UNIVERSITY OF CALIFORNIA
Los Angeles
Tracking Oppositional Identity among United States Youth
A dissertation submitted in partial satisfaction of the
requirements for the degree Doctor of Philosophy
in Sociology
by
David C. Medina, Jr.
2012
© Copyright by
David C. Medina, Jr.
2012
ABSTRACT OF THE DISSERTATION
Tracking Oppositional Identity among United States Youth
by
David C. Medina, Jr.
Doctor of Philosophy in Sociology
University of California, Los Angeles, 2012
Professor Vilma Ortiz, Chair
The source of the racial achievement gap in this country continues to elude researchers.
One popular explanation for divergent educational outcomes between racial/ethnic groups is
John Ogbu’s oppositional culture thesis that differences in academic outcomes can be traced to differences in academic orientations (attitudes and behaviors). Using the National Educational
Longitudinal Survey of 1988 (NELS), this dissertation addresses various substantive and methodological questions within the oppositional culture framework through studying Mexican
American, African American, Asian American, and White youth.
Chapter 2 examines a number of measurement issues regarding oppositional identity.
One of these issues is the consistency of teacher versus student-reported oppositional identity. I find that when examining the same behaviors, there are differences in what teachers and students report. Thus, I argue that teacher and student reports of oppositional identity should not be used interchangeably. A second issue concerns the number of sub-dimensions that make up oppositional identity. I employ factor analysis to validate my oppositional identity construct. I find that oppositional identity can have as many as five sub-dimensions, but they all fall under a
ii global construct of oppositional identity. A third issue is the stability of oppositional identity over time. I conduct longitudinal invariance analyses and find that oppositional identity is not stable, but rather exhibits change over time. Thus, I argue that research must take into account who is doing the reporting, what dimension of oppositional identity is being examined, and when oppositional identity is studied.
Researchers have studied differences in oppositional identity by race/ethnicity. One omission in such analyses is the role of ability grouping (tracking) within the school system. A second omission is the role of family factors in the adoption of oppositional identity. Chapter 3 seeks to fill these gaps by introducing track placement, family background, and parental involvement as predictors of oppositional identity. I find that the effect of these covariates depends on who is reporting, the survey wave being studied, and the dimension of oppositional identity of interest. I also find that racial/ethnic differences are reduced when these covariates are added, while gender differences are consistent and robust. Thus, I argue that oppositional identity is more of a gendered than racial phenomenon.
Chapter 4 examines the effects of oppositional identity on both achievement and dropping out of school. I find that oppositional identity is a significant predictor of both reading and mathematics achievement. I also find that oppositional identity is a significant predictor of dropping out, yet is not able to account for the racial/ethnic gap in these educational outcomes.
I conclude my dissertation with a discussion of my findings and the limitations of my analyses. I also discuss areas for future research on oppositional identity.
iii The dissertation of David C. Medina Jr. is approved.
Min Zhou
Darnell Hunt
José Felipe Martinez
Vilma Ortiz, Committee Chair
University of California, Los Angeles
2012
iv
To my family, who define who I am and make me better.
v
TABLE OF CONTENTS
Acknowledgements: ______vii
Vita: ______xi
Chapter 1: Introduction______1
Chapter 2: The Measurement of Oppositional Identity ______36
Chapter 3: Predictors of Oppositional Identity ______102
Chapter 4: Oppositional Identity and Educational Outcomes______145
Chapter 5: Discussion and Conclusion ______168
Appendix ______176
References ______197
vi ACKNOWLEDGEMENTS
This dissertation is the culmination of many years of being surrounded by supportive individuals in all facets of my life.
I would like to thank all my dissertation committee members. Min Zhou provided me with valuable feedback that ultimately strengthened my dissertation. Her work on segmented assimilation and questioning the status quo ultimately served as motivation and inspiration for my own work. Darnell Hunt’s guidance was also invaluable in strengthening and clarifying my dissertation’s goals. His research and role in the public sphere inspires me to continue actualizing social change through rigorous research. I would like to thank José Felipe Martínez for giving me the opportunity to enroll in my first Social Research Methodology course during my third year of graduate school. I quickly realized during my first day in attendance that it was not the content matter I had envisioned, but I nonetheless remained in the course out of curiosity.
This decision set the foundation for all my subsequent methodological training and newfound passion during a time in graduate school when I really needed it. José has subsequently provided me with guidance and support beyond academics and I am truly grateful for this support.
Vilma Ortiz, my committee chair, has served as my advisor since my first year of graduate school. She has provided me with advice, motivation, and mentorship throughout these six years, and particularly during my dissertation. She has pushed me to reach my full potential during these last few months and there were times when I doubted myself. Nonetheless, she coupled her demand for increased productivity on my end with belief in my ability on hers, thereby offsetting any doubts on my ability to finish. Her office was a safe zone where we could talk about anything that was on my mind. I always came out of her office happy and hopeful
vii regardless of my mood going in. This made graduate school as enjoyable as possible and I would not have reached such an important milestone without her support and dedication. And it is for these reasons that I am truly grateful that I met her so early in my career and was able to interact with her since then.
I am grateful to have been given the opportunity to work at UCLA’s Institute for
Democracy, Education, and Access for the past six years. I would like to thank Veronica
Terriquez for giving me access to join the UCLA IDEA team during my first year. I would also like to thank Siomara Valladares for having the patience as I learned SPSS during my first couple of months on the job. I did not know what I was doing for most of that time. I would have probably fired me if I were her. Nonetheless, she kept pushing and training me, and I could not have reached the level of proficiency I now have today without her help. I would also like to thank my colleagues and friends at IDEA—Rhoda Freelon, Mel Bertrand, Jaime L. Del Razo,
Bryan Ventura, and Jared Planas—all of whom provided me with invaluable support and friendship that made working at IDEA enjoyable.
Other individuals and organizations further supported my tenure at UCLA. I would like to thank the Graduate Division for providing funding for me in the form of the Eugene V. Cota-
Robles fellowship and other summer funding. I would like to thank members of the Department of Sociology for all of their help. I am grateful for Wendy Fujinami’s help with the dissertation process and submitting paperwork for me when I was out of town. I would also like to thank Li
Cai, Michael Seltzer, and Noreen Webb from Social Resarch Methodology for their support and valuable training that I am now in a position to put to good use. Bengt O. Muthen, Linda
Muthen, and the support staff at Mplus also provided me with much statistical guidance
viii throughout my dissertation and I could not have completed my work without their instrumental help.
I have also received a lot of support from graduate students at UCLA. I would like to thank Anthony C. Ocampo for serving as a type of mentor during my first years at UCLA. He provided me with balance in life whether it was heading to the gym, grabbing dinner, or just hanging out. Some of my most memorable experiences at UCLA can be traced to Anthony.
And now as an Assistant Professor, he has provided me with a role model that I look up to and respect. I would also like to thank Dwight Davis, Sarah Schlabach, Hyeyoung Oh, Kjerstin
Elmen-Gruys, Elena Shih, Rennie Lee, Larry Thomas, Forrest Stuart, Gabriel Nelson, Anthony
Alvarez, Yana Kucheva, Erica Morales, Rocio Rosales, and Sylvia Zamora. Your friendships are invaluable and every single interaction was one I will treasure forever.
I would also like to thank my girlfriend and life partner, Queenie X. Zhu. She has provided me with unwavering support during this period that enabled me to retain perspective and ultimately, my sanity. I cannot count the countless nights we stayed up working together to meet external or personal deadlines. While this process has been challenging and has pushed me to my limits, her laughter and smile led me to smile as well even when everything seemed to be falling apart.
I also owe a debt of gratitude to my two sisters, Ana and Linda Jeanette. Ana endured a lot growing up to ensure that Linda and I enjoyed a normal life. She has also paved the way for me and taught me what it means to be successful and is my role model. I am grateful to Linda for her support and for always being a positive force in my life. She reminds me of the meaning of perseverance every day she continues to excel in every facet of her life. I am also grateful for
ix her cookie-making ability and for allowing me to eat more than the two-cookie limit she set for others.
My mom and dad provided me with the solid foundation that enabled me to reach this milestone. My father, David Medina Sr., always emphasized that getting an education was our only “job,” so we had to make sure we put in as much effort as possible. My mother, Maria C
Medina, inculcated in me a sense of humility that I have tried to maintain throughout my career and interactions with others. I am also grateful for her great cooking, particularly on those long days or nights where I otherwise would be eating some pre-packaged snack. Both my mom and dad also provided me with a mold for being a good man and I strive to fit that mold everyday of my life.
Ultimately, it is all those listed above who shaped me into the individual I am today. The saying that we “stand on the shoulders of giants” truly applies to me. I have been blessed with great people that have provided me with a solid foundation for all facets of my life. I am truly grateful and will always remember each contribution as I move forward to the next phase of my career.
x VITA
EDUCATION
2006 B.A., Sociology and Criminology The Ohio State University Columbus, Ohio
2008 M.A., Sociology University of California, Los Angeles Los Angeles, CA
2010 C.Phil., Sociology University of California, Los Angeles Los Angeles, CA
RESEARCH EXPERIENCE
2008 Graduate Student Researcher Institute for Research on Labor and Employment University of California, Los Angeles
2011 Research Consultant Boyle Heights Promised Neighborhood Initiative Los Angeles, CA
2006-2012 Graduate Student Researcher Institute for Democracy, Education, & Access University of California, Los Angeles
2012 Director of Resarch and Analysis Alliance College-Ready Public Schools Los Angeles, CA
FELLOWSHIPS
2005 CIC Summer Research Opportunity Fellow The Ohio State University Columbus, OH
2007 NSF UC DIGSSS Fellow University of California, Los Angeles Los Angeles, CA
xi
2006-2011 Eugene V. Cota-Robles Fellow Department of Sociology University of California, Los Angeles
PUBLICATIONS
Rogers, J., Fanelli, S., Freelon, R., Medina, D., Bertrand, M., Del Razo, M. (2010). Educational Opportunities in Hard Times: The Impact of the Economic Crisis on Public Schools and Working Families. Los Angeles: UCLA IDEA, UC/ACCORD.
Rogers, J., Fanelli, S., Medina, D., Zhu, Q., Freelon, R., Bertrand, M., Del Razo, J. (2009). California Educational Opportunity Report: Listening to Public School Parents. Los Angeles: UCLA/IDEA, UC/ACCORD.
Rogers, J., Oakes, J., Fanelli, S., Medina, D., Valladares, S., Terriquez, V. (2007). California Educational Opportunity Report: The Racial Opportunity Gap. Los Angeles: UCLA/IDEA, UC/ACCORD.
xii CHAPTER 1
INTRODUCTION
U.S.-born minority youth constitute a growing proportion of students in our nation’s schools. This demographic reality is accompanied by concern over the eventual academic success of U.S.-born minority youth because they are at a high risk of dropping out (Rumberger
1993). High school dropouts have limited employment opportunities in an increasingly service- oriented economy (Portes and Rumbaut 2001). The high dropout rate exhibited by this population poses serious problems for the future of these children since education is the most common means of achieving intergenerational mobility. In response, scholars seek to unravel the determinants of academic success for this population.
One common argument traces the failure of U.S.-born minority students to the formation of an oppositional identity (Ogbu and Fordham 1986). This identity formation is a response to
“the way white Americans treat them [minorities] in economic, political, social, and psychological domains, including white exclusion of these groups from true assimilation” (Ogbu and Fordham 1986:181). The creation of an oppositional reference point serves to reinforce the boundary between the dominant group and subordinate minorities. The end result is that
“subordinate minorities regard certain forms of behavior and certain activities or events, symbols, and meanings as not appropriate for them because those behaviors, events, symbols, and meanings are characteristic of white Americans” (Ogbu and Fordham 1986:181).
This study focuses primarily on Mexican American youth, but includes African
Americans, Asian Americans, and Whites as comparisons. Both African American and White youth are included because much of the literature on oppositional identity involves comparisons between these two groups (e.g. Ainsworth-Darnell and Downey 1998). This comparison is
1 common because Black students are generally considered to exhibit oppositional identity, but
White students are not (Ogbu and Fordham 1986). The segmented assimilation literature also highlights the existence of an oppositional identity among U.S.-born students, which includes not only African American students, but third generation Mexican American youth as well. Asian
Americans are included because the segmented assimilation literature suggests they are likely to exhibit lower rates of oppositional identity. Thus, both the oppositional identity and segmented assimilation literatures highlight the existence of an oppositional identity.
The existence of oppositional identity among minority youth is a subject of debate among scholars. Some argue that this orientation exists (e.g. Ogbu and Fordham 1986; Farkas 2008), while others argue that it does not (Ainsworth-Darnell and Downey 1998). Part of the divergence may be due in part to different studies employing different measures in their operationalization of oppositional identity. The assumption in such analyses is that those measures tap into some latent variable that truly captures oppositional identity. Furthermore, studies employ measures that vary in who is reporting—some rely on student self-reports, others rely on teacher self-reports, while another relies on reports by parents (Farkas 2008). The assumption is that each of these sources on their own is sufficiently reliable for reporting the true phenomena at work. To date, no study has developed a comprehensive measure of oppositional identity, and then used this measure to answer various questions relevant to the oppositional identity literature and other relevant literatures.
I begin this chapter by listing the specific research questions I have addressed in my dissertation. I then discuss the relevant literature as it relates to race/ethnicity, assimilation, and oppositional identity. I conclude this chapter by providing an overview of the organization of my dissertation and a short discussion of my findings for each chapter.
2
RESEARCH QUESTIONS
Chapter 2: The Measurement of Oppositional Identity
i. How consistent are students and teachers in reporting oppositional identity behaviors?
ii. Is oppositional identity best measured by individual factors or in combined scales?
iii. Is oppositional identity stable over time?
Chapter 3: Predictors of Oppositional Identity
i. Are there differences in oppositional identity among Mexican American, African
American, Asian American, and White youth?
ii. Are there differences in oppositional identity by gender?
iii. Are prior achievement, parental socioeconomic status, and parental involvement
significant predictors of oppositional identity?
iv. Are track placement, attending a poorer school (with a majority receiving free-reduced
lunch), attending a homogenous White school, and attending public school significant
predictors of oppositional identity?
v. How do the initial racial/ethnic and gender differences in oppositional identity change
with the inclusion of other covariates?
Chapter 4: Oppositional Identity and Educational Outcomes
i. Does oppositional identity affect educational achievement?
ii. Does oppositional identity affect dropping out of school?
3 OPPOSITIONAL IDENTITY, ASSIMILATION, AND RACE/ETHNICITY
Oppositional Identity and Race/Ethnicity
There are fewer studies on oppositional identity among Mexican origin youth than their
African American counterparts. Taylor (2008) studies both Latinos and African Americans to determine if group differences in oppositional identity exist. She measures the degree to which these two groups exhibit differences in: cultural inversion, cultural mistrust, perceptions of mobility/barriers, experiences with discrimination and ethnic identity, ethnic identity, and attitudes towards schooling (Taylor 2008). She finds that African Americans report significantly higher perceptions of victimization than Latinos (Taylor 2008). On the other hand, “despite
Latinos’ perception of personal obstacles to success, they do not seem to develop a generally negative worldview as do African Americans” (Taylor 2008:491). Taylor also suggests that future research should look into the possibility that “oppositional identity might be composed of different dimensions across ethnic groups and possibly across generations” (Taylor 2008:491).
As Waldinger and Feliciano (2004) point out, “the mix of working/lower class boys taught by female, middle-class teachers proves particularly combustible” (381). Working class boys are taught to express a masculine identity in all aspects of daily life. This creates conflict within the classroom given that teachers are more likely to be female and/or middle-class, thereby expressing more femininity (Waldinger and Feliciano 2004). The end result is that boys may exhibit more adversarial behaviors, thereby leading to even more confrontation. This would also be consistent with boys being more likely to be marginalized than girls because school officials see them as threatening, particularly those boys that wear attire attributed to gang affiliation such as baggy pants (Valenzuela 1999).
4 Valenzuela (1999) argues that a process of subtractive schooling takes place within the
U.S. educational system. Mexican American youth are stripped of valuable resources as they progress through school. The institutional policies and practices within schools coupled with a lack of resources serve to alienate Mexican American youth (Valenzuela 1999). Thus, the poor environments minority youth are exposed to may be the source of an oppositional identity and/or poor academic achievement. Ogbu (2003) makes a similar argument when he traces the source of oppositional identity (in part) to the perception among African American students that their teachers and/or counselors simply do not care about them. He argues that White students do not have a similar expectation; the only expectation White students have of the teacher/student relationship is that the teacher be qualified/knowledgeable.
Classical/Neoclassical Assimilation
Park and Burgess (1921) define assimilation as the process by which outsiders take on the characteristics of the host society. This early conception of assimilation was based primarily on cultural attributes, whereby outsiders adopt the cultural orientation (e.g. values) of the host society. Park (1950) argues that assimilation is the last stage of a “race relations cycle”, which begins when newcomers come in contact with members of the host society. Park’s framework acknowledges that assimilation may be a gradual process, but asserts that it is irreversible once it takes place (Park 1950). Gordon builds on Park’s work by identifying seven dimensions of assimilation: cultural, structural, marital, identity, prejudice, discrimination, and civic (Gordon
1964:71). Acculturation, or cultural assimilation, is the first (and inevitable) stage that takes place once two groups come into contact.
Alba and Nee (2003) modify classical assimilation theory to account for factors that are unique to post-1965 immigrants. One such factor is that today’s immigrants “are much more
5 visible by their skin color and other phenotypical differences” (Lucassen 2005:5). These phenotypical differences are likely to persist across generations. The disappearance of mid-level jobs (hourglass economy) makes it difficult for immigrants with low levels of human capital to achieve intragenerational mobility; the intergenerational mobility of their children is jeopardized as well (Portes and Zhou 1993). Also, post-1965 immigrants are also more economically heterogeneous than their European counterparts. Thus, all groups may not share a similar process and/or trajectory of assimilation.
Alba and Nee (2003) also steer towards a more dynamic view of the “mainstream.”
Classical assimilation theorists (Park 1950 and Gordon 1964) assumed that immigrants ultimately take on the cultural orientation of the host society. Such a model views the
“mainstream” as fixed and unable to be redefined by immigrants as they enter society. Alba and
Nee (2003) acknowledge that the newly incorporated members can in fact play a role. They define the mainstream as a place where “ethnic and racial origins have at most minor impacts on life chances or opportunities” (Alba and Nee 2003:12).
Alba and Nee (2003) describe assimilation as an involuntary process as well. This process is characterized as a rational choice that leads individuals/groups to perform certain activities in an attempt to succeed in the host society. Thus, unlike scholars of classical assimilation, Alba and Nee steer away from making the assumption that there exists a unitary
WASP culture to which immigrant groups should assimilate. The assumption that a unitary culture does remain though, but it shifts from WASP to what the authors define as the mainstream (Alba and Nee 2003).
Segmented Assimilation
6 Segmented assimilation theorists argue that modifications to classical theory are not enough to account for the contemporary trajectory for children of immigrants. Segmented assimilation theory stresses the role of both family background and context of reception in determining the trajectory of immigrant groups (Portes and Zhou 1993; Portes and Rumbaut
2001; Zhou and Kim 2006). For example, Zhou and Kim (2006) argue that the intersection between culture and structure facilitates the educational attainment of Chinese and Korean children: “the cultural attributes of a group feed on the structural factors” (24). In other words, ethnic institutions (of supplementary education) within Chinese and Korean communities reinforce their respective pro-schooling cultural orientations. Thus, segmented assimilation stresses that both what immigrants bring with them and how they are received impacts observed trajectories.
Portes and Zhou (1993) argue that there are different segments of society in which groups incorporate. This departs from an assumption of a single culture in the classical and neoclassical assimilation frameworks. Some groups may travel the conventional path of advancement across generations, while other groups may experience a decline in status through downward assimilation. Downward assimilation, as indicated by not completing high school, is problematic given that today’s hourglass-shaped economy places a premium on higher education (Portes and
Rumbaut 2001). Zhou and Bankston point out that “economic restructuring has undone the traditional ladder of social mobility, creating new obstacles for those newcomers who are poorly educated and lack marketable skills” (Zhou and Bankston 1998:3). One consequence of downward assimilation is the creation of a rainbow underclass at the bottom of society (Portes and Rumbaut 2001).
7 Waldinger and Feliciano (2004) challenge this notion by pointing out that the term downward assimilation distorts reality. They look to employment rates, average weeks worked, and long-term joblessness between first and second-generation Mexican origin individuals and whites. The focus on those of Mexican origin stems from low levels of human capital and negative context of reception experienced by this group, making it a perfect example with which to examine if downward assimilation is taking place. The patterns across all three indicators of employment suggest parity of the second generation with whites, particularly for women. They therefore conclude that “the experience of today’s second generation is consistent with the earlier pattern, in which children of immigrants progressed by moving ahead within the working class”
(Waldinger and Feliciano 2004:395).
Waldinger and Feliciano (2004) also point out that second generation Mexicans demonstrate higher educational attainment relative to their parents. While their educational attainment may not be as high as that of other groups, it is still an improvement compared to the immigrant generation. Telles and Ortiz (2008) show a similar trend, whereby the second generation improves relative to their parents. In contrast, the third and subsequent generations appear to have stalled and are doing no better than their parents (Telles and Ortiz 2008).
However, no generation fares well in educational attainment when compared Whites. Thus, while there is no evidence of downward assimilation, there is also little indication of straight-line assimilation.
Portes and Zhou (1993) identify various mechanisms within schools that may lead to downward assimilation: the exposure of children of immigrants to U.S.-born minorities (Blacks and Chicanos), discrimination, and inadequate resources. The exposure of immigrant children to oppositional minorities puts them at-risk of adopting the adversarial stance towards education.
8 Also, immigrant youths may also be discriminated against by virtue of racial group membership or lack of English proficiency. Finally, academic success is hindered by the poor conditions (e.g. lack of credentialed teachers, poor facilities) of inner-city schools to which immigrant youth are exposed (Portes and Zhou 1993).
The retention of ethnicity through selective acculturation insulates immigrant children from this negative context of reception (Portes and Zhou 1993). Zhou and Bankston (1998) argue that the elaborate social networks of Vietnamese immigrants promote academic achievement. Educational attainment is facilitated through the resources generated from the ongoing interaction between children, parents, and the community (Zhou and Bankston 1998).
These resources, commonly referred to as social capital, serve to reinforce social relations and norms that effectively insulates Vietnamese children from negative aspects of their environment, like poor schools and/or adversarial U.S.-born students.
Segmented assimilation makes reference to an oppositional culture among U.S.-born students (Zhou and Bankston 1998). “They (the Cholos) were also native-born Mexican
Americans, easily identifiable by their deliberate manner of dress, walk, speech, and other cultural symbols” (Portes and Zhou 1993:98). This oppositional schooling orientation is what immigrant children are supposed to be insulated from by their elaborate social networks (Zhou and Bankston 1998). While many U.S.-born groups are assumed to possess an oppositional culture, African Americans and Chicanos are usually the reference group within the segmented assimilation framework (Portes and Zhou 1993).
Segmented assimilation theory therefore highlights that group differences in oppositional culture exist; both Mexican origin and African American youth are expected to exhibit higher
9 rates of oppositional attitudes and behaviors than Whites. On the other hand, Asian Americans are expected to exhibit lower rates of oppositional identity than Whites.
Oppositional Identity and Segmented Assimilation
The segmented assimilation literature assumes that an oppositional identity exists, even if oppositional identity has not been directly examined. An exception is Kroneberg (2008), who questions whether “school performance among the new second generation are indeed due to differential exposure to the oppositional cultures of domestic minorities, the ways immigrant groups have been received by the U.S. government and society, and/or the capability of ethnic communities to act as social capital” (140). He models standardized achievement as a function of community-based social capital (e.g. community educational expectations), family background, and school characteristics. Kroneberg finds no support for the argument that immigrant children are at risk of developing an oppositional orientation when in contact with
U.S.-born minorities. As he points out, “this mechanism is hardly responsible for the interethnic differences in school performance…the corresponding variables have no net association with schooling performance” (Kroneberg 2008:149).
Zhou and Bankston (1998) are another exception in that they offer insights into oppositional identity by examining delinquency among Vietnamese youth. They utilize cluster analysis to classify Vietnamese youth into two distinct groups: delinquents and non-delinquents.
They then compare rates of various behaviors between delinquent and non-delinquent
Vietnamese youth. These behaviors include general tastes and preferences, activities carried out at home, and activities carried out in school. They find that “the delinquents tend to dislike listening to traditional Vietnamese music, helping around the house, reading, participating in
10 school clubs…[and] favor gold necklaces, hanging out, jewelry in pierced noses, and rap music
(Zhou and Bankston 1998:198).
This account highlights that there are striking similarities between delinquency and oppositional identity. While listening to traditional music and helping around the house are not common measures of school engagement, other measures used by Zhou and Bankston are similar measures to those used within the oppositional identity framework. For example, excessively watching television and/or hanging out can negatively impact the amount of homework completed by students, and homework completion is a common measure within the oppositional identity framework. Similarly, delinquency itself can be seen as an extreme form of oppositional identity.
OPPOSITIONAL IDENTITY- METHODOLOGICAL DIFFICULTIES
Measures of Oppositional Identity
The ideal way of measuring oppositional identity would be for independent observations of behaviors by researchers and observations in a range of settings and across time. In such an ideal world, there would be minimal measurement error as compared to when students report on their own behavior and/or other authority figures report on students’ behavior. Unfortunately, it is not possible to achieve this gold standard of validity given the limitations of existing survey data and the complexity/expense of collecting original data. The reality is that studies on oppositional identity use many different measures to capture the oppositional identity construct.
These measures vary in many respects and this variation could explain differences in findings and conclusions in prior research.
11 Reports about both abstract ideals and behaviors have been used in studies of oppositional culture. Abstract ideals involve the orientation(s) of students or what they believe.
For example, students may say that they believe education is important for their future, which may or may not correspond with what they actually do. Thus, it is imperative to also look at actual behaviors by students. In other words, do students who report that education is important engage in behaviors that foster academic achievement (such as turning in homework)?
Similarly, there is wide variation in the specific attitudes and/or behaviors that are used in studies. For example, Ainsworth-Darnell and Downey (1998) gauge students’ agreement with the following statement: I get a feeling of satisfaction from doing what I’m supposed to do in class. This statement provides a glimpse into the orientation of respondents towards schooling.
Farkas (2008) asks students on their level of agreement to the following statement: My friends make fun of people who try to do well. The preceding statement assumes that the students share the same ideology as their peers. In other words, the statement my friends make fun people who try to do well potentially says more about students’ peers than it does about the students’ orientations.
Measures also vary in how directly they are asking about an oppositional orientation. For instance, Ainsworth-Darnell and Downey (1998) also include the following two items:
Discipline is fair and I think it is OK to disobey school rules. If students disagree with the first statement, students may simply be reporting that there are unfair rules at school. Thus, this statement would have to assume that schools have fair discipline in order to potentially indicate oppositional identity. The second statement gives more direct insight into oppositional identity.
Thus, whether these two examples capture the same construct (or oppositional identity in general) is debatable.
12 Another example of an indirect measure of oppositional identity is the amount of time spent on homework (e.g. Harris and Robinson 2007). The problem with this item is that the amount of time spent on homework is partly dependent on the amount of homework that is assigned. A more accurate measure of oppositional behaviors would be to ask directly about completing homework or to combine measures of time spent on homework and the amount of homework given by the teacher. Similarly, the focus on “attentiveness” in class disregards the possibility that the material presented by teachers does not engage the student. Thus, it is imperative to examine items closely to remove any factors that may bias the item’s inclusion as a measure of oppositional identity.
In sum, the assumption in these various measures is that they exhibit construct validity in measuring oppositional identity. If this assumption does not hold, the different conclusions reached by researchers may be an artifact of methodology. In other words, inconsistent findings in prior research may be due to variation in the operationalization of oppositional identity utilized. Even if we were to assume these items measure the same underlying construct, an additional issue is whether the different measures equally capture oppositional identity. In other words, does agreement with the statement my friends make fun people who try to do well capture oppositional identity to the same degree as agreement with the statement it is ok to cheat on tests? Furthermore, using too few items may not fully capture oppositional identity (for instance,
Kroneberg 2008 only used two items). The multidimensionality and/or complexity of oppositional identity may require multiple measures to fully capture the phenomena at work.
Similarly, the introduction of multiple measures is key to determining whether oppositional identity is a matter of degree. Oppositional identity is likely made up of different components; some students may possess more of these “components” than other students (Taylor
13 2008). Thus, oppositional identity is potentially best thought of as a continuum where some students exhibit some oppositional behaviors/attitudes, while others potentially exhibit many more. It is therefore imperative to develop comprehensive measures of oppositional identity to tap into the potential variation in oppositional culture across students. These measures can then be used to answer other questions relevant to both the oppositional identity and segmented assimilation literature as well.
Construct Validity of Oppositional Identity
Using one or just a few indicators may be inadequate in fully capturing oppositional identity. Lundy and Firebaugh (2005) point out that previous studies (like Kroneberg’s) used a limited number and not the best measures to capture oppositional identity. They introduce multiple items and create factor-analyzed scales to capture oppositional identity. One scale involves peer relations in which students are asked about relationships with peers both within and outside of school. The second scale created seeks to capture “academic orientation” based on behaviors and attitudes towards education. Lundy and Firebaugh (2005) therefore hypothesize that oppositional identity is made up of both peer relation and academic orientation dimensions.
Ainsworth-Darnell and Downey (1998) introduce many more items of oppositional identity in their attempt to debunk oppositional identity theory. They create multiple scales capturing perceptions of future opportunity, skills (habits), concrete attitudes, and popularity among peers. These scales consist of the summed responses to the items for each construct. For example, the effort scale (a subset of the skills construct) was constructed by adding the responses to four questions that asked: how often the student worked hard for good grades, how often the student completed homework assignments, and how often the student is attentive in
14 class. No factor analysis is conducted, but alpha coefficients are reported; the alpha coefficients are uniformly high across all scales.
