Racial and Ethnic Differences in Homework Time among U.S. Teens

Abstract

Along with intensified competition for college admissions, U.S. teens increasingly spend more time on educational activities. Homework can be a particularly important component of educational time for economically disadvantaged and racial/ethnic minority students who have limited access to private sources of learning beyond the classroom. This study uses data from the American Time Use Survey and the Programme for International Student Assessment to compare homework time by race/ethnicity and examine the factors that explain these differences. We extend existing literature to consider explanations beyond demographic and family background. Our ordinary least squares (OLS) results show that family background accounts for the difference in homework time between Hispanic and White students and partially explains the difference between Black and White students, with students’ academic characteristics or school fixed effects explaining the remaining gap. While these factors partially account for Asian students’ greater time spent on homework than their White peers, a substantial gap remains.

Historically, teens in the United States have spent less time on schoolwork (such as time spent on homework, studying and preparation for standardized tests) and have had more free time than their counterparts in other parts of the world, particularly in East Asia (Larson 2001). Reflecting the minor role of academic-oriented schoolwork in teens’ daily activities, relatively little attention has been paid to who spends more or less time on academic-oriented schoolwork. Instead, research has focused more on other aspects of time use among U.S. teens, including their participation in non-academic extracurricular activities such as sports, arts, music, and clubs (Dotterer, McHale, and Crouter 2007; Dumais 2009; Gabler and Kaufman 2006; Haghighat and Knifsend 2019) as well as high schoolers’ involvement in paid work (Mortimer 2010).

However, recent research suggests that compared with previous decades, U.S. teens are spending more time on homework and are increasingly involved in commercial SAT/ACT prep services and private tutoring after regular school hours, but are spending less time on paid work and on socializing (Buchmann, Condron, and Roscigno 2010; Ho, Park, and Kao 2019; Livingston 2019). These patterns point to a broad expansion in the amount of time U.S. teens spend on academically oriented work and are consistent with larger trends of intensified competition for college admissions and increased parental time and financial investment in children’s education (Kornrich and Furstenberg 2013; Park et al. 2016; Schneider, Hastings, and LaBriola 2018).

Recent scholarship has considered racial and ethnic inequalities in students’ access to and engagement with private supplementary education as well as the implications of these inequalities for racial and ethnic differences in educational outcomes more broadly (e.g., Ho et al. 2019; Park et al. 2016). However, relatively little scholarly attention has been paid to racial and ethnic inequalities in homework time. Although some descriptive portraits of racial and ethnic variation in homework time are available, they often distinguish only between non-Hispanic White and all racial and ethnic minority students combined, failing to explore variation across the diverse groups of students (Porterfield and Winkler 2007). Other studies treat race and ethnicity merely as controls in multivariate analysis to examine other factors such as paid work time in relation to homework time (Kalenkoski and Pabilonia 2009, 2012). More fundamentally, while some research has documented racial and ethnic differences in homework time (e.g., Hansen and Quintero 2017), little research has systematically investigated the individual and structural factors that might explain these observed differences.

Understanding inequalities in homework time is important because consistent evidence suggests a significant relationship between students’ homework time and improved educational outcomes for high school students (although this relationship is contested at younger ages). Despite some debates on the magnitude and potential variation by different characteristics of students, the weight of empirical evidence supports the conclusion that the average effect of homework time on achievement is positive among high school students (see, e.g., Cooper 1989; Cooper, Robinson, and Patall 2006; Daw 2012). Racial and ethnic differences in homework time, therefore, may be a channel through which racial and ethnic inequalities in educational outcomes are produced.

In this study, we draw on time use data from the American Time Use Survey (ATUS) and survey data from the Programme for International Student Assessment (PISA) to examine how homework time varies by race and ethnicity among U.S. high schoolers and to assess the extent to which family, students’ academic and school factors explain these differences. Using ATUS data, we first compare how average daily educational time (including time spent in class, on homework and on other educational activities such as SAT preparation) varies among non-Hispanic Asian, non-Hispanic Black, Hispanic, and non-Hispanic White high schoolers aged 15 to 19. We show that substantial racial and ethnic differences in homework time drive racial and ethnic differences in daily education time, which further highlights the significance of research on homework time.

We then use ATUS and PISA data to examine the extent to which diverse aspects of family background (family structure, family socioeconomic status, and location) and students’ academic background (test scores and students’ educational expectations), in addition to students’ demographic characteristics (age, gender, and nativity), contribute to racial and ethnic differences in homework time. Although ATUS’s time diary design allows for robust estimates of teens’ homework time, analysis of the factors explaining racial and ethnic differences in homework time with this household-based survey is limited to family and demographic factors. PISA measures homework time using stylized survey questions; however, the study’s school-based design allows for greater exploration of how students’ academic background mediates the relationship between race/ethnicity and homework time.

Finally, recognizing that the schools that racial and ethnic minority and White students attend may vary in academic climates including homework practices, we use school IDs provided in PISA to estimate school-fixed effect models to assess racial and ethnic differences within schools.¹ By assessing racial and ethnic variation in homework time after taking into account students’ academic characteristics and school fixed effects, our study extends current understandings of racial and ethnic differences in homework time based on studies that consider solely students’ demographic characteristics and family background factors available in ATUS.

Background

Racial and Ethnic Differences in Homework Time and Family Background

Numerous studies have assessed racial and ethnic differences in a variety of characteristics of students, families, and schools in attempt to account for differences in various educational outcomes among different racial and ethnic groups of students (American Psychological Association 2012; Gamoran 2001; Kao and Thompson 2003). Surprisingly, however, there is a dearth of research examining racial and ethnic variation in homework time. Although homework time may not have a uniform relationship across different educational outcomes or by student characteristics, evidence of a positive relationship with academic achievement (especially among high school students) suggests that homework time may be relevant for explaining racial and ethnic differences in students’ educational outcomes (Cooper et al. 2006; Eren and Henderson 2008; Kalenkoski and Pabilonia 2017). Some research has examined Asian American students’ considerable academic effort and work ethic as well as their time spent participating in after-school supplementary education (Asakawa and Csikszentmihalyi 1998; Byun and Park 2012; Hsin and Xie 2014); however, scholarly attention to homework time across other racial and ethnic minority groups is limited.

In a study assessing how students’ attitudes, skills, habits, and styles explain racial differences in school performance, James Ainsworth-Darnell and Douglass Downey (1998) showed that African American high school sophomores spent less time on homework than their non-Hispanic White counterparts, while Asian American students spent more than both racial groups of students. They found that family income, parental occupation, and parental education partly accounted for the difference in homework time between African American and White students. Demographic characteristics and family structure, rather than socioeconomic background, partly accounted for the difference between Asian American and White students. Although this study provides important evidence to inform our current study, it did not examine homework time among Hispanic students and did not consider within-school relationships between race and homework time. Moreover, current overall patterns of homework time among U.S. teens may differ substantially from those among their study sample of high school sophomores in 1990.

Michael Hansen and Diana Quintero (2017) provide a more recent descriptive analysis of racial and ethnic differences in homework time. Drawing on pooled data from ATUS 2003–2013, they found that Asian American high schoolers spent nearly two hours on homework per day—almost twice the amount of time spent by White and Hispanic students. Black students spent the least amount of time on homework at around 30 minutes per day. However, the authors of this brief report did not seek to examine the factors driving these differences in homework time, except for noting that “gaps by race still existed, even holding parental education constant” (p. 5).

