Notes on: Dunatchik, A. & Park, H. (2022) Racial and Ethnic Differences in Homework Time among U.S. Teens Sociological Perspectives 1–25 DOI: 10.117 7/07311214221101422

Dave Harris

[A very thorough classic quantitative methods piece. Shows the art of doing 'positivist' work, devising ways round operational problems and making the most of what data you can get, formulatingmodelas and hypotheses and testing them then checking for robustness]

There has been intensified competition for college admissions among US teens  who spend more time on educational activities including homework. There is an American Time Use Survey (ATUS) apparently and a Programme for International Student Assessment (PISA) which looks at comparisons of race and ethnicity. Apparently family background accounts for differences in homework time between Hispanic and White and partly explains the difference between Black and White. Academic characteristics or school effects explain the remaining gaps.

Historically, teens in the US have spent less time on schoolwork than their counterparts in other parts of the world especially East Asia, and research has tended to focus on participation in nonacademic extracurricular activities such as sports, music and clubs as well as involvement in paid work. However recently they are spending more time on homework and are involved in commercial prep services and private tutoring, spending less time unpaid work and and socialising. This is consistent with more intensified competition for college admissions and increased parental time and investment in children's education.

There might be racial and ethnic inequalities in engagement with private supplementary education, but there could also be inequalities in homework time. Those that exist usually distinguish only between non-Hispanic White and all other ethnic minority groups combined, while other studies take race and ethnicity as controls in multivariate analysis to look at factors such as paid work time.

Consistent evidence, however suggest a significant relationship between the amount of time spent on homework and improved educational outcomes, for high school students although not necessarily for younger ones, although there is some debate about how big this effect is [references p.2]. The two surveys mentioned in the abstract provide data on average daily educational time, including time spent in class on homework and and other educational activities such as 'SAT preparation' and how this varies. There are 'substantial racial and ethnic differences'. They go on to examine how diverse aspects of family background including SES, family structure and location, and students' academic background, test scores and expectations and the demographic characteristics, age, gender and 'nativity' contribute to these differences. We have to use both measures to do this. School effects are examined because they might vary including homework practices [not sure what they do exactly — examined to see if schools are successful in promoting ethnic minority students?] They take these other factors to go back to try and explain successful students as identified in the first study which focused only on demographic characteristics and family background factors. Homework time might not have a uniform relationship, but it seems to be generally positive and a possible explanation of racial and ethnic differences, especially when examining Asian-Americans, their effort and work ethic. One particular study [cited on p3] showed that African-American sophomores spent less time on homework than their non-Hispanic White counterparts and Asian American students spent most time, family income occupation in education partly accounted for this difference so did family structure, but this study did not examine homework time specifically nor consider school effect in the form of school relationships between race and homework time in each school. The sample might be limited as well. Another study found that Asian American high school students spend nearly 2 hours per day on homework, almost twice the amount spent on White and Hispanic students, and Black students only spent 30 minutes per day, but this study did not examine factors. A third study used data from the time use survey and found that native born Hispanic and native born Black adolescents spent 'significantly less time studying than their non-Hispanic nativeborn White peers' but that the differences disappeared after taking family SES and demographic factors into account. The most time studying was by Asian immigrant adolescents and this did not substantially change even after controlling for SES and demographic variables. Overall, these studies do show that Black and to a lesser extent Hispanic students spend less time on homework than White peers, but Asian American students spend most time of all, and that this can be seen as independent to some extent of SES and demographic background.

We also need to try and pin down students academic characteristics, however, their prior performance and their educational expectations. Some studies show this might particularly favour Asian-Americans and their efforts and persistence. These are positively associated with homework time, especially when they expect to receive college degree. There is also 'between-school variation' (4), with significant differences between the schools attended by White students and racial and ethnic minority students — 'socio-economic contexts, overall academic performance, teacher experience and resources… Academic climates, including the importance the schools and teachers attached homework and homework related practices (e.g. the frequency and quantity of homework assignments)'. Peer groups also have an effect on academic climates, and some can reinforce academic norms, while others can produce effects despite individuals characteristics and expectations.

