An Empirical Investigation of Relationships between Junior ...
Transcript of An Empirical Investigation of Relationships between Junior ...
147
1 Background of the study
U.S. studies on parental involvement (PI) indicate that parenting practices vary by families’ socioeconomic status (SES) (e.g., Lareau 2003) and that different degrees of PI differentiate students’ academic achievement (e.g., Hill and Tyson 2009). PI differences based on parents’ SES are considered
one source of the achievement gap. While some scholars (e.g., Honda 2008) address this critical topic in
Japanese society, existing studies using regional and/or retrospective data without a rigorous indicator
of students’ academic abilities fall short of investigating relationships between students’ family SES,
Article
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status,
Parental Involvement, and Academic Performance in Japan
理論と方法 (Sociological Theory and Methods) 2014, Vol.29, No.1:147-165
数理社会学を社会調査の授業に埋め込む?
Abstract
U.S. studies on parental involvement (PI) indicate that parenting practices vary by families’ socioeconomic
status (SES) (e.g., Lareau 2003) and that different degrees of PI differentiate students’ academic achievement
(e.g., Hill and Tyson 2009); PI differences based on parents’ SES are considered one source of the achievement
gap. While some scholars (e.g., Honda 2008) address this critical topic in Japanese society, existing studies
using regional and/or retrospective data without a rigorous indicator of students’ academic abilities fall short
of investigating relationships between students’ family SES, the degree of PI, and their achievement at one of
the most important stages of education: compulsory education. This study is therefore intended to empirically
investigate these relationships by analyzing nationally representative data of Japanese eighth-grade students.
This study’s results indicate that (1) higher SES parents tend to more frequently ask their children what they
study in school; (2) the school-level PI indicator is not equally distributed socioeconomically, and School SES
relates to the degree of PI in school activities; and (3) the degree of PI and school PI in school activities are
associated with students’ mathematics achievement. Contrary to expectations, however, PI mediates small parts
of SES effects, especially at the student level; only some of the relationships between SES, PI, and achievement
are verified empirically.
Keywords and phrases: Parental involvement, concerted cultivation, compulsory education, multilevel, TIMSS
Ryoji MATSUOKA
Waseda University
理論と方法
148
the degree of PI, and their achievement at one of the most important stages of education: compulsory
education. Since students’ academic performance in the ninth grade influences their subsequent
educational and occupational achievement (e.g., Honda 2008), it is important to empirically test whether
PI varies with families’ SES and if its disparity relates to students’ academic performance. This study is
therefore intended to empirically investigate these relationships by analyzing nationally representative
data of Japanese eighth-grade students.
1.1 The relationship between family SES and parental involvement
Arguably, “Unequal Childhoods” by Lareau (2003, 2011) is one of the most influential studies in PI.
In this qualitative study, she coins the term “concerted cultivation,” which is a cultural logic of middle-
class mothers’ parenting practices. These mothers structure children’s daily lives (e.g., by scheduling
extracurricular activities for their time outside school), emphasize the importance of language use by
reasoning and negotiating with their children, and actively interact with social institutions (e.g., school)
to develop their children’s cognitive and social abilities. Meanwhile, working-class mothers follow the
“accomplishment of natural growth” logic, which emphasizes children’s development without rigorous
guides. Disadvantaged mothers do not structure children’s time as much as middle-class mothers do, use
directive and restricted language codes when talking to their children, and tend to avoid interacting with
social institutions. Studies using longitudinal U.S. data empirically support Lareau’s qualitative findings.
In fact, strong relationships between parents' social class and concerted cultivation were found for
elementary school years (e.g., Cheadle and Amato 2011).
Building on Lareau’s study, Bennett, Lutz, and Jayaram (2012) interviewed parents at two urban
middle schools, finding that middle-class and working-class youth participate in different types of
school activities. Specifically, middle-class parents attempt to customize their children’s participation in
extracurricular activities to develop their talents and interests, while working-class parents emphasize
the importance of safety. In addition, working-class children are involved in fewer non-school activities.
These class differences stem from financial and institutional constraints (i.e., less access to institutions
other than school and church) (Bennett, Lutz, and Jayaram 2012). This argument is consistent with Chin
and Phillips (2004), who, based on ethnographic data, report social class differences in the quality and
quantity of fourth-grade children’s involvement in activities during summer stemming from parents' different levels of access to various resources (e.g., money and networks).
Studies conducted in Japanese society also identified relationships between social class and parental
educational involvement/strategies (e.g., Kataoka 2001) and between social class and child rearing (e.g.,
Kanbara and Takata 2000). While these studies focus on each specific relation (i.e., if parenting style
differs by SES group), following Lareau’s study, Honda (2008) assesses the relationships between SES,
parenting practices, and educational outcomes. Specifically, she interviewed 39 mothers whose children
were elementary school students (from fourth–sixth grades), then analyzed survey data on youth (aged
15–29 years) and their mothers (1890 pairs) in Japan. Honda’s (2008) interviews revealed that mothers
with college degrees have higher expectations, actively intervene in home education, and intensively
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
149
use shadow education services (e.g., lessons at juku that should be paid for). In contrast, other mothers
seem to implement home education “naturally.” Honda (2008) contends that these different parenting
practices are aligned to Lareau’s observations of the “concerted cultivation” practiced by middle-class
mothers and “accomplishment of natural growth” by working-class mothers. Then, Honda’s (2008)
quantitative analysis of mothers’ retrospective responses about their child-rearing styles when their child
was attending elementary school showed that mothers in higher SES families tend to engage in various
aspects of parenting practices: both “rigorous” child rearing (i.e., enforcing discipline at home, high
expectations for children’s academic performance, and using shadow education services) and “natural” parenting (i.e., listening to children’s desires, allowing children to play outside, and letting them to have
various sorts of experiences). However, in contrast to their lower SES counterparts, higher SES mothers
practice the former style more often.
