Eren and Henderson (2011) found, using within-student, within-teacher variation in assigned homework for US eighth graders (NELS:88), that additional mathematics homework substantially raises mathematics test scores while additional homework in science, English, and history has little to no effect on the corresponding scores, and that black students and students with low parental education realize no measurable return to additional math homework. Later literature has substantially examined the adjacent questions of other grade levels, newer data, and stronger identification (high-school time-diary evidence, elementary twin fixed-effects evidence, cross-country within-student designs, a field experiment, and causal machine learning). What remains open is the mechanism core: why the returns to homework time differ by subject and by student group — testing with direct measures of homework content and cognitive demand whether the math advantage reflects problem-solving rather than memorization work, and identifying why disadvantaged-group students show null returns (completion, parental support, or time constraints).

Background

Economists model student achievement with an education production function in which inputs such as class time, teacher quality, and study time combine to produce test scores; homework time is one input whose causal effect is hard to measure because students and teachers who assign or complete more homework differ in unobserved ways (a selection problem). Eren and Henderson (2011, Economics of Education Review 30:950–961) addressed this for US eighth graders in the National Education Longitudinal Study of 1988 (NELS:88), a nationally representative longitudinal study that tested eighth graders in math, science, English, and history and surveyed two of each student's subject teachers. Their identification strategy exploited the matched-pair feature of these data — each student was observed in two of the four subjects, and the surveyed teachers usually taught multiple classes — so that first-differencing a student's two contemporaneous subject observations yields within-student, within-teacher comparisons that largely eliminate unobserved student and teacher traits; these estimates were supplemented by instrumental-variable (IV) estimates in which teacher-reported weekly assigned homework hours (the treatment variable, taken from the subject-specific teachers' reports rather than from students' own reports) are instrumented with a teacher-reported indicator for whether the textbook provides good suggestions for homework assignments, interacted with subject fixed effects. Their headline result is a subject-specific asymmetry: math homework has a large, statistically meaningful effect on math scores, while additional homework in science, English, and history shows little to no effect on those subjects' scores; effects are essentially null for black students and for students whose parents have low education. In the authors' version of the article (IZA Discussion Paper 5547), the conclusion raises a single forward-looking research item — the caveat that the two-decades-old cohort may not generalize to the current educational environment ('Perhaps future work can re-approach this issue in more detail') — together with the policy caution that additional homework may increase the relative performance gap for black students and, similarly, for students from less educated families. The mechanism questions the paper leaves open sit in its results and heterogeneity discussions rather than in a concluding list: the authors conjecture that math homework requires solving problems and not simple memorization while homework in the other subjects contains larger percentages of 'memorizing exercises', but NELS:88 holds no direct measures of homework content or cognitive demand, so they could not test it; for the subgroup nulls they offer candidate explanations — the largest discrepancy among racial groups between math homework assigned and completed for black students (1.11 hours per week) and a large discrepancy for students whose parents have less than a high school diploma (2.40 hours assigned versus 1.03 hours completed), difficulty obtaining help on assignments from parents, and possibly hitting a time constraint or 'give-up' limit — none of which their data can identify. The paper also acknowledges, as internal caveats, that the within-teacher variation in assigned homework is not argued to be inherently random, that its IV estimates are imprecise, and that there is no strong evidence of nonlinearities. Subsequent work has addressed several adjacent questions — other grade levels, newer cohorts, and stronger identification: Kalenkoski and Pabilonia (2017) used time-diary data from the Panel Study of Income Dynamics to study high school homework and found positive effects on GPA and college attendance for boys; Jerrim, López-Agudo, and Marcenaro-Gutiérrez (2019) used twin fixed-effects models (comparing identical twins in the same Spanish elementary classrooms, which controls for family and genetic factors) and found no association between homework time and achievement in elementary grades; Grodner and Rupp (2013) ran a field experiment on homework in a college course; McJames, Parnell, and O'Shea (2024) applied causal machine-learning estimation (flexible statistical-learning estimators adapted to recover causal effects rather than plain associations) and found benefits from short, frequent homework in mathematics and science; Flannery, Palcic, and Clavel (2026) exploited within-student, between-subject variation in homework assignment across 23 countries in TIMSS 2019 (the Trends in International Mathematics and Science Study, a cross-national assessment) and found a small positive effect driven by assignment frequency rather than volume; and Fan and colleagues (2017) meta-analyzed roughly 30 years of homework–achievement studies in math and science. Daw (2012) showed with US secondary-school panel data that higher-income students gain more from homework time in all subjects except history, especially in math, implying more homework can widen socioeconomic achievement gaps. What no credible causal study has yet delivered is a direct test of why the returns differ: measures of what the assigned homework actually consists of (problem-solving versus memorization, alignment with the tested material), and direct evidence on which barrier — incomplete completion, weak parental support, or competing time constraints — makes additional math homework unproductive for black and low-parental-education students.