A study by Yelizavetta Kofman and Suzanne Bianchi (2012) most closely reflects the aims and focus of our current study. Using the data from ATUS 2003–2010, the authors examined nativity and racial/ethnic differences in time spent on various activities including “studying time” among 15- to 17-year-olds.² They found that native-born Hispanic and native-born Black adolescents spent significantly less time studying than their non-Hispanic native-born White peers but that these groups no longer differed after taking family socioeconomic and demographic factors into account. Asian immigrant adolescents spent the most time studying among all groups, and the gap between Asian immigrants and native-born White peers did not substantially change after controlling for family socioeconomic and demographic variables.³ Latin American immigrant adolescents also spent more time than their native-born White peers in the multivariate model, although the gross difference was not significant.

In sum, the few studies examining racial and ethnic differences in homework time show that Black, and to a lesser extent Hispanic, students tend to spend less time on homework than their White peers, whereas Asian American students spend more time on homework than White students. Furthermore, studies consistently find that Black-White and Hispanic-White differences in family socioeconomic and demographic background account for some proportion of the observed differences in homework time but that these factors do not sufficiently explain the gap in homework time between Asian and White students.

Students’ Academic Background and School Setting

Prior research has shed some light on racial and ethnic differences in homework time, pointing to the importance of family background in partially explaining some of these differences. However, partially due to reliance on ATUS as a main data source, existing studies do not examine other individual and structural factors beyond family background in assessing racial and ethnic differences in homework time. Among various potential factors, students’ academic characteristics are likely to be highly relevant. Studies have found substantial racial and ethnic differences in prior academic performance and children’s educational expectations, particularly favoring Asian American students (Goyette and Xie 1999; Hsin and Xie 2014; Kao and Thompson 2003; Qian and Blair 1999; Turcios and Milan 2013). These characteristics are likely to affect students’ efforts and persistence to study. Prior achievement may indicate academic ability and motivation and is positively associated with homework time (Gershenson and Holt 2015). Students’ own expectations for educational attainment are also likely to be associated with students’ homework time. Students who expect to receive a college degree or higher are likely to spend more time on homework than those who do not. In short, prior academic achievement and educational expectations may mediate the relationship between race/ethnicity and homework time. We expect this mediating role of prior academic achievement and educational expectations to be particularly evident for the difference in homework time between White and Asian American students whose academic achievement and educational expectation are notable.

Another factor that has largely been overlooked in prior literature on racial and ethnic differences in homework time is between-school variation. Evidence shows that the schools that racial and ethnic minority students attend differ significantly in a variety of characteristics compared with the schools that White students attend, such as socioeconomic contexts, overall academic performance, teacher experience and resources (e.g., Condron 2009; Fiel 2013; Goyette and Lareau 2014; Logan, Minca, and Adar 2012). Furthermore, schools may differ in academic climates, including the importance that schools and teachers attach to homework and homework-related practices (e.g., the frequency and quantity of homework assignments). Academic climates of schools are determined not only by schools’ policies and practices but also by peer groups. When students with strong academic motivation and high academic standards attend the same school, they may enforce academic norms among themselves (Jiménez and Horowitz 2013). If racial and ethnic minority students are more likely to attend schools that place less emphasis on study time and homework than White students, racial and ethnic differences in homework time may persist even after considering students’ individual characteristics. Peer group effects in those less academically demanding schools will further contribute to racial and ethnic differences in homework time.

However, isolating specific school-related characteristics among many possible school factors that influence students’ engagement with homework presents challenges and few school-level data sets contain detailed information on schools’ homework policies and practices. It is challenging to estimate the effects of school-level variables on students’ outcomes separately from students’ own observed and unobserved characteristics without randomly assigning students to schools (Raudenbush and Willms 1995). Instead, following Seth Gershenson and Stephen Holt (2015), we use school-fixed effect models to estimate racial and ethnic differences in homework time within schools. If racial and ethnic differences in homework time persist while accounting for school fixed effects, this indicates that these differences are not due to the fact that students of different racial and ethnic backgrounds tend to attend different schools.

Hypotheses

In this study, we draw on data from the ATUS and the PISA to build on the few existing studies of racial and ethnic differences in homework and daily education time among U.S. high school students. First, we use ATUS to examine racial and ethnic differences in high schoolers’ daily education time (including time spent in class, on homework and on other educational activities such as SAT preparation), highlighting that racial and ethnic differences in overall educational time are largely due to differences in homework time. We then document how racial/ethnic patterns of homework time have changed over the last decade. Finally, drawing on both ATUS and PISA, we use regression analysis to assess the factors that account for the racial and ethnic differences in homework time identified. We not only provide an updated and detailed portrait of racial/ethnic variation in homework time but, importantly, we extend existing research by examining the extent to which racial and ethnic differences in students’ academic characteristics and school setting explain differences in homework time in addition to previously established explanations regarding family background and demographic factors.

Drawing on prior research on racial ethnic differences in education and academic outcomes, we hypothesize the following:

• Hypothesis 1 (H1): Racial and ethnic differences in family background characteristics such as family structure, family resources, and location explain differences in homework time between Black and White and between Hispanic and White students.

• Hypothesis 2 (H2): Differences in family background do not sufficiently account for differences in homework time between Asian and White students. However, this gap in homework time is further explained by differences in students’ academic characteristics between Asian and White students.

• Hypothesis 3 (H3): Between-school differences partially explain differences in homework time for all racial and ethnic minority groups compared with White students.

Data and Methods

Data and Sample

We use two different sources of data to examine racial and ethnic differences in homework time: the ATUS (Hofferth et al. 2020) from 2003 to 2019 and the U.S. component of the PISA 2012 (Organization for Economic Co-Operation and Development [OECD] 2014). Each data set possesses strengths and weaknesses that complement each other, allowing us to better fill the research gaps we have identified above. Although ATUS provides time diary estimates of teens’ time use as well as family and household characteristics, it does not collect information on respondents’ academic background or schools they attend. By contrast, while PISA contains only stylized survey estimates of students’ time spent on homework, the survey contains school-related information as well as students’ academic characteristics.

ATUS is a nationally representative cross-sectional study, with a sample drawn from households that have completed their final month of interviews for the Current Population Survey (CPS). One individual 15 years of age or older from each household is randomly selected to complete the ATUS time diary interview, which provides a detailed account of the respondent’s activities over one 24-hour period. Because the ATUS sample is drawn from the CPS, some information on other household members can be linked from the CPS interview. Yearly sample sizes for 15- to 19-year-olds, particularly for Asians, are small and therefore we pool data across 2003 through 2019 with survey years included as a control.

PISA is an international student assessment of 15-year-olds collected and administered by the OECD every three years. The study includes knowledge and skills assessments across reading, mathematics, and science as well as some year-specific subjects, a student survey covering students’ family backgrounds, study habits and attitudes toward learning, and a survey of the principles of participating schools providing school-level information about students’ learning environment. PISA collects data using a two-stage stratified sampling design where schools are first sampled then students at age 15, regardless of their grade, are selected within sample schools. In 2012, PISA was administered in 65 countries and economies including the United States.⁴ The 2012 student survey asked students about the amount of time they spend on homework, allowing for comparison of homework time collected in ATUS.

