1. Introduction
“The differences among families seem to cluster together in meaningful patterns”
—Lareau (2011, p. 3)
Parent personality, educational beliefs, parenting styles, and parent–child interactions affect parental involvement, which can affect children’s academic learning [
1]. However, most past studies ignored differences across profiles of parents within a population (assuming homogeneity) and only considered linear or quadratic relations between parental involvement behaviors and their children’s academic learning. These assumptions can distort the relations between parental involvement and children’s academic learning.
Hence, this study aims to extend the results of past studies on parental involvement in children’s learning at home in two directions: (a) specifying profiles of parental involvement in children’s learning at home and (b) determining these parental involvement profiles’ antecedents (parent education and family income) and consequences (students’ academic achievement). This study used latent class analysis (LCA) on data from 12,575 seventh- and eighth-grade Chinese students’ parents to explore their patterns of parental involvement in their children’s learning at home.
1.1. Definitions of Parental Involvement
Parental involvement is defined as a parent’s engagement in activities related to his or her child’s academic achievement (parental involvement has been used interchangeably with parent engagement, parent participation, home literacy practices, home experiences, parent–school relations, parent tutoring, parent curriculum, and parent–child reading [
2]). Regarding their children’s education, many parents act as guardians, teachers’ assistants, and voluntary tutors [
3]. Parental involvement includes home-based involvement and school-based involvement [
1,
4,
5]. Home-based parental involvement includes parent education and parental involvement in children’s learning at home [
6]. Specifically, regarding support for their child’s learning, parental involvement at home includes homework-related supervision (e.g., supervision of schoolwork, checking homework, and homework assistance), reading with children, education expectations and aspirations, and parental attitudes toward education [
7,
8,
9]. School-based involvement includes communication between parents and schools, parents’ voluntary participation in schoolwork, and parent participation in school policy decision making [
4]. The present study focused on home-based involvement.
1.2. Inconsistent Relation between Parental Involvement and Academic Achievement
Previous studies showed mixed effects of parental involvement on student learning outcomes. Some studies showed that parental involvement was positively related to students’ academic achievement [
10,
11,
12,
13,
14,
15]. The effect sizes between parental involvement and academic achievement were found to be 0.25 to 0.30 [
11,
16], 0.22 to 0.62 for minority groups [
12], 0.52 for Latino students [
17], 0.29 for kindergarteners and primary school students, and 0.12 to 0.35 for secondary school students [
10,
16]. In addition, Wilder performed a meta-synthesis across meta-analysis studies and concluded that parental involvement had a positive impact on children’ academic achievement overall [
18].
However, other studies showed nonsignificant or even negative links with students’ academic achievement. For example, home-based involvement had very little effect (
r = 0.03) on students’ academic achievement, no significant effect according to the data of the National Educational Longitudinal Study of 1988 (CI was −0.17 to 0.08), and a negative effect (CI was −0.14 to −0.11) when using the Longitudinal Study of American Youth data [
19]. Mattingly et al. evaluated different programs of parental involvement from kindergarten to secondary education and found no evidence of a connection between promoting parental involvement and improvement in students’ learning [
20]. Hill and Tyson found a positive relation between parental involvement and academic achievement, but the effect size was small (0.04) [
19]. Taking these results together, the relation between parental involvement and students’ achievement is inconsistent.
Different definitions of parental involvement (typically, specific behaviors or activities) across studies might account for these inconsistent results. Helping with homework had a negative effect on academic achievement (
r = −0.11) [
19], while supervising homework had little effect on students’ academic achievement (
r = 0.02) [
10]. Listening to children read had a much greater positive effect on students’ academic achievement (d = 0.51) than reading to them (
r = 0.17; d = 0.18) [
10,
21]. Among different parental involvement behaviors, parent expectations also showed a positive link with students’ academic achievement (
r = 0.33;
r = 0.22) [
10,
11]. In addition, parent–child interactions could improve a child’s learning outcomes [
22]. Overall, the relation between parental involvement and students’ academic achievement varied according to the definition of parental involvement in specific studies.
1.3. Profiles of Parental Involvement in Students’ Learning at Home
To determine how parental involvement is related to children’s learning outcomes, we can evaluate the effects of overall parental involvement or of each of its components on children’s academic achievement, but this variable-centered approach cannot address the complex interactions among the variables. As the number of variables increases, the number of higher-order interactions (three-way, four-way, etc.) increases, which reduces the statistical power to detect significant effects.
Furthermore, the variable-centered approach assumes a sample from a single population, so its analytic results (e.g., reading to children at home is positively correlated with their reading performance) are assumed to be generalizable to that entire population [
23]. When the sample has multiple distinct subpopulations, the relations can differ in size or even have opposite directions across subpopulations. Hence, differences in subpopulations can contribute substantial error to variable-centered studies. Hence, family involvement studies need a method for identifying subpopulations [
24].
