Measuring Social Support and School Belonging in Black/African American and White Children

Abstract

Objective:

To determine the suitability of the Elementary School Success Profile for Children (ESSP-C) for assessment and comparison of social support and school belonging between Black/African American and White students.

Methods:

Multiple-group confirmatory factor analysis and invariance testing were conducted to determine the ESSP-C’s validity for use with Black/African American and White students. Latent mean comparisons were performed to determine statistically significant differences in school belonging and social support between racial/ethnic groups.

Results:

The ESSP-C demonstrated partial measurement invariance at a level (93% invariant) that supports the validity of the measure for Black/African American and White students. Black/African American students reported a significantly higher mean level of school belonging compared to White students.

Conclusion:

The ESSP-C can be used to make valid assessments and comparisons of social support and school belonging between Black/African American and White students, which may be useful in guiding school social work practice and intervention.

Keywords

factor analysis psychometric study children population Blacks African Americans quantitative methodological article schools social support school belonging

Social support’s strong associations with various aspects of health and well-being have made fostering socially supportive relationships a key component of many social work interventions. Social support refers to a social network’s ability to provide both psychological and material resources that enhance coping, learning, and general well-being (Cohen, 2004). As alluded to in the definition, social support is a multidimensional construct, and resources provided under its aegis can take the form of instrumental support (provision of material support or task assistance), informational support (providing relevant information), or emotional support (Cohen, 2004).

For children, parents or caregivers, teachers, and peers represent the most important sources of social support (Cauce & Srebnik, 1990). Increasing children’s sense of social support can be a powerful mechanism through which school social workers can reach myriad goals of school social work practice, such as removing barriers to learning, creating a safe school environment for all students, and preventing individual-level personal and school maladjustment (Demaray & Malecki, 2002, 2006; National Association of Social Workers, 2012). School-aged children who perceive high levels of social support from key adults and from peers are more likely to experience positive mental health (Stewart & Suldo, 2011), resilience in the face of peer victimization and other challenges (Demaray & Malecki, 2002, 2006), increased academic achievement (Estell & Perdue, 2013; Furrer & Skinner, 2003), and higher levels of school engagement (Estell & Perdue, 2013; Furrer & Skinner, 2003; Hamre & Pianta, 2005).

Consistent with a systems theory perspective (Bronfenbrenner, 1979), supportive social relationships, student engagement, and academic achievement mutually influence each other in a feedback loop (Hughes, Luo, Kwok, & Loyd, 2008). Enhancing one component of the process, such as increasing emotional engagement in school by fostering supportive peer relationships, can affect the other components and trigger a cycle of recursive benefits. For example, when supportive teacher–student relationships are established early in a child’s academic career, they not only improve the quality of daily classroom interactions, but actually reduce the risk of early and persistent underachievement (Hamre & Pianta 2005). The mutual, recursive nature of the relationships between social support and beneficial outcomes provides multiple points and mechanisms that can be leveraged in interventions to boost children’s academic achievement and psychosocial well-being.

Influence of Parent or Caregiver Support on Academic Success and Child Well-Being

Although parent/caregiver social support is a strong predictor of academic achievement for all students (Estell & Perdue, 2013; Furrer & Skinner, 2003), high levels of parent or caregiver support may be especially important for students coming from low-socioeconomic status (SES) families due to its ability to buffer the negative impact of factors such as poverty on academic achievement (Malecki & Demaray, 2006). In addition to its positive direct association with academic achievement, parent/caregiver social support is also significantly associated with behavioral engagement at school (Estell & Perdue, 2013). Because engagement is a main mechanism through which motivation affects learning, an increase in behavioral engagement or learning behaviors may provide an additional pathway through which parent or caregiver support promotes academic success (Estell & Perdue, 2013). Parent or caregiver support is sometimes operationalized as parental warmth, a definition that emphasizes the provision of emotional support. High parental warmth may help to promote positive social and academic outcomes through its demonstrated ability to reduce anxiety. Conversely, parent–child relationships that are low in parental warmth are associated with poor academic achievement and behaviors (Bodovski & Youn, 2010). Another important feature of parent or caregiver support is that it may promote a “readiness for socialization” that enables children to develop and benefit from supportive relationships with other adults and peers (Furrer & Skinner, 2003).

Influence of Peer Support on Academic Success and Child Well-Being

Peer support is particularly associated with increases in affective or emotional engagement, which consists of children’s feelings and perceptions about school and learning (Estell & Perdue, 2013; Furrer & Skinner, 2003). However, peer support also has a unique and direct impact on academic achievement (Cappella, Kim, Neal, & Jackson, 2013). Apart from generating positive emotions about school, relationships with peers may support achievement by facilitating access to learning-related resources and activities (Wentzel & Wigfield, 1998) and by contributing to the sense that the classroom is an emotionally safe space to take the risks necessary for learning (Duke, Borowsky, & Pettingell, 2011; Smith, Osgood, Caldwell, Hynes, & Perkins, 2013).

Peer social support also maintains several important associations with aspects of peer victimization and bullying. Both victims of peer aggression and bully victims report lower levels of social support from peers, and several widely used bully violence prevention programs focus on improving peer support to prevent and intervene on bullying incidents (Demaray & Malecki, 2005).

Influence of Teacher Support on Academic Success and Child Well-Being

Of the three major social support sources in children’s lives (parent or caregiver, peer, and teacher), teacher support demonstrated the strongest effect on children’s academic achievement and overall social well-being in Chu, Saucier, and Hafner’s (2010) meta-analysis of social support effects. Social support from teachers consists of two components: emotional support and instructional support. In the classroom context, emotional support consists of an overall feeling of classroom warmth, sensitivity and responsivity to individual children, positive affect and feedback, and a child-centered philosophy. Instructional support involves intensive, task-focused teacher–student interactions that facilitate children’s higher order thinking and cognition (Hamre & Pianta, 2005). Similar to other forms of support, teacher support not only directly affects student achievement, but also does so indirectly by promoting student engagement. Teacher support has been demonstrated to increase effortful engagement, the type of school engagement that consists of focusing, persisting, and putting forth one’s best effort on a task (Hughes et al., 2008) as well as behavioral and emotional engagement (Furrer & Skinner, 2003).