Farkas (2008) also creates a scale using multiple items of what he calls “child’s school engagement.” The specific questions used to capture this construct are how often the child: cares about doing well in school, only works on schoolwork when forced to, does just enough school-work to get by, and always does homework. He sums the responses to these four questions and creates a 16-point scale, “in which higher scores indicate greater school engagement” (Farkas 2008:325). Unlike other researchers (Lundy and Firebaugh 2005;
Ainsworth-Darnell and Downey 1998), there is no reference to an alpha coefficient.
Both Lundy and Firebaugh (2005) and Ainswell-Darnell and Downey (1998) employ a commonly used metric in to determine the unidimensionality of scales: the alpha coefficient (a measure of reliability). Gardner (1995) points out that Cronbach’s alpha coefficient is often misused in applied research, in part because internal consistency is confused with unidimensionality. He points out that Cronbach’s alpha only captures internal consistency: the degree to which “every item in a scale shares common variance with at least some other items in the scale” (Gardner 1995:286). On the other hand, unidimensionality means that items in scale all share common variance. As is evident, “a scale can be internally consistent and display a high alpha value, but this does not necessarily mean that it is unidimensional” (Gardner
1995:286).
Thus, analyses that rely on Cronbach’s alpha to justify unidimensionality are inadequate.
While it is true that unidimensional scales can exhibit high internal consistency, the reverse is not true. Thus, the only way to establish unidimensionality is to employ factor analysis on the items believed to measure an underlying latent variable. Factor analysis is used to establish factorial
15 validity: a form of construct validity. Factorial validity is achieved by examining the interrelationships among a set of observed variables; the observed variation in the observed variables can potentially be explained by unobserved factor(s). There are two types of factor analysis: confirmatory and exploratory. Exploratory factor analysis is conducted when there is little theory governing the number of factors to be extracted; it is therefore largely data-driven.
Confirmatory factor analysis employs theoretical justification for the number of factors to be identified; the researcher must therefore specify the model prior to conducting the analysis.
Lundy and Firebaugh (2006) and Taylor (2008) both employ factor analysis before creating the scales used in their respective analyses. The main shortcoming is that too little information is provided to determine the procedure that was actually conducted. While Lundy and Firebaugh (2005) point out that they employ “factor analysis,” they do not provide enough information on their procedure. For example, a common mistake among researchers is to take
Pearson correlations of categorical data, then using the correlation matrix in the factor analysis procedure. This is problematic given that the factor analysis model assumes continuous random variables. The correct procedure is to calculate a tetrachoric/polychoric correlation matrix derived from the underlying continuous response variables given the observed categorical responses (Wirth and Edwards 2007). They also do not highlight whether they conducted exploratory or confirmatory factor analysis.
Both studies also employ varimax rotation after having obtained the factor loading estimates. Varimax rotation is a type of orthogonal rotation: the underlying assumption is that the common factors are uncorrelated in the population. The problem with employing orthogonal rotation is that the researcher does not know whether the factors are uncorrelated prior to fitting the model. Also, given the nature of oppositional identity, it stands to reason that the different
16 factors are in fact correlated. An alternative is to employ oblique rotation: this technique allows the common factors to be correlated. In carrying out factor analysis with oblique rotation, one would get the same result as when employing orthogonal rotation if the common factors are uncorrelated. Thus, it makes sense to employ oblique rotation over orthogonal rotation.
Irving and Hudley (2008) hypothesize that oppositional identity is composed of four dimensions: cultural mistrust, outcome expectations and outcome value, resistant cultural attitudes, and ethnic identity affirmation (Irving and Hudley 2008). They conduct factor analysis and create four one-factor scales to capture each dimension. The cultural mistrust scale is designed to gauge the degree of mistrust in the dominant culture. The outcome expectations and outcome value scale “investigates the benefits that respondents expect to gain from academic achievement, as well as how much they value these outcomes” (Irving and Hudley 380:2008).
The resistant cultural attitudes and ethnic identity affirmation scales are designed to capture negative beliefs about schools and ethnic identity status, respectively.
Student and Teacher Reports of Oppositional Identity
It is also important to examine the inconsistencies that are introduced due to employing different sources of information in reporting on behaviors. Students are commonly used as sources of information. Lundy and Firebaugh (2005) use student self-reports of behavior and attitudes to examine differences among African Americans, Whites, Asians, and Hispanics.
Lundy and Firebaugh do not discuss the extent to which student self-reports may be unreliable in reporting the extent of their oppositional behavior. Taylor (2008) also uses student self-reports to identify the degree of cultural inversion, cultural mistrust, perceptions of barriers, and experiences with discrimination among respondents. Given that she deals exclusively with attitudes, the focus on student self-report makes sense.
17 In addition to student reports, teacher reports on students’ behaviors are commonly used.
Ainsworth-Darnell and Downey (1998) use both student and teacher self-reports in their analysis. They examine perceptions of future opportunity, concrete attitudes, and popularity among peers. The use of student self-reports when gauging attitudes makes sense given that only students can report this type of information. They then use both student self-reports and teacher- reports about students’ behavior. The degree of effort a student engages in is determined from teacher report. Homework patterns and other behavioral attributes (e.g. number of times suspended) rely on student self-reports. The underlying assumption when using teacher and student self-reports interchangeably is that teachers and/or students are capturing the same phenomena at work.
Teacher reports may be viewed as more objective since they are outsiders, yet they may be biased in other ways. For instance, teacher reports may vary by teachers’ characteristics (e.g., race). Farkas et al. (1997) point out that white teachers are more likely than black teachers to give black students negative evaluations. Downey and Pribesh (2004) seek to determine why black students receive harsher evaluations from white teachers than they do from black teachers.
They point to two competing hypotheses that explain this observed pattern. The first is that black students misbehave more when placed with white teachers versus black teachers because of the strained schooling relationship, which would be consistent with oppositional culture framework. The second hypothesis is that white teachers simply discriminate against black youth by evaluating their behavior more harshly than they would a white student. They find limited support for the second hypothesis: “negative evaluations of black students are restricted to those matched with white teachers” (Downey and Pribesh 2004:275).
18 Farkas (2008) attempts to overcome the potential bias of both teachers and students by using parental reports. As he points out, “presumably, parents are not biased against their own children…they also should be better able than their children to accurately report on their children’s behavior, avoiding the wishful thinking that children may indulge in when reporting their own behavior” (Farkas 2008:321). While this explanation may be true, it is also possible that parental reports are unreliable as well. For example, it assumes that parents are knowledgeable of their children’s behaviors in school. It also assumes that parents are not overly favorable or engage in wishful thinking when describing their own child’s behavior.
Furthermore, shifting to parental reports says nothing about the reliability of student and/or teacher self-reports. Thus, it is imperative to delve into this issue further to come up with a better understanding of reporting practices and their substantive implications.
The source of information in seminal works of oppositional identity does not come from students, teachers, and/or parents, but rather is based on the direct observation by anthropologists
(Ogbu and Fordham 1986; Ogbu 2003). Objective observers may be the best source of information for this type of phenomena since it is not based on the interests/bias from teachers, parents, and/or students. Nonetheless, observations are still subject to bias in that the students may engage in socially desirable behaviors during observation; this bias is minimized as the researcher observe behavior for longer periods. Furthermore, there is the issue of whether the experiences in a specific setting generalize to other settings. Even within the same setting, it is possible that the observations may not generalize over time for behaviors and attitudes that are inconsistent.
The use of students and/or teachers as sources of information on oppositional identity across studies may contribute to the divergent findings. Thus, it is important to account for the
19 potential sources of error in the reporting of “true” behaviors. At the most basic level, it is important to determine how consistent students and teachers are in reporting behavior during a given survey wave. The next step is to identify the key sources of error variance and attempt to reduce this variance to better capture true variation from person to person (universe score variance). In the situation where this error variance is unable to be reduced, it is important to acknowledge that the use of different sources of information may lead to divergent outcomes. In this regard, presenting the divergent results is important in order to gauge how different the conclusions are when student self-reports are used instead of teacher reports. For example, it is possible that oppositional culture does not “exist” when using student self-reports, but does
“exist” when using teacher reports.
Presence and Persistence of Oppositional Identity
Ogbu (1991) questions why immigrant minorities are more successful than non- immigrant minorities. He puts forth a cultural model that highlights how various ethnic/racial/immigrant groups perceive the inner-workings of society differently. The perception adopted by groups governs how members of the group behave in society. He applies this model to the educational attainment of ethnic/racial/immigrant groups, citing how different orientations lead to divergence in achievement across groups. For example, he points out that immigrant groups succeed at higher rates because they are able to remain optimistic given their dual point of reference. This dual point of reference is non-existent for the U.S.-born, whose reference is whites. Thus, even though both immigrant and non-immigrant minorities may experience discrimination, the divergent ideologies lead to divergent outcomes.
He refers to the ideology adopted by U.S.-born minorities as a secondary cultural system.
This secondary cultural system governs the behavior that is appropriate for that group in any
20 given situation. These behaviors are unique to the group and are “symbols of identity to be maintained” (Ogbu 1991:15). Thus, there is stigma associated for minorities that excel academically because other minority students deem such behavior as “acting white.” In order to avoid this stigma, minority students are said to develop an oppositional identity that sees education as not being instrumental for the future. In developing that identity, behaviors that are detrimental to academic achievement (e.g. cheating, cutting class) surface (Ogbu 1991).
Ainsworth-Darnell and Downey (1998) challenge Ogbu by arguing that African
Americans are not only likely to have high expectations, but also view success in school in a positive manner. In fact, they are more likely than white students to report that education is a key determinant of future occupational attainment. African Americans are also less likely to think that “cheating is ok” and are more likely to see themselves as good students. They then move beyond abstract ideals to directly model behaviors using both student and teacher self- reports. Unlike abstract ideals, they find support for oppositional behaviors. For instance,
African Americans report doing less homework and are rated as more disruptive by teachers
(Ainsworth-Darnell and Downey 1998).
Alexander et al. (1997) highlight that the academic engagement of black students is similar to that of other racial groups early on but diverges as black students progress through their educational career. This is consistent with oppositional culture theory: minority students develop an adversarial stance as they grow older because the perception of limited opportunities becomes more salient over time. One would therefore expect an oppositional identity to increase as minority students progress through their academic career (Ogbu 1991). Nonetheless, the pattern of adoption of oppositional identity has not been fully developed in the academic
21 literature. For example, once a student adopts an oppositional identity, does it persist for the remainder of their academic career?
Farkas (2008) sheds further light on oppositional culture theory by looking to trends in school engagement among different racial/immigrant groups. He finds that school engagement
“declined significantly with grade level for both male and female White and African Americans, but for neither of the Hispanic groups” (Farkas 2008:328). Additionally, he finds that oppositional culture has a significant negative effect on reading achievement for all children
(Farkas 2008). This analysis was based on students between first and seventh grade, so it remains to be seen whether such trends hold for students in later grades. Furthermore, Farkas does not disaggregate the Hispanic group by national origin, which is problematic given the heterogeneity of Hispanic origin groups and the emphasis within the segmented assimilation theory on Mexican origin children.
Lundy and Firebaugh (2005) explore gender differences in oppositional culture. They find that males are in fact more resistant to schooling than females across all racial groups. In turn, this resistance to schooling leads to lower academic achievement for boys than girls. They therefore conclude that “the oppositional identity notion has merit, but that the oppositional identities are associated with gender, not with race or ethnicity” (Lundy and Firebaugh
2005:243).
Mickelson (2008) looks into the effect of the structure of opportunity by looking at middle school and high school students in Charlotte, North Carolina. She examines the relationship of family background and schooling variables on the abstract, concrete, and oppositional attitudes of Latino, African American, and White youth. She conducts factor analysis and confirms that “students simultaneously hold abstract, concrete, and oppositional
22 attitudes” (Mickelson 2008:365). She finds minimal differences in oppositional attitudes between groups at the high school level, but finds significant differences in middle school. This highlights either that oppositional attitudes is not a stable trait, or that those students who exhibit these behaviors drop out of school (so self-selection may be at work).
Thus, Ogbu argues not only that oppositional culture exists, but also that both African
Americans and Mexican Americans are likely to exhibit higher rates of oppositional ideology/behaviors than Whites. These hypotheses regarding the existence and group differences of oppositional culture are consistent with the predictions from segmented assimilation theory. Furthermore, it appears that there is a gendered aspect to oppositional culture that must be taken into account as well; boys are expected to exhibit higher rates of oppositional attitudes/behaviors than girls (Lundy and Firebaugh 2005).
In sum, the evidence on oppositional identity is mixed. Ainsworth-Darnell and Downey
(1998) find little evidence that an oppositional identity exists among African American tenth graders. Farkas (2008) finds that oppositional culture does exist among first to seventh graders.
Mickelson (2008) argues that self-selection may be at work at later stages of schooling, since students who engage in oppositional attitudes/behaviors may have dropped out at that point.
This self-selection argument might explain the conflicting findings between Ainsworth-Darnell and Downey (1998) and Farkas (2008). Longitudinal data on schooling behaviors/attitudes would allow us to identify if self-selection is at work and if oppositional culture is stable and persists over time.
PREDICTORS OF OPPOSITIONAL IDENTITY
Parental Involvement
23 Ogbu (2003) argues that families are important in creating/reinforcing oppositional identity. He points out how African American parents do not monitor their children enough, particularly in behaviors like doing homework. Similarly, Ogbu contends, while African
American parents may have high expectations for their children, they are not engaged in their children’s education (Ogbu 2003). Thus, Ogbu (2003) argues that parental involvement, as a measure of social capital within the family, is a source of oppositional identity. Coleman (1988) defines social capital as “some aspect of social structure…that facilitates actions of certain actors…within the structure…making possible the achievement of certain ends that in its absence would not be possible” (S98). He points out that social capital within the family exists only when strong relations between child and parents are present. Thus, the physical presence of adults is not enough; there must be an ongoing interaction between parent(s) and child.
A similar account is provided when describing the success of Vietnamese students in the
United States (Zhou and Bankston 1998). The ongoing engagement between child, parents, and the community fosters an environment that leads to positive educational outcomes. Zhou and
Bankston (1998) point out that families are able to utilize resources generated within the community to monitor the behavior of their children. This social control is pivotal to countering the negative aspects of the social context to which the second generation are exposed (Zhou and
Bankston 1998). Hao and Bonstead-Bruns (1998) similarly find that “frequent and high-quality parent-child interactions strengthen parent-child bonds, normative expectations, and a sense of obligation” (Hao and Bonstead-Bruns 1998). Thus, the lack of parental involvement is a potential reason that oppositional identity develops.
Parental involvement encompasses many interactions between children and parents. One such interaction is that of parents checking their children’s homework on a consistent basis.
24 Another interaction involves parents setting up a system to reward their children when they succeed academically (or consequences when they do not succeed). This type of social control not only encourages learning, but also enables parents to gain some working knowledge of their children’s education. Parental involvement in their children’s education has a positive effect on educational outcomes (Morrison 1978). According to Sui-Chu and Willms (1996), discussing school-related activities at home has the strongest relationship to academic achievement.
The experience of Vietnamese students highlights the positive effects of parental involvement on academic achievement (Zhou and Bankston 1998). This stands in contrast to the
African American experience. African American parents have high expectations, but do not become involved in their children’s education (Ogbu 2003). A similar argument is that Mexican
American parents do not value their children’s education. This translates into academic failure because “they [parents] fail to inculcate this value [education] in their children via academic socialization, and seldom participate in parental involvement activities in their home or the school” (Valencia and Black 2002). Thus, academic failure among African American and
Mexican American children is credited to deficiencies within their families. These deficiencies fail to foster the resources within the family necessary to ensure academic success. This argument may be inadequate because focusing on students and their families disregards structural forces in everyday life.
Portes and Rumbaut (2001) highlight the positive effects of parental involvement. They point out that immigrant girls are more likely to succeed given that they are more susceptible to their parents’ control (assuming that parents have high academic expectations for their daughters). This social control also protects girls “from assimilating oppositional views and behaviors at variance with those of their parent and home cultures (Rodriguez 34:2003). The
25 greater liberty afforded to males puts them at a higher risk of adopting an adversarial stance towards education (Portes and Rumbaut 2001).
Thus, the lack of social capital, and as indicated the lack of parental involvement, may lead to oppositional identity. First, the positive effect of parental involvement, an indicator of social capital, has been well documented (Portes and Rumbaut 2001; Zhou 1997; Zhou and
Bankston 1998). Second, Ogbu (2003) finds support for the prediction that parental involvement decreases oppositional identity. But Ogbu’s data is qualitative and the effect of parental involvement on oppositional attitudes and behaviors has yet to be examined using quantitative data. Nonetheless, it stands to reason that parental involvement should decrease oppositional identity among youth. Furthermore, the effect of oppositional identity on academic achievement may be upwardly biased with the omission of parental involvement.
Socioeconomic Status
Another potential source of oppositional identity involves the child’s socioeconomic status. Irving and Hudley (2008) find that an increase in socioeconomic status is associated with a decrease in oppositional identity. But they caution that “some students from high-SES households [that] still maintain the characteristics of oppositional identity” (Irving and Hudley
2008:390). This is consistent with Ogbu’s findings of the existence of oppositional identity in an affluent black community (Ogbu 2003). It remains to be seen whether such patterns are found among other ethnic/racial groups. Furthermore, the persistence of oppositional identity among high-SES minority youth may indicate that another contributing factor to oppositional attitudes/behaviors lies in the characteristics of the schools these children attend.
Blau and Duncan (1967) demonstrate that education is the main source of transmitting status from parent to child. Treiman and Yip (1989) argue that education is both the main source
26 of social reproduction and the primary engine of social mobility. Parental education has direct effects on their children’s educational attainment. Highly educated parents serve as role models for their children. Their children are able to see firsthand the “social and economic payoffs that good education could provide” (Ozturk and Singh 2006). Thus, it makes sense that socioeconomic status has an impact on the adoption of oppositional identity. In effect, highly educated parents reproduce in their children the linguistic and cultural competence valued in society (DiMaggio 1982).
Track Placement
One important structural component within the educational apparatus is tracking or the grouping students of similar ability. Tracking is an attempt by school administrators to match students’ abilities to a specific curriculum. Ability grouping is believed to maximize learning by providing students with appropriate content and pace of learning. Historically, social Darwinism justified tracking as the means of socializing children to meet the diverse demand of the economy—filling positions ranging from manual labor to those in need of a college education
(Oakes 2005). Today, track placement decisions are influenced by a variety of factors, including grades, language proficiency, teacher recommendations, scores on standardized exams, and social biases (Oakes and Guiton 1995; Xiong and Zhou 2006).
Oakes (2005) shows that teachers in low-track classes feel that the need to instill values and discipline is more important than the expansion of knowledge. Teachers also provide students in lower tracks with less content than students are capable of handling, thereby limiting the amount of learning (Good 1981; Ansalone 2001). Oakes (2005) shows the process by which these limited opportunities to learning emerge in that the time devoted to actual instruction was greater for those in higher-track classrooms than in low-track ones. While these observed
27 differences in instruction constituted only a few minutes per day, they quickly add up to a significant amount throughout the secondary education of students (Oakes 2005).
Teachers’ expectations conform to track level; teachers have higher expectations at the higher tracks and lower expectations in lower tracks (Wheelock 1992). Part of this may be due to the assumption by teachers that prior achievement is the only criteria by which track placement is determined. This reinforces the notion that those in the higher tracks are smarter and able to achieve better academically. These expectations then impact academic success by taking on the form of a self-fulfilling prophecy for their students (Good 1981). Engaging students differently according to preconceived notions only reinforces the perceived outcome.
Teachers’ credentials vary across track levels as well (Darling-Hammond 2004).
The aforementioned factors translate into distinct experiences for students in different tracks; students in lower tracks are offered a less effective curriculum (Xiong and Zhu 2006).
The distinct pattern of socialization across tracks may have consequences for students and their relationships to schooling. Ansalone (2001) highlights a potential mechanism: “by fostering a negative self-concept in a student, the tracking system hinders the self-actualization of the student and impacts negatively on the development of his/her personality” (44). Thus, it is possible that tracking may impact oppositional identity.
Mickelson (2008) examines the effect of tracking on the oppositional attitudes of 8th and
12th graders. He finds significant differences in oppositional attitudes by track such that students in more academically oriented tracks exhibited lower rates of oppositional attitudes net of race/ethnicity (Mickelson 2008). Interestingly, “Whites in regular and advanced levels held stronger oppositional attitudes than African Americans in these tracks” (Mickelson 2008:361).
28 Track placement also has a direct effect on academic achievement. Students in higher tracks find themselves in an environment that fosters greater academic success. Oakes (2001) points out that “track location has a far-reaching influence—with college track students (relative to their otherwise comparable non-college-track peers) enjoying better prospects for high school completion, college attendance, grades, and graduation” (239). Gamoran and Mare (1989) find a net advantage for those in the academic track of 20 percent for mathematics achievement and 50 percent for graduation (even after controlling for self-selection bias).
Gamoran (1987) argues that those in the academic track enjoy better educational opportunities that ultimately translate to better educational outcomes (through better instruction, qualified teachers, etc.). He concludes that “there is considerable evidence that within-school differences in student experiences contribute to variation in student achievement” (Gamoran
1987:153).
In sum, track placement has a significant effect on both oppositional identity and academic achievement. While there is an extensive documentation of track placement’s effects on academic achievement, the literature on the effect of tracking on oppositional identity is scarce (e.g. Mickelson 2008). Similarly, the effect of track placement on oppositional behaviors has not been examined to date. Moreover, the previous findings suggest that the effect of oppositional identity on academic achievement may decrease once track placement is controlled.
Therefore, it is worthwhile to examine the ways in which track placement impacts both the attitudes and behaviors of Mexican American, African American, Asian American, and White youth.
School Effects
29 A second structural component to take into account is schools. Ryabov and Van Hook
(2007) look at the impact of racial and socioeconomic school composition on academic achievement. Racial composition of the school does not have a significant impact on academic achievement, but socioeconomic composition does (Ryabov and Van Hook 2007). One explanation for this relationship revolves around peer networks established within schools.
Students in high-SES schools are more likely to be exposed to students with high expectations or other positive role models (Ryabov and Van Hook 2007). This suggests that an oppositional identity may be less likely to surface in high-SES schools relative to low-SES schools.
Nonetheless, Ogbu (2003) finds that this is not the case in that Black students in affluent schools are disengaged from education, which suggests that the racial composition of schools has no effect. On the other hand, Farkas (2008) finds that schools with a higher proportion of African
Americans and Hispanics exhibit higher rates of oppositional culture for students of all ethnicities.
Ogbu (2003) examined the extent of oppositional identity among African American students in a wealthy suburb in Ohio. This study is unique in that in takes place in one of the best school districts in the nation, which should be conducive to learning. Furthermore, Ogbu makes it a point to emphasize that most of the mechanisms most often cited in underachievement are non-existent in these schools. He finds that African Americans simply are not taking full advantage of the opportunities they are given. For example, he finds that some refuse to enroll in advanced courses because they felt they would be the only African American in that class. He also reports that academic disengagement among African Americans is common, even though they report high academic expectations. Thus, he finds that an oppositional identity still exists in this setting as well. Whether or not one can generalize to other similar schools remains to be
30 seen, but this highlights the possibility that school-level factors may play a minor role in the incidence of oppositional identity.
Fryer and Torelli (2005) offer an empirical analysis of the acting white hypothesis advanced by oppositional culture theory. They find that “high achieving students and students in environments with more interracial contact are most burdened by acting white” (2005:31). They argue that acting white does not seem to exist in schools that are predominantly African
American. This contradicts the findings by Ogbu and Fordham (1986), based in a predominantly
African American school yet are consistent with Ogbu’s most recent study in a racially diverse affluent suburb (Ogbu 2003).
CONSEQUENCES OF OPPOSITIONAL IDENTITY
Lundy and Firebaugh (2005) test the effect of oppositional identity on academic achievement. They find that “African Americans and Hispanics tend to have favorable academic attitudes but unfavorable academic behaviors” (Lundy and Firebaugh 2005:236). They conclude that the racial gap in behaviors does not explain the racial gap in academic achievement.
Mickelson (2008) examines how the structure of opportunity affects the academic achievement of African Americans and Whites. A key component of the structure of opportunity involves oppositional attitudes held by youth. She finds that “oppositional attitudes [have] a significant, direct, negative effect on achievement for twelfth and eighth graders” (Mickelson
2008:364). She measures academic achievement by using both fourth quarter GPA in eighth grade and cumulative GPA at 12th grade. The negative effect of oppositional attitudes persists even after controlling for track placement and having attended a highly segregated school.
31 Irving and Hudley (2008) use structural equation modeling to examine the effect of oppositional identity on academic achievement among African American males. They find support for the cultural mistrust, outcome expectations and outcome value, and resistant cultural value dimensions of oppositional identity. The ethnic identity affirmation factor was non- significant and they subsequently drop this factor. They find that “oppositional identity had an inverse relationship with academic achievement” (Irving and Hudley 2008:389). They operationalize academic achievement as the youth’s overall GPA during high school. It remains to be seen whether the effects of oppositional identity are similar when looking at other educational outcomes. Similarly, the focus on Black males precludes generalization to other racial/ethnic groups and females.
Harris and Robinson (2007) argue that previous studies of oppositional identity have not controlled for prior achievement; in doing so, the authors shed light on the potential bias of past research due to an omitted variable. They find that schooling behaviors are a function of prior achievement (Harris and Robinson 2007). Similarly, they find that “prior skills explain a greater proportion of the Black/White gap than do schooling behaviors” (149). Thus, it appears that the omission of prior achievement in previous studies overstates the effect of an oppositional orientation on academic achievement. They also find that prior skills do not have the same effect on Asians, highlighting the existence of an interaction between prior skills and racial/ethnic group on academic achievement.
The shortcoming of Harris and Robinson’s analysis is that the causal ordering may be reversed. An equally feasible alternative is that an adversarial stance in school leads students to under-perform in school. In turn, the continued under-performance leads to further oppositional behaviors/attitudes. If this scenario holds, the effect of oppositional culture is confounded with
32 that of prior skills in their analysis. Thus, it is imperative to disentangle the potential feedback loops between previous achievement and that of oppositional identity. The use of structural equation modeling facilitates these kinds of analyses. Similarly, it is imperative to examine if other omitted variables lead to overstating the effects of oppositional attitudes and/or behaviors on educational outcomes. Family background (e.g. social capital) and structural conditions within schools (e.g. tracking) are ideal candidates in this regard.
Similarly, it is important to look beyond achievement at the potential long-term effects of oppositional identity. Graduating from high school is important because it sets the foundation for subsequent life chances (Kao and Thompson 2003). Students who drop out from high school have diminished life chances. “The negative consequences of dropping out range from limited job opportunities, unemployment, and welfare dependency to criminal behavior” (Garner et al.
1997:36). Continuing in high school has the opposite effect, increasing the occupational opportunities and life chances for students.
DATA
This analysis will use the National Educational Longitudinal Survey of 1988 (NELS).
This survey began in 1988 by establishing benchmark information on a national sample of 7th graders. Parents, students, and teachers were interviewed during this initial phase of the survey.
Four follow-ups were conducted in 1990, 1992, 1994, and 2000. Students participated in all four follow-ups, teachers in two, and parents in just one subsequent administration of the survey. I will use data up until the third follow-up, which gives me longitudinal data from 8th grade, 10th grade, 12th grade, and around the time most of the students had already begun some kind of post- secondary education.
33 ORGANIZATION OF THE DISSERTATION
As I have argued, the oppositional identity literature is riddled with substantive and methodological issues. Studies have included various measures of oppositional identity, but it is unclear whether such measures are valid or reliable. In order to address this issue, recent studies
(e.g. Lindy and Firebaugh 2005; Taylor 2008) have employed factor analysis, but have not provided enough information about their procedures to evaluate their effectiveness. Even if the measures used do in fact capture oppositional identity, the source from which they are derived
(e.g. teachers, students, and/or parents) may lead to different conclusions.
Chapter 2 examines the consistency, construct validity, and longitudinal stability of oppositional identity. I find that there is a mismatch between student and teachers reporting on the same behaviors. I also find that oppositional identity is not unidimensional as has been assumed in previous studies, but can have as many as five sub-dimensions. Finally, I find that oppositional identity changes over time in multiple ways. This reality reinforces the fact that who is doing the reporting, what dimension is being examined, and when the study takes place matters.
Chapter 3 examines predictors of oppositional identity. I begin by examining whether there are differences in the degree of oppositional identity by race/ethnicity and gender. I also examine the effect of prior achievement, familial socioeconomic status, parental involvement, track placement, and school-level covariates on oppositional identity. I find that racial/ethnic differences fluctuate depending on the covariates included in the model, but that gender differences are consistent and significant.
Chapter 4 examines the consequences of oppositional identity. I examine dropping out as an outcome to explore the long-term effects of oppositional identity. I study the short-term
34 effect(s) of oppositional identity by looking at achievement in various both mathematics and reading achievement. I find that oppositional identity is a strong predictor of academic outcomes by lowering achievement and increasing the probability of dropping out. Nonetheless, I also find that while this effect of oppositional identity is strong, it does not account for much of the racial/ethnic gap in achievement.
My final chapter, Chapter 5, discusses my findings in more detail and situates them in the relevant literature. I also discuss the limitation of my analyses and future directions for the oppositional identity literature.
35 CHAPTER 2
THE MEASUREMENT OF OPPOSITIONAL IDENTITY
TEACHER AND STUDENT CONSISTENCY
Introduction
The consistency of oppositional identity measures is important because the source of information (e.g. students, teachers) is different across studies. Thus, discrepant results across studies may be due to method rather than the actual true construct (see Ainsworth-Darnell and
Downey 1998; Harrison and Robinson 2007). I begin by examining the consistency of student and teacher reports of oppositional identity behaviors. Following from past researchers’ conventions and given the data available, I use being late, being absent, and homework completion as measures of oppositional identity for this research question; I examine whether these measures are indicative of oppositional identity in subsequent analyses. I begin by exploring the differences between what teachers and students report regarding behavioral attributes that have traditionally been identified as indicative of oppositional identity.
