The relationship between homework time and academic performance among K‐12: A systematic review

Abstract Background Homework is a common educational task given to students around the world. It demands mental exertion, but staying focused can be challenging, especially for K‐12 students. Too much homework can increase their cognitive load and mental fatigue, leading to decreased motivation and performance. This can cause boredom with homework and learning. To lessen their load and make homework more effective, it is important to establish the connection between homework duration and academic achievement. Objectives To evaluate the relationship between homework time and academic performance among K‐12 students. Search Methods On November 5, 2021, we retrieved articles from a variety sources. Firstly, we searched 10 electronic databases for related publications, including Academic Search Premier, APA PsycArticles, APA PsycInfo, Business Source Premier, Education Resources Information Center (ERIC), Journal Storage (JSTOR), Learning and Technology Library (LearnTechLib), OCLC FirstSearch, Social Sciences Citation Index (Web of Science), and Teacher Reference Center. We also searched two publisher platforms: ScienceDirect and Taylor & Francis Online Database. Secondly, we consulted five educational organization website such as, American Educational Research Association, Best Evidence Encyclopedia, Education Endowment Foundation, European Educational Research Association, What Works Clearinghouse, and the Open Grey database for unpublished studies. We then searched Open Dissertations and ProQuest Dissertations & Theses Global databases to locate the relevant dissertations and theses. Additionally, we hand‐searched seven educational journals to identify unpublished documents, reports, and potential studies not indexed in the databases. Lastly, we searched Campbell Library to identify relevant reviews and primary (and nearly eligible studies) in these reviews. We also searched Google Scholar for related studies and checked the citations of eligible studies as well as their bibliographies. Selection Criteria Studies with the following criteria were included: − Population: K‐12 school students with no disabilities or not attending special education schools;− Intervention: Homework assigned regularly by schoolteachers to students to complete during non‐school hours;− Comparison: Different time spent on the homework;− Outcomes: Academic performance was the primary outcome. The secondary outcomes were academic motivation and the quality of homework;− Study design: Treatment‐control group design or comparison group design studies. Data Collection and Analysis We reviewed the titles, abstracts, and full texts of the retrieved records. Our team extracted and coded all relevant information from the studies that met our inclusion criteria. To evaluate the risk of bias, we used the Cochrane Risk of Bias tool for randomized controlled trials and ROBINS‐I for non‐randomized controlled trials. A random‐effect meta‐analysis was conducted to determine the effect of homework on academic achievement as compared to no homework. A funnel plot, trim‐and‐fill method and Egger's test were used to test for any publication bias. Due to the insufficient data on homework duration and academic achievement, we analyzed these data using qualitative synthesis. Main Results Eleven publications were identified that examined the relationship between homework duration and academic outcomes using an experimental design. Based on their focus, we categorized them into two groups: comparisons of homework with no homework and comparison of homework with less homework. There were 10 articles with 14 independent reports that compared academic performance between students who did homework and those who did not. Overall, the meta‐analysis revealed that the students who did homework had better academic performance than that those who did not (n = 14; g = 0.45, 95% confidence interval [CI]: 0.24–0.66; Q = 454.30, I 2 = 71.30%, τ 2 = 0.11), especially in arithmetic computation (n = 5; g = 0.46, 95% CI: 0.17–0.75; Q = 13.03, I 2 = 69.29%, τ 2 = 0.07) and arithmetic problems solving (n = 6; g = 0.17, 95% CI: 0.02–0.33; Q = 6.87, I 2 = 27.17%, τ 2 = 0.01), but not in arithmetic concepts (n = 3, g = −0.02, 95% CI: −0.22–0.18; Q = 1.46, I 2 = 0.00%, τ 2 = 0.00). Two experiments explored the effectiveness of homework moderated by homework time. In Koch (1965), the effects of long daily homework (20–30 min) and short daily homework (10–15 min) were compared. The authors found that achievement in arithmetic concepts was higher with long homework assignments every day. Recently, Dolean and Lervag (2021) confirmed the effect of homework on writing skills, and their findings were consistent with those of Koch (1965), who found that increasing time spent on homework was associated with greater writing achievement (average 20 min each time). Authors' Conclusions Homework could be used as a supplement to enhance the academic performance of primary school students. However, the optimal amount of time they should dedicate each day to homework to achieve the best results remains uncertain. More high‐quality experiments are needed to determine the ideal homework duration for these students. Furthermore, additional research is required to understand the impact of homework on secondary school students.

Selection Criteria: Studies with the following criteria were included: Data Collection and Analysis: We reviewed the titles, abstracts, and full texts of the retrieved records.Our team extracted and coded all relevant information from the studies that met our inclusion criteria.To evaluate the risk of bias, we used the Cochrane Risk of Bias tool for randomized controlled trials and ROBINS-I for nonrandomized controlled trials.A random-effect meta-analysis was conducted to determine the effect of homework on academic achievement as compared to no homework.A funnel plot, trim-and-fill method and Egger's test were used to test for any publication bias.Due to the insufficient data on homework duration and academic achievement, we analyzed these data using qualitative synthesis.
Main Results: Eleven publications were identified that examined the relationship between homework duration and academic outcomes using an experimental design.
Authors' Conclusions: Homework could be used as a supplement to enhance the academic performance of primary school students.However, the optimal amount of time they should dedicate each day to homework to achieve the best results remains uncertain.More high-quality experiments are needed to determine the ideal homework duration for these students.Furthermore, additional research is required to understand the impact of homework on secondary school students.
academic performance, homework time, K-12, systematic review

| Limited experimental studies of the optimum time of homework among K-12
The available studies suggested large effects of homework on academic performance among primary school students, but few studies evaluate the optimum time spent on homework to improve academic achievement for K-12 students.