The Role of School Belonging

In addition to their direct effects on children’s well-being and academic performance, supportive relationships in school settings also have an indirect effect on well-being and academic success by influencing students’ sense of school belonging (McMahon, Wernsman, & Rose, 2009). School belonging, defined as “a student’s felt experience of acceptance, respect, and inclusion by adults and peers within the school social environment” (McMahon et al., 2009: 269), combines the nature of these relationships with features of the school environment. Phrased differently, school belonging can also be thought of as a student’s sense of community regarding the school context (Osterman, 2000). As with social support and relationships, it is a child’s perception of school belonging that is most strongly associated with outcomes, rather than observed indicators of belonging (Osterman).

By satisfying the fundamental human need for belonging (Baumeister & Leary, 1995), school belonging plays a key role in learning and academic performance by influencing cognitive processes, behaviors, and emotions. Specifically, school belonging has been demonstrated to be an important predictor of academic success and well-being at all ages and levels of education (Osterman, 2000). A strong sense of school belonging may be particularly important in elementary school to establish a positive academic trajectory and prevent students from falling behind (McMahon et al., 2009).

Although school belonging has been shown to directly influence academic achievement (Battistich, Solomon, Watson, & Schaps, 1997; Goodenow, 1993), it also interacts with several related constructs to further influence academic learning and performance. School belonging is associated with good feelings about school and school satisfaction (McMahon et al., 2009), lowered anxiety (Osterman, 2000), higher expectations for success at school (Goodenow & Grady, 1993), reduced personal risk and increased participation in school activities (Osterman), and increased frequency of prosocial and helping behaviors (Osterman). A key pathway through which school belonging may influence academic achievement is through its effects on motivation and self-efficacy. Stronger senses of school belonging are associated with higher levels of intrinsic academic motivation (Solomon, Watson, Battistich, Schaps, & Delucchi, 1996) and greater academic self-efficacy (McMahon, Parnes, Keys, & Viola, 2008). Students who reported higher levels of school belonging also reported stronger senses of personal identity (Osterman).

Social Support and School Belonging as Stereotype Threat Buffers

In addition to the many documented associations of social support with academic achievement, school engagement, and general well-being, social support, and school belonging may play important roles in preventing and buffering stereotype threat among elementary school children. Stereotype threat is a social psychological phenomenon affecting task performance, notably academic achievement, for people belonging to negatively stereotyped social groups (Steele & Aronson, 1995). Although stereotype threat among elementary school students has not been extensively studied, a recent systematic review found strong evidence that stereotype threat does affect the academic performance of elementary school students and that social support and school belonging may serve as important buffers of negative threat effects (Wegmann, 2014). Strambler and Weinstein (2010), for example, found that students who reported higher perceived teacher caring, one form of social support, also reported lower academic devaluing, an established indicator of identity conflict triggered by stereotype threat (Steele, 1997). Similarly, Wasserberg (2009) concluded that supportive teacher–student relationships may buffer against stereotype threat, although this conclusion was not empirically tested in his study. Syed, Azmitia, and Cooper (2011) noted the important but often overlooked role that family support may play in the academic success of ethnic minority and immigrant students, recognizing that families of color and immigrant families may use “invisible” strategies to promote academic achievement that differ from traditional forms of family engagement recognized by schools.

With respect to school belonging, Gillen-O’Neel, Ruble, and Fuligni (2011) found that although Dominican students reported higher stigma awareness (a recognized prerequisite of stereotype threat) compared to European American students, they also reported higher intrinsic motivation for academic tasks. Subsequent analyses revealed that the Dominican students’ levels of intrinsic motivation were positively associated with their reported feelings of school belonging. Gillen-O’Neel and colleagues hypothesized that school belonging might buffer the effect of ethnicity-related stigma on academic anxiety or might enhance the motivation of ethnic minority students by directly reducing academic anxiety and increasing intrinsic motivation.

Purpose

From an exploratory perspective, better understanding of potential differences in the way school belonging and social support are perceived by different social groups of students is necessary to leverage the concepts’ preventive and buffering potentials for design and implementation of psychosocial interventions. In addition, the ability to validly measure and compare levels of social support and school belonging between different social groups of students is important for intervention selection as well as for general assessments of the school and classroom social environments.

While social support in general is well researched, there are currently few comprehensive measurement tools that have been validated for use and comparison of children’s perceptions of school belonging and various forms of social support between racial/ethnic groups. Many measures are limited to assessing just one form or source of social support, providing an incomplete picture of a child’s social context. Even the most widely used measures addressing multiple sources and forms of support, such as the Social Support Scale for Children (Harter, 1985) and the Child and Adolescent Social Support Scale (Malecki, Demaray, & Elliott, 2000), have not had their validity established with diverse populations or undergone measurement invariance testing (Gordon, 2011; Lipski, Sifers, & Jackson, 2014). If social support and school belonging are to be leveraged for intervention in diverse settings, it is necessary to have a measure that has been determined to provide valid assessment and comparison across racial/ethnic groups. The current study seeks to establish the suitability of an existing ecological assessment measure, the Elementary School Success Profile (ESSP) for Children (ESSP-C), to assess and compare perceived levels of social support and school belonging between Black/African American and White elementary school students.

Research questions

The following specific research questions guided the current study:
Are measures of social support and school belonging in the ESSP-C invariant between Black/African American and White students?

Do Black/African American and White students demonstrate statistically significant differences in levels of school belonging and social support?

Methods

Measures

The ESSP is a holistic assessment of an elementary school student’s social environment, measuring home, school, and neighborhood characteristics. Because of children’s developmental limitations in providing valid and reliable self-report data, three components comprise the ESSP: the ESSP-C, a child self-report instrument; the ESSP for families, completed by each child’s primary caregiver; and the ESSP for teachers, completed by each child’s classroom teacher. Together, the three pieces of the ESSP provide a well-rounded picture of a child’s social world by incorporating information from three key sources. The ESSP was based on the School Success Profile (SSP), an analogous instrument for middle and high school students (Bowen, Richman, & Bowen, 2002; Bowen, Rose, & Bowen, 2005).