In addition to behavioral attributes, abstract ideals—such as college expectations and believing that education is important for future—have been used to measure oppositional identity, but only students can report on these for themselves. The question of student-teacher consistency therefore does not apply to these abstract measures of ideals and feelings.
Accordingly, I focus on behaviors since behavioral attributes are observable by both students and teachers. I compare student reports of behaviors to that of reports by two teachers in the base year and first follow-up. Student reports are compared to reports by one teacher surveyed in the second follow-up. There are differences in the response categories for questions asked of teachers and students, which makes comparisons difficult. Therefore, my discussion focuses on
36 responses at the upper and lower ends of the response categories. For example, while the responses for the student question “How often comes to class without homework” include never, seldom, often, and usually, the responses for the teacher question “How often student completes homework” include all of the time, most of the time, some of the time, rarely, and never. These distributions make it difficult to compare nuanced “middle” answers, but the extremes (e.g. never) are more easily compared. Nonetheless, I provide tables with the full distributions of response categories for each question.
Student-Teacher Cross Tabulations
Tables 2.1 to 2.3 present consistency between teachers’ and students’ reports in the base year for three behaviors: homework completion, tardiness, and absenteeism. This wave of the survey provided teachers with limited response categories, which limits comparability with students’ responses.
Table 2.1 compares teacher versus student report homework completion. Teachers’ reports of whether students complete homework is compared with students’ responses of whether they come to class without homework. This table shows that roughly 19-20 percent1 of teachers report that students rarely complete homework, while roughly 21 percent of students report that they often/usually come to class without homework. Teachers’ and students’ reports indicate some relationship. For instance, 35-37 percent of teachers report that the student rarely completes his/her homework when the student reports often/usually coming to class without his/her homework. In contrast, only nine percent of teachers report that the student rarely completes homework when the student reports that he/she never comes to class without their homework.
1 I include a numerical range to account for the fact that there are two teachers reporting for each behavior in each table. 37 Table 2.2 compares teachers’ and students’ reports of being late. Teachers were asked whether a student is frequently tardy, which is best compared to the upper response of being tardy three or more days in the students’ responses. Table 2.2 shows that only 5-6 percent of teachers agreed with a student being frequently tardy, but roughly twice as many students (11 percent) reported being late three or more days. Table 2.2 indicates a weak relationship between teachers’ and students’ reports. Only 16-17 percent of teachers say the student is frequently tardy when the student reports being late three or more days, and only 3 percent of teachers report the student is frequently tardy when the student reports never being late for school.
Table 2.3 compares teachers’ and students’ reports of school absences. Table 2.3 shows a similar pattern for absenteeism, with teachers reporting that 10-11 percent of students are frequently absent versus 20 percent of students reporting that they were absent three or more days. Table 2.3 further shows that only 26-27 percent of teachers agree a student is frequently absent when students report missing school three or more day, while 4 percent of teachers report the student is frequently absent when the student reports never missing school. Thus, the relationship depicted in Table 2.3 is a bit stronger than Table 2.2, but still weak. Part of the reason for this weak relationship may not be due to actual disagreement, but inconsistent response categories given to teachers and students and the limited (true/false) response categories provided to teachers during this survey wave.
Tables 2.4 to 2.6 present consistency in behaviors as reported by teachers and students in the first follow-up. The response categories in this survey wave were not identical but were more similar than was used in base year survey. These tables show that there are sharp differences in what students and teachers report when looking across all three behaviors as well.
38 Table 2.4 shows that about 20 percent of students report that they never come to class without homework during the first follow-up, while both teachers reported that roughly 34 percent of students always complete homework. Thus, taking student reports at face value, we observe that teachers exhibit a bias toward more positive responses. Fifty percent of teachers report that the student completes homework all of the time when the student reports never coming to class without homework. Nineteen percent of teachers report that the student completes homework all of the time when the student reports often/usually coming to class without homework. At the other end of the continuum, 44-45 percent of teachers report that the student completes homework some, most, or all of the time when the student reports completing homework often/usually. And 16 percent of teachers report that the student completes homework some, rarely, or never when the student reports never coming to class without homework. Thus, Table 2.4 shows a moderate relationship between teacher and student reports in that some agreement between teacher and students exists.
This positivity bias is even higher when students and teachers are asked about being late.
Table 2.5 shows that roughly 27-28 percent of students report never being late while 54 percent of teachers report that students never being late. Table 2.5 also shows an overwhelming 69 percent of teachers report the student never being tardy when the student reports never being late for school. While this indicates consistency, 30-31 percent of teachers report the student never being tardy when the student reports being late over 7 times. At the other end of the continuum,
33 percent of teachers report that the student is tardy some, most, or all of the time when the student reports being late over 7 times. And few teachers (4-5 percent) report that the student is tardy some, most, or all of the time when the student reports never being late. This indicates some but not complete consistency between teachers’ and students’ reports.
39 Table 2.6 shows that roughly 15-16 percent of students report that they were never absent, while 10-11 percent of teachers report that students are never absent. Furthermore, 23-24 percent of teachers report students are never absent when students report they were never absent from school. And almost no teachers (1-2 percent) report that the student is never absent when the student reports being absent 11 or more days. At the other end of the continuum, 69-72 percent of teachers report that the student is absent some, most, or all of the time when the student reports being absent 11 or more days. And few teachers (10 percent) report that the student is absent some, most, or all of the time when the student reports never being absent.
Thus, overall there is disagreement between student and teacher reports across items for the first follow-up as well. This disagreement is more pronounced when looking at homework patterns and tardiness, while there appears to be more agreement in absenteeism. Similarly, there are stronger relationships during the first follow-up than the base year. Part of the reason for these stronger relationships is that the response categories for teachers in first follow-up nicely match the response categories for students, unlike the limited (two-category) responses in the base year.
Tables 2.7 to 2.9 shows similar patterns for teachers and students during the second follow-up. Again, there are differences between student and teacher reports. Again, the positive bias on the part of teachers observed during the first follow-up persists during the second follow- up as well.
Table 2.7 shows that about 23 percent of students report that they never come to class without homework during the first follow-up, while teachers reported that 29 percent of students always complete homework. Thus, similar percentages of students and teachers report on completing homework in the second follow-up. Nonetheless, an examination of the entire
40 distribution shows inconsistencies. For example, 45 percent of teachers report students complete homework all of the time when students report that they never come to class without homework, and 16 percent of teachers report that the student completes homework all of the time when the student reports coming to class without homework often/usually. At the other end of the continuum, 45 percent of teachers report that the student completes homework some, rarely, or never when the student reports coming to class without homework often/usually, and 16 percent of teachers report that the student completes homework some, rarely, or never when student reports never coming to class without homework.
Table 2.8 shows that 21 percent of students report never being late for school, while 46 percent of teachers agree with that statement. Sixty-four percent of teachers report that students are never tardy when students report they were never late for school, and 28 percent of teachers report that students are never tardy when students report being late over 7 times. At the other end of the continuum, 32 percent of teachers report that the student is tardy some, most, or all of the time when the student reports being late over 7 times, and few teachers (4-5 percent) report that the student is tardy some, most, or all of the time when the student reports never being late.
This indicates some but not complete consistency between teachers’ and students’ reports.
Finally, Table 2.9 shows that both students and teachers report that 10 percent of students were never absent. While this hints at consistency, there are inconsistencies when examining the entire table. For example, only 25 percent of teachers report students are never absent when the student reports never being absent, and almost very few teachers (3 percent) report that the student is never absent when the student reports being absent over 7 times. At the other end of the continuum, 57 percent of teachers report that the student is absent some, most, or all of the time when the student reports being absent 11 or more days, and 10 percent of teachers report
41 that the student is absent some, most, or all of the time when the student reports never being absent.
Overall, there appears to be more agreement between students and teachers during the second follow-up. Similarly, the greatest level of agreement occurs when reporting absenteeism, while the most disagreement occurs when reporting on tardiness.
While there are differences in what teachers and students report, some of the conclusions drawn are limited by both the way the questions were asked and the response categories attached to each question. This is most problematic during the base year with the true/false response categories provided to teachers, but was less of a limitation with the introduction of more categories during subsequent waves. Thus, I was able to compare the ends of the continuums
(never versus ever dichotomy) in the first and second follow-up, which provides a more robust comparison and allows for a more robust conclusion. Nonetheless, an examination of the full distribution of teacher and student reports show clear differences in what students and teachers report during all survey waves. Similarly, an interesting finding is that, while teachers were likely to concentrate their responses in the “positive” categories, students were more likely to use the full spectrum when reporting on all behaviors. In other words, teachers show systematic positivity bias for all behaviors during each survey wave.
Another systematic pattern is that the gap in disagreement narrows over time. One could argue that this is the case in comparing the base year and first follow-up simply because of the improved response categories in the first follow-up. But this does not explain all discrepancies since there are inconsistencies between teachers’ and students’ reports in second follow-up with a broader set of response categories.
42
Student-Teacher Relationships
In order to further show the relationship between teachers and students, Table 2.10 includes Spearman correlations in order to assess strength of association between teacher and student reports. I use Spearman correlations given the ordinal nature of the survey responses and the problems associated with different response categories discussed above. Overall, the correlations between students and teachers are low. This pattern is consistent when comparing students with both teachers. For example, the correlations between students and teachers for homework, tardiness, and absenteeism during the first follow-up are .23-.24, .29, and .41, respectively. This pattern of low correlations between student and teachers is consistent during the base and second follow-up as well. The only behavior where the correlation is of modest strength involves absenteeism at .41 during the first follow-up, but this correlation drops to .37 during the second follow-up.
I continue my analyses by comparing student responses to the average responses of both teachers. This comparison provides a more conservative estimate of students’ “overall” behaviors because it encompasses two different teachers reporting on two different classrooms and will therefore reduce measurement error associated with teacher reports. Table 2.11 shows the Spearman correlations for this analysis for both the base year and first follow-up, but not the second follow-up, because only one teacher was surveyed during this wave. I find moderate increases in the correlations for all behaviors during both survey waves using teacher averages rather than individual teacher reports. For example, the correlations between students and teachers for homework, tardiness, and absenteeism during the first follow-up are .27, .35, and .48 respectively, which are slightly higher than the original correlations (in Table 2.10) of .23-.24,
.29, and .41.
43 The results above indicate that there are clear differences in what students and teachers report. Consistency is weakest during the base year but these inconsistencies are evident even when analyzing subsequent survey waves. Similarly, the low correlations for all survey waves
(even after reducing measurement error) further indicates that the lack of consistency. Thus, it is important to take into consideration who is doing the reporting. These differences are due in part to the positivity bias exhibited by teachers. Similarly, there appears to be a temporal component because consistency increases over time. Nonetheless, it is not possible to determine what proportion of the difference is attributable to time, positivity bias, or simply measurement error.
However, the most important element is that there are significant inconsistencies. This suggests that teachers and student reports should not be used interchangeably. Thus, I will run parallel analyses using both students and teacher reported behaviors given the discrepancy in reporting between students and teachers. In doing so, I will be able to determine whether the observed differences between student and teacher reports lead to significantly different substantive conclusions.
VALIDITY OF OPPOSITIONAL IDENTITY MEASURES
I begin by examining the face validity of items used in previous studies of oppositional identity that used NELS. Table 2.12 shows a full list of these items, including whether the information comes from the student or teacher. I argue that sixteen out of the forty items that have been used in past studies are not actually measuring oppositional identity. These sixteen items are flagged with a “NO” in the face validity column. I then proceed to incorporate other items that I think capture the oppositional identity construct; Table 2.13 lists these items. Thus,
44 the full set of items that are used in these analyses are found in Table 2.13 and a subset of Table
2.12.
I begin my statistical analyses by splitting my sample into two equal halves. I conduct exploratory factor analysis using the program CEFA on the first half of the data to determine what factor model(s) best explains the data. I then run confirmatory factor analysis on the second-half of the data using the information I gathered from my exploratory factor analysis
(namely the number of factors and the items to use). I use the statistical program MPLUS
(Version 6.11) created by Muthen and Muthen (2010) to run the confirmatory factor analyses models. I use the diagonally weighted least squares estimator (WLSMV) to estimate my models. WLSMV uses a diagonal weight matrix to generate parameter estimates. This estimation method is preferable to traditional weighted least squares (WLS) when using ordinal data because WLS becomes unstable as number of items and model complexity increases. Both the Root Mean Square Error of Approximation (RMSEA) and Comparative Fit Index (CFI) are used to examine model fit because they have been found to perform well with ordinal data and
WLSMV (Yu 2002). RMSEA is a measure of discrepancy (model error) between the factor analysis model and population covariance matrix; RMSEA penalizes less parsimonious models.
The CFI places a given factor analysis model on a spectrum between a null model and an ideal model. Yu (2002) recommends a cutoff value of .05 for RMSEA and .95 for CFI as indicators of good model fit. I take these cutoffs into consideration when evaluating fit for all of my factor analysis models. Finally, I re-ran the confirmatory analyses after cross-validation so that the results presented here apply to the entire sample.
Teachers
45 I begin by examining the factor analysis model that fits best teacher reports during the base year. Table 2.14 shows results for the five behaviors reported by two teachers during base year. I find that a one-factor solution fits well for both teachers, with items measuring homework, tardiness, attentiveness making-up the non-compliant behaviors factor. The CFI is
1.0 for both teachers and the RMSEA is .042 for the first teacher and .037 for the second teacher.
Both teachers also exhibit similar trends in factor loadings, with most loadings being almost identical. Table 2A (in appendix) includes an alternative model that includes whether a student is frequently absent. This second model has an acceptable CFI value, but the RMSEA (.076,
.075) lies outside the acceptable level of .05. Exploratory analyses highlighted that the absent item did not correlate as well as the other items, but I included it in my confirmatory model to reinforce this point. Thus, I adopt a one-factor, four-item model for both teachers during the base year.
The first follow-up introduced three more items in the teacher reports. These items are whether teachers spoke to parents about behavior, homework, and/or absenteeism. Table 2.15 shows results for teachers during this wave of the survey. I find that a two-factor solution fits well for both teachers. The two-factor model has a CFI of 1.0 and a RMSEA of .043 and .044 for teacher one and two, respectively. Given the high degree of similarity between both factors, it is likely that they both capture non-compliance. Nonetheless, the newly introduced items are likely to be slightly more severe because they involve teachers having to be in contact with parents regarding the behaviors. Thus, items indicating whether a student usually does not work hard, does not complete homework, is tardy, and is inattentive make up the original non- compliant behaviors factor. The three newly introduced items make up the severe non-compliant behaviors factor because they are likely to involve more problematic behaviors.
46 The first follow-up also included other items that did not fit well with the two-factor solution. These alternative models are found in Tables 2B through 2D (in appendix). These alternative models were developed to highlight lack of fit of two items: absenteeism and disruptiveness. Table 2B includes the disruptive item in the first factor and as a result the
RMSEA increases above .05 (.059, .057). Table 2C incorporates the absenteeism measure, which leads to an even greater increase in RMSEA (.068, .075). Finally, the inclusion of both items in Table 2D leads the RMSEA to climb to .072 for teacher one and .076 for the second teacher. Thus, I adopt a bi-factor model for teachers during the first follow-up.
The second follow-up includes identical items to the ones found during the first follow- up. The only difference is that there is only one teacher reporting on students during the second follow-up. Table 2.16 shows the results for teachers during this wave of the survey. I find that a two-factor solution fits well with a CFI of .99 and a RMSEA of .048. This bi-factor solution is consistent with the results during the first follow-up. Thus, both the non-compliant behaviors and severe non-compliant behaviors factors remain during the second follow-up. Similarly, both the item measuring disruptiveness and the item measuring absenteeism do not fit the bi-factor solution well. Tables 2E and 2F show that the inclusion of the disruptiveness item leads to a
RMSEA of 0.071, while the inclusion of the absenteeism item leads to an even larger RMSEA at
0.086. The final model (Table 2G) shows that the inclusion of both items leads to the highest
RMSEA value at 0.089. These alternative models can be found in the appendix. Thus, I adopt a bi-factor model for teachers during the second follow-up as well.
These results highlight a few general themes. The first is that given the complexity of oppositional identity, one underlying factor does not appear to account for all the items that passed the face validity test. Similarly, other items still were unable to be accounted for with
47 either a single (base) or bi-factor (first and second follow-ups) solution. For example, an item measuring absenteeism did not fit well during all waves of the survey. Similarly, an item measuring disruptiveness did not fit during the first and second follow-ups with the introduction of an item measuring student effort. A related point is that an item measuring disruptiveness fit well during the base year, but not subsequent survey waves. This highlights the importance of validating items through factor analysis (thereby establishing factorial validity) even when face validity has been met.
Finally, these findings touch on an issue that surfaces in previous studies of oppositional identity: the assumption that various items carry an equal weight in gauging the degree of oppositionality among youth. My findings suggest that different items carry a different weight relative to others. For example, being tardy has the lowest factor loading during each wave.
Similarly, the absenteeism item had an even lower factor loading- presumably one of the reasons it did not fit the data well. Thus, it is important to take into consideration not only the factor structure underlying oppositional identity, but also the degree to which each individual item contributes to any given factor.
Students
Table 2.17 shows the results of my confirmatory factor analysis model for students during the base year. I find that a three-factor solution fits the data well with a CFI of 0.98 and
RMSEA of 0.036. The first factor consists of items measuring whether the student got into a fight, was sent to office for misbehaving, sent to office with schoolwork problems, and whether parents received a warning about behavior. This factor highlights delinquent behaviors that are most likely to capture the most severe forms of oppositional identity. The second factor consists of less severe items measuring absenteeism, tardiness, and truancy. This factor can be best
48 described as measuring non-compliant behaviors. The third factor involves three more items measuring how often the student comes to class without pencil/paper, books, and homework.
This is the least severe of the factors, likely capturing general schooling unpreparedness. The correlation between all three factors ranges from a high of .69 between the delinquent and non- compliant behaviors factors, to a low of .52 between the delinquent behaviors and school unpreparedness factors. Thus, these factors appear to be considerably correlated with one another.
Table 2.18 shows the results of my confirmatory factor analysis model for students during the first follow-up. This wave of the survey not only introduced more behavioral items, but also included measures of ideology. I find that a five-factor solution fits the data well with a
CFI of 0.96 and RMSEA of 0.029. Thus, the introduction of these measures of ideology added two more factors to the existing three-factors from the base year. The first ideological factor involves attitudes towards teachers and school, or deviant attitudes. This factor includes much more severe attitudes like thinking it is ok to destroy school property, abuse teachers, and/or steal belongings from school. The second factor involves how a student feels about being late, cutting class, and skipping class, which can be best described as non-compliant attitudes. The second follow-up also included items measuring whether students think it is ok to cheat or copy.
Exploratory analyses showed that these two items formed its own factor, but I did not include them in my results because a minimum of three items per factor is recommended when fitting confirmatory models to avoid identification issues (Brown 2006).
Table 2.19 shows the results of my confirmatory factor analysis model for students during the second follow-up. I find that a three-factor solution fits the data well with a CFI of
0.99 and RMSEA of 0.015. The first factor, delinquent behaviors, consists of items measuring
49 whether the student got into trouble, was put on in-school suspension, was suspended, transferred for disciplinary reasons, and being arrested. The second factor, non-compliant behaviors, consists of less severe items measuring absenteeism, tardiness, and truancy. The third factor, schooling unpreparedness, also emerges during the second follow-up with the same items from the base year and first follow-up. The correlation between all three factors ranges from a high of
.69 between the delinquent behaviors and non-compliant behaviors factors, to a moderate of .47 and .46.
My findings for students offer a much more complicated picture than when teacher reports are used. This is partly because there are more items for students than teachers. These results reinforce the notion that oppositional identity does not involve an underlying factor, but rather is a multi-dimensional construct. I find as many as three factors at work when I examine behavioral measures of oppositional identity. The first factor (delinquent behaviors) in all of the models is the most severe and likely the most indicative of extreme oppositional identity. The second factor contains items that capture less severe non-compliant behaviors, which may nonetheless negatively impact academic achievement. This second factor is comprised of items measuring cutting class, being late, and absenteeism. The third factor is the least severe with items measuring whether a student comes to class prepared. It remains to be seen how important this school unpreparedness factor is for academic achievement.
The introduction of ideological items during the first follow-up further complicates the matter with two other factors (for a total of five). The deviant attitudes factor highlights ideology about severe behaviors such as destroying school property, disobeying school rules, and abusing teachers. The non-compliant attitudes factor captures less severe ideology revolving around skipping class and being late to school. The moderate to high correlations between all
50 factors (behavioral and ideological) reinforces the notion of multiple dimensions, rather than distinctiveness in measurement. In other words, these models may be measuring distinct factors, but these factors as a whole appear to be related with each other.
I explicitly test the assertion regarding multi-dimensionality by employing a second-order factor analysis model. Table 2.20 shows the results for this second order capturing a global oppositional identity. This table highlights that a second-order factor model fits well during the base year with a RMSEA of 0.036 and CFI of 0.98. The second-order model fits well during the second follow-up as well. But is on the cusp of significance during the first follow-up showing that the RMSEA value is excellent at 0.035 during the first follow-up, but the CFI value of 0.94 is not above the 0.95 critical value cutoff recommended by Yu (2002), but close enough to be meaningful and warrant mentioning. Thus, while all factors might capture oppositional identity in some ways, the factor loadings suggest that using the factors interchangeably2 appears to be a bad idea (particularly the schooling unpreparedness factor with the lowest loading).
A related issue involves whether the ideological and behavioral factors can and should be used interchangeably as well. Both ideological factors are highly correlated with each other and all three behavioral factors are correlated with each other as well. The weakest correlations come from the schooling unpreparedness factor, which reaches a correlation of .44 with all other factors during the first follow-up. There are strong correlations between behavioral and ideological factors as well. For example, there is a strong correlation of .77 between the non- compliant behaviors factor and the non-compliant attitudes factor during the first follow-up.
There is a similar correlation of .69 between the delinquent behaviors and deviant attitudes factors. This suggests that while ideology and behaviors may be distinct factors, they potentially
2 I use the term “interchangeably” to refer to the practice of using different factors/items as if they were measuring the same thing and/or measuring it to the same degree. 51 gauge oppositional identity to the same degree (at the factor level). The second-order analyses above suggest that this might be the case3. In fact, the ideological factors have the highest factor loadings out of all the factors, even the behavioral ones. This might be due to the fact that
“oppositional identity” is in fact more of an ideological phenomenon, but the delinquent behaviors and non-compliant behaviors factors’ loadings are high as well.
A final theme regarding interchangeability involves the degree to which specific items within each factor capture oppositional identity. This is important to note because many of the measures that have been used in previous studies of oppositional identity find themselves loading on different factors in this study. The previous analyses highlights that differences in degree remain even when looking at items that capture the same factor. For example, while being absent is an item in the non-compliant behaviors factor, it is clearly the weakest of the three items in capturing this factor. Similarly, completing homework is the weakest of the items in the schooling unpreparedness factor. Thus, these issues of interchangeability both at the factor level and at the item level must be taken into account when looking at a multi-dimensional construct like oppositional identity.
This discussion highlights the importance of what is being used in measuring oppositional identity. Nonetheless, it is important to point out that what is being reported on is complicated by who is doing the reporting. That is, teachers and/or students may be in a better position to capture varying dimensions of oppositional identity. The results presented in the previous section indicated that teacher and student reports should not be used in place of each other because of the high degree of disagreement. A related point is that both teacher factors are comparable to the least severe of the student-derived factors, but not the most severe factors. In
3 This particular conclusion is tentative because the CFI value for the second-order model during the first follow-up was close (by .01), but did not meet, the recommended cutoff value for adequate model fit. 52 other words, teachers are only able to capture non-compliance and parts of school unpreparedness (homework) given they are only able to observe classroom-related processes.
Similarly, as mentioned previously, only students can report on internal attitudes and feelings.
This highlights that students may be in a better position to capture more of the complexity of oppositional identity than teachers. Nonetheless, the limited scope of using teacher reports does not negate the fact that they are still able to capture some aspect of oppositional identity.
STABILITY OF OPPOSITIONAL IDENTITY
Conflicting results across studies of oppositional identity may be due to examining oppositional identity during different points in students’ educational trajectories. In other words, oppositional identity may “look” different over time. Thus, a researcher examining oppositional identity during the early high school years may reach a different conclusion when examining high school seniors. This is a possibility for two reasons. The first is that students may exhibit some level of oppositional identity early on, but this level may increase or taper off over time.
The second is that the degree to which items capture oppositional identity may change over time.
My previous analyses find that the number of factors appears consistent over time, but it is still possible that the underlying structure of each factor has changed over time. For example, being constantly late and/or absent may “capture” oppositional identity differently over time.
Correlations over Time
I begin by examining the correlations among factors from the base year until the second follow-up for teachers. This portion of the analysis uses not only all the factors discussed in the previous section, but also all of the items. Table 2.21 presents these correlations using only one
53 teacher report4 given the consistency in teacher reporting cited above. The correlations of interest are those involving two different time points, which are enclosed with dotted lines. In other words, these particular correlations assess consistency for the same factors over time.
Nonetheless, I present all correlations for completeness and to serve as a reference point in my discussion (to be explained shortly).
The first pattern that becomes apparent is that the strongest relationship is among the same factor over time, particularly when examining adjacent survey waves. For example, students who scored high on the non-compliant behaviors factor during the base year are likely to score high in the same factors in subsequent waves. Specifically, the correlations between the base year non-compliant behaviors factor and the first and second follow-ups for the non- compliant behaviors factor are .54 and .46, respectively. The correlation between the first and second follow-up for the non-compliant behaviors factor is moderate as well, at .51.
Nonetheless, this pattern of high correlations does not always hold. For example, the correlation between the first and second follow-ups for the severe non-compliant behaviors factor is .25.
The second pattern that is apparent is that the numbers are much lower when we examine the other correlations. The exception is the two correlations (underlined) that capture relationships among factors within the same time point. For example, the correlation between the non-compliant behaviors factor and the severe non-compliant behaviors factor during the second follow-up is equal to .60. The reason why I included these intra-wave correlations is to provide a relevant reference point between factors. For example, we can see that the correlation between the non-compliant behaviors factor during the first follow-up and the severe non- compliant behaviors factor during the second follow-up is equal to .30. This is important
4 It is important to note that while I use only one teacher report during each survey wave, the teachers changed from year to year. 54 because it highlights the importance of both what one examines and when this examination takes place.
For example, given the correlation of .60 between the non-compliant behaviors factor and the severe non-compliant behaviors factor, it is likely that two researchers studying oppositional identity through using these two different constructs may reach different conclusions. Moreover, a researcher examining the non-compliant behaviors factor during the first follow-up and a researcher examining the severe non-compliant behaviors factor during the second follow-up
(instead of during the first follow-up) would reach even more diverging conclusions given the even lower correlation of 0.30 between the non-compliant behaviors factor during the first follow-up and the severe non-compliant behaviors factor during the second follow-up. The diverging conclusions these two researchers may reach are likely to be attributable to differences in both the factors and the time period they are studying.
Table 2.22 shows a similar table of correlations, but these correlations are using student reports. This table highlights a stronger relationship between the same factors over time than we see with teachers. For example, the correlation between the delinquent behaviors factor between the base year and first follow-up is equal to .70. Similarly, there is a correlation of .80 between the first and second follow-ups for the non-compliant factor. This table also highlights that the same factor always exhibits a stronger relationship with the adjacent survey wave. For example, even though the correlation for the delinquent behaviors factor between the base year and second follow-up is equal to .60, this is still lower than the same correlation between the base year and first follow-up (adjacent survey waves) at .72.
55 The low cross-factor correlations5 we see with teachers also exist for students, but we can see that this is not always the case. There appears to be moderate cross-factor correlations (in bold) between two different factors at two different time points. Interestingly enough, it appears that most of these correlations (except two) involve the delinquent behaviors factor. For example, there is a moderate correlation of .51 between the delinquent behaviors factor during the first follow-up and the non-compliant behaviors factor during the base year. Similarly, there is a moderate correlation of .54 between the non-compliant behaviors factor during the second follow-up and the non-compliant attitudes factor during the first follow-up. This suggests that student reports of oppositional identity are much more stable over time than those of teachers.
This assertion is further reinforced when we compare all student and teacher correlations. Table
2.21 shows that the lowest correlation is equal to .16 and the highest correlation is equal to .54.
These correlations are much lower than what we see in Table 2.22 using student reports, with the lowest and highest correlations equaling .24 and .80, respectively.
These results highlight some general themes. The first is that there appears to be considerable differences over time when examining oppositional identity. This change is particularly pronounced using teacher reports given the uniformly low correlations. One possible reason for this is that the same teacher is not surveyed in different waves, so there may be different teacher-student relationships emerging at each time point. Thus, it becomes impossible to disentangle “true” change from the error introduced by utilizing different teachers across waves. Even though I cannot disentangle the source, it is still important to highlight this change. After all, most survey data utilizes different teachers as students progress throughout their academic career. If this is the case, I argue that student reports might actually incorporate less error because we keep the “rater” consistent over time.
5 A cross-factor correlation refers to the correlation between two different factors at two different times. 56 The main drawback of these analyses is that each factor does not consist of the same items, so the variation might not be due to change, but rather to different items being used. I test this possibility by examining correlation among factors using only items that were consistent over time. The shortcoming of this approach is that I am not able to examine all possible interrelationships because similar items only surfaced for adjacent survey waves. In other words,
I am only able to present correlations between the base year and first follow-up or first follow-up and second follow-up. Nonetheless, this limited comparison will allow me to get a sense of how much variation observed in Tables 2.21 and 2.22 is due to utilizing different items and how much is due to actual change.