The school level variables are hard to estimate in their effect, and to separate them from students and characteristics. They have used 'school – fixed effect models' to estimate these effects.

They proceeded by drawing on the existing two surveys to examine racial and ethnic differences in daily educational time, and how they have changed over the last decade. Then they use regression analysis to identify the factors and account for any differences, especially in homework time. They focus especially on students academic characteristics and school setting, to supplement the previous explanations which relied on family background and demographic factors. They then formulated three hypotheses:

1.Racial and ethnic differences in family background characteristics (family structure, family resources and location) 'explain differences in homework time between Black and White and between Hispanic and White students
2. Differences in family background alone do not sufficiently account for differences in homework time and we need further explanations 'differences in students' academic characteristics between Asian and White students'
3. 'Between school differences partially explain differences in homework time for all racial and ethnic minority groups compared with White students' (5)

They used American Time Use Survey (ATUS) and the US version of PISA 2012. The first has diary estimates of time use, family and household characteristics but nothing on academic background or schools, while the second does have school-related information and students' academic characteristics but only 'stylised survey estimates' of time spent on homework. The first is a nationally representative cross section and one (15 year old at least) individual from each household completes a time diary. Yearly samples especially for Asians are rather small so they pooled the data across 16 years of results. PISA includes test results on basic subjects and surveys family background study habits, attitudes towards learning and the principles of participating schools and their learning environments. It's quite a large survey across 65 countries — 4978 students [ATUS had 6791 students, although they excluded some and ended with 3165]

They wanted to measure the number of minutes respondents spent doing homework per day. The surveys measured this differently — the first one is a better measure because it used time diary data, while the second one reports average amounts spent on homework each week. Categories of race and ethnicity also differ — the PISA data has six categories (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian, multiracial, and other — they collapse the last two into one other category).ATUS has a five category variable, including a rather heterogeneous other category. They control for gender and whether the respondent was born in the USA. Family background measures include a binary of whether respondents live in a two-parent household, and whether they have at least one parent with a bachelors degree. They measured family income. Missing information was fairly small. They had some wealth data and data about the size of the community in which people lived, again a binary, Metropolitan or not. For academic characteristics and school IDs, they used PISA assessment scores which were classified into 'tertiles' [so were other variables like income and wealth], and average scores for maths science and reading as a measure of prior academic achievement. They use the scores 'to predict homework time' [?], So they cannot use PISA data to get at the relationship between homework time and test scores. They also included a measure of the level of education that students expected to complete, and whether this was at degree level.

[Classic workmanlike way to manage this data, lots of combinations and simplifications are needed]

For the first survey they did some regressions to estimate racial and ethnic differences in time spent on homework. The particular model was 'nested ordinary least squares' and this has been criticised for time use data because there are a lot of zeros. Nevertheless they began with gross differences then added various controls, like basic demographic characteristics, 'gender, age, nativity, survey years and the day of the time diary (weekend versus weekday)' (7). In the second model they added family background measures '(family structure, parental education, family income, and metropolitan location)' to examine the role of family background. For the PISA data they did something similar estimating gross differences and then adding basic demographic characteristics and further measures of family background. However they had some additional models with this survey and added 'students to schools and educational expectations' to the third model. Then they had two additional models focusing on between school differences. They  are able to modify their own model 2 with school fixed effects to try to see the extent to which racial and ethnic differences in homework time are 'robust to school fixed effects' even after we've allowed for demographic characteristics and family background. They were able to modify their own model 3 with school effects similarly. They weren't able to identify specific school level variables but they did get 'improved estimates of racial and ethnic differences among students within the same schools by controlling for unobserved school characteristics' these characteristics 'may affect students homework time' however, this can obviously only deal with characteristics that are observed. Overall, including these effects 'increase confidence that the 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'. [This is odd though and seems to be contradicted below --maybe they are talking about diferent school characteristics in eachcase?]