Because the predictive power of social support and school belonging is tied to a person’s own perceptions of the constructs rather than observed indicators, only child self-report data were included in the proposed analyses, although the ESSP also collects information from parents/caregivers and teachers. As a holistic assessment of a student’s social environment, the ESSP-C includes items that assess key aspects of school belonging and social support from family and friends as well as a general sense of social support. The ESSP-C also measures children’s perceptions of their relationships with teachers and friends at school and positive feelings about school. The version of the ESSP-C available in 2008–2009 consisted of 83 items assessing 12 dimensions along 5 domains spanning a child’s individual and social–environmental circumstances. Items selected for the current analysis belonged to six established scales of the ESSP-C: School Is A Fun Place to Learn, Teachers Who Care, Friends Who Care, Family Who Care, Good Adjustment, and Knows Where to Get Support. The complete text of all included items can be found in Table 1. The 4 items selected from the School Is A Fun Place to Learn scale, hypothesized to reflect school belonging, assessed children’s positive feelings about going to school and their sense of connection to peers at school. Items from the Teachers Who Care scale were selected to represent children’s perceptions of supportive relationships with their teachers, including whether teachers engage in specific behaviors to facilitate learning and validate the importance of children’s contributions to the classroom. Items from the Friends Who Care scale assessed children’s perceptions of peer support (both in and out of school), based on specific supportive behaviors such as, “My friends listen to me when I have something to say,” as well as through more general statements about the nature of the relationship (e.g., My friends and I have fun together). Similar to the teacher support and peer support items, chosen items from the Family Who Care scale reflected children’s perceptions of caregiver support. The items assessed whether caregivers encouraged children to do their best in school, provided emotional support, and fostered a healthy sense of self-esteem in their children. Items from the Good Adjustment and Knows Where to Get Support scales were hypothesized to jointly measure children’s perceptions of general social support. The 2 items from the Good Adjustment scale are reverse-coded items inquiring about a lack of general social support (Do you ever feel nobody cares about you? and Do you ever feel nobody listens to you?). The 2 items from the Knows Where to Get Support scale ask whether children have someone in their lives who listen to them and “is on their side.” These items are similarly worded to items from the caregiver, friend, and teacher scales; however, no specific source of support is mentioned. Therefore, these items are intended to capture other sources of social support that may not have been assessed in other items.

Table 1.
List of ESSP-C Items Used in Current Study.

ESSP-C Scale Item Text Potential Factors

School is a Fun Place to Learn

C8 I think school is fun School belonging

C9 I look forward to going to school School belonging

C12 I have friends to talk to at school School belonging

C14 I have fun with other kids at my school School belonging

C15 I have friends to play with at school School belonging

C16 I have friends to eat lunch with at school School belonging

Teachers Who Care

C17 My teacher and I get along well Teacher relations

C18 My teacher listens to what I have to say Teacher relations

C19 When I try hard or do a good job, my teacher makes me feel good Teacher relations

C20 When I raise my hand, my teacher calls on me Teacher relations

C21 My teacher lets me know that he or she cares about my schoolwork Teacher relations

C22 When I don’t understand something, my teacher helps me Teacher relations

Friends Who Care

C24 When I am upset, my friends help me Friend support

C25 My friends listen to me when I have something to say Friend support

C26 My friends are happy when something good happens to me Friend support

C27 My friends and I have fun together Friend support

C28 My friends are on my side Friend support

Family Who Care

C45 When I am upset, my family helps me Family support

C46 My family listens to me when I have something to say Family support

C49 Adults in my home make me feel special Family support

C50 Adults in my home are nice to me Family support

C51 Adults in my home tell me I did a good job Family support

C52 When I am scared, worried, or upset, I can talk to an adult at home Family support

C53 Adults in my home tell me to try my best in school or to do well in school Family support

Knows Where to Get Support

C71 (reverse coded) Do you ever feel nobody cares about you? General social support

C75 (reverse coded) Do you ever feel no one listens to you? General social support

C80 Someone tells me they are on my side General social support

C82 I can talk to someone who sees things the way I do General social support

Note. ESSP = Elementary School Success Profile.

All items in the ESSP-C are assessed via an ordinal response scale, with most items using a 4-point scale asking children to indicate how often a particular situation occurs: “never,” “sometimes,” “often,” or “always.” The ESSP-C has undergone rigorous cognitive and psychometric testing throughout its development and subsequent use, consistently demonstrating its ability to collect valid and reliable data from a middle childhood population (Bowen, 2008, 2011; Woolley, Bowen, & Bowen, 2006). The ESSP-C has an α coefficient of .77 and a test–retest reliability of .71, both of which compare favorably with similar instruments used with the target population (Bowen, 2008).

In addition to the substantive items chosen for analysis, two important demographic variables were also measured via the ESSP. Children’s race/ethnicity was indicated by a categorical item on the ESSP for Families, completed by children’s parents or caregivers. Because of the nature of the research questions and the fact parents/caregivers had to complete their portion of the ESSP in order for this information to be present, only child cases that had race/ethnicity information were included in the current analysis.

A second categorical variable indicated in which of the sample schools each child was enrolled. This information was provided by the sample schools and stored as a numeric code. The presence of a school enrollment code was important in order to control for potential autocorrelation of cases from children attending the same school, which can bias results if not sufficiently addressed in analysis (Muthén & Muthén, 1998–2012).

Sample

Data were collected from 1,251 third- through fifth-grade students in 13 elementary schools in four school districts in a mid-Atlantic state. Schools were participating in four concurrent ESSP projects. Data were collected during the 2008–2009 school year: during the fall semester from two urban school districts and during the spring semester from two rural school districts.

On the ESSP, key demographic information such as child gender, race/ethnicity, and family economic status is provided by parents/caregiver respondents. Because race/ethnicity is a primary variable of interest to the proposed study, only the 690 child cases that included a completed parent/caregiver component of the ESSP could be included in the analyses. (This group of cases will be referred to as the “study sample” throughout this document.) Mann–Whitney U tests were performed to determine whether statistically significant differences existed between the distributions of the full sample and the study sample on the child report variables used in the current study. In general, students whose parents/caregivers did not complete their portion of the ESSP reported statistically significant lower levels of social support in all evaluated domains (peer, parent/caregiver, teacher, and general social support) and statistically significantly lower levels of school belonging. Given the limited information available on these cases due to the missing parent/caregiver component of the ESSP, it would be premature to draw conclusions about why these differences exist.

The study sample, or cases including race/ethnicity information, remained diverse across several dimensions. Forty-six percent of caregivers in the study sample responded that their children were White, 39% answered that their children were Black/African American, 7% indicated Latino ethnicity, 2% identified as Asian, and 1% reported Native American ethnicity. The remaining 5% of caregivers classified their children’s race/ethnicity as “multiracial” or “other.” Because the available sample sizes for Hispanic/Latino, Asian, Native American, multiracial, and “other” students were well under the minimum of 100 cases recommended for confirmatory factor analysis (CFA; n < 50 for each group, Kline, 2005), Black/African American and White students were the only racial/ethnic groups large enough to be compared in the analysis.