Table 2.23 shows the correlations among factors using teacher reports when the same items are used (unlike Table 2.21 which uses different items). The interesting thing is that this table is almost identical to Table 2.21, with most correlations deviating by .01. This suggests that the interrelationship between factors holds when using only a subset of the items. In other words, the low correlations seen in Table 2.23 confirm that there is disagreement in reporting and that the observed differences in Table 2.21 are not due to different items being used.
Table 2.24 shows similar correlations using student reports. Nonetheless, while there appear to be minor deviations of around .01 overall, there are some larger differences when compared to Table 2.22. For example, the correlation between the non-compliant behaviors factor between the first and second follow-up increases from .80 to 85. Similarly, the correlation between the non-compliant behaviors factor between the base year and first follow-up increases from .71 to 73. This suggests that part of the variation observed in Table 2.22 is due to having utilized different items. Nonetheless, as a whole, it appears that the interrelationship between factors holds when using only a subset of the items.
57 While there are clear limitations to the preceding analyses, my results suggest that the change exhibited by the same factors across time is not due to different items being used, but is indicative of actual change. Nonetheless, whether this change is true (alpha) change is tentative until I conduct longitudinal invariance analyses on these factors. While these limitations exist, it is also important to highlight is that these analyses have given me a sense on how measures of oppositional identity behave over time beyond what I could get from only focusing on the same items and same factors over time. For example, I am able to see that the full battery of items behave almost identical to the subset of items that are consistent over time. This suggests that the results of my forthcoming longitudinal invariance analyses might be generalizable to all items, not just the subset used. Similarly, I was able to examine the relationship of teacher factors during the first and second follow-ups with the base year, which is impossible in invariance analyses due to the different categories and items during the base year.
I am also able to examine cross-factor correlations to get a sense on how one factor is related with another factor during a different survey wave. For example, I am able to determine how the relative standing in the delinquent behaviors factor during one survey wave corresponds to the relative standing in the non-compliant behaviors factor during another survey wave. I find moderate cross-factor correlations using student reports, but these correlations are low when using teacher reports. This is important because the moderate to low correlations reinforces the pitfalls of not taking into account both what is being reported and when it is being reported6, particularly when using teacher reports. In sum, looking at these cross-factor correlations has allowed me to get a sense on how all items and factors behave over time.
Finally, these analyses have allowed me to examine how oppositional identity behaves over three periods. This is significant considering that my forthcoming longitudinal invariance
6 See “illustrative example” above. 58 models will only compare two survey waves at a time. The only three-wave longitudinal invariance model will model the schooling unpreparedness factor using student reports. I have found that adjacent survey waves have the strongest relationship, but that the same factor has moderate correlations later in time as well. This highlights that I will likely find change when examining two adjacent time points, but that I would likely find greater change taking place if I were in a position to examine longitudinal invariance from the base year to the second follow-up.
I now proceed to determine whether this tentative conclusion of change over time holds as I examine longitudinal measurement invariance.
Measurement Invariance
Measurement invariance analyses are necessary before one can determine if true change has taken place for a given construct. Measurement invariance is also able to examine whether an item (or items) that does a good job of capturing oppositional identity early on may lose its utility when students are surveyed at an older age. I examine these possibilities by looking at the invariance (or lack thereof) of items over time for both teachers and students. I examine invariance only for items that were asked for at least two time points. Thus, most analyses involve only two time points, with the exception of one student model that is able to utilize all three longitudinal time points. Similarly, I utilize only those models that the exploratory factor analyses and subsequent confirmatory analyses suggested were good fits for the data. I also focus on only one teacher given the consistent results across teachers for the previous confirmatory factor analysis models.
Chan (1998) argues that there are three types of change that may be encountered in studies of change over time: alpha, beta, and gamma. Alpha change refers to what most applied researchers are usually interested in: true change in a given construct over time. Beta change
59 occurs when the measurement properties of a construct are different over time, even though the underlying construct remains consistent. Finally, gamma change involves when the actual underlying construct changes over time. For example, gamma change is said to take place when a different amount of factors are needed to capture the underlying construct at different time points. Thus, accounting for both beta and gamma change is essential before one is able to measure true (alpha) change over time. I begin by examining gamma change by running the same factor model (and same items) during the two different time points. The goal here is to determine whether the same factor structure is similar at both time points. Nonetheless, I will also examine whether there are considerable similarities in factor loadings and/or correlations that might hint at beta change.
Table 2.25 shows the results for the only teacher model that compares factor models during the first and second follow-ups. The results indicate that the two-factor model is a good fit during the first and second follow-ups. Nonetheless, there are some minor changes in regards to fit, with the first follow-up solution exhibiting a better fit. Specifically, the RMSEA value is
.031 during the first follow-up, but increases to .049 during the second follow-up. The factor loadings also exhibit some change, particularly in regards to the second factor. Finally, the correlation between factors increases dramatically during the second follow-up by almost a factor of two (from .32 to .61). Thus, there appears to be no gamma change (factor structure is the same), but beta change is possible.
Table 2.26 shows results from the first student model that compares the base year with the first follow-up. The results indicate that the two-factor solution fits very well during both the base year and the first follow-up. The RMSEA is 0.033 during the base year and increases slightly to 0.035 during the first follow-up. Nonetheless, factor loadings and correlations exhibit
60 some change between the base year and first follow-up. One shortcoming of this comparison is that while the questions asked during both waves were similar, the response periods are different.
The questions during the base year asked students to report on whether they engaged in an activity “in the first four weeks,” while the first follow-up encompassed “the first half of the current school year.” Thus, this particular model is likely to overestimate the amount of variation across time. The results suggest that gamma change is non-existent, but beta change is definitely possible.
Table 2.27 shows results for the second student model that compares factor models from the first and second follow-ups. This comparison does not suffer the same shortcomings associated with the base year and first follow-up comparison. The CFI, RMSEA, and factor correlations are very similar for both time points. Similarly, while there is slight variation in factor loadings, the changes appear to be both minor and/or restricted to a few items. Thus, there is no gamma change and there also appears to be an absence of beta change during the second follow-up. This result is more robust when compared to the results from the earlier student model (base year and first follow-up) and the sole teacher model (first and second follow-ups).
Table 2.28 shows the final model for that examines invariance for students from the base year to the second follow-up. Unfortunately, there were only three items in this analyses that were consistent for all three points. Therefore, my ability to comment on fit is non-existent because a minimum of four items is needed for a one-factor solution to be properly identified
(Brown 2006). Nonetheless, I am still able to comment on the factor loadings in order to attempt to detect invariance. Table 17 shows that the factor loadings increase from the base year to the first follow-up, but stabilize between the first and second follow-ups.
61 The results of this section indicate that gamma change has not occurred for any model for both teachers and students. Nonetheless, the existence of beta change is a possibility for some of these models. For example, the greater amount of change in factor loadings and correlations exhibited between teachers’ measures during the first and second follow-ups hints that beta change may be taking place. This might be due in part to the fact that the same teachers are not surveyed in both waves and different teachers could have different relationships to students.
Nonetheless, I cannot argue that beta change has occurred until formal tests are conducted.
Brown (2006) points out, “If markedly disparate measurement models are obtained between groups [when analyzing them separately], this outcome would contraindicate further invariance evaluation” (271). None of my models meet this criteria; I therefore perform further evaluation of invariance for all previous analyses.
I continue my measurement invariance analyses by testing for beta change with an unrestricted model that contains data from both time points. This first model does not impose any constraints on the coefficients, so it is similar to the previous analyses of running them independently. I then compare the fit of the unrestricted model with a second model that tests for equality of factor loadings. The acceptance of this model is often referred to as establishing weak factorial invariance. The third and final model I test involves placing equality constraints on not only factor loadings, but on thresholds and intercepts. The acceptance of this model is often referred to as establishing strong factorial invariance. Only after establishing strong factorial invariance can I continue to examine alpha (true) change.
I look to differences in the Comparative Fit Index (CFI) to evaluate the relative fit of alternative models, which deviates from the traditional approach of utilizing chi-square differences. I employ this alternative strategy given chi-square’s sensitivity to large sample sizes
62 (Cheung and Rensvold 2002). I reject models that deviate by more than 0.01 (∆ .01) in CFI from the unconstrained model; this threshold is consistent with previous research (Cheung and
Rensvold 2002; Chen 2007). In other words, I reject the weak factorial invariance model (which tests that the factor loadings are equal at both time points) if the difference in CFI values is greater than 0.01 when compared to the unconstrained model. If this were the case, I would not continue any further and conclude that beta change has taken place. Nonetheless, if this were not the case, I would then proceed to compare the CFI values between the unrestricted model and the strong factorial invariance model.
I begin by examining temporal factorial invariance for the sole teacher model for both the first and second follow-ups. Table 2.29 shows that the unconstrained model fits the data well, with a RMSEA of 0.030 and CFI of 0.991. The weak factorial invariance model fits the data just as well, with identical RMSEA and similar CFI value (0.001 difference). I therefore find support for weak factorial invariance, which means I can proceed to test strong factorial invariance. The final model’s RMSEA value rises to 0.038, while the CFI value drops to 0.982, suggesting that it too fits the data well. Having established strong factorial invariance, I proceed to examine mean differences over time. I find significant differences over time for both factors.
The coefficient of change is equal to 0.261 for the non-compliant behaviors factor and -0.243 for the severe non-compliant behaviors factor. This suggests that both factors have an inverse relationship, where the first factor increases over time, while the second one decreases.
I now turn to the student model that tests for factorial invariance between the base year and first follow-up. Table 2.30 shows that the unconstrained model fits the data well, with a
RMSEA of 0.022 and CFI of 0.973. The weak factorial invariance model fits the data just as well, with a RMSEA of 0.021 and CFI of 0.974. I now proceed to test strong factorial invariance
63 having established weaker forms of invariance. The final model’s RMSEA value rises to 0.033, while the CFI value drops to 0.925, suggesting that this third model does not fit the data well.
This result suggests that the measurement properties of the items have changed over time. In other words, the underlying structure of the construct itself has changed (beta change). Thus, while I find significant differences over time for both the non-compliant behaviors and schooling unpreparedness factors (1.402 and -0.271, respectively), I cannot determine what proportion of this change is true (alpha) or beta change.
The second student model tests for factorial invariance between the first and second follow-ups. Table 2.31 shows that the unconstrained model fits the data well with a RMSEA of
0.020 and CFI of 0.968. The weak factorial invariance model fits the data just as well, with a
RMSEA of 0.019 and CFI value of 0.970. I therefore find support for weak factorial invariance, which means I can proceed to test strong factorial invariance. The final model’s RMSEA value remains stable at 0.019, while the CFI value drops to 0.964, suggesting that it too fits the data well. Having established strong factorial invariance, I proceed to examine mean differences over time. I significant differences for the delinquent behaviors and non-compliant behaviors factors, but non-significant differences over time for the schooling unpreparedness factor. The coefficients of change are equal to -0.102, 0.504, and 0.025, respectively. This result is consistent with the first student model in that the non-compliant behaviors factor is positive and significant, but different in that the difference in the first model is due to both alpha and beta change.
Table 2.32 shows final student model tests for factorial invariance between the base year and second follow-up. The unconstrained model fits the data reasonably well with a RMSEA of
.050 and CFI of .959. The weak factorial invariance model fits slightly better, with a RMSEA of
64 0.044 and CFI of 0.963. I now proceed to test strong factorial invariance having established weaker forms of invariance. The final model’s RMSEA value rises dramatically to 0.102, while the CFI exhibits a similar drop to 0.706, suggesting that this third model definitely does not fit the data well. This result suggests that the measurement properties of the items have changed from the base year to the second follow-up. Thus, while I find significant differences over time for the schooling unpreparedness factor across two waves (-0.297 and -0.076, respectively), I cannot determine what proportion of this change is true (alpha) change over time.
Overall, I find that oppositional is not a stable trait in multiple ways. The first way surfaced when I examined the correlations of factors over time for both teachers and students.
These analyses highlighted that the degree of oppositional identity exhibited by students may depend on the time point one examines. Again, I cannot trace the source of the variation, but its existence is what matters. Thus, the temporal aspect of oppositional identity must be taken into account before comparisons are made over time. Similarly, it is important to reemphasize what measures one uses may interact with when these measures are used. After all, I also found that the same set of correlations during one survey wave decrease dramatically with the introduction of cross-factor correlations. The hypothetical scenario discussed above reinforces this point.
The measurement invariance analyses also highlight that real (alpha) change takes place over the academic trajectory of students. For example, I find significant mean differences in factor means when examining oppositional identity during the first and second follow-ups using teacher reports. Nonetheless, while it is important to highlight that I found mean differences over time in some cases, I did not in others. For example, the examination of oppositional identity using student reports for this same period (first and second follow-ups) yielded non- significant results for one of the three factors.
65 The measurement invariance analyses also revealed a second way oppositional identity changes over time: beta change. The comparison between the base year and first follow-up using student reports highlighted that beta change took place between these two time points. Similarly, the sole three-wave model (using student reports) yielded similar results. Chan (1998) best summarizes beta change when he points out that “the observed change results from an alteration in the respondent's subjective metric or evaluative scale rater than an actual change in the construct of interest” (414). Thus, this must be taken into account when choosing measures to capture oppositional identity during a specific time period (e.g., 12th grade), but particularly when comparisons over time are to be made (e.g., 8th grade to 12th grade).
Finally, it is also important to highlight that I found evidence of gamma change in my analyses. Part of the reason the base year factor using teacher reports was not included in the analyses because the questions asked during this wave do not match those found during the two subsequent follow-ups. Nonetheless, another reason was that while the items themselves were similar, the factor structure was slightly different during the base year and subsequent survey waves as well. In other words, a different set of items clustered to create the non-compliant behaviors factors during the base year relative to the first and second follow-ups. Specifically, an item measuring disruptiveness fit well during the base year, but did not fit well in subsequent analyses. This fits the definition of gamma change put forth by Chan (1998) in that a different set of factors is necessary to capture the underlying construct.
CONCLUSION
This chapter has allowed me to get a better sense of “how” oppositional identity exists.
My results highlight the importance of who is doing the reporting in studies of oppositional
66 identity. I find low to moderate correlations between teachers and students in regards to reported behaviors. While it is possible that these patterns may not hold for all items, it is important to note that the items I used to gauge consistency between teachers and students can be found in two different factors for students. Thus, two out of three behavioral factors in my confirmatory factor analysis models have an item where students and teachers disagree. Thus, I argue that this disagreement is likely to hold across other items and factors.
My results also indicate that multiple items are necessary to capture the multidimensional construct of oppositional identity. There are some items that are severe, while others may exhibit noncompliance and/or lack of effort on the part of students. This points to the possibility that the different conclusions that have been drawn in previous research is driven by what items are chosen to capture oppositional identity. Results indicate that all of these factors are in fact correlated, but the degree to which these factors are related may vary considerably. Thus, the degree to which conclusions are affected may vary depending on the items chosen.
Finally, I find support for the notion that measures of oppositional identity vary over time. This is important because it highlights that even the same items can lead to different conclusions depending on the time period. The reasons for this are twofold. The first is that some items may do an adequate job in capturing oppositional identity at one time point, but the underlying structure of this item may be different at other times. Furthermore, even in the presence of factorial invariance, I find that there are differences in oppositional identity over time. Both of these findings highlight the importance of taking into account the when in studies of oppositional identity.
My findings have significant ramifications for my future analyses. First and foremost, as a result of the discrepancy in reporting, I argue that teacher and student reports should not be
67 used interchangeably. Thus, it is best to conduct separate analyses using student and teacher reports. By doing so, I will be able to determine whether the observed differences between these two sets of analyses lead to significantly different substantive conclusions. The existence of multiple dimensions will also allow me to examine whether differences exist when looking at different dimensions. For example, I may find no differences across racial groups when examining one dimension, but this difference may surface when looking at another dimension.
This point is particularly relevant considering that many of the items used in previous studies ended up being a part of related, but distinct factors.
I will also take advantage of the fact that I have multiple time points when examining different aspects of oppositional identity. For example, the inclusion of items that measure ideology during the first follow-up will allow me to capture another dimension that focusing on the base year or second follow-up would not allow me to do. Similarly, it is possible that substantive conclusions will be different when examining different time points individually or longitudinally. Ultimately, I will take into account who, what, and when individually, but also how they may interact to create potentially divergent substantive conclusions.
68
Table 2.1 Student versus Teacher Reported Homework Patterns, Base Year
Teacher One: True if student rarely completes homework
Student: How often comes to class without No Yes Column % (N) homework?
Never 90.9 9.1 28.2 (5819) Seldom 82.2 17.8 50.9 (10505) Often/Usually 64.9 35.1 20.9 (4313) Total 81.0 19.0 100 (20637)
Teacher Two: True if student rarely completes homework
Student: How often comes to class without No Yes Column % (N) homework?
Never 90.6 9.4 28.2 (5802) Seldom 81.7 18.3 51.0 (10476) Often/Usually 63.3 36.7 20.8 (4282) Total 80.4 19.6 100 (20560)
69
Table 2.2 Student versus Teacher Reported Late/Tardy Patterns, Base Year
Teacher One: True if student frequently tardy
Student: How many times No Yes Column % (N) late for school?
None 96.9 3.1 63.1 (13032) 1 or 2 days 93.3 6.7 25.4 (5246) 3 or more days 84.4 15.6 11.4 (2359) Total 94.6 5.4 100 (20637)
Teacher Two: True if student frequently tardy
Student: How many times No Yes Column % (N) late for school?
None 97.1 2.9 63.3 (13021) 1 or 2 days 92.6 7.4 25.3 (5200) 3 or more days 83.2 16.8 11.4 (2339) Total 94.4 5.6 100 (20560)
70
Table 2.3 Student versus Teacher Reported Absenteeism, Base Year
Teacher One: True if student is frequently absent
Student: Number of days No Yes Column % (N) missed from school
None 96.3 3.7 46.7 (9639) 1 or 2 days 89.9 10.1 33.1 (6828) 3 or more days 74.0 26.0 20.2 (4170) Total 89.6 10.4 100 (20637)
Teacher Two: True if student is frequently absent
Student: Number of days No Yes Column % (N) missed from school
None 95.7 4.3 46.8 (9625) 1 or 2 days 89.7 10.3 33.1 (6810) 3 or more days 72.9 27.1 20.1 (4125) Total 89.1 10.9 100 (20560)
71
Table 2.4 Student versus Teacher Reported Homework Patterns, First Follow-Up
Teacher One: How often student completes homework? Some of Student: How often comes All of Most of the the time, Column to class without (N) the time time Rarely, % homework? Never
Never 49.9 33.8 16.2 19.6 (2943) Seldom 33.1 40.4 26.5 62.3 (9345) Often/Usually 18.9 35.6 45.5 18.1 (2719) Total 33.8 38.2 27.9 100.0 (15007)
Teacher Two: How often student completes homework? Some of Student: How often comes All of Most of the the time, Column to class without (N) the time time Rarely, % homework? Never
Never 50.3 34.0 15.7 19.6 (2240) Seldom 33.8 40.9 25.3 62.7 (7167) Often/Usually 18.8 37.0 44.2 17.6 (2016) Total 34.4 38.9 26.8 100 (11423)
72
Table 2.5 Student versus Teacher Reported Late/Tardy Patterns, First Follow-Up
Teacher One: How often is student tardy?
Some, Student: How many times Never Rarely Most, All Column % (N) late for school? of the time
Never 68.7 26.9 4.4 26.8 (4021) 1 to 2 times 56.6 33.9 9.5 37.8 (5679) 3 to 6 times 43.6 37.5 18.9 22.4 (3368) Over 7 times 30.2 37.1 32.7 12.9 (1939) Total 53.5 33.2 13.3 100 (15007)
Teacher Two: How often is student tardy?
Some, Student: How many times Never Rarely Most, All Column % (N) late for school? of the time
Never 68.9 25.8 5.3 27.5 (3145) 1 to 2 times 58.0 33.2 8.8 37.7 (4310) 3 to 6 times 42.7 37.6 19.7 22.4 (2561) Over 7 times 30.9 36.0 33.0 12.3 (1407) Total 54.2 32.5 13.3 100 (11423)
73
Table 2.6 Student versus Teacher Reported Absenteeism, First Follow-Up
Teacher One: How often is student absent?
Some, Student: How many days Never Rarely Most, All Column % (N) absent for school? of the time
Never 23.8 65.8 10.4 15.2 (2276) 1 or 2 days 15.0 69.6 15.4 24.6 (3691) 3 or 4 days 7.8 65.7 26.5 27.5 (4129) 5 to 10 days 4.3 49.8 45.9 22.9 (3438) 11 or more days 1.1 27.2 71.7 9.8 (1473) Total 10.5 59.3 30.2 100 (15007)
Teacher Two: How often is student absent?
Some, Student: How many days Never Rarely Most, All Column % (N) absent for school? of the time
Never 22.5 68.7 8.8 15.5 (1774) 1 or 2 days 14.0 71.1 14.9 25.0 (2857) 3 or 4 days 7.2 65.9 26.9 27.6 (3156) 5 to 10 days 4.2 50.0 45.7 22.5 (2573) 11 or more days 1.7 29.4 69.0 9.3 (1063) Total 10.1 60.6 29.2 100 (11423)
74
Table 2.7 Student versus Teacher Reported Homework Patterns, Second Follow-Up
Teacher One: How often student completes homework? Some of Student: How often comes All of Most of the time, to class without Column % (N) the time the time Rarely, homework? Never
Never 44.5 39.2 16.4 23.0 (2146) Seldom 27.2 43.0 29.8 61.1 (5703) Often/Usually 15.9 39.4 44.7 15.9 (1484) Total 29.4 41.6 29.0 100 (9333)
75
Table 2.8 Student versus Teacher Reported Late/Tardy Patterns, Second Follow-Up
Teacher One: How often is student tardy?
Student: How many Some, Most, Never Rarely Column % (N) times late for school? All of the time
Never 64.0 31.2 4.8 20.6 (1926) 1 to 2 times 52.2 38.4 9.5 33.0 (3078) 3 to 6 times 39.6 41.6 18.9 25.1 (2346) Over 7 times 27.9 40.4 31.6 21.2 (1983) Total 46.3 38.1 15.6 100 (9333)
76
Table 2.9 Student versus Teacher Reported Absenteeism, Second Follow-Up
Teacher One: How often is student absent?
