VON E. NEBBITT AND AJITA M. ROBINSONINTRODUCTION
SINCE ITS INCEPTION, THE CENTRAL theme of research on low-income urban African American youth has been deficits and pathologies, with a relative exclusion of capacities and strengths. In 1999, Zimmerman, Ramirez-Valles, and Maton noted that much more is known about the causes of psychopathology among African American youth than about how and why some of these youth become well-functioning citizens. A review of empirical papers suggests that research, with few exceptions, has not deviated from this trend. One exception has been the proliferation of studies on self-efficacy and its effects on reducing risk behavior (Goh, Primavera, & Bartalini 1996; Jonson-Reid et al. 2005) and increasing well-being (Connell, Spencer, & Aber 1994) in African American youth. Studies have also emerged that attempt to identify familial and community correlates of self-efficacy in African American adolescents (Lombe, Nebbitt, & Mapson 2009; Nebbitt 2009).
This emerging body of research on African American youth has the potential to move the discussion beyond a focus on deficits to include a focus on positive aspects in these youth. Still, an obvious gap in this literature is research that attempts to identify characteristics in samples of urban African American youth that have the potential to increase their life chances. It is very likely that prosocial attitudes and beliefs co-occur in youth, as mental health symptomology (e.g., depression, anxiety) also co-occurs in youth.
Using latent profile analysis, this chapter explores whether there are subgroups of youth based upon their attitudes toward deviance and their efficacious beliefs. This chapter also assesses how membership in these subgroups is influenced by environmental factors (community and family), which in turn influences youths’ symptoms and behavior.
SELF-EFFICACY AND ATTITUDES TOWARD DEVIANCE
Self-efficacy is defined as a person’s belief about their ability to organize and execute courses of action necessary to achieve a specific goal (Bandura 1977). Individuals with strong efficacious beliefs are more confident in their capacity to accomplish their desired goals. Efficacious beliefs have a significant impact on youths’ goals and accomplishments by influencing their personal choices, motivation, patterns, and emotional reactions. Generalized perceived self-efficacy also determines the level of effort and persistence a person demonstrates in the face of adversity. Self-efficacy is positively related to persistence—a trait that allows us to gain corrective experiences that reinforce our sense of self-efficacy.
In addition to efficacious beliefs, adolescents’ attitudes toward deviant behaviors are critical to their involvement in antisocial and health-risk behaviors. Attitude toward deviant behavior refers to the degree to which a person has a favorable or unfavorable evaluation or appraisal of the behavior in question (Ajzen 1991). The likelihood of an adolescent performing behaviors of different kinds can be predicted with high accuracy from attitudes toward the behavior. However, the failure of attitudes to predict specific behaviors directed at the target of the attitude has produced calls for abandoning the attitude concept (Wicker 1969). However, research by Kenneth Miller (1975) suggested that a situational multiattribute attitude model may allow more accurate assessment of behavior prediction on the basis of attitudes.
Because of the aggregated nature of adolescents’ attitudes, their views toward deviance and their efficacious beliefs may predict specific behaviors, which in turn are predicted by contextual factors.
CORRELATES OF SELF-EFFICACY AND ATTITUDES TOWARD DELINQUENCY
Efficacious Beliefs
Evidence suggests that self-efficacy is inversely related to delinquent behavior in youth. Specifically, youth who reported higher self-efficacy exhibited lower delinquent behavior (Chung & Elia 1996; Donnellan et al. 2004). Research has also found individual characteristics related to self-efficacy in youths (Jenkins, Goodness, & Buhrmester 2002). For example, a youth’s mental health status has been indicated as a strong correlate of self-efficacy (Francis et al. 2007; Maciejewski, Prigerson, & Mazure 2000). These studies indicate that higher self-efficacy is associated with lower scores of depression. Likewise, self-efficacy is said to be inversely related to stressful life events, exposure to violence, attachment anxiety, and feelings of loneliness (Wei, Russell, & Zakalik 2005).
