The Best WritersRacial and Gender Identity Among Black Adolescent Males: An IntersectionalityPerspectiveAuthor(s): Leoandra Onnie Rogers, Marc A. Scott and Niobe WaySource: Child Development, Vol. 86, No. 2 (MARCH/APRIL 2015), pp. 407-424Published by: Wiley on behalf of the Society for Research in Child DevelopmentStable URL: https://www.jstor.org/stable/24696597Accessed: 19-03-2022 00:45 UTCJSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a widerange of content in a trusted digital archive. We use information technology and tools to increase productivity andfacilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available atTerms and Conditions of UseSociety for Research in Child Development, Wiley are collaborating with JSTOR todigitize, preserve and extend access to Child DevelopmentThis content downloaded from on Sat, 19 Mar 2022 00:45:01 UTCAll use subject to https://about.jstor.org/termsChild Development, March/April 2015, Volume 86, Number 2, Pages 407^124Racial and Gender Identity Among Black Adolescent Males: An Intersectionality PerspectiveLeoandra Onnie Rogers Marc A. Scott and Niobe Way University of Washington New York UniversityA considerable amount of social identity research has focused on race and racial identity, while gender iden tity, particularly among Black adolescents, remains underexamined. The current study used survey data from 183 Black adolescent males (13-16 years old) to investigate the development and relation between racial and gender identity centrality and private regard, and how these identities impact adjustment over time. It was found that dimensions of racial and gender identity were strongly correlated. Levels of racial centrality increased over time while gender centrality, and racial and gender private regard declined. In addition, racial and gender identity uniquely contributed to higher levels of psychological well-being and academic adjust ment. These findings are discussed within the context of existing identity theories and intersectionality theory.The social groups that youth belong to, and the socially constructed norms and expectations that accompany them, provide critical information for how youth make sense of who they are and want to be. Social identity refers to the sense of “we-ness” that develops from being the member of a social group, such as being Black or male (Ruble et al., 2004). Social identities are important because they can influence how we evaluate ourselves (e.g., Crocker & Major, 1989) and judge our own abilities (e.g., Bouchey & Harter, 2005) as well as how we interact with others (e.g., Tajfel & Turner, 1986). Much of the social identity research has focused on race (e.g., Crocker, Major, & Steele, 1998), and racial identity is oft cited as one of the keys to promoting positive development among Black youth (Quin tana et al., 2006; Spencer & Markstrom-Adams, 1990; Syed, Azmitia, & Cooper, 2011). However, race is not the only relevant social identity. Gender is one of the first social identities to emerge early in childhood (e.g., Martin & Ruble, 2009) and, like racial identity, has been shown to be important for adolescents’ well-being and adjustment (e.g., Halim & Ruble, 2010; Perry & Pauletti, 2011). Yet, com pared to racial identity, there are very few studies of adolescent gender identity development, particu larly among Black youth (e.g., Corby, Hodges, & Perry, 2007). We know even less about the natureThis research was supported by dissertation fellowships awarded to Leoandra Onnie Rogers from the Spencer Founda tion and the Ford Foundation.Correspondence concerning this article should be addressed to Leoandra Onnie Rogers, UW I-LABS, PBB, Rm 464, 1715 Colum bia Rd. North, Seattle, WA 98195. Electronic mail may be sent to [email protected]of the relation between racial and gender identity— whether they follow a similar trajectory over time, the extent to which they are related to each other, or the relative and joint influence of each type of identity on adjustment. In this article, we apply an intersectionality lens to psychological theories of identity explore the patterns of change over time and correlates of racial and gender identity among Black adolescent males.Identity TheoriesThis article draws upon social identity theory and developmental theories of identity. The social identity approach focuses on the content of identity —the features that make up an identity, such as affect, attitudes, and behaviors (e.g., Ashmore, Deaux, & McLaughlin-Volpe, 2004; Tajfel & Turner, 1986)—while the developmental approach focuses on the process of identity formation—when and how one forms an identity (e.g., Cross, 1991; Phinney, 1989). Other scholars have noted this “content” ver sus “process” distinction in the identity literature and called for more work that blends the views (e.g., Syed & Azmitia, 2008). We employ both theoretical orientations to ask: How does the content of racialand gender identity change during adolescence and to what extent are identities processes interrelated?Identity, according to social identity theory, refers to the “value and emotional significance” that one© 2014 The AuthorsChild Development © 2014 Society for Research in Child Development, Inc. All rights reserved. 