Cram Schools, Kumon, and Online Learning — What the Evidence Actually Says

Lead

Around the end of third grade, the question surfaces among parents: "Shouldn't we be thinking about cram school?" The major test-prep cram schools (juku) structure their curricula with a fourth-grade start in mind — so by this point in third grade, a sense of falling behind is easy to generate.

Does cram school actually improve outcomes? Which format suits which child? Does the starting age matter? These are questions that educational economics and developmental psychology have addressed directly, and the answers are more complicated than the enrollment calendars suggest.

Background: Japan's "Shadow Education" System

Juku (Japanese cram schools), Kumon (the subscription drill-based program), and correspondence-based distance learning belong to what researchers call shadow education — the vast, privately funded supplemental learning industry that operates parallel to formal schooling [6].

Japan's private middle school entrance exam system is unusual by international standards. The curriculum at major test-prep cram schools is designed to compress roughly six years of content into three years (grades 4–6). A "fourth-grade start" is a feature of that design — not a developmentally derived optimal entry point. In the greater Tokyo metropolitan area, approximately 24% of students entered private or national middle schools in 2024; in rural areas, the rate is in the low single digits.

The shadow education industry is large and growing — but Bray (1999) notes that this growth does not necessarily mean equal access to learning opportunities. The expansion of private supplemental education tends to track economic resources and reinforce inequality [6].

What Research Shows About Cram School Effectiveness

Studying the effects of juku empirically is methodologically difficult. Children who attend cram school differ from those who do not in family socioeconomic status (SES), parental educational aspirations, and the child's own academic motivation — all of which are also predictors of academic outcomes.

Stevenson and Baker (1992) used Japanese national data and found a positive correlation between juku attendance and academic achievement and educational attainment. But when family SES and parental educational expectations were statistically controlled, the effect shrank substantially [1]. (selection bias is a key concern here.) The observation "attends cram school and does well in school" is difficult to disentangle from "comes from a family where both academic success and cram school attendance are expected."

Dang and Rogers (2008), reviewing the global evidence on private tutoring for the World Bank, reached similar conclusions: when selection bias is accounted for, the effect of private supplemental instruction tends to diminish, and the practice may amplify inequality rather than reduce it [5].

A variable that consistently emerges as more predictive than attendance per se is clarity of purpose. When parents and children can articulate why they are attending (for a specific entrance exam, to address a specific subject weakness), continuation rates are higher and outcomes more clearly positive [2]. Starting without a clear stated purpose tends to produce high cost and moderate results.

Kumon's Specific Evidence Base

Kumon is a drill-and-advance program originating in Japan that has expanded to more than 50 countries. Its approach combines repetitive practice with learning ahead of the school curriculum.

The U.S. Institute of Education Sciences' What Works Clearinghouse (WWC) reviewed the Kumon Mathematics Program in 2009 and rated it as having "potentially positive effects" — but noted that the evidence was not based on rigorous randomized controlled trials, and that the evidentiary confidence was therefore limited [3].

González-Castro et al. (2019) presented a more favorable assessment, citing research suggesting that repetitive practice and curriculum advancement can improve procedural fluency (the speed and accuracy of calculation) [4]. Long-term follow-up data and evidence for higher-order problem-solving skills remain limited.

The clearest case for Kumon is in procedural fluency — calculation speed and accuracy. That is a different cognitive target from conceptual understanding or applied problem solving. Whether Kumon matches what a family is actually trying to achieve depends on aligning those goals with that specific tool.

Correspondence Learning: The Autonomy Question

Correspondence and online learning programs have the least robust evidence base — randomized trial data are sparse, and the outcome research is fragmented. Industry-side statistics suggest completion rates around 50–60% for major correspondence programs in Japan — which means that for a substantial fraction of enrolled children, the effectiveness question never becomes relevant.

The conditions under which correspondence learning tends to work are similar to those identified by Patall et al. (2008) for homework more broadly: high self-regulation on the child's part, and appropriately calibrated parental support rather than close monitoring. Children who can manage their own pace tend to do better; those who cannot may find the format more frustrating than productive.

Practical Takeaways

Name the purpose before enrolling. "Everyone's going" is not a purpose. Specifying — "we want to prepare for a particular middle school entrance exam" or "we want to address slow calculation speed" — creates a criterion for evaluating whether the format is working and whether to continue.

Match the format to how the child learns. Group-format cram schools require processing speed to follow the classroom pace, and some children find the competitive dynamic motivating. Individual tutoring is better suited for targeting specific weaknesses. Kumon targets procedural fluency but does not address conceptual understanding or word problems. Correspondence learning is self-paced but requires the self-regulation to use that freedom. None of these formats is universally better; each fits a different profile.

Treat the first three to six months as a trial. Rather than continuing indefinitely because you've started, define a review point early: "After one semester, we'll assess whether this is working." That structure reduces the sunk-cost pull and makes evaluation part of the plan.

Monitor fatigue, not just grades. When a child starts a cram school schedule, the signs of being overloaded are easy to miss. Whether the added load is compressing sleep, meals, unstructured play time, or friendships is worth watching — those are the prerequisites for the learning to stick.

Summary

"Starting in fourth grade is the right time" is a product of how test-prep cram school curricula are designed, not a finding from developmental research. Cram school effects, when selection bias is accounted for, are smaller than raw correlations suggest; clarity of purpose and format-to-child fit are the larger variables. Kumon improves calculation fluency but requires separate approaches for conceptual understanding. Correspondence learning works best for children with high self-regulation.

Any supplemental learning format tends to work when the purpose is clear. Starting without a clear purpose tends to accumulate cost without proportionate benefit.

References

1. Stevenson DL, Baker DP. Shadow education and allocation in formal schooling: transition to university in Japan. Am J Sociol. 1992;97(6):1639–1657. doi:10.1086/229942
2. Sasaki M, Knight J. Cram schools in Japan: the need for research. TLT (JALT J). 2016;40(1):19–26. URL: https://jalt-publications.org/sites/default/files/pdf-article/39.1tlt_art5.pdf
3. Institute of Education Sciences — What Works Clearinghouse. Kumon Mathematics Program Intervention Report. 2009. URL: https://ies.ed.gov/ncee/wwc/Docs/InterventionReports/wwc_kumon_031009.pdf
4. González-Castro P, Cueli M, Rodríguez C, García T, Álvarez L. The Kumon method: its importance in the improvement of the teaching and learning of mathematics from early education. Mathematics. 2019;7(1):109. doi:10.3390/math7010109
5. Dang H-A, Rogers FH. The growing phenomenon of private tutoring: does it deepen human capital, widen inequalities, or waste resources? World Bank Res Obs. 2008;23(2):161–200. doi:10.1093/wbro/lkn004
6. Bray M. The Shadow Education System: Private Tutoring and its Implications for Planners. Paris: UNESCO/IIEP; 1999.