They then did something clever 'to generate nationally representative estimates with correct standard errors' (8)

They offer 'descriptive portraits'[for idiots like me]. They categorised educational activities from the first survey into three — daily time spent in high school classes, time spent on homework, specifically '"research in homework for class or degree certification or licensure"', and time spent on supplementary educational activities '(such as SAT preparation, participation in academic clubs and additional classes)'. Black students spent less time overall compare with White students — 4 hours 45 minutes compared to 5 hours and 1 minute, while Asian students spent 6 hours and 36 minutes and Hispanic 5 hours and 10 minutes, although this was down to greater time spent in classes.

There were 'substantial differences' in the time spent on homework. Black students spent 20 fewer minutes a day on homework compared with White students, whereas Hispanic students spent 6 fewer minutes, and Asian students spent twice as much time on homework, 2 hours 14 minutes a day compared with 56 minutes for White students. Asian students also spent more time on supplementary activities, although this was only a small proportion of time overall

This proportion spent on homework has' changed considerably over the past 15 years', for all racial and ethnic groups, probably after increased competition for college admission, but increases in homework time 'was steepest among Asian students'. Differences in homework time between White, Hispanic and Black students narrowed over time. Small sample sizes make generalisations difficult, though

Considering data from both surveys produces similar reports of average daily time, although estimates tended to be lower than the time diary records, except for Black students, and this resulted in smaller racial and ethnic differences in home time. However the relative magnitude was consistent. [There may be other reasons for smaller differences, including measurement errors and ethnic heterogeneity, which only appear here and are not discussed in the other data! (10)].

Family characteristics were significantly different, with White and Asian students more likely to live in two-parent households and Black students least likely [below 50% in both surveys]. White and Asian students were more likely to have a parent with a college degree or higher, to come from more financially advantaged backgrounds, to be least likely to live in large metropolitan communities, and least likely to be born outside of the USA especially compared with Asian students. Apart from that their education experiences were fairly similar across racial and ethnic characteristics. Most expected to obtain a BA, although Hispanic students less likely to [still 73%, but compared to 83% of White students, Asian students 90%].

More specific testing of the models and the first survey showed substantial gross differences in homework time [1st model] — Asian students 78 more minutes a day than White students, Black and Hispanic students 20 and  6 fewer minutes on homework per day. By the time they had added the variables detailed in the second model, the basic demographics, they found the differences between Black and White and Hispanic and White increased to 21 minutes and 11 minutes respectively, but the gap between Asian and White reduced by 10% to 70 minutes. This may be starting to measure the impact of foreign-born students who tend to spend more minutes per day than their native peers. Female students do as well. The model including racial and ethnic differences in family background explained 'a substantial proportion of the observed gaps in homework time' — taken together 43% of the difference in homework time between Black and White students, 84% between Hispanic and White, and 6% of the gap between Asian and White [this is the one that takes to into account family structure, parental education, family income, and living in the metropolitan area]

Testing models with the second survey, PISA, there were smaller racial and ethnic gaps in the first place, although the trends were consistent. The gross model showed that Black students spent 10 minutes less per day on homework compared with White, and Asian students 35 minutes a day more than White students, Hispanic students 3 minutes fewer than White, although this was not statistically significant. Adding demographic characteristics explain 6% of the differences in homework time between Black and White and reduced differences to 9 and 33 minutes between Black and White and Asian and White. Adding family background together with demographic characteristics accounted for 29% of the difference in homework time between Black and White, and 43% of the difference between Hispanic and White,  16% the difference between Asian and White — Black students spent 7 minutes less on homework compared with White students and Asian students 29 more minutes on homework, Hispanic students 2 fewer minutes on homework than White students, but not statistically significant.