Our ATUS sample consists of 6,791 teenagers aged 15 to 19 years old who were enrolled full-time in high school and who completed their diary interviews between September and May (to exclude time diaries collected during summer break). The U.S. component of PISA 2012 consists of 4,978 students from 162 schools. Homework-related items were included in the student questionnaire for a random two thirds of the sample and therefore 1,775 students who did not receive homework-related questions were excluded from our analysis. We excluded an additional 38 students missing information on educational expectations (explained below), resulting in an analytic sample of 3,165 students from 161 schools.

Variables

Our outcome measure is a continuous measure of the number of minutes respondents spend doing homework per day. In ATUS, this measure is a summation of the total time over the 24-hour diary day that respondents reported doing “research or homework for class or degree, certification, or licensure” (activity code 060301). In PISA, this measure comes from a student survey question asking students to record the total number of hours per week they spend, on average, on “homework or other material assigned by [their] teachers.” To aid comparability with ATUS, we divided responses by 7 to obtain a daily average.

It is worth noting the differences between the ATUS and PISA measures of homework time. Time diary data (like that provided in ATUS) is considered the “gold standard” method of measuring time allocation (Cornwell, Gershuny, and Sullivan 2019) as other methods of measurement such as stylized survey questions about time allocation (like those provided in PISA) are more prone to measurement error (e.g., Schulz and Grunow 2012). In ATUS, respondents report their activities within 24 hours from 4:00 a.m. the previous day and time use is recorded in minutes. In PISA, by contrast, students report the average amount of time they spend on homework each week, recorded in hours. These differences in time collection methods and the unit of measurement indicate that ATUS should provide more precise and reliable estimates on daily homework time than PISA does. However, as we emphasized above, information on schools and students’ academic characteristics is available only in PISA.

Our key independent variable is race/ethnicity. In the publicly available PISA data for the United States, the race/ethnicity measure has six categories, distinguishing between non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian, Multiracial, and Other. Due to small sample sizes, we combined students who identified as “Multiracial” and “Other” into the “Other” category, resulting in a five-category race/ethnicity variable. Seventeen cases missing race/ethnicity information were also coded as “Other.” In ATUS, we constructed a five-category race/ethnicity variable comparable to PISA’s measure using separate race and ethnicity variables. Although we include the category of “Other” in our regression analysis, we do not give any substantive interpretation for the category given its heterogeneity. There are no missing cases on race and ethnicity in the ATUS analytic sample of 6,791.

In assessing the differences in homework time across racial and ethnic groups of students, we control for gender and whether the respondent was born in the United States. In analysis of ATUS, we also control for survey year, a binary indicator of whether the respondents’ diary was collected on a weekend or a weekday, and respondents’ age, centered at 17. In analysis of PISA, we control for respondents’ grade in school relative to the modal year (10th grade). There are no missing cases on these control variables in ATUS or PISA.

For family background measures, we include binary measures of whether the respondent lives in a two-parent household and whether the respondent has at least one parent with a bachelor’s degree. In analysis of ATUS, we also include a measure of family income, measured in tertiles calculated in each survey year separately. The share of students missing information on parental education and family income is fairly small (2.3 percent and 7.0 percent, respectively) and we include those missing students into the models with a separate category for “missing.” For PISA, which did not collect information on family income, we include a measure of family wealth also measured in tertiles.⁵ One case missing family wealth data was recoded to the middle tertile. Finally, we also control for the size of community in which respondents live. In ATUS, this is a binary variable indicating whether a respondent lives in a metropolitan area (as defined by the Census Bureau at the time of survey) or not, while in PISA this distinguishes whether the respondents’ school is located in a town, city, or large city (referred to as “metropolitan” for consistency with ATUS) versus a small town or rural community. Forty-eight cases (1.5 percent of the analytic sample) from three schools were missing school location information in PISA and were recoded to the modal category, metropolitan.

In PISA, which contains students’ academic characteristics and school IDs, we classify students into tertiles according to their PISA assessment scores (averaging assessment scores in math, science and reading) to measure prior academic achievement. That this measure of academic achievement was measured simultaneously with homework time as part of the PISA study is a limitation of this analysis. We use PISA assessment scores to predict homework time; however, it is also possible that increased homework time improves students’ test scores. Given the cross-sectional data from PISA, we do not attempt to address the nature of the relationship between homework time and test scores. In addition to students’ test scores, we also include a categorical measure of the level of education students report that they expect to complete, distinguishing between those who expect to complete less than a BA, those who expect to complete a BA and those who expect to complete graduate school.

Methods

In ATUS, we estimate a series of nested ordinary least squares (OLS) regression models to examine racial and ethnic differences in daily time spent on homework among high schoolers. Although some have questioned the use of OLS models in the analysis of time use data due to the frequency of zeros, research has found that OLS estimates of time use data are generally robust (Stewart 2013) and these models are commonly used in time use research.⁶ Our first OLS regression model (Model 0) estimates gross differences in homework time across racial and ethnic groups of students with no controls. In Model 1, we add to Model 0 controls for basic demographic characteristics such as gender, age, nativity, survey years, and the time diary day (weekend vs. weekday).⁷ In Model 2, we add family background measures (family structure, parental education, family income, and metropolitan location) to Model 1 to examine the role of family background in explaining racial and ethnic differences in homework time.

In PISA, we first present Model 0, estimating gross differences in homework time across racial and ethnic groups of students. Similar to analysis in ATUS, we then present Model 1, which adds controls for basic demographic characteristics (gender, nativity, and grade relative to the mode), and Model 2 which further adds measures of family background (family structure, parental education, family wealth, and metropolitan location) to Model 1. Next, we present additional models that are not possible in ATUS. Model 3 adds students’ test scores and educational expectations to Model 2 in order to examine the extent to which students’ academic characteristics account for racial and ethnic differences in homework time in addition to family background and students’ demographic characteristics. Finally, we estimate two additional models that take into account between-school differences using PISA. Model 2* repeats Model 2 (including demographic and family background characteristics only) with school fixed effects utilizing school IDs available in PISA. The comparison between Models 2 and 2* informs the extent to which racial and ethnic differences in homework time net of demographic characteristics and family background are robust to school fixed effects. Finally, Model 3* repeats Model 3 (demographic characteristics, family background and academic background) with school fixed effects.

Although these school-fixed effect models do not allow us to identify specific school-level variables that are responsible for racial and ethnic differences in homework time, they provide improved estimates of racial and ethnic differences among students within the same schools by controlling for unobserved school characteristics that may affect students’ homework time. Although we could include some observed school characteristics in the regression model and estimate the relationship between those specific school characteristics and student’s homework time, the strategy could only control for school characteristics that are observed. The school-fixed effect models increase confidence that racial and ethnic differences in homework, if any, are not due to differences in school quality or any other characteristics between schools that minority students typically attend and those attended by White students.

To account for the complex survey design of ATUS and PISA, all analyses use respondent and replicate weights to generate nationally representative estimates with correct standard errors.

Results

Descriptive Portraits

Before examining homework time specifically, in Figure 1 we show the average amount of time per day students spend on different types of educational activities including homework. We categorize educational activities available in ATUS into three categories: (1) daily time spent in high school classes, (2) time spent on homework (described as “research and homework for class or degree, certification, or licensure” in the activity code), and (3) time spent on supplementary educational activities beyond usual classes and study time (such as SAT preparation, participation in academic clubs and additional classes outside of usual school hours). Black students reported spending less time overall on educational activities per day compared with White students – 285 (4 hours 45 minutes) versus 301 minutes (5 hours 1 minute). Meanwhile, Asian students reported spending substantially more time on educational activities per day – 396 minutes (6 hours 36 minutes). Hispanic students spent slightly more time on educational activities per day than White students – 310 minutes (5 hours 10 minutes), owing to greater time spent in classes – 249 minutes (4 hours 9 minutes).