Problem Statement

Within a credible causal framework, explain the heterogeneous returns to homework time documented by Eren and Henderson (2011) for US eighth graders — large positive effects of additional mathematics homework on mathematics scores, near-zero effects of additional homework in science, English, and history on their respective scores, and null returns for black students and students with low parental education. Concretely: (i) test, using direct measures of assigned homework content and cognitive demand, whether the subject-specific pattern reflects the problem-solving character of mathematics homework relative to memorization-oriented homework in other subjects, or instead arises from teacher sorting, grading practices, or test alignment; and (ii) identify which barrier — incomplete homework completion, weaker parental support, or competing time constraints — drives the null returns to additional math homework for black students and students with low parental education. The mechanisms or barriers in (i) and (ii) may be multiple, with several contributing jointly rather than a single exclusive cause, and the rival explanations named in (i) are part of the target, extending beyond the content and cognitive-demand measures to whatever evidence bears on them. Extension to other grade levels (elementary and high school) is not part of this problem's core: the target is defined by the source's eighth-grade NELS:88 design, and subsequent literature has examined those levels directly (see Background).

References

  1. Eren, Ozkan; Henderson, Daniel J. (2011). 'Are we wasting our children's time by giving them more homework?' Economics of Education Review 30(5):950–961. DOI: 10.1016/j.econedurev.2011.03.011. https://doi.org/10.1016/j.econedurev.2011.03.011
  2. Grodner, Andrew; Rupp, Nicholas G. (2013). 'The Role of Homework in Student Learning Outcomes: Evidence from a Field Experiment.' The Journal of Economic Education 44(2):93–109. DOI: 10.1080/00220485.2013.770334. https://doi.org/10.1080/00220485.2013.770334
  3. Daw, Jonathan (2012). 'Parental income and the fruits of labor: Variability in homework efficacy in secondary school.' Research in Social Stratification and Mobility 30(3):246–264. DOI: 10.1016/j.rssm.2012.01.004. https://doi.org/10.1016/j.rssm.2012.01.004
  4. Kalenkoski, Charlene Marie; Pabilonia, Sabrina Wulff (2017). 'Does high school homework increase academic achievement?' Education Economics 25(1):45–59. DOI: 10.1080/09645292.2016.1178213. https://doi.org/10.1080/09645292.2016.1178213
  5. Jerrim, John; López-Agudo, Luis Alejandro; Marcenaro-Gutiérrez, Oscar D. (2019). 'The relationship between homework and the academic progress of children in Spain during compulsory elementary education: A twin fixed-effects approach.' British Educational Research Journal 45(5):1021–1049. DOI: 10.1002/berj.3549. https://doi.org/10.1002/berj.3549
  6. Fan, Huiyong; Xu, Jianzhong; Cai, Zhihui; He, Jinbo; Fan, Xitao (2017). 'Homework and students' achievement in math and science: A 30-year meta-analysis, 1986–2015.' Educational Research Review 20:35–54. DOI: 10.1016/j.edurev.2016.11.003. https://doi.org/10.1016/j.edurev.2016.11.003
  7. McJames, Nathan; Parnell, Andrew; O'Shea, Ann (2024). 'Little and often: Causal inference machine learning demonstrates the benefits of homework for improving achievement in mathematics and science.' Learning and Instruction 93:101968. DOI: 10.1016/j.learninstruc.2024.101968. https://doi.org/10.1016/j.learninstruc.2024.101968
  8. Flannery, Darragh; Palcic, Dónal; Clavel, Jose G. (2026). 'Who Benefits From the Assignment of Homework?' Oxford Bulletin of Economics and Statistics. DOI: 10.1111/obes.70119. https://doi.org/10.1111/obes.70119
  9. Eren, Ozkan; Henderson, Daniel J. (2011). 'Are We Wasting Our Children's Time by Giving Them More Homework?' IZA Discussion Paper No. 5547, March 2011 — the authors' version of reference 1, published in Economics of Education Review 30(5):950–961. https://docs.iza.org/dp5547.pdf