| What is this review about?
Homework is a common activity among K-12 students, and numerous observational meta-analyses have confirmed the positive relationship between homework and academic performance.However, considering children's limited energy and attention, several researchers have proposed an optimum time to spend on homework.This Campbell review examines the effects of homework on academic performance among K-12 students with experimental studies and estimates the optimum time that should be spent on it.

| What studies are included?
This review included eleven experimental studies, eight of them evaluated the effects of homework for K-12 students and two studies estimated the time effect of homework.Only one study was published online recently, in 2021, and the remaining studies spanned the period from 1939 to 1995.

| What is the relationship between homework and academic performance?
Homework has a beneficial effect on primary school student's academic achievement, especially in improving their arithmetic computational skill and arithmetic problem-solving skills.However, the optimum duration spent on homework is not clear.

| What do the findings of this review mean?
Homework is recommended as a supplement to improve academic performance in primary school students.However, scarce and outdated empirical studies included in this review suggest an urgent need for high-quality experiments to explore the optimal duration to make full use of homework.

| How up-to-date is this review?
The review authors searched for studies up to November 5, 2021.

| Description of the condition
Homework is defined as "any task assigned by schoolteachers intended for students to carry out during non-school hours" (Cooper, 1989).This definition explicitly excludes (a) in-school guided study; (b) home study courses delivered through the mail, television, audio or videocassette, or the internet; and (c) extracurricular activities such as sports and participation in clubs (Cooper et al., 2006).Nowadays, thanks to technology, web-based homework has become increasingly popular among teachers.Platforms like Google Classroom, 1 Firefly, 2 eSchools, 3 and Moodle 4 (Mendicino et al., 2009) allow students to complete their homework online and teachers to provide immediate feedback (Callahan, 2016;Lucas, 2012;Mendicino et al., 2009).Therefore, in this systematic review, homework also includes online tasks performed outside school.
The purpose of homework can be divided into instructional and non-instructional objectives (Lee & Pruitt, 1979).The most common instructional purpose of homework includes review, preview, and extension (Becker & Epstein, 1982;Lee & Pruitt, 1979;Muhlenbruck et al., 1999).Review assignments allow students to practise newly acquired skills or review material learned in class.Preview assignments introduce new skills or materials before the class to help students prepare for unfamiliar knowledge (Muhlenbruck et al., 1999).Extension assignments involve transferring previously learned skills to new situations (Cooper et al., 2006;Lee & Pruitt, 1979).The noninstructional purpose of homework varies, including forming better study habits, increasing students' sense of responsibility, enhancing awareness of independent learning, and building communication among parents, children, and teachers (Becker & Epstein, 1982;González et al., 2001;Lee & Pruitt, 1979;Muhlenbruck et al., 1999;Van Voorhis, 2003).However, homework can also be used as a form of punishment (Epstein & Van Voorhis, 2001).
Students worldwide are frequently assigned homework as part of their educational activities, which is considered a hallmark of the educational excellence movement.Homework policies vary globally.In China, Primary 1 and 2 students are not assigned any homework, and homework for Primary 3 to 6 students is limited to maximum of 60 min, while junior high school students are allowed up to 90 min per day (Ministry of Education of the People's Republic of China, 2021).In Singapore, the general guideline for time spent on homework is about 30 min to an hour for Primary 1-2 students, 1-1.5 h for Primary 3 and 4 students and 1.5-2 h for Primary 5 and 6 students (South View Primary School, 2024).
The UK had specific mandates for homework ranging from no homework for kindergarteners to a maximum of 2.5 h per day for Years 10 and 11 (Cooper & Nye, 1994), but these guidelines were scrapped in 2012 and homework frequency and duration, with recommended daily averages of 40 min for primary school students, 70 min for junior high school students, and 100 min for senior high school students (Roderique et al., 1994).Recently, some US schools have adopted no-homework policies in response to the Common Core curriculum (Mae Gambong Luengas & Deloy, 2022).Finland, on the other hand, has no national homework policy, and Finnish students have significantly less homework than other nations, with an average of less than 30 min per day (Federick, 2020).
Initially, the level of homework was widely accepted by parents.
However, as the volume of homework increased, parents and scholars began to realize the burden it placed on students.Parents complained that children lost their childhood and called for less homework (Gill & Schlossman, 2003).Similarly, in the mid-19th century, homework became commonplace in the United Kingdom, and its increased level became a topic of much debate in the 1880s in response to the introduction of payment according to results and other factors for teachers (Hallam, 2004).In 2016, the World Health Organization (WHO) reported that students feel pressured from schoolwork (World Health Organization Regional Office for Europe, 2016).Meanwhile, parents continued to express concerns about excessive homework assigned to their children (Gill & Schlossman, 2003, 2004;Jerrim et al., 2019;Xue & Zhang, 2019).
The topic of whether or not homework improves academic performance has been a subject of debate for over a century (Cheema & Sheridan, 2015;Cooper et al., 2006;Cooper, 1989;Kitsantas et al., 2011;Kralovec & Buell, 2000;Trautwein & Lüdtke, 2007).Although several meta-analyses of the relationship between homework and performance have shown a positive correlation between homework time and academic performance (Baş et al., 2017;Cooper et al., 2006;Cooper, 1989;Fan et al., 2017), it is challenging to establish causality.It is possible that more academically inclined students who typically score better grades complete their homework more thoroughly and conscientiously.On the other hand, students who struggle academically may put in more effort to catch up on their studies at home.Regardless, it is possible that the effects of homework are not linear.Some evidence suggests that increasing homework duration can improve academic performance, but there is a point where too much homework can actually lead to a decline in performance (Ackerman et al., 2011;Krejtz et al., 2018;Reteig et al., 2019).For example, based on data from the National Center for Education Statistics (NCES) survey of 58,000 high school students in grades one and two, Keith (1982) found that increasing the amount of homework can improve performance for individuals at any level of ability.However, there is a limit to how much homework can be beneficial, and exceeding this limit would lead to a decline in performance.Homework can play a compensatory role, but it must be increased moderately and not beyond a certain duration (Keith, 1982).

