Coleman argued that high school students face a zero-sum situation in which time and energy devoted to sports comes at the cost of other activities, such as academic performance. However, subsequent research has found that high school sports participation is associated with several positive academic outcomes such as higher GPA, fewer disciplinary referrals, lower absenteeism and dropout rates, higher college aspirations and attendance, and stronger commitment to school (Eccles & Barber, 1999; Marsh, 1993; Sabo, Melnick, & Vanfossen, 1989; Snyder & Spreitzer, 1992; Whitley, 1998). Upon finding that athletic participation had significant positive effects on 14 of 22 senior and postsecondary educational outcomes and no negative impact on the others, Marsh concluded that, contrary to Coleman’s zero-sum expectations, “participation in sport apparently adds to–not detracts from–time, energy, and commitment to academic pursuits” (1993:35).

It has also become increasingly apparent, however, that the apparent protective effects of sports with respect to academic outcomes are neither universal nor indisputably causal in nature. Involvement in athletics, like academics, is subject to strong selection effects; for example, athletes disproportionately hail from socioeconomically advantaged backgrounds (Crosnoe, 2002). Moreover, students with better grades tend to self-select into high school sports programs (Sabo et al., 1989), although some studies have also identified a positive academic trajectory for athletes over time, relative to their nonathletic peers (Crosnoe, 2002; Eccles & Barber, 1999; Marsh & Kleitman, 2003). The reasons for this linkage are likely to be, at least to some degree, pragmatic; that is, student-athletes who have disciplinary problems or who perform poorly in the classroom risk being dropped from the team. It is likely that institutional eligibility requirements accounted for Laughlin’s (1978) finding that athletes had lower rates of absenteeism and higher GPAs during their playing seasons than out of season. Similarly, Larson (1994) found no support for the supposition that athletic participation reduces delinquent behavior, concluding instead that delinquent adolescents were more likely to voluntarily self-select out of sports.

Several researchers have noted that the relationships between adolescent athletic involvement and academic outcomes cross racial and gender lines (e.g., Marsh, 1993; Whitley, 1998). However, most agree that the effects are not identical or equally strong across these demographic categories. For example, Wells and Picou (1980) found that athletic participation was consistently associated with socialization for educational achievement (e.g., educational ambition, better academic performance, or association with a college-oriented crowd) only for white male adolescents; the link was weaker for white females and black males, and weakest of all for black females. Sabo and his colleagues (1989) reached a similar conclusion with respect to post-high school educational mobility. Although black and Hispanic athletes reported better grades and greater involvement in school activities than their nonathletic peers, the effects were more short-lived than for whites, for whom high school sports participation was associated with higher rates of college attendance and completion. Finally, Crosnoe (2002) found that gender and athlete status both predicted initial high school academic performance, with female athletes reporting the highest GPAs and male nonathletes reporting the lowest. Over the course of the high school career, both gender and athlete status had protective effects on academic achievement; only male nonathletes saw overall declines in achievement over time.

Although a small number of studies have explored the link between peer crowd identification as a “jock” and school outcomes (Brown, Mounts, Lamborn, & Steinberg, 1993; Clasen & Brown, 1985; Urberg, Degirmencioglu, Tolson, & Halliday-Scher, 2000), few researchers have attempted to disaggregate the academic effects of objective athletic participation (what one does) from the effects of subjective athletic identity (whom one perceives oneself to be). The notable exceptions were Eccles and her colleagues (Eccles & Barber, 1999; Barber, Eccles, & Stone, 2001), who found that both high school athletic participation and identification as a “jock” predicted higher levels of post-secondary educational attainment. However, the researchers noted that educational achievement was more closely associated with athlete status, whereas jock identity was more closely related to positive psychological adjustment. These differences suggest that measures of “athlete status” and “jock identity” tap distinct, if overlapping, constructs.

The fact that high school athletes tend to perform better in school than their peers is well established, but the parameters of this relationship remain obscured in several important respects. The present analysis seeks to untangle the gender- and race-specific linkages between adolescent athletic involvement and subsequent academic outcomes. Specifically, we address three research questions: (1) Is the relationship stable over time (i.e., are athletic involvement and academic performance positively linked when measured several years apart)? (2) Does the relationship differ by dimension of athletic involvement (athlete status vs. jock identity)? (3) Does the relationship operate differently by gender and/or race?