Worth mentioning is that Sugihara (2011), based on data from four cities collected from parents of
fifth-grade students, also reports differences in mothers’ parenting (e.g., giving picture books to their
children) based on education qualifications. She found differences in children’s enrollment in enrichment
lessons offered by shadow education institutions, children’s learning time, and education expenses
according to both parents’ education backgrounds.
In the Japanese education context, it is important to point out that Honda (2008) considers the use of
shadow education services as part of the “rigorous” parenting style, which is similar to Lareau’s “concerted
cultivation” in terms of organizing extracurricular activities for children. A number of studies (e.g.,
Holloway et al. 2008; Matsuoka 2013; Yamamoto and Brinton 2010) found that family SES influences
children’s participation in additional lessons at shadow education institutions (e.g., juku and yobiko) in both
elementary and secondary education. Furthermore, studies have also indicated that parental school choice
differs according to family SES (e.g., Oshio 2012).
1.2 The relationship between parental involvement and students’ achievement
Lareau (2003; 2011) contends that parenting differences between middle-class and working-class
mothers contribute to students’ levels of engagement in structured activities, and these differences result
in an achievement gap and different life trajectories, reproducing social class advantages for the middle
class and disadvantages for the working class. Furthermore, a series of quantitative studies show that
PI positively relates to various aspects of educational outcomes,1) while these outcomes are differently
defined and measured: academic achievement (e.g., Hill and Tyson 2009), reduction of problem
behaviors (e.g., Domina 2005), middle-school students’ placement in ability groups (Useem 1992), self-
efficacy and intrinsic motivation toward English and mathematics (Fan and Williams 2010), school
persistence at upper secondary education level (e.g., McNeal 1999), or choice of majors in college (Ma
2009). As for PI in shadow education, a recent study by Park and his colleagues (Park, Byun and Kim
2011), who analyzed longitudinal data collected in South Korea over two years, found that parents’ efforts with regard to choosing and monitoring private tutoring services are related to middle-school
students’ performance in math and English.
理論と方法
150
Likewise, studies conducted in Japan show that PI is associated with educational outcomes. For
example, Uzuki (2004) shows that mothers’ attitudes toward education influence elementary school
students’ learning hours, and mothers’ expectations for their children’s education shape whether their
children desire college education; parents’ daily approaches and high expectations for their children are
meaningful. Moreover, Honda (2008) tests relationships between the two parenting styles and various
aspects of educational outcomes by analyzing mother–child pair data.2) The results of a series of multiple
regression analyses indicate that “rigorous" parenting relates to children’s academic performance in the
ninth grade. Importantly, ninth-grade academic performance is associated with whether children received
four-year college education or higher, which in turn determined if they were employed full time, which
afforded them a higher income (Honda 2008). Honda (2008) concludes that the “rigorous” parenting
style is critical in shaping children’s academic performance at the ninth-grade level, which subsequently
influences their academic background, employment status, and income level. As such, parenting style
during children’s elementary school years directly influences their achievement at the compulsory
education level and indirectly affects their subsequent educational and occupational accomplishments.
Other studies (e.g., Katase and Hirasawa 2008) also indicate that parental strategies (i.e., using shadow
education in the ninth grade) relate to subsequent educational achievement.
Some U.S. studies (e.g., Desimone 1999) show that PI effects also vary with SES group. For
instance, lower SES students obtain less benefits from discussions with their parents, even when they
have the same level of parent-child discussion as higher SES counterparts do. Conversely, higher SES
students are less likely to drop out of high school because of discussions with parents, but this effect
was not observed among low SES students (McNeal 1999). Furthermore, equal PI does not produce
identical education results (e.g., Dumais, Kessinger and Ghosh 2012). Park (2008), based on a large-
scale international examination (PISA2000) of 14 countries, reported a positive association between
PI and high school freshmen’s reading literacy in Japan. In addition, Park (2008) identified a negative
statistically significant interaction between an index of child–parent communication and students’ SES
(p < 0.1). Essentially, when PI is equal, the effect of PI is stronger for low SES parents. However, for
more specific child–parent communication regarding schooling, while it remains negative, the interaction
between PI and students’ SES is not significant statistically (Park 2008). This result seems to indicate no
difference in SES effects on specific PI in students’ educational outcomes in Japan.
2 Rationale of the study
The literature indicates that parenting practices are likely to vary according to family SES, and
differences in PI contribute to the achievement gap. These relationships, however, have not been
rigorously studied in lower secondary education, the last stage of Japanese compulsory education
before students are sorted into different school-based tracks that shape their education trajectories.
In addition, previous research uses regional and/or retrospective data that do not include rigorous
information on students’ academic performance. Furthermore, the studies rely on single-level analyses,
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
151
even though students are nested in schools; student characteristics in one school tend to be similar, likely
producing spurious significant estimates (e.g., Hox 2010). To overcome these issues, this study analyzes
nationally representative large-scale data that include detailed information regarding students’ academic
performance by employing multilevel techniques that consider its nested structure.
It should also be highlighted that this is the first empirical attempt to address school-level PI and its
relation to students' academic performance in Japanese society. This is critically important because
using multilevel techniques and assessing school-level factors could clarify the achievement gap between
schools and the school factors related to the gap in compulsory education. Certainly, previous studies
do address school differences in Japanese compulsory education, but these compare private/national
schools with public counterparts; higher SES students tend to attend private junior high schools (e.g.,
Kataoka 2009), and those attending non-public schools demonstrate higher academic performance (Taki
2012). Since most students attend neighborhood public junior high schools in Japan and lower SES
students who cannot afford shadow education institutions depend on public schooling (Kariya 2004),
it is imperative to investigate whether school factors relate to or explain the achievement gap between
schools, while controlling for types of schools (i.e., public and private/national schools). If PI at the
school level is unequally distributed and this partly explains the achievement gap between schools, it
could be considered as a mechanism of how the gap emerges and persists.
3 Research Questions
This study attempts to unravel the relationships between family SES, PI, and student achievement by
addressing the following four research questions.