To overcome this limitation of a variable-centered analysis, we can instead use a person-centered analysis. Specifically, LCA can be used to identify how dimensions are organized within individuals and how values along these dimensions are distributed across a heterogeneous population to separate it into more homogeneous subpopulations [
23,
25]. Then, we can describe the characteristics of each subpopulation and study their antecedents and consequences [
25].
1.4. The Present Study
This study addresses three issues. First, we integrated the indicators of parental involvement used in empirical studies. The LCA identified groups (styles) of parental involvement in learning at home to identify different subpopulations (i.e., classes) of parents based on their levels of parental involvement behaviors. Second, past studies showed that parents’ education, income, or social class can affect parental involvement [
1,
6,
26]. Therefore, we examined whether these factors were linked to profiles of parental involvement. Third, as past studies showed inconsistent links between parental involvement and academic achievement, we examined whether these profiles of parental involvement were related to academic achievement.
3. Results
Table 1 reports the means and standard deviations of each parental involvement item (see the correlations in
Table 2).
3.1. Latent Class Analysis
The latent class analysis tested five different models of one to five classes. Although the five-class model had the lowest BIC value (386,371), the four-class model had the highest entropy (0.760) and latent class probabilities exceeding 0.83, and the LMRT (LMR = 1654.60,
p = 0.016) showed that the five-class model was not superior to the four-class model (
p > 0.01); together, they indicated that the four-class model was the best model (see
Table 3). These four classes were: permissive (54%), supportive (20%), neglectful (18%), and restrictive (8%).
3.2. Classification Accuracy
The discriminant analysis and MANOVA results supported the four-class model. The discriminant analysis indicated that the classification of group membership was 93% correct for permissive (91.1%), supportive (99.7%), neglectful (96.5%), and restrictive (79.0%; see
Table 4). The MANOVA results showed that the latent classes significantly predicted all 17 indicator variables (F = 577.40, df = 51,
p = 0.0001, η
2 = 0.59, and power = 1.0; see
Table A1).
The MANOVA post-hoc tests showed that 91% (93/102) of the indicators for the pairs of classes (permissive vs. supportive, permissive vs. neglectful, etc.) differed significantly (see
Table A1 and
Table A2). The three nonsignificant differences were for permissive versus supportive (action during conflict with child,
p = 0.196) and permissive versus restrictive (supervising homework,
p = 0.054; assigning homework,
p = 0.371). In addition, some pairs also showed low effect sizes: permissive vs. supportive (angry about unfinished homework, ES = −0.10; scolding child for poor scores, ES = −0.09; talking with child about poor scores, ES = −0.06; conflict with child? ES = −0.08), permissive vs. restrictive (setting TV and game time, ES = 0.17), and supportive vs. neglectful (coolly treating poor-scoring child, ES = −0.11). Hence, these items were not well differentiatd within these pairs of parental involvement types.
3.3. External Validation of Latent Classes
Parent education differed significantly across the four latent classes (MANOVA results: F [3, 12489] = 394.0, p = 0.0001) and across all pairs (post-hoc comparisons all showed p < 0.0001), showing a clear ordering from highest to lowest with substantial effect sizes: supportive > permissive > restrictive > neglectful. Likewise, family income differed significantly across the four latent classes (F [3, 12206] = 121.8, p = 0.0001) and across all pairs (all post-hoc comparisons were significant), showing a clear ordering from highest to lowest with substantial effect sizes: supportive > permissive > restrictive > neglectful. In addition, students’ academic achievement differed significantly across the four latent classes (F [3, 12387] = 43.5, p = 0.0001) and across all pairs (all post-hoc comparisons were significant), showing a clear ordering from highest to lowest with substantial effect sizes: supportive > permissive > neglectful > restrictive.
Together, these results show somewhat similar orders.
Parent education: Supportive > Permissive > Restrictive > Neglectful.
Family income: Supportive > Permissive > Restrictive > Neglectful.
Academic achievement: Supportive > Permissive > Neglectful > Restrictive.
These results support those of past studies showing how parent education and family income are linked to parental involvement, along with its link to student academic achievement. As a result, these results help validate the four parental involvement classes.
3.4. The Relations between Indicators of Parental Involvement and Academic Achievement in Each Latent Class
For the four parental involvement profiles, the correlations between indicator and academic achievement showed mixed results (see
Table A3). Furthermore, the correlations between an indicator and academic achievement could differ across parental involvement profiles. For example, the correlation between examining homework and academic achievement was significantly positive overall (0.13,
p < 0.01), as it was for restrictive parental involvement (0.09,
p < 0.01), it was but significantly negative for neglectful parental involvement (−0.06,
p < 0.01) and nonsignificant for both permissive and supportive parental involvement. As these correlations differ across different latent classes of parental involvement, ignoring these distinct subpopulations yields incorrect, oversimplified results for a heterogeneous population.