After weighting to compensate for the oversampling of black families, the characteristics of the overall sample closely matched census demographics in the area. After excluding respondents who had dropped out of school by the third wave of data collection, our wave-1 sample included 586 adolescent subjects ranging in age from 14 to 19, with a mean age of 14.4 years. Approximately 55 percent of the respondents were female. Black adolescents made up 30 percent of the unweighted sample and 15 percent of the weighted sample.

The analysis included four sociodemographic measures, two dimensions of athletic involvement, and two academic outcomes. All independent measures were derived from wave 1 data; all dependent variables were derived from wave 3. Gender was coded 0=male, 1=female. Race was coded into two categories: black (=1) and white/other (=0). Because they collectively constituted too small a subsample to analyze independently (n=13), respondents identified as Asian American, Hispanic, Native American, or “other race” were coded with white subjects.

Family socioeconomic status was assessed using a measure that combined family income, mother’s education, and father’s education (alpha =.74). Family income was reported by the adolescent’s parents; if only one parent was available for the interview, the other provided an estimate of the income of the absent parent. Family income categories included (1) $0–$14,999; (2) $15,000–$34,999; (3) $35,000–$49,999; and (4) $50,000 or more. Parental education categories included (1) 0–11 years; (2) 12 years; (3) 13–15 years; and (4) 16 or more years. We calculated the mean of these three measures in order to derive a comprehensive indicator of family socioeconomic status, in which SES ranged from a low of 1.00 to a high of 4.00.

Gender, race, age, and socioeconomic status are collectively powerful predictors of adolescent academic performance. However, they cannot adequately account for the selection process identified previously; namely, the tendency for “good” students to self-select into athletic programs. In order to minimize the impact of this effect, a wave-one measure of each of the two school outcome variables (GPA and school misconduct) was also included in analyses predicting the same outcome variables at wave three. The inclusion of these measures as wave-one controls means that in effect, the analysis assesses the impact of athletic involvement on changes in GPA and school misconduct between waves one and three.

Athletic involvement was measured in two ways. First, to measure school athlete status, respondents were asked about participation in a list of “school activities.” One of these activities was designated as “sports (football, basketball, baseball, swimming, track, etc.)”, with responses coded 0=no, 1=yes. Adolescents who responded affirmatively to this question were coded positively for athlete status. A second measure of athletic involvement emphasized respondent self-perception rather than behavior. Respondents were asked, “Teenagers sometimes characterize one another on the basis of their attitudes toward school, clothes, music, partying, and so forth. Some people give names to these types, such as jocks, preps, air heads, burnouts and so forth. How well does each type fit you?” Those responding that the “jock” label fit them “very well” or “somewhat” were coded as having a jock identity; those who responded “a little,” “not at all,” or “never heard of this group” were coded as not having a jock identity.

Student self-reports of grades and school misconduct in the past year served as the dependent variables in this analysis. To estimate GPA, respondents were asked their grade point average over the year prior to the survey, with responses ranging from 1 (mostly Fs, <65) to 7 (mostly As, 90–100). School misconduct (alpha =.64) summed responses to four questions about skipping school, cutting a class, having someone called to the school from home for disciplinary reasons, and being sent to the principal’s office. Categorical responses were recoded to midpoint values on each question. For the question about skipping school, responses included 0 (never), 1 (once), 2.5 (2–3 times), 4.5 (4–5 times), 7.5 (6–9 times), and 15 (10+ times). For each of the remaining questions, the available responses were 0 (never), 1 (once), 2.5 (2–3 days), 6.5 (4–9 days), 14.5 (10–19 days), 29.5 (20–39 days), and 48 (40+ days). Summing responses to the four school misconduct questions yielded a potential response range from 0 occasions of misconduct to 159 occasions of misconduct. Because many students reported no occasions of misconduct, we performed a log transformation on this variable to approximate a more normal distribution (skewness = 2.93 before transformation, −0.69 afterward; kurtosis = 10.27 before transformation, −1.056 afterward). Descriptive statistics (Table 1) employed the untransformed version of the school misconduct variable, for ease of interpretation; however, all multivariate analyses were conducted using the log transformed variable.