The study sample maintained an approximately even split on participant sex, with 47.9% boys and 52.1% girls. All participants were within the target grade level range of the ESSP (third through fifth grade). Almost half of the study sample was enrolled in fourth grade (48.8%), with 27.9% of the sample enrolled in third grade and 23.3% enrolled in fifth grade. The mean age of the participants was 9.47 years, with 45.3% of the sample at 9 years old, 26.7% at 10 years old, 13% at 11 years old, and 12.3% at 8 years old. Over half (56%) of caregiver respondents indicated that their children received free or reduced price lunch at school. In addition, 41% of caregivers reported some degree of financial instability in their families by agreeing with the statement that it was difficult “to make ends meet” during at least some months of the year.

Data Collection

All data were collected through the online version of the ESSP-C, the child self-report component of the tripartite ESSP. Data collection in the urban schools took place during the fall semester, and data collection in the rural schools occurred in the spring semester of the same academic year. To encourage maximum participation, most schools scheduled times for entire classes of children to come to the school computer lab and complete the survey at once.

Because all of the indicator variables included in the analysis were ordered categorical and not normally distributed, the maximum likelihood estimator typically employed in CFA was inappropriate (Jöreskog, 2005). Ordered categorical variables do not have an established metric, meaning that the means, variances, and covariances of the variables cannot be used to calculate a Pearson moment correlation matrix. Instead, a threshold model is used to link ordinal responses to underlying latent continuous distributions, which are assumed to be normal. Polychoric correlations between all pairs of latent response variables are then calculated (Bollen, 1989; Jöreskog, 2005; Muthén & Asparouhov, 2002). Because of its generation and use of the polychoric correlation matrix, Weighted Least Squares Means and Variances (WLSMV) estimation is the recommended method for analysis of ordered categorical data (Bollen, 1989; Kline, 2005).

Analysis

In order to determine whether the selected items from the ESSP-C perform equivalently for different racial/ethnic groups of students (Research Question 1), invariance testing was performed through multiple group CFA. Multiple group invariance testing allows researchers to determine whether measurement instruments function in the same way for different groups of respondents, enabling valid comparison of scores between the groups. Because factor loadings and intercepts for ordered categorical data both influence the item characteristic curve, the process for invariance testing under WLSMV is slightly different from the process used with the traditional maximum likelihood estimator. The steps for invariance testing under WLSMV estimation are as follows (Byrne, 2012; Muthén & Muthén, 1998–2012):
establishment of group-specific baseline models;

testing of configural invariance: number of latent factors, patterns of factor loadings constrained to be equal in both groups; and

factor loadings and thresholds simultaneously constrained to be equal.

Partial measurement invariance

If full measurement invariance does not hold for a particular measure, the option to identify noninvariant parameters and continue testing for partial measurement invariance (PMI) exists (Byrne, Shavelson, & Muthén, 1989). To test for PMI, potentially noninvariant parameters are identified through modification indices and successively freed. The effect of releasing the constraint on a particular parameter is then evaluated by comparing the fit of the model with the released parameter to the configural invariance model, just as when testing for full measurement invariance. If few parameters are found to be noninvariant while the vast majority of parameters perform equally between groups, the measure can be considered to possess PMI. Sass (2011) notes that the presence of a small percentage of noninvariant items (20% or fewer; Dimitrov, 2010) has little to no effect on a measure’s ability to draw valid comparisons between the groups. If PMI is established for a given measure, comparison of latent means between groups is also possible (Byrne, Shavelson, & Muthén, 1989). Although invariance of latent variances and covariances is required to compare latent means when using other estimation methods, under WLSMV estimation such tests are not required (Bovaird & Koziol, 2012).

Analysis procedure

The multiple group CFA was performed in Mplus version 7.11 (Muthén & Muthén, 2014), using the WLSMV estimator. The Mplus default of delta parameterization was used for all models because the research questions and practical applications of the study did not require the stringency of invariance of residual error terms (Wang & Wang, 2012). Missing data were handled through the pairwise present approach, Mplus’s default missing data procedure associated with WLSMV. Mplus output indicated that a total of 26 cases (12 Black/African American, 14 White) had missing data on all variables included in the analysis. With respect to missing data on individual items, covariance coverage rates ranged from 94% to 100% for Black/African American students and 96% to 100% for White students. Mplus reported 22 missing data patterns for White students and 27 missing data patterns for Black/African American students. Potentially inflated standard errors caused by autocorrelated data, such as the school groups represented in the current study sample, was addressed through use of Mplus’s clustered data feature.

Model quality was evaluated through selected fit statistics, the magnitude and statistical significance of factor loadings and variances, and the substantive rationale for patterns of loadings and correlations between error variances. Commonly recommended fit indices for WLSMV analysis (Bowen & Guo, 2012) were used to evaluate model fit: the robust WLSMV χ², the Tucker–Lewis Index (TLI), the Comparative Fit Index (CFI), and the Root Mean Square Error of Approximation (RMSEA). Hu and Bentler’s (1999) recommendations for acceptable fit index values were followed for the TLI and CFI, with values equal to or greater than 0.90 indicating adequate fit, and values equal to or greater than 0.95 considered indicative of good fit. Kline’s (2005) standard for RMSEA was followed, with values below .05 indicating good fit. Factor loadings were evaluated both on their statistical significance and the strength of the factor loading. Consistent with Comrey and Lee’s (cited in Tabachnick & Fidell, 2013, p. 654) guidelines for exploratory factor analysis, the minimum acceptable value for a standardized factor loading will be 0.32, with greater loading values suggesting stronger relationships between the indicator and latent variable (Bowen & Guo, 2012).

The robust WLSMV χ² is a variation of a traditional χ² test, adjusted for the WLSMV estimation method. Like a traditional χ² test, nonsignificant values indicate good model fit, although they may be difficult to achieve with the sample size required for CFA (Thompson, 2004, pp. 128–129). Furthermore, WLSMV-estimated models should not be evaluated solely by the adjusted χ² score or significance, with Bovaird and Koziol (2012) even suggesting that it should not be interpreted at all. For models of categorical data using the WLSMV estimator, the real value of the adjusted χ² is in difference testing between nested models, such as the successively constrained models used to test for invariance between groups (Muthén & Muthén, 1998-2012; Sass, 2011). In the context of invariance testing, a nonsignificant adjusted χ² difference test indicates that whatever equality constraints were applied in the most recent model did not produce statistically significant worse fit compared to the configural model used as a baseline.