The relationship between peer affiliation and an adolescent’s behavior is well documented (Boyer, Tschann, & Shafer 1999; Henrich et al. 2000). Although it is clear that peers play an important role in adolescents’ involvement in delinquent behavior, little is known about the role peers play in influencing an adolescent’s self-efficacy. Evidence does suggest, however, that self-efficacy waxes and wanes over time, by social comparison with peers, especially peers with similar capabilities (Center for Positive Practices 2000). Indeed, group norms, aspiration, and performance have been found to influence collective and individual efficacy (Prussia & Kinicki 1996). Moreover, evidence suggests that youth tend to affiliate with peers who share the same interests and values, thus ensuring the promotion of self-efficacy in directions of mutual interest (Bandura 1994; Center for Positive Practices 2000).
Attitudes Toward Deviance
Previous research among adolescents has found favorable attitudes toward deviance associated with problem behavior (Heimer & Matsueda 1994; Huesmann & Guerra 1997; Zhang, Loeber, & Stouthamer-Loeber 1997). Existing evidence suggests that adolescents with predispositions to delinquency are more likely to engage in delinquent behavior (Moffitt & Caspi 2001). The general assumption is that an increase in tolerance toward delinquency often precedes the initiation of delinquent acts (Pardini, Loeber, & Stouthamer-Loeber 2005). Evidence also indicates that the influence of attitudes on behavior may be contingent upon a number of covariates, including the environment and the mental health status of the youth (Nebbitt, Lombe, & Williams 2008). Indeed, the environment in poor neighborhoods is often charged with factors that may heighten mental health symptoms, such as depression (Nebbitt & Lombe 2007). For example, feelings of depression may be deepened in situations that are perceived as threatening, such as witnessing community violence, the death of a friend or family member, verbal insults, or physical assault. Such incidences, in certain individuals, may increase pressure to cope; engagement in a health-risk behavior, such as delinquency, may be a possible response.
Co-Occurring Efficacious Beliefs and Attitudes Toward Deviance
Research has found individual characteristics associated with self-efficacy in youths (Francis et al. 2007; Jenkins, Goodness, & Buhrmeister 2002; Maciejewski, Prigerson, & Mazure 2000). Evidence suggests that an adolescent’s attitude toward deviance is a gauge of their efficacious beliefs (Huesmann & Guerra 1997; Zhang, Loeber, & Stouthamer-Loeber 1997). Youth who report higher levels of self-efficacy exhibited lower levels of delinquent behavior (a good proxy of an adolescent’s norm-violating attitude; Conner et al. 2004; Galilner, Evans, & Weiser 2007). Considering the co-occurrence of self-efficacy and attitudes toward deviance, it is likely that there are typologies of youth based upon their attitudes toward norm-violating acts and their efficacious beliefs. The evidence reviewed previously provides a sound empirical foundation for an investigation into variations in urban youth based upon their attitudes and beliefs.
THEORETICAL ORIENTATION
Social context plays a critical role in an adolescent’s development. African American youth living in public housing face several challenges due to a high concentration of poverty and social problems that exist in many public housing neighborhoods. It is important to note, however, that despite living in public housing, many African American youth manage to do well (Furstenberg et al. 1999). Most remain in school, graduate, and avoid significant life-compromising behavior (Smith et al. 1995). Coll et al. (1996) argued that low-income urban communities are simultaneously promoting and inhibiting, which contributes to the multifinality exhibited by youth in public housing. Indeed, African American adolescents in urban public housing are influenced by positive and negative aspects of their community. Surely, they need an array of internal resources in addition to social support to navigate their living environments and to survive (Dodge & Frame 1982).