(1) Does the degree of PI vary according to family SES?
Following the literature (e.g., Honda 2008) based on regional and/or retrospective data without
rigorous information regarding students’ academic skills, this study hypothesizes that higher SES
students receive a higher degree of PI.
(2) Does the degree of PI at the school level differ by school SES?
Lareau (2003) shows disparities between PI at schools (e.g., middle-class mothers tend to actively
interact with schools); PI in school activities is another aspect of “concerted cultivation.”3) Based on her
study, disparities are expected between schools in terms of the degree of PI at the school level in Japan as
well. A hypothesis for this question is that school SES shapes the degree of school-level PI; higher SES
schools tend to have a higher degree of PI at the school level.
(3) Are the two PI indicators related to students’ academic performance?
Student-level PI is hypothesized to relate to students’ achievement in accordance with the literature (e.g.,
Honda 2008) indicating this association. Likewise, this is the case for the school level as well, since U.S.
studies show that school-based PI moderately influences students’ achievement (Hill and Tyson 2009).
Specifically, students receiving a higher degree of PI at home and attending schools with a higher degree
of PI in school activities tend to demonstrate higher academic achievement.
理論と方法
152
(4) Are the effects of family and school SES mediated by the two PI indicators?
This question clarifies the associations between SES, PI, and achievement. It is hypothesized that the
effects of SES are mediated by PI at each level; students’ achievement is shaped by both student- and
school-level degrees of PI that vary by Student/School SES. If so, this implies that the effect of different
parental practices should be considered as a source of inequality. The frequency of PI in their children’s
school issues differs by SES and differentiates students’ academic performance, which creates, maintains,
and widens the achievement gap.
4 Method
4.1 Data
This study uses the Japanese sample of the Trends in International Mathematics and Science Study
2011 (TIMSS 2011), which was designed and conducted by the International Association for the
Evaluation of Educational Achievement (IEA). TIMSS employed a two-stage sampling process: schools
were randomly chosen for each category (i.e., by area and type of school), and a class was selected from
the sampled schools. The sampled students completed a test and a student questionnaire. This nationally
representative data include 4414 second-year junior high school students in 138 schools (National
Institute for Educational Policy Research 2012).
The test was administrated in March̶the end of the Japanese academic year̶meaning that the
sampled second-year junior high school students were about to become third-year junior high school
students (ninth-grade students). In less than a year, these students will take high-stakes high school
entrance examinations. As high schools are hierarchically ranked and function as academic tracks that
shape students’ academic trajectories (e.g., Kariya 2011), students’ academic performance at the time of
TIMSS-test is important.
4.2 Variables
Student-Level Variables
Student SES. Second-year junior high school students were asked to report (a) the number of books
in their homes and (b) any specific items at home (e.g., own room and Internet connection). First, 11
items were added to develop a “home possessions” index.4) Then, the number of books at home and the
“home possessions” index were combined through a principal component analysis to create a continuous
variable. This independent variable should indicate students’ SES.5)
Student Score/Standardized Math Score. Five plausible values were used to represent students’ performance in mathematics, while standardized scores were also created and utilized in the first
set of the analyses. Of the two subjects tested in TIMSS, mathematics was selected over science, as
mathematics is the most popular subject taken by junior high school students in the shadow education
industry (MEXT 2008). It is likely that PI and parental encouragement shape students’ achievements in
this most studied subject.
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
153
Female (Gender). Female students were indicated as 1. Male counterparts were coded as 0.
PI. Among many PI aspects, this study focuses on parent–child communication/discussion on academic
issues, which U.S. literature indicates as being effective (e.g., Hill and Tyson 2009; Sui-Chu and Willms
1996). More specifically, this study tests if students’ (family) SES relates to the frequency of parent–
child communication about schoolwork and investigates if the communication frequency is related to
students’ achievement. Therefore, “PI” was created from students’ responses to Q11 of the IEA Student
Questionnaire (2011b: 9): “11. How often do the following things happen at home? a) My parents ask me
what I am learning in school.” Sampled students were asked to select one of the following four options:
“Every day or almost every day,” “Once or twice a week,” “Once or twice a month,” and “Never or almost
never.” These responses were recoded from 3 (Every day or almost every day) to 0 (Never or almost
never). This variable represents the degree of parent–child communication on academic matters. Through
communication, parents can notice if their child begins to disengage in school, and “the importance of
schooling and education is conveyed to the child” (McNeal 1999:124). This variable also likely works
as a proxy for parents’ daily approaches and the high expectations of their children, which Uzuki (2004)
contends as meaningful.
School-Level Variables
School SES. Student SES was averaged at each school to create this variable to determine a school
socioeconomic compositional effect (e.g., Raudenbush and Bryk 2002) on students’ performance.
School PI in School Activities (School PI in School Activities). This variable was based on school
principals’ or department heads’ responses to the IEA School Questionnaire (2011a: 6): “11. How
would you characterize each of the following within your school?” “f) Parental involvement in school
activities.” Responses were coded as “Very high” (4), “high” (3), “medium” (2), “low” (1), and “very low” (0).
Private/National. Using TIMSS classifications of schools, private and national schools were coded as
1, and public schools as 0. This needs to be controlled because private/national schooling is a possible
mechanism for the SES-based achievement gap in Japan (Taki 2012).
Urban. This variable was based on responses to the IEA School Questionnaire (2011a: 2): “5. B. Which
best describes the immediate area in which your school is located?” (2) “Urban–Densely populated” was
coded as 1 and others (e.g., medium-size city and small town) as 0.
School Score. Each plausible value was used at each school to indicate school performance.
