A HLM Analysis of the Effects of Parental Involvement on Academic Achievement in Rural Gansu, China

Abstract

 Presenter (s) Xiaoran Yu, Lehigh University; Peggy A. Kong, Lehigh University
Title A HLM Analysis of the Effects of Parental Involvement on Academic Achievement in Rural Gansu, China

Introduction
Overall, research has shown that parental involvement in children’s education contributes to students’ educational success (Chowa, Masa, & Tucker, 2013; Rothon, Goodwin, & Stansfeld, 2012; Sargent, Kong, & Zhang, 2014). Yet, families do not live in isolation. Home, school, and community are immediate contexts that influence a child’s development (Bronfenbrenner, 1986; Shin, 2009). Families are embedded in communities, which exert a strong influence on children’s educational outcomes.

The influence of parental involvement on student academic achievement has been examined in various ways, one of which is the social capital theory (Rothon et al., 2012; Sargent et al., 2014). Coleman (1988) defines social capital as social networks, relations, and interactions between and among actors. Social capital is valuable to people as far as how ‘aspects of social structure can serve as resources that they can use to achieve their interests’ (Coleman, 1988, p.101). Although many studies have focused on the relationship between the amount of social capital and reading achievement of children (e.g., Gilleece, 2015; Harris & Graves, 2010), there is a lack of research in the context of rural China, where parents tend to support their children in non-traditional ways that are invisible to teachers and schools (Kong, 2016). Compared to school-based formal involvement such as volunteering and talking with teachers, Rural Chinese parents are more likely to engage in informal parental involvement such as monitoring and discussion. Additionally, different from western countries, China has a long history of punitive parenting. Parental punishment is considered normal, appropriate, and necessary to discipline children (Chen & Liu, 2012; Hester, He, & Tian, 2009). In contrast to western cultures that emphasize individuality, China has a collective culture. A form of social capital that Coleman (1988) discusses, and which exists with special strength in rural China, is accumulated in the community, which includes social norms and sanctions. China is a collectivist society where families develop close social bonding and networks (Adams, 2006), and people monitor each other’s children and support each other (Dion & Dion, 1996; Liu et al., 2005). Thus, to understand social capital between parents and children is of particular importance in the rural China context (Sun, 1999).

Methods
To investigate the influence of community on children’s academic achievement in rural China, we draw from the 2000 Gansu Survey of Children and Families (GSCF). The 2000 GSCF includes linked target child data with mother, child, household, teacher, school, village leader questionnaires that allow examination of children’s schooling, educational achievement, and social welfare. It draws a multi-stage cluster sample of 2000 children aged 9-12 years old in 100 villages in Gansu. The present study draws information from mother, child, household, and village questionnaires. Parents responded to self-report surveys administered to them about their parenting practices. The analytic sample in this study is restricted to 568 parents of children who were attending grades 1-3 and received Yuwen (Chinese) tests.

In this study, we developed a two-level HLM model to examine how the student-level (Level-1) parenting practices and village-level (Level-2) social capital influences student Yuwen achievement. The outcome variable is student Chinese achievement. Student-level predictor variables include two parenting styles: constructive parenting and punitive parenting. Student-level control variables include child gender, child age, child grade, and mother’s years of education. Village-level predictor variables include village average communication, village average punishment, and community closure. Village average mother’s education is used as the village-level control variable.

Research Questions
In the present study, we intend to answer the following questions:
a. Do individual-level communicative parenting and punitive parenting predict Chinese language achievement?
a.1. If yes, do the effects vary by villages?
b. Do village average communicative parenting and punitive parenting predict Chinese language achievement?
c. Does village community closure predict Chinese language achievement?

Findings
Fully Unconditional Model
Test of the fully unconditional model was conducted, which found that a significant proportion of variance in Chinese achievement (ICC =44%; χ2(97) = 482.54, p < .001) occurred between villages. 56% of the variance in Chinese achievement is due to variability within villages. The design effect of 2.76, which is higher than 2.0, confirmed the appropriate usage of HLM.

Adding Predictors to Level-1 Model
The Level-1 models with child gender, age, grade, mother’s years of education, constructive parenting, and punitive parenting were computed to determine which Level-1 variables predicted student Chinese achievement. Results showed that constructive parenting (grand mean centered) was a positive predictor of Chinese achievement, t(399) = 2.49, p = .01. Controlling for child age and grade, for every unit increase in parent-child communication, Chinese achievement goes up 1.97 points.

To test whether the relationship between constructive parenting and Chinese achievement varied by village, we changed constructive parenting to group mean centered and entered it as a random effect. The results of the chi-square test did not find a significant relationship between village ID and the slope of the relationship between constructive parenting and Chinese achievement, χ2(89) = 75.36, p > .05. Therefore, we decide to keep constructive parenting grand mean centered and as a fixed effect.

Adding Predictors to Level-2 Model
Complete two-level models were then conducted to test for the effects of village average constructive parenting, village average punitive parenting, community closure, and village average mother’s education.

Village-average punitive parenting was grand mean centered, and was found to be a significant predictor of student Chinese achievement, t(96) = -2.84, p = .01. This indicated that students in villages where on average more parents practice punitive parenting have lower Chinese achievement. village-average mother’s education (t(95) = 4.67, p < .001) is positively associated with student Chinese achievement.

The two-level model supports the hypothesis that students from villages with more parents who practice punitive parenting have lower Chinese achievement. The study confirms the importance of parent-child constructive communication on reading achievement at the individual level in the context of China. Yet, the study did not find support for the hypothesis that village-average constructive parenting lead to higher Chinese achievement, nor the hypothesis that students from villages with more parents knowing each other have higher Chinese achievement.

2 Responses

  1. M. Najeeb Shafiq

    Thanks for the clear poster and interesting study! The results are somewhat unexpected (punitive is not statistically associated but constructive is statistically associated) but make sense. A basic question: How did you construct the “constructive” and “punitive” parental approaches? That is, what survey questions did you use, and what did the variables look like (Likert or dummy)? Also, would you describe the statistical associations sizes as large or small?

    Best wishes,
    M. Najeeb Shafiq
    University of Pittsburgh

  2. Xiaoran Yu

    Hi Najeeb,

    Thank you for your comment! The items used to construct the “constructive communication” domain include ones that ask parents “If your child has done something wrong, you ask for reasons and then discuss,” and “When you ask the child to do something, you explain to the child why,” on a 3-point Likert scale (1 = never, 2 = sometimes, 3 = often). In the domain punitive parenting, parents responded to questions such as, “If your child has done something wrong, you beat her/him,” and “if your child has done something that annoys you, you purposely ignore the child,” on a 3-point Likert scale (1 = never, 2 = sometimes, 3 = often). Your comment on statistical associations sizes is greatly appreciated. The intraclass correlation coefficient is 44%, but we have not yet estimated the standardized coefficients or the proportional reduction in variance. Will definitely examine the effect sizes!

Leave a Reply