The Integrated Model of Adolescent Development in Public Housing Neighborhoods (detailed in chapter 3) provides a framework for explicating and investigating how positive outcomes in youth are achieved within the context of public housing neighborhoods. A section of the Integrated Model posits that adolescents’ internalized resources (e.g., self-efficacy, attitudes toward deviance) are directly influenced by the positive and negative aspects of the public housing neighborhood. The model argues that the impact of negative aspects of the neighborhood depends on the level of protective factors available to the youth. The model further posits that an adolescent’s internalized resources (e.g., efficacious beliefs, attitudes toward deviance), in turn, are related to symptoms and behaviors. This chapter tests these hypothesized relationships of the Integrated Model on Adolescent Development in Public Housing Neighborhoods.
This chapter has three goals. First, it explores whether there are latent classes of youth based on their attitudes toward deviance and their efficacious beliefs (e.g., self-efficacy). Second, it examines how class membership is associated with adolescents’ symptoms, behaviors, and affiliates. Third, it assesses whether contextual risk and protective factors and their interactions predict youths’ membership in each latent class. This chapter advances three research questions:
1. What are the underlying variations in efficacious beliefs and attitudes toward deviance among African American adolescents living in urban public housing?
2. How are variations in efficacious beliefs and attitudes toward deviance related to African American adolescents’ symptoms and behavior?
3. How are variations in efficacious beliefs and attitudes toward deviance associated with perceptions of peers, family, and community?
METHODS
The sample for this chapter includes youth from Washington DC, Philadelphia, and New York City. St. Louis was excluded from this analysis, as there were no data on self-efficacy from the St. Louis sample.
Indicators of Efficacious Beliefs and Attitudes Toward Deviance
As table 5.1 summarizes, 24 items were used to identify latent classes across the pool of study participants based on their self-efficacy and attitudes toward deviance. To assess efficacious beliefs, youth completed the General Perceived Self-Efficacy Scale (Schwarzer & Jerusalem 1995). To assess attitudes toward deviance, youth completed the National Youth Survey’s Attitudes Toward Delinquency Subscale (Elliot 1987).
TABLE 5.1 Summary of Attitudes and Efficacy Indicators (n = 660)
| ATTITUDES TOWARD DEVIANCEa (RANGE, 1–4) |
MEAN |
SD |
| 1. Use marijuana? |
1.75 |
1.01 |
| 2. Damage or destroy property that does not belong to you? |
2.04 |
1.14 |
| 3. Steal something worth less than $5? |
2.15 |
1.24 |
| 4. Hit or threaten to hit someone for no reason? |
2.70 |
1.29 |
| 5. Use alcohol? |
1.39 |
0.80 |
| 6. Break into a vehicle or building to steal something? |
1.75 |
1.06 |
| 7. Sell drugs such as heroin, cocaine, and crack? |
1.62 |
0.99 |
| 8. Steal something worth $5? |
1.71 |
1.00 |
| 9. Get drunk once in a while? |
1.59 |
0.99 |
| 10. Give or sell alcohol to kids under 18? |
1.57 |
0.96 |
| 11. Attack someone with the idea of seriously hurting or killing them? |
1.48 |
0.92 |
| 12. Exceed the speed limit by 10 or 20 mph? |
1.55 |
0.89 |
| 13. Use force to get money or things from people? |
1.91 |
1.19 |
| 14. Hit and injure their girlfriend or boyfriend? |
2.34 |
1.27 |
| GENERALIZED SELF-EFFICACY (RANGE, 1–4) |
|
|
| 1. I can always manage to solve difficult problems if I try hard enough. |
2.40 |
1.15 |
| 2. If someone goes against me, I can find a way to get what I want. |
2.23 |
1.05 |
| 3. I am sure that I can accomplish my goals. |
2.78 |
1.22 |
| 4. I am confident that I could handle unexpected events. |
2.46 |
1.13 |
| 5. Thanks to my abilities, I can handle unexpected situations. |
2.45 |
1.13 |
| 6. I can solve most problems if I put in the necessary effort. |