Student: How many times Some, Most, Never Rarely Column % (N) missed school? All of the time
Never 25.4 64.5 10.1 10.2 (951) 1 to 2 times 12.9 67.9 19.2 33.3 (3106) 3 to 6 times 6.4 60.0 33.6 33.6 (3139) Over 7 times 3.0 40.0 57.0 22.9 (2137) Total 9.7 58.5 31.8 100 (9333)
77
Table 2.10 Spearman Correlations for Homework, Tardiness, and Absenteeism between Students and Teachers, All Waves
Teacher One Base Year First Follow-Up Second Follow-Up Student Homework Late Absent Homework Late Absent Homework Late Absent
Homework 0.22 0.24 0.23 Tardy 0.16 0.29 0.30 Absent 0.26 0.41 0.37
Teacher Two Base Year First Follow-Up Second Follow-Up Student Homework Late Absent Homework Late Absent Homework Late Absent
Homework 0.23 0.23 * Tardy 0.18 0.29 * Absent 0.26 0.41 *
78
Table 2.11 Spearman Correlations for Homework, Tardiness, and Absenteeism between Students and Teachers' Average, All Waves
Both Teachers’ Average Base Year First Follow-Up Second Follow-Up Student Homework Late Absent Homework Late Absent Homework Late Absent
Homework 0.27 0.27 * Tardy 0.21 0.35 * Absent 0.30 0.48 *
79
Table 2.12 Items of Oppositional Identity from Previous Studies with Face Validity, NELS Report Previously Face Type Used Validity Good grades are important to me Student YES YES Try as hard as you can in math, English, history, and science Student YES YES Come to class without pencil Student YES YES Come to class without paper Student YES YES Come to class without books Student YES YES Come to class without homework done Student YES YES Other students see me as a troublemaker Student YES YES Think it is OK to cut a couple of classes Student YES YES Think it is OK to skip school for a whole day Student YES YES Think it is OK to talk back to teachers Student YES YES Think it is OK to disobey school rules Student YES YES Think it is OK to cheat on tests Student YES YES Think it is OK to copy someone else’s homework Student YES YES Feel it is ok for you to cut a couple of classes? Student YES YES Feel it is ok for you to be late for school? Student YES YES Feel it is ok for you to copy someone else’s homework? Student YES YES I got in trouble for not following school rules Student YES YES I was put on an in-school suspension Student YES YES I was suspended or put on probation from school Student YES YES I was transferred to another school for disciplinary reasons Student YES YES I was arrested Student YES YES
How often is this student attentive in class? Teacher YES YES Does this student usually work hard for good grades? Teacher YES YES How often does this student complete homework assignments? Teacher YES YES How often is this student disruptive in class? Teacher YES YES
80
Table 2.12, contd. Items of Oppositional Identity from Previous Studies with No Face Validity, NELS Report Previously Face Type Used Validity Education is important to getting a job later on Student YES NO Other students see me as a good student Student YES NO I get a feeling of satisfaction from doing what I’m supposed to do in class Student YES NO When you work hard on schoolwork, your teachers praise your efforts Student YES NO Most of your teachers really listen to what you have to say Student YES NO The teaching is good at this school Student YES NO Teachers are interested in students Student YES NO Discipline is fair Student YES NO In class you often feel “put down” by your teachers Student YES NO Other students see me as popular Student YES NO Other students see me as socially active Student YES NO Other students see me as part of the leading crowd Student YES NO I make friends easily with girls/boys Student YES NO I have good friends who are members of my own sex Student YES NO I feel well accepted among peers Student YES NO Occupation expected at age 30 Student YES NO
81
Table 2.13 Items of Oppositional Identity Not Previously Used and with Face Validity, NELS Previously Face Report Type Used Validity
Respondent sent to office for misbehaving Student NO YES Respondent sent to office with school work problems Student NO YES Parents received warning about attendance Student NO YES Parents received warning about grades Student NO YES Parents received warning about behavior Student NO YES Respondent got into fight with another student Student NO YES How often respondent cuts or skip classes Student NO YES How many times was respondent late for school Student NO YES Respondent feels it's ok to work hard for good grades Student NO YES Respondent feels it's ok to get into physical fights Student NO YES Respondent feels it's ok to steal belongings from school Student NO YES Respondent feels it's ok to destroy school property Student NO YES Respondent feels it's ok to abuse teachers Student NO YES
Student rarely completes homework Teacher NO YES Student is frequently absent Teacher NO YES Student is frequently tardy Teacher NO YES Student exceptionally passive/withdrawn Teacher NO YES How often student is absent Teacher NO YES How often student is tardy Teacher NO YES Student is exceptionally passive Teacher NO YES
82
Table 2.14 Factor Loadings of Items From Teachers, Base Year
Confirmatory Factor Analysis
Teacher One Teacher Two Non-Compliant Non-Compliant Item Behaviors Behaviors
True if student rarely completes homework 0.83 0.84 True if student is frequently tardy 0.67 0.69 True if student is inattentive in class 0.97 0.97 True if student is frequently disruptive 0.74 0.75
CFI 1.00 1.00 RMSEA 0.042 0.037
83
Table 2.15 Factor Loadings of Items From Teachers, First Follow-Up
Confirmatory Factor Analysis
Teacher One Teacher Two Severe Severe Non- Non- Non- Non- Compliant Compliant Compliant Compliant Behaviors Behaviors Item Behaviors Behaviors
True if student usually does NOT work hard 0.93 0.93 How often student does NOT complete homework 0.90 0.90 How often student is tardy 0.53 0.52 How often student is inattentive in class 0.85 0.85
True if spoke to parents about student behavior 0.91 0.91 True if spoke to parents about student homework 0.95 0.95 True if spoke to parents about student absenteeism 0.80 0.79
Correlation between factors 0.35 0.36
CFI 1.00 1.00 RMSEA 0.043 0.044
84
Table 2.16 Factor Loadings of Items From Teachers, Second Follow-Up
Confirmatory Factor Analysis
Severe Non- Non-Compliant Compliant Behaviors Item Behaviors
True if student is NOT motivated to work hard for good grades 0.91 How often student does NOT complete homework 0.89 How often student is tardy 0.55 How often student is inattentive in class 0.83
True if spoke to parents about student behavior 0.88 True if spoke to parents about student homework 0.86 True if spoke to parents about student absenteeism 0.84
Correlation between factors 0.60
CFI 0.99 RMSEA 0.048
85
Table 2.17 Factor Loadings of Items From Students, Base Year
Confirmatory Factor Analysis
Non- Delinquent Schooling Compliant Behaviors Unpreparedness Item Behaviors
Respondent sent to office for misbehaving 0.90 Respondent sent to office with school work 0.74 problems Parents received warning about behavior 0.87 Respondent got into fight with another student 0.70
Parents received warning about attendance 0.77 How many times respondent cuts or skip classes 0.80 How many times respondent is late to class 0.57 How many times respondent missed school 0.38
How often respondent comes to class without 0.71 pencil/paper How often respondent comes to class without books 0.76 How often respondent comes to class without 0.70 homework
Correlation between factors (1:2, 1:3, 2:3) 0.69 0.52 0.54
CFI 0.98 RMSEA 0.036
86
Table 2.18 Factor Loadings of Items From Students, First Follow-Up Confirmatory Factor Analysis
Non- Non- Delinquent Schooling Deviant Compliant Compliant Behaviors Unpreparedness Attitudes Item Behaviors Attitudes
Parents received warning about behavior 0.76 Respondent got into fight with another student 0.68 How many times respondent got into trouble 0.86 How many times respondent was put on in-school 0.72 suspension How many times respondent was suspended from school 0.74 How many times respondent was transferred for 0.66 disciplinary reasons How many times respondent was arrested 0.75
Parents received warning about attendance 0.62 How many times respondent cut/skip class 0.91 How many times respondent was late for school 0.63 How many times respondent missed school 0.48
How often respondent comes to class without pencil/paper 0.71 How often respondent comes to class without 0.80 books How often respondent comes to class without homework 0.68
Respondent feels it is OK to steal belongings from school 0.78 Respondent feels it is OK to destroy school 0.86 property Respondent feels it is OK to abuse teachers 0.76 Respondent feels it is OK to talk back to teachers 0.74 Respondent feels it is OK to disobey school rules 0.84 Respondent feels it is OK to get into physical fights 0.72
Respondent feels it is OK to be late for school 0.67 Respondent feels it's ok to cut a couple of classes 0.91 Respondent feels it's ok to skip school the whole 0.77 day
Correlation between factors (2:1) 0.67 Correlation between factors (3:1, 3:2) 0.45 0.44 Correlation between factors (4:1, 4:2, 4:3) 0.69 0.49 0.47 Correlation between factors (5:1, 5:2, 5:3, 5:4) 0.48 0.77 0.43 0.77
CFI 0.96 RMSEA 0.029
87
Table 2.19 Factor Loadings of Items From Students, Second Follow-Up
Confirmatory Factor Analysis
Non- Delinquent Schooling Compliant Behaviors Unpreparedness Item Behaviors
How many times respondent got into trouble 0.83 How many times respondent was put on in-school 0.78 suspension How many times respondent was suspended from 0.80 school How many times respondent was transferred for 0.88 disciplinary reasons How many times respondent was arrested 0.68
How many times respondent cut/skip classes 0.78 How many times respondent late for school 0.71 How many times respondent missed school 0.52
How often respondent comes to class without 0.82 pencil/paper How often respondent comes to class without books 0.84 How often respondent comes to class without 0.61 homework
Correlation between factors (1:2, 1:3, 2:3) 0.69 0.47 0.46
CFI 0.99 RMSEA 0.015
88
Table 2.20 Factor Loadings of Second-Order Model From Students, Base Year - Second Follow-Up
Confirmatory Factor Analysis
Item Base Year First Follow-Up Second Follow-Up
Delinquent Behaviors 0.82 0.76 0.84 Non-Compliant Behaviors 0.85 0.78 0.83 Schooling Unpreparedness 0.63 0.55 0.56 Deviant Attitudes * 0.83 * Non-Compliant Attitudes * 0.87 *
CFI 0.98 0.94 0.99 RMSEA 0.036 0.035 0.015
89
Table 2.21
Correlations among Factors, Teachers, Base Year through Second Follow-Up Severe Non-Compliant Non-Compliant Behaviors Behaviors BY F1 F2 F1 F2
Non-Compliant Behaviors, Base Year * Non-Compliant Behaviors, First Follow-Up 0.54 * Non-Compliant Behaviors, Second Follow-Up 0.46 0.51 * Severe Non-Compliant Behaviors, First Follow-Up 0.19 0.32 0.16 * Severe Non-Compliant Behaviors, Second Follow- 0.25 0.30 0.60 0.25 * Up
90
Table 2.22 Correlations among Factors, Students, Base Year through Second Follow-Up Non- Non-Compliant Schooling Deviant Delinquent Behaviors Compliant Behaviors Unpreparedness Attitudes Attitudes BY F1 F2 BY F1 F2 BY F1 F2 F1 F1
Delinquent Behaviors, Base Year * Delinquent Behaviors, First Follow-Up 0.70 * Delinquent Behaviors, Second Follow-Up 0.60 0.72 * Non-Compliant Behaviors, Base Year 0.70 0.51 0.39 * Non-Compliant Behaviors, First Follow-Up 0.43 0.64 0.43 0.71 * Non-Compliant Behaviors, Second Follow- 0.35 0.46 0.66 0.54 0.80 * Up Schooling Unpreparedness, Base Year 0.51 0.41 0.35 0.53 0.36 0.32 * Schooling Unpreparedness, First Follow-Up 0.35 0.46 0.35 0.33 0.43 0.38 0.55 * Schooling Unpreparedness, Second Follow- 0.27 0.33 0.47 0.24 0.26 0.46 0.43 0.58 * Up Deviant Attitudes, First Follow-Up 0.43 0.70 0.48 0.36 0.48 0.38 0.33 0.49 0.35 * Non-Compliant Attitudes, First Follow-Up 0.26 0.48 0.35 0.43 0.77 0.54 0.30 0.43 0.31 0.75 *
91
Table 2.23 Correlations among Factors using Same Items, Teachers, Base Year through Second Follow-Up Severe Non-Compliant Non-Compliant Behaviors Behaviors BY F1 F2 F1 F2
Non-Compliant Behaviors, Base Year * Non-Compliant Behaviors, First Follow-Up - * Non-Compliant Behaviors, Second Follow-Up - 0.50 * Severe Non-Compliant Behaviors, First Follow-Up - 0.32 0.16 * Severe Non-Compliant Behaviors, Second Follow- - 0.30 0.61 0.25 * Up
92
Table 2.24 Correlations among Factors using Same Items, Students, Base Year through Second Follow-Up Non- Non-Compliant Schooling Deviant Delinquent Behaviors Compliant Behaviors Unpreparedness Attitudes Attitudes BY F1 F2 BY F1 F2 BY F1 F2 F1 F1
Delinquent Behaviors, Base Year * Delinquent Behaviors, First Follow-Up - * Delinquent Behaviors, Second Follow-Up - 0.73 * Non-Compliant Behaviors, Base Year - - - * Non-Compliant Behaviors, First Follow-Up - 0.68 0.43 0.73 * Non-Compliant Behaviors, Second Follow- - 0.49 0.67 - 0.85 * Up Schooling Unpreparedness, Base Year - - - 0.48 0.33 - * Schooling Unpreparedness, First Follow-Up - 0.48 0.35 0.31 0.44 0.39 0.55 * Schooling Unpreparedness, Second Follow- - 0.34 0.47 - 0.26 0.46 0.44 0.57 * Up Deviant Attitudes, First Follow-Up ------* Non-Compliant Attitudes, First Follow-Up ------*
93
Table 2.25 Factor Loadings of Items From Teachers, First & Second Follow-Ups
Confirmatory Factor Analysis
First Follow-Up Second Follow-Up Severe Severe Non- Non- Non- Non- Compliant Compliant Compliant Compliant Behaviors Behaviors Item Behaviors Behaviors
How often student does NOT complete homework 0.85 0.85 How often student is tardy 0.52 0.57 How often student is inattentive in class 0.88 0.86
True if spoke to parents about student behavior 0.91 0.88 True if spoke to parents about student homework 0.97 0.85 True if spoke to parents about student absenteeism 0.78 0.85
Correlation between factors 0.32 0.61
CFI 1.00 0.99 RMSEA 0.031 0.049
94
Table 2.26 Factor Loadings of Items From Students, Base Year & First Follow-Up
Confirmatory Factor Analysis
Base Year First Follow-Up Delinquent Schooling Delinquent Schooling Behaviors Preparedness Behaviors Preparedness Item
Parents received warning about attendance 0.70 0.72 How many times respondent late for school 0.63 0.66 How many times respondent missed school 0.40 0.59
How often respondent comes to class without pencil/paper 0.69 0.73 How often respondent comes to class without books 0.77 0.83 How often respondent comes to class without homework 0.68 0.60
Correlation between factors 0.48 0.39
CFI 0.99 0.98 RMSEA 0.033 0.035
95 Table 2.27 Factor Loadings of Items From Students, First & Second Follow-Ups
Confirmatory Factor Analysis
First Follow-Up Second Follow-Up Non- Non- Delinquent Schooling Delinquent Schooling Compliant Compliant Behaviors Unpreparedness Behaviors Unpreparedness Item Behaviors Behaviors
How many times respondent got into trouble 0.78 0.84 How many times respondent was put on in-school suspension 0.75 0.78 How many times respondent was suspended from school 0.78 0.77 How many times respondent was transferred for disciplinary reasons 0.86 0.85 How many times respondent was arrested 0.67 0.72
How many times respondent cut/skip classes 0.80 0.76 How many times respondent late for school 0.67 0.69 How many times respondent missed school 0.48 0.51
How often respondent comes to class without pencil/paper 0.73 0.80 How often respondent comes to class without books 0.81 0.84 How often respondent comes to class without homework 0.62 0.59
Correlation between factors (1:2, 1:3, 2:3) 0.69 0.48 0.45 0.67 0.47 0.46
CFI 0.98 0.99 RMSEA 0.020 0.020
96
Table 2.28 Factor Loadings of Items From Students, Base Year - Second Follow-Up
Confirmatory Factor Analysis
Schooling Schooling Schooling Unpreparedness Unpreparedness Unpreparedness (First Follow- (Second Follow- (Base Year) Item Up) Up)
How often respondent comes to class without pencil/paper 0.69 0.76 0.79 How often respondent comes to class without books 0.78 0.82 0.87 How often respondent comes to class without homework 0.67 0.57 0.56
CFI * * * RMSEA * * *
97
Table 2.29 Measurement Invariance for Teachers, First and Second Follow-Ups CFI Δ CFI RMSEA
No measurement invariance 0.991 0.030 Factor loading invariance 0.990 0.001 0.030 Factor loading + intercept invariance 0.982 0.009 0.038
Mean Difference 0.261 (Non-Compliant Behaviors) Mean Difference -0.243 (Severe Non-Compliant Behaviors)
98
Table 2.30 Measurement Invariance for Students, Base Year and First Follow-Up CFI Δ CFI RMSEA
No measurement invariance 0.973 0.022 Factor loading invariance 0.974 0.001 0.021 Factor loading + intercept invariance 0.925 0.048 0.033
Mean Difference 1.402 (Non-Compliant Behaviors) Mean Difference -0.271 (Schooling Unpreparedness)
99
Table 2.31 Measurement Invariance for Students, First and Second Follow-Ups CFI Δ CFI RMSEA
No measurement invariance 0.968 0.020 Factor loading invariance 0.970 0.002 0.019 Factor loading + intercept invariance 0.964 0.004 0.019
Mean Difference -0.102 (Delinquent Behaviors) Mean Difference 0.504 (Non-Compliant Behaviors) Mean Difference 0.025 (Schooling Unpreparedness)
100
Table 2.32 Measurement Invariance for Students, Base Year, First, and Second Follow- Ups CFI Δ CFI RMSEA
No measurement invariance 0.959 0.050 Factor loading invariance 0.963 0.004 0.044 Factor loading + intercept invariance 0.706 0.253 0.102
Mean Difference -0.297; -0.076 (Schooling Unpreparedness)
101 CHAPTER 3
PREDICTORS OF OPPOSITIONAL IDENTITY
INTRODUCTION
Scholars proposing and documenting oppositional identity posit that that minority and
White students differ in the degree to which they exhibit oppositional behaviors and attitudes.
To investigate this claim, I compare behaviors and attitudes among African American, Asian
American, Mexican American, and White youth. I also examine the possibility that oppositional identity varies more by gender than by race/ethnicity (Lundy and Firebaugh 2005). I then examine the relationship the effect of race/ethnicity and gender on oppositional identity controlling for prior achievement—an often-omitted variable—to avoid overstating gender and race/ethnic differences (Harris and Robinson 2007). Finally, I incorporate parental involvement, parental socioeconomic status, track placement, and school-level covariates as predictors of oppositional identity.
I begin with a methodological section operationalizing variables used in the following analyses. I present a short summary of how I derived the oppositional identity measures that are used as dependent variables in this chapter. I then discuss the independent variables. I then conduct measurement invariance using both student and teacher reports to make sure that the measurement properties of my confirmatory factor analysis results are similar across groups.
This ensures that the observed differences are due to alpha change, not beta or gamma change7.
The second section of this chapter will be devoted to answering three interrelated questions using both student and teacher-reported oppositional identity. The first is to highlight that there might be differences by race/ethnicity and/or gender when examining the individual
7 A full discussion of alpha, beta, and gamma change can be found in Chapter 2. 102 factors of the oppositional identity construct. For example, it is possible that both Mexican
American and African American youth both exhibit a higher degree of deviant attitudes than
Asian American and White youth, yet are not different from Asian American and White youth in schooling unpreparedness. Second, I examine whether group differences in oppositional identity persist or change over time. Finally, I include other covariates typically overlooked in existing research to determine (1) the direct effects of these covariates and (2) if the racial/ethnic and gender differences persist after controlling for these potentially confounding variables. These covariates include: parental involvement, parental socioeconomic status, track placement, and school-level variables.
The third section of this chapter shifts from focusing on measures that are available in all waves to those only available during the first follow-up in order to incorporate ideological factors available only during this survey wave (using student reports). Thus, the three behavioral factors available for all three survey waves remain, but there are now two ideological factors included. The goal of this section is to determine how racial/ethnic and gender differences differ when examining similar behaviors and attitudes. Similarly, I examine how these differences change with the inclusion of the covariates mentioned above.
The final section compares results from student and teacher reports. This analysis shows how conclusions about oppositional identity would change depending on who is surveyed. This comparison helps explain the results presented in Chapter 2 and support the conclusion that teacher and student reports of oppositional identity should not be used interchangeably.
ANALYSIS STRATEGY
Dependent Variables
103 The dependent variables for these analyses are the behavioral and ideological sub- dimensions of oppositional identity. The specific dependent variables that are used throughout the analyses depend on which survey wave is available and whether teacher and/or student reports are available. Thus, some factors are included in some waves, but not others.
Nonetheless, the following three factors are available during all survey waves using student reports: delinquent behaviors, non-compliant behaviors, and schooling unpreparedness.
Additional items included in the first follow-up using student reports resulted in two additional sub-dimensions: delinquent attitudes and non-compliant attitudes. A non-compliant behaviors factor derived using teacher reports of items available in all survey waves as well. Finally, items included in the severe non-compliant behaviors factor result from teacher reports that are available in the first and second follow-ups. All of these factors were created by first employing exploratory factor analysis and were cross-validated using confirmatory factor analysis. A more detailed description of the methodology and all these factors is available in Chapter 2.
Independent Variables
Prior achievement is operationalized as eighth grade test scores in reading, math, science, and history. I include test scores on multiple subjects to reduce the measurement error associated in using only one test score (Jencks 1985).
Students’ race/ethnicity and gender rely on student self-reports. The term
“race/ethnicity” is used throughout this discussion since both racial (White, Asian American, and
African American) and ethnic (Mexican American) variables utilized. These racial/ethnic variables are dichotomous with White youth serving as the reference category. Gender is dichotomized with females serving as the reference category.
104 I create my track placement indicator by using both transcript data and teacher-reported track placement. NELS researchers coded the transcript track placement variable using course descriptions for each student throughout their academic career. If transcript information is available, I use that information to assign track into advanced, academic, and vocational categories.
If the transcript information is not available, I assign a track placement as advanced, academic, and vocational based on the teacher-reported data. Teachers were asked: “Which of the following best describes the 'track' this class is considered to be?” The answers ranged from
“remedial” to “AP.” Teachers were asked this question at two time points. They responded to this question for two courses during the first wave and for one course during the second wave of the survey. Students are assigned to the “track” based on the three courses that teachers used for their reports. If two out of the three courses are classified in one track, then students are put in that track. For instance, students are classified in the academic track if two courses are considered “academic” and one considered “advanced.” If there is only information for one class, students are assigned to the track of that class. If there is information for two courses, but there is disagreement between the courses, the track of the math or science course is given preference, since math and science courses exhibit greater differentiation (Kelly 2004).
Information derived from transcripts is the most accurate in that it accounts for changes over time and is not biased by student or teacher perceptions. The teacher reports have more errors than transcript data, but less so than the student reports, since students often do not know what track they are in. This strategy will allow me to gauge a fairly accurate representation of track placement for the entire academic career of most students and at two time points for other students.
105 Parental involvement is measured by whether parents engage in the following: discuss programs at school, discuss school activities, discuss things studied in class, check homework, require chores done, limit time watching TV, and limit going out with friends. I conducted exploratory and confirmatory factor analyses with these items (as I did the oppositional identity measures). The results indicate that there are two factors at work: parental communication and parental limits. The parental communication factor consists of three items measuring whether parents and students: discuss programs at school, discuss school activities and discuss things studied in class. The parental limits factor consists of four items measuring whether parents: check homework, require chores done, limit time watching TV, and limit going out with friends.
I also include parental socioeconomic status, measured with a NELS-specific composite that combines information on parents’ education (mother and father), family income, and occupational prestige (mother and father). These five items are standardized and then combined to produce the SES composite used in this study.
I incorporate the following school-level variables: percentage of students who qualify for free or reduced meals (FRPM), percentage of White students, and whether a school is public or private. Both percent free or reduced meal and percent White are coded as dichotomous variables given the restrictions placed on the public use data set in NELS. Specifically, percent free or reduced meal is coded “1” if the proportion of free or reduced meal students in the school is greater than or equal to 50% and “0” if not. This dichotomous free or reduced meal indicator signifies a “majority free-reduced meal” school. Similarly, the percent White variable is coded as “1” if the proportion of White students is greater than or equal to 90% and “0” if not. This dichotomous variable is coded in this manner in order to capture a “homogenous White school.”
106 Finally, the school type variable is coded as “1” if the school is public and “0” if the school is non-public.
Measurement Invariance
I begin by testing for measurement invariance across race, gender, track placement, and school-level covariates. I do this to ensure that the observed differences across groups in all subsequent analyses are due to true (alpha) change. I test an unconstrained model (measuring non-invariance) versus the constrained model that tests for measurement invariance by holding both factor loadings and thresholds equal across groups. I continue to look to differences in the
Comparative Fit Index (CFI) to evaluate the relative fit of alternative models, where I reject the constrained model if it deviates by more than 0.01 (∆ .01) in CFI. Tables 3A to 3F (in appendix) describe the results of the unconstrained model relative to the constrained model across all groups (e.g. race, gender, school type) using student reports. Tables 3G to 3L replicate the analyses in Tables 3A to 3F using teacher reports. These tables highlight that the critical value of .01 is not exceeded in most cases, but there are two exceptions. Table 3B shows that the coefficient obtained when measuring invariance across school type (private versus public) during the first follow-up exceeds the critical value of .01 by .002. Similarly, Table 3D indicates that the coefficient measuring invariance across gender during the second follow-up exceeds the critical value by .001. This indicates that the results for these two cases are likely to be biased upward8. Nonetheless, my results indicate that any observed differences in every other situation are due to true (alpha) change.
SECTION II: STUDENT-REPORTED OPPOSITIONAL IDENTITY
8 Subsequent analyses reveal that the public school coefficient for the first follow-up and the male coefficient for the second follow-up are in fact slightly higher than those of other waves. 107
Introduction
This section of the paper utilizes all the independent variables cited above and the three behavioral oppositional identity factors derived from student reports. The goal of this section is threefold: (1) examine racial/ethnic and gender differences, (2) determine if these group differences are stable over time, and (3) determine the effects of independent variables and whether these independent variables alter (1) racial/ethnic and gender differences and (2) group differences over time.
Model I: Race/Ethnicity and Gender
Table 3.1 shows racial and gender differences from the base year through the second follow-up in oppositional identity measures using student reports. The reference category in this table for race is White youth. The results are standardized and therefore can be interpreted as z- scores.
I begin by examining the results for the delinquent behaviors factor. This table indicates that there are significant differences by race/ethnicity and gender in this regard. For example,
Mexican American students were .32 standard deviations higher than White students on this factor during the base year. African American students were higher than Whites as well, but
Asian American students exhibited a lower degree of delinquent behaviors. Similarly, both
African American and Mexican American youth exhibit a significantly higher degree of delinquent behaviors during the first follow-up, while Asian American students are lower than
White students in this regard. Mexican American youth are indistinguishable from Whites during the second follow-up, while African American and Asian American youth exhibit
108 divergent paths (higher and lower than White youth, respectively). Finally, males exhibit a significantly higher degree of delinquent behaviors than females during all three survey waves.
These results highlight that the magnitude of racial/ethnic differences change when examining the delinquent behaviors factor over time. For example, Asian American youth are
.16 standard deviations lower than White students on the delinquent behaviors factor in the base year, but this difference doubles to .32 in the second follow-up. Similarly, the magnitude of differences is so large that the statistical significance of the coefficient changes when comparing different survey waves. For example, Mexican American youth exhibit a significantly higher degree of delinquent behaviors compared to White youth in the base year and first follow-up, but this difference is negligible in the second follow-up. The gender coefficient exhibits minor change in that boys exhibit consistently higher delinquent behaviors than girls for all three survey waves (.73, .65, and .77, respectively).
This table further shows that both Mexican American and African American youth are higher than White youth for the non-compliant behaviors factor during the base year, while
Asian Americans are not statistically different from Whites. Furthermore, Mexican American youth continue to exhibit a higher degree of non-compliant behaviors during the first follow-up, but African Americans are now indistinguishable from White youth. Asian Americans also exhibit change as well in that they are .15 standard deviations lower than Whites during the first follow-up, but similar during the base year. The second follow-up shows even less differences across race/ethnicity because only Mexican Americans are significantly different than White youth (.33 standard deviations higher). Finally, boys are .12 standard deviations higher than girls during the second follow-up, but statistically indistinguishable during the other two survey waves.
109 These results suggest that the Mexican American coefficient is largely stable when examining the non-compliant behaviors factor. On the other hand, both African Americans and
Asian Americans exhibit more change over time. For example, the difference in magnitude dropped by a factor of four (.23 to .05) between the base year and first follow-up when examining the non-compliant behaviors factor among African American youth. Similarly, there is a large drop for Asian American youth from -.15 to -.02 between the first and second follow- ups. Thus, there appears to be a steady decline in racial group differences over time for these two groups when examining this particular factor. Finally, there is a negligible effect of gender during the base year and first follow-up, but the effect is statistically significant during the second follow-up.
A close examination of the schooling unpreparedness factor indicates that there are few racial/ethnic differences. Specifically, Mexican American youth are .16 standard deviations higher than White youth, while Asian American youth are roughly .15 standard deviations lower than White youth in this factor during the base year. Furthermore, Asian Americans are .11 standard deviations lower than Whites, while both African American and Mexican American youth are indistinguishable from White youth during the first follow-up. There are no statistically significant racial/ethnic differences during the second follow-up of this factor.
Finally, boys exhibit a higher degree of schooling unpreparedness than girls by .34, .39, and .46 standard deviations during each survey wave.
A general theme that emerges across most factors is that there are strong gender effects across survey waves, while racial/ethnic differences tend to fluctuate (but primarily decline). For example, boys are .73 standard deviations higher than girls in the delinquent behaviors factor during the base year. This effect is consistent across survey waves with a difference of .65 and
110 .77 standard deviations during the first and second follow-ups, respectively. The effect of gender is much smaller when examining the schooling unpreparedness factor during all survey waves, but consistent nonetheless. Thus, the gender coefficients appear to be stable over time, unlike the decline in group differences exhibited by racial/ethnic groups. Nonetheless, while the effect of gender is much more pronounced over time and across factors, this is not the case for the non- compliant behaviors factor. This is the only factor where racial/ethnic differences are stronger relative to gender differences. Finally, there are also more significant racial/ethnic differences during the base year than any other survey wave. The first follow-up has some significant differences as well, but the second follow-up has almost no racial/ethnic differences.
Model II: Incorporating Prior Achievement
Extant research suggests that prior achievement is a significant predictor of oppositional identity; students who exhibit higher prior achievement exhibit lower rates of oppositional identity (Harris and Robinson 2007). Table 3.2 combines the race/ethnicity and gender results from Model I with the results of including prior achievement in the model (Model II). Model I is included in the top portion of the table, while Model II is presented in the bottom portion of
Table 3.2. The table shows a noticeable reduction in the coefficients of race/ethnicity, but not gender, across all factors of oppositional identity. For example, the schooling unpreparedness coefficient for Mexican American youth during the base year drops from .155 to approximately zero (.025) in the final model. Similarly, the Mexican American and African American coefficients for the delinquent behaviors factor during the base year drop significantly after the inclusion of prior achievement. Thus, prior achievement appears to be an omitted variable that overstates and/or alters the effects of race/ethnicity on oppositional identity, yet has a negligible effect on gender.
111 A second observation is that prior achievement has less of an effect for some groups than others. For example, Table 3.2 highlights that the Asian American coefficients are largely unchanged across factors and over time when comparing both models. This stands in contrast to both African American and Mexican American youth in that there exists more fluctuation in the coefficients. A related pattern is that some of the fluctuations are so large that some coefficients become non-significant after incorporating prior achievement. For example, there was a shift to non-significance for Mexican American youth when examining the schooling unpreparedness factor during the base year. African American youth stand out as a special case because they exhibit both fluctuations in the coefficients, but also see a higher proportion of significance changes. For example, African American youth exhibit significant differences in the non- compliant behaviors factor during the base year in the initial model, but not Model II. This situation is reversed when examining the schooling unpreparedness factor during this same survey wave in that a significant coefficient surfaced only during Model II.
An interesting observation is that prior achievement not only alters significance, but also the direction of the observed relationships. All of the significant coefficients for African
American and Mexican American youth in the initial model are positive, but this is no longer the case after the inclusion of prior achievement. For example, two out of the three significant
African American coefficients in Model II are negative. Similarly, two out of six significant
Mexican American coefficients are also negative. This suggests that some of the initial racial/ethnic differences in oppositional identity are in fact due to differences in prior achievement among racial/ethnic groups.
Final Model
112 The final model includes parental involvement, parental socioeconomic status, track placement, and school-level covariates to investigate not only its effects on oppositional identity but also its impact on the previous racial/ethnic and gender differences. Table 3.3 shows the results after incorporating all of these covariates and prior achievement. Most of the prior achievement measures are statistically significant during the base year, but the effect dissipates when examining the first and second follow-up years. Thus, there appears to be a temporal component, which makes sense given that the prior achievement measures are derived from the base year. Similarly, prior achievement appears to have the most consistent effect on the delinquent behaviors and non-compliant behaviors factors. For example, all of the prior achievement measures are statistically significant on the delinquent behaviors factor during the base year, three out of four during the first follow-up, and two out of four during the second follow-up. This stands in contrast with the schooling unpreparedness factor, which exhibits mostly non-significant prior achievement effects during the first and second follow-ups.
This table also highlights that parental involvement is significantly associated with all dimensions of oppositional identity and across most survey waves. Both parental involvement factors lead to a significant reduction in oppositional identity, but the strength of each parental involvement factor depends on the dimension of oppositional identity under examination. For example, a one standard deviation change in parental communication is associated with a .17 standard deviation decrease in schooling unpreparedness during the base year, while a one standard deviation change in parental limits has roughly one-quarter of that effect (-.04).
Similarly, there were some survey waves where only one parental involvement factor had a significant effect. For example, the parental limits factor was not a significant predictor of the
113 delinquent behaviors factor during the second follow-up and the schooling unpreparedness factors during the first and second follow-ups.
On the other hand, parental socioeconomic status is associated with a significant decrease in oppositional identity in only a few occasions and primarily during the base year. For example, parental socioeconomic status has a non-significant effect for the schooling unpreparedness factor during all survey waves. In the few cases where parental socioeconomic status exhibits a significant effect, it is in the expected direction in that it reduces oppositional identity. The negligible effect of parental socioeconomic status is likely due to the inclusion of both parental involvement factors given that high socioeconomic parents are also likely to be highly involved
(Sui-Chu and Willms 1996).
Table 3.3 also shows that students in the advanced track were roughly .12 to .45 standard deviations lower than students in the academic track across dimensions of oppositional identity.
This stands in contrast to students in the vocational track, who are not statistically different than students in the academic track in the majority of cases involving behavioral factors, with two exceptions. These exceptions involve students in the vocational track exhibiting a statistically significant lower degree of delinquent behaviors and non-compliant behaviors than students in the academic track during the base year.
As with most predictors, there is some variation in the effects of school-level covariates across oppositional identity dimensions and survey waves. The most temporally consistent and strongest school-level variable is whether students attended a public school. Students who attended a public school are .08 to .27 standard deviations higher than students who attended a non-public school on most factors from the base year to the second follow-up. The variable signifying White student body homogeneity is inconsistent in statistical significance, but it
114 exhibits significant effects for the non-compliant behaviors factor during all survey waves.
Students attending a homogenously White school are .11 to .24 standard deviations lower on this factor compared to students attending a non-homogenously White school. Finally, attending a school with a majority free and reduced meal population is associated with a decrease in oppositional identity only during the base year.
Summary
This chapter began by examining whether there are significant differences across race/ethnicity and gender when examining the various dimensions of student-reported oppositional identity. I find that there are significant differences across race/ethnicity when other covariates are omitted, but that these significant differences are inconsistent once covariates such as prior achievement, parental involvement, parental socioeconomic status, track placement, and school-level variables are incorporated into the model. On the other hand, the effects of gender are largely stable both before and after the inclusion of other covariates.
Examination of Table 3.3 highlights that prior achievement, parental involvement, track placement, and attending a public school are very strong predictors of oppositional identity.
Track placement is by far the most consistent of the school environment variables, but the impact of attending a public school is important to acknowledge as well given its significant effect in some situations. Finally, parental socioeconomic is not statistically significant in most cases.
Table 3.4 compares the coefficients from Model I (top portion) and my final model
(bottom portion). Table 3.4 suggests there are noticeable changes in the racial/ethnic differences of the initial model versus the final model. First and foremost, the racial/ethnic differences are smaller in the final model than in the initial model. For example, the initial model highlights that
Mexican American youth exhibit much higher rates of non-compliant behaviors (.45 standard
115 deviations higher) than White youth during the base year. Nonetheless, the final model highlights that this relationship is much lower (.11). Similarly, the African American coefficient is equal to .35 for the delinquent behaviors factor during the base year in the initial model, but it drops to .07 in the final model. A close inspection of the previous models highlights that part of this shift occurred after the inclusion of prior achievement, suggesting that the initial results were due in part to both Mexican American and African American youth exhibiting lower rates of achievement than Whites prior to beginning high school. The shift for African Americans was also due in part to the inclusion of school-level variables, which highlights the importance of school environment in contributing to oppositional identity.