Figure 1. Average daily time spent on educational activities by race/ethnicity (minutes).

Source. American Time Use Survey 2003–2019.

Note. Results for students of “Other” races and ethnicities (including multiracial and other racial/ethnic students) not shown.

* and ^† indicate differences from White students are statistically significant using a two-tailed t test: ^†p < .1. *p < .05.

Looking across the types of educational activities students engaged in, differences by race and ethnicity in overall education time appear to be driven by substantial differences by race and ethnicity in the amount of time students spend on homework. Black students spent 20 fewer minutes per day on homework compared with White students, whereas Hispanic students spent 6 fewer minutes. By contrast, Asian students spent more than twice as much time on homework than White students, averaging 2 hours 14 minutes per day compared with 56 minutes among White students. In addition to spending more time on homework, Asian students also spent more time on other supplementary educational activities (27 minutes) compared with White, Black and Hispanic students (14, 11, and 11 minutes, respectively). However, supplementary educational activities comprised a small proportion of overall educational time among high schoolers.

The relevance of homework time for racial and ethnic differences in educational outcomes is demonstrated not only because it drives racial and ethnic variation in students’ overall educational activities time but also because daily time spent on homework—and racial and ethnic differences in homework time—has changed considerably over the past 15 years. As explained above, the small yearly sample sizes in ATUS preclude a systematic analysis of the temporal trend. However, we can show rough trends in homework time by race and ethnicity by presenting five-year moving average minutes spent on homework. As can be seen in Figure 2, the amount of time that high school students spend on homework per day has increased over time, to varying degrees, across all racial and ethnic groups, perhaps reflecting increased competition for college admission during this time period. Increases in homework time were steepest among Asian students, who already spent considerably more time on homework than other students in the mid 2000s, resulting in an expanding gap in homework time. By contrast, differences in homework time between White, Hispanic and particularly Black students appear to have narrowed over time. The overall increase in homework time with racial and ethnic variation in the magnitude of the increase corroborates this study’s motivation to examine homework time as a relevant channel leading to racial and ethnic differences in educational outcomes. Given the small sample sizes in each year, however, we are cautious in drawing a strong conclusion on the trend.⁸

Figure 2. Average daily time spent on homework by race/ethnicity (minutes), five-year moving average with 95 percent confidence intervals.

Source. American Time Use Survey 2003–2019.

Note. Results for students of “Other” races and ethnicities (including multiracial and other racial/ethnic students) not shown.

Table 1 presents descriptive statistics for all measures included in our regression analysis for both ATUS and PISA data. Reports of average daily time spent on homework were similar across data sets, although estimates for White, Hispanic and especially Asian students tended to be lower in PISA (collected via a stylized survey question) than the time diary estimates from ATUS. By contrast, PISA estimates for Black students were slightly higher than for ATUS. Together, this resulted in smaller racial and ethnic differences in homework time in PISA than in ATUS, although the relative magnitude of these differences between groups was consistent across data sets. These differences in the average homework time between ATUS and PISA are not unexpected and consistent with prior research comparing time diary estimates (as provided in ATUS) with those derived from stylized survey questions about time use (as provided in PISA), which may be subject to greater measurement error (e.g., Schulz and Grunow 2012). Differences in homework time may also relate to age differences in the ATUS and PISA samples: while the ATUS sample includes students aged 15 to 19, PISA students are primarily 15 years old. The relatively large discrepancy in homework time of Asian Americans between ATUS and PISA may be also due to considerable ethnic heterogeneity of Asian Americans (East Asian, South Asian, Southeast Asian, and other Asian). However, information on ethnicity among Asian respondents is not available in PISA and is only available in ATUS from 2013.

Table 1.

Sample Descriptive Statistics.

Variables	White	Black	Hispanic	Asian
ATUS
Daily time spent on homework (minutes)	56.3 (90.9)	36.6* (60.2)	50.0 (77.9)	134.1* (137.3)
Dual parent household (%)	79.5	43.5*	70.4*	80.3
Parent’s highest education: (%)
Less than BA	46.1	64.0*	78.7*	35.6*
BA or higher	51.6	26.2*	18.4*	61.2*
Missing	2.4	9.8*	3.0	3.2
Family income tertile (%)
Lowest	21.9	51.1*	57.7*	25.4
Middle	40.0	29.6*	27.4*	33.6*
Highest	32.1	10.6*	10.6*	33.9
Missing	6.1	8.8*	4.3*	7.1
Metropolitan (%)	79.3	89.7*	92.8*	96.8*
Foreign born (%)	1.6	9.4*	19.5*	30.1*
Female (%)	48.3	45.5	45.2*	50.0
Age	16.3 (1.0)	16.4 (1.0)	16.3 (1.0)	16.4 (1.0)
Weekday diary (%)	71.9	72.1	72.0	72.6
N	4,303	767	1,213	277
PISA
Daily time spent on homework (minutes)	52.5 (44.8)	42.5* (47.3)	49.1 (48.2)	87.3* 66.7)
Family type (%)
Single parent or other household	14.9	32.7*	21.5*	10.9
Dual parent household	78.1	45.7*	68.5*	81.2
Missing	7	21.6*	10*	7.9
Parent’s highest education: (%)
Less than BA	43.3	47.5	72.3*	47.8
BA or higher	47.9	34.8*	17.3*	45.5
Missing	8.8	17.7*	10.4	6.8
Family wealth tertile (%)
Lowest	24.0	49.0*	49.4*	33.1*
Middle	41.4	35.0*	36.7*	34.3
Highest	34.6	16.0*	13.9*	32.6
Metropolitan (%)	65.9	82.9*	90.1*	97.2*
Nativity (%)
U.S. born	97.9	90.8*	80.6*	72.0*
Foreign born	1.1	4.0*	15.9*	23.1*
Missing	0.1	5.2*	3.6*	4.9*
Female (%)	50.3	46.1	49.4	43.5
Grade relative to mode	0.05 (0.48)	0.03 (0.59)	0.05 (0.62)	0.19 (0.62)
Educational expectation (%)
Less than BA	17.3	19.5	26.7*	10.0*
BA	38.7	33.0*	29.5*	27.4*
Graduate school	44.1	47.5	43.8	62.7*
Average test score tertile (%)
Lowest	21.5	62.0*	45.7*	12.6*
Middle	35.1	26.6*	34.4	25.8*
Highest	43.4	11.4*	19.9*	61.6*
N	1,653	395	748	141

Note. Standard deviations shown in parentheses. Results for multiracial and other racial/ethnic students not shown. ATUS = American Time Use Survey; PISA = Program for International Student Assessment.

Asterisks indicate differences from White students are statistically significant using a two-tailed t test (for continuous measures) or a proportion test (for categorical measures): *p < .05.

