obese students obtain lower school grades but equivalent test scores

International Journal of Obesity (2013) 37, 40 -- 46 & 2013 Macmillan Publishers Limited All rights reserved 0307-0565/13 www.nature.com/ijo

PEDIATRIC ORIGINAL ARTICLE

Just as smart but not as successful: obese students obtain lower school grades but equivalent test scores to nonobese students C MacCann1 and RD Roberts2 OBJECTIVE: The obesity epidemic in industrialized nations has important implications for education, as research demonstrates lower academic achievement among obese students. The current paper compares the test scores and school grades of obese, overweight and normal-weight students in secondary and further education, controlling for demographic variables, personality, ability and well-being confounds. PARTICIPANTS: This study included 383 eighth-grade students (49% female; study 1) and 1036 students from 24 community colleges and universities (64% female, study 2), both drawn from five regions across the United States. MEASURES: In study 1, body mass index (BMI) was calculated using self-reports and parent reports of weight and height. In study 2, BMI was calculated from self-reported weight and height only. Both samples completed age-appropriate assessments of mathematics, vocabulary and the personality trait conscientiousness. Eighth-grade students additionally completed a measure of life satisfaction, with both self-reports and parent reports of their grades from the previous semester also obtained. Higher education students additionally completed measures of positive and negative affect, and self-reported their grades and college entrance scores. RESULTS: Obese students receive significantly lower grades in middle school (d ¼ 0.83), community college (d ¼ 0.34) and university (d ¼ 0.36), but show no statistically significant differences in intelligence or achievement test scores. Even after controlling for demographic variables, intelligence, personality and well-being, obese students obtain significantly lower grades than normal-weight students in the eighth grade (d ¼ 0.39), community college (d ¼ 0.42) and university (d ¼ 0.31). CONCLUSION: Lower grades may reflect peer and teacher prejudice against overweight and obese students rather than lack of ability among these students. International Journal of Obesity (2013) 37, 40--46; doi:10.1038/ijo.2012.47; published online 24 April 2012 Keywords: academic achievement; conscientiousness; well-being; prejudice

INTRODUCTION Claims from multiple sources indicate an ‘obesity epidemic’ among adolescents of industrialized nations over the last decade,1 with the proportion of overweight 12 -- 19 year olds in the United States increasing from 5.0% in 1976 -- 1980 to 16.3% in 2003 -- 2006.2 Aside from the obvious negative health consequences, the obesity epidemic may also have important consequences for academic achievement. Research demonstrates that obese students show lower achievement at school than healthy-weight students, with some longitudinal designs in the early school grades suggesting a causal relationship.3 -- 7 A closer examination of the literature suggests that there may be different findings for different indices of achievement. There are few significant differences in mathematics or reading test scores for overweight and obese students versus nonobese students, suggesting that overweight and obese students are no less able than nonobese students.6,8,9 Despite this, obese students tend to obtain lower school grades than nonobese students.7,10,11 However, existing research has not controlled for ability test scores when examining differences in grades. Obese students may obtain lower grades either because they are less academically able (in which case, differences in grades would be accounted for by ability), or because of other systematic differences between groups unrelated to ability (for example, prejudice against heavier students by teachers, peers or parents12 -- 17). The current research

1

is the first study to control for both mathematics and vocabulary scores when examining differences in grades. In addition to ability, several other confounds may influence the relationship between obesity and school grades. First, demographic factors may confound the relationship between obesity and academic performance. Both school performance and obesity differ by socioeconomic status (SES), ethnicity and sex.2,18 Geographic region of the United States also relates to both test scores and obesity.19,20 Second, personality traits may play a causal role in both school achievement and obesity. The personality trait Conscientiousness has a strong relationship to academic achievement, is inversely related to unhealthy eating behaviors and relates to longevity through health-protection behaviors.21 -- 23 Third, emotional well-being relates to both students’ weight and their achievement. Overweight and obese adolescents tend to have lower self-esteem and experience a greater degree of negative emotions, which may affect both test scores and grades.24,25 Existing studies do not control for the contributions of ability, personality or well-being as measured in proctored psychometric tests.6,7,24 The current study addresses both this need and the need to control for ability test scores when examining the link between obesity and students’ grades. Specifically, there are two hypotheses that are expected to hold across both secondary and higher education students:

School of Psychology, University of Sydney, Sydney, New South Wales, Australia and 2Educational Testing Service, Princeton, NJ, USA. Correspondence: Dr C MacCann, School of Psychology, University of Sydney, 449 Brennan MacCallum Building (A18), Sydney, New South Wales 2006, Australia. E-mail: [email protected] Received 8 December 2011; revised 2 March 2012; accepted 4 March 2012; published online 24 April 2012

