported a family income less than $12,000. Eighty-eight percent of the adolescents' primary caregivers were currently married, 12% were divorced, and 1% were widowed. Approximately 76% of high school students eligible for the study participated. Comparisons of alcohol use for the sample used in this study with alcohol use among adolescents participating in national (Johnston, O'Malley, & Bachman, 1991) and New York State (Barnes & Welte, 1986) surveys indicated high levels of similarity (see Note 1).