In order to achieve satisfactory model fit, modification indices provided in the Mplus output, the residual correlation matrix, and R ²/squared multiple correlation values were used to identify potential improvements to the model. In accordance with the best practices for improving model fit suggested by Byrne, Shavelson, and Muthén (1989) and Bowen and Guo (2012), modifications were considered only when prespecified fit criteria were not met, the modifications were theoretically supported, and they did not substantially alter other parameter estimates.

If the tests described previously indicated full or PMI, factor means on the constructs of interest could be compared across the two groups. In multiple group invariance testing, invariance of latent means is determined through latent mean difference testing between the analyzed groups (Byrne, 2012). Therefore, one group serves as a reference group with all latent means fixed at zero, and the latent means of the other group describe the difference in latent means of the second group relative to the first group. The p value provided in the Mplus output indicates whether differences between the groups’ means on each latent variable are statistically significant.

Results

Group-Specific Models

In accordance with the recommendations of Byrne (2012) and Sass (2011), analysis began with fitting models to each racial/ethnic group separately. The initially hypothesized five-factor model based on substantive knowledge of school belonging and social support fit the Black/African American group’s data well, requiring no changes to achieve good fit according to the fit criteria (reported in Table 2).

Table 2.
Model Fit Statistics.

Model χ² TLI CFI RMSEA Δχ^2a

Group-specific Black/African American students 396.835* 0.964 0.960 0.026 —

Group-specific White students 496.556 0.941 0.947 0.040 —

Configural invariance 881.090 0.950 0.955 0.033 —

Measurement invariance initial 951.177** 0.955 0.956 0.031 108.374*

Partial measurement invariance Model 1 _{(RELEASE C9$1, C9 LOADING)} 947.603** 0.955 0.956 0.031 102.516*

Partial measurement invariance Model 2 _{(RELEASE C9$3)} 944.937** 0.956 0.956 0.031 95.584*

Partial measurement invariance Model 3 _{(RELEASE C20$3; C20 LOADING)} 942.021** 0.956 0.956 0.031 92.520*

Partial measurement invariance Model 4 _{(RELEASE C14$1, C14 LOADING)} 937.634** 0.956 0.957 0.031 84.174

Note. TLI = Tucker–Lewis Index; CFI = Comparative Fit Index; RMSEA = Root Mean Square Error of Approximation.

^aDifference in χ² values between tested measurement invariance model and configural model.

*p < .05, **p < .001.

The model for White students only needed two modifications, both in the form of an error correlation, to achieve good fit as indicated by the CFI, TLI, and RMSEA reported in Table 2. Error correlations reflect the possibility of an unexplored association between 2 items that results in a common source of measurement error. Any error correlations included in a model should represent possible substantively sound associations between the items involved (Bowen & Guo, 2012; Kline, 2005). An error correlation was added between items C8 (I think school is fun) and C9 (I look forward to going to school), which was substantively justified because both indicators describe general positive feelings about school. A second error correlation was added between items C8 (I think school is fun) and C17 (My teacher and I get along well). Because literature on teacher–student relationships has shown that children have more positive attitudes toward school when they have positive relationships with their teacher, the second error correlation was also considered theoretically justified.

Configural Invariance

When run simultaneously, the two group models demonstrated configural invariance. As with the group-specific models, the adjusted χ² was significant, but all other fit indices showed good fit, as reported in Table 2. The good fit of the simultaneous group models demonstrates that the same latent factors, indicators, and paths are shared between the two groups. The demonstration of configural invariance justified progressing to more stringent tests of measurement invariance.

Measurement Invariance

Following Muthén and Muthén’s (1998-2012) and Sass’s (2011) recommendations, both the factor loadings and the item thresholds were constrained simultaneously in one test of strong measurement invariance. The initial measurement invariance model resulted in acceptable fit statistics but also indicated a statistically significant worsening of fit compared to the configural model.

After concluding that the selected ESSP-C items did not demonstrate full measurement invariance, PMI testing was conducted per the examples of Byrne et al. (1989) and Sass (2011). As suggested by Byrne (2012) and Muthén and Muthén (1998–2012), when a constraint on a potentially invariant item threshold was released, the accompanying item factor loading was also freed. Only the potentially noninvariant threshold was affected by this method; other thresholds for the item remained constrained. A sequence of PMI models was tested, with each iteration freeing a threshold suggested to be potentially noninvariant by the modification indices and its corresponding factor loading. Thresholds 1 and 3 for item C9 (I look forward to going to school), threshold 1 for item 14 (I have fun with other kids at my school), and Threshold 3 for item C20 (When I raise my hand, my teacher calls on me) were successively freed during the PMI testing procedure, along with the factor loadings for items C9, C14, and C20. The model with these freed parameters demonstrated overall good fit (CFI = .957, TLI = .956, and RMSEA = .031) with no statistically significant difference in fit compared to the configural invariance model, Δχ²(65) = 84.174, p > .05. In combination with the previously relaxed constraints on item C20, the third PMI model demonstrated overall good fit (CFI = .957, TLI = .955, and RMSEA = .031) and no significant difference in fit compared to the configural invariance model, Δχ²(60) = 79.075, p > .05.

Comparison of Latent Means

Establishing PMI in the absence of full measurement invariance allows for further testing of parameters associated with the structural model, such as the comparison of latent means (Byrne, 2012; Byrne et al., 1989; Sass, 2011). The latent means of the White group model were considered the reference variables and fixed at zero. The latent means of the Black/African American group then represented the relative difference in mean levels of Black/African American students compared to White students. All latent mean values are listed in Table 3. The only latent mean value found to be significantly different was that of School Belonging, for which the mean for Black/African American students was .207 unit higher than for White students, which was significant at the p < .05 level. The final models for each group are depicted in Figures 1 and 2.

Table 3.
Latent Mean Differences Between White^a and Black/African American Students.

Latent Factor White Mean Black/African American Mean p Value

School belonging .000 .207 .044*

Teacher–student relationships .000 .147 .094

Friend support .000 −.061 .500

Family support .000 .067 .641

General social support .000 .157 .231

^aWhite students were the reference group for the analysis; therefore, latent mean values were fixed at zero. Mean values for Black/African American students represent change relative to White students.

*p < .05.

Figure 1.
Partial measurement invariance model and standardized parameter estimates for Black/African American students. Unstandardized estimates are equal to those for White students.

Figure 2.
Partial measurement invariance model and standardized parameter estimates for White students. Unstandardized estimates are equal to those for Black/African American students.