4. Discussion
Across the last several decades, studies of the impact of parental involvement on student achievement have shown mixed results, in part because of different definitions of parental involvement [
10,
11,
15]. This study showcased a methodology of LCA, discriminant analysis, MANOVA, and correlation analysis in order to determine and validate subpopulations of distinct types of parental involvement.
Specifically, we used LCA on 17 commonly studied measures of parental involvement to identify four latent classes of parental involvement in children’s home learning: permissive, supportive, neglectful, and restrictive. The discriminant analysis showed satisfactory accuracy for the classifications of permissive, supportive, and neglectful parental involvement (all exceeding 91%), but with less accuracy for classifying restrictive parental involvement (only 79%), indicating the need to improve the indicators for distinguishing it from the other three parental involvement styles. The mean effect sizes of three indicators—“I treat my child deliberately coolly when s/he scores poorly” (0.57); “I talk with my child when he/she scores poorly” (0.55); “Do you have conflicts with your child” (0.42)—were far below 0.80, indicating that they had unacceptably low power for distinguishing restrictive parental involvement from the other parental types of involvement [
36]. Future studies can examine these items and, perhaps, revise them to improve their accuracy. The other indicators showed sufficient effect sizes and, thus, power for differentiating the four latent classes.
Supportive parental involvement had the highest values regarding engagement with children’s homework, parent–child interactions, emotional support, and setting clear rules for children at home, along with the lowest rates of conflict with children. These parents were highly responsive to their children, highly respected them, and made high demands of them (e.g., setting clear rules for children), similarly to the attributes of authoritative parenting [
38]. Supportive parents had the highest parent education level and highest family income, and their children had the highest academic achievement. These results suggest that these parents had the knowledge, skills, and resources (e.g., to buy educational materials or learning opportunities/experiences) to effectively help their children learn more. Other studies suggested that their abilities and resources contributed to their greater efficacy in helping their children [
39].
Permissive parents were highly responsive to their children and highly respected, them but made few demands of them [
40]. Compared to supportive parents, permissive parents had substantially lower involvement, lower education levels, and lower income, while their children performed worse academically than those of supportive parents. In turn, permissive parents had more involvement, more education, and more income than both neglectful and restrictive parents, and children of permissive parents showed higher academic achievement than those of restrictive parents (not significantly differently from those of neglectful parents).
Restrictive parents placed high demands on their children, but were not responsive to them, consistently with the authoritarian parenting style [
41]. The emotional support for children’s homework was lowest for this type of parental involvement. Although parent education and family income were higher for restrictive parents than for neglectful parents, academic achievement was lower for children of restrictive parents than for children of neglectful parents. These results are consistent with those of studies showing that restrictive parenting reduces children’s learning motivation and learning outcomes [
42].
Neglectful parents had the lowest parental involvement in this study. They had low demands and low responsiveness to their children’s learning at home, similarly to a neglectful (or uninvolved) parenting style [
38]. Parents in this latent class had the lowest parent education and lowest family income, though their children’s academic achievement was similar to that of the children of permissive parents and higher than that of the children of restrictive parents. These results suggest that high demands without emotional support harm children’s learning outcomes more than the absence of demands or support. Future studies can examine this issue further.
The correlations between indicators of parental involvement and academic achievement varied, sometimes with opposite directions, across latent parental involvement profiles. This result shows that the relation between such an indicator and academic achievement is not universal, but dependent on the parental involvement context [
11,
19]. Hence, ignoring the parental involvement context can yield inaccurate results [
43,
44].
6. Future Research
Future studies could include questions related to parental involvement with children that complete homework on electronic devices [
45,
46,
47,
48]. The use of the four parenting styles would also permit statistical analyses with the variables of parent gender, subject area, and course type. Furthermore, this study showed how parent education and family income (components of family social economic status (SES)) are related to parental involvement, so future studies can examine this relation and its possible causal mechanisms [
6]. As this study indicates links between parental involvement and student academic achievement, future studies can discern its causal mechanisms and mediating variables, such as students’ emotions or parental attachment [
49]. Future research can explore how parents’ interactions in schools and with teachers and parental involvement styles with their children mutually influence each other [
50]. Future studies can also examine these relations in other countries. Clarification of the definitions of the parental involvement types delineated in this study will facilitate future research on parental involvement (e.g., comparisons across demographic attributes).