Journal of Educational Psychology © 2015 American Psychological Association

2015, Vol. 107, No. 4, 1075–1085 0022-0663/15/$12.00 http://dx.doi.org/10.1037/edu0000032

Adolescents’ Homework Performance in Mathematics and Science:

Personal Factors and Teaching Practices

Rubén Fernández-Alonso Javier Suárez-Álvarez

Government of the Principality of Asturias, Oviedo, Asturias, University of Oviedo
Spain, and University of Oviedo

José Muñiz
University of Oviedo and Biomedical Research Network in Mental Health (CIBERSAM), Madrid, Spain

Classical educational research provides empirical evidence of the positive effect of doing homework on
academic results. Nonetheless, when this effect is analyzed in detail there are inconsistent, and in some
cases, contradictory results. The central aim of this study was to systematically investigate the effect of
homework on performance of students in mathematics and science using multilevel models. The original
sample consisted of 7,725 Spanish adolescents with a mean age of 13.78 (⫾0.82) of which 7,451 were
evaluated after purging the sample of the students who did little to no homework. A 2-level hierarchical-
linear analysis was performed, student and class, with 4 individual adjustment variables: gender,
socioeconomic and cultural level, year repetition, and school grades, which were used to reflect previous
student achievement. The individual level examined time spent, effort made, and the way homework was
done. The class level considered frequency of assignment and quantity of homework. Prior knowledge,
estimated using school grades, is shown to be the most important predictor of achievement in the study.
Its effect is greater than the combined effect of all the other variables studied. Once background factors
are controlled, the homework variables with most impact on the test are student autonomy and frequency
of homework assignment by teachers. Autonomy when doing homework was shown to be the most
important individual-level variable in both mathematics and science, and not effort and or time spent
doing homework. The optimum duration of homework was found to be 1 hr a day.

Keywords: homework, mathematics, science, academic performance, multilevel models

The characteristics and effects of homework have been widely Cooper, & Robinson, 2008; Pomerantz, Moorman, & Litwack,
studied, both from a general perspective (Blazer, 2009), and spe- 2007), the role of teachers (Epstein & van Voorhis, 2001), stu-
cific areas such as the relationship between homework and aca- dents’ perceptions (Warton, 2001), and the way the students man-
demic performance (Cooper, 1989a; Cooper, Robinson, & Patall, age and do their homework (Xu, 2013). Research about the rela-
2006; Cooper & Valentine, 2001; Cooper, Steenbergen-Hu, & tionship between homework and academic performance has
Dent, 2012; Goldstein, 1960; Paschal, Weinstein, & Walberg, evolved from positivist and experimentalist positions (Goldstein,
1984; Trautwein & Köller, 2003), family involvement (Hoover- 1960) toward more integrated, comprehensive approaches (Coo-
Dempsey, Battiato, Walker, Reed, DeJong, & Jones, 2001; Patall, per, 1989a; Cooper et al., 2006; Paschal et al., 1984). The general
conclusion of this research is that there is a positive relationship
between homework and academic performance but recent work
This article was published Online First March 16, 2015. has led to a profound change in this general conclusion. Trautwein
Rubén Fernández-Alonso, Department of Education, Principality of As- and Köller (2003) severely criticized large parts of the previous
turias Government, Oviedo, Asturias, Spain, and Department of Education research on this topic and laid the foundations for a new generation
Sciences, University of Oviedo; Javier Suárez-Álvarez, Department of
of studies, signaling the need to use more sophisticated statistical
Psychology, University of Oviedo; José Muñiz, Department of Psychology,
University of Oviedo, and Biomedical Research Network in Mental Health
models to allow a more rigorous analysis of the variables involved.
(CIBERSAM), Madrid, Spain. With this new focus, the most commonly used models have been
This research was funded by the Ministerio de Economía y Competi- two-level hierarchical-linear models, generally student and class,
tividad del Gobierno de España. References: PSI2011-28638, BES2012- although other designs are possible. These multilevel models have
053488. We would like to express our utmost gratitude to the Government been used to predict academic performance (De Jong, Westerhof,
of the Principality of Asturias (Spain), without whose collaboration this & Creemers, 2000; Dettmers, Trautwein, & Lüdtke, 2009; Núñez,
research would not have been possible. Finally, we would like to express
Vallejo, Rosario, Tuero, & Valle, 2014; Trautwein, Köller,
our sincerest gratitude to the reviewers; their contributions have helped to
improve considerably the final manuscript.
Schmitz, & Baumert, 2002; Tymms & Fitz-Gibbon, 1992), the
Correspondence concerning this article should be addressed to Javier time spent on homework (Trautwein, 2007), the effort made (Trau-
Suárez-Álvarez, Department of Psychology, University of Oviedo, 33003 twein, Niggli, Schnyder, & Lüdtke, 2009; Trautwein, Schnyder,
Oviedo, Spain. E-mail: [email protected] Niggli, Neumann, & Lüdtke, 2009), self-regulation (Xu, 2010a),
1075
1076 FERNÁNDEZ-ALONSO, SUÁREZ-ÁLVAREZ, AND MUÑIZ