Family characteristics differed significantly among students of different races and ethnicities. White and Asian students were most likely to live in two-parent households (both 80 percent in ATUS; 78 percent and 81 percent in PISA) while Black students were least likely (44 percent in ATUS; 46 percent in PISA). White and Asian students were also more likely to have a parent with a college degree or higher than Hispanic and Black students. White and Asian students came from more financially advantaged backgrounds with approximately one-third of White and Asian students represented in the top third of the family income (ATUS) and family wealth (PISA) distributions compared with less than one-sixth of Black and Hispanic students. Most students of all races and ethnicities lived in larger, metropolitan communities but White students were least likely to do so (79 percent in ATUS; 66 percent in PISA) and Asian students were most likely (97 percent in both ATUS and PISA). White students were also less likely to have been born outside of the United States (2 percent in ATUS; 1 percent in PISA), particularly compared with Asian students (30 percent in ATUS; 23 percent in PISA).

Additional student characteristics available in PISA indicate that students’ educational expectations were fairly similar across students of different races and ethnicities. Most students reported that they expected to obtain a BA or higher, although Hispanic students were least likely to share this expectation (73 percent compared with 83 percent of White students) and Asian students were most likely (90 percent). Asian students were also most likely to expect to attain a graduate degree (63 percent compared with 44 percent of White students). Despite these similarities in educational expectations, White and especially Asian students tended to have higher average PISA assessment scores compared with Hispanic and Black students.

Regression Results: ATUS

Table 2 presents the results from our OLS models examining differences in daily homework time by race and ethnicity using ATUS data. Model 0, our null model, shows substantial differences in homework time by race and ethnicity. Black and Hispanic students spent 20 and 6 fewer minutes on homework per day compared with White students, respectively (although the difference between Hispanic and White students was only statistically significant at the 0.1 level) and Asian students spent 78 more minutes per day on homework than White students. After including basic demographic controls in Model 1, the difference in homework time between Black and White and Hispanic and White students increased to 21 minutes and 11 minutes, respectively. By contrast, the gap in homework time between Asian and White students reduced by 10 percent to 70 minutes per day. The increased gap in homework time between White and Hispanic students and the reduced gap between White and Asian students reflect the greater share of foreign-born students among Hispanic and Asian students than among White students. Besides race/ethnicity, the results of Model 1 show that foreign-born students spent 23 more minutes per day than their native peers. Female students also spent more time on homework than their male counterparts.

Table 2.

Ordinary Least Squares Regression Results: Daily Homework Time, American Time Use Survey 2003–2019.

Variables	Model 0	Model 1	Model 2
Race (ref = White)
Black	−19.72** (3.40)	−20.59** (3.13)	−11.67** (3.22)
Hispanic	−6.30† (3.25)	−10.92** (3.54)	−1.71 (3.86)
Asian	77.78** (10.65)	70.03** (10.72)	65.57** (10.69)
Other (multiracial and other racial/ethnic)	−12.50† (6.35)	−13.98* (6.14)	−8.03 (6.34)
Dual parent household (ref = single parent or other)			8.10** (2.78)
Parent’s highest education (ref = less than BA)
BA or higher			21.63** (3.25)
Missing			8.95 (6.04)
Family income tertile (ref = lowest)
Middle			5.36 (3.30)
Highest			11.17* (4.51)
Missing			2.99 (4.48)
Metropolitan (ref = not metropolitan)			15.92** (3.08)
Foreign born (ref = U.S. born)		22.64** (6.33)	21.91** (6.21)
Female (ref = male)		18.03** (2.35)	17.58** (2.35)
Weekday diary (ref = weekend diary)		10.99** (2.52)	11.27** (2.47)
Survey year		1.23** (0.26)	0.96** (0.28)
Age (centered at 17)		−2.75* (1.28)	−2.46† (1.25)
Constant	56.29** (1.71)	26.62** (3.10)	−7.10 (4.63)
N	6,791	6,791	6,791
R 2	.04	.06	.09

Note. Standard errors in parentheses.

^†p < .1.

^*p < .05.

^**p < .01 (two-tailed).

The results from Model 2 show that racial and ethnic differences in family background characteristics explain a substantial proportion of the observed gaps in homework time between Black and White students and Hispanic and White students. Together, family structure, parents’ education, family income, and whether students live in a metropolitan area accounted for 43 percent of the difference in homework time between Black and White students identified in Model 1, 84 percent of the gap between Hispanic and White students, and 6 percent of the gap between Asian and White students. This substantial reduction in homework time between Black and White and between Hispanic and White students after taking into account family background characteristics supports our H1.

Because controlling for demographic characteristics in Model 1 increases, rather than decreases, the homework gap between Black and White and between Hispanic and White students, it is useful to compare Model 0 and Model 2 as well. Both basic demographic characteristics and family background factors—family structure, parents’ education, family income, and whether students live in a metropolitan area—together accounted for 41 percent of the gross difference in homework time between Black and White students identified in Model 0, 73 percent of the gap between Hispanic and White students, and 16 percent of the gap between Asian and White students. According to Model 2, Black students spent 12 fewer minutes on homework per day compared with White students, while Asian students spent 66 more minutes on homework per day. The difference in homework time between Hispanic and White students, however, was reduced to −2 minutes and was not statistically significant.

Regression Results: PISA

The patterns observed in ATUS are supported in our analysis of PISA data, presented in Table 3. As noted earlier, although the PISA data show smaller racial and ethnic gaps in homework time, the relative size and direction of the gaps across groups is consistent with ATUS. The results for Model 0 show that Black students spent 10 fewer minutes per day on homework compared with White students and Asian students spent 35 minutes per day more than White students. Hispanic students spent 3 fewer minutes per day on homework than White students; however, this difference was not statistically significant. Controlling for basic demographic characteristics in Model 1 accounted for 6 percent of the differences in homework time identified in Model 0 between Black and White students and also between Asian and White students, reducing differences to 9 and 33 minutes, respectively. The difference in homework time between Hispanic and White students increased slightly to 4 minutes, although this difference remained statistically non-significant.

Table 3.

Ordinary Least Squares Regression Results: Daily Homework Time, Program for International Student Assessment 2012.

Variables	Model 0	Model 1	Model 2	Model 3	Model 2*	Model 3*
Race (ref = White)
Black	−10.02** (3.79)	−9.39** (3.34)	−7.07† (3.55)	1.82 (3.30)	−1.72 (3.94)	3.39 (3.93)
Hispanic	−3.43 (2.87)	−4.34 (2.92)	−1.96 (2.90)	3.60 (2.67)	1.06 (3.68)	3.84 (3.63)
Asian	34.79** (8.48)	32.63** (8.29)	29.33** (8.28)	24.85** (7.53)	21.55* (8.39)	18.83* (8.20)
Other (multiracial and other racial/ethnic)	−0.26 (4.76)	−0.13 (4.63)	2.97 (4.04)	7.80* (3.86)	7.20* (3.45)	9.07** (3.39)
Family type (ref = other)
Dual parent household			1.02 (2.95)	0.79 (2.76)	0.62 (2.72)	0.10 (2.64)
Missing			−6.94† (3.50)	−5.35 (3.60)	−7.07† (3.96)	−6.61 (4.00)
Parent’s highest education (ref = less than BA)
BA or higher			9.19** (2.09)	2.80 (2.14)	2.29 (1.92)	−1.86 (1.88)
Missing			4.49 (3.77)	4.11 (3.56)	1.65 (3.02)	1.78 (2.91)
Family wealth tertile (ref = lowest)
Middle			3.94* (1.80)	2.66 (1.75)	2.54 (1.98)	1.76 (1.94)
Highest			12.52** (2.55)	9.29** (2.56)	7.72** (2.49)	5.84* (2.46)
Metropolitan (ref = not metropolitan)			13.76** (2.86)	10.91** (2.65)	−0.15 (2.25)	−9.77** (2.71)
Highest grade expected to complete (ref = less than BA)
BA				9.64** (2.01)		6.52** (2.04)
Graduate school				13.63** (2.29)		12.48** (2.27)
Average test score tertile (science, math, reading) (ref = lowest)
Middle				13.63** (2.37)		9.77** (2.31)
Highest				29.08** (2.94)		23.61** (2.70)
Nativity (ref = U.S. born)
Foreign born		7.54 (5.13)	8.21 (5.11)	8.68* (4.29)	8.53 (5.13)	8.07† (4.39)
Missing		−1.68 (5.45)	−0.50 (5.91)	4.77 (5.44)	3.17 (6.26)	6.08 (5.95)
Female (ref = male)		14.80** (1.81)	15.24** (1.81)	14.32** (1.72)	13.98** (1.75)	13.41** (1.67)
Grade compared to modal grade in country		11.41** (1.68)	9.34** (1.57)	3.64** (1.37)	6.32** (1.63)	1.57 (1.68)
Constant	52.55** (2.10)	44.50** (1.82)	24.21** (3.05)	4.44 (3.00)	35.87** (3.15)	23.93** (3.22)
School fixed effects?	No	No	No	No	Yes	Yes
N	3,165	3,165	3,165	3,165	3,165	3,165
R 2	.03	.07	.11	.18	.25	.29