4.3 Analysis
To test the first research question on the relationship between students’ SES and the degree of PI, a
multilevel ordinal regression analysis was conducted. An ordinal regression analysis was then employed
to assess whether School SES is associated with School PI in School Activities. Finally, multilevel
regression analyses were carried out to empirically test whether student- and school-level PI indicators
relate to students’ performance, and if the effect of SES are mediated by PI at each level. For the first
and last part of the analyses, multilevel modeling techniques were applied as the data with student and
理論と方法
154
school levels being better suited to multilevel investigations, which may more accurately capture the
school effects embedded in school settings on students’ performance.6) For each multilevel analysis,
a model including both student- and school-level predictors is presented in this study.7) In addition,
to observe if the two PI indicators mediate the effects of Student and School SES, two models were
created (with/without the PI indicators) for the last part of the analyses. The following is an example of
modeling showing the last multilevel analysis: Level-1 (Student level) Model: Student Math Score ij = β0j
+ β1j(Student SES ij) + β2j(Female ij) + β3j(PI ij) + rij Level-2 (School level) Model: β0j = γ00 + γ01(School
SES j) + γ02(Private j) + γ03(Urban j) + γ04(School PI in School Activities j) + u0j, β1j = γ10 + u1j, β2j =
γ20, β3j = γ30
In the first set of the multilevel analyses, random intercept models are specified as all slopes do not
vary, indicating that the effects of Student SES, Female (gender), and Math Score are the same across
schools. Meanwhile, all models of the last analysis are random intercept and slope models. The variation
in the levels of the intercepts was estimated to reflect between-school differences in students’ math
performance, while the slope of Student SES varies between schools, which indicates that the effect of
Student SES differs between schools.8) The other within slopes were fixed, given the results that random
slopes of the other variables became insignificant, suggesting no between-school differences in the
effects of Female and PI.
5 Results
5.1 Descriptive Statistics
Descriptive statistics of three student-level and two school-level continuous variables are shown
in Table 1.9) Table 2 is a frequency table of the two student-level and three school-level categorical
variables.10) These results clarify that the PI is unequally distributed at each level.
Table 1. Descriptive Statistics for Continuous Variables
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
155
5.2 Correlation matrices
The correlation matrix of the variables is shown in Table 3.11) Student SES significantly correlates to
PI (.220), which means that the parents of higher SES students often ask what they learn in schools. At
the school level, School SES correlates with School PI in School Activities (.413), indicating that school
principals in higher SES schools perceive a higher degree of PI in school activities.
Table 2. Frequency Distribution of Categorical Variables
Table 3. Correlations between Variables at Each Level
理論と方法
156
5.3 Assessing the relationship between Student SES and PI
Table 4 shows the results of a multilevel ordinal regression analysis.12) Due to space limitations, only a
final model is presented (not only for this analysis but for all the analyses). The ordinal outcome ranges
from 0 to 3; the four categories are ordered from the lowest (never or almost never) to the highest (every
day or almost every day). Table 4 provides an estimation of the log odds that a student reports receiving
PI (the highest category versus combined lower categories). Intercept log odds are negative because it is
more likely that most students report lower categories of PI.
Table 4. Factors Differentiating the Degree of PI at Home
The results show that, while there are differences in the degree of PI between schools, most
differences are at the student level; only School Score appears to be significant (p < 0.1). In fact, intra-
class correlation (ICC) of Model 1 (without any predictors) is 0.032 (3.2%) and that of Model 3 is
0.022 (2.2%); only 3.2% of the total student differences is at the school level, while the remainder is
at the student level where Student SES, Standardized Math Score, and Female are significant positive
predictors of the PI degree. As Table 4 shows, Student SES influences the PI degree that students receive
at home. More specifically, when holding other variables constant, 1-standard deviation (1-SD) increase
in Student SES increases the odds of being at the highest category of receiving PI versus the combined
lower categories by a factor of 1.470 (or 47.0%), compared to their average SES peers.
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
157
5.4 Testing School-level relationships between SES and PI
The results of the school-level ordinal regression analysis in Table 5 indicate that all independent
variables including School SES are significant predictors of School PI in School Activities. These results
seem consistent with those in Table 4: higher performing schools likely have a higher PI degree in school
activities, while private/national schools and those in urban areas are less likely to have a higher PI
degree in school activities.
Table 5. Relationship between SES and PI at the School Level
5.5 Investigating the Relationships between SES, PI, and Score
For this analysis, two sets of results are presented in Table 6. The left side of the table shows model’s
results without two PI indicators, while the right side presents findings of the final model with two PIs to
investigate if these PI indicators mediate effects of Student SES and School SES.13)
For relationships between PIs and achievement, all student-level predictors, namely Student SES,
Female, and PI, significantly relate to students’ achievement in mathematics. For Student SES, when
other variables are held constant, students with an SES 1-SD above the mean score around 20.9 points
higher than the grand mean. Female students are likely to obtain a slightly lower score (approximately
-5.6 points) than male counterparts. Results for PI indicate that students whose parents ask them about
schoolwork tend to demonstrate higher math ability. As this variable ranges from 0 to 3, students whose
parents ask them what they learn in school every day or almost every day score an additional 12.57
points (4.190 × 3) compared to those whose parents “never or almost never” do so, while the other
variables including Student SES are controlled. Although its effect is relatively weak, this result shows
the significant relationship between the degree of PI and students’ math achievement.
理論と方法
158
In addition, three school-level variables appear to be significant predictors of students’ achievement:
School PI in School Activities, Private/National, and Urban. The results indicate that students who
attend schools with a higher PI degree in school activities demonstrate higher academic ability. As the
variable ranges from 0 (very low) to 4 (very high), students attending schools with “very high” PI in
school activities have 34.08 points (8.520 ×4), compared to those who go to schools with “very low” PI
regarding school activities. Attending private and national schools also significantly relates to students’ math achievement; students in private/national schools score 46.5 points higher than their counterparts
in public schools when other variables are held constant. ICC is 0.128 (12.8%) for Model 1 (without
any predictors) and 0.042 (4.2%) for the final model. These mean that 12.8% of the variance in math
performance is at the school level, and it decreased to 4.2% when (primarily) school-level explanatory
variables were added to the model.15)
Comparing the two models indicates that the effects of Student SES and School SES weaken with PI
indicators, implying that PIs mediate some SES effects. Specifically, because of School PI in school
activities, the effect of School SES decreases (approximately 61.12% = (6.009- 2.336)/6.009), although
the effect of School SES itself is small. With regard to Student level, only 3.3% of Student SES effect
(=(21.564‒20.847)/ 21.564) is mediated by the degree of PI.