interest in the assignments (Xu, 2008, 2011), motivation (Dett- negative. It seems that effort is more related to motivation and
mers, Trautwein, Lüdtke, Kunter, & Baumert, 2010; Trautwein, personality traits (Trautwein et al., 2006), the role of the teacher
Lüdtke, Schnyder, & Niggli, 2006), reasons for completing home- (Patall et al., 2010; Xu, 2011), and previous performance (Xu, 2008).
work (Xu, 2010b), emotions felt (Dettmers, Trautwein, Lüdtke, Another key variable related to homework is the ability to work
Goetz, Pekrun, & Frenzel, 2011), behavioral strategies used (Lub- autonomously, that is, without parental involvement. De Jong et al.
bers, Van Der Werf, Kuyper, & Hendriks, 2010), and the attitude (2000) demonstrate that, after controlling adjustment variables, the
toward subjects (Farrow, Tymms, & Henderson, 1999). Three- effect of family help on results in mathematics is no longer statisti-
level hierarchical models have also been used, although less fre- cally significant. In addition, family help does not seem to be asso-
quently, with fundamental designs such as student, school and ciated with either effort (Trautwein & Lüdtke, 2007) or expectations
system (OECD, 2013), student, class, and school (Murillo & of success, (Trautwein & Lüdtke, 2009) nor with interest in or
Martínez-Garrido, 2013), and other three-level combinations (Pat- enthusiasm for doing homework (Xu, 2008). Xu (2010a) found that
tal, Cooper, & Wynn, 2010; Trautwein & Lüdtke, 2007, 2009). students who needed less help, that is, those who were more auton-
In the first multilevel study centered on the relationship between omous when doing homework, achieved better results, whereas Trau-
homework and academic performance Tymms and Fitz-Gibbon twein and Lüdtke (2007) observed that students who receive more
(1992) found differing results depending on the analysis level. In help also spend more time and, as already noted, get worse results.
general the data indicate that, at an individual level, time spent has Taken together, this data seems to indicate that students who need
little effect on academic performance (De Jong et al., 2000), and in more help and therefore, are less autonomous when doing homework,
some cases there was even a negative correlation, so that the more tend to demonstrate more difficulties with learning, motivation and
time spent on homework, the worse the academic performance (Trau- concentration, and fewer self-regulating strategies. Other studies, not
twein, 2007; Trautwein, Schnyder et al., 2009). Nevertheless, these using multilevel approaches, also confirm the relationship between
three studies found a statistically significant, positive relationship at autonomy in homework and academic achievement (Cooper, Lind-
the class level, that is, the students in classes that are assigned more say, & Nye, 2000).
homework tend to have better academic results. De Jong et al. (2000) The field of the study of homework is a crossroads where psycho-
note the possibility that students with lower levels of previous knowl- logical research and the study of school effectiveness and educational
edge need more time to complete their assignments, which leads to improvement meet. Thanks to this combination we know that maxi-
the negative correlation. Trautwein (2007) and Trautwein and Ludtke mizing the homework-result relationship requires a mix of certain
(2007) empirically confirmed this hypothesis, and found that the individual (personal) characteristics and certain teaching practices.
students that were more able, who knew more, and who had higher Psychology gives unified models that combine input from neurobiol-
expectations of success tended to spend less time on their homework. ogy with research on cognitive factors and personality involved in
On the other hand, it seems clear that the relationship between time learning, highlighting the importance of prior knowledge as the prin-
spent doing homework and academic results is not linear (Blazer, cipal predictive variable of future achievement (Shell, Brooks,
2009; Cooper & Valentine, 2001; Daw, 2012; Gobierno del Princi- Trainin, Wilson, Kauffman, & Herr, 2010). Failure to consider prior
pado de Asturias, 2013; Gobierno Vasco, 2012; Ministerio de Edu- knowledge in research into homework could undermine any conclu-
cación, Cultura y Deporte, 2003). sions made (De Jong et al., 2000). At the same time, research into
Frequency of homework, in the same way as time spent on it, personality factors and student attitudes has underscored the influence
presents two hallmark characteristics of a multilevel situation. The of motivation and effort, autonomy and regulation of behavior, and
meaning it has depends on the measurement level; at an individual expectations about homework and the role of conscientiousness when
level it indicates the student’s work habits, and taken as a class it comes to optimizing the homework-result relationship (Lubbers et
average it is an estimator of the frequency that the teacher assigns al., 2010; Trautwein et al., 2006; Trautwein & Ludtke, 2009; Xu,
homework (Trautwein & Köller, 2003). In addition, it demonstrates 2008; Zimmerman, 2001; Zimmerman & Kitsantas, 2005). Research
differential effects depending on the level of analysis. With data on school effectiveness has also been concerned with the effect of
aggregated by class, Farrow et al. (1999) found that frequency had a teaching practices related to assignment, frequency, and other char-
small positive effect on performance in mathematics and science, but acteristics of homework. Scheerens, Witziers, and Steen (2013) used
at the student level the effect was negligible. In any case, the data a concept of curriculum quality that includes three variables: oppor-
point to the fact that the teacher regularly assigning homework matters tunity to learn, effective learning time, and homework. Although in
more than the quantity. Trautwein (2007) and Trautwein et al. (2002) their meta-analysis the estimated mean effect size of homework is
not only reported a positive association between the frequency of moderate (0.073), the value placed on homework in teaching practices
homework assignments and performance in mathematics, they also is notable. As Dettmers et al. (2010) noted, homework works only so
indicated that when the frequency is considered, the amount of the far as the teacher assigns work of sufficient quality.
homework is no longer significant. However, frequency and time are Within this context the general aim of our work was to systemat-
not the only variables that affect academic performance. Dettmers et ically study the effect of doing homework on academic results in
al. (2010, 2011) and Trautwein and Lüdtke (2007) indicate that effort mathematics and science subjects in adolescents. Multilevel models
is a better predictor of results in mathematics than either frequency or were used with two levels: students and classes. The student level
time spent. In fact effort has a central role in the theoretical model examined time spent on homework, effort made and how it was done
proposed by the Max Planck Institute (Trautwein et al., 2006). The while the class level evaluated the frequency and quantity of home-
relationship between effort and academic results generated a series of work set. In addition the effects of four background variables were
work aimed at studying the factors that predict effort. Curiously time controlled: gender, socioeconomic and cultural level, and two mea-
does not appear to be among them, and as Trautwein (2007) indicated, sures of previous academic performance (school grades and repetition
the correlation between time spent on homework and effort is low and of school year). The use of the latest generation hierarchical-linear
 EFFECT OF HOMEWORK PERFORMANCE IN MATH AND SCIENCE 1077

models will allow a much more precise, rigorous analysis of the almost always. Analysis of the principal components showed that
complex relationships between homework and academic performance these items conform to an essentially unidimensional scale (the first
(Vallejo Seco, Ato García, Fernández García, & Livacic Rojas, 2013). factor accounts for almost 64% of the common variance). In addition,
This is the first study on homework carried out on a large sample of the Cronbach’s ␣ coefficient (␣ ⫽ .71) seems appropriate considering
Spanish adolescents that combines the application of multilevel mod- that the scale is only composed of three items.
els and the psychometric advantages of Item Response Theory mod- The way homework was done was evaluated using the question
els. Homework is a significant part of the day to day lives of millions “How do you do your school assignments?” The options were (a)
of students, parents, and teachers all over the world. Providing new without help, (b) I occasionally need help, (c) I often need help, and
data on homework in a southern European population gives added (d) I always need help. A binary variable was constructed from this
confirmatory value, bearing in mind the current debate in psychology item where 1 signifies doing homework without help or occasional
and education on the need for accumulative and replicable science help and 0 signifies those who need frequent or constant help.
(Koole & Lakens, 2012; Nosek & Lakens, 2014). We hope, therefore, Tests of academic performance: Mathematics and science.
that our work can contribute to a crosscultural consolidation of current Academic performance was measured using a test that evaluated
research into homework. mathematics and science, the detailed description of the content may
be found in Gobierno del Principado de Asturias (2012). An item
Method bank was created that was purified through a pilot study and following
the protocol established by Fernández-Alonso (2005). Finally, 192
Participants items were selected (96 mathematic and 96 science). There were three
item formats used: multiple choice, short open-ended answer, and
The sample was made up of 7,725 students in the second year of essay. The item pool was distributed in eight test booklets in accor-
obligatory secondary education in the Principality of Asturias, a dance with the matrix design given in Fernández-Alonso and Muñiz
region in the north of Spain. The mean age was 13.78 with a SD of (2011). Each student was presented with a test booklet containing 48
0.82. Girls made up 47.2% of the sample; 90.6% were Spanish items (24 mathematics and 24 sciences) that they completed in two
nationals; and 72.9% were in the standard school year for their age sessions of 50 min each with a break in between.
whereas 27.1% were 1 or 2 years behind in relation to their age. The Given that the group of items was constructed to measure two
students were from 353 classes belonging to 148 schools. Technically competences (mathematics and science), the item pool was adjusted
this is the population as the evaluation was carried out on all of the to a two-dimensional item response model. It was assumed that each
registered students at the level studied in this geographical region in item saturated in one and only one of the dimensions, for which a
the academic year 2010 –2011. two-dimensional model among items was chosen (multidimensional
between-item tests). The fitting of the bank was conducted with the
Procedure ConQuest 2.0 program (Wu, Adams, Wilson, & Haldane, 2007). For
the initial calibration of items, 2,000 subjects were selected from the
The tests were performed within the official annual Principality of
total sample. This ensured that each item was responded to by 500
Asturias educational system evaluation program. Test management
subjects. Once the adjustment of the items to the model was checked,
within each school was provided by the school authorities. The
the parameters of these were estimated. These parameters were used
questionnaires and tests were given by active teachers in accordance
as initial values to calculate the score of the subjects on each compe-
with the Instructions for the development of diagnostic evaluation, set
tence in function of the response to the items. Finally, the marks were
by the Department of Education of the Principality of Asturias Gov-
transformed to a scale with a mean of 500 and a SD of 100. The
ernment. Each student completed a mathematics and science test with
difficulty indexes of the tests comprise the entire possible rank of
24 items for each subject, as well as a questionnaire for homework
scores, both easy items and difficult items, although most of them
evaluation.
(around 40%) were concentrated on the middle range of the distribu-
tion of difficulty. The mean Cronbach’s ␣ coefficient of the eight
Instruments booklets was 0.85 (with a minimum of 0.82 and a maximum of 0.88).
Questionnaire for homework evaluation. The total daily time Adjustment variables. Four adjustment variables were in-
dedicated to homework was calculated from two multiple choice cluded. Two variables were used to look at sociological background:
items on the questionnaire. The first asked about the frequency that gender and Socioeconomic and Cultural Index (ISEC) of the partic-
homework was done, the options were (a) never, (b) 2 or 3 days per ipants. To ensure the quality of the ISEC, teachers were asked to
week, (c) almost every day, and (d) every day. The second item asked report on the educational attainment and professions of the students’
about the total time spent on homework considering all subjects, the families. This information was reduced, using an analysis of principal
options were (a) less than 30 min, (b) between 30 and 60 min, (c) components, to an index N(0,1) following the procedure described in
between 60 and 120 min, and (d) more than 120 min. For the second Peña-Suárez, Fernández-Alonso, and Muñiz (2009). The remaining
level analysis, the means of these variables for each class were variables were used as measures of prior knowledge. School history
calculated as estimators of the frequency of homework assignment by was estimated using the variable Repetition of school year, a binary
teachers and the amount of homework, respectively. variable where 0 meant being in the school year corresponding to age
Effort was measured using three items (I make an effort to get good and 1 signified those who were repeating a year. Finally, the prior
marks, I finish my homework even if they are difficult or they take me knowledge variable was measured using the school grades of the
a long time, and I am careful to keep my notebooks and work neat and students in mathematics and science. The school grades are presented
tidy) where 1 ⫽ means never or almost never and 4 ⫽ always or in the following measurement scale (based on the Spanish grade
1078 FERNÁNDEZ-ALONSO, SUÁREZ-ÁLVAREZ, AND MUÑIZ