Obese students obtain lower grades C MacCann and RD Roberts

1. Test scores (mathematics, vocabulary and SAT and ACT scores) will not differ for obese versus normal-weight students after appropriately accounting for differences in demographic variables (sex, ethnicity, SES and geographical location), conscientiousness and students’ well-being. 2. Obese students will obtain significantly lower school grades than healthy-weight students, even after controlling for differences in mathematics and vocabulary scores (as well as demographic variables, conscientiousness and well-being). STUDY 1: MIDDLE-SCHOOL STUDENTS METHOD Procedure This study uses data from 383 middle-school students from five US cities (Atlanta, Chicago, Denver, Fort Lee (NJ) and Los Angeles), as well as data reported by their parents on a range of the variables of interest. Participants completed two testing sessions, 14 months apart. Vocabulary test scores and SES indices were obtained at time 1 and all other information was obtained at time 2. Students completed a proctored, computerized test battery in testing rooms. A parent of each student also completed paperand-pencil test protocol in a separate room. Each session took B1.5 h to complete and included a rest-break mid-way through the test battery. All tests and protocols were approved under the Educational Testing Service human subjects review committee and fairness review process. Participants Student sample. The student sample consisted of 383 middleschool students (49% female) beginning the eighth grade at time 2. Participants’ ages ranged from 12 to 15 years, with most aged 13 (73.6%) or 14 (22.7%) years. Reported ethnicities were 13.3% Black, 15.1% Hispanic, 68.9% White non-Hispanic and the remaining 2.3% as ‘Other’, in reasonably close alignment with the US school population.26 Parent sample. The parent sample consisted of 376 individuals (parent data were not available for 7 students), most of whom were the mother (83.3%) or the father (12.5%) of the child. Most parents were aged in their 30s (20.4%), 40s (58.5%) or 50s (17.5%). Student measures Mathematics. A total of 19 items no longer in operational use were drawn from the NAEP (National Assessment of Educational Progress) mathematics test. For example, Which of the following numbers is five million, eighty thousand? (a) 5,800,000; (b) 5,008,000; (c) 5,000,008; (d) 5,080,000; (e) 580,000. Scores are expressed in terms of percentage correct (0 to 100). Vocabulary. A subset of 18 items (6 for each of the first three levels of difficulty) was taken from the Vocabulary Levels Test.27 Each item consisted of three words that had to be matched to one of six synonyms. Scores are expressed in terms of percentage correct (0 to 100). Parent measures Household income. Parents reported their annual household income in thousands of dollars as one of seven income categories: o20; 21 -- 35; 36 -- 50; 51 -- 65; 66 -- 80; 81 -- 95; and 495. Income was available for 343 cases. Level of education. The accompanying care-giver reported their own highest obtained education level, which was classified into six categories: High school or less; some college/community college; Associate/technical degree; Bachelor’s degree; some & 2013 Macmillan Publishers Limited

graduate school; or Master’s degree or above. Parent education was available for 371 cases. Combined parent -- student variables Both parents and students reported the students’ height, weight and report-card grades, and both completed assessments (either self-report or other-report) of the students’ life satisfaction and conscientiousness. These variables were combined into a student--parent aggregate to reduce error and also for the purposes of reporting the current findings efficiently (note that when hypotheses were first tested in two separate analyses of studentreported and parent-reported variables, results were similar). The correlation between student- and parent-reported Conscientiousness was r ¼ 0.48 (Po0.01) and the correlation between student- and parent-reported life satisfaction was r ¼ 0.34 (Po0.01). Student grades. Both students and parents reported the students’ grades (from A þ , A, A and so on to Fail) from the previous semester for English, mathematics, science and social studies. Grades were converted to a 13-point scale from 0 (E or Fail) to 12 (A þ ). Grades were aggregated by taking the first principal component of the eight subject grade variables across self-reports and parent reports. The resulting ‘overall grade’ variable was thus standardized to a mean of 0, the s.d. of 1.00, with a sample size of 324. Student BMI. Both parents and students reported the student’s height (in feet and inches) and weight (in pounds), and the average of these reports was used to calculate students’ body mass index (BMI; weight in kg divided by height in m squared). Self-reports and parent reports correlated at 0.87 for height and 0.91 for weight, with self-reports 0.59 cm taller (t ¼ 2.171, P ¼ 0.031) and 0.14 kg lighter (t ¼ 0.519, P ¼ 0.604) than parent reports (similar to discrepancies between self-reports and physical measures).28 The 5th, 85th and 95th percentile values from the WHO Reference 2007 male and female growth charts of BMI by age in months were used to delineate underweight, healthy weight, overweight and obesity.29 Students’ conscientiousness. Both students and parents rated the student on 18 conscientiousness-related items (for example, ‘I make an effort’ or ‘My child makes an effort’) using a five-point scale. Items were drawn from the Industriousness and Perfectionism facets of the Comprehensive Conscientiousness Scale, as these facets showed the highest correlations with test scores and other academic outcomes.30 The mean score across self-ratings and parent ratings was used to indicate conscientiousness. Students’ life satisfaction scale. Students responded to the seven items from the students’ life satisfaction scale,31 indicating their satisfaction with life on a 6-point scale (for example, ‘My life is going well’). Parents responded to the same seven items, rephrased as other-reports (for example, ‘My child’s life is going well’). The mean score across self-ratings and parent ratings was used to indicate life satisfaction.