0009-3920/2015/8602-0006 DOI: 10.1111/cdev.12303This content downloaded from on Sat, 19 Mar 2022 00:45:01 UTCAll use subject to https://about.jstor.org/terms408 Rogers, Scott, and Wayderives from his or her membership in, and attach ment to, a social group (Tajfel, 1981, p. 63). Thus, the relevance of a social identity is determined by (a) the level of importance an individual places on a social group and (b) the affective evaluation one attaches to that group. In the multidimensional model of racial identity (Sellers, Smith, Shelton, Rowley, & Chavous, 1998), these dimensions are termed centrality and private regard, respectively. This view acknowledges that not all Black people view race as equally important to their identities: Some view it as highly central while others give it less primacy (Chavous et al., 2003; Sellers, Cope land-Linder, Martin, & L’Heureux Lewis, 2006). Likewise, individuals vary in how they regard their racial group—some feel positively while others feel more neutral or negative (e.g., Seaton, Scottham, & Sellers, 2006). A similar story exists for gender: Multiple dimensions, including centrality and pri vate regard, comprise gender identity (Ashmore et al., 2004; Egan & Perry, 2001; Luhtanen & Croc ker, 1992). But, unlike racial identity, research on gender identity rarely includes youth of color (Cor by et al., 2007).Research on gender identity among Black males, specifically, focuses on adherence to gender-typed beliefs such as “hyper masculinity” (e.g., Cunning ham, 1999; Cunningham, Swanson, & Hayes, 2013; Davis, 2001; Ferguson, 2000; Spencer, Fegley, Harpalani, & Seaton, 2004; Stevenson, 1997). This gender identity literature indicates that levels of adherence to norms of masculinity vary consider ably across individuals, with some boys clinging tightly to traditional masculine ideologies and oth ers deviating from these gender expectations (Cunningham, 1999; Ferguson, 2000; Santos, Galli gan, Pahlke, & Fabes, 2013; Santos, Way, & Hughes, 2011; Way, 2011). In the social identity framework, gender-typed beliefs constitute one dimension of gender identity, which varies along side gender centrality and private regard (e.g., Ash more et al., 2004; Egan & Perry, 2001). Thus, one could expect similar variability in the centrality and private regard dimensions of gender identity. Indeed, research using Egan and Perry’s (2001) multidimensional measure of gender identity for children shows significant variance in how much value youth place on gender and how positively (or negatively) they feel about their gender group membership (e.g., Carver, Yunger, & Perry, 2003; Lurye, Zosuls, & Ruble, 2008).While the social identity perspective on racial and gender identity captures variation between indi viduals at a specific moment in time, it does notexplain whether (or how) identities vary within individuals over time. This is the work of develop mental theories, such as Cross’s (1991) racial iden tity model or Phinney’s (1989) three-stage ethnic identity model. In these models, racial identity begins “unexamined,” as youth express little aware ness of or interest in the meaning or significance of race in their lives (i.e., low centrality). With advances in social-cognitive development, aware ness of social group norms and expectations, and broadening social experiences (e.g., transitioning to high school), youth begin to “explore” their racial identities (e.g., asking questions, seeking informa tion about their racial heritage). The result of such active exploration is an “achieved” or committed racial identity, where youth view race as a positive and important aspect of the self (e.g., Cross, 1991; Phinney, 1989). Thus, developmentally, the central ity of race is assumed to increase with age as affec tive or evaluative ties are formed. Empirical data generally support this idea: Levels of racial central ity and private regard are higher among older youth compared to younger (Quintana, 2007; Yip, Seaton, & Sellers, 2006), with midadolescence mark ing a key period for identity growth (French, Seid man, Allen, & Aber, 2006; Pahl & Way, 2006).The literature on gender identity development during adolescence is less established. Generally, models of gender identity development focus on early childhood identification and behavioral wchoices such as the age at which children label themselves and others as “boys” and “girls” or associate with gender typical activities (e.g., girls play with dolls, boys play with trucks; girls like pink, boys like blue; Halim & Ruble, 2010; Martin & Ruble, 2009). The “gender intensification hypothe sis” (Hill & Lynch, 1983), however, argues that with the physical and social changes that accompany adolescence comes a heighted awareness of and interest in gender as youth