One-way analysis of variance tests were performed to test the significance of gender and race differences in reported grades and school misconduct. As shown in Table 1, gender and racial variations in these school outcomes were in keeping with prevailing patterns. Boys reported lower grades and more school misconduct than girls. Boys were more likely to skip school (p<.01), have someone called from home (p<.01), and be sent to the principal’s office (p<.001) than girls. Blacks reported lower grades than whites (p<.001). There were no significant race differences in school misconduct overall, although when the components of the scale were disaggregated, blacks reported fewer illegal absences but more disciplinary referrals than whites. Specifically, blacks were marginally less likely to cut class (p<.10), but more likely to have someone called from home (p<.01) and marginally more likely to be sent to the principal (p<.10) than whites.

Table 1 also shows self-reported athletic involvement. Consonant with existing evidence that athletic participation is the single most popular school-sponsored extracurricular activity among adolescents regardless of gender, race, and ethnicity (e.g., U.S. Department of Education, 1995; Eccles & Barber, 1999), adolescents in our sample reported high rates of athletic involvement at wave 1. Nearly two thirds (65%) reported participating in school sports. More than one third (35%) identified themselves as “jocks.” In general, athletic involvement was higher for boys than girls. Black and white respondents reported comparable rates of participation in school sports (65% and 65% respectively), but self-identification as a jock was markedly more prevalent among whites (37%) than blacks (22%).

Comparisons of athletes and nonathletes (data not shown) revealed several key gender differences, with female (but not male) athletes reporting a higher GPA at wave 3. Female athletes also engaged in significantly more wave-3 school misconduct than female nonathletes, particularly with respect to cutting class. In contrast, male athletes engaged in less misconduct than male nonathletes; there were significant differences on three of the four misconduct measures (skipping school, someone called from home, and sent to the principal’s office).

Comparisons of jocks and nonjocks (data not shown) revealed significant differences with respect to GPA but not school misconduct. Female jocks reported higher GPA at wave 1 (concurrent with measurement of jock identity) than female nonjocks, but lost this advantage by wave 3. (Waves 1 and 3 may be understood as roughly corresponding to median grade 9 and median grade 11, although the sample actually spanned a more than 4-year age distribution at any given point in time.) Black jocks reported significantly lower wave-3 GPAs than black nonjocks.

Multiple regression analyses were employed to predict adolescent GPA and school misconduct in wave 3. In order to control for the potentially confounding effects of age and socioeconomic status in predicting the race- and gender-specific relationships between athletic involvement and school outcomes, two regression equations were modeled for each academic outcome variable. Model One included main effects of gender, race, age, socioeconomic status, jock identity, and athlete status at wave 1, as well as a wave-1 measure of the academic outcome in question. In the second model, 2-way interactions of each athletic involvement variable with race and with gender were added to the equation. A third model including three-way product terms for race, gender, and each of the athletic involvement variables was tested but discarded due to the unreliability of any results based on such small cell sizes. Because the effects of race were controlled for in each of these models, all multivariate analyses were performed on the unweighted sample (Winship & Radbill, 1994). Results are presented in Table 2. Where significant cross-product terms were found among gender, race, and/or athletic involvement, additional group-specific analyses were conducted to probe the interactions.

Considerable scholarly attention has been devoted to the relationship between adolescent athletic participation and educational outcomes such as academic achievement, absenteeism, misconduct, and school attachment. Most studies can be clearly located on either side of a longstanding theoretical divide, favoring or opposing interscholastic sports (see Braddock, 1981; Marsh, 1992 for review of the debate). Developmental theorists argue that athletic participation contributes to better academic performance by developing skills, habits, and values transferrable to the classroom, integrating students into a prosocial network of adults and peers, providing tangible incentives to stay in school and get good grades, and increasing commitment to the school (Marsh & Kleitman, 2003; Snyder & Spreitzer, 1990). Zero-sum theorists counter that the resources adolescents devote to competing pursuits are finite; when young athletes’ time and energy, as well as the resources of their schools and communities, are diverted from the classroom to extracurricular activities such as sports, academic objectives are undermined (Coleman, 1961; Hauser & Lueptow, 1978).