system): fail (0 points), pass or good (1 point), very good (2 points), at the true scale. The normalized data are presented in the tables and
and outstanding (3 points). is interpreted as a standardized coefficient of ordinary regression
analysis. The range of missing values in the variables varied between
Data Analyses 5% and 12%. To impute missing values, two methods were used
depending on the type of missing data presented: when a case was
First, to study students’ homework behavior, descriptive statistics incomplete the data point was replaced by the mean of the person,
and Pearson correlations were calculated. Second, a multilevel anal- whereas for completely missing data we used a procedure of multiple
ysis was performed to study the effect of homework on academic imputation with auxiliary variables in the SPSS program (IBM Corp,
performance. With the aim of improving the representation of stu- 2011). Fernández-Alonso, Suárez-Álvarez, and Muñiz (2012) found
dents who actually do the assigned homework, the 2.4% of students that use of this combination is especially recommended when data
who never do homework were removed from the analysis (see Figure loss is greater, and its loss mechanism is more conditioned.
1). For this reason, the multilevel analyses looked at 7,451 students
rather than the 7,725 originally evaluated. A sequence of five
hierarchical-linear models was constructed by subject (mathematics Results
and science). These are models of random two-level coefficients:
student and class (Raudenbush & Bryk, 2002). Each new model Completion of Homework by the Students
increases in complexity given that new explanatory variables are
added while keeping those in the previous model. The modeling Figure 1 shows the frequency distribution of the items describ-
strategy used HLM 6.03 software (Raudenbush, Bryk, Cheong, & ing the completion of homework by the students. As can be seen,
Congdon, 2004). HLM does not give standardized regression coeffi- the students spend, on average, between 1 and 2 hr doing home-
cients, therefore the data presented in this study are presented in two work from all subjects.
ways: the original scale N(500, 100) and the normalized scale N(0, 1). The total amount of homework assigned by teachers is a little
The former is used in the graphics and shows the predictions of results over 70 min a day (see Table 1). According to Murillo and

Frequency of doing homework (N = 7,725) Total daily time spent on homework (N = 7,541)
60 60
53.7%

50 50 47.9%

40 40
35.1%

30 30
26.1%
20.7%
20 20

10
8.8% 10
5.3%
2.4%
0 0
Never Every few Almost every Every day Less than 30 Between 30 and Between 60 and More than 120
days day minutes 60 minutes 120 minutes minutes

Effort made on homework (N = 7,541) Frequency that student needed help (N = 7,541)
60 60
54.4%

50 50
44.1%

40 40
32.7% 33.5%

30 30

20.2%
20 20

9.5%
10 10
3.0% 2.7%
0 0
Never or almost Sometimes Often Always or Never or almost Sometimes Often Always or
never almost always never almost always

Figure 1. Completion of homework by students.

 EFFECT OF HOMEWORK PERFORMANCE IN MATH AND SCIENCE 1079

Table 1
Descriptive Statistics and Pearson Correlations of the Variables

M SD 1 2 3 4 5 6 7 8 9 10 11

1. Homework time (total daily minutes) 73.24 39.09 —

2. Effort 2.08 0.71 .51 —
3. How homework is done (1 ⫽ without help) 0.88 0.33 ⫺.05 .08 —
4. Sex (1 ⫽ girl) 0.48 0.50 .15 .19 .00 —
5. Socioeconomic and cultural index (SEC) 12.41 3.51 .14 .19 .06 ⫺.02 —
6. Repeater (1 ⫽ yes) 0.26 0.44 ⫺.28 ⫺.36 ⫺.11 ⫺.07 ⫺.34 —
7. Prior knowledge in mathematics 1.36 0.88 .21 .38 .17 .03 .32 ⫺.46 —
8. Prior knowledge in science 1.59 0.92 .28 .46 .17 .06 .38 ⫺.55 .70 —
9. Score in mathematics 501.7 99.9 .13 .26 .18 ⫺.03 .35 ⫺.39 .57 .55 —
10. Score in science 501.3 99.9 .07 .18 .16 ⫺.11 .32 ⫺.32 .47 .50 .66 —
11. Frequency of homework assignment (class) 2.39 0.27 .23 .21 .02 .04 .17 ⫺.19 .15 .20 .20 .17 —
12. Homework time (class) 72.48 13.46 .32 .21 ⫺.02 .05 .18 ⫺.18 .11 .17 .18 .14 .76

Martinez-Garrido (2013) in Spanish-speaking countries, including Multilevel Models to Predict Academic Performance
Spain, homework assignments are usually paper and pencil tasks
such as complementary lectures (81.5% of teachers assign this Tables 2 and 3 show the results of the hierarchical-linear
kind of homework often or very often), problems to solve (66.3% models. The null model, without independent variables, indi-
of teachers assign often or very often), write an essay or research cates that 82% of the variance of results in mathematics and
(53.1% often or very often), and small research or experiments 87.1% of results in science are found in Level 1 (students),
(44%). According to this data, the teachers show a certain unifor- while the variance between classes is 18% in mathematics and
mity in the practice of assigning homework meaning that the 12.9% in science. Model 1 includes the variables time spent,
difference in the time taken to do homework variable is more about frequency of assignment, and amount of homework. In the first
the student’s level (i.e., 94.6% of the variance is because of the place, it confirms that both the time-result relationship and the
difference in student commitment and only 6.4% is because of amount-result relationship are curvilinear. In both cases the
differing homework practices on the part of the teachers). This effect of the quadratic element is statistically significant (95%
latter value is slightly lower than the international mean in PISA confidence interval [CI]) albeit negative. In other words, there
2003 (7.6%) and slightly higher than the Spanish estimate (5.4%) is an optimal amount of, and time to spend on, homework and
in that same study (Dettmers et al., 2009). once that threshold is passed, the time spent and the amount of