engage what it means to belong to a gender, the extent to which one “fits” with their gender, and how they feel about it. From this view, one would expect gender identity central ity and private regard to increase during adoles cence, similar to racial identity. We do not know whether racial and gender identity journey the same trajectory or if one identity takes primacy over the other. For example, racial minority youth are aware of race earlier and tend to score higher on measures of racial identity that racial majority peers (e.g., Fra ble, 1997; Phinney, 1990; Turner & Brown, 2007), and thus they may experience race as more central to their identity than gender and evaluate race more positively. Yet, if attention to gender is heightenedThis content downloaded from on Sat, 19 Mar 2022 00:45:01 UTCAll use subject to https://about.jstor.org/termsduring adolescence (Hill & Lynch, 1983), this pat tern of the primacy of racial identity may not be evident during middle adolescence.An Intersectionality LensBecause midadolescence is a peak period for identity exploration and development (Erikson, 1968; French et al., 2006), youth are not only mak ing sense of their membership in a social group but also weaving together how their identity group memberships fit together into a coherent whole. The literatures on racial identity and gender iden tity are largely distinct, with very few empirical studies that investigate both identities concurrently (e.g., Frable, 1997; Settles, 2006; Turner & Brown, 2007). Yet, intersectionality theorists argue that social categories (and thus identities) are “mutually constitutive.” As such, the experience of race is filtered through the lens of gender (e.g., Collins, 1999; Shields, 2008) such that the social identity “Black male” becomes the unit of investigation rather than the separate identities of “Black” and “male.” An alternative interpretation, however, is that “Black,” “male,” and “Black male” each repre sents three distinct (though perhaps related) social identities. For example, Ghavami and Peplau (2013) found that college students associate qualitatively distinct stereotypes to each “race,” “gender,” and “Race x Gender” category. Their findings support the idea that racial identity and gender identity may be related but distinct identity constructs.To this point, Settles (2006) found that Black women reported equivalent levels of racial and gen der centrality and that levels of centrality were pos itively correlated. Turner and Brown (2007) found that Black children (but not White children) rated race as more important than gender. These findings also support the idea that identities are distinct but related, and that relative importance of each iden tity may depend on developmental stage and race and gender demographics of the sample. For Black adolescent males, a demographic group for whom the intersection of race and gender is particularly salient (e.g., Dottolo & Stewart, 2008; Ghavami & Peplau, 2013), race and gender may rank as highly important and be strongly correlated. In fact, Ghav ami and Peplau (2013) found the highest level of correspondence in stereotype descriptions for the categories “Black” and “Black male.” A similar argument could be made for private regard—how one feels about being Black (racial private regard) ought to be related, in part, to how he feels about being male (gender private regard). Although theRacial and Gender Identity 409cultural construction of social identity categories are linked, whether young Black males also experience their racial and gender identities in this same over lapping fashion is an empirical question.Identity and AdjustmentThere is extensive empirical evidence that racial identity is related to the psychological and aca demic lives of Black youth (e.g., Phinney, 1990; Smalls, White, Chavous, & Sellers, 2007; Syed et al., 2011; Yip et al., 2006). For example, Black youth with a more positive evaluation of their racial group (private regard) report higher levels of self esteem and lower levels of depressive symptoms (Phinney, 1990; Sellers et al. 2006; Yip et al., 2006). Similarly, stronger racial group ties (centrality and regard) support academic engagement and perfor mance (Altschul, Oyserman, & Bybee, 2006; Smalls et al., 2007). In other words, the stronger connection youth have with their racial group, the better they feel about themselves and the better they do in school-related activities.For gender identity, there is similar evidence. Youth who view gender as important to the self and feel positively about their gender group score higher on measures of self-esteem and exhibit fewer depressive symptoms (e.g., Carver et al., 2003; Egan & Perry, 2001). In a review of gender identity research, Perry and Pauletti (2011) sum marize findings that suggest that gender identities marked by positive regard and centrality relate to positive youth outcomes. Most of these gender findings, however, are based on samples of chil dren and early adolescents, and