Discussion and Applications to Practice

The establishment of PMI supports the use of the ESSP-C to make valid comparisons between self-reported levels of school belonging, positive teacher–student relationships, and various forms of social support between Black/African American and White elementary school students. Although full measurement invariance did not hold for the selected ESSP-C items, the presence of a very small percentage of noninvariant parameters has a minimal effect on the validity of a measure for making group comparisons (Sass, 2011). In the current study, only 7, or approximately 4%, of the 179 estimated parameters in the final PMI model were not invariant, well below the 20% noninvariance cutoff suggested by Dimitrov (2010). Future analysis using the ESSP-C social support and school belonging scales can proceed as if the scales were fully invariant or researchers can choose to exclude the 3 noninvariant items if they are not shown to adversely affect the measure’s psychometric properties (Sass, 2011).

The establishment of configural invariance between the group-specific models for White and Black/African American students serves as evidence that the basic structure of the social support and school belonging model is the same for both groups, such that there is the same number of latent factors for each group (Sass, 2011), and the same observed indicators are associated with each latent factor (Cheung & Rensvold, 2002). The establishment of PMI across the two groups signifies that in addition to sharing the same basic model structure, the majority of relationships between the observed indicators and the latent variables are equivalent in magnitude for both White and Black/African American students (Cheung & Rensvold).

As noted by Byrne et al. (1989), the presence of PMI also facilitates valid comparison of the levels of latent variables between groups. Black/African American students reported a statistically significantly higher level of school belonging compared to White students. In light of the evidence associating school belonging with positive academic outcomes, as well as the hypothesis that the relationships and sense of community associated with school belonging may protect students from the effects of stereotype threat (Gillen-O’Neel, Ruble, & Fuligni, 2011), this is an encouraging finding for the academic success of the Black/African American students in the sample. The finding is also in accordance with previous literature that has noted significantly higher levels of self-reported school belonging for African American students compared to their White peers (Booker, 2006; Voelkl, 1997). Although the possibility that Black/African American students in the study sample may be able to benefit from strong feelings of school belonging is promising, previous literature has shown a tendency for the positive association between school belonging and academic achievement to weaken as Black/African American students advance through school (Voelkl, 1997). Ironically, higher self-reported levels of belonging or identification with school may render Black/African American students more vulnerable to stereotype threat, as the identity conflict at the heart of stereotype threat is rooted in perceived dissonance between ascribed identity characteristics and a strong sense of identification with a particular task domain such as school (Steele, 1997). Maintaining Black/African American students’ strong feelings of school belonging and translating them into a buffer against stereotype threat remains a challenge for school social workers, teachers, and other school personnel.

Study Limitations

A limitation related to the study sample is that data from about half of the students in the original data set could not be used for the study because of missing race/ethnicity information. Cases that were excluded due to missing demographic information generally showed lower mean values on the study variables compared to cases with race/ethnicity information. Although the subset of cases used in the current analysis demonstrated some significant differences on the analyzed group of study variables compared to the subset of cases without race/ethnicity information, the subset of cases used in the study maintained several important forms of diversity, such as geographic diversity (urban vs. rural) and diversity of SES, increasing the generalizability of conclusions drawn from the eligible cases.

Implications for Practice

A well-established body of literature emphasizes the importance of strong senses of social support and belonging within a school community, with benefits relating to the emotional well-being, healthy identity development, and academic achievement of students. Because of the socio-emotional nature of these constructs, school social workers are likely to be at the center of efforts to enhance school belonging and feelings of social support, whether directing interventions at the school level, working with students individually or in small groups, or guiding classroom teachers in best practices. A psychometrically sound assessment of school belonging and social support is critical to enable school social workers to identify particular strengths and challenges within a school community or an individual student’s social environment. Social support and school belonging assessment may be useful in selecting interventions to prevent or promote myriad outcomes, including peer victimization and school violence, improved child mental health and well-being, and increased academic achievement.

As noted earlier, many of the widely used measures of school belonging and social support have not been validated for use with students of different races/ethnicities. The current study has established that ESSP-C can be used to accurately assess levels of school belonging and social support for both Black/African American and White students as well as to make valid comparisons of the levels reported by these groups. As such, the ESSP-C may be a valuable tool for school practitioners seeking to enhance social support and school belonging among their students.

Implications for Future Research

Although the sample in the current study was diverse with respect to several characteristics, the total numbers of students from racial/ethnic backgrounds other than White or Black/African American were too small to permit invariance testing for other racial/ethnic groups. Given the increasing Latino population in the study state as well as in the United States overall, it would have been especially useful to test whether the ESSP-C could also be used to validly measure and compare levels of school belonging and various forms of social support for Hispanic/Latino students.

Establishing the suitability of the ESSP-C for valid use with racially/ethnically diverse populations would increase its utility for in guiding socio–environmental interventions to prevent and reduce stereotype threat. Because stereotype threats also exist in relation to other characteristics such as sex and socioeconomic class, establishing invariance for these social identities as well would also allow use of the ESSP-C to assess and guide stereotype threat prevention and reduction interventions with a wider audience.

Although the validation of use of the ESSP-C to guide social interventions related to stereotype threat is a definite step in the right direction, development, and validation of a direct measure of stereotype threat for use with children in the middle childhood stage would be an invaluable tool for addressing stereotype threat in elementary school. Such a measure might be used with the ESSP-C or incorporated into the instrument. In addition to the typical best practices involved in scale development, qualitative research is necessary to better understand how children’s developmental stages may uniquely influence the existence and perception of stereotype threat in this age-group.

Conclusion

Recent research has suggested that one pathway through which school belonging and social support foster academic achievement is by protecting or buffering students from the negative effects of stereotype threat. Selected items on the ESSP-C related to school belonging and social support have demonstrated both configural and PMI for White and Black/African American students, permitting valid comparisons of school belonging and social support to be made. Aligned with the ESSP’s primary purpose of using socio-environmental data to guide intervention, elementary schools can use information on school belonging and social support derived from the ESSP-C to plan targeted interventions utilizing these concepts to prevent and reduce stereotype threat.