Table 6. Association between PI and Score/Mediated Effect of SES
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
159
6 Discussion and implications
6.1 Discussion
This study’s results support the first three hypotheses: (1) higher SES parents tend to more frequently
ask what their children study in school, (2) the school-level PI indicator is not equally distributed
according to socioeconomic lines, and (3) the degrees of PI and school PI in school activities are
associated with students’ mathematics achievement. Contrary to the fourth hypothesis, however, PIs mediate
small parts of SES effects, especially at the student level. Only some relationships between SES, PI, and
achievement are verified empirically.
Although there is not much support for the last hypothesis, the study still provides evidence of each
relation between SES and PI and between PI and achievement based on nationally representative data.
Students with lower family SES experience not only low SES but also less PI about their schoolwork.
This seems to be a mechanism for inequality; higher SES students tend to have greater academic ability
facilitated through early family socialization and then continuously receive attention from their parents
with regard to schoolwork when they have less than one year before taking high-stakes examinations for
upper secondary education. In addition, School SES differentiates School PI in School Activities. This
result is consistent with Lareau’s observation of “concerted cultivation.” By being involved in school
activities, parents get to know teachers and other parents, helping them monitor their children’s schooling
(McNeal 1999). Furthermore, by having opportunities to communicate with other parents, mothers might
obtain information about shadow education institutions to boost their children’s likelihood of gaining
admission to selective high schools, as Park, Byun, and Kim (2011) found that selecting and monitoring
shadow education services may improve academic performance in South Korea.
While PIs mediate only small parts of SES effects, the study empirically demonstrates the relationship
between PIs and achievement; students receiving PI demonstrate higher achievement. It should be
noted that, since interaction terms (e.g., between PI and student/school SES) were insignificant, no
difference is evident between SES groups in terms of PI effects on students’ math achievement.16) This
finding parallels results obtained by Park (2008), who assessed specific child–parent communication
regarding schooling of Japanese high school freshmen. Considering that Park’s results are similar to the
PI definition in this study, the results of his study and those obtained in this study indicate that the PI
effect (i.e., the specific type of child–parent communication) on students’ achievement does not vary with
family SES for both lower and upper secondary education in Japan.
6.2 Policy Implications
Parents should be informed that the PI degree significantly relates to students’ achievement. However,
it is likely that higher SES parents will respond to any recommendation as they tend to value academic
achievement. More specifically, policy emphasis on home education and social pressures could widen
disparities, as advantaged parents are already active in their children’s education and can employ more
resources to maximize their children’s potential (Honda 2008). Moreover, this study finds no interaction
理論と方法
160
effect between the PI degree and student SES; the effect of PI does not vary with family SES in Japan.
When PI uniformly benefits students’ achievement regardless of family SES, encouraging parents to
get involved in their children’s education may not reduce the achievement gap (Park 2008) because
compared to lower SES counterparts, higher SES parents are more likely to (afford to) respond to the
importance of home education advocated by policies and the media.
The importance of academic performance at the time of TIMSS administration must be emphasized;
tested students were about to enter ninth grade. Since academic achievement in the ninth grade is
associated with long-term educational and occupational achievement (Honda 2008), policies should
focus on narrowing the achievement gap in compulsory education. Honda (2008) proposes that public
school education be improved (e.g., by lowering student–teacher ratios) and learning opportunities
outside schools be increased (e.g., by allowing students to have various experiences and distributing
vouchers that can be used for enrichment courses outside schools).
Essentially, simply promoting more PI may expand differences in parenting practices along with
socioeconomic lines, reinforcing the unequal distribution of PI. To rectify the trend and avoid victim
blaming, students from low SES families and their parents need support; it is unrealistic to expect the
new involvement of these parents in their children’s education without any support from schools or
other social institutions. Consequently, as the literature (e.g., Chin and Phillips 2004; Bennett, Lutz and
Jayaram 2012; Honda 2008; Lareau 2003) indicates, their relatively lower degree of involvement in
children’s education stems from fewer resources including economic means and networks.
6.3 Research Implications
If variables had represented various aspects of students’ SES and PI indicating “concerted cultivation” in more detail, it would have been more rigorous to test the relationship between students’ family SES
and PI to assess if PI mediates SES effects on students’ achievement. There may be other unobserved
factors that change the relationship between the degree of PI and student achievement. For example,
if the data had indicated student enrollment in shadow education institutions (juku) or private tutoring
services at the time of TIMSS administration, the PI degree (i.e., parent–child communication on
academic matters and PI in school activities) could have accounted for less variance in students’ academic
performance, and SES effects on math scores could be mediated by shadow education participation. This
is a limitation also faced by Honda (2008). In her study, shadow education and other types of lessons
during a child’s elementary school years are significant predictors of students’ academic performance at
the ninth-grade level, while “rigorous rules at home” is weakly related to achievement and the other PI (i.e.,
passionately leading a child to have better academic performance) is insignificant. These results could
indicate that using shadow education services substantially mediates SES effects on students’ academic
performance, and specific parenting styles may not matter much, at least, in Japanese society. If this is
the case, it would be necessary to test if the cultural logic of “accomplishment of natural growth” (Lareau
2003) relates to the use of shadow education services; following this logic, lower SES parents may avoid
structuring their child’s time using shadow education services. It is also important to investigate the
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
161
effect of cultural logic when parents decide to apply or not apply for private elementary, junior high, and
high schools, since attending top private high schools significantly increases the likelihood of gaining
admission to competitive universities (Kariya 2011). Finally, further studies, ideally using longitudinal
data, should be conducted to assess relationships between SES, PI, and different aspects of educational
outcomes (e.g., students’ behavior problems), as McNeal (1999) found, with U.S. students, that parent–
child communication might reduce the likelihood of behavioral issues.
