Examining academic success among African American high school students

ABSTRACT

We aimed to identify individual (family) and school protective factors that are critical to academic success among African American students, using Bronfenbrenner’s ecological systems theory. A total of 2020 African American students in 463 schools from the Educational Longitudinal Study provided data. We developed a multilevel logistic model (students nested within schools) to predict the probability of academic success based on variables at the student and school levels. The probability was 11% for the nationally typical African American student. At the student level, the effects of socioeconomic status on academic success were stronger for students with lower peer academic commitment, and we identified school involvement, teacher expectation, and time spent on homework as protective factors that increased the probability of academic success. At the school level, we identified one protective factor of academic climate and found that strong school remediation effort tended to signal schools where academic success was lacking.

KEYWORDS:

• Academic success
• African American students
• protective factors
• multilevel modelling
• assessment

Disclosure of potential conflicts of interest

No potential conflict of interest was reported by the author(s).

Notes

1. A similar term to protective factors is present in the literature, referred to as promotive factors. Both come from the success approach or perspective. Promotive factors are defined as environmental, social, and individual factors that interrupt the trajectory from risk to pathology (Fergus and Zimmerman 2005). They are the factors that are associated with positive development and help students overcome adversities and disruptions. As such, the similarity between protective factors and promotive factors is obvious. Even Fergus and Zimmerman did not avoid the term, protective, in their argumentation for the success approach; instead, they identified compensation, protective factor, and challenge as the three cornerstone concepts for the success approach.

2. All estimates of internal consistency (i.e. Cronbach’s alpha) were sample based.

3. At the student level, missing values ranged from zero to 28.7% (with self-efficacy missing at 43.0%). Missing data were therefore mild, and they did not show any serious pattern (i.e. missing by block). Our analytical strategy (see the upcoming Procedures of Analyses) of pairing variables between individual characteristics (e.g. gender) and protective factors (e.g. parent expectation) also helped in the case of self-efficacy. This variable, because its pairing with individual characteristics did not indicate any statistically significant results, was removed very early from data analysis (i.e. not in the final model). Overall, missing data among students did not appear to threaten the ecological validity at the student level.

4. Teacher-student relationship was measured at the student level. Its relation to the outcome within the context of schools can be explored in two ways. One is to aggregate teacher-student relationship from students within each school to be used as a school-level variable. The other is to use teacher-student relationship as a student-level variable allowing its slope (i.e. its relation to the outcome) to vary at the school level. Aggregation is the standard practice when examining school effects (see Ma, Ma, and Bradley 2008). The distinction is mainly conceptual. Teacher-student relationship is considered an aspect of school climate rather than a personal trait of an individual student. Ma (1999) made a similar argument for parental involvement.

5. We also considered the 75th percentile of 57.81 to operationalise academic success. This cut-off point identified only 125 of 2020 African American students as academically successful. The 75th percentile was then deemed too extreme that may speak to academic condition much higher than academic success (e.g. academic excellence).

6. We converted academic achievement from a continuous variable to a dichotomous variable. Although in doing so we lost some information, we improved the conceptual clarify of the outcome. For example, with the dichotomous outcome, increase in teacher expectation was related to how many times more likely for a student to become academically successful; whereas with the continuous outcome, the same increase in teacher expectation was related to certain amount of improvement in academic achievement score. Even though the latter information can be important, it does not directly inform (i.e. relate to) academic success.

7. We made a minor modification within the boundary of Bronfenbrenner’s ecological systems theory. Interaction effects at the student level are not inferred in Bronfenbrenner and Morris (2006). Because these interactions were between individual characteristics (e.g. gender) and protective factors (e.g. parent expectation), they did not change the nature of, say, family protective factors but simply indicated that there might be some individual differences (e.g. gender differences) in the effects of family protective factors.

8. Seeking a parsimonious model within the boundary of Bronfenbrenner’s ecological systems theory was the other minor modification that we made. This practice simply recognised that there might be competitions among protective factors. As such, some protective factors might be more important than others, and they might be emphasised by remaining in the final model from which secondary (statistically insignificant) protective factors were removed.

9. These tables contain the estimates on skewness and kurtosis as part of a preliminary analysis on data distribution. Based on the thresholds of |2.00| and |9.00| respectively (Schmider et al. 2010), neither student-level nor school-level variables indicated any serious abnormalities. In addition, we checked for univariate outliers concerning both student-level and school-level variables (except dummy or count variables). No serious outliers were detected. Finally, because logistic regression is non-parametric, we did not examine normality and homogeneity of variance concerning the dependent variable.

10. In general, effect size is not calculated for statistically significant logistic (regression) coefficients because odds ratio (expressed as EXP in Table 5) can be considered as a measure of effect size with the interpretation of how many times more likely.

11. This estimate also meant that 11% of African American students were academically successful. National (probability) measures concerning African American students are very limited in the literature. Graduation from high school was at 78% for African American students (compared with the national average of 85%) (U.S. Department of Education 2018).

12. This notion of nullification assumed that, although schools low in academic achievement offered more remediation efforts to help students improve, positive results had not yet occurred. A possible alternative to interpret the negative effects was that there were fewer remediation programmes (efforts) in schools where students achieved well because those programmess were not needed. In this sense, the negative effects were not something that needed nullification (i.e. schools operated based on their needs).

Additional information

Notes on contributors

Natasha Murray

Natasha Murray graduated with a doctoral degree of educational psychology from the Department of Educational, School, and Counseling Psychology at the University of Kentucky. Her main research interests include socialization processes and educational outcomes of African American students.

Xin Ma

Xin Ma is Professor of Quantitative and Psychometric Methods in the Department of Educational, School, and Counseling Psychology at the University of Kentucky. His main research interests include advanced statistical methods, large-scale assessments, programme evaluation, policy analysis, school effectiveness and improvement, and psychology of mathematics education.

Kenneth Tyler

Kenneth Tyler is Professor of Educational Psychology in the Department of Educational, School, and Counseling Psychology at the University of Kentucky. His main research interests include culture and cognitive development, race and racism, identity development, school- and community-based learning and socialization processes, motivation, school attachment, and African American student achievement.