ESSP-C Scale	Item Text	Potential Factors
School is a Fun Place to Learn
C8	I think school is fun	School belonging
C9	I look forward to going to school	School belonging
C12	I have friends to talk to at school	School belonging
C14	I have fun with other kids at my school	School belonging
C15	I have friends to play with at school	School belonging
C16	I have friends to eat lunch with at school	School belonging
Teachers Who Care
C17	My teacher and I get along well	Teacher relations
C18	My teacher listens to what I have to say	Teacher relations
C19	When I try hard or do a good job, my teacher makes me feel good	Teacher relations
C20	When I raise my hand, my teacher calls on me	Teacher relations
C21	My teacher lets me know that he or she cares about my schoolwork	Teacher relations
C22	When I don’t understand something, my teacher helps me	Teacher relations
Friends Who Care
C24	When I am upset, my friends help me	Friend support
C25	My friends listen to me when I have something to say	Friend support
C26	My friends are happy when something good happens to me	Friend support
C27	My friends and I have fun together	Friend support
C28	My friends are on my side	Friend support
Family Who Care
C45	When I am upset, my family helps me	Family support
C46	My family listens to me when I have something to say	Family support
C49	Adults in my home make me feel special	Family support
C50	Adults in my home are nice to me	Family support
C51	Adults in my home tell me I did a good job	Family support
C52	When I am scared, worried, or upset, I can talk to an adult at home	Family support
C53	Adults in my home tell me to try my best in school or to do well in school	Family support
Knows Where to Get Support
C71 (reverse coded)	Do you ever feel nobody cares about you?	General social support
C75 (reverse coded)	Do you ever feel no one listens to you?	General social support
C80	Someone tells me they are on my side	General social support
C82	I can talk to someone who sees things the way I do	General social support

Model	χ²	TLI	CFI	RMSEA	Δχ^2a
Group-specific Black/African American students	396.835*	0.964	0.960	0.026	—
Group-specific White students	496.556**	0.941	0.947	0.040	—
Configural invariance	881.090**	0.950	0.955	0.033	—
Measurement invariance initial	951.177**	0.955	0.956	0.031	108.374*
Partial measurement invariance Model 1 _{(RELEASE C9$1, C9 LOADING)}	947.603**	0.955	0.956	0.031	102.516*
Partial measurement invariance Model 2 _{(RELEASE C9$3)}	944.937**	0.956	0.956	0.031	95.584*
Partial measurement invariance Model 3 _{(RELEASE C20$3; C20 LOADING)}	942.021**	0.956	0.956	0.031	92.520*
Partial measurement invariance Model 4 _{(RELEASE C14$1, C14 LOADING)}	937.634**	0.956	0.957	0.031	84.174

Latent Factor	White Mean	Black/African American Mean	p Value
School belonging	.000	.207	.044*
Teacher–student relationships	.000	.147	.094
Friend support	.000	−.061	.500
Family support	.000	.067	.641
General social support	.000	.157	.231

Footnotes

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

References

Battistich

Solomon

Watson

Schaps

(1997). Caring school communities. Educational Psychologist, 32, 137–151.

Baumeister

R. F.

Leary

M. R.

(1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117, 497–529.

Bodovski

Youn

M.-J.

(2010). Love, discipline, and elementary school achievement: The role of family emotional climate. Social Science Research, 39, 585–595. doi:10.1016/j.ssresearch.2010.03.008

Bollen

K. A.

(1989). Structural equations with latent variables. New York, NY: John Wiley & Sons.

Booker

K. C.

(2006). School belonging and the African American adolescent: What do we know and where should we go? The High School Journal, 89, 1–7. doi:10.1353/hsj.2006.0005

Bovaird

J. A.

Koziol

N. A.

(2012). Measurement models for ordered categorical indicators. In Hoyle

R. H.

(Ed.), Handbook of structural equation modeling (pp. 495–511). New York, NY: Guilford Press.

Bowen

G. L.

Richman

J. M.

Bowen

N. K.

(2002). The School Success Profile: A results management approach to assessment and intervention planning. In Roberts

A. R.

Greene

G. J.

(Eds.), Social workers’ desk reference (pp. 787–793). New York, NY: Oxford University Press.

Bowen

G. L.

Rose

R. A.

Bowen

N. K.

(2005). The reliability and validity of the School Success Profile. Philadelphia, PA: Xlibris.

Bowen

N. K.

(2008). Cognitive testing and the validity of child-report data from the Elementary School Success Profile. Social Work Research, 32, 18–28.

10.

Bowen

N. K.

(2011). Child-report data and assessment of the social environment in schools. Research on Social Work Practice, 21, 476–486. doi:10.1177/1049731510391675

11.

Bowen

N. K.

Guo

(2012). Structural equation modeling. New York, NY: Oxford University Press.

12.

Bronfenbrenner

(1979). The ecology of human development: Experiments by nature and design. Cambridge, MA: Harvard University Press.

13.

Byrne

B. M.

(2012). Structural equation modeling with Mplus. New York, NY: Routledge.

14.

Byrne

B. M.

Shavelson

R. J.

Muthén

(1989). Testing for the equivalence of factor covariance and mean structures: The issue of partial measurement invariance. Psychological Bulletin, 105, 456–466. doi:10.1037//0033-2909.105.3.456

15.

Cappella

Kim

H. Y.

Neal

J. W.

Jackson

D. R.

(2013). Classroom peer relationships and behavioral engagement in elementary school: The role of social network equity. American Journal of Community Psychology, 52, 367–379. doi:10.1007/s10464-013-9603-5

16.

Cauce

A. M.

Srebnik

D. S.

(1990). Returning to social support systems: A morphological analysis of social networks. American Journal of Community Psychology, 18, 609–616.

17.

Cheung

G. W.

Rensvold

R. B.

(2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233–255. doi:10.1207/S15328007SEM0902_5

18.

Chu

P. S.

Saucier

D. A.

Hafner

(2010). Meta-analysis of the relationships between social support and well-being in children and adolescents. Journal of Social and Clinical Psychology, 29, 624–645.

19.

Cohen

(2004). Social relationships and health. American Psychologist, 59, 676–684.

20.

Demaray

M. K.

Malecki

C. K.

(2002). The relationship between perceived social support and maladjustment for students at risk. Psychology in the Schools, 39, 305–316. doi:10.002/pits.10018

21.

Demaray

M. K.

Malecki

C. K.

(2006). A review of the use of social support in anti-bullying programs. Journal of School Violence, 5, 51–70. doi:10.1300/J202v05n03_05

22.

Dimitrov

D. M.

(2010). Testing for factorial invariance in the context of construct validation. Measurement and Evaluation in Counseling and Development, 43, 121–149.

23.

Duke

N. N.

Borowsky

I. W.

Pettingell

S. L.

(2011). Adult perceptions of neighborhood: Links to youth engagement. Youth & Society, 44, 408–430. doi:10.1177/0044118X11402852.

24.

Estell

D. B.

Perdue

N. H.

(2013). Social support and behavioral and affective school engagement: The effects of parents, peers, and teachers. Psychology in the Schools, 50, 325–339. doi:10.1002/pits

25.