At the high school level, links between sports involvement and academic outcomes have for the most part been largely positive; student athletes tend to have higher grades, less absenteeism, fewer discipline referrals, stronger internal locus of control, and better odds of aspiring to–and completing–a college education (Eccles & Barber, 1999; Fejgin, 1994; Marsh, 1993; Videon, 2002). However, this body of findings has been plagued by small effect sizes and, in some cases, puzzling inconsistencies. For example, Hauser and Lueptow (1978) found that, although athletes’ grades improved over the course of their high school careers, the gains were smaller than those of nonathletes. More recently, Hanson and Kraus (1998) observed that athletic participation had a positive effect on the science-related experiences of white female adolescents but the opposite effect for black female adolescents. Fisher, Juszczak, & Friedman (1996) found no association at all between athletics and the academic performance of inner-city adolescents.

One reason for weak or inconsistent findings may be that the relationships between athletic participation and positive school outcomes are selective rather than causal; that is, those adolescents who do well in high school are also those who choose to participate in school sports (Barron, Ewing, & Waddell, 2000). Studies that take into account background characteristics tend to find weaker correlations between sports participation and GPA (Holland & Andre, 1987). There is also considerable uncertainty regarding the shelf life of athletic effects. Strong positive associations between sports participation and concurrent academic performance may dissipate or even turn negative if predictor and outcome variables are measured several years apart.

A second explanation is that the relationship between sports participation and school outcomes is not monolithic. Some studies have examined the role of gender (Hanson & Kraus, 1998; Videon, 2002) or race (Melnick, Sabo, & Vanfossen, 1992; Snyder & Spreitzer, 1990; Wells & Picou, 1980) as potential moderators of the link between athletic participation and academics. Eide and Ronan (2001) found disparate effects of high school sports participation on educational attainment for white males (negative impact), white females and black males (positive impact), and black females and Hispanics of either gender (no significant impact). However, except for Eccles and her colleagues (Eccles & Barber, 1999; Barber et al., 2001), most previous research on academic performance has not addressed the distinction between what athletes do (e.g., participation in sports) and how they perceive themselves (e.g., identification as a “jock”). The present study has examined how gender and race interact with dimensions of athletic involvement to predict academic outcomes.

Several tentative conclusions may be derived from careful examination of these findings. First, the lagged effects of athletic involvement on academic outcomes several years later were far less favorable than contemporaneous, cross-sectional effects widely documented in the extant literature. Neither jock identity nor athlete status was unequivocally associated with improved subsequent academic performance. Developmental theorists posit that participation in organized sports enhances attachment to school, provides constructive guidance and adult supervision, reinforces prosocial values, and teaches skills that spill over into the classroom. In light of our results, however, it may be that the short-term athletic benefits identified by developmental researchers derive more from the immediate context of participation (e.g., team rules about absences or minimum GPA requirements) than from longer-term developmental processes. Another powerful influence on positive school outcomes lies in the preexisting characteristics of adolescents who self-select into athletic programs.

Second, the potentially beneficial effects of athletic involvement appear to be contingent on the dimension of involvement under consideration. For example, female athletes enjoyed a small but significant GPA advantage over female nonathletes; but female and black respondents who identified themselves as jocks reported lower GPAs than those who did not. That a jock identity is not conducive to enhanced academic performance for these adolescents is not entirely surprising. A limited body of previous research has documented troubling links between jock identity and such problem behaviors as heavy drinking and binge drinking (Ashmore, Del Boca, & Beebe, 2002; Miller, Hoffman, Barnes, Farrell, Sabo, & Melnick, 2003), violence and bullying (Miller, Melnick, Farrell, Sabo, & Barnes, forthcoming; Wilson, 2002), and sexual risk-taking (Miller, Farrell, Barnes, Melnick, & Sabo, in press). However, to date, this emerging picture has seemed to suggest a “toxic jock” effect that is strongest for white boys. In the present analysis, conversely, white boys actually proved the exception; a jock identity had no significant adverse effect on grades for this subgroup. This finding was unexpected. It may be that where academic performance is concerned, the jock label constitutes less of a departure from the norm for white boys than it does for female or black adolescents, thus weakening its negative impact on their educational outcomes.¹