RESULTS AND DISCUSSION Descriptive statistics Table 1 provides internal consistency estimates (Cronbach’s a) and descriptive statistics for vocabulary, mathematics, conscientiousness and life satisfaction scores for the eighth-grade students. All measures show adequate reliability. The mathematics test is more difficult than the vocabulary test (average scores were 58% versus 87%). International Journal of Obesity (2013) 40 -- 46

41


42

Relationship between overweight and academic performance Table 2 presents the relationship between weight and achievement in two ways: (1) the correlation of BMI with mathematics, vocabulary and grade point average (GPA), and (2) the mean scores on mathematics, vocabulary and GPA for each weight category (underweight, healthy weight, overweight and obese). In both cases, this relationship is shown both before and after controlling for covariates (that is, as covariate-adjusted means and as partial correlations). The significance of the mean differences was calculated using simple contrasts comparing other groups with the healthy-weight group (using an analysis of variance procedure for unadjusted means, and an analysis of covariance procedure for the covariate-adjusted means). The effect size for the differences between weight groups was calculated using Cohen’s d.32 Table 2 also shows the power of each contrast to detect a medium-sized effect (equivalent to d ¼ 0.50, calculated based on a two-tailed t-test). Correlations. There was no significant correlation between BMI and vocabulary, either before or after accounting for covariates. There was a very small significant correlation between BMI and mathematics scores, with heavier students showing slightly poorer mathematics scores. However, this was no longer significant after controlling for covariates. Results were different for GPA: a small

Table 1.

Reliability and descriptive statistics for eighth graders

Grade Mathematics Vocabulary Conscientiousnessa Life satisfactiona

Score range

Cronbach’s a

0 -- 100 0 -- 100 1 -- 5 1 -- 6

0.72 0.85 0.88 0.86

Mean (s.d.) 0.00 57.95 86.71 3.14 4.78

(1.00) (17.85) (11.69) (0.43) (0.72)

a

Reliability estimates are calculated for the self-report items only (estimates for the parent-report items were similar).

significant correlation between BMI and GPA remained significant after controlling for covariates. Mean differences. Obese students had significantly lower mathematics and vocabulary scores than healthy-weight students, with a medium effect size in both cases. However, after controls were accounted for, the effect size for these differences was small, and no longer significant. This decrement in mathematics and vocabulary scores was not observed for overweight students: there was no significant difference between overweight and healthy-weight students on either the vocabulary or the mathematics test. Trends for GPA were different: both obese and overweight students obtained poorer GPA than healthy-weight students, with a large effect size for obese students, and a moderate effect size for overweight students. After controls were accounted for, this difference was still significant and of moderate size for both overweight and obese students. Conclusions Results from study 1 support both hypotheses. Intelligence test scores for obese students did not differ from healthy-weight students when covariates were accounted for. However, both overweight and obese students obtained lower grades than their healthy-weight counterparts. This finding cannot be attributed to lower intelligence. Nor can this result be accounted for by lower levels of conscientiousness on the part of the overweight and obese students. However, the sample size for obese students was small. Study 2 will attempt to replicate results in a larger sample, as well as examine whether this trend holds in higher education. STUDY 2: STUDENTS IN HIGHER EDUCATION METHOD Procedure All participants undertook proctored computerized testing sessions at a local site, and were reimbursed for their participation in

Table 2.

Mean test scores and grades for eighth graders by weight category, with and without controlling for confounds (d-scores calculate the effect size for healthy-weight vs other groups) Marginal means (after controls) a

Means (before controls) N

Mean (s.d.)

