Methods
The participants in this study came from a large longitudinal study (Child Academic and Psychological Development Study-CAPS) that examined the influence of family, school, and individual factors on children’s academic and mental health development. The project began in the fall term of 2016 (the fall semester of the fourth grade), and each test interval is half a year. The data used in the current study were gathered in five waves (the five waves were given in the spring semester of the fourth grade, the fall semester of the fifth grade, the spring semester of the fifth grade, the fall semester of the sixth grade, and the spring semester of the sixth grade) from children and their parents (both mothers and fathers) of 36 public primary schools in Baoding City, Hebei Province, China. The economic level of Hebei Province was slightly lower than the national average level, with a national per capita disposable income of 25,973.8 yuan for the national average and 21,484.1 yuan for Hebei Province in 2017 (National Bureau of Statistics of China, 2018).
The class was regarded as a unit for the test at each time point in the students’ tests. The trained supervisors (graduate students and undergraduate students) read the uniform instruction and told the students that the test was only for scientific research, the teachers and parents could not see their answers, and the questionnaires would be collected immediately after the test. The parents’ questionnaire and informed consent were brought home by the students the day before the test. After the parents agreed to participate in the test, they answered the father’s or mother’s questionnaire independently, and data were collected the next day. These procedures were approved by the Institutional Review Board of the Collaborative Innovation Center of Assessment toward Basic Education Quality, Beijing Normal University.
A total of 4,004 children were enrolled in the initial sample, and questionnaires with a miss rate of more than one in 10 were removed; therefore, 3,798 children were involved in the study at Time 1. Due to the transfer of students, sick leave, or other reasons, there was a certain loss in the data. The final sample included 3,627 children who had completed at least three tests, among whom 1,885 were boys (52%) and 1,742 were girls (48%), with an average age of 10.421 years (SD = 0.518) at Time 1. We examined whether systematic differences were observable in the children involved and those who dropped out of the study. Chi-square tests revealed no differences between the two groups in the core variable of growth mindset at Time 1, χ2(1) = 1.284, p = 0.257, children’s gender, χ2(1) = 0.928, p = 0.336, children’s age, χ2(1) = 0.256, p = 0.613, and annual household income, χ2(1) = 3.430, p = 0.064. However, there were significant differences in mothers’ education level, χ2(1) = 27.116, p < 0.001, and fathers’ education level, χ2(1) = 34.717, p < 0.001. Little’s missing completely at random test indicated that missing data were consistent with the pattern of the missing completely at random test (χ2 = 34.912, p = 0.424). Among these students, their parents also had to complete three or more tests, and 3,436 fathers and 3,509 mothers were included in the analysis. Nearly 14.3% of fathers and 19.2% of mothers completed primary school education or below, 67.1% of fathers and 60.7% of mothers completed middle school education or secondary vocational school education, 11.6% of fathers and 12.5% of mothers completed junior high school education or technical secondary school education, and the remaining parents completed a 3-year college or bachelor’s degree or higher.
Measures
Growth mindset
Growth mindset was independently self-reported by children, their fathers, and their mothers at five time points by using the Mindset subscale, which was originally developed by Dweck (1999). The original scale contains 8 items, and we used four of these questions to measure the growth mindset. An example item was “Everyone can significantly change his or her ability.” The questions were graded on a 5-point scale, with 1 for “strongly disagree” and 5 for “strongly agree.” Higher scores indicated a higher growth mindset. The Cronbach’s alpha coefficients of the five tests ranged from 0.704 to 0.811 for children, from 0.765 to 0.827 for fathers, and from 0.744 to 0.836 for mothers.