They then went on to look at school effect and academic background characteristics. Here, the full model showed that the difference between Black and White students 'disappears once students experience prior to schools and educational expectations are taken into account in addition to family background' (15). There is still a difference between Asian and White students, but it is reduced by 15%, from 29 minutes to 25 minutes, although this is still significant, and supports the second hypothesis, but the result for Black and White students did contradict their expectations and shows a more substantial role than they expected for students' academic background and explaining Black and White differences.

The modified models take into account school fixed effects and do not show a difference in homework time between Black and White students, [so there is a school effect?] — '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' (15), so they expect these factors to be 'likely overestimated'. Gaps in homework time between Asian and White students also decreased, implying 'that Asian students are more likely than White students to attend schools more conducive to homework time' [in a real weasel here — 'ignoring between-school differences overestimates Asian students' greater time spent on homework compared with their White peers' (16)].

The third modified model produced statistically insignificant differences between Black and White and Hispanic and White homework times and did not change the differences between Black Hispanic and White students. It didn't reduce the difference between Asian and Whites by a further 24%. Overall, they think there is support for the third hypotheses, that 'between school differences partially explain differences in homework time for all racial and ethnic minority groups compared with White students… School setting is particularly important in explaining the difference in homework time between Asian and White students'.

Overall they think that the gap in homework time between Hispanic and White students was 'not substantial', even not significant when family background differences were held constant. The difference in homework time between Black and White students remains significant after controlling for family background but became negligible and statistically insignificant once 'either students' academic background or school fixed effects were considered in addition to family background… Differences in academic characteristics and school environment between Black and White students play an important role in explaining the gap in homework time'. Asian students spend 'significantly more time on homework than their White counterparts even after controlling for family background. Both students 'academic background and school fixed effect partially accounted for the difference… However substantially difference [sic] in homework time remained'.

They conducted several 'supplementary analyses' to test the results and how run they were, rerunning their main models limiting the sample to those who were active for at least 10 minutes of homework per day. The results were similar. They tested several models better at handling data with large numbers of zero observations, and again found robust findings and conclusions although differences in the magnitude of their variations. They then reran all the PISA models choosing only schools with at least one Asian student represented in the students sample, again find inconsistencies with those data they have reported.

In conclusion, they found they were substantial variations in daily homework time, summarised as 56 minutes a day for White students, 37 minutes for Black, 50 minutes for Hispanic, 134 minutes for Asians. There are also differences in time spent on daily educational activities. All groups have increased the time spent on homework, and differences between ethnic minorities have narrowed, but those between White and Asian students have widened. However their analysis is 'only suggestive'.

Their regression analyses do reflect prior research findings and indicate that family background characteristics also matter and explain differences in homework time, but not fully especially when comparing the differences between White and Black and White and Asian. Looking at academic characteristics and school setting further reduces the differences in homework time so the school environment, and differences in time between Black and White are reduced to statistical insignificance, although differences remain between Asian and White students and 'substantial gaps in homework time remain'

So a number of factors at a variety of levels influence homework behaviour. When we compare those who are similar in terms of 'family background, academic characteristics and school setting, Black, Hispanic and White students did not differ substantially in their homework time, while Asian students still spend more time on homework per day' (17 ), evidence of a potential mechanism for the '"Asian advantage"'. In my also be evidence for the strategies cultural ethos among marginalised ethnic communities that see educational success as a main tool for social mobility to overcome racism. They might also indicate the successful mentoring ethnic minority youth.

[So hang on,the ony real differences after we have taken into account all the background factors are between Asians and the rest?]

There are limitations. Qualitative measures of homework time might be more important, including how well they are integrated into the curriculum. People might benefit differently from the same amount of homework and there may be racial and ethnic variations There are additional variables such as 'network effects, competition and pressure, and achievement norms'. Cross-sectional approaches have made it difficult to focus on specific factors such as prior academic achievement. Nevertheless, early research has been updated and homework is important as an indicator of academic effort or commitment. It might be as important as differences in family background, especially for Asian-White differences. Reducing racial and ethnic differences in educational performance 'should take a multidimensional approach'.