Table 2
Hierarchical-Linear Models to Predict Mathematics Performance

Null model: Model 1: Time, Model 2: Effort and Model 3: Adjustment Model 4: Prior
Without predictors quantity, and frequency autonomy variables knowledge
B (SE) B (SE) B (SE) B (SE) B (SE)

Level 1
Total daily minutes — 0.23 (0.04) 0.02 (0.04) 0.01 (0.04) ⫺0.07 (0.04)
Total daily minutes ˆ 2 — ⫺0.15 (0.04) ⫺0.02 (0.04) ⫺0.01 (0.04) 0.03 (0.04)
How homework is done (1 ⫽ without
help) — — 0.16 (0.01) 0.15 (0.01) 0.09 (0.01)
Effort — — 0.19 (0.01) 0.18 (0.01) 0.04 (0.01)
Sex (1 ⫽ girls) — — — ⫺0.07 (0.01) ⫺0.06 (0.01)
Socioeconomic and cultural index (SEC) — — — 0.23 (0.01) 0.11 (0.01)
Repeater — — — — ⫺0.09 (0.01)
Grades in mathematics — — — — 0.43 (0.01)
Level 2
Frequency of homework assignment — 0.13 (0.03) 0.11 (0.03) 0.10 (0.03) 0.07 (0.03)
Quantity: Average minutes per day per
class — 0.29 (0.14) 0.30 (0.13) 0.17 (0.12) 0.06 (0.11)
Quantity: Average minutes per day per
class ˆ 2 — ⫺0.25 (0.12) ⫺0.24 (0.12) ⫺0.15 (0.10) ⫺0.03 (0.10)
Variance
Between classes 0.180 0.136 0.125 0.086 0.079
Within classes 0.820 0.813 0.759 0.723 0.561
Percentage of explained variance
Between classes 24.4% 30.6% 55.2% 56.1%
Within classes 0.9% 7.4% 11.8% 31.6%
Total 5.1% 11.6% 19.1% 36.0%
1080 FERNÁNDEZ-ALONSO, SUÁREZ-ÁLVAREZ, AND MUÑIZ

Table 3
Hierarchical-Linear Models to Predict Science Performance

Null model: Model 1: Time, Model 2: Effort Model 3: Adjustment Model 4: Prior
Without predictors quantity, and frequency and autonomy variables knowledge
B (SE) B (SE) B (SE) B (SE) B (SE)

Level 1
Total daily minutes — 0.17 (0.04) 0.02 (0.04) 0.01 (0.04) ⫺0.09 (0.04)
Total daily minutes ˆ 2 — ⫺0.14 (0.04) ⫺0.04 (0.04) ⫺0.04 (0.04) 0.02 (0.04)
How homework is done (1 ⫽ without help) — — 0.15 (0.01) 0.13 (0.01) 0.08 (0.01)
Effort — — 0.14 (0.01) 0.13 (0.01) ⫺0.01 (0.01)
Sex (1 ⫽ girls) — — — ⫺0.13 (0.01) ⫺0.12 (0.01)
Socioeconomic and cultural index (SEC) — — — 0.24 (0.01) 0.12 (0.01)
Repeater — — — — ⫺0.04 (0.01)
Grades in science — — — — 0.42 (0.01)
Level 2
Frequency of homework assignment — 0.13 (0.03) 0.11 (0.03) 0.10 (0.03) 0.06 (0.02)
Quantity: Average minutes per day per class — 0.21 (0.11) 0.21 (0.11) 0.07 (0.10) 0.03 (0.08)
Quantity: Average minutes per day per class ˆ 2 — ⫺0.18 (0.10) ⫺0.17 (0.10) ⫺0.06 (0.08) 0.01 (0.07)
Variance
Between classes 0.129 0.099 0.094 0.064 0.057
Within classes 0.871 0.869 0.832 0.777 0.650
Percentage of explained variance
Between classes 23.3% 27.1% 50.4% 55.8%
Within classes 0.2% 4.5% 10.8% 25.4%
Total 3.2% 7.4% 15.9% 29.3%

homework stop being effective and end up being detrimental similar: around 1 hr a day of homework seems to be sufficient time
(see Figure 2). to achieve satisfactory results.
The top of the curve is between 90 and 100 min per day of As can be seen in Tables 2 and 3, the frequency of assignment
homework for all subjects combined and this may be considered gives results consistent with Model 1: more frequent homework
the most effective amount of time, as from this point on results assignment, better results. In contrast, the amount of homework
begin to decline. However, the most effective is not always the gives more erratic results. Although in mathematics the effect of
most efficient. For example, for students in classes where the homework volume is significant, in the case of science, the value
amount of homework is 50 min, results in the middle of the scale is at the limit of statistical significance (p ⫽ .063). In short, Model
(i.e., 504 points in science) are predicted. The gain expected in 1 indicates that, keeping the other variables constant, a difference
science for students in classes where the amount of homework is of near 50 points is predicted between those students in classes
70 min is around an eighth of a SD (517 points) but the gains where the teachers systematically assign homework and those
between 70 and 90 min barely reach 3% of a SD. That small gain students in classes where homework is assigned occasionally or
requires 2 hr more homework per week, which is a large time sporadically. The predictions of Model 1 seem reasonable, al-
investment for such small gains. For that reason, assigning more though the percentage of variance it explains (especially the vari-
than 70 min homework per day does not seem very efficient, as the ance within the classes) is small. That suggests the existence of
expectation of improved results is very low. In the case of math- other factors related to homework that account for the differences
ematics the curve is more pronounced but the interpretation is between the classes and the students.
Model 2 explores the role of the two individual variables not
related to time: effort and needing help to do homework. This
model shows that both effort and autonomy in doing homework
have more explanatory power than the variables considered in the
previous model. In fact, time spent loses statistical significance
whereas the variables related to teaching practice in terms of
homework hardly change compared with the previous model.
Finally, Model 2 explains twice the total variance that Model 1
does. This explanatory gain is produced in the variance within the
centers, which confirms that the effort and the way the homework
is done is more important than the time spent, the frequency or the
amount of homework. Figure 3 shows the prediction for mathe-
matics according to the need for help with homework, supposing
constant effort by students and homework assigned daily by the
teacher.
Figure 2. Prediction of performance as a function of total daily time spent Figure 3 allows two interpretations to be made. First, student
on homework. autonomy is more effective as it predicts 54 points more than for
 EFFECT OF HOMEWORK PERFORMANCE IN MATH AND SCIENCE 1081