very few samples include Black youth. Corby et al.’s (2007) study of gender identity among ethnic minority preado lescents (fifth-graders) actually found no associa tion between gender identity measures and well being for Black children. They conclude that racial identity may impact well-being but not gender identity. Still, there is evidence that Black boys who hold less rigid gender-type beliefs fare better on indices of well-being and score higher on school engagement and achievement (Cunningham et al., 2013; Santos et al., 2011; Santos et al., 2013). Thus, like racial identity, gender identity may play an important role in the adjustment outcomes of Black adolescents.The limitation of these studies on the linksbetween racial and gender identity and adjustment is that they are isolated so we are unable to deter mine the effect of both identities on adjustment. Each identity may play a specialized role. ForThis content downloaded from on Sat, 19 Mar 2022 00:45:01 UTCAll use subject to https://about.jstor.org/terms410 Rogers, Scott, and Wayexample, given a history of race-based discrimina tion and unequal education, a strong, positive racial identity that buffers negative expectations may be more relevant than gender identity in an academic setting. In contrast, male and academic success is constructed as congruent; therefore, identifying strongly with gender might also support academic identification or success. At the intersection of raceand gender, Black males are negatively stereotyped in academic domains (Ferguson, 2000; Polite & Davis, 1999); as such, one might observe an inter active effect of racial and gender identity on aca demic outcomes. From the perspective of social identity theory (Tajfel & Turner, 1986) in which social group affiliation supports global well-being, both racial and gender identity (specifically private regard) should be linked to positive self-esteem. Based on existing studies, it is unclear whether each identity functions uniquely (i.e., explains unique variance in the outcome) or, as Corby et al. (2007) suggest, racial identity functions in this capacity for Black youth. Taking an intersectional ity approach to the study of identity and adjust ment fills these gaps and advances our understanding of identity processes.Current StudyThe current study sought to extend our under standing of adolescent identity development by investigating the intersections of racial and gender identity and patterns of change over time during middle adolescence. Three questions guided our analysis: (a) What is the association between and dif ferences in levels of racial identity (centrality and pri vate regard) and gender identity (centrality and private regard)? (b) How do racial identity and gen der identity (centrality and private regard) change over time and are these patterns associated with each other? (c) Do racial identity and gender identity pre dict psychological well-being (self-esteem, depressive symptoms) and academic adjustment (academic engagement, school value), independently and jointly, concurrently and over time? We expected lev els of racial and gender identity to be equivalent and positively correlated (e.g., Settles, 2006), and antici pated a slight increase in both identities over time (e.g., Hill & Lynch, 1983; Pahl & Way, 2006). We also expected both racial identity and gender identity to predict better adjustment outcomes independently (Perry & Pauletti, 2011; Quintana, 2007), and explored whether the interaction between racial and gender identity contributed above and beyond the main effects.MethodData were drawn from the first author’s dissertation, a longitudinal mixed-method study of identity development in an all-Black male high school. Employing a within-group design, the study sought to capture the variability in the identity experiences of Black males, a demographic group that is often cast as monolithic. In contrast to a comparative approach, which functions to diminish within group variation and emphasize group differences (e.g., differences between Black and White males), a within-group focus can highlight strengths rather than deficits, and push the conversation beyond the surface of group differences to the nuances of iden tity processes.Data, including observations and field notes as well as surveys and interviews with students, teachers, and administrators, were collected across 2 academic years. Wave 1 was collected the sum mer prior to the start of 9th grade (August 2008), Wave 2 at the end of 9th grade (May 2009), and Wave 3 at the end of 10th grade (May 2010). Here we the report survey data.Research ContextData were collected at Black Male Charter School(BMCS), an all-Black, all-male high school located in a Midwest urban center. Black Male CharterSchool is a public charter school, with free tuition and lottery admission. (Note: Students pay a $200 school uniform fee.) At the time of data collection, the school was in its 3rd year of operation, serving approximately 400 students in Grades 