| Description of the intervention
The intervention we consider in this review is homework assigned by schoolteachers to students during their non-school hours, without additional teaching or support such as activities in a study club.We excluded flipped learning because the homework in flipped classrooms typically consists of instructional videos, rather than problem sets without additional instruction; Moreover, the students benefit from traditional homework as it extends the school day, while the homework in a flipped classroom is still considered part of teaching (Blazer, 2009;de Araujo et al., 2017).The comparison condition was different amounts of time spent on the homework; we divided it into groups of 0-15, 16-30, 31-45, 46-60min, 61-90, 90-120 min, and more than 120 min.
Any type of homework was included, like written, oral, or practical homework.We excluded homework sources like parents or teachers from extracurricular activities such as sports and participation in clubs, as well as in-school guided study and home study courses.We also excluded homework related to psychotherapy.

| How the intervention might work
The conceptual framework for this review was the theory of change, which describes how homework may affect academic performance.understand the learning content more rapidly (Muhammad, 2015).
Second, the law of exercise suggests that repetition of relevant actions in practice will enhance learned connections.Third, the law of effect indicates that all types of positive or negative feedback that learners receive while learning will strengthen or weaken the cognitive connections that they have formed.
However, excessive homework can lead to cognitive overload and mental fatigue, which may reduce students' readiness for learning.This can result in tiredness and anxiety, leading to decreased performance and inadequate feedback (Ackerman et al., 2011;Krejtz et al., 2018;Reteig et al., 2019).Moreover, if students spend excessive time on homework, they may lose time to participate in other activities that could contribute to their overall development.
It is also essential to recognize that spending more time on homework does not always lead to better academic performance.
Once students have fully understood the homework content, completing additional homework may not be beneficial.Rather, there may be diminishing returns on the time spent on homework, which can eventually reach zero (Ackerman et al., 2011;Bartelet et al., 2016).
Additionally, if students miss out on sleep due to homework, they may experience reduced performance in class or tests, further undermining their learning readiness (Reteig et al., 2019).

| Previous reviews
Several systematic reviews have explored the effectiveness of homework in improving students' performance.However, they assumed a linear relationship between the time spent on homework and performance, without considering how homework duration affects students' autonomous motivation (Bartelet et al., 2016).Even a summary of the evidence by Hallam (2004) did not suggest how much time students should spend on homework (Hallam, 2004).The UK Education Endowment Foundation's (EEF) toolkit entry for secondary school homework emphasized the importance of quality over quantity, but quantity still is a key factor that affects its quality.Therefore, extant reviews left the practical question of homework duration unanswered.
In 1989, Cooper conducted a review of the relationship between homework and performance.The results showed that the average correlation for students in primary-, middle-, and high-school between the amount of homework and performance was zero; for students in middle school, the correlation was 0.07, whereas for high school students, the correlation was 0.25 (Cooper, 1989).In 2006, Cooper et al. (2006) conducted another systematic review of the effectiveness of homework at improving academic performance.The results showed that the correlation between homework time and performance for high-school students was 0.25, but for middleschool students, it was nearly zero.However, they did not explicitly consider homework time (Cooper et al., 2006).Fan et al.(2017) conducted a meta-analysis of the studies published between 1986 and 2015, and demonstrated a small but positive relationship between homework duration and math/science attainment, with a correlation of 0.145 (Fan et al., 2017).
All studies above assumed that more homework always leads to better performance, with a linear correlation between the two.However, it is important to consider that homework can be tedious and mentally taxing, and there are limits to one's ability to concentrate.To maximize the effectiveness of homework, it is important for teachers, school administrators, parents, and students to determine the optimal amount of time to spend on homework.In previous studies, the correlation coefficient was used as a measure of the strength of the linear relationship between homework and performance.

| The contribution of this review
Homework can be a valuable tool for improving students' academic performance, whether it is for preparation, practice, extension, or application.Previous reviews have shown that homework is effective in this regard.However, it is important to note that more homework does not necessarily equate to better results.The amount of homework assigned should be limited by students' ability to concentrate for extended periods.This systematic review categorizes participants into groups based on the amount of time spent on homework (0-30, 31-60, 61-90, 90-120 min, and more than 120 min) and compares their test scores to identify the relationship between homework time and academic performance.This information will be useful for teachers and parents to better understand the importance of homework and will provide guidance for teachers in assigning appropriate amounts of homework.