possibly those who have bigger gaps in their learning or problems

concentrating. Autonomy, that is, doing homework without paren-
tal involvement was shown to be the most important individual
level variable, it is, after all, the only variable related to student
behavior toward homework that is significant, both in mathematics
and in science. The homework effort variable gave an unexpected
result: in science it lacks statistical significance and in mathemat-
ics its effect is significantly reduced on the inclusion of the
previous achievement variables. Finally, at a classroom level, the
frequency of homework assignment, while having a reduced ef-
fect, retains its significance.
As a whole the results may be interpreted as follows: once the
adjustment variables are controlled, the homework variables with
Figure 3. Prediction of mathematics performance as a function of quan- most impact on the test are doing homework without help and
tity, time, and how the homework was done. frequency of homework assignment by teachers. Discounting the
adjustment factors and all the homework related variables (includ-
ing doing homework autonomously) the results of the students
those students who need frequent or constant help. In addition, who attend classes in which homework is assigned daily are
dependent students hardly score more than the average despite predicted to exceed the forecasted mean for students in classes
showing significant commitment. Second, autonomous students with less frequently assigned homework by 15% of the SD. If the
are more effective, especially at higher levels of ability. For condition of doing homework autonomously is added to that, this
example, to score 475 points, a dependent student would need, on difference is increased over a quarter (27%) of the SD.
average, 30 min per day more than an autonomous student. How- The previous estimations do not take into account that compared
ever, to score 500 points the dependent student would have to with students who need frequent or constant help, students who do
spend 41 min more than the autonomous student. In other words, homework without parental involvement make more effort and
a dependent student who spends 70 min a day on homework would spend less time on homework (a variable that is negatively asso-
expect results in the 50th percentile whereas an autonomous stu- ciated with success). Figure 4 looks at all of these circumstances.
dent who spends the same amount of time would expect to score It is clear that, once the adjustment variables are controlled, the
in the 70th percentile. In science, the curve follows the same difference in terms of frequency of homework assignment is close
pattern as in mathematics, while noting that students who do to 20 points, whereas the difference between students who need
homework without help are predicted to score 500 points with less help with homework and those who do not is around 30 points.
than 25 min per day, the difference between student profiles is Therefore, the combined effect of variables related to student
greater: to score 500 points, a student who needs help with home- behavior toward homework and teaching practices on school
work spends 50 min more per day than a student who does grades may cause differences in the region of 50% of the SD in the
homework without help. scale of school results. The results in science are not presented in
Model 3 (see Tables 2 and 3) includes the two sociological the chart but the combined effect of these variables is a little more
adjustment variables. SEC is shown to be the variable with the moderate and is about 40% of the SD.
most explanatory power, as the homework variables at an individ-
ual level see little variation. At the classroom level, the amount of
Discussion and Conclusions
homework loses its statistical significance definitively, and only
the frequency of assignment retains its positive, significant effect The results of the study agree with those found in studies from
(95% CI) in both subjects. It should also be noted that, with the Germany (Trautwein, 2007; Trautwein et al., 2002), Great Britain
inclusion of sociological adjustment variables, the total percentage (Farrow et al., 1999), The Netherlands (De Jong et al., 2000),
of variance explained increases significantly in both mathematics Switzerland (Trautwein, Schnyder et al., 2009), and others in
and science, doubling in the latter case. Spain (Gobierno de Aragón, 2010; Gobierno Vasco, 2012). The
Finally, including the variables of previous academic achieve- results of the multilevel models have important practical implica-
ment in Model 4, the results vary substantially. Prior knowledge, tions. First, Model 1 confirms that the relationship between home-
estimated using school grades, is shown to be the most important work and academic results is not linear, something already raised
predictor of achievement in the test. Its effect is greater than the by Cooper and Valentine (2001) and Daw (2012). The data suggest
combined effect of all the other variables in the previous models. that spending 60 min per day doing homework is a reasonable and
Furthermore, on the inclusion of variables of previous achieve- effective time. Furthermore, the results indicate that both the
ment, the effect of SEC is halved, which serves as a note of caution quantity of homework and the frequency of assignment are related
about the relationship between SEC and school grades. In terms of to academic results. In any case, the predominance of frequency
the homework variables, it should be noted that first, the relation- over quantity of homework is consistent with the data from Trau-
ship between results and homework time is, once again, negative, twein (2007) and Trautwein et al. (2002) and is closer to the
beyond statistical doubt in the case of science, and on the limit of standpoint of Farrow et al. (1999), showing that the effect of
statistical significance in the case of mathematics (p ⫽ .07). These homework quantity is small. It is worth stressing that the explan-
results may be interpreted in the following way: once background atory power of this model is moderate (little more than 6% of the
factors are controlled, those students who spend more time are total variance), in line with that noted by De Jong et al. (2000).
1082 FERNÁNDEZ-ALONSO, SUÁREZ-ÁLVAREZ, AND MUÑIZ

530

525

Effect of teaching practice: groups that do

Score on mathematics test daily homework almost 20 more points are
predicted 515

506

500
Effect of student behavior: Students who do 496
not need help get almost 30 points higher than
students who need help (considering that
students who do not need help make more
effort and spend less time) 486

Dependent
Need help
478
Without help
Autonomous

470
Sometimes Often Always

Frequency of homework assignments

Figure 4. Prediction of mathematics performance as a function of frequency of assignment and how the
homework was done.

Model 2 looks at the impact of effort and the amount of help vary significantly, suggesting that behavior toward homework is
needed to do homework. The conclusion is that when it comes to more closely related to school experience than sociological vari-
homework how is more important than how much. Once individual ables. Second, it should be pointed out that the inclusion of
effort and autonomous working is considered, the time spent previous achievement reduces the effect of SEC to half its previous
becomes irrelevant. This data is consistent with the central position value. The data seems to indicate that, within the adjustment
of effort in predicting academic performance (Dettmers et al., variables, a student’s school history has a larger effect on behavior
2010, 2011; Trautwein & Lüdtke, 2007). In addition, autonomy toward homework than social origin, and this data is consistent
when doing homework may be linked to processes of self- with that found in Suárez-Álvarez, Fernández-Alonso, and Muñiz
regulation and these have been found to be positively associated (2014).
with performance (Xu, 2010a). Our multilevel analysis confirms Some conclusions may be drawn in relation to the results from
correlations that have been published by some Spanish educational Model 4. From the Level 2 factors, the principal conclusion is that
authorities, which indicate that students who need frequent or the frequency of homework assignment (but not the amount of
constant help with homework have worse academic results (Go- homework) has statistical significance both in mathematics and in
bierno de Aragón, 2010; Gobierno Vasco, 2012). science. This result is consistent with that described by Farrow et
This reading, together with Models 3 and 4 confirms that to al. (1999) and with the findings of Trautwein (2007) and Trau-
study the relationship between homework and academic perfor- twein et al. (2002), which showed that when frequency is consid-
mance, it is necessary to include adjustment variables as they give ered, the effect of the amount of homework loses statistical sig-
the models the added value of school effectiveness (Schereens, nificance. This data is also compatible with evidence on the
1992). The inclusion of adjustment variables in studies of home- potentially positive effects of teachers’ homework policies (Buijs
work is essential for at least two reasons. First, because of their & Admiraal, 2013; Dettmers et al., 2010; Epstein & van Voorhis,
explanatory power, something that is clear when comparing the 2001; Patall et al., 2010).
variance explained by the latter model compared with the previous In terms of Level 1 variables, three conclusions stand out. First,
ones. Second, because our data agree with that of De Jong et al. when background factors are considered, the time spent on home-
(2000) and Dettmers et al. (2009), which showed that not consid- work becomes negative, which is in line with the results from
ering adjustment variables may undermine the results of the study. Trautwein (2007), Trautwein and Ludtke (2007), and Trautwein,
Regardless, separating the adjustment variables into two different Schnyder et al. (2009). In addition, autonomy, that is, doing
models (on the one hand, sociological factors, and on the other, homework without parental involvement, is the homework
previous achievement) allows two interpretations of the data that behavior-related variable with the best connection to results. This
would have been difficult to see if all of the background variables data compares with conclusions by Xu (2010a), and is consistent
had been together in one model. First, it can be seen that the with theories that emphasize the role of self-control in the learning
inclusion of sociological variables in Model 3 has hardly any process (Zimmerman & Kitsantas, 2005). Finally, it should be
impact on the homework variables (with the exception being the noted that once prior knowledge is discounted, the homework
definite loss of significance of the amount of homework). How- effort variable presents a small effect in mathematics and no effect
ever, when previous achievement is included in Model 4 the results in science. This data contradicts the central position of effort in the
 EFFECT OF HOMEWORK PERFORMANCE IN MATH AND SCIENCE 1083