9 through 11, and grappling with an overflow of students on their waiting list. (Note: BMCS has since opened two more campuses to accommodate the demand.) The leaders describe BMCS as a space designed to provide its students with “positive counterimages of successful Black men” and give them a “vision for college.” They refer to the students as “scholars” and college graduation (not merely acceptance or attendance) defines the school’s primary mission.ProceduresDuring summer orientation at BMCS, the first author gave a brief invitation speech to the incom ing freshmen class (N = 225) about the study. An invitation letter was also mailed to all enrolled students’ homes along with parental consent forms describing the project as a research study about “what it means to be a young Black male in today’sThis content downloaded from on Sat, 19 Mar 2022 00:45:01 UTCAll use subject to https://about.jstor.org/termssociety.” Additional recruitment materials were dis tributed at school and students returned signed parent consent forms to their homeroom teachers or main office. Surveys were administered in home room classes at the end of the school day and took 30-40 min to complete. Participation was com pletely voluntary. Students were given candy and allowed computer time when finished.ParticipantsOne hundred ninety-five Black male students consented to participate, 12 of whom were excluded from the analyses because they did not provide data beyond names and birthdates. The final sam ple (N = 183) includes participants who provided data on the primary study variables during at least one wave of data collection. Demographics for the sample are presented in Table 1. All participants were Black males and ranged in age from 13 to 16 years (MTimei = 14.2, SD = .56). Approximately half (58%) of the students lived in single-parent homes, 34% lived with both parents, and a small percentage was under nonparental care (8%), living with a relative (e.g., grandmother, aunt/uncle, older sibling) or foster parent. Based on school records, these students come from predominantly from low-income backgrounds and live within aRacial and Gender Identity 4114-mile radius of the school campus. According to student reports, the majority of their mothers (75%) and fathers (66%) were employed during the first wave of data collection. At this same period, 22.5% of the sample had at least one parent who had gone to college and 52.5% had a parent who completed some college or earned a 4-year degree. Of the remaining, 18% had earned a high school diploma and 7% had not completed high school.MeasuresIdentityRacial identity was measured using two sub scales from the Multidimensional Inventory of Black Identity-Teen (Scottham, Sellers, & Nguyen, 2008): Centrality measures the importance of race to the self (“If I were to describe myself one of the first things I would say is that I’m Black”), and Private Regard measures the personal evaluation of race (“I’m happy that I’m Black”). Each subscale consists of three items rated on 4-point Likert scale of agreement. Parallel items were adapted for gen der identity (e.g., “If I were to describe myself one of the first things I would say is that I’m a male”). This method of parallel items mirrors prior work on multiple identities (Settles, 2006). On the raceTable 1Descriptives for Sample Demographics and Primary Study Variables by WaveOriginal data sets Imputed data setsWave 1 Wave 2 Wave 3 Wave 1 Wave 2 Wave 3Age 14.20 (.56) 14.90 (.68) 15.90 (.70) 14.17 (.04) 14.87 (.05) 15.87 (.05) Family structureNonparental care 8.2% — — 10.37% — —Single parent 57.70% — — 56.47% — —Dual parent 34.10% — — 33.16% — — Parent educationNo high school 7.00% — — 1.78% — —High school 18.00% — — 14.45% — —Some college 22.50% — — 30.40% — —College degree 52.50% — — 53.37% — —Primary variablesRacial centrality 2.95 (.68) 3.03 (.77) 3.12 (.70) 2.96 (.05) 3.02 (.08) 3.08 (.07) Racial private regard 3.54 (.60) 3.23 (.77) 3.46 (.67) 3.54 (.05) 3.23 (.08) 3.39 (.07) Gender centrality 3.52 (.55) 3.16 (.61) 3.38 (.69) 3.49 (.05) 3.14 (.05) 3.53 (.06) Gender private regard 3.78 (.45) 3.53 (.69) 3.66 (.61) 3.76 (.04) 3.51 (.06) 3.65 (.05) Self-esteem 3.37 (.49) 3.08 (.55) 3.31 (.51) 3.36 (.04) 3.06 (.07) 3.28 (.05) Depressive symptoms 1.17 (.24) 1.24 (.40) 1.29 (.38) 1.18 (.02) 1.24 (.04) 1.31 (.05) Academic engagement 3.20 (.38) 3.20 (.41) 3.23 (.56) 3.21 (.03) 3.20 (.03) 3.19 (.06) School value 3.50 (.39) 3.56 (.40) 3.32 (.65) 3.50 (.03) 3.56 (.03) 3.28 (.06)This content downloaded from on Sat, 19 Mar 2022 00:45:01 UTCAll use subject to https://about.jstor.org/terms412 Rogers, Scott, and Waymeasure, reliability for racial centrality was low to ade quate (as = .55, .81, .74) and strong for racial private regard (as = .81, .91, .85). Reliability on the gender measure was marginal to adequate for gender central ity (as = .64, .67, .76) and strong for gender private regard (as = .84, .91, .89). The lower alpha levels in some waves are consistent with prior research on