Mathematics Healthy weight Underweight Overweight Obese

223 15 53 31

57.71 62.11 62.96 50.42

(18.26) (18.15) (16.51) (18.92)

Vocabulary Healthy weight Underweight Overweight Obese

223 15 53 31

86.79 89.88 88.09 81.18

(11.67) (10.09) (10.20) (16.60)

Grades Healthy weight Underweight Overweight Obese

193 15 45 24

0.18 0.26 0.30 0.59

(0.89) (0.87) (1.03) (1.15)

d

Mean

0.29 0.29 0.40*

57.47 63.29 62.29 52.68

0.27 0.11 0.45*

86.47 88.69 88.50 83.50

0.09 0.52** 0.83**

0.15 0.01 0.31 0.21

Power to detect medium effect

db

0.32 0.27 0.26

0.19 0.18 0.24

0.15 0.50** 0.39*

Correlations with body mass index r

Partial r

0.09

0.02

0.12*

0.03

0.25**

0.18**

0.46 0.90 0.74

0.46 0.90 0.74

0.46 0.85 0.63

a

Sex, ethnicity, household income, parent education, geographical region, Conscientiousness and life satisfaction are entered as covariates for all analyses of covariance (ANCOVAs) and partial correlations. For grades, vocabulary and mathematics scores are also entered as covariates. bTo ensure that d-scores are comparable before and after covariate adjustments, Cohen’s d is calculated using the pooled s.d. before adjustment to compare marginal means (rather than the covariate-attenuated standard error). *The t-statistic significant at Po0.05 (based on simple contrasts with healthy weight as the reference). **The t-statistic significant at Po0.01 (based on simple contrasts with healthy weight as the reference).

International Journal of Obesity (2013) 40 -- 46

& 2013 Macmillan Publishers Limited


the study. All tests and protocols were approved under the Educational Testing Service human subjects review committee and fairness review process. Participants This sample comprises 1036 students (64% female) enrolled in university (n ¼ 574) or community college courses (n ¼ 462) across 24 institutions in five regions of the United States (East, West, midWest, South and Alaska). Students were aged between 15 and 85 years, with most (80.8%) in the typical college age range of 18 to 25 years. Self-reported ethnicities were White Non-Hispanic (52.5%), African American (23.1%), Hispanic Non-White (13.6%) or Asian (5.6%). Measures Mathematics. Participants completed 15 items comprising Part 1 of the ‘Necessary Arithmetic Operations Test’ from the ETS Kit of Factor-Referenced Cognitive Tests.33 For each item, test-takers select one of four sets of mathematical operations that must be performed to obtain the answer. For example: ‘A sweater marked $40 was sold for $29.95 during a sale. What was the percent reduction? 1 -- divide and add, 2 -- subtract and divide, 3 -- multiply and subtract, 4 -- add and divide’. Vocabulary. Participants completed a 25-item version of the Vocabulary Test II from the ETS Kit of Factor-Referenced Cognitive Tests.33 Test takers were required to select from five choices the synonym closest in meaning to a target word. For example: ‘Compatible: 1 -- abridged, 2 -- congenial, 3 -- compelling, 4 -- related, 5 -- combined’. HEXACO-PI-R. This 96-item test assesses six dimensions of personality, with 16 items per dimension: Honesty/Humility, Emotional Stability, Extraversion, Agreeableness, Conscientiousness and Openness to Experience.34 The Conscientiousness total score was used as a measure of Conscientiousness (for example, ‘I pay attention to details’). Participants rated how accurately each item described them on a five-point scale from ‘Very Inaccurate’ (1) to ‘Very Accurate’ (5). Positive and negative feelings. This 33-item test measures the extent to which students have felt positive and negative emotions in the last month, when engaging in typical school activities.35 Positive emotions are captured by 14 items (for example, ‘I have felt HAPPY’) and negative emotions by 19 items (for example, ‘I have felt ANXIOUS’). Participants rated how often they felt each emotion in a four-point scale from ‘Never or rarely’ (1) to ‘Usually or Always’ (4). Socioeconomic status. SES was derived from taking the first principal component of four variables (mother’s education, father’s education, number of books in one’s childhood home and whether there were encyclopedias in one’s childhood home). These questions were drawn from the background variables of the NAEP. Missing cases were replaced with the mean value for each variable. Population density. Participants were asked to report whether their family home was in a city, suburb, town or rural area. High-stakes test scores. Participants self-reported the scores they obtained on standardized college entrance tests: EITHER their SAT scores (the sum of the mathematics and verbal components) or their total ACT scores. Student grades. Participants reported their GPA in the previous semester of college (n ¼ 536) or community college (n ¼ 409) on a 0 to 4 scale. & 2013 Macmillan Publishers Limited