| OBJECTIVES
This review synthesized the results from publications focused on homework duration and academic performance and assessed the relationship between the two factors.Our objectives were the following: 1.To identify the extent of the relationship between homework time and students' academic performance; 2. To analyze differences in the effectiveness of homework time across genders, grades, subjects, and regions; and 3. To identify the potential factors that affect homework time, such as academic subject, task difficulty, type of homework, mode of homework, parental involvement, and feedback on homework.

| Types of studies
According to our protocol (Guo et al., 2021), we included treatmentcontrol group designs or comparison group designs, namely randomized controlled trials (RCTs) and non-randomized controlled trials (NRCTs), such as cohort studies, controlled pre-and post-treatment studies, and interrupted time-series studies.To be included, studies needed to explore the effect of homework duration on academic performance by comparing test score differences between different groups before and after the intervention, and explicitly report the time spent on homework, as well as the mean and standard deviation of academic achievement.
Due to the language restrictions within our team, we included only studies published in English.

| Types of participants
This review focused on K-12 school students, excluding those with disabilities and those attending special education schools.If a primary study included a mixed sample of students, we only considered the sub-sample without special needs students, if that information was provided.This is because the effects and context may differ for individuals with disabilities.

| Types of interventions
We explored the relationship between homework duration and academic performance by comparing academic scores obtained given different homework duration.The eligible intervention studies needed to explicitly state that the intervention was regular homework assigned by schoolteachers to students to complete during non-school hours, with the aim of improving academic achievement, regardless of the nature of the homework content.Furthermore, we included only school-based interventions; homework was excluded if allocated by other people, such as parents or teachers from extra-curricular schools, or study clubs, or was associated with extracurricular activities such as sports and participation in clubs.Homework related to psychotherapy was also excluded.
The comparison condition was the different duration of homework.

| Types of outcome measures
The objective of the review was to explore the impact of homework on students' academic outcomes.We extracted homework time and academic performance from the studies.Homework duration was an exact time or a time frame reported by students or parents.Academic performance was measured by a teacher, exam results, and/or by the research team using any valid standardized test.
Valid standardized tests were considered norm-referenced tests (e.g., Gates-MacGinitie Reading Tests and Star Math), state-wide tests (e.g., Iowa Test of Basic Skills), and national tests (e.g., National Assessment of Educational Progress).If the nature of the test was not clear from the description of outcome measures in the studies, we used electronic sources to determine whether the test was standardized.
The outcome was academic performance (test score and standard deviation); studies that measured academic performance and homework time were included.

| Duration of follow-up
There is no restriction on the duration of follow-up.

| Types of settings
Only the studies conducted in K-12 schools were included.
Compared with protocol, we conducted an extra search of the LearnTechLib on November 5, 2021.

Unpublished studies
We searched the following sources to identify relevant unpublished studies and reports using the keyword "homework."Searches were conducted up to November 5, 2021.

net/)
• Best Evidence Encyclopedia (https://bestevidence.org/) T A B L E 1 The search strategy and results for each electronic database/platform.

Dissertations and theses
We searched OpenDissertations (via EBSCOhost) and ProQuest Dissertations & Theses Global databases for dissertations on November 5, 2021.The search strategies are shown in Table 1.

Hand searching
Based on the scope of journals and their 5-year impact factors, we selected the following international journals and hand-searched them for relevant studies on November 5, 2021.
• American Educational Research Journal We searched the Campbell Library (https://www.campbellcollaboration. org/better-evidence.html) to identify previous systematic reviews related to homework, and then scanned the primary studies included (and nearly eligible studies listed in excluded section) in these reviews.
We also searched Google Scholar using the keyword "homework" on November 5, 2021, and all the titles of results and abstracts were screened and evaluated based on our inclusion and exclusion criteria.We stopped the search when there were five consecutive pages with no relevant studies.Additionally, we checked the forward citations of eligible studies as well as their bibliographies.

| Selection of studies
The selection of studies was performed independently by the first two reviewers (Guo LP and Jieyun Li) in Rayyan (https://rayyan.qcri.org/).All titles and abstracts of the records identified after retrieval were screened, the potentially relevant references were retrieved in full-text, and the primary studies that met our criteria were included for further data extraction.Any discrepancies between the reviewers were resolved by consensus with another reviewer (Kehu Yang).The study screening process was based on the PRISMA guidelines (Moher et al., 2009).

| Data extraction and management
According to our protocol (Guo et al., 2021) "primary school student" OR "pupil" OR "schoolchild" OR "junior high school student" OR "schoolage"#4 #1 AND #2 AND #3 ProQuest Dissertations & Theses Global https://www.proquest.com(ab("k-12" OR "preschool student" OR "pre-school student" OR "kindergarten student" OR "middle school student" OR "high school student" OR "senior school student" OR "primary school student" OR "pupil" OR "schoolchild" OR "junior high school student" OR "school-age") AND ab ("achievement" OR "performance" OR "grade" OR "score" OR "academic achievement" OR "gpa" OR "academic performance")) AND (su("homework" OR "home-work") OR ti("homework" OR "home-work")) Total 5179 a This number includes the records from the databases of Academic Search Premier, Business Source Premier, Teacher Reference Center, APA PsycArticles, APA PsycInfo (via EBSCOhost) and OpenDissertations, which were searched via EBSCOhost (http://search.ebscohost.com/)using the same search strategy.
− For academic performance, means, standard deviations (or information by which to estimate standard deviations), and the number of participants in each group were extracted.If more than one measure was reported, we extracted all measures and planned to analyze the measurement method as a moderator.
− We extracted the homework duration reported in the primary studies and coded the data as presented, either as categorical or continuous data.We created a data set of continuous variables, using the mid-point for data reported in categorical form, and at least two categorical data sets.The multiple categorical data sets were used to test sensitivity for the chosen thresholds; the mean and standard deviations were calculated in each group.If the weekly homework duration was reported instead of the daily duration, we divided the total homework duration by five.If the homework duration was listed in hours, we converted it to minutes.
− For controlled pre-and post-intervention studies, mean or median change from baseline scores were extracted.If change scores were unavailable or could not be computed, post-intervention values were extracted by the reviewers.