model of Trautwein et al. (2006), and the conclusions drawn by emotions, and self-concept in four different subjects. Bearing in
Dettmers et al. (2010, 2011). Looking at the correlation matrix (see mind that our target variables are achievement in subjects that are
Table 1) it is worth noting that our effort measure is high and somewhat associated with each other, mathematics and science, it
positively correlated with homework time (r ⫽ .51), whereas is not unreasonable to consider that the results would not have
Trautwein (2007) indicated that these variables are independent. In varied much even if the homework related variables had been
addition, effort shows a better correlation with the teachers’ scores separated by subject. Nevertheless, in future studies it will be
than with the results in the test. This is more pronounced in worth distinguishing between subjects for homework related vari-
science. Although colinearity between homework effort and teach- ables.
ers’ scores has been ruled out, there is covariation between these Furthermore, it should be mentioned that the analysis models
two variables. The teachers very probably consider the students’ presented were focused on the study of students’ behavior when
efforts in their scores that would explain why this effect loses doing homework and in the teachers’ homework assignment prac-
significance when school grades are controlled for. This result tices and it might be worth exploring a study based on a compre-
indicates that school grades may produce some bias when used as
hensive theoretical framework as proposed by Cooper (1989a);
estimators of previous knowledge. According to data from the
Trautwein et al. (2006) and Xu (2008), which would allow the
Gobierno del Principado de Asturias (2010) only a small propor-
inclusion of all the relevant variables to academic performance,
tion of teachers assign school grades based exclusively or mainly
such as motivation and interest in the homework (Dettmers el al.,
on results of standardized achievement tests. The majority also
2010; Xu, 2008, 2011), self-sufficiency (Zimmerman & Kitsantas,
consider other aspects, such as behavior, interest, participation,
and effort. For 40% of teachers these aspects together are given as 2005), and self-regulation (Rosário, Lourenço, Paiva, Rodrigues,
much or more weight than standardized achievement tests when Valle, & Tuero-Herrero, 2012; Xu, 2010a). The inclusion of these
assigning school grades. Furthermore, as indicated by Trautwein, variables will certainly improve the explanatory power of the
Schnyder et al. (2009), school grades are not strictly comparable model.
between classrooms, as teachers assign them with reference to the Our data sends a clear message to teaching professionals: well
class group level and not the total student population. Although used, homework remains a vital tool for comprehensive education
school grades have the advantage of being easy to obtain, objective of adolescents. The results indicate that regular assignment of
measures of previous achievement based on ad hoc tests seem homework and how it is carried out are two key aspects, which
more appropriate (Murillo & Martínez-Garrido, 2013; Trautwein, have a series of implications for educational practice. School
2007). homework has traditionally been considered as an opportunity to
Finally, to appropriately interpret the results, some consider- augment teaching and the time spent on a subject (Paschal et al.,
ation must be made of the study’s limitations. As Trautwein and 1984), but this argument loses force once the limited predictive
Lüdtke (2009) noted, the word “effect” must be understood as capacity of homework time is seen (Trautwein, 2007; Trautwein,
“predictive effect.” In other words, it is possible to say that effort Schnyder et al., 2009). Nevertheless, far from being devalued,
doing homework is connected to performance; however, it is not school homework has demonstrated new, rich values and is cur-
possible to say in which direction the association runs. In fact the rently seen as an opportunity to improve the student’s engagement
data presented are correlations and as such one can only speak of with their own learning (Xu, 2013; Zimmerman & Kitsantas,
an association between variables and not of directionality in the 2005). This is surely our work’s main message to teachers, that
analysis. In addition, the sample was restricted to students around doing homework is still valid nowadays in technically sophisti-
13-years old and as noted by Cooper (1989a, 1989b), the strength cated learning contexts. Our data indicate that it is not necessary to
of the homework-results relationship depends in large part on the assign huge quantities of homework, but it is important that as-
student’s age. signment is systematic and regular, with the aim of instilling work
One of the limitations of the study is that the items referring to habits and promoting autonomous, self-directed learning. Home-
homework are generic, they do not distinguish between the two
work should not exclusively aim for repetition or revision of
subjects being evaluated, mathematics and science. There is evi-
content, as this type of task is associated with less effort and lower
dence that indicates that behavior, motivation and emotions about
results (Trautwein et al., 2002). On the contrary, homework should
homework can vary between subjects (Goetz, Nett, Martiny, Hall,
present a certain level of challenge or difficulty, without being so
Pekrun, Dettmers, & Trautwein, 2012; Trautwein & Lüdtke, 2007;
challenging that it discourages effort (Dettmers et al., 2010). In the
Trautwein et al., 2006). Therefore, for example, motivation has a
pattern associated with gender; boys demonstrate more expectancy last decade, school homework has been connected to principle
and task value in physics-mathematics subjects, and girls demon- variables such as effort, responsibility, study techniques, causal
strate more effort in communicative-linguistic subjects (Trautwein attributions before the task, and self-efficacy (Kitsantas & Zim-
& Lüdtke, 2009). merman, 2009). Stoeger and Ziegler (2008) consider that self-
Nonetheless, when the dependent variables are school results the regulation is something which can be trained when doing home-
differences between subjects are not so clear. Lubbers et al. (2010) work, whereas providing data that indicates that the students who
separated homework time in language and mathematics and found took best advantage of the experimental program were precisely
that the effects of both variables were equal for predicting achieve- those who had the least self-efficacy at the beginning, which seems
ment in the two subjects. Chang, Wall, Tare, Golonka, and Vatz to be evidence that teachers’ homework policies do not only have
(2014) have highlighted the same pattern in relation to homework positive effects on students’ motivation, but also compensatory
time-achievement in a variety of foreign languages, and Goetz et effects. In summary, our data encourages teachers to use home-
al. (2012) reported similar correlations between academic results, work as an effective tool for students’ education and training.
1084 FERNÁNDEZ-ALONSO, SUÁREZ-ÁLVAREZ, AND MUÑIZ

References Fernández-Alonso, R., & Muñiz, J. (2011). Diseño de cuadernillos para la

