Data
The data in this paper is from the Chinese Family Panel Studies 2010 baseline survey data (CFPS2010). CFPS2010 covered 14,960 households in 25 provinces, municipalities, and autonomous regions in China involving three questionnaire surveys for each household: namely the family questionnaires, adult questionnaires for those aged 16 and above, and the children’s questionnaires for those aged 16 and under. The children’s questionnaire was divided into the part reported by the parents and by the part by children themselves (10–15 years old). The research object of this article is children aged 10–15 years who are having compulsory education and who have filled in self-administered questionnaires. We matched the data obtained from the children’s questionnaire with the data from the family and parent questionnaires while removing samples containing missing variables. Finally, we obtained 2750 cases for analysis in the paper.
Measurement
Family SES is one of the key explanatory variables of this article. The following three indicators were used in the analysis for measurement. The first indicator is the net income of households per capita in 2009; the second is the years of education of the father; the third indicator is the years of education of the mother.
Parental participation in their children’s education is an important mediator of the influences of family SES on the academic achievement of children. In the surveys, four interview questions were engaged to measure the parents’ participation in their children’s education. First, “when the child is learning, will you always cease watching your favorite TV programs?” Second, “have you often discussed what happens in school with the child since the beginning of this school year?” Third, “Do you often ask the child to complete his homework?” Fourth, “Do you often check the child’s homework?”. The measures are ordered from 1 to 5, indicating never, rarely (once per month), occasionally (1–2 times per week), frequent (2–3 times per week), and very often (6–7 times a week). In the multiple regression analysis, we took the average of these measures as the value of parents’ educational participation for analysis.
The quality of the school that children attend has a very important influence on their learning behavior and academic achievement. Four measurements were used to measure the quality of children’s school attendance: first, children’s satisfaction with the school; second, children’s satisfaction with the class advisor; third, children’s satisfaction with the Chinese teacher; and fourth, children’s satisfaction with the Mathematics teacher. The scale of these indicators ranged from 1 to 5. The higher the value means the higher level of the satisfaction. In the multiple regression analysis, we take the average of these four as the value of the school quality. Although the subjective evaluation of children may not fully reflect the quality of the school they attend, it still reflects to a great extent their perception and evaluation of the quality of the school. We look forward to further studies that can make up for the deficiencies in the school’s quality measurement in this paper.
The educational services that children received in the market are measured by the following two indicators: first, whether the children participated in a remedial class in the previous semester, and, second, children’s extracurricular tutoring/tutoring expenditures last year.
The measurement of children’s learning behavior, including their daily learning habits, was surveyed with the following four interview questions. First, “I study very hard”; second, “I concentrate on learning in class”; third, “I only play after completing my homework”; and fourth, “I check it several times to make sure it is correct after finishing my homework.” The measurements of the variable range from 1 to 5, representing very disagree, disagree, neutral, agree, and agree very much respectively.
The measurement of children’s academic achievement involves two types of indicators. First, the parental assessments of language and mathematics scores, which were surveyed with “What do you know about the language/mathematics scores of your child last semester”. It is an ordinal variable ranging from 1 to 4, with 1 poor, 2 medium, 3 good, and 4 excellent. The second category includes the CFPS2010 benchmark scores of children’s words and math ability, with the degree of difficulty adjusted based on the level of children’s education. The scores were standardized according to the province of the child and the grade of enrollment in the analysis.
In studies of the relationship between children’s academic achievement and family background, the ranking of family socio-economic status is usually measured at the national level. It is necessary to pay special attention to the fact that the opportunities of secondary education for children in China are rather regional, and the selection of middle schools from elementary schools, of high schools from middle schools, and of colleges from high schools is implemented based on the regional (county, city, and province) processes gradationally. The access to educational opportunities at a higher level does not depend on the children’s ranking at the national level, but on their relative location within the region. In the same way, their competitors are also not country-level students but the peer group in that specific region.
Therefore, both the influence of family background and the measurement of academic achievement should be relative and regional based. In the multiple regression analysis, we controlled the regional differences in children’s academic achievement and family socioeconomic status by adding provincial dummy variables. In the structural equations, we also standardized measures such as children’s academic achievements, remedial class expenses, and family socioeconomic statuses according to provinces and grades, that is, controlling for the differences in grades and regions in the analysis. For that, the control variables also include gender and ethnicity.
Table 1 reports the sample distribution and descriptive statistics of each of the measured and latent variables. In our sample, urban samples took 38.3%, rural samples 61.7%, boys accounted for 50.6%, and girls 49.4%; 63.7% of children enrolled in primary school and 36.3% enrolled in middle school.
Table 1 Descriptive statistics of the main variables (N = 2750) Method
To simultaneously estimate the relationship between observable indicators and latent variables and the relationship within these latent variables themselves, structural equation model is used to estimate the relationship between family background variables and children’s academic achievement. Based on the analysis framework (Fig. 1) and research hypotheses of this paper, the structural equation model was set as follows (see Fig. 2). For the corresponding relationship between latent variables and measured indicators, please refer to Table 1.
First, the socio-economic status of exogenous latent variables has a direct impact on children’s quality of school attendance, education services children receiving on market, parental education participation, and children’s academic behavior, and indirectly affects children’s academic achievement. We set the socio-economic status of the family as the only exogenous variable other than gender, ethnicity, and region. Past research shows that parents’ parenting style, the quality of children’s school, and children’s own educational expectations and learning behaviors are all affected by the socio-economic status of the family extensively.
Second, key schools typically have excellent teachers and students, which not only has a direct impact on children’s academic achievements, but also affects their learning attitudes and behaviors through teachers and peers. We propose that the quality of children’s school and parental education participation can directly affect children’s academic achievement and can also have an indirect effect on children’s academic achievement through the mediator of children’s academic behavior.
Third, there is no direct measure for laten variable children's academic achievement in Fig. 2. Instead, in the model, it is regarded as a high-level latent variable measured by the children’s benchmark test (Test) and performance ranking (Rank).
Fourth, as it can be arbitrary to assume the correlation between the measurement error terms of the variables which is to be adjusted according to LISREL, it is assumed that the error terms of all endogenous variables are not relevant.
Fifth, the urban-rural differences in the mechanisms of family background affecting children’s academic achievement are examined by comparing the urban sample with the rural sample.