| Assessment of risk of bias in included studies
For randomized controlled trials, the Cochrane Risk of Bias tool 2.0 (ROB 2) was used to assess the quality of the method and potential limitations (Higgins & Green, 2011).For non-randomized studies (including cohort studies, controlled pre-and post-intervention studies, and interrupted time-series studies), the risk of bias in non-randomized studies of interventions (ROBINS-I) was used to check the quality of the individual study (Sterne et al., 2016).The risk of bias assessment was conducted by two of the authors (Zheng Xu and Xing Xin).Any disagreement was resolved by discussion with another author (Xiuxia Li).
The results of the ROB analysis will be used to assess the credibility of the review findings, and as a moderator in the analysis of heterogeneity.

| Measures of treatment effect
We used Hedges' g to determine the magnitude of intervention effects.
We calculated Cohen's d with the primary data (e.g., means and standard deviations, mean difference) reported for experimental and control groups in studies using Campbell web-based effect size calculator,6 and then calculated Hedges' g with Comprehensive Meta Analysis V2.

| Unit of analysis issues
We considered the unit of analysis of the studies to determine whether individuals were randomized in groups (i.e., cluster-randomized trials), book-based) were included in this review, although only intervention and control groups that meet the eligibility criteria will be used in the data synthesis.

| Dealing with missing data
If there were any missing data, we contacted the author at least twice to obtain additional information, if the correspondence address was available.If these data were unavailable, we only analyzed the available data; studies with missing data are described in the results section.Furthermore, the potential impact of missing data on comprehensive estimates are considered in the discussion.

| Assessment of heterogeneity
Forest plots were used to visually investigate overlaps in the confidence intervals (CIs) of the results of the individual studies.The χ 2 test was performed, and the Q statistics, I 2 , and τ 2 index were adopted to evaluate heterogeneity across studies.For Q statistics, a p-value of 0.05 was used as a threshold for statistical significance.
The I 2 index refers to the truly observed variation ratio (Borenstein et al., 2009), and 25%, 50%, and 75% of the I 2 indicate low medium, and high heterogeneity (Higgins & Thompson, 2002).And τ 2 is the between-studies variance (the variance of the effect size parameters across the population of studies), that is, the variance of true effect sizes (across an infinite number of studies).

| Assessment of reporting biases
Visual funnel plots with trim and fill and Egger's test of funnel plot symmetry were performed to evaluate potential publication bias.

| Data synthesis
Due to the small number of included studies, we conducted a meta-analysis to evaluate the effectiveness of homework as compared to no-homework.We also used a narrative approach to describe the findings, which indicated that the effect of homework depended on the time spent on it (more homework compared to less homework).
In our meta-analysis, we encountered a variety of outcome indices.To address this, we extracted all the raw data and calculated Cohen's d using the Campbell web-based effect size calculator, and then adopted Hedges's g as the effect size index when performing meta-analysis.A random effects model (REM) was adopted to estimate the effect of homework compared to no-homework.All these processes were performed using Comprehensive Meta Analysis V2.

| Subgroup analysis
According to our protocol, we planned to conduct the subgroup analysis by gender, grade level, region, publication year, mode of homework, type of homework, the measure of academic performance, and the subject (Guo et al., 2021).However, due to the limited studies included in each group, we were unable to conduct all the planned subgroup analyses.

| Sensitivity analysis
As per the protocol, we used the "one-leave-out" method to check the outliers that potentially influence the effect of homework compared to no-homework.

| Treatment of qualitative research
All qualitative research was excluded.

| Summary of findings and assessment of the certainty of the evidence
All the primary studies included in this review were rated as low quality, and as such we have refrained from assessing the overall quality of evidence.

| Characteristics of included studies
Of the 4904 studies identified as related to homework time and academic achievement, the titles and abstracts of 4769 were doublescreened, after removing duplicates; 127 were selected for full-text review, and 11 primary studies were included in the systematic review (see Figure 2).The characteristics of the included studies, including study design, research setting, participants, interventions and comparisons, and outcomes, are shown in Table 2.
The average sample size of the remaining studies was 53.4.
Four experiments also considered participants' intelligence.In Koch (1965), the mean IQ scores for experimental and control groups were similar in both language (control group: 111, half group: 109, and full group: 112) and non-language groups (control group: 109, half group: 113, and full group: 109) (Koch, 1965).Intelligence was excluded from the potential confounding factors in Foran and Weber (1939), who reported the mean of intelligence test scores in two groups as 105.55 (SD = 9.20) and 107.71 (SD = 9.17), respectively (Foran & Weber, 1939).Maertens (1969) included the level of intelligence as a classification variable in the data analysis.Students were divided into three groups based on their IQ: high (IQ of 112 and above), middle (IQ of 103 through 111), and low group (IQ of 102 and below).The median IQ of the total sample was 106 (Maertens, 1969).Tupesis (1972) tested the IQ in two groups using Henmon-Ne1s and found no difference between them (108.5 vs. 109.1)(Tupesis, 1972).