Body mass index. Participants reported their height (in feet and inches) and weight (in pounds), and these reports were used to calculate a BMI score for each individual (kg m -- 2). BMI cutoffs of 18.5, 25 and 30kg m -- 2 were used to define weight categories of underweight (o18.5 kg m -- 2), normal weight (18.5 -- 25 kg m -- 2), overweight (25 -- 30 kg m -- 2) and obese (430 kg m -- 2). In total, 4% of participants were underweight, 50% were of normal weight, 25% were overweight and 21% were obese. RESULTS AND DISCUSSION Descriptive statistics Table 3 provides internal consistency estimates (Cronbach’s a) and descriptive statistics for intelligence test scores, achievement test scores, students’ GPA, conscientiousness and both positive and negative feelings toward school. All measures show adequate reliability. Relationship between overweight and academic performance As in study 1, Table 4 presents the relationship between achievement and weight in two ways: (1) as the correlation of BMI with each dependent variable (mathematics, vocabulary, SAT and ACT scores as well as GPA in community college, and at university); and (2) as the mean score on each dependent variable for underweight, healthy-weight, overweight and obese students, including whether scores for healthy-weight students are significantly different than those for underweight, overweight or obese students. As in study 1, both forms of the weightachievement relationship are shown with and without accounting for covariates. Table 4 also shows the power to detect a mean difference of medium effect size (d ¼ 0.50). Correlations. Before controlling for covariates, BMI was significantly correlated with mathematics scores, SAT scores and university GPA. However, after controlling for covariates, BMI was significantly related to both university and community college GPA, but was not significantly related to any of the test scores (mathematics, vocabulary and SAT or ACT scores). Mean differences. For both the intelligence test scores (mathematics and vocabulary) and the entrance exam scores (SAT and ACT), there were no differences between healthy-weight and obese students, even before controlling for covariates. There were also no significant differences between overweight and healthyweight students. As for the eighth graders, the relationship for grades was different: obese students received significantly lower grades than healthy-weight students both before and after controlling for ability, personality and well-being covariates. Overweight students also received lower grades than healthyweight students at community college, but only after controlling for covariates. Table 3.

Reliability and descriptive statistics for higher education

students

GPA (university, n ¼ 563) GPA (community college, n ¼ 409) Mathematics Vocabulary SAT score (n ¼ 295) ACT score (n ¼ 386) Conscientiousness Positive feelings Negative feelings

Score range

Cronbach’s a

Mean (s.d.)

0 -- 4 0 -- 4

-----

3.12 (0.57) 3.11 (0.61)

0 -- 100 0 -- 100 400 -- 800 3 -- 36 1 -- 5 1 -- 4 1 -- 4

0.77 0.77 ----0.80 0.92 0.92

62.78 64.89 931.79 21.83 3.59 2.53 3.06

(22.25) (16.07) (265.77) (6.52) (0.56) (0.61) (0.55)


43


44 Table 4. Mean test scores and grades for higher education students by weight category, with and without controlling for confounds (d-scores calculate the effect size for healthy-weight vs other groups) Marginal means (after controls)a

Means (before controls) N

Mean (s.d.)

Mathematics Healthy weight Underweight Overweight Obese

501 41 258 217

9.53 9.44 9.29 9.33

(3.31) (3.26) (3.26) (3.46)

Vocabulary Healthy weight Underweight Overweight Obese

501 41 258 217

16.15 15.98 16.56 16.20

(4.02) (4.16) (3.65) (4.31)

SAT Healthy weight Underweight Overweight Obese

152 13 75 55

940.95 964.23 917.09 918.82

ACT Healthy weight Underweight Overweight Obese

189 19 93 79

22.31 20.89 21.99 21.29

(6.33) (7.45) (6.60) (6.40)

GPA (CC) Healthy weight Underweight Overweight Obese

176 16 88 116

3.18 3.34 3.10 2.97

(0.54) (0.52) (0.58) (0.71)

GPA (Uni) Healthy weight Underweight Overweight Obese

296 25 152 85

3.17 3.17 3.12 2.95

(0.55) (0.47) (0.53) (0.70)

(262.71) (274.98) (277.72) (261.00)