evaluación de las competencias básicas [Booklet designs for the evalu-
Blazer, C. (2009). Literature review: Homework. Miami, FL: Miami Dade ation of basic skills]. Aula Abierta, 39, 3–34.
County Public Schools. Retrieved from http://drs.dadeschools.net/ Fernández-Alonso, R., Suárez-Álvarez, J., & Muñiz, J. (2012). Imputación
LiteratureReviews/Homework.pdf de datos perdidos en las evaluaciones diagnósticas educativas [Imputa-
Buijs, M., & Admiraal, W. (2013). Homework assignments to enhance tion methods for missing data in educational diagnostic evaluation].
student engagement in secondary education. European Journal of Psy- Psicothema, 24, 167–175.
chology of Education, 28, 767–779. http://dx.doi.org/10.1007/s10212- Gobierno de Aragón. (2010). Evaluación de diagnóstico en Aragón 2010
012-0139-0 [Diagnostic evaluation in Aragon 2010]. Zaragoza: Departamento de
Chang, C. B., Wall, D., Tare, M., Golonka, E., & Vatz, K. (2014). Educación, Cultura y Deporte.
Relations of attitudes toward homework and time spent on homework to Gobierno del Principado de Asturias. (2010). Evaluación de Diagnóstico
course outcomes: The case of foreign language learning. Journal of Asturias 2009 [Diagnostic evaluation in Asturias 2009]. Oviedo: Con-
Educational Psychology, 106, 1049 –1065. http://dx.doi.org/10.1037/ sejería de Educación y Ciencia. Retrieved from http://www.educastur
a0036497 .es/media/institucional/calidad/diagnostico_ast09/ED_2009.pdf
Cooper, H. (1989a). Homework. White Plains, NY: Longman. http://dx.doi Gobierno del Principado de Asturias. (2012). Evaluación de Diagnóstico
.org/10.1037/11578-000 Asturias 2011 [Diagnostic evaluation in Asturias 2011]. Oviedo: Con-
Cooper, H. (1989b). Synthesis of research on homework. Educational sejería de Educación, Cultura y Deporte.
Leadership, 47, 85–91. Gobierno del Principado de Asturias. (2013). Evaluación de Diagnóstico
Cooper, H., Lindsay, J. J., & Nye, B. (2000). Homework in the home: How Asturias 2012 [Diagnostic evaluation in Asturias 2012]. Oviedo: Con-
student, family, and parenting-style differences relate to the homework sejería de Educación, Cultura y Deporte.
process. Contemporary Educational Psychology, 25, 464 – 487. http://dx Gobierno Vasco. (2012). Evaluación diagnóstica. Informe de resultados y
.doi.org/10.1006/ceps.1999.1036 análisis de variables 2° educación secundaria obligatoria [Diagnostic
Cooper, H., Robinson, J. C., & Patall, E. A. (2006). Does homework evaluation. Report results and analysis of variables 2nd compulsory
improve academic achievement? A synthesis of research, 1987–2003. secondary education]. Bilbao: Instituto Vasco de Evaluación e Investi-
Review of Educational Research, 76, 1– 62. http://dx.doi.org/10.3102/ gación Educativa.
00346543076001001 Goetz, T., Nett, U. E., Martiny, S. E., Hall, N. C., Pekrun, R., Dettmers, S.,
Cooper, H., Steenbergen-Hu, S., & Dent, A. L. (2012). Homework. In & Trautwein, U. (2012). Students’ emotions during homework: Struc-
K. R. Harris, S. Graham, & T. Urdan (Eds.), APA educational psychol- tures, self-concept antecedents, and achievement outcomes. Learning
ogy handbook, Volume 3: Application to learning and teaching. (pp. and Individual Differences, 22, 225–234. http://dx.doi.org/10.1016/j
475– 495). Washington, DC: American Psychological Association. .lindif.2011.04.006
Cooper, H., & Valentine, J. C. (2001). Using research to answer practical Goldstein, A. (1960). Does homework help? A review of research. The
questions about homework. Educational Psychologist, 36, 143–153. Elementary School Journal, 60, 212–224. http://dx.doi.org/10.1086/
http://dx.doi.org/10.1207/S15326985EP3603_1 459804
Daw, J. (2012). Parental income and the fruits of labor: Variability in Hoover-Dempsey, K. V., Battiato, A. C., Walker, J. M. T., Reed, R. P.,
homework efficacy in secondary school. Research in Social Stratifica- DeJong, J. M., & Jones, K. P. (2001). Parental involvement in home-
tion and Mobility, 30, 246 –264. http://dx.doi.org/10.1016/j.rssm.2012 work. Educational Psychologist, 36, 195–209. http://dx.doi.org/10.1207/
.01.004 S15326985EP3603_5
De Jong, R., Westerhof, K. J., & Creemers, B. P. M. (2000). Homework IBM Corp. (2013). IBM SPSS missing values 20 [Software manual].
and student math achievement in junior high schools. Educational Re- Retrieved from ftp://service.boulder.ibm.com/software/analytics/spss/
search and Evaluation, 6, 130 –157. http://dx.doi.org/10.1076/1380- documentation/statistics/22.0/en/client/Manuals/IBM_SPSS_Missing_
3611(200006)6:2;1-E;F130 Values.pdf
Dettmers, S., Trautwein, U., Lüdtke, M., Kunter, M., & Baumert, J. (2010). Kitsantas, A., & Zimmerman, B. J. (2009). College students’ homework
Homework works if homework quality is high: Using multilevel mod- and academic achievement: The mediating role of self-regulatory be-
eling to predict the development of achievement in mathematics. Journal liefs. Metacognition and Learning, 4, 97–110. http://dx.doi.org/10.1007/
of Educational Psychology, 102, 467– 482. http://dx.doi.org/10.1037/ s11409-008-9028-y
a0018453 Koole, S. L., & Lakens, D. (2012). Rewarding replications: A sure and
Dettmers, S., Trautwein, U., & Lüdtke, O. (2009). The relationship simple way to improve psychological science. Perspectives on Psy-
between homework time and achievement is not universal: Evidence chological Science, 7, 608 – 614. http://dx.doi.org/10.1177/
from multilevel analyses in 40 countries. School Effectiveness and 1745691612462586
School Improvement, 20, 375– 405. http://dx.doi.org/10.1080/ Lubbers, M. J., Van Der Werf, M. P. C., Kuyper, H., & Hendriks, A. A. J.
09243450902904601 (2010). Does homework behavior mediate the relation between person-
Dettmers, S., Trautwein, U., Lüdtke, O., Goetz, T., Frenzel, A., & Pekrun, ality and academic performance? Learning and Individual Differences,
R. (2011). Students’ emotions during homework in mathematics: Test- 20, 203–208. http://dx.doi.org/10.1016/j.lindif.2010.01.005
ing a theoretical model of antecedents and achievement outcomes. Ministerio de Educación. Cultura y Deporte. (2003). Evaluación de la
Contemporary Educational Psychology, 36, 25–35. http://dx.doi.org/ Educación Secundaria Obligatoria 2000: Informe final [Evaluation of
10.1016/j.cedpsych.2010.10.001 Secondary Education 2000: Final report]. Madrid: Instituto Nacional de
Epstein, J. L., & van Voorhis, F. L. (2001). More than minutes: Teachers’ Evaluación y Calidad del Sistema Educativo.
roles in designing homework. Educational Psychologist, 36, 181–193. Murillo, F. J., & Martínez-Garrido, C. (2013). Homework influence on
http://dx.doi.org/10.1207/S15326985EP3603_4 academic performance. A study of iberoamerican students of primary
Farrow, S., Tymms, P., & Henderson, B. (1999). Homework and attain- education. Revista de Psicodidáctica, 18, 157–171. http://dx.doi.org/
ment in primary schools. British Educational Research Journal, 25, 10.1387/RevPsicodidact.6156
323–341. http://dx.doi.org/10.1080/0141192990250304 Nosek, B. A., & Lakens, D. (2014). Registered reports. A method to
Fernández-Alonso, R. (2005). Assessment of mathematical performance increase the credibility of published results. Social Psychology, 45,
(Unpublished doctoral dissertation). University of Oviedo, Spain. 137–141. http://dx.doi.org/10.1027/1864-9335/a000192
 EFFECT OF HOMEWORK PERFORMANCE IN MATH AND SCIENCE 1085