Note. Standard errors in parentheses.

^†p < .1.

^*p < .05.

^**p < .01 (two-tailed).

Consistent with the results in ATUS, Model 2 shows that family background, in addition to basic demographic characteristics, accounted for a greater share of the difference in homework time between Black and White and between Hispanic and White students compared with the difference between Asian and White students. Family background and basic demographic characteristics accounted for 29 percent of the difference in homework time between Black and White students observed in Model 0, 43 percent of the difference between Hispanic and White students and 16 percent of the difference between Asian and White students. Net of these characteristics, Black students spent 7 fewer minutes per day on homework compared with White students (statistically significant at the 0.1 level), and Asian students spent 29 more minutes on homework per day. Hispanic students spent 2 fewer minutes on homework than White students but this difference was not statistically significant.

With consistent results across PISA and ATUS for Models 0 to 2, we then extended our analysis in PISA to examine how students’ academic background characteristics and schools may additionally explain racial and ethnic differences in homework time beyond the demographic characteristics and family background factors available in ATUS. Model 3 in Table 3 shows that the difference between Black and White students disappears once students’ (prior) test scores and educational expectations are taken into account in addition to family background, reducing from −7.1 minutes in Model 2 to 1.8 in Model 3. Note that the difference between Hispanic and White students was already small and statistically insignificant in Models 0 to 2. Accounting for the higher test scores and higher educational expectations of Asian students over their White peers reduced the difference in homework time between the two groups of students by 15 percent from Model 2 (29 minutes) to Model 3 (25 minutes). Notably, the difference between Asian and White students was still significant in Model 3. The further reduction in the Asian-White difference in homework time after students’ academic background characteristics, in addition to demographic characteristics and family background, supports our H2. However, we also expected that the difference in homework time between Black and White students would be explained mostly by family background and that the additional portion of the difference explained by students’ academic background would be relatively small. Contrary to this expectation, the results from Model 3 show an even more substantial role for students’ academic background in explaining Black-White differences in homework time than Asian-White differences.

Model 2* is equivalent to Model 2, which controlled for demographic characteristics and family background, but it estimates racial and ethnic differences within the same schools by taking into account school fixed effects. The results of Model 2* show that Black and White students attending the same schools did not differ in homework time, contrasting the result of Model 2, which showed that Black students spent significantly less time on homework than White students, albeit at the 0.1 level, when school fixed effects were not considered. The comparison between Models 2 and 2* indicates that between-school differences between Black and White students, in addition to demographic and family background differences, help explain the Black-White difference in homework time observed in the population. The difference between the two models with and without school fixed effects indicates that the schools attended by Black students tend to be less conducive to homework than those attended by White students, although the school-fixed effect model cannot identify specific school-level factors that are responsible for such between-school differences. Therefore, without controlling for between-school differences, the Black-White difference in homework time is likely overestimated. Model 2* also shows that the gap in homework time between Asian and White students decreased by 27 percent from Model 2, a larger decline than the corresponding change between Model 2 and Model 3 (which additionally accounted for academic background). This result implies that Asian students are more likely than White students to attend schools more conducive to homework time, and therefore ignoring between-school differences overestimates Asian students’ greater time spent on homework compared with their White peers.

Model 3* extends Model 3 by taking into account school fixed effects. The Black-White difference and the Hispanic-White difference in homework time were statistically insignificant in Model 3 with academic background taken into account. Additionally controlling for school fixed effects in Model 3* did not change the corresponding differences between Black or Hispanic and White students. However, the difference in homework time between Asian and White students in Model 3 was further reduced by 24 percent in Model 3*. Even after controlling for students’ academic background, in addition to demographic characteristics and family background, between-school differences still play some role in accounting for Asian students’ greater time spent on homework than their White peers. Overall, the comparison between Models 2 and 2* or the comparison between Models 3 and 3* provides support for our H3: between-school differences partially explain differences in homework time for all racial and ethnic minority groups compared with White students. Our results suggest that school setting is particularly important in explaining the difference in homework time between Asian and White students.

Taken together, the key results in Tables 2 and 3 can be summarized in three points. First, the gap in homework time between Hispanic and White students was not substantial to the extent that it was not significant when family background differences between the two groups were held constant. Second, the difference in homework time between Black and White students, which remained significant even after controlling for family background, became negligible and statistically insignificant once either students’ academic background or school fixed effects were considered in addition to family background. This suggests that differences in academic characteristics and school environment between Black and White students play an important role in explaining the gap in homework time between Black and White students. Finally, Asian students spent significantly more time on homework than their White counterparts even after controlling for family background. Both students’ academic background and school fixed effects partially accounted for the difference in homework time between Asian and White students remaining after family background taken into account; however, a substantially difference in homework time remained.

Supplementary Analysis: Robustness of Results

We conducted several supplementary analyses to test the robustness of our results. To examine whether our results are driven by racial and ethnic differences in non-participation in homework, we re-ran our main models limiting our sample to those who reported spending at least 10 minutes of homework per day in ATUS and PISA. These results, presented in Appendix Tables A1 and A2, are similar in magnitude and statistical significance to our main results except for a loss of statistical significance for the Black-White difference in PISA due to the smaller sample size. We also tested several models that are designed to better handle data with a large number of zero observations (e.g., students completing zero minutes of homework) such as Tobit and zero-inflated count models (Poisson and negative binomial regression) (Long and Freese 2003; Stewart 2013). Our main findings and conclusions are robust to these alternative specifications (although the magnitude of racial and ethnic differences and the extent to which each factor explained such differences varied across models), and these results are available upon request.

Second, reflecting the small share of Asian American student population in the United States, many schools sampled in PISA included no Asian students in their PISA student samples. To assess whether the inclusion of schools with no Asian students affects the Asian-White differences in homework time revealed in Table 3, we re-ran all PISA models restricting our sample to schools with at least one Asian student represented in the student sample, shown in Appendix Table A3. With this restriction, the sample size was reduced from 3,165 to 1,141 and the number of schools reduced from 161 to 58. However, the changing patterns of the Asian-White difference in homework time across models, observed with restricted samples, were consistent with those reported in Table 3, although the gap in homework time between Asians and White students was somewhat smaller in the restricted sample than in the original sample.