d

Mean

0.03 0.07 0.06

9.48 9.29 9.27 9.49

0.04 0.11 0.01

16.12 15.98 16.50 16.37

0.09 0.09 0.08

936.91 940.50 903.38 954.31

0.22 0.05 0.16

22.09 20.53 22.15 21.72

0.30 0.14 0.34**

3.21 3.44 3.06 2.95

0.00 0.09 0.36**

3.15 3.12 3.16 2.97

Power to detect medium effect

db

0.06 0.06 0.00

0.03 0.10 0.06

0.01 0.13 0.07

0.24 0.01 0.06

0.43 0.27* 0.42**

0.06 0.02 0.31**

Correlations with BMI r

Partial r

0.23*

0.01

0.92 1.00 1.00 0.19

0.04

0.31**

0.16

0.02

0.02

0.07

0.12**

0.17**

0.15**

0.92 1.00 1.00

0.53 0.94 0.89

0.66 0.96 0.96

0.60 0.97 0.99

0.77 1.00 0.98

Abbreviations: BMI, body mass index; CC, community college; GPA, grade point average; Uni, university. aSex, socioeconomic status (SES), dummy-coded ethnicity, geographical region, Conscientiousness and both positive and negative feelings toward school and negative are entered as covariates for all analyses of covariance (ANCOVAs) and partial correlations. For grades, vocabulary and mathematics scores are also entered as covariates. bTo ensure that d-scores are comparable before and after covariate adjustments, Cohen’s d is calculated using the pooled s.d. before adjustment to compare marginal means (rather than the covariate-attenuated standard error). *The t-statistic significant at Po0.05 (based on simple contrasts with healthy weight as the reference). **The t-statistic significant at Po0.01 (based on simple contrasts with healthy weight as the reference).

Both correlations and mean differences analyses suggested a suppressor effect for community college GPA: there was a significant relationship between weight and GPA before accounting for covariates, but this decreased to nonsignificance after covariates were accounted for. We checked whether the suppressor effect was because of cognitive factors (that is, we controlled only for demographics, vocabulary and mathematics in the analysis of covariance) or noncognitive factors (that is, we controlled only for demographics, conscientiousness and feelings toward school in the analysis of covariance). The variables responsible for the suppressor effect are vocabulary and mathematics scores (not conscientiousness, nor positive or negative feelings). That is, differences between students’ intelligence are suppressing the effect of obesity on grades in this sample. Conclusions Results for study 2 thus provide support for both hypotheses. Although there was no difference in intelligence or achievement tests for obese and healthy-weight students (after controlling for covariates), both community colleges and universities awarded lower grades to obese students than to healthy-weight students. International Journal of Obesity (2013) 40 -- 46

These lower grades are not because of intelligence, SES, psychological factors like work ethic/achievement drive (that is, conscientiousness) or because of different emotional reactions to school for obese versus healthy-weight students. Given the diversity of the sample (covering 24 different institutions in disparate geographic regions across the United States), the results suggest that this is a real social phenomenon occurring in higher education in the United States at present. GENERAL DISCUSSION Results indicate support for both hypotheses: school grades showed much larger discrepancies between healthy-weight and obese students than did test scores, either for intelligence tests (vocabulary and mathematics) or university entrance exams. In fact, both studies showed that there were no differences in test scores between obese and healthy-weight students when covariates were appropriately controlled. For study 2, there were not even any differences in test scores before covariates were considered. Such a result informs the seeming contradiction in previous findings: a difference in achievement is found for grades (replicating Sabia7), but not for test scores (replicating Kaestner and Grossman6). & 2013 Macmillan Publishers Limited


One possible explanation for our results is that GPA and test scores simply reflect different types of ability, which are open to different influences. Perhaps the cognitive processes required to obtain high grades may be impaired by obesity, but the processes required for obtaining vocabulary and mathematics scores are not. However, considering the nature of the cognitive tests used in both studies makes this argument less compelling. Modern intelligence theory distinguishes between acculturated knowledge (Gc) and fluid reasoning ability (Gf), and both mathematics and vocabulary are closer to Gc rather than Gf.36 GPA also more strongly reflects Gc than Gf. That is, results are even more compelling given that grades and test scores are both based on the same underlying ability (Gc). After controlling demographic and personality variables, overweight students have comparable intelligence and scholastic aptitude to healthy-weight students, such that their significantly lower grades do not seem to reflect an inability to master the scholastic material. Several researchers have suggested that overweight students are discriminated against by teachers and other students.6,37,38 This discrimination may lower grades through either direct or indirect pathways. In a direct pathway, teachers may award lower grades to overweight and obese students based on a holistic judgment or attitude toward the student. Neumark-Sztainer et al.13 demonstrate that a substantial proportion of adolescent educators hold negative beliefs about obese people, believing that obese workers cannot be successful, obese people are untidy and that becoming obese is one of the worst things that can happen to a person. In addition, studies demonstrate that physical education teachers, in particular, have negative beliefs about obese students.14 Research demonstrates that low teacher expectations can predict student outcomes, and this may well apply in the case of obese students.39 In an indirect pathway, obese students may obtain less of the teacher’s time and resources, get less help from their peers or family, be more frequently excluded from group projects or may be bullied in ways that affect school grades (such as having homework assignments or resources stolen). In support of some of these indirect pathways, research has found that: (1) obese students are more frequently bullied than nonobese students;12 (2) fifth- and sixth-grade children show a stronger bias against obesity than against any other physical characteristic;37 and (3) obese students receive less financial support for college from their parents than their nonobese counterparts, irrespective of family economic circumstances.15,16 Taken together, it seems plausible that obese students are receiving lower grades because of peer, teacher and even family discrimination and stigma. One alternative explanation is that obesity may lead to health problems that cause school absences for medical reasons (thus lowering grades through absenteeism).40 However, absenteeism and lack of learning opportunities should also affect students’ scores on mathematics, vocabulary, SAT and ACT tests (as students would not then learn the scholastic content involved in these tests). As there is little difference between obese and healthy-weight students in mathematics, vocabulary, SAT or ACT scores, it is unlikely that absenteeism is the cause of differences in school grades. A further alternative explanation relates to affective variables such as depression and related negative feelings or lack of positive affect. Depression is a causal factor in overweight and obesity,41 and negative affect may also be a causal factor in low achievement.35 However, this pathway was to some extent controlled for by accounting for variables such as life satisfaction (in study 1), and positive and negative affect (in study 2). It is unlikely that depression or negative emotionality account for the lower grades among obese and overweight students, given that broadly similar constructs were controlled for. An important goal for future research would be to ascertain the causal pathway between overweight/obesity and school grades. & 2013 Macmillan Publishers Limited