Núñez, J. C., Vallejo, G., Rosário, P., Tuero, E., & Valle, A. (2014). Trautwein, U., & Lüdtke, O. (2007). Students’ self-reported effort and time
Student, teacher, and school context variables predicting academic on homework in six school subjects: Between-student differences and
achievement in Biology: Analysis from a multilevel perspective. Revista within-student variation. Journal of Educational Psychology, 99, 432–
de Psicodidáctica, 19, 145–172. http://dx.doi.org/10.1387/RevPsicodidact 444. http://dx.doi.org/10.1037/0022-0663.99.2.432
.7127 Trautwein, U., & Lüdtke, O. (2009). Predicting homework motivation and
OECD. (2013). PISA 2012 results: What makes schools successful? Re- homework effort in six school subjects: The role of person and family
sources, policies and practices (Vol. IV). Paris: OECD Publishing. characteristics, classroom factors, and school track. Learning and In-
Paschal, R. A., Weinstein, T., & Walberg, H. J. (1984). The effects of struction, 19, 243–258. http://dx.doi.org/10.1016/j.learninstruc.2008.05
homework on learning: A quantitative synthesis. The Journal of Edu- .001
cational Research, 78, 97–104. http://dx.doi.org/10.1080/00220671 Trautwein, U., Lüdtke, O., Schnyder, I., & Niggli, A. (2006). Predicting
.1984.10885581 homework effort: Support for a domain-specific, multilevel homework
Patall, E. A., Cooper, H., & Robinson, J. C. (2008). Parent involvement in model. Journal of Educational Psychology, 98, 438 – 456. http://dx.doi
homework: A research synthesis. Review of Educational Research, 78, .org/10.1037/0022-0663.98.2.438
1039 –1101. http://dx.doi.org/10.3102/0034654308325185 Trautwein, U., Niggli, A., Schnyder, I., & Lüdtke, O. (2009). Between-
Patall, E. A., Cooper, H., & Wynn, S. R. (2010). The effectiveness and teacher differences in homework assignments and the development of
relative importance of providing choice in the classroom. Journal of students’ homework effort, homework emotions, and achievement. Jour-
Educational Psychology, 102, 896 –915. http://dx.doi.org/10.1037/ nal of Educational Psychology, 101, 176 –189. http://dx.doi.org/
a0019545 10.1037/0022-0663.101.1.176
Peña-Suárez, E., Fernández-Alonso, R., & Muñiz, J. (2009). Estimating Trautwein, U., Schnyder, I., Niggli, A., Neumann, M., & Lüdtke, O.
value added of schools. Aula Abierta, 37, 3–18. (2009). Chameleon effects in homework research: The homework–
Pomerantz, E. M., Moorman, E. A., & Litwack, S. D. (2007). The how, achievement association depends on the measures used and the level of
whom, and why of parents’ involvement in children’s academic lives: analysis chosen. Contemporary Educational Psychology, 34, 77– 88.
More is not always better. Review of Educational Research, 77, 373– http://dx.doi.org/10.1016/j.cedpsych.2008.09.001
410. http://dx.doi.org/10.3102/003465430305567 Tymms, P. B., & Fitz-Gibbon, C. T. (1992). The relationship of homework
to A-level results. Educational Research, 34, 3–10. http://dx.doi.org/
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models:
10.1080/0013188920340101
Applications and data analysis methods (2nd ed.). Newbury Park, CA:
Vallejo Seco, G., Ato García, M., Fernández García, M. P., & Livacic
Sage.
Rojas, P. E. (2013). Multilevel bootstrap analysis with assumptions
Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., & Congdon, R. T. (2004).
violated. Psicothema, 25, 520 –528.
HLM6: Hierarchical linear and nonlinear modeling. Chicago, IL: Sci-
Warton, P. M. (2001). The forgotten voices in homework: Views of
entific Software International.
students. Educational Psychologist, 36, 155–165. http://dx.doi.org/
Rosário, P., Lourenço, A., Paiva, O., Rodrigues, A., Valle, A., & Tuero-
10.1207/S15326985EP3603_2
Herrero, E. (2012). Predicción del rendimiento en matemáticas: Efecto
Wu, M. L., Adams, R. J., Wilson, M. R., & Haldane, S. A. (2007). ACER
de variables personales, socioeducativas y del contexto escolar [Predic-
ConQuest 2.0: Generalised item response modelling software. Camber-
tion of mathematics achievement: Effect of personal, socioeducational
well, Victoria: Australian Council for Educational Research.
and contextual variables]. Psicothema, 24, 289 –295.
Xu, J. (2008). Models of secondary school students’ interest in homework:
Scheerens, J., Witziers, B., & Steen, R. (2013). A meta-analysis of school
A multilevel analysis. American Educational Research Journal, 45,
effectiveness studies. Revista de Educación, 361, 619 – 645. 1180 –1205. http://dx.doi.org/10.3102/0002831208323276
Schereens, J. (1992). Effective schooling, research theory and practice. Xu, J. (2010a). Predicting homework time management at the secondary
London: Cassell. school level: A multilevel analysis. Learning and Individual Differences,
Shell, D. F., Brooks, D. W., Trainin, G., Wilson, K. M., Kauffman, D. F., 20, 34 –39. http://dx.doi.org/10.1016/j.lindif.2009.11.001
& Herr, L. M. (2010). The unified learning model. Dordrecht: Springer, Xu, J. (2010b). Homework purposes reported by secondary school stu-
Netherlands. http://dx.doi.org/10.1007/978-90-481-3215-7 dents: A multilevel analysis. The Journal of Educational Research, 103,
Stoeger, H., & Ziegler, A. (2008). Evaluation of a classroom based training 171–182. http://dx.doi.org/10.1080/00220670903382939
to improve self-regulation in time management tasks during homework Xu, J. (2011). Homework completion at the secondary school level: A
activities with fourth graders. Metacognition and Learning, 3, 207–230. multilevel analysis. The Journal of Educational Research, 104, 171–
http://dx.doi.org/10.1007/s11409-008-9027-z 182. http://dx.doi.org/10.1080/00220671003636752
Suárez-Álvarez, J., Fernández-Alonso, R., & Muñiz, J. (2014). Self- Xu, J. (2013). Why do students have difficulties completing homework?
concept, motivation, expectations and socioeconomic level as predictors The need for homework management. Journal of Education and Train-
of academic performance in mathematics. Learning and Individual Dif- ing Studies, 1, 98 –105.
ferences, 30, 118 –123. http://dx.doi.org/10.1016/j.lindif.2013.10.019 Zimmerman, B. J. (2001). Theories of self-regulated learning and academic
Trautwein, U. (2007). The homework–achievement relation reconsidered: achievement: An overview and analysis. In B. J. Zimmerman & D. H.
Differentiating homework time, homework frequency, and homework Schunk (Eds.), Self-regulated learning and academic achievement: The-
effort. Learning and Instruction, 17, 372–388. http://dx.doi.org/10.1016/ oretical perspectives (2nd ed., pp. 1–38). Mahwah, NJ: Erlbaum, Inc.
j.learninstruc.2007.02.009 Zimmerman, B. J., & Kitsantas, A. (2005). Homework practices and
Trautwein, U., & Köller, O. (2003). The relationship between homework academic achievement: The mediating role of self-efficacy and per-
and achievement: Still much of a mystery. Educational Psychology ceived responsibility beliefs. Contemporary Educational Psychology,
Review, 15, 115–145. http://dx.doi.org/10.1023/A:1023460414243 30, 397– 417. http://dx.doi.org/10.1016/j.cedpsych.2005.05.003
Trautwein, U., Köller, O., Schmitz, B., & Baumert, J. (2002). Do home-
work assignments enhance achievement? A multilevel analysis in 7th Received July 9, 2014
grade mathematics. Contemporary Educational Psychology, 27, 26 –50. Revision received January 22, 2015
http://dx.doi.org/10.1006/ceps.2001.1084 Accepted January 22, 2015 䡲