These results suggest that students subjected to unfavorable schooling conditions are likely to develop oppositional identity regardless of race/ethnicity. Thus, researchers who ignore prior achievement and/or schooling variables will erroneously conclude that African American and Mexican American youth exhibit more oppositional identity behaviors than Whites.
Similarly, these results highlight that oppositional identity is governed more by gender than race, since the gender coefficients are the strongest predictors of oppositional identity across survey waves and most oppositional identity dimensions. Ultimately, a holistic examination of Table
3.4 highlights that (for the most part) all non-Whites exhibit comparable and/or lower oppositional behaviors than their White counterparts in the final model, which contradicts common arguments and previous research. Similarly, while there are differences in race/ethnicity, these are largely overshadowed by the strength of gender effects.
A related observation is that each racial group has a unique story when comparing the initial and final models. Asian Americans exhibit a lower degree of oppositional identity relative to all three groups in both the initial and final model. Nonetheless, there are some instances
116 where omitting covariates leads to different conclusions when comparing the Asian American coefficient in both models. For example, the Asian-American coefficient for the non-compliant behaviors factor during the base year becomes statistically significant in the final model.
Mexican-American youth exhibit higher oppositional identity than Whites for most coefficients in the initial model. In final model, however, Mexican American youth are largely indistinguishable from White youth in most cases and even less oppositional in others. African
American youth provide an even greater departure from the initial model at the final model.
Almost all of the African American coefficients are positive in the initial model, yet exhibit lower oppositional identity by the final model. In fact, three of the only four significant coefficients for African Americans are negative in the final model.
Furthermore, the observed differences are a function of what factor one chooses to focus on. Specifically, Table 3.4 shows that some factors highlight racial/ethnic differences, while others do not. Similarly, while gender plays a highly significant role overall, it does little when examining the non-compliant behaviors factor. Ultimately, this underscores the importance of considering the questions of what is being measured in oppositional identity research.
There is also a temporal component that is observable in Table 3.4. The final model highlights that most of the significant racial differences that remain are restricted to the base year and first follow-up. Thus, there are minimal racial/ethnic differences by the second follow-up.
Table 3.5 seeks to determine the source of the temporal component of oppositional identity using student reports. Specifically, does the lack of racial differences during the second follow-up highlight that oppositional identity is something youth outgrow, or is it simply a “cohort effect”9?
9 “Cohort effect” refers to the fact that the lack of observed differences during the second follow-up is due to students who exhibited a higher degree of oppositional identity no longer being enrolled (dropped out) by then. 117 Table 3.5 compares the final model for all students (in top portion of table) with a new model that only incorporates students who were continuously enrolled during all three survey waves. This table shows that the exclusion of students who dropped out alters the patterns exhibited by the cross-sectional results. In other words, the observed decline in group differences over time in the cross-sectional results is likely driven by a cohort effect given that this same decline is not observable when using the same students over time. Nonetheless, regardless of the source of the decline observed in my original model, the important aspect is that these changes highlight the importance of when one examines oppositional identity.
SECTION II: TEACHER-REPORTED OPPOSITIONAL IDENTITY
Introduction
This section of the paper utilizes all the independent variables cited above and the two behavioral oppositional identity factors derived from teacher reports. The goal of this section is threefold: (1) examine racial/ethnic and gender differences, (2) determine if these group differences are stable over time, and (3) determine the effects of independent variables and whether these independent variables alter (1) racial/ethnic and gender differences and (2) group differences over time.
Model I: Race/Ethnicity and Gender
Table 3.6 highlights differences across racial/ethnic groups and gender during the base year through the second follow-up, this time using teacher reports. This table indicates that there are significant differences by race/ethnicity and gender across both the non-compliant behaviors and severe non-compliant behaviors factors. For example, Asian American youth are .40, .42, and .16 standard deviations lower than White students in the non-compliant behaviors factor
118 during the base year, first follow-up, and second follow-up, respectively. Mexican American youth exhibit consistently higher oppositional identity than Whites when examining non- compliant behaviors, but are indistinguishable from Whites when examining severe non- compliant behaviors. This stands in contrast with African Americans, who exhibit consistently higher oppositional identity when examining both factors and all survey waves. The male coefficients are also positive across factors and all survey waves, indicating that boys are more oppositional than girls.
Model II: Incorporating Prior Achievement
Table 3.7 combines Model I with the one presented here (Model II). As the table suggests, there is a noticeable reduction in the coefficients of race/ethnicity, but not gender, across all factors of oppositional identity. The exception to this trend is Asian American youth in that the coefficients remain largely unchanged when comparing the two models in this table.
On the other hand, both African American and Mexican American youth see a reduction in racial/ethnic differences relative to Whites after controlling for prior achievement. Also, unlike the results gleaned from student reports of oppositional identity, there is no temporal component using teacher reports. The inclusion of prior achievement appears to reduce racial differences over time uniformly. Similarly, the racial/ethnic and gender differences are also similar in strength with each other across both factors, unlike the inconsistency seen using student reports.
Final Model
The final model incorporates parental involvement, parental socioeconomic status, track placement, and school-level variables. Table 3.8 shows that prior achievement is a significant predictor of oppositional identity. There is a negative relationship between prior achievement
119 and oppositional identity. For example, 8th grade reading, math, science, and history scores are all statistically significant for the non-compliant behaviors factor during the base year. Three out of the four prior achievement measures are also statistically significant for this factor during the first follow-up as well. Nonetheless, an interesting trend is that a larger proportion of measures of prior achievement are not statistically significant for the severe non-compliant behaviors factor. Thus, it appears that prior achievement has less predictive power for this factor than the non-compliant behaviors factor, particularly during the second follow-up10.
This table also highlights that parental socioeconomic status leads to a decrease in oppositional identity. In fact, parental socioeconomic status is statistically significant for both dimensions across all survey waves. Nonetheless, a distinct story surfaces when examining the effect of parental involvement on oppositional identity: in most cases, there is no parental involvement effect. For example, the parental limits factor has non-significant effects on both dimensions of oppositional identity across all survey waves. Similarly, the parental communication factor is only a significant predictor of the non-compliant behaviors factor during the base year and first follow-up.
This table also indicates that students in the advanced track are .27 to .40 standard deviations lower on the non-compliant behaviors factor than students enrolled in the academic track. The effect of being in the advanced track is smaller when examining the severe non- compliant behaviors factor, with coefficients of roughly half the magnitude relative to the non- compliant behaviors factor. The results of being in the vocational track versus the academic track are statistically indistinguishable. Finally, this table highlights that the school-level
10One can argue that this is a temporal effect. Nonetheless, the first follow-up had a majority of significant effects for the non-compliant behaviors factor during this time, but not for the severe non-compliant behaviors factor. Thus, while a temporal effect is likely influencing the outcomes of the second follow-up, it is likely that the predictive power of prior achievement has an effect as well (otherwise we would see the same pattern for both factors). 120 predictors are not statistically significant in most instances. One exception is that students attending public schools versus non-public schools are .14 standard deviations higher on the severe non-compliant behaviors factor during the second follow-up. Similarly, there is also one instance in which attending a homogenous White school and a majority-free and reduced lunch school is statistically significant.
Summary
Table 3.9 highlights the coefficients that correspond to the initial and final model. As mentioned previously, the initial model only incorporated race/ethnicity and gender, while the final model included all the covariates cited above. This table emphasizes the stability of gender even after controlling for other covariates remains. Nonetheless, there are differences when examining the race/ethnicity coefficients. The first difference is that the race/ethnicity coefficients are much lower in the final model. Second, the change in the coefficients is so large in some cases that there are no longer significant differences in the final model. For example, the initial model highlights that Mexican American, Asian American, and African American youth exhibited significantly different levels (relative to Whites) in the non-compliant behaviors factor during the first follow-up. Nonetheless, while the Asian-American coefficient remains statistically significant in the final model, this is no longer the case for African Americans and
Mexican Americans. Thus, some of the initial differences in race/ethnicity are in fact due to other factors like prior achievement and/or schooling environment.
Another observation is that although some group differences disappear in the final model, the final model still highlights that significant differences remain across race. For example, both
African American and Asian American youth are significantly different from White youth in the non-compliant behaviors factor during the base year. Specifically, African American youth
121 exhibit a higher degree of non-compliant behaviors, while Asian American youth exhibit a lower degree of non-compliant behaviors than Whites. Similarly, African American youth are significantly higher than Whites on the non-compliant behaviors factor during the second follow- up.
A related point is that the initial model’s story is one of race/ethnicity. After all, the initial model not only highlights that racial/ethnic differences exist, but also that these differences are larger than those of gender. On the other hand, the story of the final model is one of gender. Boys exhibit more oppositional identity behaviors than girls across all dimensions and survey waves. Similarly, gender is a much stronger predictor than race/ethnicity in the final model, with one exception. This exception is the coefficient of Asian Americans, which is comparable to that of gender in some instances.
SECTION III: OPPOSITIONAL ATTITUDES VERSUS OPPOSITIONAL BEHAVIORS
Introduction
The goal of this section is to determine how racial/ethnic and gender differences differ when examining similar behaviors and attitudes during the first follow-up. Similarly, I examine how covariates impact these five oppositional identity factors and how the initial group differences change with the inclusion of these independent variables.
Model I: Race/Ethnicity and Gender
Table 3.10 shows differences across race/ethnicity and gender during the first follow-up.
This table shows that there are racial/ethnic and gender differences across all groups and five factors. Mexican American youth exhibit a higher rate of delinquent behaviors, non-compliant behaviors, and schooling unpreparedness relative to White youth. Nonetheless, they are no
122 different than White youth in non-compliant attitudes and lower than White youth in delinquent attitudes. African American youth have a similar trajectory to that of Mexican American youth, with two exceptions. The first is that African Americans are indistinguishable from Whites for the non-compliant behaviors factor. The second is that African-Americans exhibit a lower degree of non-compliant attitudes relative to Whites. Asian American youth exhibit a lower degree of oppositional identity than Whites on all factors except the schooling unpreparedness factor. Boys exhibit a higher amount of oppositional identity relative to girls on all factors as well, with the exception of the non-compliant behaviors factor.
The gender effects are largely dominant, but there are some exceptions. The race/ethnicity differences are greater than those of gender for the non-compliant behaviors factor. Similarly, the African American coefficient for the non-compliant attitudes factor is much larger, while the Asian American coefficient is comparable to the gender coefficient. The strongest gender effects exist for both of the most severe factors: delinquent behaviors and attitudes. Similarly, the two strongest racial/ethnic differences are for the non-compliant behaviors and attitudes factors. Specifically, Mexican American youth are .45 standard deviations higher than Whites for the non-compliant behaviors factor, while African Americans are .42 standard deviations lower than Whites for the non-compliant attitudes factor.
Model II: Incorporating Prior Achievement
Table 3.11 shows the racial/ethnic and gender results after the inclusion of prior achievement. The dimensions of oppositional identity that is most affected by prior achievement are the three behavioral factors (the ideological factors are largely unchanged). For example, the schooling unpreparedness factor sees a significant reduction in the effect of race/ethnicity when comparing both models in Table 3.11. In fact, there are no longer any significant racial/ethnic
123 differences for this factor in Model II. Similarly, there are significant changes for the delinquent behaviors and non-compliant behaviors factors. Mexican American youth exhibit more delinquent behaviors than Whites, but this relationship is reversed in Model II. African-
American youth exhibit a similar trend between the initial model and Model II in that a non- significant difference becomes negative and significant. African American youth are also significantly different than White youth in delinquent behaviors in the initial model, but not
Model II. This is in contrast with both the deviant attitudes and non-compliant attitudes factors, which see minor change in racial/ethnic differences after incorporating prior achievement. Thus, as a whole, the primary difference between both models is that non-Whites are overwhelmingly lower than Whites on almost each factor in Model II, but not the initial model. Finally, the gender coefficients remain stable after the inclusion of prior achievement as well.
Final Model
Table 3.12 shows the final model that incorporates parental involvement, parental socioeconomic status, track placement, and school-level covariates. This table shows that prior achievement has positive significant effects for the delinquent behaviors and non-compliant behaviors factors, but not the schooling unpreparedness factor. This is likely due to the inclusion of track placement given the fact that the schooling unpreparedness factor captures typical schooling behaviors. Similarly, prior achievement has a negligible effect on the two ideological factors in most cases. Thus, it appears prior achievement is most significant for behavioral factors and not the ideological ones.
Both parental involvement factors have a negative effect on oppositional identity.
Specifically, a one standard deviation change in parental communication is associated with a
124 -.05 to -.10 standard deviation change in oppositional identity. The parental limits factor has a slightly higher effect that ranges from -.09 to -.15 standard deviations. Interestingly, both parental involvement factors have a slightly stronger effect on ideology than behaviors. On the other hand, parental socioeconomic status has no effect on oppositional identity with the exception of the delinquent behaviors factor. It is likely that this non-significance of parental socioeconomic status is due to the inclusion of parental involvement in the model.
Being in the advanced track rather than the academic track is associated with a decrease in oppositional identity across all factors. For example, students in the advanced track are .45 standard deviations lower for the non-compliant behaviors factor. Being in the vocational track rather than the academic track is not statistically significant for any of the behavioral factors as well. The one consistent area where both levels of track placement are a significant predictor involves ideology. For example, students in the advanced track were roughly .19-.20 standard deviations lower than students in the academic track on both the deviant attitudes and non- compliant attitudes factors. On the other hand, students in the vocational track exhibited higher rates of oppositional identity on both of these dimensions relative to students in the academic track.
Attending a public school is the most consistent predictor of oppositional identity with three out of five coefficients being statistically significant. The effect of attending a public school leads to an increase in both non-compliant behaviors and attitudes, but a .12 standard deviation drop in delinquent behaviors. Students who attend a majority free-reduced meal school are not statistically different from those who attend a non-majority free-reduced meal school for any of the five factors. Finally, attending a homogenous White school is associated with a .24
125 standard deviation decrease in the non-compliant behaviors factor and a .09 standard deviation decrease in the schooling unpreparedness factor.
Summary
Table 3.13 shows the initial and final model of both oppositional ideology and behaviors during the first follow-up. As depicted in Table 3.13, the initial model shows that one area where African American, Mexican American, and Asian American youth exhibit consistently lower levels of oppositional identity is in ideology. Specifically, the coefficients for all non-
White groups are statistically significant and negative, with the exception of the Mexican
American coefficient for the non-compliant attitudes factor during the second follow-up. This result is consistent with that of previous research that has found that African Americans exhibit more positive attitudes than Whites (Ainsworth-Darnell and Downey 1998) and other immigrant literatures (Portes and Zhou 1993).
However, an inspection of the final model offers a different story than the one portrayed in the initial model. One difference is that there are no statistically significant differences for the schooling unpreparedness factor in the final model. Thus, it appears that all the schooling covariates and prior achievement are omitted variables that overstate the effects of this particular factor. A second difference involves the fact that the initial model has only Asian American youth as less oppositional than White youth, but this is no longer the case by the final model.
In fact, all non-Whites exhibit lower oppositional identity on all four factors (minus schooling unpreparedness because there are no statistically significant differences). The exception to this is Mexican American youth for the non-compliant behaviors factor where they are .20 standard deviations higher than White youth. Thus, overall, my results suggest that oppositional identity is more of a gendered phenomenon given the strength of gender in my
126 analyses. Furthermore, I find that while racial/ethnic differences exist, they are overwhelmingly against oppositional identity theory with Asian American, African American, and Mexican
American youth exhibiting lower oppositional behaviors and attitudes than White youths after controlling for various covariates.
SECTION IV: STUDENT VERSUS TEACHER-REPORTED OPPOSITIONAL IDENTITY
The use of both student and teacher reports has allowed me to get a better sense of how oppositional identity exists. Table 3.14 combines both teacher-reported factors with the most comparable student-reported factor of non-compliant behaviors. This table derives estimates from both the student and teacher final models. One difference between using teachers or students is that the magnitude of gender differences is consistently much higher when using teacher reports. For example, boys were .10 standard deviations higher than girls during the base year when examining the non-compliant behaviors factor using student reports. This difference drastically increases to .42 when using teacher reports. Similarly, examination of the non- compliant behaviors factors during the base year highlights that the gender coefficient is not statistically significant using student reports, but is highly significant (.37) using teacher reports.
There are also differences when examining the different racial groups. The Asian
American coefficients are about twice as large when using teacher reports rather than student reports. For example, the Asian American coefficient for the non-compliant behaviors factor during the base year is equal to -.12 using student reports, but is equal to -.34 using teacher reports. The Mexican American coefficients are more mixed in that there is stability in some cases, but changes in others. For example, there are no significant differences in the non- compliant behaviors factor during the first follow-up using teacher reports, but this difference is
127 significant using student reports. African American youth exhibit a similar trend to that of
Mexican American youth.
Temporally, the results using student-reported oppositional identity indicate a decline in oppositional identity across survey waves, but teacher-reported oppositional identity measures still indicate racial/ethnic differences during the second follow-up. This finding reinforces the importance of when one examines oppositional identity. Similarly, it highlights that the impact of when oppositional identity is reported depends on who is doing the reporting.
This table also shows that there are both similarities and differences in the effect of other covariates when using student or teacher-reported oppositional identity. For example, the effect of parental involvement is much stronger using student reports than teacher reports. Similarly, parental socioeconomic status has a non-significant effect when examining student-reported results, but the opposite is true when examining teacher-reported results. The results also indicate that track placement is a significant predictor of oppositional identity when we use both student and teacher reports of oppositional identity. Specifically, being in the advanced versus the academic track lowers oppositional identity using both student and teacher reports. The advanced track coefficient is largely indistinguishable across student and teacher reports.
Finally, it is important to point out that most of the significant school-level effects are restricted to the results involving student-reported oppositional identity.
CONCLUSION
The aforementioned analyses highlight the importance of who does the reporting, what specific facet of oppositional identity is being examined, and when reports are collected
(heretofore called “the who, the what, and the when). The importance of who is doing the
128 reporting surfaces when comparing the independent results of using teacher versus student reports. This point is further reinforced by the divergent conclusions drawn when examining similar factors of teacher versus student-reported oppositional identity. In other words, although
Table 3.14 highlights a dimension of oppositional identity that both teacher and student reports are able to capture, it is important to reemphasize that this dimension only captures a subset of the oppositional identity construct. Thus, given that there are inconsistencies that lead to different conclusions using this limited focus, it stands to reason that the who, what and when play an even greater role in the conclusions drawn when examining the full spectrum of oppositional identity dimensions. In fact, a close examination of Tables 3.3 and 3.8 reinforces this reality.
The importance of what is being examined and when it is analyzed surfaces throughout my independent analyses of student and teacher-reported oppositional identity. I find that whether racial/ethnic and/or gender differences exist depends very much on what factor I am examining and the survey wave11 under examination. Thus, a comparison of Tables 3.3 and 3.8 is necessary in achieving a complete and comprehensive understanding of the complexities of oppositional identity. These tables highlight that whether a researcher finds racial/ethnic and gender differences “exist” will depend on the source of information (teacher or student), but also the factor being examined and the time frame being analyzed.
The effect(s) of covariates are governed by the same phenomenon in that the effect of any given variable depends on the specific situation as well. Some of the covariates exhibited consistent effects for some dimensions of oppositional identity, but there were always exceptions. The end result of this is that the initial model looked much different than the final model in most cases. Perhaps the clearest example of this reality involves the comparison of the
11 As mentioned previously, this is due to a “cohort effect” and not to a decline in oppositional identity over time. 129 initial and final models when examining behaviors and attitudes during the first follow-up.
Table 3.12 highlights that the final model shows all non-White groups exhibit a lower degree of oppositional identity on most factors and survey waves relative to White youth, but the opposite is true in the initial model.
Ultimately, it appears that these specific contingencies are less likely to impact researchers focusing on Asian American youth given the consistency exhibited by this group.
This is likely to hold for those only focusing on a gendered interpretation of oppositional identity. Nonetheless, given the focus on this literature on minority groups (particularly African
Americans), the results of this chapter highlight the importance of taking into account who, what, and when in oppositional identity research.
130
Table 3.1 Model I: Predictors of Oppositional Behaviors using Student Reports, Base Year through Second Follow-Up
Delinquent Behaviors Non-Compliant Behaviors Schooling Unpreparedness BY F1 F2 BY F1 F2 BY F1 F2
Mexican American 0.315 0.102 -0.012 0.447 0.449 0.326 0.155 0.088 0.011 African American 0.352 0.262 0.233 0.228 0.045 0.023 0.047 0.087 -0.004 Asian American -0.155 -0.295 -0.317 -0.101 -0.146 -0.019 -0.150 -0.111 -0.047
Male 0.728 0.646 0.773 0.089 -0.006 0.123 0.338 0.389 0.457
Statistically significant at .05
131
Table 3.2 Comparison of Model I and Model II of Oppositional Behaviors using Student Reports, Base Year through Second Follow-Up
Delinquent Behaviors Non-Compliant Behaviors Schooling Unpreparedness BY F1 F2 BY F1 F2 BY F1 F2 Model I
Mexican American 0.315 0.102 -0.012 0.447 0.449 0.326 0.155 0.088 0.011 African American 0.352 0.262 0.233 0.228 0.045 0.023 0.047 0.087 -0.004 Asian American -0.155 -0.295 -0.317 -0.101 -0.146 -0.019 -0.150 -0.111 -0.047
Male 0.728 0.646 0.773 0.089 -0.006 0.123 0.338 0.389 0.457
Delinquent Behaviors Non-Compliant Behaviors Schooling Unpreparedness BY F1 F2 BY F1 F2 BY F1 F2 Model II
Mexican American 0.077 -0.122 -0.180 0.210 0.310 0.265 0.025 0.049 -0.012 African American 0.077 -0.001 0.036 -0.035 -0.119 -0.058 -0.104 0.043 -0.036 Asian American -0.142 -0.296 -0.316 -0.092 -0.125 -0.015 -0.136 -0.101 -0.038
Male 0.725 0.648 0.780 0.102 0.030 0.150 0.346 0.402 0.462
Model I: No Controls Model II: Controlling for 8th Grade Achievement (Reading, Math, Science, and History)
Statistically significant at .05
132
Table 3.3 Final Model: Oppositional Behaviors using Student Reports, Base Year through Second Follow-Up
Delinquent Behaviors Non-Compliant Behaviors Schooling Unpreparedness BY F1 F2 BY F1 F2 BY F1 F2
Mexican American 0.063 -0.170 -0.142 0.111 0.201 0.205 0.025 0.021 -0.022 African American 0.071 -0.037 0.061 -0.126 -0.207 -0.119 -0.107 0.021 -0.053 Asian American -0.126 -0.311 -0.313 -0.122 -0.149 -0.055 -0.129 -0.110 -0.048
Male 0.723 0.639 0.760 0.107 0.025 0.143 0.343 0.399 0.459
8th Grade Reading Score -0.060 -0.034 -0.001 -0.021 0.078 0.062 -0.011 0.033 0.032 8th Grade Math Score -0.140 -0.131 -0.119 -0.122 -0.085 -0.024 -0.080 -0.029 -0.039 8th Grade Science Score -0.083 -0.076 -0.043 -0.091 -0.113 -0.062 -0.028 -0.004 0.002 8th Grade History Score -0.060 -0.063 -0.101 -0.088 -0.028 -0.049 -0.061 -0.032 -0.022
Parental Communication -0.085 -0.043 -0.092 -0.143 -0.054 -0.035 -0.173 -0.100 -0.165 Parental Limits -0.108 -0.089 -0.036 -0.138 -0.108 -0.086 -0.041 -0.043 0.035
Parental Socioeconomic -0.054 -0.047 0.027 -0.040 0.009 0.026 -0.004 -0.004 0.016 Status
Advanced Track -0.282 -0.380 -0.294 -0.317 -0.445 -0.363 -0.158 -0.192 -0.123 Vocational Track -0.082 0.037 0.071 -0.141 -0.013 -0.067 0.037 0.045 0.004
Public School 0.080 -0.116 -0.089 0.244 0.266 0.134 0.184 0.043 0.001 Majority-FRPM School -0.105 -0.066 -0.102 0.019 -0.034 -0.033 -0.080 -0.069 -0.009 Homogenous White School -0.023 -0.074 -0.029 -0.111 -0.240 -0.183 -0.034 -0.085 -0.058
Statistically significant at .05
133
Table 3.4 Comparison of Model I and Final Model of Oppositional Behaviors using Student Reports, Base Year through Second Follow-Up Delinquent Behaviors Non-Compliant Behaviors Schooling Unpreparedness BY F1 F2 BY F1 F2 BY F1 F2 Model I
Mexican American 0.315 0.102 -0.012 0.447 0.449 0.326 0.155 0.088 0.011 African American 0.352 0.262 0.233 0.228 0.045 0.023 0.047 0.087 -0.004 Asian American -0.155 -0.295 -0.317 -0.101 -0.146 -0.019 -0.150 -0.111 -0.047
Male 0.728 0.646 0.773 0.089 -0.006 0.123 0.338 0.389 0.457
Delinquent Behaviors Non-Compliant Behaviors Schooling Unpreparedness BY F1 F2 BY F1 F2 BY F1 F2 Final Model
Mexican American 0.063 -0.170 -0.142 0.111 0.201 0.205 0.025 0.021 -0.022 African American 0.071 -0.037 0.061 -0.126 -0.207 -0.119 -0.107 0.021 -0.053 Asian American -0.126 -0.311 -0.313 -0.122 -0.149 -0.055 -0.129 -0.110 -0.048
Male 0.723 0.639 0.760 0.107 0.025 0.143 0.343 0.399 0.459
Model I: No Controls Final Model: All Controls
Statistically significant at .05
134
Table 3.5 Comparison of Final Model of Oppositional Behaviors using Student Reports for Cohort versus Longitudinal Students, Base Year through Second Follow-Up
Delinquent Behaviors Non-Compliant Behaviors Schooling Unpreparedness BY F1 F2 BY F1 F2 BY F1 F2
Model For All Students
(Cohort)
Mexican American 0.315 0.102 -0.012 0.447 0.449 0.326 0.155 0.088 0.011 African American 0.352 0.262 0.233 0.228 0.045 0.023 0.047 0.087 -0.004 Asian American -0.155 -0.295 -0.317 -0.101 -0.146 -0.019 -0.150 -0.111 -0.047
Male 0.728 0.646 0.773 0.089 -0.006 0.123 0.338 0.389 0.457
Delinquent Behaviors Non-Compliant Behaviors Schooling Unpreparedness BY F1 F2 BY F1 F2 BY F1 F2
Model For Students Present in
All Waves (Longitudinal)
Mexican American 0.173 -0.091 -0.138 0.132 0.240 0.226 -0.044 0.006 0.002 African American 0.168 0.010 0.001 -0.123 -0.225 -0.219 -0.083 0.005 -0.100 Asian American -0.062 -0.267 -0.264 0.053 -0.143 0.063 -0.132 -0.071 -0.008
Male 0.749 0.634 0.767 0.160 -0.004 0.112 0.321 0.397 0.486
Statistically significant at .05
135
Table 3.6 Model I: Predictors of Oppositional Identity using Teacher Reports, Base Year through Second Follow-Up
Severe Non-Compliant Non-Compliant Behaviors Behaviors
BY F1 F2 F1 F2
Mexican American 0.400 0.236 0.296 -0.057 0.108 African American 0.459 0.373 0.349 0.170 0.296 Asian American -0.404 -0.417 -0.155 -0.281 -0.181
Male 0.422 0.370 0.458 0.208 0.226
Statistically significant at .05
136
Table 3.7 Comparison of Model I and Model II of Oppositional Identity using Teacher Reports, Base Year through Second Follow-Up
Severe Non-Compliant Non-Compliant Behaviors Behaviors
BY F1 F2 F1 F2 Model I
Mexican American 0.400 0.236 0.296 -0.057 0.108 African American 0.459 0.373 0.349 0.170 0.296 Asian American -0.404 -0.417 -0.155 -0.281 -0.181
Male 0.422 0.370 0.458 0.208 0.226
Severe Non-Compliant Non-Compliant Behaviors Behaviors
BY F1 F2 F1 F2 Model II
Mexican American 0.102 -0.039 0.068 -0.150 -0.022 African American 0.120 0.053 0.108 0.052 0.157 Asian American -0.338 -0.357 -0.139 -0.257 -0.170
Male 0.428 0.380 0.472 0.218 0.241
Model I: No Controls Model II: Controlling for 8th Grade Achievement (Reading, Math, Science, and History)
Statistically significant at .05
137
Table 3.8 Final Model: Predictors of Oppositional Identity using Teacher Reports, Base Year through Second Follow-Up
Severe Non-Compliant Non-Compliant Behaviors Behaviors
BY F1 F2 F1 F2
Mexican American 0.064 -0.055 0.068 -0.114 -0.022 African American 0.083 0.058 0.121 0.051 0.153 Asian American -0.340 -0.322 -0.117 -0.273 -0.176
Male 0.419 0.373 0.466 0.212 0.237
8th Grade Reading Score -0.060 -0.045 -0.002 -0.004 0.003 8th Grade Math Score -0.221 -0.196 -0.150 -0.134 -0.099 8th Grade Science Score -0.069 -0.030 -0.031 -0.052 -0.039 8th Grade History Score -0.067 -0.081 -0.077 0.002 -0.039
Parental Communication -0.099 -0.065 -0.026 -0.001 0.041 Parental Limits 0.005 0.003 -0.020 -0.005 0.013
Parental Socioeconomic Status -0.018 -0.048 -0.029 0.114 0.061
Advanced Track -0.349 -0.404 -0.270 -0.092 -0.197 Vocational Track 0.016 -0.012 -0.019 -0.069 -0.030
Public School 0.048 0.007 0.049 0.070 0.136 Majority-FRPM School -0.043 -0.030 -0.098 0.035 0.039 Homogenous White School -0.097 0.005 0.005 -0.066 -0.048
Statistically significant at .05
138
Table 3.9 Comparison of Model I and Final Model of Oppositional Identity using Teacher Reports, Base Year through Second Follow-Up
Severe Non-Compliant Non-Compliant Behaviors Behaviors BY F1 F2 F1 F2 Model I
Mexican American 0.400 0.236 0.296 -0.057 0.108 African American 0.459 0.373 0.349 0.170 0.296 Asian American -0.404 -0.417 -0.155 -0.281 -0.181
Male 0.422 0.370 0.458 0.208 0.226
Severe Non-Compliant Non-Compliant Behaviors Behaviors
BY F1 F2 F1 F2 Final Model
Mexican American 0.064 -0.055 0.068 -0.114 -0.022 African American 0.083 0.058 0.121 0.051 0.153 Asian American -0.340 -0.322 -0.117 -0.273 -0.176
Male 0.419 0.373 0.466 0.212 0.237
Model I: No Controls Final Model: All Controls
Statistically significant at .05
139
Table 3.10 Model I: Predictors of Oppositional Identity using Student Reports, First Follow-Up
Delinquent Non-Compliant Schooling Delinquent Non-Compliant
Behaviors Behaviors Unpreparedness Attitudes Attitudes
Mexican American 0.102 0.449 0.088 -0.155 -0.068 African American 0.262 0.045 0.087 -0.235 -0.416 Asian American -0.295 -0.146 -0.111 -0.194 -0.143
Male 0.646 -0.006 0.389 0.613 0.172
Statistically significant at .05
140
Table 3.11 Comparison of Model I and Model II of Oppositional Identity using Student Reports, First Follow-Up Delinquent Non-Compliant Schooling Delinquent Non-Compliant
Behaviors Behaviors Unpreparedness Attitudes Attitudes Model I
Mexican American 0.102 0.449 0.088 -0.155 -0.068 African American 0.262 0.045 0.087 -0.235 -0.416 Asian American -0.295 -0.146 -0.111 -0.194 -0.143
Male 0.646 -0.006 0.389 0.613 0.172
Delinquent Non-Compliant Schooling Delinquent Non-Compliant
Behaviors Behaviors Unpreparedness Attitudes Attitudes Model II
Mexican American -0.122 0.310 0.049 -0.170 -0.069 African American -0.001 -0.119 0.043 -0.259 -0.433 Asian American -0.296 -0.125 -0.101 -0.188 -0.126
Male 0.648 0.030 0.402 0.623 0.192
Model I: No Controls Model II: Controlling for 8th Grade Achievement (Reading, Math, Science, and History)
Statistically significant at .05
141
Table 3.12 Final Model: Oppositional Identity using Student Reports, First Follow-Up Delinquent Non-Compliant Schooling Delinquent Non-Compliant
Behaviors Behaviors Unpreparedness Attitudes Attitudes
Mexican American -0.170 0.201 0.021 -0.151 -0.031 African American -0.037 -0.207 0.021 -0.232 -0.406 Asian American -0.311 -0.149 -0.110 -0.162 -0.097
Male 0.639 0.025 0.399 0.619 0.184
8th Grade Reading Score -0.034 0.078 0.033 0.037 0.074 8th Grade Math Score -0.131 -0.085 -0.029 -0.009 -0.043 8th Grade Science Score -0.076 -0.113 -0.004 -0.029 -0.052 8th Grade History Score -0.063 -0.028 -0.032 0.012 0.037
Parental Communication -0.043 -0.054 -0.100 -0.090 -0.064 Parental Limits -0.089 -0.108 -0.043 -0.111 -0.152
Parental Socioeconomic Status -0.047 0.009 -0.004 -0.024 0.026
Advanced Track -0.380 -0.445 -0.192 -0.192 -0.195 Vocational Track 0.037 -0.013 0.045 0.105 0.126
Public School -0.116 0.266 0.043 -0.044 0.216 Majority-FRPM School -0.066 -0.034 -0.069 -0.080 -0.061 Homogenous White School -0.074 -0.240 -0.085 0.024 -0.002
Statistically significant at .05
142
Table 3.13 Comparison of Model I and Final Model of Oppositional Identity using Student Reports, First Follow-Up Delinquent Non-Compliant Schooling Delinquent Non-Compliant
Behaviors Behaviors Unpreparedness Attitudes Attitudes Model I
Mexican American 0.102 0.449 0.088 -0.155 -0.068 African American 0.262 0.045 0.087 -0.235 -0.416 Asian American -0.295 -0.146 -0.111 -0.194 -0.143
Male 0.646 -0.006 0.389 0.613 0.172
Delinquent Non-Compliant Schooling Delinquent Non-Compliant
Behaviors Behaviors Unpreparedness Attitudes Attitudes Final Model
Mexican American -0.170 0.201 0.021 -0.151 -0.031 African American -0.037 -0.207 0.021 -0.232 -0.406 Asian American -0.311 -0.149 -0.110 -0.162 -0.097
Male 0.639 0.025 0.399 0.619 0.184
Model I: No Controls Final Model: All Controls
Statistically significant at .05
143
Table 3.14 Final Models for Predictors of Non-Compliant Behaviors, Base Year through Second Follow-Up, Student and Teacher Reports Base Year First Follow-Up Second Follow-Up Severe Severe Non-Compliant Non-Compliant Non- Non-Compliant Non- Behaviors Behaviors Compliant Behaviors Compliant Behaviors Behaviors Student Teacher Student Teacher Teacher Student Teacher Teacher
Mexican American 0.111 0.064 0.201 -0.055 -0.114 0.205 0.068 -0.022 African American -0.126 0.083 -0.207 0.058 0.051 -0.119 0.121 0.153 Asian American -0.122 -0.340 -0.149 -0.322 -0.273 -0.055 -0.117 -0.176
Male 0.107 0.419 0.025 0.373 0.212 0.143 0.466 0.237
8th Grade Reading Score -0.021 -0.060 0.078 -0.045 -0.004 0.062 -0.002 0.003 8th Grade Math Score -0.122 -0.221 -0.085 -0.196 -0.134 -0.024 -0.150 -0.099 8th Grade Science Score -0.091 -0.069 -0.113 -0.030 -0.052 -0.062 -0.031 -0.039 8th Grade History Score -0.088 -0.067 -0.028 -0.081 0.002 -0.049 -0.077 -0.039
Parental Communication -0.143 -0.099 -0.054 -0.065 -0.001 -0.035 -0.026 0.041 Parental Limits -0.138 0.005 -0.108 0.003 -0.005 -0.086 -0.020 0.013
Parental Socioeconomic Status -0.040 -0.018 0.009 -0.048 0.114 0.026 -0.029 0.061
Advanced Track -0.317 -0.349 -0.445 -0.404 -0.092 -0.363 -0.270 -0.197 Vocational Track -0.141 0.016 -0.013 -0.012 -0.069 -0.067 -0.019 -0.030
Public School 0.244 0.048 0.266 0.007 0.070 0.134 0.049 0.136 Majority-FRPM School 0.019 -0.043 -0.034 -0.030 0.035 -0.033 -0.098 0.039 Homogenous White School -0.111 -0.097 -0.240 0.005 -0.066 -0.183 0.005 -0.048
Statistically significant at .05
144 CHAPTER 4
OPPOSITIONAL IDENTITY AND EDUCATIONAL OUTCOMES
INTRODUCTION
This chapter examines the consequences of oppositional identity on academic outcomes.