Discussion

This study sought to describe patterns of racial and ethnic differences in homework time among U.S. high schoolers and examine how family background, students’ academic characteristics and school setting account for these differences. Consistent with prior research, our analysis of time diary data from ATUS finds substantial variation in daily homework time among U.S. high schoolers by race and ethnicity (e.g., Hansen and Quintero 2017). White students reported spending an average of 56 minutes per day on homework compared with 37 minutes among Black students and 50 minutes among Hispanic students. Meanwhile, Asian students spent substantially more time per day on homework than students of any other racial or ethnic group at 134 minutes per day. Furthermore, in comparing the total amount of time students spend on academically oriented activities per day (including attending school classes, completing homework and participating in extracurricular educational activities such as prep-courses for standardized tests), we find that racial and ethnic differences in homework time drive overall differences in total daily educational time among U.S. teens. The analysis based on five-year moving average estimates suggests that all racial and ethnic groups of students have increased time use for homework. However, with some different trends by race and ethnicity: differences between White, Hispanic and particularly Black students have narrowed since the mid-2000s while differences between White and Asian students have widened. Due to small yearly sample sizes, our trend analysis is only suggestive and further analysis is needed.

The results from regression analysis of ATUS and PISA data reflect prior research findings that family background characteristics such as family structure, parental education, and family economic resources matter for racial and ethnic differences in homework time (Ainsworth-Darnell and Downey 1998; Kofman and Bianchi 2012). However, we find that these factors do not fully explain racial and ethnic differences in homework time, particularly when comparing homework time between White and Black students and between White and Asian students.

Examining the role of students’ academic characteristics and school setting, a key contribution of this study provides greater insight into racial and ethnic difference in homework time. Taking into account racial and ethnic differences in students’ prior academic achievement and their own expectations regarding academic futures further reduce Black-White and Asian-White differences in homework time. So too does school environment. After accounting for students’ academic characteristics or school fixed effects, differences in homework time between Black and White students are statistically insignificant. Our models are less successful in explaining racial and ethnic differences in homework time between Asian and White students. Although controlling for family background, students’ academic characteristics, and school fixed effects partially account for the difference in homework time between Asian and White students, substantial gaps in homework time remain.

Taken together, these results demonstrate how racial and ethnic inequalities at a variety of levels combine to influence homework behavior. When we compare homework time among Black, Hispanic, Asian, and White students who are similar in terms of family background, academic characteristics, and school setting, Black, Hispanic, and White students do not differ substantially in their homework time, while Asian students still spend more time on homework per day. These findings help to illuminate previous studies that have identified an “Asian advantage” across a wide range of academic outcomes (see, e.g., Hsin and Xie 2014). The greater daily time spent on homework among Asian students found in this study may be a potential mechanism through which this academic advantage may occur. Furthermore, these findings may also reflect strategies or cultural ethea among marginalized racial and ethnic minority families and communities that emphasize educational success as a tool for social mobility and over-coming adverse effects of racism and discrimination (Bauman 1998; Merolla 2013; Xie and Goyette 2003). Prior research has highlighted the importance of mentoring for increased homework time (Gaddis 2012). Understanding the extent to which availability of mentors for youth in Asian American communities contributes to greater time spent on homework among Asian students is a potentially fruitful area of future research on racial and ethnic difference in homework time.

A key limitation of this study is that the data available only allow for consideration of the quantity of homework time students complete, while other more qualitative measures of homework time are likely to be important for capturing the quality and nature of homework. Future qualitative work examining other characteristics of homework time such as the quality of homework completed and how well homework assignments are integrated into the curriculum would further illuminate racial and ethnic differences in homework. In addition, while this study aimed to examine two key explanations for racial and ethnic differences in homework time that have thus far been absent from the literature—students’ academic characteristics and school setting—there are further explanations that warrant attention. In particular, peer group effects such as network effects, competition and pressure, and achievement norms may influence students’ homework behavior and may differentially affect students of different races and ethnicities (e.g., Jiménez and Horowitz 2013; Workman 2020). Finally, the cross-sectional nature of the data analyzed in this study limits our ability to systematically examine how prior academic achievement shapes homework time. Future research utilizing longitudinal data with better measures of prior academic performance would help disentangle the nature of this relationship.

Despite these limitations, the current study updates existing research on racial and ethnic differences in academically oriented time, with a particular focus on homework, and extends our understanding of the factors that drive observed racial and ethnic differences in homework time. Given that homework is often interpreted as an indicator of academic effort or commitment that is tied to educational performance, the results of this study have important implications for understanding sources of racial and ethnic gaps in homework time and their contribution to racial and ethnic differences in educational performance and attainment. Our findings suggest that family background, students’ academic background, and school setting each play somewhat varying roles in explaining homework time of Black, Hispanic, and Asian students compared with their White counterparts. Differences in family background explain a substantial portion of Black-White and Hispanic-White differences in homework time but they play a relatively small role in accounting for Asian-White differences. Instead, differences in academic background and school setting between Asian and White students explain a greater proportion of this difference. These factors, in addition family background, also help explain the difference in homework time between Black and White students, but appear less important in explaining Hispanic-White differences in time after accounting for differences in family background.

The varying roles of each factor for different racial and ethnic groups suggest that efforts to reduce racial and ethnic differences in homework time—and thus ultimately to reduce differences in educational performance and attainment—should take a multidimensional approach that acknowledges the individual, family, and structural factors that shape racial and ethnic differences in homework behavior. Moreover, different racial and ethnic groups of students may differ not only in their time spent on homework but also in the extent to which they benefit from the same amount of homework time for improved educational outcomes (cf. Cooper et al. 2006; Daw 2012). A comprehensive study of racial and ethnic variation in both determinants and consequences of homework time is warranted.

Biographies

Allison Dunatchik is a PhD candidate in Sociology and Demography at the University of Pennsylvania. Her research interests include family processes and social stratification, with a particular focus on how inequalities within and between families are reflected in how individuals spend their time.

Hyunjoon Park is Korea Foundation Professor of Sociology and Director of the James Joo-Jin Kim Center for Korean Studies at the University of Pennsylvania. He research interest includes educational inequality, social stratification, family, and transition to adulthood.

Appendix

Robustness Analysis

Table A1.

Ordinary Least Squares Regression Results: Daily Homework Time among Those Reporting 10+ Minutes per Day, American Time Use Survey 2003–2019.

Variables	Model 0	Model 1	Model 2
Race (ref = White)
Black	−25.74** (5.71)	−26.62** (5.41)	−15.35* (5.91)
Hispanic	−11.16* (4.75)	−12.47* (5.38)	0.69 (6.13)
Asian	67.46** (12.12)	62.45** (12.57)	61.84** (12.42)
Other (multiracial and other racial/ethnic)	−3.35 (11.08)	−6.45 (11.04)	1.49 (10.62)
Dual parent household (ref = single parent or other)			10.60* (4.99)
Parent’s highest education (ref = less than BA)
BA or higher			19.50** (5.03)
Missing			13.15 (11.44)
Family income tertile (ref = lowest)
Middle			12.63* (6.05)
Highest			15.53* (7.26)
Missing			10.86 (7.39)
Metropolitan (ref = not metropolitan)			13.29* (6.55)
Foreign born (ref = U.S. born)		7.97 (7.99)	7.53 (8.01)
Female (ref = male)		18.44** (4.22)	18.60** (4.18)
Weekday diary (ref = weekend diary)		−30.93** (4.34)	−29.26** (4.29)
Survey year		1.41** (0.45)	1.12* (0.50)
Age (centered at 17)		1.00 (2.17)	1.76 (2.15)
Constant	119.18** (2.74)	122.23** (6.00)	80.72** (8.87)
N	2,827	2,827	2,827
R 2	.04	.07	.10

Note. Standard errors in parentheses.

