Dettmers, Trautwein, and Lüdtke (2009) analyzed data from the Programme for International Student Assessment (PISA) 2003 - an OECD assessment of 15-year-olds - covering 231,759 students in 9,791 schools across 40 countries and found that school-average mathematics homework time was positively associated with achievement but shrank markedly under socioeconomic and tracking controls, while the student-level association was mixed in sign across countries, leaving the causal direction of the homework-achievement relation unresolved. The open problem is to determine, with designs that separate selection from causation, whether and how much homework time causally improves achievement at the student and class/school levels, and to explain why the association varies across countries. Later quasi-experimental work - instrumental-variables and difference-in-differences estimates on TIMSS (the Trends in International Mathematics and Science Study), twin and student fixed-effects studies, longitudinal panels, and a field experiment on homework frequency - yields conflicting results in sign and magnitude, so the causal-direction question and the mechanisms behind the cross-country heterogeneity remain open.

Background

Homework is one of the most widely used and most debated instructional practices in school systems worldwide, and whether the time students spend on it actually raises achievement has been contested since the mid-twentieth century. Early syntheses (notably Cooper's meta-analyses) reported positive homework-achievement associations, especially in secondary school, but almost all primary studies were correlational and treated the student as the unit of analysis. Two insights reshaped the field. First, educational data are nested - students sit in classes within schools within countries - and multilevel models (also called hierarchical linear models, which estimate separate regression coefficients at each level of the nested structure) revealed that the homework-achievement association depends drastically on the level of analysis: within students or classes it is often negative or near zero, plausibly because weaker students need more time for the same assignment (reverse causality), whereas between classes and schools it tends to be positive. Trautwein, Schnyder, Niggli, Neumann, and Lüdtke (2009) called these "chameleon effects": the association changes appearance with the measure used and the level of analysis chosen. Second, self-reported homework time is a noisy indicator entangled with homework frequency, length, quality, and effort; Flunger, Trautwein, Nagengast, Lüdtke, Niggli, and Schnyder (2015), using latent profile analysis (a person-centered clustering method that groups students by response patterns), showed that homework time predicts achievement only in combination with high effort. Against this backdrop, Dettmers, Trautwein, and Lüdtke (2009) fitted multilevel models to the Programme for International Student Assessment (PISA) 2003, an OECD assessment of 15-year-olds, covering 231,759 students in 9,791 schools in 40 countries, with mathematics homework time measured by a single self-report item. School-average homework time was positively associated with mathematics achievement in most countries, but the coefficients decreased considerably once the Index of Economic, Social and Cultural Status (ESCS, PISA's composite socioeconomic measure) was controlled at the student and school levels, and fell further in the six systems (Austria, Flanders, Germany, Japan, Korea, and the United States) where tracking - the sorting of students into academically differentiated school types or course levels - could be controlled explicitly. In a German extension, adding a figural-reasoning measure of cognitive ability rendered the student-level homework coefficient non-significant. The authors concluded that the homework time-achievement relationship "is not universal" and left its causal direction open: longer homework time may raise achievement, but higher-achieving or more academically oriented classes may also be assigned more homework, and reciprocal effects are possible. Because the design was cross-sectional, relied on self-reported time, used the school rather than the classroom as sampling unit, and proxied school quality and tracking with socioeconomic composition, the mechanisms behind the cross-country differences remained unidentified. Resolving the question requires identification strategies from causal inference - instrumental variables (a variable that affects homework time but plausibly affects achievement only through it), fixed-effects designs exploiting within-student or within-twin comparisons over time or across subjects, difference-in-differences over time (comparing achievement changes across units whose homework time changed by different amounts), or randomized assignment of homework amounts - rather than further cross-sectional correlations.

Problem Statement

Does time spent on homework causally improve mathematics achievement, or does the observed association arise because higher-achieving, more academically oriented, or socioeconomically advantaged classes and school systems receive or complete more homework - or do both directions operate reciprocally? Dettmers, Trautwein, and Lüdtke (2009) established that the association between homework time and achievement is not universal across the 40 PISA 2003 countries: positive between schools, mixed in sign at the student level, and strongly confounded with socioeconomic status and tracking. The problem they left open, at the cross-national scope they studied, is to identify the direction and magnitude of the causal effect of homework time on achievement at the student level and at the class/school level using a design that rules out selection and reverse causality, and, jointly, to identify the mechanisms that produce the cross-country differences in the association, including but not limited to tracking structures, school and class composition, homework frequency as distinct from homework length, homework quality and effort, parental assistance, and country-specific instructional practices. Both components are jointly required and neither alone resolves the problem; the listed mechanisms are illustrative rather than exhaustive. A satisfactory answer must not rest on the cross-sectional, self-reported correlational structure of the original study; because the open question concerns the cross-country pattern documented across the 40 systems, claims about cross-country heterogeneity are claims about differences between school systems.