the construct validity of the racial centrality measure (as = .55; see Scottham et al., 2008), and may reflect the limited number of items or suggest some general instability in the construct of racial centrality in these middle years of adolescence.Psychological Well-BeingMeasures of psychological well-being included self-esteem and depressive symptoms. Global self esteem was measured on the 10-item Rosenberg Self-Esteem scale (“I am satisfied with myself”), which uses a 4-point Likert scale of agreement (Rosenberg, 1965). Internal consistency for the scale was strong for each wave (as = .78, .80, .77). The Children’s Depression Inventory-Short Form (CDI S) is a 10-item scale that measures nonclinical depressive symptoms (Kovacs, 1992). The CDI-S is measured on a 3-point Likert scale ranging from 0 (never) to 2 (always), and includes emotional (sad ness), cognitive (negative self-attitudes), and behav ioral (lack of energy) components. Reliability was strong for each wave of measurement (as = .74, .82, .85).Academic AdjustmentMeasures of academic adjustment included engagement and school value. Engagement (e.g., “When I’m working on something I care about, nothing can distract me”) was measured on an eight-item Likert scale of agreement where high scores indicate higher levels of engagement (Suarez Orozco, Suarez-Orozco, & Todorova, 2009). Reli ability was adequate (as = .65, .77, .89). School value was measured using a subscale from Walton and Cohen’s (2005) measure of social fit. This five item subscale measures students’ perceived impor tance of school (e.g., “It is important to me to do well in school”). Internal consistency was adequate (as = .69, .85, .86). For analyses that use the aca demic variables, parental education was included as a control variable given the known association between parent education and child academic out comes (Sirin, 2005). Parental education was taken from students’ reports of the highest level of educa tion of the person(s) living with them.Data AnalysisFirst, analyses were conducted to assess missing data. For the primary outcome variables, the per centage of complete cases across the three waves ranged from 28% to 44%, with most of the missing data occurring in Waves 2 and 3. We compared outcomes and predictors across waves to assess the impact of the missing-person-period cases. Only one demographic variable, parent education, was noticeably different across available cases between waves; specifically, the percentage of the sample whose parents did not have a high school diploma was significantly lower in Wave 3 (4%) compared to prior waves (7% and 8.5%, respectively). Part of our strategy to address missing data was to reduce potential bias due to attrition.Missing data were addressed using multiple imputation (MI; Little & Rubin, 1987). In MI, a series of predicted values are imputed for each missing value generating multiple complete data sets. For our analyses, 50 imputations were generated using the multi-imputation program MI in STATA (StataCorp, 2013). Specifically, we primarily used a series of “chained regressions” in which variables are imputed in order from least to most missing (Royston, 2004; van Buuren, 2012). A few predictors were ordinal or nominal, and as such were handled slightly differ ently, but the basic paradigm of chained regressions still applies. All analyses were conducted on both the original and the imputed data sets, the latter’s esti mates being summarized using the methods described by Rubin and Schenker (1986). Both origi nal and Mi-based analyses rely on the assumption that the variables are missing at random (MAR; see Little & Rubin, 1987). While the MAR assumption cannot be assessed directly, MAR is threatened when there are insufficient predictors to reliably model the missing mechanism and the expected value of the missing variables. In our data, we can use parent education from early waves to model missing out comes for subjects with similar characteristics in sub sequent waves. Missing values for parent education were minimal (6.6%) and these were imputed as well, during the chained iterations. The advantage to the MI approach is that the uncertainty associated with missing data is accounted for in the estimation proce dure. Imputed data were compared to nonimputed distribution, and matched very well.Growth ModelingGrowth models—extensions of regression models to allow for subject-specific differences in trajectoriesThis content downloaded from on Sat, 19 Mar 2022 00:45:01 UTCAll use subject to https://about.jstor.org/terms(see Raudenbush & Bryk, 2002; Singer & Willett, 2003)—were estimated using STATA statistical software. We report findings from models based on imputed data in this article, as it utilizes as much information in the data as possible. For each out come, we first estimated a baseline model, which includes the outcome variable and time (dummy coded: Wave 2 and Wave 3). From the baseline model we calculated the intraclass correlation (ICC), P (P =