Demographics
In this study, the school provided the children’s gender and birth date of the students. Parents reported their education level and family annual income at time point 1. The average annual household income reported by the father and mother was taken as the annual household income. The education level of the parents was coded as follows: (1) primary school and below; (2) middle school education or secondary vocational school; (3) junior high school or technical secondary school; (4) 3-year college; (5) a bachelor’s degree; and (6) a master’s degree or above. Family annual income was coded as follows: (1) less than 3,600 yuan; (2) 3,601–7,200 yuan; (3) 7,201–14,000 yuan; (4) 14,001–30,000 yuan; (5) 30,001–50,000 yuan; (6) 50,001–100,000 yuan; (7) 100,001–200,000 yuan; (8) 200,001–300,000 yuan; (9) 300,001–500,000 yuan; and (10) more than 500,001 yuan. Therefore, in this study, the average father’s education level was 2.14 (SD = 0.813), the average mother’s education level was 2.12 (SD = 0.880), and the average family annual income was 4.063 (SD = 1.863).
Data analytic plan
First, SPSS version 21.0 was used to conduct the descriptive statistics on the research variables and calculate the correlation coefficient to observe not only the relationship between parents’ growth mindset and children’s growth mindset but also the stability of their growth mindset. To test whether the change in growth mindset was due to true changes in the participants’ beliefs rather than the change in the scale structure, we performed a longitudinal measurement invariance analysis. The configuration invariance model, weak invariance model, and strong invariance model were established in turn, and then, these nested models were compared by using the chi-square difference test. If the chi-square difference test between the two models is significant, then the measurement invariance is not satisfied (Satorra and Bentler, 2001). However, previous studies have found that the chi-square test is easily affected by the sample size; with an increase in the sample size, even a small difference will result in a significant difference, thus, the Fitting Index Difference method can also be used to test measurement invariance (Cheung and Rensvold, 2002; Meade et al., 2008).
To examine the mean growth pattern in the growth mindset for children, fathers, and mothers, we used Mplus version 7.11 (Muthén et al., 2013) to establish three separate univariate latent growth curve models to investigate the initial level and development trend of the growth mindset and whether there are individual differences. There were two latent factors, namely, an intercept factor (which described the average initial level) and a slope factor (which described the development and change). The initial level (intercept) and growth (slope) were allowed to correlate. The load on the intercept factor of the repeated measurement was fixed at 1 at five time points. As we were concerned about the long-term impact of parental mindset on children’s mindset, the load on the slope was fixed at 0, 1, 2, 3, and 4 according to the time interval of the test for fathers and mothers from T1 to T5, while it was fixed at −4, −3, −2, −1, and 0 for children from T1 to T5 (Biesanz et al., 2004). Moreover, the variance of the intercept and slope described the variability of the initial level and change separately.
Then, a parallel process latent growth curve model was established by using the abovementioned models to examine the associations between developmental changes in growth mindset for children and their fathers and mothers. In this model, the latent factors (intercept and/or slope) of parental growth mindset were used to predict the latent factors (intercept and/or slope) of children’s growth mindset while controlling for the correlation between parent–child’s growth mindset in the same period. We established two models with or without control variables. Children’s gender and age, father’s education level, mother’s education level, and annual family income were added to the conditional parallel process latent growth model (Figure 1).
Figure. Parallel process latent growth curve model of children’s and father’s and mother’s growth mindset. For clarity reasons, intercept and slope factor loadings for variable and correlation coefficients are not depicted. GM, growth mindset; S, slope. (See the original article for the figure.)
Finally, to test whether the paths from parents’ growth mindset to children’s growth mindset differed by children’s gender, we conducted multiple group analyses. Specifically, we divided boys and girls into two groups and performed a series of Wald tests for different paths from parents’ growth mindset to children’s growth mindset.
In this study, model fit was assessed by the comparative fit index (CFI) and root mean square error of approximation (RMSEA; Hu and Bentler, 1999). A maximum likelihood estimation (ML) was used for model estimation, and the full-information maximum likelihood method (FIML) was used for missing data processing.
CC BY 4.0. "The longitudinal association between children's growth mindset in senior primary school and their parents' growth mindset" by Chen J, Liu C, Frontiers in psychology (2023). https://doi.org/10.3389/fpsyg.2023.1110944. Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Peer-reviewed and open access; full text from PubMed Central (PMC10043423). Headings and layout adapted for reading; figures and tables are in the original.