Interventions and comparisons
All 11 interventions were pencil-paper homework designed by teachers or experimenters, but the content varied depending on the target subject.None of the studies regarded homework duration as the primary variable, and only two studies described the exact duration in the experimental groups.Gray and Allison (1971) clearly defined that the students in the intervention group received three 20-min homework assignments per week, and the students in the control group received no homework (Gray & Allison, 1971).Dolean and Lervag (2021) reported the homework time in three groups (writing group, balanced group, and math group) as an average of 27, 20, and 21 min, respectively (Dolean & Lervag, 2021).Foran and Weber (1939)  Two studies distinguished the different types of homework.Foyle (1990) designated practice homework and preparation homework, which were randomly assigned to two of three treatment groups (Foyle, 1990).Maertens and Johnston (1972) considered the time of feedback and then proposed two homework treatments.One was homework with immediate feedback: children were assigned daily homework and received the answer to each problem or exercise as it was completed.The other was homework with delayed feedback: children were assigned homework identical to that in the immediate feedback treatment but did not received the answers until after the entire homework assignment was completed (Maertens & Johnston, 1972).
They further considered the effect on writing and used editing and spelling performance as their outcomes (Dolean & Lervag, 2021).Townsend (1995) researched the effect of homework on vocabulary lessons, using vocabulary knowledge and understanding as outcomes (Townsend, 1995).Foyle (1990) studied the effectiveness of practicing homework on performance in the social studies lessons (Foyle, 1990).Ewart Anderson (1946) also explored the average effect of homework on English, social studies, and math, and adopted a unified test score as the final outcome (Ewart Anderson, 1946).McGrath (1992) estimated the effectiveness of homework on English literature (McGrath, 1992).

| Excluded studies
Studies were excluded for the following reasons: (1) they did not include a homework intervention, (2) they did not assess academic achievement, (3) they did not include K-12 students, or (4) did not use an experimental design.We scanned 42 primary studies included in the EEF toolkit on the homework and 40 of them were excluded.The details are shown in Table 3.

| Risk of bias of RCTs
Two of 11 studies used randomized controlled trials, both of which were evaluated as having high risk of bias based on the ROB2 (Dolean & Lervag, 2021;Maertens & Johnston, 1972), especially in terms of bias in the measurement of the outcome (details are shown in Figure 3).

| Effects of interventions
Eleven publications were included in the present review, and the outcomes were divided into two groups based on the comparison: one comparison is homework versus no homework and another is more homework versus less homework.Fewer studies were included in the second comparison and could not be synthesized quantitatively.Thus, we provide a meta-analysis of the first comparison outcome and a narrative description of the second.
The other three suggested no difference was present between experimental and control groups (Koch, 1965;Maertens, 1969;McGrath, 1992).

Publication bias
As shown in Figure 7, all effect size was symmetrically distributed, and the results of the meta-analysis had no obvious change before and after trim and fill.The result of Egger's test also showed no significant publication bias (t = 2.20, p = 0.06).

Sensitivity test
We performed a sensitivity test using the "one-leave-out" method, and found no outliers.The effect sizes were from 0.383 to 4.484 (Figure 8).

| More homework versus less homework
Two publications explored the effectiveness of homework according to its duration (Dolean & Lervag, 2021;Koch, 1965).As mentioned above, Koch (1965) compared homework with no homework and the Compared with the writing group, the short-term effect of less homework on editing and spelling was worse in the Math group (d = −0.31and -0.346), but the effect was no longer clear after 4 months.However, increasing or decreasing the amount of math homework did not alter its impact on math skill (Dolean & Lervag, 2021).
Overall, the experimental findings provide partial evidence that appropriate homework contributes positively to academic achievement among K-12 students.However, the optimal time allocation for homework is still unclear.Unfortunately, the planned subgroup analyses were not feasible due to insufficient data.Consequently, potential differences across genders, grades, subjects, and regions remain untested, and the moderating and mediating variables remain unidentified.

| Summary of main results
The present review identified 11 publications that used experimental methods to assess the relationship between homework duration and academic outcomes.We grouped the studies into two categories: homework versus no homework, and long homework versus less homework.Given the limited number of studies and the variability in outcomes, we employed two approaches to present the results: a meta-analysis for comparison between assigning homework and not assigning homework, and a narrative synthesis for the comparison between assigning longer homework and less homework.
Two experiments explored the effectiveness of homework moderated by the homework time (Dolean & Lervag, 2021;Koch, 1965).In Koch (1965), the effect of long-daily homework (20 to 30 min) and short-daily homework (10-15 min) were compared.He suggested that achievement in arithmetic concepts following daily long homework assignments was higher than that with short homework assignments.
Recently, Dolean and Lervag (2021) confirmed the effect of homework on writing skill.Their results also support Koch (1965), in that increased time spent on homework (average 20 min) fostered higher gains in writing achievement (Dolean & Lervag, 2021).
We planned to conduct a subgroup analysis to examine the differences in homework effectiveness across factors such as gender, grade level, subject matter, and geographical regions.However, due to the scarcity of studies available for each subgroup, we could not execute all the planned subgroup analyses.
Additionally, we aimed to identify potential factors on the duration of homework, including the academic subject, task difficulty, type of homework, mode of homework, parental involvement, and feedback on homework.However, only two studies investigated the factors that could affect the effectiveness of homework.Foyle (1990) explored the difference between types of homework and found no significant disparities between groups that received either preparation-or practice-oriented homework (Foyle, 1990).Similarly, Maertens and Johnston (1972) examined the moderated effect of feedback and observed no obvious distinction in the effects of homework between students who received immediate feedback (per problem) and those who received delayed feedback (delayed until after the homework was fully complete) (Maertens & Johnston, 1972).