^†p < .1.

^*p < .05.

^**p <.01 (two-tailed).

Table A2.

Ordinary Least Squares Regression Results: Daily Homework Time among Those Reporting 10+ Minutes per Day, Program for International Student Assessment 2012.

Variables	Model 0	Model 1	Model 2	Model 3	Model 2*	Model 3*
Race (ref = White)
Black	−6.75† (3.63)	−7.20* (3.35)	−5.99 (3.81)	1.61 (3.50)	−2.52 (4.51)	2.38 (4.50)
Hispanic	−1.97 (2.80)	−2.98 (2.88)	−0.24 (2.92)	4.49† (2.70)	2.44 (3.75)	4.93 (3.72)
Asian	36.80** (7.48)	34.47** (7.43)	32.41** (7.26)	28.23** (6.75)	26.02** (8.34)	23.35** (8.20)
Other (multiracial and other racial/ethnic)	2.87 (5.21)	2.96 (5.07)	5.57 (4.56)	9.43* (4.27)	9.73* (4.20)	11.08** (4.11)
Family type (ref = other)
Dual parent household			1.10 (3.31)	0.57 (3.22)	0.48 (3.19)	−0.24 (3.13)
Missing			−6.57 (3.96)	−5.74 (4.13)	−7.96† (4.45)	−7.94† (4.51)
Parent’s highest education (ref = less than BA)
BA or higher			8.74** (2.14)	2.55 (2.20)	2.90 (2.11)	−1.27 (2.14)
Missing			7.57† (4.15)	6.84† (3.95)	4.82 (3.38)	4.55 (3.31)
Family wealth tertile (ref = lowest)
Middle			3.33 (2.13)	2.39 (2.11)	1.55 (2.39)	1.01 (2.35)
Highest			9.67** (2.62)	7.12** (2.68)	4.12 (2.65)	2.80 (2.68)
Metropolitan (ref = not metropolitan)			11.81** (2.97)	9.10** (2.74)	−7.35* (2.98)	−18.13** (3.67)
Highest grade expected to complete (ref = less than BA)
BA				8.76** (2.24)		5.15* (2.24)
Graduate school				13.57** (2.85)		11.34** (2.76)
Average test score tertile (science, math, reading) (ref = lowest)
Middle				11.47** (2.64)		8.57** (2.72)
Highest				26.82** (2.93)		22.54** (2.89)
Nativity (ref = U.S. born)
Foreign born		7.25 (5.57)	7.55 (5.56)	7.99† (4.72)	8.72 (5.85)	8.19 (5.13)
Missing		0.51 (6.07)	1.88 (6.26)	6.91 (5.85)	5.16 (6.71)	8.08 (6.32)
Female (ref = male)		12.03** (1.88)	12.73** (1.86)	11.92** (1.76)	11.82** (1.83)	11.39** (1.77)
Grade compared to modal grade in country		11.10** (1.82)	9.43** (1.70)	4.28** (1.56)	5.80** (1.91)	1.45 (2.04)
Constant	60.22** (1.86)	53.08** (1.67)	34.34** (3.28)	15.76** (3.58)	49.60** (3.57)	38.95** (3.81)
School fixed effects?	No	No	No	No	Yes	Yes
N	2,653	2,653	2,653	2,653	2,653	2,653
R 2	.03	.06	.09	.15	.23	.27

Note. Standard errors in parentheses.

^†p < .1.

^*p < .05.

^**p < .01 (two-tailed).

Table A3.

Ordinary Least Squares Regression Results: Daily Homework Time among Students in Schools with at Least One Asian Student Represented in the Student Sample, Program for International Student Assessment 2012.

Variables	Model 0	Model 1	Model 2	Model 3	Model 2*	Model 3*
Race (ref = White)
Black	−5.01 (4.87)	−7.92 (5.17)	−4.78 (6.15)	5.17 (6.03)	−2.82 (7.01)	3.46 (7.01)
Hispanic	−8.00 (4.86)	−11.18* (5.14)	−4.87 (5.51)	2.56 (5.16)	−2.88 (6.99)	−0.11 (6.88)
Asian	26.20** (8.34)	23.05** (8.58)	23.86** (8.51)	19.18* (7.85)	19.08* (9.30)	15.22† (9.04)
Other (multiracial and other racial/ethnic)	4.60 (6.93)	3.61 (6.53)	6.93 (6.79)	13.50* (6.57)	9.29 (6.50)	12.90* (6.03)
Family type (ref = other)
Dual parent household			0.51 (6.23)	−0.05 (5.41)	0.84 (5.75)	0.27 (5.24)
Missing			−4.43 (6.95)	−2.83 (7.43)	−6.27 (8.27)	−5.67 (8.39)
Parent’s highest education (ref = less than BA)
BA or higher			11.14* (4.32)	1.81 (4.25)	3.31 (3.79)	−2.72 (3.60)
Missing			5.12 (8.28)	3.54 (7.81)	1.94 (6.27)	1.08 (5.93)
Family wealth tertile (ref = lowest)
Middle			4.96 (3.56)	4.11 (3.46)	4.02 (3.94)	4.30 (3.76)
Highest			11.16* (4.50)	8.28† (4.47)	6.41 (4.42)	5.54 (4.17)
Metropolitan (ref = not metropolitan)			4.72 (6.46)	4.33 (5.52)	−22.20** (3.05)	−23.45** (3.06)
Highest grade expected to complete (ref = less than BA)
BA				11.32** (4.17)		5.46 (3.73)
Graduate school				15.05** (4.58)		12.35** (4.23)
Average test score tertile (science, math, reading) (ref = lowest)
Middle				18.45** (5.07)		13.95* (5.29)
Highest				38.16** (5.05)		31.72** (4.66)
Nativity (ref = U.S. born)
Foreign born		14.09† (7.23)	16.10* (7.55)	18.87** (5.68)	16.38† (8.38)	17.38* (6.59)
Missing		5.71 (9.44)	9.74 (10.26)	15.23† (9.05)	16.19 (10.44)	19.05† (9.66)
Female (ref = male)		18.91** (3.55)	19.63** (3.68)	19.50** (3.44)	19.09** (3.75)	19.29** (3.58)
Grade compared to modal grade in country		13.23** (3.32)	12.70** (3.33)	6.18* (2.63)	10.04** (3.23)	4.17 (3.17)
Constant	61.14** (3.43)	50.07** (3.20)	32.99** (6.67)	3.53 (6.92)	47.34** (3.42)	22.15** (4.81)
School fixed effects?	No	No	No	No	Yes	Yes
N	1,141	1,141	1,141	1,141	1,141	1,141
R 2	.03	.09	.11	.19	.24	.29

Note. Standard errors in parentheses.

^†p < .1.

^*p < .05.

^**p < .01 (two-tailed).