*The author gratefully acknowledges the valuable suggestions and comments from anonymous
reviewers.
*This study was supported by JSPS KAKENHI Grant Number 24830009.
Notes
1) While cited studies show significant relationships between some types of PI and educational outcomes, the PI
effects are not conclusive. Domina (2005) argues that findings of studies on effects of PI have been inconsistent
at middle and high school levels, while his longitudinal data analysis shows that some aspects of PI (i.e., parents
volunteering at school and helping/checking their child’s homework) prevent children’s behavioral problems at the
elementary school level. PI effects seem to significantly vary on the basis of how PI and educational outcomes are
measured. In relation to this study, which assesses junior high school students, Hill and Tyson (2009) find positive
associations between PI and achievement through meta-analysis of 50 studies.
2) It should be noted that a small number of pairs were used in some of her analyses.
3) As TIMSS 2011 did not ask students about the degree of PI in school activities, this aspect of PI was assessed
with variables obtained from the school questionnaire completed by school principals.
4) The 11 items are from BSBG05 A to K, including 6 country-specific items.
5) In total, 25.8% of students chose “I don’t know” with regard to their father’s highest education level, while
33.3% did not know their mother’s education background. In addition, there is no reasonable way of conducting
multiple imputation analyses because of data limitations (student questionnaire items). Thus, to avoid a large
number of missing values, parents’ education backgrounds were not included in Student SES. See note 15 for
more detail in this regard.
6) It should be noted that TIMSS 2011 is a cross-sectional examination. Had it been designed as a longitudinal
study, alternative explanations for the results of the study could have been excluded more rigorously. All
multilevel models were created on the basis of the major literature in the field (e.g, Hox 2010; Raudenbush and
Bryk 2002; Raudenbush, Bryk, Cheong, Congdon and Mathilda 2011; Heck and Thomas 2009).
7) While interaction terms (e.g., PI and Student/School SES) were created and tested, they were all insignificant.
The models in this study include only primary effects. These results show no differential effect of PI by SES
groups on eighth-grade students’ achievement in Japan.
8) No school variable was found to explain the variation in the SES-achievement slope at the individual level.
Thus, why the relationship between Student SES and Student Math Performance varies between schools is
unknown. See Hox (2010) for detailed explanations of fixed and random effects.
9) All continuous variables, except Math Score, were standardized to facilitate the interpretation of results. IEA
IDB Analyzer produced the descriptive statistics for these variables with Total Student Weight (TOTWGT) for
the student level and School Level Weight (SCHWGT) for the school level. For Math Score, five “Plausible
Value Mathematics” (BSMMAT01 to 05) were employed and then averaged. See Olson, Martin and Mullis
理論と方法
162
(2008) for a detailed discussion on plausible value. N in the table was not weighted to show the number of
original cases. Only cases used for all analyses are included; 110 students (2.49%) and 1 school (0.72%) are
missing because of lack of data.
10) These variables were not weighted to show the original distributions of variables. Valid percentages are shown
in this table. The number of private/national schools is only 11 in the dataset; this is an accurate reflection
of the population. MEXT (2012) shows that there were 73 national and 763 private schools of 10751 junior
high schools in Japan: 0.7% national and 7.1% private schools in the academic year 2011–12 (the year of
TIMSS2011 administration).
11) IEA IDB Analyzer was used to analyze five “Plausible Value Mathematics” with Total Student Weight. In
the correlation matrices, non-continuous variables were also included to indicate a sense of the strength of
relationships between the variables used in the analyses.
12) HLM7.01 was utilized for all multilevel analyses. Five “Plausible Value Mathematics” were separately included
in each model, while Student Weight Adjustment (WGTADJ3) and School Level Weight (SCHWGT) were
applied. Then, five sets of results were averaged. In an ordinal model, thresholds (or cutpoints) are used to
determine which response category is observed, with one less threshold than the number of ordered categories
(C ‒ 1) in the outcome required to specify estimated probabilities (Hox 2010). It is common to treat the first
threshold (δ0) as the intercept, and the second and third thresholds can be designated as δ1 and δ2, respectively.
The model is specified such that the intercept can vary as a random parameter at the school level of the model.
All other thresholds are fixed to 0 between groups for model identification purposes. While the thresholds are
useful in determining predicted probabilities, they have no substantive meaning because they are unaffected by
the levels of X predictors for individual cases in this study.
13) The two ordinal PI indicators are treated as continuous in the last analyses, since the outcomes (five PVs)
increase over the ordinal categories. More specifically, the means of PI are 552.28 (never or almost never),
570.12 (once or twice a month), 576.68 (once or twice a month), and 589.18 (every day or almost every
day). For School PI in School Activities, a standardized school score (mean = 0, SD = 1) was used to obtain
the means: ‒ 0.48 (very low), ‒ 0.22 (low), ‒ 0.10 (medium), 0.11 (high), and 0.24 (very high). As the means
increase over the categories of the ordinal variables, they were included in the models as if they were
continuous. See Agresti (2013) for a detailed discussion on defining ordinal predictors as intervals.
14) A deviance of Model 1 is 50483.207 and that of Model 2 is 49290.587. The final model’s deviance is
49211.057, as shown in Table 6 (a smaller deviance means a better model fit). As for the multilevel ordinal
regression model presented in Table 4, HLM7.01 provides no model-fit indicator including a deviance. When
the dependent variable was treated as continuous, a deviance of each model is 12050.229 (Model 1), 11805.874
(Model 2), and 11801.129 (Model 3), while the results did not change: a better model fit as variables were
added.
15) Ad-hoc multilevel analyses were conducted to verify this study’s results. First, parents’ highest education level
(BSDGEDUP with missing value of 22.2%), number of books at home, and “home possessions” were combined
with a principal component factor analysis. This newly created Student SES index and its averaged school-level
variable were included in the final set of the multilevel models instead of the originally coded SES variables.