Specifically, I will investigate high school dropout and achievement in mathematics and reading as key outcomes of interest. Graduating from high school is particularly relevant considering it is commonly used as a measure of downward assimilation within the segmented assimilation framework. This chapter is divided into four sections. The first section discusses the analysis strategy used in this chapter. The second section presents findings based student reports, while the third section presents findings based on teacher-reports. The final section summarizes the findings and provides a conclusion to this chapter.
The second section on student reports and third section on teacher reports are further divided into sub-sections. The first sub-section presents dropping out as an outcome, while the second sub-section presents achievement in reading and mathematics. These analyses are guided by two questions. What is the effect of oppositional identity on each of these academic outcomes? Does oppositional identity explain the racial/ethnic and gender gap in education outcomes and achievement?
ANALYSIS STRATEGY
Dependent Variables
Dropping-Out
This outcome variable of interest, dropping-out of school, comes from school records. If the respondent has ever dropped out, he or she is considered a dropout even if he or she 145 subsequently returned to school. The final variable is coded as dropped out of school (coded as
“1”) or did not drop out of school (coded as “0”).
Reading and Mathematics Achievement
Both reading and math achievement measures are derived from tests taken by students at different survey waves. These tests were derived and scored using IRT (Item Response Theory).
Independent Variables
Oppositional Identity
The key independent variables in these analyses are the behavioral and ideological indicators of oppositional identity. The specific independent variables used in the analyses depend on which survey wave is available and whether teacher and/or student reports are available since some factors are included in some waves, but not others. A detailed description of how these measures were developed these measures is presented in Chapter 2.
Student Characteristics
Prior achievement is operationalized as eighth grade test scores in reading, math, science, and history. I use the achievement measures in the multiple subjects to reduce the measurement error associated in using only one test score (Jencks 1985).
Students’ race/ethnicity and gender rely on student self-reports. The term
“race/ethnicity” is used throughout this discussion since both racial (White, Asian American, and
African American) and ethnic (Mexican American) variables utilized. These racial/ethnic variables are dichotomous with White youth serving as the reference category. Gender is dichotomized with females serving as the reference category.
146
Curriculum Characteristics: Track Placement
I create the indicator of track placement by using both transcript data and teacher- reported track placement. NELS researchers coded the transcript track placement variable using course descriptions for each student throughout their academic career. If transcript information is available, I used that information to assign track into advanced, academic, and vocational categories. More detailed information on this variable is available in the analysis strategy section of Chapter 3.
Parent Characteristics
Parental involvement is measured by two factors: parental communication and parental limits. The parental communication factor consists of three items measuring whether parents and students: discuss programs at school, discuss school activities and discuss things studied in class. The parental limits factor consists of four items measuring whether parents: check homework, require chores done, limit time watching TV, and limit going out with friends.
I also include parental socioeconomic status, measured with a NELS-specific composite that combines information on parents’ education (mother and father), family income, and parents’ occupational prestige (mother and father). These five items are standardized and then combined to produce the SES composite used in this study.
School Characteristics
I incorporate the following school-level variables: majority free-reduced meal school, homogenous White school, and public school. The dichotomous measure of majority free- reduced meal school is created from percent free or reduced meal and is coded “1” if the proportion of free or reduced meal students in the school is greater than or equal to 50% and “0”
147 if not. Similarly, the dichotomous measure of homogenous White school is created from percent
White and is coded as “1” if the proportion of White students is greater than or equal to 90% and
“0” if not. Finally, public school is coded as “1” if the school is public and “0” if the school is non-public.
SECTION II: STUDENT-REPORTED OPPOSITIONAL IDENTITY
Dropping Out
Oppositional Identity
Table 4.1 presents the independent effect12 of oppositional identity factors on dropping out from multiple models. In this analysis, the oppositional identity factors are from student reports. Similarly, each of the oppositional identity factor estimates are derived by carrying out separate analyses where they each serve as predictors. The logic behind this approach is to determine the independent effect of each factor without controlling for the other oppositional identity factors. This table presents both the base year and first follow-up cohorts. The dependent variable is whether any student enrolled during the base year (base year cohort) or first follow-up (first follow-up cohort) had dropped out by the second follow-up. Model I includes no covariates, Model II controls for prior achievement, and the final model includes all covariates.
This table shows a strong relationship between oppositional identity and dropping out of school. First, we observe that a one standard deviation increase in delinquent behaviors is associated with a .50 standard deviation decrease in dropping out of high school for the base year cohort. The effect of oppositional identity is higher for both the non-compliant behaviors and
12 I use the term “independent effect” in this chapter to refer to the fact that each oppositional identity factor estimate is derived by carrying out separate analyses where only one factor at a time serves as a predictor. The logic behind this approach is to determine the “independent effect” of each factor without controlling for the other oppositional identity factors. 148 schooling unpreparedness factors at .63 and .62, respectively. Adding 8th grade achievement decreases the effects of non-compliant behaviors and schooling unpreparedness, but does little to alter the effect of the delinquent behaviors factor. Thus, the coefficients for all three factors in
Model II are largely indistinguishable and hover around half a standard deviation (range from .43 to .50). In Model III with the inclusion of all other covariates, the effect of all three oppositional identity factors is further reduced. The strongest effect on dropping out is now the non- compliant behaviors factor at .45 standard deviations, followed by the schooling unpreparedness and delinquent behaviors factors. Thus, the effect of oppositional identity remains even after controlling for individual-level and school-level variables.
Table 4.1 shows the findings for the first follow-up cohort. In Model I with no covariates, we observe strong effects of oppositional identity for all five factors. The strongest effect on dropping-out is for the non-compliant behaviors factor, while the weakest is the delinquent attitudes factor. With the inclusion of 8th grade achievement in Model II, the effect of all five factors decreases although the relationships continue to be statistically significant. On the other hand, the inclusion of other covariates in Model III does little to alter the relationships between oppositional identity and dropping out.
Racial Gap
Table 4.2 presents the race/ethnicity and gender differences in the various models. Of key interest is the change in race/ethnicity and gender effects when the oppositional identity factors are added in Model III. The first model for the base year cohort shows that both African
American and Mexican American youth are more likely to drop out, while Asian American youth are less likely to drop out than Whites. Boys are also more likely to drop out than girls.
The racial/ethnic and gender differences decline sharply when all controls except oppositional
149 identity factors are introduced in Model II. Mexican American youth are now indistinguishable from Whites, while African American youth are now less likely to drop out. Specifically,
African American youth are .32 standard deviations lower than Whites in the probability of dropping out. This coefficient of -.32 for African Americans is comparable to that of Asian
Americans at -.29. The effect of gender roughly doubles, but is in the same direction with boys being more likely to drop out than girls. Interestingly, when oppositional identity is included in
Model, the relationship between race/ethnicity and dropping out remains the same.
The results for the first follow-up cohort are also presented on Table 4.2. Both African
American and Mexican American youth are more likely to drop out than Whites, but Asian
Americans are indistinguishable from Whites in the initial model. Similarly, boys and girls are not statistically different either. The inclusion of prior achievement drastically in Model II alters these relationships such that there are no statistically significant differences by race/ethnicity.
Adding oppositional identity in Model III does not change the relationship between race/ethnicity or gender and dropping out. Thus, the final model shows that there are no racial/ethnic or gender differences in dropping out.
Reading and Mathematics Achievement
10th Grade Oppositional Identity, 12th Grade Math and Reading Achievement
Table 4.3 presents the independent effect of 10th grade oppositional identity factors on
12th grade mathematics and reading achievement. The first model does not include any controls, while the second model incorporates 8th grade math achievement or 8th grade reading achievement when examining math or reading achievement as outcomes, respectively. The third model includes all other 8th grade math achievement measures (reading, science, and history) and the final model includes all other covariates.
150 I begin by discussing the results for math achievement. In the first model, a one standard deviation increase in the delinquent behaviors factor is associated with a .22 standard deviation decrease in mathematics achievement. The coefficients for the non-compliant behaviors, and schooling unpreparedness factors are very similar at -.20 and -.19, respectively. The weakest effect is for both attitudinal factors at -.15. The second model shows a significant reduction of the coefficients for all factors, but particularly the behavioral ones. Thus, all the coefficients are roughly -.10 in the second model. In the third and fourth models, the coefficients for oppositional identity remain essentially the same.
This table further shows the results for 12th grade reading achievement. In Model I when no covariates are included, a one standard deviation increase in oppositional identity factors is associated with a .14 to .22 decrease in reading achievement. The inclusion of 8th grade reading achievement decreases all the coefficients significantly, particularly the effect of delinquent behaviors factor which drops from -.22 in the first model to -.11 in the second model. The coefficients for the attitudinal factors decrease less between the first and second models.
Including the other 8th grade achievement measures in Model III has negligible effects on the coefficients of oppositional identity. In Model IV the other covariates are added and the strong effects oppositional identity are reduced almost by a factor of two.
Racial Gap, Reading Achievement
Table 4.4 presents racial/ethnic and gender differences in reading achievement during the first and second follow-ups. The first model includes no controls, the second includes all covariates except oppositional identity, and the final model incorporates oppositional identity factors. Model I with the first follow-up cohort shows that both Mexican American and African
American youth score roughly six to seven points lower on reading achievement tests than White
151 youth; Asian American youth are not statistically significant than White youth. Boys also score
2.45 points lower than girls in reading achievement when examining the first model. The inclusion of all other controls (except oppositional identity) reduces all racial/ethnic differences to non-significance, but the gender gap remains significant. The inclusion of oppositional identity in Model III does little to alter the results of Model II.
The findings with the second follow-up cohort are similar for race/ethnic groups and gender. After the inclusion of all covariates in Model III, African-Americans continue to have significant lower reading scores than Whites.
Racial Gap, Mathematics Achievement
Table 4.5 presents a similar analysis for mathematics achievement. Model I shows significant differences for Mexican Americans and African Americans in both survey waves.
Similarly, the gap in mathematics achievement between African American and White youth drops substantially between Models I and II. The inclusion of oppositional identity in Model III does little to alter the coefficients of Model II for both survey waves. The primary difference between both survey waves concerns the gender coefficients. Boys are indistinguishable from girls during the first follow-up, but score higher than girls in math achievement during the second follow-up.
SECTION III: TEACHER-REPORTED OPPOSITIONAL IDENTITY
Dropping Out
Oppositional Identity
Table 4.6 presents the independent effect of oppositional identity factors on dropping out of school for the both the base year and first follow-up cohorts. The first model has no controls,
152 the second controls for 8th grade achievement, while the final model includes all covariates. A one standard deviation change in non-compliant behaviors is associated with a .49 standard deviation increase in dropping out in the base year cohort. This coefficient drops slightly to .34 in Model II and to .32 in Model III. Thus, the non-compliant behaviors factor is a significant predictor of dropping out for the base year cohort, even after controlling for all covariates.
Similarly, both the non-compliant behaviors and severe non-compliant behaviors factors significantly increase the probability of dropping out in Model I for the first follow-up cohort.
Both Model II and Model III show a drop in the strength of the relationship although they both remain significant. Specifically, the non-compliant behaviors coefficient is equal to .42, while the severe non-compliant behaviors factor is equal to .39 in the final model (Model III). One key difference across survey waves is that the strength of the non-compliant behaviors factor is slightly higher in all models during the first follow-up.
Racial Gap
Table 4.7 presents the racial/ethnic and gender gap in dropping out for both the base year and first follow-up cohorts. The first model includes only group comparisons, the second model incorporates all other covariates except oppositional identity, and the final model includes oppositional identity. Model I with the base year cohort shows that there are significant differences by race/ethnicity, but no differences by gender. Specifically, Asian American youth are less likely than Whites to drop out of school, while both African American and Mexican
American youth are more likely to drop out. Nonetheless, the inclusion of covariates in Model II alters these relationships. Mexican American youth have similar drop out rates than Whites, while both Asian American and African American youth have lower drop out rates than Whites.
153 These race/ethnicity and gender differences remain essentially the same when controlling for oppositional identity in Model III.
Model I with the first follow-up cohort shows that the three racial/ethnic groups differ statistically. Both African Americans and Mexican Americans are more likely drop out than
Whites and Asian Americans are less likely to drop out than Whites. In Model II, all three non-
White groups are less likely to drop out than Whites. Model III does not alter the coefficients, although the African-American coefficient is no longer significant in this final model. In Model
I, boys are indistinguishable from girls. Nonetheless, boys are more likely to drop out than girls in Model II; this difference remains in Model III.
The data for the two survey waves are consistent in showing that the racial/ethnic differences are statistically significant in the initial models with no controls. Both African-
American and Mexican-American youth have a higher likelihood of dropping out than Whites, while Asian-American youth have a lower likelihood. One difference is seen for African-
Americans and Mexican-Americans. Mexican-Americans are similar to Whites for the base year cohort, but have a lower rate of dropping out in the first follow-up cohort. The opposite is true for African-Americans, who experience lower rates of dropping out in the base year cohort, but similar rates in the first follow-up cohort. The gender findings are also consistent, in that boys are no different than girls in Model I for both survey waves.
Reading and Mathematics Achievement
10th Grade Oppositional Identity, 12th Grade Math and Reading Achievement
Table 4.8 presents the independent effect of 10th grade oppositional identity on 12th grade mathematics and reading achievement. A one standard deviation change in non-compliant behaviors is associated with a .39 standard deviation decrease in mathematics achievement in
154 Model I. Similarly, a one standard deviation change in severe non-compliant behaviors is associated with a comparable .34 standard deviation decrease in mathematics achievement. The inclusion of 8th grade math achievement leads to a dramatic decrease in the effect of both oppositional identity factors. The inclusion of all other 8th grade measures of achievement
(Model III) and all other covariates (Model IV) further reduces the coefficients. Thus, the final model shows that the effect of oppositional identity is reduced by about two-thirds. Specifically, the final model has a coefficient of -.11 for the non-compliant behaviors factor and -.09 for the severe non-compliant behaviors factor.
Table 4.8 also shows that a one standard deviation in non-compliant behaviors and severe non-compliant behaviors is associated with a .33 and .28 decrease in 12th grade reading achievement, respectively. The most significant decline in coefficients comes between Model I and Model II, but there is a drop in strength across all four models. The final model shows that a one standard deviation increase in non-compliant behaviors is associated with a .06 standard deviation decrease in reading achievement. Similarly, a one standard deviation increase in severe on-compliant behaviors is associated with a .05 standard deviation decrease in reading achievement.
One key difference between both achievement outcomes is that the effect of both oppositional identity factors is slightly stronger for mathematics than reading. There is also a larger drop for reading achievement than mathematics achievement when comparing Model I and Model IV. Mathematics achievement sees a drop of about two-thirds between these two models, while reading achievement experiences a drop of about four-fifths.
Racial Gap, Reading Achievement
155 Table 4.9 presents the racial/ethnic gap in reading achievement during the first and second follow-ups. Both Mexican American and African American youth score 6.10 to 6.75 points lower than Whites during the first follow-up when examining Model I. Asian American youth score similar to Whites, while boys score 1.83 points lower than girls during this same time period and model. The inclusion of all other covariates except oppositional identity erases all racial/ethnic differences, but the gender gap remains at 1.10 points. The inclusion of oppositional identity does not change the coefficients of Model II.
The racial/ethnic gap is very similar when examining the first model during the second follow-up. Mexican Americans score 6.68 points lower than Whites, African Americans score
7.51 points than Whites, and Asian Americans do not score significantly differently from Whites.
Model II shows that Mexican Americans score similar to Whites, while African American score lower and Asian Americans score higher than Whites. These relationships hold for Model III as well. Finally, the gender gap decreases by roughly half from Model I to Model III, with boys scoring 1.31 points higher than girls in the final model. Thus, both survey waves provide a similar story when examining reading achievement.
Racial Gap, Math Achievement
Table 4.10 shows racial/ethnic differences in math achievement during the first and second follow-ups. All non-White groups are statistically significant from Whites in Model I during both survey waves. African Americans have higher achievement than Whites in Model
II, while Asian American youth exhibit higher achievement during both survey waves as well.
There is no gender gap for all models during the first follow-up, but boys exhibit higher math achievement than girls for all models during the second follow-up. The final model is identical to that of Model II in regards to both race/ethnicity and gender.
156
CONCLUSION
This chapter has examined the effect of oppositional identity on both dropping out of school and achievement. I find that while the effect of oppositional identity is significant and strong when omitting other covariates, a large portion of the effect is due to prior achievement.
The inclusion of prior achievement drastically alters the effect of oppositional identity on academic outcomes. Nonetheless, I find that oppositional identity continues to be a significant predictor of both dropping out and achievement (reading and math) even after controlling for not only prior achievement, but other relevant covariates as well. The significant effect of oppositional identity is so robust that all models and oppositional factors using both student and teacher reports are statistically significant.
The second question guiding this chapter asks whether accounting for oppositional identity explains the racial/ethnic and gender gap in achievement. The answer to this question is a resounding no. A consistent finding is that while oppositional identity is a significant predictor of achievement, it is not able to account for the achievement gap by race/ethnicity or gender after controlling for all other covariates (e.g. track placement, homogenous White school, etc). Thus, overall, oppositional identity is a predictor of dropping out and achievement (it is always significant), but its effect does little to explain racial/ethnic achievement gaps or gender gaps after controlling for other factors.
157
Table 4.1 Predictors of Dropping Out of School, Student Reports
Base Year First Follow-Up Model I Model II Model III Model I Model II Model III
Delinquent Behaviors 0.496 0.430 0.340 0.530 0.441 0.461 Non-Compliant Behaviors 0.629 0.499 0.454 0.654 0.572 0.567 Schooling Unpreparedness 0.622 0.439 0.367 0.617 0.496 0.486 Delinquent Attitudes - - - 0.414 0.330 0.345 Non-Compliant Attitudes - - - 0.504 0.409 0.413
Model I: No Controls Model II: 8th Grade Achievement (Reading, Math, Science, and History) Model III: All Controls
Statistically significant at .05
158
Table 4.2 Racial/Ethnic and Gender Differences in Dropping Out of School, Student Reports
Base Year First Follow-Up Model I Model II Model III Model I Model II Model III
Mexican American 0.465 -0.071 -0.071 0.395 -0.117 -0.116 African American 0.200 -0.320 -0.320 0.309 -0.143 -0.143 Asian American -0.255 -0.293 -0.293 -0.187 -0.270 -0.270
Male 0.063 0.102 0.102 0.053 0.072 0.072
Model I: No Controls Model II: All controls except oppositional identity factors Model III: All Controls plus oppositional identity factors
Statistically significant at .05
159
Table 4.3 Effect of 10th Grade Oppositional Identity on 12th Grade Math and Reading Achievement, Student Reports
Mathematics Model I Model II Model III Model IV
Delinquent Behaviors -0.222 -0.109 -0.102 -0.104 Non-Compliant Behaviors -0.202 -0.112 -0.105 -0.101 Schooling Unpreparedness -0.194 -0.112 -0.108 -0.103 Delinquent Attitudes -0.150 -0.092 -0.089 -0.091 Non-Compliant Attitudes -0.145 -0.096 -0.093 -0.093
Reading Model I Model II Model III Model IV
Delinquent Behaviors -0.216 -0.111 -0.103 -0.066 Non-Compliant Behaviors -0.180 -0.103 -0.091 -0.056 Schooling Unpreparedness -0.189 -0.107 -0.103 -0.057 Delinquent Attitudes -0.152 -0.093 -0.091 -0.058 Non-Compliant Attitudes -0.136 -0.091 -0.087 -0.054
Model I: No Controls Model II: 8th Grade Math/ Reading Achievement Model III: 8th Grade Achievement (Reading, Math, Science, and History) Model IV: All Controls
Statistically significant at .05
160
Table 4.4 Racial/Ethnic and Gender Differences in Reading Achievement, First and Second Follow-Ups, Student Reports
First Follow-Up Second Follow-Up Model I Model II Model III Model I Model II Model III
Mexican American -5.856 0.123 0.097 -6.001 -0.574 -0.588 African American -6.954 -0.134 -0.139 -6.981 -1.265 -1.277 Asian American -0.155 -0.372 -0.366 -0.502 -0.431 -0.428
Male -2.451 -1.192 -1.196 -2.229 -1.089 -1.092
Model I: No Controls Model II: All controls except oppositional identity factors Model III: All Controls plus oppositional identity factors
Statistically significant at .05
161
Table 4.5 Racial/Ethnic and Gender Differences in Math Achievement, First and Second Follow-Ups, Student Reports
First Follow-Up Second Follow-Up Model I Model II Model III Model I Model II Model III
Mexican American -8.534 0.242 0.242 -9.620 -0.523 -0.532 African American -10.380 -0.730 -0.733 -11.592 -1.426 -1.434 Asian American 2.933 1.316 1.341 3.448 1.597 1.590
Male 0.123 0.151 0.162 0.802 1.393 1.388
Model I: No Controls Model II: All controls except oppositional identity factors Model III: All Controls plus oppositional identity factors
Statistically significant at .05
162
Table 4.6 Predictors of Dropping Out of School, Teacher Reports
Base Year First Follow-Up Model Model Model Model Model Model I II III I II III
Non-Compliant Behaviors 0.488 0.339 0.317 0.528 0.431 0.417 Severe Non-Compliant Behaviors - - - 0.527 0.408 0.387
Model I: No Controls Model II: 8th Grade Achievement (Reading, Math, Science, and History) Model III: All Controls
Statistically significant at .05
163
Table 4.7 Racial/Ethnic and Gender Differences in Dropping Out of School, Teacher Reports
Base Year First Follow-Up Model Model Model I Model II Model I Model II III III
Mexican American 0.483 -0.079 -0.079 0.414 -0.116 -0.116 African American 0.252 -0.274 -0.275 0.401 -0.107 -0.107 Asian American -0.293 -0.294 -0.294 -0.274 -0.285 -0.285
Male 0.035 0.075 0.075 0.060 0.102 0.102
Model I: No Controls Model II: All controls except oppositional identity factors Model III: All Controls plus oppositional identity factors
Statistically significant at .05
164
Table 4.8 Effect of 10th Grade Oppositional Identity on 12th Grade Mathematics and Reading Achievement, Teacher Reports
Mathematics Model I Model II Model III Model IV
Non-Compliant Behaviors -0.390 -0.126 -0.113 -0.105 Severe Non-Compliant Behaviors -0.339 -0.098 -0.088 -0.087
Reading Model I Model II Model III Model IV
Non-Compliant Behaviors -0.334 -0.117 -0.084 -0.062 Severe Non-Compliant Behaviors -0.277 -0.093 -0.064 -0.048
Model I: No Controls Model II: 8th Grade Math/Reading Achievement Model III: 8th Grade Achievement (Reading, Math, Science, and History) Model IV: All Controls
Statistically significant at .05
165
Table 4.9 Racial/Ethnic and Gender Differences in Reading Achievement, Teacher Reports
First Follow-Up Second Follow-Up Model I Model II Model III Model I Model II Model III
Mexican American -6.095 -0.091 -0.090 -6.683 -0.145 -0.146 African American -6.749 -0.183 -0.183 -7.512 -1.236 -1.235 Asian American 0.175 -0.143 -0.150 -0.507 -0.338 -0.338
Male -1.825 -1.102 -1.103 -2.333 -1.306 -1.308
Model I: No Controls Model II: All controls except oppositional identity factors Model III: All Controls plus oppositional identity factors
Statistically significant at .05
166
Table 4.10 Racial/Ethnic and Gender Differences in Math Achievement, Teacher Reports
First Follow-Up Second Follow-Up Model I Model II Model III Model I Model II Model III
Mexican American -8.901 -0.253 -0.252 -11.150 -0.386 -0.381 African American -11.350 -1.233 -1.232 -12.851 -1.772 -1.772 Asian American 4.371 1.456 1.455 3.254 1.188 1.190
Male 0.310 0.074 0.076 0.952 1.276 1.277
Model I: No Controls Model II: All controls except oppositional identity factors Model III: All Controls plus oppositional identity factors
Statistically significant at .05
167 CHAPTER 5
CONCLUSION
This study examined the methodological and substantive issues related to oppositional identity using three survey waves of the National Educational Longitudinal Survey of 1988
(NELS). The sample used in analyses consisted of White, African American, Mexican
American, and Asian American youth. The first chapter introduced the pertinent literature and gave a general overview of my research questions. Chapter 2 presented analyses that offered a general sense of how oppositional identity exists. In doing so, I created an oppositional identity construct that I used in subsequent analyses. This oppositional identity construct was used as a dependent variable in Chapter 3, as I investigated the factors that predict oppositional identity.