2.57 |
1.16 |
| 7. I remain calm when facing problems because I can rely on my coping abilities. |
2.22 |
1.08 |
| 8. When I am faced with a problem, I can find several solutions. |
2.34 |
1.05 |
| 9. If I am in trouble, I can think of a good solution. |
2.39 |
1.11 |
| 10. I can handle whatever comes my way. |
2.29 |
1.12 |
aItems began with, “How wrong is it for someone your age to …”
The covariates were as follows:
1. Exposure to Delinquent Peers and Self-Reported Delinquency (Elliot 1987)
2. Anxiety Sensitivity Index (Peterson & Reiss 1987)
3. Center for Epidemiological Study Depression Scale (Radloff 1977)
4. Alcohol, tobacco and marijuana use, assessed using the Centers for Disease Control and Prevention’s Youth Risk Behavior Survey (2011)
5. Maternal and Paternal Encouragement Parental Attitude Measure (Lamborn et al. 1991)
6. Domestic violence, assessed using the Family Conflict Scale (Barbarin, Richter, & deWet 2001)
7. Neighborhood hazard and cohesion, assessed using the Subjective Neighborhood Scale (Aneshensel and Sucoff 1996)
8. Survey of Exposure to Community Violence: Self-Report Version (Richters & Martinez 1990)
Analytic Procedures
As previously mentioned, we employed latent profile analysis (LPA) to determine the optimal number of subgroups or classes. Because our indicator variables were ordinal, LPA is the appropriate technique. If they were dichotomous, then we would have employed latent class analysis. Although housing units were large and few in number, robust standard errors were employed to correct for biasing due to any data nesting. Once the optimal number of subgroups was identified, we examined class differences based on the aforementioned covariates. Chi-square and analysis of variance (ANOVA) employing Bonferroni post-hoc tests were used for this examination. Finally, a multinomial logistic regression model was executed, with subgroup membership serving as the dependent variable to further refine and interpret the effects of covariates on the identified subgroups.
With respect to the execution of our LPA models, our analysis was carried out in an exploratory fashion using LatentGOLD version 4.5 software (Vermunt & Magidson 2005). Rather than testing a class solution specified a priori, analyses examined the fit of a series of different models. Missing values were assumed to be missing at random and were imputed using an expectation maximization algorithm. A single-class model was examined first, and classes were added one at a time until four classes were completed. The empirical fit of each model was determined on the basis of several fit indices, including the Bayesian Information Criterion (BIC) and the Akaike Information Criterion (AIC). As model fit improves, the values of the BIC and AIC decrease. The conceptual fit of models is critical. Here, it was examined by using visual representations of the indicators and assessing their theoretical interpretability and practical implications.
RESULTS
Descriptive Summary
As table 5.2 shows, the mean age of the study sample was 15.4 years (SD = 2.42). The sample was 47.7 percent female (n = 313) and was composed predominately of African Americans (82.8 percent) and biracial youth (12.9 percent). The sites from which the study sample was drawn were New York City–Queens (n = 237, 35.9 percent; and n = 110, 16.7 percent), Philadelphia (n = 149, 22.6 percent), and Washington, DC (n = 164, 24.8 percent).
TABLE 5.2 Sociodemographic Characteristics Across Research Sites (n = 660)
| CHARACTERISTIC |
DATA |
| Age, mean (SD) |
15.4 (2.42) |
| Gender: female |
313 (47.7) |
| Race/ethnicity |
|
| African American |
545 (82.8) |
| Biracial |
85 (12.9) |
| Othera |
28 (4.2) |
| City |
|
| New York City (North Site) |
237 (35.9) |
| New York City (South Site) |
110 (16.7) |
| Philadelphia |
149 (22.6) |
| Washington, DC |
164 (24.8) |
Data are n (%) unless otherwise noted.
aIncludes white, Latino/Hispanic, Native American, and Asian.