PI at both student and school levels became insignificant with the analysis (i.e., t-value of PI at the student
level is 1.519). A total of 22.2% of students omitted from the analysis did not know either parent’s education
background. Many of these were from low SES families, did not demonstrate high mathematical ability, and
indicated a lower degree of PI. In fact, means for Math Score, (originally coded) Student SES, and PI of these
students are lower than those for the entire sample. Moreover, in preliminary analyses, a missing variable flag
was created to indicate that students who reported parents’ education background scored significantly higher in
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
163
math performance than those with missing values (p < .001). This implies that the missing data may be missing
not at random (MNAR). To further examine this possibility, preliminary analyses of a number of different
hypothetical distributions of missing information on parental education were also examined (i.e., from 100%
low education to 0% low education), and in each case, except the most extreme (i.e., 100% low education), the
results did not change the effect of PI on the outcome. These preliminary results indicate that in the extreme
case where all missing cases were from low education families, the analysis likely produces a downward bias,
resulting in the non-significance of PI variables. In any other case, however, results showing the significance of
PI could be considered valid.
16) Results of tested interaction terms are not included in the paper because of space limitations.
References
Agresti, Alan. 2013. Categorical Data Analysis. Hoboken, N.J.: Wiley-Interscience.
Bennett, Pamela R., Amy C. Lutz and Lakshmi Jayaram. 2012. "Beyond the Schoolyard: The Role of Parenting
Logics, Financial Resources, and Social Institutions in the Social Class Gap in Structured Activity
Participation." Sociology of Education 85(2): 131-57.
Cheadle, Jacob E. and Paul R. Amato. 2011. "A Quantitative Assessment of Lareau's Qualitative Conclusions About
Class, Race, and Parenting." Journal of Family Issues 32(5): 679-706.
Chin, Tiffani and Meredith Phillips. 2004. "Social Reproduction and Child-Rearing Practices: Social Class, Children's
Agency, and the Summer Activity Gap." Sociology of Education 77(3): 185-210.
Desimone, Laura. 1999. "Linking Parent Involvement with Student Achievement: Do Race and Income Matter?"
Journal of Educational Research 93(1): 11-30.
Domina, Thurston. 2005. "Leveling the Home Advantage: Assessing the Effectiveness of Parental Involvement in
Elementary School." Sociology of Education 78(3): 233-49.
Dumais, Susan A., Richard J. Kessinger and Bonny Ghosh. 2012. "Concerted Cultivation and Teachers' Evaluations
of Students: Exploring the Intersection of Race and Parents' Educational Attainment." Sociological
Perspectives 55(1): 17-42.
Fan, Weihua and Cathy M. Williams. 2010. "The Effects of Parental Involvement on Student's Academic Self-
Efficacy, Engagement and Intrinsic Motivation." Educational Psychology 30(1): 53-74.
Heck, Ronald H. and Scott L. Thomas. 2008. An Introduction to Multilevel Modeling Techniques. New York:
Routledge.
Hill, Nancy E. and Diana F. Tyson. 2009. "Parental Involvement in Middle Achool: A Meta-Analytic Assessment of
the Strategies That Promote Achievement." Developmental Psychology 45(3): 740-63.
Holloway, Susan D., Yoko Yamamoto, Sawako Suzuki and Jessica Mindnich. 2008. "Determinants of Parental
Involvement in Early Schooling: Evidence from Japan." Early Childhood Research and Practice 10(1).
Honda, Yuki. 2008. Katei Kyōiku No Airo [Obstacles to Home Education]. Tokyo: Keiso shobo.
Hox, J. J. 2010. Multilevel Analysis: Techniques and Applications (2nd Edition). New York, NY: Routledge.
IEA. 2011a. "Timss2011 School Questionnaire<Grade 8>." (http://timssandpirls.bc.edu/timss2011/downloads/
T11_SchQ_8.pdf).
IEA. 2011b. "Timss2011 Student Questionnaire <Grade 8>." (http://timssandpirls.bc.edu/timss2011/downloads/
T11_StuQ_8.pdf).
Kanbara, Fumiko and Yoko Takata. 2000. Kyoikuki No Kosodate to Oyako Kankei: Oya to Ko No Kakawari O
Aratana Kanten Kara Jissho Suru [Child Rearing and Parent-Child Relationship During Years of Schooling:
Demonstrating Parent-Child Relationship from New Perspectives]. Kyoto: Mineruba shobo.
Kariya, Takehiko. 2004. "Gakuryoku No Kaisosa Ha Kakudai Shitaka [Is Disparity of Academic Achievement
理論と方法
164
Based on Social Class Widen] " in Gakuryoku No Shakaigaku: Chosa Ga Shimesu Gakuryoku No Henka to
Gakushu No Kadai [Sociology of Academic Achievement: Changes of Academic Achievement and Problems
of Learning Shown by Surveys] edited by T. Kariya and K. Shimizu. Tokyo: Iwanami Shoten.
Kariya, Takehiko. 2011. "Japanese Solutions to the Equity and Efficiency Dilemma? Secondary Schools, Inequity
and the Arrival of ‘Universal’ Higher Education." Oxford Review of Education 37(2): 241-66.
Kataoka, Emi. 2001. "Kyoikutassei Katei Ni Okeru Kazoku No Kyoiku Senryaku: Bunka Shihon Koka to Gakkogai
Kyoiku Toshi Koka No Gienda Sa O Chushin Ni [Family Strategy in Educational Attainment Process
in Japan: Effects of Cultural Capital and Investment in Extra-School Education]." Japanese Journal of
Educational Research 68(3): 259-73.
Kataoka, Emi. 2009. "Kakusa Shakai to Shochugaku Juken--Juken O Tsujita Shakaitekiheisa, Risukukaihi, Ishitsuna
Tasha Heno Kanyosei [Class Closeness and Parental School Choice--Sociological Analysis Concerning
'O-Juken' and Junior High School Examinations in Japan." Japanese Journal of Family Sociology 21: 30-44.