Furrer

Skinner

(2003). Sense of relatedness as a factor in children’s academic engagement and performance. Journal of Educational Psychology, 95, 148–162. doi:10.1037/0022-0663.95.1.148

26.

Gillen-O’Neel

Ruble

D. N.

Fuligni

A. J.

(2011). Ethnic stigma, academic anxiety, and intrinsic motivation in middle childhood. Child Development, 82, 1470–1485. doi:10.1111/j.1467-8624.2011.01621.x

27.

Goodenow

(1993). The psychological sense of school membership among adolescents: Scale development and educational correlates. Psychology in the Schools, 30, 79–90.

28.

Goodenow

Grady

K. E.

(1993). The relationship of school belonging and friends’ values to academic motivation among urban adolescent students. Journal of Experimental Education, 62, 60–71.

29.

Gordon

A. T.

(2011). Assessing social support in children: Development and initial validation of the Social Support Questionnaire for Children (Unpublished doctoral dissertation). Lousiana State University, Baton Rouge, LA.

30.

Hamre

B. K.

Pianta

R. C.

(2005). Can instructional and emotional support in the first-grade classroom make a difference for children at risk of school failure? Child Development, 76, 949–967.

31.

Harter

(1985). Manual for the social support scale for children. Denver, CO: Author.

32.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. doi:10.1080/10705519909540118

33.

Hughes

J. N.

Luo

Kwok

O. M.

Loyd

L. K.

(2008). Teacher-student support, effortful engagement, and achievement: A 3-year longitudinal study. Journal of Educational Psychology, 100, 1.

34.

Jöreskog

K. G.

(2005). Structural equation modeling with ordinal variables using LISREL. Lincolnwood, IL: Scientific Software International, Inc.

35.

Kline

R. B.

(2005). Principles and practice of structural equation modeling. New York, NY: Guilford Press.

36.

Lipski

D. M.

Sifers

S. K.

Jackson

(2014). A study of the psychometric properties of the Social Support Scale for Children. Journal of Psychopathology and Behavioral Assessment, 36, 255–264.

37.

Malecki

C. K.

Demaray

M. K.

(2006). Social support as a buffer in the relationship between socioeconomic status and academic performance. School Psychology Quarterly, 21, 375–395. doi:10.1037/h0084129

38.

Malecki

C. K.

Demaray

M. K.

Elliott

S. N.

(2000). The child and adolescent social support scale. DeKalb, IL: Northern Illinois University.

39.

McMahon

S. D.

Parnes

A. L.

Keys

C. B.

Viola

J. J.

(2008). School belonging among low-income urban youth with disabilities: Testing a theoretical model. Psychology in the Schools, 45, 387–401.

40.

McMahon

S. D.

Wernsman

Rose

D. S.

(2009). The relation of classroom environment and school belonging to academic self-efficacy among urban fourth- and fifth-grade students. The Elementary School Journal, 109, 267–281.

41.

Muthén

B. O.

Asparouhov

(2002). Latent variable analysis with categorical outcomes: Multiple-group and growth modeling in Mplus. Retrieved from http://www.statmodel.com/download/webnotes/CatMGLong.pdf

42.

Muthén

L. K.

Muthén

B.O.

(1998–2012). Mplus user’s guide (7th ed.). Los Angeles, CA: Muthén & Muthén.

43.

Muthén

L. K.

Muthén

B. O.

(2014). Mplus Version 7.11 [computer software]. Los Angeles, CA: Muthén & Muthén.

44.

National Association of Social Workers. (2012). NASW standards for school social work services. Washington, DC: NASW Press.

45.

Osterman

K. F.

(2000). Students’ need for belonging in the school community. Review of Educational Research, 70, 323–367. doi:10.3102/00346543070003323

46.

Sass

D. A.

(2011). Testing measurement invariance and comparing latent factor means within a confirmatory factor analysis framework. Journal of Psychoeducational Assessment, 29, 347–363. doi:10.1177/0734282911406661

47.

Smith

Osgood

Caldwell

Hynes

Perkins

(2013). Measuring collective efficacy among children in community-based afterschool programs: Exploring pathways toward prevention and positive youth development. American Journal of Community Psychology, 52, 27–40. doi:10.1007/s10464-013-9574-6.

48.

Solomon

Watson

Battistich

Schaps

Delucchi

(1996). Creating classrooms that students experience as communities. American Journal of Community Psychology, 24, 719–748.

49.

Steele

C. M.

(1997). A threat in the air: How stereotypes shape intellectual identity and performance. American Psychologist, 52, 613–629.

50.

Steele

C. M.

Aronson

J. A.

(1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology, 69, 797–811.

51.

Stewart

Suldo

(2011). Relationships between social support and early adolescents’ mental health: The moderating effect of student achievement level. Psychology in the Schools, 48, 1016–1032. doi:10.1002/pits

52.

Strambler

M. J.

Weinstein

R. S.

(2010). Psychological disengagement in elementary school among ethnic minority students. Journal of Applied Developmental Psychology, 31, 155–165. doi:10.1016/j.appdev.2009.11.006

53.

Syed

Azmitia

Cooper

C. R.

(2011). Identity and academic success among underrepresented ethnic minorities: An interdisciplinary review and integration. Journal of Social Issues, 67, 442–468.

54.

Tabachnick

B. G.

Fidell

L. S.

(2013). Using multivariate statistics (6th ed.). Upper Saddle River, NJ: Pearson.

55.

Thompson

(2004). Exploratory and confirmatory factor analysis: Understanding concepts and applications. Washington, DC: American Psychological Association.

56.

Voelkl

K. E.

(1997). Identification with school. American Journal of Education, 105, 294–318. doi:10.1086/444158

57.

Wang

(2012). Structural equation modeling: Applications using Mplus. West Sussex, England: John Wiley & Sons.

58.

Wasserberg

M. J.

(2009). Stereotype threat and standardized testing experiences of African American children at an urban elementary school. (Unpublished PhD). Florida International University, Miami, FL.

59.

Wegmann

K. M.

(2014). An exploration of stereotype threat in middle childhood. (Unpublished doctoral dissertation). University of North Carolina, Chapel Hill, NC.

60.

Wentzel

K. R.

Wigfield

(1998). Academic and social motivational influences on students’ academic performance. Educational Psychology Review, 10, 155–175.

61.

Woolley

M. E.

Bowen

G. L.

Bowen

N. K.

(2006). The development and evaluation of procedures to assess child self-report item validity. Educational and Psychological Measurement, 66, 687–700. doi:10.1177/0013164405282467