Latent Profile Analysis
All 24 indicator variables were significant contributors to distinguishing classes. The empirical fit indices reported in table 5.3 show that BIC and AIC values decrease as additional classes are added. A four-class solution exhibited the best fit with respect to BIC and AIC values, entropy, and low class error. To test whether the three-class solution was a better fit to the data, a conditional bootstrap simulation with 1,000 iterations was executed to compare the four-class solution with the three-class solution. Results (table 5.3) showed that the four-class solution was a superior fit for the data (−2LL differential = 1158.82, p < .0001).
TABLE 5.3 Fit Indices for Latent Classes (1–4) (n = 660)
BIC, Bayesian Information Criterion; AIC, Akaike Information Criterion; L2, log squared; NA, not applicable.
aModel solution chosen.
The conceptual fit of the models was determined through visual inspection and meaningfulness. This involved plotting the estimated mean values for each indicator variable by each class. Results (figure 5.1) show that classes are clearly distinguishable and are composed of a moderate attitudes and moderate efficacy subgroup (class 1, n = 199), a high attitudes and high efficacy subgroup (class 2, n = 185), a high attitudes and low efficacy subgroup (class 3, n = 148), and a low attitude and low efficacy subgroup (class 4, n = 128). In sum, the four-class model was conceptually clear and theoretically important.
FIGURE 5.1 Mean plots for scores on indicators of efficacy and attitudes across classes.
Comparisons of Subgroups
As shown in table 5.4, chi-square tests revealed several proportional differences in class composition. Classes differ to a statistically significant degree in gender (χ2 [3] = 18.84, p < .000), having tried alcohol (χ2 [6] = 22.13, p < .001), and having tried marijuana (χ2 [6] = 15.00, p < .05). There were no compositional differences across classes with regard to currently being in school or having tried tobacco. With respect to gender, results show that adolescents in class 2 (high efficacy and high unfavorable attitudes) were predominately female (59 percent), and class 4 (low efficacy and low favorable attitude) were predominately male (66 percent). With regard to marijuana use, adolescents in class 2 (high efficacy and high unfavorable attitudes) reported the highest percentage of youth who had not tried marijuana (54 percent) and class 4 (low efficacy and low favorable attitude) reported the highest percentage of youth who had tried marijuana (59 percent).
TABLE 5.4 Tests of Differences Among the Four-Class Solution Using Chi-Square and One-Way ANOVA (n = 660)
Class 1 was moderate attitudes and moderate efficacy, class 2 was high attitudes and high efficacy, class 3 was high attitudes and low efficacy, and class 4 was low attitudes and low efficacy. Post-hoc comparisons (Bonferroni) were conducted for all ANOVAs.
ANOVA, analysis of variance; NS, not significant.
aPercentage not in school and have not used.
bClasses 1 and 2 are different.
cClasses 2 and 3 are different.
dClasses 1 and 3 are different.
eClasses 1 and 4 are different.
fClasses 2 and 4 are different.
One-way ANOVA detected numerous mean differences across latent classes (table 5.4). Delinquent behavior (F = 35.14, p < .001), exposure to deviant peers (F = 25.37, p < .001), depressive symptoms (F = 9.13, p < .000), and adultification (F = 7.96, p < .000) differ significantly across latent classes. Anxiety sensitivity did not differ across latent classes. Bonferroni post-hoc comparisons revealed significant differences between classes. That is, class 2 (high efficacy and high unfavorable attitudes) reported significantly lower mean scores on delinquency (M = 18.03, SD = 4.66) compared with classes 1 and 4. Delinquency did not differ between class 2 and 3. However, class 2 adolescents also reported significantly lower exposure to deviant peers (M = 22.46, SD = 7.34) compared with classes 1 and 4; exposure to delinquent peers did not differ between class 2 and 3. Adolescents in class 2 (M = 14.92, SD = 8.80) reported significantly lower depressive symptoms compared with classes 1 and 4, but not class 3. With respect to adultification, class 2 (M = 7.11, SD = 2.34) and class 3 (M = 7.21, SD = 2.37) reported significantly higher levels of adultification than classes 1 and 4. Class 2 did not differ from 3, and class 1 did not differ from class 4.