Third, the relationships among athlete status, gender, and adolescent school misconduct were inconsistent. Participation in sports was associated with marginally more misconduct for girls and marginally less misconduct for boys; although neither finding alone was statistically remarkable, the gender difference was statistically significant. This unexpected finding has no clear precedent; in fact, the few previous studies that have directly examined links between female athletic participation and school misconduct found that female high school athletes were less likely to break school rules and regulations than their nonathlete peers (Buhrmann, 1977; Buhrmann & Bratton, 1978; Fejgin, 1994). Assuming that our results can be replicated in future studies, they may indicate the interplay of several developmental and psychosocial processes. First, school-based sports place the participant at the center of a social network that reinforces commitment to the school (Marsh, 1992, 1993) and mandates conformity to conventional expectations. Because nonconformity with school and team norms may result in suspension or even expulsion from the team, misconduct thus becomes a criterion for selection out of sports. This combination of restraints helps explain the negative relationship between athletic participation and school misconduct for boys.

However, the processes that may account for reduced misconduct by male athletes do not explain why female athletes actually tended toward more frequent misconduct than female nonathletes. We speculate that the institutional fit between adolescent girls and the athlete role may be less comfortable than it is for boys. Mainstream acceptance of girls as bona fide athletes is a relatively new phenomenon. Traditional, hegemonic cultural scripts for feminine behavior–what Connell (1995) described as “emphasized femininity”–left little room for the dirt, sweat, and overt physical competition of organized sports. Even though such behavior has grown more acceptable in recent years, it may well be that female athletes still experience more role conflict between the demands of the playing field and the demands of the classroom than their male counterparts do, one manifestation of that conflict being school misconduct. Ironically, it may also be the case that, whereas boys who get into trouble are selectively filtered out of sports, girls who get into trouble are selectively filtered into sports because they are disproportionately amenable to violating conventional gender norms regarding assertiveness, competition, and physicality.

In the absence of more data, this interpretation remains purely speculative. The present study is however consistent with previous findings that other, nonacademic concommitants of school athletic participation and jock identity differ by gender (e.g., Miller, Sabo, Farrell, Barnes, & Melnick, 1998; Sabo, Miller, Farrell, Melnick, & Barnes, 1999) and race (e.g., Miller et al., in press). Future researchers will need to address how gender and/or racial differences specifically impact the relationship between athletic involvement and academic outcomes.

This study also confirms the importance of distinguishing among dimensions of athletic involvement. It also highlights the need to develop better instruments for doing so. In particular, jock identity is a more nebulous construct than most other sport-related measures. Unlike athlete status or frequency of athletic activity, which are subject to faulty recollection but nevertheless lend themselves to objective assessment, “jock identity” relies on the adolescent’s more subjective, self-reported perception. Furthermore, although there are distinct and mutually exclusive literatures which examine the behavioral implications of athletic participation (e.g., Crosnoe, 2002; Marsh & Kleitman, 2003; Snyder & Spreitzer, 1990), athletic identity (e.g., Brewer, Van Raalte, & Linder, 1993; Horton & Mack, 2000; Webb, Nasco, Riley, & Headrick, 1998), and affiliation with the “jock” peer crowd (e.g., Brown, Eicher, & Petrie, 1986; Eckert, 1989; La Greca, Prinstein, & Fetter, 2001), little formal theorization (and almost no empirical research) has been devoted to understanding the differences between an “athlete” and a “jock.” The few researchers who have explicitly drawn this distinction have generally found that the two constructs overlap less than might be expected (e.g., Barber et al., 2001; Miller et al., in press).

Athletes and jocks are not the same. Athletes are valorized in popular culture; in contrast, the label of jock is perceived by many as a derogatory term that connotes ignorance (e.g., “dumb jock”). Together they represent the two faces of sport: one ascetic and disciplined, the other gregarious and risk-oriented. In this study, we have examined some of the contrasting implications of these distinct and often conflicting constructs. However, our measure of jock identity–while conventionally used in peer crowd research–did not directly probe the subjects’ interpretation of the jock label, nor its confirmation by others. We were thus unable to explore how or if the meanings conventionally assigned to this label might differ between genders or races, or indeed even within a single gender (Miller et al., in press; Miller et al., forthcoming; Pascoe, 2003). Future data collection might profitably frame jock identity by developing a multi-item indicator comparable to Brewer et al.’s 1993 Athletic Identity Measurement Scale. Likewise, our dichotomous measure of objective athlete status could not capture nuances such as the intensity, frequency, or type of athletic activity, all of which might condition the relationship between sports participation and educational outcomes.