References

  1. Dettmers, S., Trautwein, U., & Lüdtke, O. (2009). The relationship between homework time and achievement is not universal: evidence from multilevel analyses in 40 countries. School Effectiveness and School Improvement, 20(4), 375-405. DOI: 10.1080/09243450902904601. https://doi.org/10.1080/09243450902904601
  2. Trautwein, U., Schnyder, I., Niggli, A., Neumann, M., & Lüdtke, O. (2009). Chameleon effects in homework research: The homework-achievement association depends on the measures used and the level of analysis chosen. Contemporary Educational Psychology, 34(1), 77-88. DOI: 10.1016/j.cedpsych.2008.09.001. https://doi.org/10.1016/j.cedpsych.2008.09.001
  3. Gustafsson, J.-E. (2013). Causal inference in educational effectiveness research: A comparison of three methods to investigate effects of homework on student achievement. School Effectiveness and School Improvement, 24(3), 275-295. DOI: 10.1080/09243453.2013.806334. https://doi.org/10.1080/09243453.2013.806334
  4. Fan, H., Xu, J., Cai, Z., He, J., & Fan, X. (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
  5. Jerrim, J., López-Agudo, L. A., & Marcenaro-Gutierrez, O. 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. Jerrim, J., López-Agudo, L. A., & Marcenaro-Gutierrez, O. D. (2020). The association between homework and primary school children's academic achievement: International evidence from PIRLS and TIMSS. European Journal of Education, 55(2), 248-260. DOI: 10.1111/ejed.12374. https://doi.org/10.1111/ejed.12374
  7. Flunger, B., Trautwein, U., Nagengast, B., Lüdtke, O., Niggli, A., & Schnyder, I. (2015). The Janus-faced nature of time spent on homework: Using latent profile analyses to predict academic achievement over a school year. Learning and Instruction, 39, 97-106. DOI: 10.1016/j.learninstruc.2015.05.008. https://doi.org/10.1016/j.learninstruc.2015.05.008
  8. Fernández-Alonso, R., Álvarez-Díaz, M., Suárez-Álvarez, J., & Muñiz, J. (2017). Students' achievement and homework assignment strategies. Frontiers in Psychology, 8, 286. DOI: 10.3389/fpsyg.2017.00286. https://doi.org/10.3389/fpsyg.2017.00286
  9. Al-Bahrani, A., Apostolova-Mihaylova, M., & Marshall, E. C. (2022). Helping some and harming others: Homework frequency and tradeoffs in student performance. The Journal of Economic Education, 53(3), 197-209. DOI: 10.1080/00220485.2022.2075506. https://doi.org/10.1080/00220485.2022.2075506
  10. Zhang, Y., Li, T., Xu, J., Chen, S., Lu, L., & Wang, L. (2024). More homework improve mathematics achievement? Differential effects of homework time on different facets of students' mathematics achievement: A longitudinal study in China. British Journal of Educational Psychology, 94(1), 181-197. DOI: 10.1111/bjep.12640. https://doi.org/10.1111/bjep.12640
  11. Wu, X., Wu, R., Hanley, C., Liu, H., & Liu, J. (2023). How to better balance academic achievement and learning anxiety from time on homework? A multilevel and classification and regression tree analyses. Frontiers in Psychology, 14, 1130274. DOI: 10.3389/fpsyg.2023.1130274. https://doi.org/10.3389/fpsyg.2023.1130274
  12. Ozyildirim, G. (2021). Time spent on homework and academic achievement: A meta-analysis study related to results of TIMSS. Psicología Educativa, 28(1), 13-21. DOI: 10.5093/psed2021a30. https://doi.org/10.5093/psed2021a30
  13. López-Agudo, L. A., & Marcenaro-Gutierrez, O. D. (2025). The relationship between weekly instruction, homework time and students' competences. The Journal of Early Adolescence. DOI: 10.1177/02724316251379765. https://doi.org/10.1177/02724316251379765
  14. Wang, L., Li, M., Wu, Y., Hu, Y., & Luo, L. (2025). Persistent effectiveness of mathematics homework time from primary school to early middle school: A longitudinal study from China. Learning and Individual Differences, 118, 102642. DOI: 10.1016/j.lindif.2025.102642. https://doi.org/10.1016/j.lindif.2025.102642
  15. Qin, K., & Wang, Y. (2026). Time and quality in and out of class: The roles of instruction and homework in mathematics achievement and interest. Frontiers in Psychology, 17, 1870861. DOI: 10.3389/fpsyg.2026.1870861. https://doi.org/10.3389/fpsyg.2026.1870861