Covariate Effects on Class Membership
To further explore and refine the adolescent development model, we used theoretically proposed community- and family-level risk (e.g., exposure to community violence, neighborhood hazard, domestic conflict) and protective (e.g., social cohesion, extended and fictive family, maternal and paternal encouragement) factors across classes in a multinomial logistic regression analysis to assess their ability to predict class membership. This regression, using a simultaneous entry, facilitated direct and indirect tests of variables in predicting class membership. Class 2 (high efficacy and high unfavorable attitudes) serves as the reference group.
The results of the multinomial logistic regression are displayed in table 5.5. Results indicate several statistically significant predictors of class membership (χ2 [18] = 144.47, −2LL = 1.55, p < .000). The model correctly classified 42 percent of the classes, with class 2 (67 percent) representing the class with the highest percent of cases correctly classified. Members in class 1 were significantly less likely to have fictive kinship present in the housing neighborhood compared with members of class 2. Compared with class 2, classes 1, 3, and 4 were significantly less likely to report maternal encouragement, and class 3 was significantly more likely to report receiving paternal encouragement compared with class 2. Domestic conflict increased the likelihood of membership in class 3 compared with class 2. Classes 1, 3, and 4 were significantly more likely to have been victimized by community violence compared with class 2. On the other hand, classes 3 and 4 were significantly less likely to have witnessed community violence.
TABLE 5.5 Multinomial Logistic Regression: Criterion—Class Membership
OR, odds ratios; CI, 95% confidence intervals; *p < .05, **p < .01, ***p < .001.
aReference: no fictive kinships in the neighborhood.
DISCUSSION
Historically, research has focused on the pathologies and deficits exhibited by African American youth while ignoring the signs of resiliency evident in this population. A section of the adolescent development model posits that promotive aspects of public housing neighborhoods contribute to increased self-efficacy and less favorable attitudes toward deviance; efficacious beliefs and conventional attitudes, in turn, are associated with positive emotionality and prosocial behavior. To test this section of the adolescent development model, this chapter explored variations in latent classes of adolescents based on their self-efficacy and attitudes toward deviance; assessed how adolescents differ on their depressive symptoms, anxiety sensitivity, delinquent behavior, and exposure to deviant peers based on their class membership; and examined how inhibiting and promotive aspects of public housing neighborhoods predict membership in each latent class.
Findings support the model in part. For example, adolescents’ underlying profile of attitudes and efficacy differentiated their behavior, mental health symptoms, and peer affiliations. Youth with conventional attitudes and more efficacious beliefs used less alcohol and marijuana than other youth. Youth with conventional attitudes and more efficacious beliefs were engaged in less delinquent behavior themselves and were affiliated with less youth involved in antisocial behavior, compared with youth with more favorable attitudes toward deviance and low efficacious beliefs. Furthermore, adolescents with less favorable attitudes toward deviance and high or low self-efficacy also experienced far fewer symptoms of depression than youth with more favorable attitudes toward deviance and moderate to low efficacious beliefs.
Higher self-efficacy appears to be a moderating effect against depressive symptoms, which is consistent with the research regarding protective factors being a mitigating factor against health risks (Zimmerman, Ramirez-Valles, & Maton 1999; DiClemente et al. 1996). Furthermore, this study suggests, although does not empirically support, that the level of exposure to deviant peers is not as significant for adolescents who are highly efficacious with high attitudes toward deviance, in comparison to moderate and low efficacious peers. The co-existence of high efficacy and high attitudes toward deviance supports the existence of prosocial attitudes and beliefs co-occurring in youth. Perhaps the most intriguing implication of this study with regard to understanding the relationship of self-efficacy and attitudes toward deviance is the identification of several classes of efficacy and attitudes toward deviance among the African American adolescents in this sample, which appear to affect the level to which youth engage in risky behaviors, the impact of peer associations, and the exhibition of depressive symptomology.