Katase, Kazuo and Kazushi Hirasawa. 2008. "Syoshika to Kyoikutoshi Kyoikutassei: Tokusyu, Jinkohendo to
Kyoikukaikaku [the Declining Birthrate and Educational Investment and Achievement: Special Issue,
Demographic Change and Educational Reform]." The journal of educational sociology 82: 43-59.
Lareau, Annette. 2000. Home Advantage: Social Class and Parental Intervention in Elementary Education. Lanham,
Md.: Rowman & Littlefield Publishers.
Lareau, Annette. 2003. Unequal Childhoods: Class, Race, and Family Life. University of California Press.
Lareau, Annette. 2011. Unequal Childhoods: Class, Race, and Family Life, Second Edition, with an Update a Decade
Later: University of California Press.
Ma, Yingyi. 2009. "Family Socioeconomic Status, Parental Involvement, and College Major Choices-Gender, Race/
Ethnic, and Nativity Patterns." Sociological Perspectives 52(2): 211-34.
Matsuoka, R. (2013). "School socioeconomic compositional effect on shadow education participation: Evidence from
Japan." British Journal of Sociology of Education. doi: 10.1080/01425692.2013.820125
McNeal, Ralph B. 1999. "Parental Involvement as Social Capital: Differential Effectiveness on Science Achievement,
Truancy, and Dropping Out." Social Forces 78(1): 117-44.
MEXT. 2008, "Kodomo No Gakkogai Deno Gakushukatsudo Ni Kansuru Jittai Chosa Hokoku [Actual Condition
Survey on Learning Activities of Children Outside of Schools]", Tokyo: Ministry of Education, Culture,
Sports, Science and Technology (MEXT). Retrieved August 1st, 2013 (http://www.mext.go.jp/b_menu/
houdou/20/08/08080710.htm).
MEXT. 2012. Gakko Kihon Chosa: Heisei 23 Nendo (Kakuteichi) Kekka No Gaiyo [2011 School Basic Survey: A
Brief Summary of Results of Academic Year 2011-12 (Definite Value)]Congress, (http://www.mext.go.jp/
b_menu/toukei/chousa01/kihon/kekka/k_detail/1315581.htm).
National Institute for Educational Policy Research. 2012. Kokusai Sugaku Rika Kyoiku Doko Chosa No
2011 Nen Chosa: Kokusai Chosa Kekka Hokoku [Timss2011: Report on Results of the International
Examination]Congress. Retrieved February 11th, 2013 (http://www.nier.go.jp/timss/2011/T11_gaiyou.pdf).
Olson, J.F., M.O. Martin and I.V.S. Mullis, eds. 2008. Timss 2007 Technical Report. Chestnut Hill, MA: TIMSS &
PIRLS International Study Center, Boston College.
Oshio, Takashi. 2012. Koritsu to Kohei O to[Inquiring About Efficiency and Equity]. Tokyo: Nihonhyoronsha.
Park, Hyunjoon. 2008. "The Varied Educational Effects of Parent–Child Communication: A Comparative Study of
Fourteen Countries." Comparative Education Review 52(2): 219-43.
Park, Hyunjoon, Soo-yong Byun and Kyung-keun Kim. 2011. "Parental Involvement and Students’ Cognitive
Outcomes in Korea: Focusing on Private Tutoring." Socioology of Education 84(1): 3-22.
Raudenbush, Stephen W. and Anthony S. Bryk. 2002. Hierarchical Linear Models: Applications and Data Analysis
An Empirical Investigation of Relationships between Junior High School Students’ Family Socioeconomic Status, Parental Involvement, and Academic Performance in Japan
165
Methods. Newbury Park, CA: Sage.
Raudenbush, Stephen W., Anthony S. Bryk, Yuk F. Cheong, Richard Congdon and Mathilda Du Toit. 2011. Hlm 7:
Hierarchical Linear and Nonlinear Modeling. Scientific Software International.
Sugihara, Nahoko. 2011. "Oya No Kyoiku Kodo to Chiikisa[Parents' Educational Behavior and Regional
Disparities]." in Kakusa Shakai O Ikiru Kazoku: Kyoiku Ishiki to Chiiki Jienda [Families Living in Unequal
Society: Attitudes toward Education, Regions and Gender], edited by Y. Ishikawa, N. Sugihara, K. Kita and Y.
Nakanishi. Tokyo: Yushindo kobunsha.
Sui-Chu, Esther Ho and Douglas Willms. 1996. "Effects of Parental Involvement on Eighth-Grade Achievement."
Sociology of Education 69(2): 126-41.
Taki, Hirofumi. 2012. "Gimukyoikudankai Niokeru Gakuryoku to Shakaikeizaitekichii No Kanrenkozo: Timss Data
O Mochiite[Associated Structure between Academic Performance and Socioeconomic Status in Compulsory
Education Using Timss Data]." Paper presented at the 85th Conference of the Japan Sociological Society,
Sapporo.
Useem, Elizabeth L. 1992. "Middle Schools and Math Groups: Parents' Involvement in Children's Placement."
Sociology of Education 65(4): 263-79.
Uzuki, Yuka. 2004. "Shochugakusei No Doryoku to Mokuhyo: Shakaiteki Sembatsuizen No Oya No Eikyoryoku
[Elementary and Junior High School Students' Effort and Aim: Parental Influence in the Pre-Selection
Stage]." in Josei No Shugyo to Oyako Kankei: Hahaoyatachi No Kaiso Senryaku[Working Women and Child-
Parent Relationship: Mothers' Strategies], edited by Y. Honda. Tokyo: Keiso Shobo.
Yamamoto, Yoko and Mary C. Brinton. 2010. "Cultural Capital in East Asian Educational Systems: The Case of
Japan." Sociology of Education 83(1): 67-83.
(Received March 8, 2013 / Accepted November 7, 2013)