| Overall completeness and applicability of evidence
While we developed the search strategy under the guidance of information specialists, certain words, such as "attainment," were not included in the keywords for academic performance.This might have resulted in missing some relevant studies.Our search did not impose any language restrictions, however we did not include non-English databases, such as the widely-used Chinese database, CNKI.This could potentially limit the applicability of our conclusion in non-English contexts.Our most recent search was conducted in November 2021, more than 2 years before the publication of this review.This time gap has affected the current relevance of our evidence.Overall, given the limited number of studies included in our review and the fact that many of them are outdated, our findings should be interpreted with caution.
6.3 | Quality of the evidence

| Quality of RCTs included in the review
The two RCTs were assessed as at high risk of bias (Dolean & Lervag, 2021;Maertens & Johnston, 1972).In each domain, both studies were at low risk of bias in randomization processes, deviations from intended interventions, and missing outcome data.However, they were assessed as having a high risk in measurement of outcomes, with some risk of bias in the selection of the reported results.
As mentioned above, the method used to measure academic performance could potentially affect its relationship with homework time (Cooper et al., 2006;Fan et al., 2017).These two RCTs neither of the two studies reported on the validity and reliability of the questionnaires (Kinyua & Odiemo, 2018;Setiabudi et al., 2019).

| Quality of NRCTs included in the review
None of nine NRCTs were at low risk of bias, four were at moderate risk (Koch, 1965;Maertens, 1969;McGrath, 1992;Tupesis, 1972), two at serious risk (Foran & Weber, 1939;Gray & Allison, 1971), and three at critical risk (Ewart Anderson, 1946;Foyle, 1990;Townsend, 1995).All of these studies were at low risk of selection bias, classification of interventions bias, reporting bias, and bias of deviations from interventions.
In contrast to RCTs, the absence of random participant assignment in quasi-experiments can lead to a higher susceptibility to bias.
This can result in the blending of intervention effects with confounding variables.Despite this, statistical techniques can be employed to control or balance these biases (Waddington et al., 2017).
The effectiveness of homework on academic achievement.
F I G U R E 7 Funnel plot of the effectiveness of homework.
Out of the nine studies included in this review that used a quasiexperimental design, only four factored in the Intelligence Quotient (IQ) before the intervention (Foran & Weber, 1939;Gray & Allison, 1971;Koch, 1965;Tupesis, 1972).Additionally, one study considered both the mental ability of students and the effectiveness of teachers (Ewart Anderson, 1946).It is important to note, however, that the impact of homework on academic achievement is influenced by a variety of confounding factors.These include the student's study skills, teacher effectiveness, socioeconomic status, parental involvement, and homework management strategies (Deysolong, 2023;Trautwein, 2003;Van Voorhis, 2003;Xu, 2009).

| Potential biases in the review process
There were several limitations with respect to the review process.
Although many studies evaluated the effectiveness of homework, few studies considered time spent on the homework.Even if studies mentioned homework duration, the time scales they covered were narrow, ranging from 10 to 30 min, which may lead to a floor effect.In addition, we tried to include all relevant studies in the selection process, but several studies were unavailable due to our limited authority over databases and lack of reply from the authors.

| Agreements and disagreements with other studies or reviews
The main findings of this present review is that homework is beneficial for primary school students, especially in arithmetic computation and arithmetic problems-solving skills.This conclusion is generally consistent with previous reviews that indicated a positive correlation between homework duration and academic performance (Baş et al., 2017;Cooper et al., 2006;Cooper, 1989;Fan et al., 2017).
However, it is important to note that due to insufficient data, this review only included students from 2nd to 8th students.Therefore, it remains unclear whether this conclusion can be generalized to other age groups.
Considering the workload of homework, some researchers have proposed a curvilinear relationship between homework duration and academic achievement (Fernández-Alonso et al., 2015;Keith, 1982;Krejtz et al., 2018;Reteig et al., 2019), For example, Fernández-Alonso et al. (2015), who surveyed 7725 Spanish adolescents with an average age of 13.78 years, concluded that the optimum duration of homework was 1 h per day for mathematics and science (Fernández-Alonso et al., 2015).However, our review included a limited number of studies with narrow homework time scales (10-30 min), we were unable to ascertain the pattern (be it linear or curvilinear) of the relationship between homework time and academic performance, much less the optimal amount of homework time.
Previous studies have also demonstrated that the relationship between homework and academic performance is influenced by factors such as the gender (Cadime et al., 2017), grade level (Baş et al., 2017), region (Zhu, 2015), publication year (Gill & Schlossman, 2004;Twenge et al., 2004), measurement tool (Fan et al., 2017), type of homework (Qiao & Fan, 2020), and subject (Fan et al., 2017;Trautwein & Lüdtke, 2007).We intended to identify the potential moderators using subgroup analysis in this review, but we were unable to do so due to insufficient data.

F I G U R E 8
The forest plot for sensitivity test.

−
Population: K-12 school students with no disabilities or not attending special education schools; − Intervention: Homework assigned regularly by schoolteachers to students to complete during non-school hours; − Comparison: Different time spent on the homework; − Outcomes: Academic performance was the primary outcome.The secondary outcomes were academic motivation and the quality of homework; − Study design: Treatment-control group design or comparison group design studies.