At present, it is not certain whether overweight/obesity causes lower school grades (through social discrimination), or whether lower school grades cause overweight/obesity. It has been suggested that students might overeat as a coping mechanism to feel better after receiving low school grades,7 although the control for well-being variables makes this a less convincing explanation (as differences in positive and negative feelings are controlled for). Nevertheless, a longitudinal research design would allow testing of causal hypotheses about whether overweight/ obesity causes school grades or vice versa. As the current data collection effort is longitudinal and ongoing, we expect to be able to answer this question as further waves of data become available. This study demonstrates that overweight and obese students are not obtaining the same grades as their healthy-weight counterparts, despite no clear differences in either mathematics or vocabulary scores. This difference is also unrelated to income level, as SES is held constant. Whatever the causal mechanism, the result is clear: overweight and obese students are not receiving the grades that would be predicted from their test scores, achievement drive and socioeconomic backgrounds. After accounting for controls, the relative difference in grades is nontrivial, with effect sizes of around a third of the s.d. CONFLICT OF INTEREST The authors declare no conflict of interest.

ACKNOWLEDGEMENTS We thank Brent Bridgeman and Cathy Wendler for insightful comments on a previous version of this manuscript.

DISCLAIMER All statements expressed in this article are of the authors and do not reflect the official opinions or policies of the University of Sydney or the Educational Testing Service.

REFERENCES 1 Pyle S, Sharkey J, Yetter G, Felix E, Furlong M, Poston W. Fighting an epidemic: the role of schools in reducing childhood obesity. Psychol School 2006; 43: 361 -- 376. 2 Ogden C, Carroll M, Flegal K. High body mass index for age among US children and adolescents, 2003-2006. JAMA 2008; 299: 2401 -- 2405. 3 Sigfu´sdu´ttir I, Kristjańsson A, Allegrante J. Health behaviour and academic achievement in Icelandic school children. Health Educ Res 2007; 22: 70 -- 80. 4 Datar A, Sturm R. Childhood overweight and elementary school outcomes. Int J Obes 2006; 30: 1449 -- 1460. 5 Datar A, Sturm R, Magnabosco J. Childhood overweight and academic performance: national study of kindergartners and first-graders. Obes Res 2004; 12: 58 -- 68. 6 Kaestner R, Grossman M. Effects of weight on children’s educational achievement. NBER Work Pap Ser 2008; 13764: 1 -- 35. 7 Sabia J. The effect of body weight on adolescent academic performance. Southern Econ J 2007; 73: 871 -- 900. 8 Carter M, Dubois L, Ramsay T. Examining the relationship between obesity and math performance among Canadian school children: A prospective analysis. Int J Pediatr Obes 2010; 5: 412 -- 419. 9 Li Y, Dai Q, Jackson J, Zhang J. Overweight is associated with decreased cognitive functioning among school-age children and adolescents. Obesity 2008; 16: 1809 -- 1815. 10 Huang T, Goran M, Spruijt-Metz D. Associations of adiposity with measured and self-reported academic performance in early adolescence. Obesity 2006; 14: 1839 -- 1845. 11 Karnehed N, Rasmussen F, Hemmingsson T, Tynelius P. Obesity and attained education: cohort study of more than 700,000 Swedish men. Obesity 2006; 14: 1421 -- 1428. 12 Janssen I, Craig W, Boyce W, Pickett W. Associations between overweight and obesity with bullying behaviors in school age children. Pediatrics 2004; 113: 1187 -- 1194.