Chapter 4 then employed the oppositional identity construct to examine its effect on two educational outcomes— high school dropout and achievement. This conclusion summarizes my findings detailed in Chapters 2-4, situates my findings in the body of extant literature on oppositional identity, highlights the limitations of my analyses, and paves the way for further research in this area.
THE MEASUREMENT OF OPPOSITIONAL IDENTITY
Three key findings emerge from the analyses conducted in Chapter 2 of this study. First, analyses reveal a substantial mismatch between teacher and student reports on the students’ behaviors. Unfortunately, a full exploration of the reasons for this mismatch is beyond the scope of this study. The extent of differences argue that teacher and student reports should not be used interchangeably, as is often the case in many studies. Interestingly, teachers tend to evaluate students more favorably than students evaluate themselves; while teachers’ reports of student
168 behavior are overwhelmingly positive, students’ reports of their own behavior are markedly more varied. In other words, teachers show systematic positivity bias for all behaviors during each survey wave. This is contrary to the expectation from previous research, particularly
Farkas’s (2008) contention that children engage in wishful thinking when reporting behaviors.
One drawback of this analysis is that I cannot adjudicate between the accuracy of teachers’ and students’ responses. Further research should compare student and teacher reports to actual school records. While some behaviors may not be available in school records, other behaviors such as attendance and tardiness are likely to be a well-documented component of student life. Similarly, future research is needed to discern whether teachers’ positivity bias is evident across students differing by race. It is possible that teachers’ positivity bias in reporting student behavior is more evident for some racial groups but not others. Consistent with this is the finding by Downey and Pribesh (2004) that White teachers judge African Americans more harshly than they judge White students engaging in the same behaviors.
Secondly, I find in Chapter 2 that the items used in previous studies of oppositional identity truly do not constitute a unidimensional construct. This is an important finding because if unidimensionality is not established, it questions the conclusions of prior studies. These conclusions are reflective of measures that purport to constitute oppositional identity rather than an indication of oppositional identity per se. I used all the items that have been used in previous studies of oppositional identity (and some additional items) to an exploratory factor analysis and found that multi-factor solutions are the norm. I found support for a second-order oppositional identity construct: while there are multiple factors, they can all be subsumed under a global oppositional identity construct. Thus, while these dimensions may be related, they should not be used interchangeably as proxies for each other. My substantive findings in Chapters 3 and 4
169 reinforce this point by highlighting that the results, whether using oppositional identity as a dependent or independent variable, differ based on the particular dimension of oppositional identity employed. In other words, the conclusions in these chapters are contingent on the particular factor of oppositional identity that is used.
Moreover, the questions of who is doing the reporting and what is being reported are serious ones to consider in oppositional identity studies. Not only were there differences when comparing student and teacher reports, there were further limits on teacher reports. One limitation is that teachers are in a good position to report on classroom behaviors but not on other dimensions of oppositional identity. Thus, indicators of severe oppositional identity (e.g. being arrested) are unlikely to be fully known or well documented by teachers. Thus, students are in the best position to report on internal attitudes and/or feelings and to capture more of the complexity of oppositional identity than teachers. Even if the accuracy analyses reveal that students' self reports are indeed more accurate, however, teacher reports would still remain a good complement in that they capture some aspect of oppositional identity.
A third and final point is that the question of when oppositional identity is studied is crucial. Researchers may reach diverging conclusions depending on when during students’ high school careers oppositional identity is being measured. I find that oppositional identity is not static, but rather, changes over the course of students’ educational trajectories. This change may be due to different sources (i.e. alpha, beta, and/or gamma change), but is nonetheless change that can lead researchers to reach different conclusions. For example, I find that oppositional identity largely declines over time.
PREDICTORS OF OPPOSITIONAL IDENTITY
170 In Chapter 3, I find that while race/ethnic differences in oppositional identity fluctuate, gender differences are robust, large, and consistent. Specifically, boys are more likely than girls to exhibit oppositional behaviors and attitudes; these gender differences persist after all controls.
This finding holds when using both reports by teachers or students, across survey waves, and when examining all oppositional identity factors. This finding is consistent with the notion that oppositional identity is more of a gendered phenomenon than a racial one (Lundy and Firebaugh
2005).
The one consistent racial/ethnic difference is that Asian American youth exhibited consistently lower levels of oppositional identity than all other groups. This is the case even after controlling for background and school environment. While current research offers little information about oppositional identity among Asian American youth, these findings on Asian
American youth are consistent with the segmented assimilation literature, which document many instances in which Asian American youth exhibit less delinquency as well as higher rates of achievement than other groups (Zhou and Bankston 1998). Thus, my findings support the segmented assimilation literature in this regard.
The racial/ethnic story for Mexican American and African American youth is more inconsistent than that of Asian Americans. Whether these two groups differ statistically from
Whites is very much contingent upon what controls are included, who is doing the reporting, what factor is being examined, and what survey wave is used. First, a consistent pattern is that both of these groups exhibit higher rates of oppositional identity than Whites when other variables are omitted. This is true regardless of whether student or teacher reports of oppositional identity are used. Once other covariates (e.g. prior achievement and schooling variables) are included in the models these patterns change. Both African Americans and
171 Mexican Americans exhibit lower levels of oppositional identity than Whites in the final model, when student reports are used. Most measures of oppositional identity among African
Americans and Mexican Americans are no longer statistically significant different once covariates like prior achievement, track placement, and parental involvement are held constant, when teacher reports are used. Some of oppositional identity measures show significantly higher rates for African Americans and Mexican Americans. Thus, researchers using teachers versus student reports would reach very different conclusions depending on which one is utilized.
A related story unfolds when comparing students’ reports of ideologies versus behaviors.
The two ideological dimensions are the only factors where all the non-White groups exhibit lower rates of oppositional identity than Whites in the first model with no controls and continue to have lower oppositional ideology in the final model with controls. Combining these findings with the findings that all non-White groups exhibit lower rates of oppositional behaviors in the final model with controls, I show that all non-White groups have lower rates of both oppositional ideology and behaviors in the final models. This is important because prior research argues that only ideology should be lower among non-White groups (Ainsworth-Darnell and Downey 1998).
This is also inconsistent with the original tenet of oppositional identity theory that Mexican
Americans and African Americans should have higher oppositional identity (Ogbu and Fordham
1986).
I also find that prior achievement is a significant predictor of oppositional identity and that omitting prior achievement would overstate the effect of racial/ethnic differences (Harris and
Robinson 2007). The most drastic reduction in racial/ethnic differences emerges after including prior achievement in the model. Nonetheless, while prior achievement affects oppositional identity, its effect is not constant for all factors or over time. For example, prior achievement
172 accounts for the racial/ethnic gap in oppositional identity for the non-compliant behaviors factor during the first follow-up using teacher reports. On the other hand, racial/ethnic differences persist for this same factor during the base year.
Notably, racial/ethnic differences in oppositional identity decline over time. This finding challenges existing research on oppositional identity, which posits that oppositionality is supposed to increase among minority youth over time as the perception of limited opportunities becomes more salient (Ogbu 1991). Investigating the reasons behind this decrease in racial/ethnic differences over time, my analyses reveal that the reason is likely a cohort effect, whereby students who are likely to exhibit more oppositionality drop out of school over time.
Nonetheless, I also find limited evidence that oppositionality itself declines over time—when examining longitudinal invariance in Chapter 2. Thus, further research is necessary to determine what proportion of the observed decline over time is attributable to a cohort effect and what proportion of the decline is due to a “true” waning effect that takes place over the high school course.
OPPOSITIONAL IDENTITY AND EDUCATIONAL OUTCOMES
I find that oppositional identity is a strong predictor of both mathematics and reading achievement. Specifically, an increase in oppositional identity behaviors and/or attitudes is associated with a decrease in both mathematics and reading achievement. The inclusion of prior achievement does decrease the effect of oppositional identity, but does not completely eliminate it. While the inclusion of other covariates further lowers the effect of oppositional identity, all of the indicators of oppositional identity were statistically significant and this was the case in every survey wave and with teacher as well as student reports). While the effect of oppositional
173 identity was significant, it did not account for more of the racial/ethnic gap than did the other covariates in the analyses.
LIMITATIONS AND FUTURE RESEARCH
Results discussed above should be interpreted with a consideration of the limitations of this study. One limitation in this study is that I do not examine generational differences among
Mexicans and Asians. Such an analysis would have addressed issues raised by the segmented assimilation framework, particularly that second-generation children of immigrants are insulated by social capital. Thus, one would expect that second-generation children would exhibit lower rates of oppositional identity. A second limitation is that the aggregation Asian Americans masks the massive heterogeneity among this group, and treats groups like Chinese students and
Cambodian students as similar, despite their drastically disparate socioeconomic conditions. The opposite is true when in my focus on only Mexican Americans rather than a more inclusive
“Latino” category. While I was restricted to Mexican Americans because of the focus on this group within the segmented assimilation literature, the applicability of my results on Mexican
American students to other Latino groups is unknown. Further research could delve into these issues to improve generalizability and/or uncover unique patterns.
Finally, I include both track placement and other school-level variables as covariates in my analyses, but the specific mechanisms behind these effects have not been addressed in my project. Future inquiry should be directed at delving into the “black box” of both academic tracking and school-level variables by investigating how these variables impact oppositional identity (and academic outcomes). Measuring the specific components of school environment instead of relying on broad measures would clarify the specific ways both track placement and
174 other schooling variables exhibits its effect(s), thereby facilitating policy interventions. For example, it may turn out that only some teacher practices are the driving forces behind these effects, making these practices the central focus of any interventions.
175 APPENDIX
FINAL MODELS FOR ALL ANALYSES
Chapter 3
Tables 3.1 to 3.14
th Oppositional Identity = β0 + β1(Race/Ethnicity) + β2(Male) + β3(8 Grade Achievement) +
β4(Parental Communication) + β5(Parental Limits) + β6(Parental Socioeconomic Status) +
β7(Track Placement) + β8(Race/Ethnicity) + β9(Public School) + β10(Majority FRPM School) +
β11(Homogenous White School)
176
Chapter 4
Tables 4.1, 4.2, 4.6, 4.7
th Dropping-Out = β0 + β1(Oppositional Identity) + β2(Race/Ethnicity) + β3(Male) + β4(8 Grade
Achievement) + β5(Parental Communication) + β6(Parental Limits) + β7(Parental Socioeconomic
Status) + β8(Track Placement) + β9(Race/Ethnicity) + β10(Public School) + β11(Majority FRPM
School) + β12(Homogenous White School)
Tables 4.3 and 4.8
th th 12 Grade Math/Reading Achievement = β0 + β1(10 Grade Oppositional Identity) +
th β2(Race/Ethnicity) + β3(Male) + β4(8 Grade Achievement) +
β5(Parental Communication) + β6(Parental Limits) + β7(Parental Socioeconomic Status) +
β8(Track Placement) + β9(Race/Ethnicity) + β10(Public School) + β11(Majority FRPM School) +
β12(Homogenous White School)
Tables 4.4, 4.5, 4.9, and 4.10
Math/Reading Achievement = β0 + β1(Oppositional Identity) + β2(Race/Ethnicity) + β3(Male) +
th β4(8 Grade Achievement) + β5(Parental Communication) + β6(Parental Limits) + β7(Parental
Socioeconomic Status) + β8(Track Placement) + β9(Race/Ethnicity) + β10(Public School) +
β11(Majority FRPM School) + β12(Homogenous White School)
177 TABLES
TABLE 2A Factor Loadings of Items From Teachers, Base Year (Alternative)
Confirmatory Factor Analysis
Teacher One Teacher Two Non-Compliant Non-Compliant Item Behaviors Behaviors
True if student rarely completes homework 0.86 0.86 True if student is frequently tardy 0.71 0.74 True if student is inattentive in class 0.94 0.93 True if student is frequently disruptive 0.73 0.73 True if student is frequently absent 0.56 0.58
CFI 0.97 0.97 RMSEA 0.076 0.075
178
TABLE 2B Factor Loadings of Items From Teachers, First Follow-Up (Alternative)
Confirmatory Factor Analysis
Teacher One Teacher Two Severe Severe Non- Non- Non- Non- Compliant Compliant Compliant Compliant Behaviors Behaviors Behaviors Behaviors Item
True if student usually does NOT work hard 0.92 0.92 How often student does NOT complete homework 0.89 0.89 How often student is tardy 0.56 0.55 How often student is inattentive in class 0.87 0.87 How often student is disruptive 0.63 0.64
True if spoke to parents about student behavior 0.92 0.93 True if spoke to parents about student homework 0.94 0.93 True if spoke to parents about student absenteeism 0.80 0.79
Correlation between factors 0.35 0.37
CFI 0.99 0.99 RMSEA 0.059 0.057
179
TABLE 2C Factor Loadings of Items From Teachers, First Follow-Up (Alternative #2)
Confirmatory Factor Analysis
Teacher One Teacher Two Severe Severe Non- Non- Non- Non- Compliant Compliant Compliant Compliant Behaviors Behaviors Behaviors Behaviors Item
True if student usually does NOT work hard 0.92 0.92 How often student does NOT complete homework 0.90 0.91 How often student is tardy 0.58 0.56 How often student is inattentive in class 0.84 0.85 How often student is absent 0.57 0.55
True if spoke to parents about student behavior 0.90 0.91 True if spoke to parents about student homework 0.95 0.94 True if spoke to parents about student absenteeism 0.83 0.84
Correlation between factors 0.35 0.37
CFI 0.98 0.98 RMSEA 0.068 0.075
180
TABLE 2D Factor Loadings of Items From Teachers First Follow-Up (Alternative #3)
Confirmatory Factor Analysis
Teacher One Teacher Two Severe Severe Non- Non- Non- Non- Compliant Compliant Compliant Compliant Behaviors Behaviors Behaviors Behaviors Item
True if student usually does NOT work hard 0.92 0.92 How often student does NOT complete homework 0.89 0.89 How often student is tardy 0.60 0.58 How often student is inattentive in class 0.86 0.87 How often student is absent 0.55 0.54 How often student is disruptive 0.62 0.62
True if spoke to parents about student behavior 0.92 0.92 True if spoke to parents about student homework 0.93 0.93 True if spoke to parents about student absenteeism 0.83 0.84
Correlation between factors 0.36 0.38
CFI 0.97 0.97 RMSEA 0.072 0.076
181
TABLE 2E Factor Loadings of Items From Teachers, Second Follow-Up (Alternative)
Confirmatory Factor Analysis
Severe Non- Non-Compliant Compliant Behaviors Item Behaviors
True if student is NOT motivated to work hard for good grades 0.90 How often student does NOT complete homework 0.87 How often student is tardy 0.57 How often student is inattentive in class 0.85 How often student is disruptive 0.62
True if spoke to parents about student behavior 0.93 True if spoke to parents about student homework 0.83 True if spoke to parents about student absenteeism 0.82
Correlation between factors 0.62
CFI 0.98 RMSEA 0.071
182
TABLE 2F Factor Loadings of Items From Teachers, Second Follow-Up (Alternative #2)
Confirmatory Factor Analysis
Severe Non- Non-Compliant Compliant Behaviors Item Behaviors
True if student is NOT motivated to work hard for good grades 0.90 How often student does NOT complete homework 0.88 How often student is tardy 0.60 How often student is inattentive in class 0.82 How often student is absent 0.62
True if spoke to parents about student behavior 0.85 True if spoke to parents about student homework 0.82 True if spoke to parents about student absenteeism 0.92
Correlation between factors 0.63
CFI 0.97 RMSEA 0.086
183
TABLE 2G Factor Loadings of Items From Teachers, Second Follow-Up (Alternative #3)
Confirmatory Factor Analysis
Severe Non- Non-Compliant Compliant Behaviors Item Behaviors
True if student is NOT motivated to work hard for good grades 0.90 How often student does NOT complete homework 0.87 How often student is tardy 0.61 How often student is inattentive in class 0.84 How often student is absent 0.61 How often student is disruptive 0.60
True if spoke to parents about student behavior 0.90 True if spoke to parents about student homework 0.80 True if spoke to parents about student absenteeism 0.90
Correlation between factors 0.65
CFI 0.96 RMSEA 0.089
184
Table 3A Measurement Invariance across Race/Ethnicity, Student Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.982 - 0.034 Measurement Invariance 0.983 0.001 0.027
First Follow-Up No measurement invariance 0.965 - 0.024 Measurement Invariance 0.969 0.004 0.021
Second Follow-Up No measurement invariance 0.985 - 0.017 Measurement Invariance 0.979 0.006 0.017
185
Table 3B Measurement Invariance across Gender, Student Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.981 - 0.033 Measurement Invariance 0.976 0.005 0.032
First Follow-Up No measurement invariance 0.953 - 0.028 Measurement Invariance 0.951 0.002 0.027
Second Follow-Up No measurement invariance 0.986 - 0.029 Measurement Invariance 0.973 0.013 0.035
186
Table 3C Measurement Invariance across Track Placement, Student Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.981 - 0.034 Measurement Invariance 0.983 0.002 0.028
First Follow-Up No measurement invariance 0.956 - 0.028 Measurement Invariance 0.952 0.004 0.028
Second Follow-Up No measurement invariance 0.987 - 0.017 Measurement Invariance 0.985 0.002 0.015
187
Table 3D Measurement Invariance across School Type, Student Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.979 - 0.035 Measurement Invariance 0.982 0.003 0.029
First Follow-Up No measurement invariance 0.969 - 0.024 Measurement Invariance 0.958 0.011 0.026
Second Follow-Up No measurement invariance 0.986 - 0.017 Measurement Invariance 0.984 0.002 0.016
188
Table 3E Measurement Invariance across Percent White Status, Student Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.980 - 0.035 Measurement Invariance 0.980 0 0.030
First Follow-Up No measurement invariance 0.958 - 0.028 Measurement Invariance 0.962 0.004 0.025
Second Follow-Up No measurement invariance 0.986 - 0.017 Measurement Invariance 0.985 0.001 0.015
189
Table 3F Measurement Invariance across Percent FRPM Status, Student Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.981 - 0.033 Measurement Invariance 0.984 0.003 0.027
First Follow-Up No measurement invariance 0.957 - 0.025 Measurement Invariance 0.966 0.009 0.021
Second Follow-Up No measurement invariance 0.988 - 0.015 Measurement Invariance 0.990 0.002 0.011
190
Table 3G Measurement Invariance across Race/Ethnicity, Teacher Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.995 - 0.046 Measurement Invariance 0.995 0 0.035
First Follow-Up No measurement invariance 0.995 - 0.044 Measurement Invariance 0.994 0.001 0.036
Second Follow-Up No measurement invariance 0.993 - 0.048 Measurement Invariance 0.994 0.001 0.033
191
Table 3H Measurement Invariance across Gender, Teacher Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.995 - 0.049 Measurement Invariance 0.992 0.003 0.051
First Follow-Up No measurement invariance 0.994 - 0.049 Measurement Invariance 0.992 0.002 0.046
Second Follow-Up No measurement invariance 0.993 - 0.048 Measurement Invariance 0.991 0.002 0.045
192
Table 3I Measurement Invariance across Track Placement, Teacher Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.996 - 0.043 Measurement Invariance 0.996 0 0.034
First Follow-Up No measurement invariance 0.995 - 0.043 Measurement Invariance 0.996 0.001 0.033
Second Follow-Up No measurement invariance 0.993 - 0.049 Measurement Invariance 0.994 0.001 0.035
193
Table 3J Measurement Invariance across School Type, Teacher Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.996 - 0.043 Measurement Invariance 0.995 0.001 0.037
First Follow-Up No measurement invariance 0.995 - 0.042 Measurement Invariance 0.995 0 0.034
Second Follow-Up No measurement invariance 0.993 - 0.048 Measurement Invariance 0.994 0.001 0.038
194
Table 3K Measurement Invariance across Percent White Status, Teacher Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.996 - 0.044 Measurement Invariance 0.995 0.001 0.038
First Follow-Up No measurement invariance 0.995 - 0.043 Measurement Invariance 0.995 0 0.037
Second Follow-Up No measurement invariance 0.993 - 0.048 Measurement Invariance 0.994 0.001 0.036
195
Table 3L Measurement Invariance across Percent FRPM Status, Teacher Reports CFI Δ CFI RMSEA
Base Year No measurement invariance 0.996 - 0.042 Measurement Invariance 0.996 0 0.035
First Follow-Up No measurement invariance 0.995 - 0.042 Measurement Invariance 0.995 0 0.033
Second Follow-Up No measurement invariance 0.993 - 0.047 Measurement Invariance 0.994 0.001 0.037
196 REFERENCES
Ainsworth-Darnell, James W., and Douglas B. Downey. 1998. “Assessing the Oppositional Culture Explanation for Racial/Ethnic Differences in School Performance.” American Sociological Review. 63:536-53.
Alba, Richard D. and Victor Nee. 2003. Remaking the American Mainstream: Assimilation and Contemporary Immigration. Cambridge: Harvard University Press.
Ansalone, George. 2001. “Schooling, Tracking, and Inequality.” Journal of Children and Poverty. 7:33-47.
Bergin, David A. and Helen C. Cooks. 2008. “High School Students of Color talk about Accusations of Acting White.” Pp. 145-66 in Minority Status, Oppositional Culture, & Schooling, edited by J.U. Ogbu. New York: Routledge.
Coleman James S. 1988. “Social Capital in the Creation of Human Capital.” American Journal of Sociology. 94:S95-S120.
Darling-Hammond, Linda. 2004. “The Color Line in American Education: Race, Resources, and Student Achievement.” Social Science Research on Race. 1:213-41.
Downey, Douglas. 2008. “A Funny Thing Happened on the Way to Confirming Oppositional Culture Theory.” Pp. 298-311 in Minority Status, Oppositional Culture, & Schooling, edited by J.U. Ogbu. New York: Routledge.
Farkas, George. 2008. “Quantitative Studies of Oppositional Culture: Arguments and Evidence.” Pp. 312-47 in Minority Status, Oppositional Culture, & Schooling, edited by J.U. Ogbu. New York: Routledge.
Ferguson, Annette. 2001. Bad Boys: Public Schools in the Making of Black Masculinity. Ann Arbor: University of Michigan Press.
Fuligni, Andrew J. 1997. “The Academic Achievement of Adolescents from Immigrant Families: The Roles of Family Background, Attitudes, and Behavior.” Child Development. 68:351- 363.
Gamoran, Adam. 1987. “The Stratification of High School Learning Opportunities.” Sociology of Education. 60:135-155.
Gamoran, Adam and Robert D. Mare. 1989. “Secondary School Tracking and Educational Inequality: Compensation, Reinforcement, or Neutrality?” American Journal of Sociology. 94:1146-1183.
Good, Thomas L. 1981. “Teacher Expectations and Student Perceptions: A Decade of Research.” Educational Leadership. 38:415-22. 197
Gordon, Milton. 1964. Assimilation in American Life. New York, Oxford University Press.
Hoover-Dempsey Kathleen V. and Howard M. Sander. 1995. “Parental Involvement in Children’s Education: Why Does it Make a Difference?” Teachers College Record 97:310-331
Irving, Anthony Miles and Cynthia Hudley. 2008. “Oppositional Identity and Academic Achievement among African American Males.” Pp. 374-94 in Minority Status, Oppositional Culture, & Schooling, edited by J.U. Ogbu. New York: Routledge.
Kelly, Sean. 2004. “Do increased levels of parental involvement account for social class differences in track placement?” Social Science Research. 33:626–59.
Kennedy, M. M. (1999). Approximations to indicators of student outcomes. Educational Evaluation and Policy Analysis, 21(4), 345-363.
Koziol, S.M. & Burns, P. (1986). Teachers' accuracy in self-reporting about instructional practices using a focused self-report inventory. Journal of Educational Research, 79(4), 205-209.
Kroneberg, Clemens. (2008). Ethnic Communities and School Performance among the New Second Generation in the United States: Testing the Theory of Segmented Assimilation. The ANNALS of the American Academy of Political and Social Science, 620: 138-160.
Mayer, D. (1999). Measuring Instructional Practice: Can Policy Makers Trust Survey Data? Educational Evaluation and Policy Analysis 21(1): 29-45.
Mickelson, Roslyn A. 2008. “The Structure of Opportunity and Adolescents’ Academic Achievement Attitudes and Behaviors.” Pp. 348-73 in Minority Status, Oppositional Culture, & Schooling, edited by J.U. Ogbu. New York: Routledge.
Morrison, George S. 1978. Parental Involvement in the Home, School and Community. Columbus, OH: Charles E. Merrill.
Oakes, Jeannie and Gretchen Guiton. 1995. “Matchmaking: The Dynamics of High School Tracking Decisions.” American Educational Research Journal. 32:3-33.
Oakes, Jeannie. 2005. Keeping Track: How Schools Structure Inequality. New Haven, CT: Yale University Press.
Ogbu John U. and Signithia Fordham. 1986. “Black Students’ School Success: Coping with the Burden of Acting White.” The Urban Review. 18:176-206.
198 Park, Robert and E.W. Burgess. 1921. Introduction to the Science of Sociology. Chicago: University of Chicago Press.
Park, Robert. 1950. Race and Culture. Glencoe, Illinois: The Free Press.
Portes, Alejandro and Min Zhou. 1993. The New Second Generation: Segmented Assimilation and Its Variants. Annals of the American Academy of Political and Social Science 530 (November): 74-96.
Portes, Alejandro and Ruben G. Runbaut. 2001. Legacies: The Story of the Immigrant Second Generation. Berkeley and Los Angeles: University of California Press.
Shannon, Sheila M., & Silvia Lojero-Latimer. 1996. “A story of struggle and resistance: Latino parent involvement in the schools.” Journal of Educational Issues of Language Minority Students. 16:301-319.
Sui-Chu, Esther H. and J. Douglas Willms. 1996. “Effects of Parental Involvement on Eight- Grade Achievement.” Sociology of Education. 69:126-41.
Taylor, April. 2008. “A Quantitative Examination of Oppositional Identity among African American and Latino Middle-School Students.” Pp. 481-95 in Minority Status, Oppositional Culture, & Schooling, edited by J.U. Ogbu. New York: Routledge.
Telles, Edward E. and Vilma Ortiz. 2008. Generations of Exclusion: Mexican Americans, Assimilation, and Race. New York: Russell Sage Foundation.
Unnever, J. D., Kerckhoff, A. C., and Robinson, T. J. 2000. District variations in educational resources and student outcomes. Economics of Education Review, 19:245–59.
Valencia, Richard R. and Mary S. Black. 2002. “‘Mexican Americans Don’t Value Education!’—On the Basis of the Myth, Mythmaking, and Debunking.” Journal of Latinos and Education, 1:81–103.
Valenzuela, Angela. 1999. Subtractive Schooling: U.S.-Mexican Youth and the Politics of Caring. Albany: State University of New York Press.
Van Laar, Colette. 2000. “The Paradox of Low Academic Achievement but High Self-Esteem in African American Students: An Attributional Account.” Educational Psychology Review. 12:33-61.
Waldinger, Roger and Cynthia Feliciano. 2004. “Will the new second generation experience ‘downward assimilation’? Segmented Assimilation re-assessed.” Ethnic and Racial Studies. 27:376-402. Xiong, Yang S. and Min Zhou. 2006. “Structuring Inequality: How California Selectively Tests, Classifies, and Tracks Language Minority Students.” California Policy Options. UCLA School of Public Affairs.
199
Zhou, Min and Carl Bankston. 1998. Growing Up American: How Vietnamese Children Adapt to Life in the United States. New York: Russell Sage Foundation Press.
Zhou, Min. 1997. “Growing Up American: The Challenge Confronting Immigrant Children and Children of Immigrants.” Annual Review of Sociology. 23:63-95.
Zhou, Min and Susan S. Kim. “Community Forces, Social Capital, and Educational Achievement: The Case of Supplementary Education in the Chinese and Korean Immigrant communities.” Harvard Educational Review. 76:1-29.
200