Figure 1
Figure 1 provides a visual representation of the framework, which shows how interventions are hypothesized to lead to the intended outcomes.As in Figure 1, the law of readiness suggests that before commencing a certain learning activity, if the learners perform well in the preparatory stages (including physiological and psychological performance) related to the corresponding learning activities, they can Social Sciences Citation Index (Web of Science, Clarivate) #1 TI=homework OR AB=homework#2 TI=home-work OR AB=home-work#3 #1 OR #2#4 TS = K-12 OR TS=preschool student* OR TS=pre-school student* OR TS=Kindergarten student* OR TS= middle school student* OR TS=high school student* OR TS=senior school student OR TS=primary school student* OR TS=pupil OR TS= schoolchild OR TS= junior high school student* OR TS=school-age#5 TS=achievement OR TS=performance OR TS=grade OR TS=score OR TS= academic achievement* OR TS = GPA OR TS=academic performance#6 #3 AND #4 AND #5 1832 Publisher platforms ScienceDirect https://www.sciencedirect.com#1 TI=homework OR AB=homework#2 TI=home-work OR AB=home-work#3 #1 OR #2#4 TS = K-12 OR TS=preschool student* OR TS=pre-school student* OR TS=Kindergarten student* OR TS= middle school student* OR TS=high school student* OR TS=senior school student* OR TS=primary school student* OR TS=pupil OR TS= schoolchild OR TS= junior high school student* OR TS=school-age#5 TS=achievement OR TS=performance OR TS=grade OR TS=score OR TS= academic achievement* OR TS = GPA OR TS=academic performance#6 #3 AND #4 AND #5 555 Taylor & Francis Online Database https://www.tandfonline.com[[Publication Title: homework] OR [Publication Title: home-work] OR [Publication Title: homework] OR [Publication Title: home-work]] AND [[Abstract: k-12] OR [Abstract: "preschool student"] OR [Abstract: "pre-school student"] OR [Abstract: "kindergarten student"] OR [Abstract: "middle school student"] OR [Abstract: "high school student"] OR [Abstract: "senior school student"] OR [Abstract: "primary school student"] OR [Abstract: pupil] OR [Abstract: schoolchild] OR [Abstract: "junior high school student"] OR [Abstract: "school-age"]] AND [[Abstract: achievement] OR [Abstract: performance] OR [Abstract: grade] OR [Abstract: score] OR [Abstract: "academic achievement"] OR [Abstract: gpa] OR [Abstract: "academic performance" , the information extraction and coding form consisted of two parts.The first was general information, including the information of primary study (publication source, year of publication, and year of data collection), sample characteristics (e.g., sample size, gender, grade level, region, family economic status, parental education level), methodological characteristics (e.g., sampling method, measures of homework duration, and measures of academic performance), and intervention characteristics (e.g., subject, mode of homework, and type of homework).The other was the effect size, including that of homework duration and test score.Before the formal data extraction, we performed three rounds of extraction with a pre-piloted data extraction form.This process was conducted independently by two of the authors (Zheng Xu and Xing Xin).Disagreements between coders were resolved by discussions with another author (Xiuxia Li).− If a study contained multiple interventions (e.g., different homework modalities such as online vs. book-based), eligible data were extracted by the reviewers; whether individuals may have undergone multiple interventions, whether there were multiple treatment groups and whether several studies are based on the same data source.− Clustered assignment of treatment.Cluster randomized trials included in this review were checked for consistency in the unit of allocation and the unit of analysis, as statistical analysis errors can occur when they are different.− Multiple intervention groups and multiple interventions per individual.Studies with multiple intervention groups with different individuals (e.g., different homework modalities such as online vs. limited homework time to within half an hour, when teachers designed the regular homework for experimental classes.Dolean and Lervag (2021) allocated students to three groups with different amounts of homework in writing and math: a writing group (students received more homework in writing and less in math), balanced group (students received similar amounts of homework in writing and math), and math group (students received more homework in math and less in writing); the balanced group was a potential comparison(Dolean & Lervag, 2021).The remaining studies simply assigned homework to experimental groups and used the students with no homework as their comparisons.
Only one study was published online recently in 2021, while the remaining studies spanned from 1939 to 1995.This may reflect the changing attitudes towards homework over time.In the early 20th century, educators widely believed that homework contributed to creating disciplined minds.However, by 1940, concern arose that homework might interfere with other home activities, leading to a backlash against its practice.This trend shifted in the late 1950s when the U.S. government was concerned that their education lacked rigor.Schools then F I G U R E 3 Risk of bias of RCTs included in the review.viewed more rigorous homework as a partial solution to this problem.However, in the 1980s, some researchers argued that excessive homework could be detrimental to students' mental health, and deprive students of essential free time(Costley, 2013;Güven & Akçay, 2019).The results of TIMSS indicates that the trend of assigning homework from 2003 to 2015 has been negative in both 4th and 8th grades in all selected countries, except in Turkey(Güven & Akçay, 2019).The attitudes toward homework have fluctuated significantly over the years, reflecting changing educational philosophies and societal concerns, which also affected the education research direction.

F
I G U R E 4 Risk of bias of non-RCTs included in the review.F I G U R E 5 The summary of risk of bias of non-RCTs.employed teacher-designed tests instead of general standardized tests.Although teacher-designed tests can be both valid and reliable assessment tools if the teachers are properly trained in test construction and understand the principles of validity and reliability,