45


46 13 Neumark-Sztainer D, Story M, Harris T. Beliefs and attitudes about obesity among teachers and school health care providers working with adolescents. J Nutr Educ 1999; 31: 3 -- 9. 14 Greenleaf C, Weiller K. Perceptions of youth obesity among physical educators. Soc Psychol Ed 2005; 8: 407 -- 423. 15 Crandall C. Do parents discriminate against their heavyweight daughters? Pers Social Psych Bull 1995; 21: 724 -- 735. 16 Crandall C. Do heavy-weight students have more difficulty paying for college? Pers Social Psych Bull 1991; 17: 606 -- 611. 17 Crandall C. Prejudice against fat people: ideology and self-interest. J Pers Soc Psychol 1994; 66: 882 -- 894. 18 Lee J. Racial and ethnic achievement gap trends: reversing the progress toward equality. Educ Researcher 2002; 31: 3 -- 12. 19 DHHS. Physical Activity and Good Nutrition: Essential Elements to Prevent Chronic Diseases and Obesity. DHHS: Atlanta, GA, 2007 (accessed 15 February 2008). 20 Kaufman A, McLean J, Reynolds C. Sex, race, region, and education differences on the 11 WAIS-R subtests. J Clin Psychol 1988; 44: 231 -- 248. 21 Bogg T, Roberts B. Conscientiousness and health-related behaviors: a metaanalysis of the leading behavioral contributors to mortality. Psychol Bull 2004; 130: 887 -- 919. 22 Kern M, Freidman H. Do conscientious individuals live longer? A quantitative review. Health Psychol 2008; 27: 505 -- 512. 23 Noftle E, Robins R. Personality predictors of academic outcomes: big five correlates of GPA and SAT scores. J Pers Soc Psychol 2007; 93: 116 -- 130. 24 Ball K, Crawford D, Kenardy J. Longitudinal relationships among overweight, life satisfaction, and aspirations in young women. Obes Res 2004; 12: 1019 -- 1030. 25 Strauss R. Childhood obesity and self-esteem. Pediatrics 2000; 105: 1 -- 5. 26 Shin H. School Enrollment -- Social and Economic Characteristics of Students: October 2003. US Department of Commerce: Washington, DC, 2005 (accessed 11 April 2008). 27 Schmitt N, Schmitt D, Clapham C. Developing and exploring the behaviour of two new versions of the Vocabulary Levels Test. Language Testing 2001; 18: I55 -- I88.


28 Elgar F, Roberts C, Tudor-Smith C, Moore L. Validity of self-reported height and weight and predictors of bias in adolescents. J Adolesc Health 2005; 37: 371 -- 375. 29 Onis M, Onyango A, Borghi E, Siyam A, Nishidaa C, Siekmanna J. Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ 2007; 85: 660 -- 667. 30 MacCann C, Duckworth A, Roberts R. Empirical identification of the major facets of conscientiousness. Learn Individ Differ 2009; 19: 451 -- 458. 31 Huebner E. Initial development of the Students’ Life Satisfaction Scale. School Psychol Int 1991; 12: 231 -- 243. 32 Cohen J. Statistical Power Analysis for the Behavioral Sciences 2nd edn. Lawrence Erlbaum Associates: Hillsdale, NJ, 1988. 33 Ekstrom R, French J, Harman H, Derman D. Manual for Kit of Factor-Referenced Cognitive Tests. ETS: Princeton, NJ, 1976. 34 Lee K, Ashton M. The HEXACO Personality Inventory: a new measure of the major dimensions of personality. Multivariate Behav Res 2004; 39: 329 -- 358. 35 Lipnevich A, MacCann C, Bertling J, Naemi B, Roberts R. Emotional reactions toward school situations: relationships with academic outcomes. J Psychoeduc Assess 2012 (in press). 36 McGrew K. CHC theory and the human cognitive abilities project: standing on the shoulders of the giants of psychometric intelligence research. Intelligence 2009; 37: 1 -- 10. 37 Latner J, Stunkard A. Getting worse: the stigmatization of obese children. Obes Res 2003; 11: 452 -- 456. 38 Strauss R, Pollack H. Social marginalization of overweight children. Arch Pediatr Adolesc Med 2003; 157: 746 -- 752. 39 Jussim L, Harber K. Teacher expectations and self-fulfilling prophecies: knowns and unknowns, resolved and unresolved controversies. Pers Soc Psychol Rev 2005; 9: 131 -- 155. 40 Taras H, Potts-Datema W. Obesity and student performance at school. J Sch Health 2005; 75: 291 -- 295. 41 Goodman E, Whitaker R. A prospective study of the role of depression in the development and persistence of adolescent obesity. Pediatrics 2002; 110: 497 -- 504.

& 2013 Macmillan Publishers Limited