Strengthening the Social Information Processing Skills of Third Graders in Rural China

Abstract

Objective:

The study aimed to describe findings from a matched clusters trial of a school-based intervention Let’s Be Friends designed to promote social competence and prevent maladaptive behavior by strengthening the social information processing (SIP) skills of third-grade children in rural China.

Method:

Using a blocked cluster design with random sampling, 13 treatment schools (n = 355) of a total of 67 schools in the study site were randomly selected, and 14 schools (n = 341) formed the control condition. All third graders from the treatment schools received 14 sessions of intervention.

Results:

The adapted program appears to have strengthened children’s SIP skills, reduced aggressive behavior, and promoted cognitive concentration. The study does not confirm that children growing up in single-child families fare worse than children living with siblings.

Conclusion:

The program has the potential to strengthen the SIP skills of children in rural China.

Keywords

aggressive behavior childhood children emotional regulation intervention research social–emotional skills social information processing

Background on Social Information Processing (SIP)

SIP skills are linked strongly to developmental outcomes in childhood and adolescence. In China, preparing children to participate meaningfully in society is a widely recognized feature of national efforts to promote social harmony, a key theme of public policies since the early 1980s. However, this task has been complicated by dramatic social and economic changes. In 2018, China observed its 40th anniversary of the launch of economic reform and open-door policies. Some 40 years of rapid economic development has made China the second largest economy in the world, and it has brought about dramatic social changes in the country. As an element of sweeping reforms associated with the adoption of a market-oriented economy, family life and work are being transformed by modernization and migration. Children who migrate with their families from rural to urban areas or who are left behind in the rural areas as their parents seek employment in urban areas face numerous adaptational challenges. The rapid growth of a marketplace economy has created conditions that make emotional regulation and SIP more challenging (Wu et al., 2016; J. Zhang, 2018; W. Zhang, 2000; Zhao, 2010). Strengthening the SIP skills of these children may promote successful adaptation and, more broadly, contribute to social harmony (Biao, 2007; Duan & Zhou, 2005; Nielsen et al., 2005; Yi, 1986).

The purpose of this article is to describe findings from a matched clusters trial of a school-based intervention—Let’s Be Friends (LBF)—designed to promote social competence and prevent maladaptive behavior by strengthening the SIP skills of third-grade children. The trial is the largest of its kind in China. It was implemented in a rural area with a study sample of 681 primary school children. The intervention is a product of a 15-year collaboration between Sino–U.S. social work researchers to adapt a successful U.S.-based program—Making Choices: Social Problem-Solving for Children (MC)—for Chinese children. Hence, the work is both translational and transcultural in nature.

The core perspective guiding the design of the MC program is SIP theory (Crick & Dodge, 1994). As a cognitive theory, the SIP model provides a coherent structure and flow for understanding what happens during social decision making. The theory is based on six steps comprising SIP: encoding cues, interpreting social and environmental cues, setting social goals, generating alternative responses, evaluating and selecting a response, and enacting a response strategy (Crick & Dodge, 1994; Dodge et al., 2002). Although these steps are ordered, information processing in real time contains feedback loops and iterations that lead nonlinearly to interpretations, response alternatives, response decisions, and behavioral enactment.

The SIP perspective provides a rich foundation for intervention research. It has led to a better understanding of how biased processing produces aggressive and delinquent behaviors. It has contributed to understanding the interpersonal decision-making processes of youths with anxiety, depression, and learning disabilities. It has provided a conceptual framework for the development of strategies for teaching social cognitive skills to prevent aggressive, delinquent, and substance abusing behaviors, and, along with other theories, it has informed the design of multielement school-based prevention and intervention programs (for a review, see Kupersmidt & Stelter, 2018).

Perhaps the most important contribution of the SIP perspective is that it shows how cognitive deficits—in the form of skills learned in early childhood—contribute to developmental trajectories that lead to antisocial and aggressive behavior in later life. According to SIP theory, information processing skills that may be functional in one setting can produce biases or errors in thinking in other settings. Children need not only to learn the core skills related to SIP (e.g., encoding cues) but also to learn in what settings information processing (e.g., interpreting cues) must be conditioned on the context. In part, SIP skills provide a foundation for understanding social competence in childhood, and, because SIP skills may be malleable, they have the potential to become a basis for interventions that disrupt negative developmental trajectories leading from early childhood maladaptive behavior to problematic behavior in early adolescence (Fraser et al., 2005, 2014).

Literature Review

Prior studies reveal that SIP skills have a strong correlational relationship with developmental outcomes in childhood and adolescence. SIP skills predict academic achievement in school (L. Chen et al., 2017; Denham et al., 2013), career success in later life (Carneiro et al., 2007; Heckman et al., 2006), and adaptation across a variety of social and environmental contexts (Dykas & Cassidy, 2011; J. Zhang, 2018). On the other hand, maladaptive behavior such as noncompliance in the home, disruptive behavior in the classroom, and delinquent or criminal involvement in the community are associated with biased information processing (Arsenio et al., 2009; Fang, 2006; Lemerise & Maulden, 2010; Zhong et al., 2015). Biased information processing is associated also with elevated risk for experimentation with harmful or illicit substances (X. Chen, 2016; Crawley et al., 2015; Nickoletti & Taussig, 2006), alcohol misuse (Ogle & Miller, 2004; Sayette et al., 1993), gambling (Smith et al., 2012), internet addiction (Jiang et al., 2016), fighting and other violent behavior (D. Liu, 2015; Terzian et al., 2014), bullying (Espelage et al., 2015; Luo, 2010), theft (Fontaine, 2006), robbery (Nas et al., 2005), depressed mood and symptoms (Li et al., 2016; Y. Liu et al., 2016; Pössel et al., 2006; Venta et al., 2014), and suicidal ideation (Fan, 2005; Yao, 2015). In sum, the evidence suggests strongly that SIP skills are related negatively to a range of behavioral health problems and positively to life course outcomes, including meaningful participation in work and community activities.

In order to adapt and update the LBF program, the research team conducted a systematic review of databases in both China and the United States. The review of the Chinese literature was performed on two databases: the China National Knowledge Infrastructure, and China’s Social Sciences Citation Index. Using four sets of key words (i.e., “social information processing 社会信息加工,” “social work + intervention 社会工作 + 干预,” “youth & adolescents + emotional regulation 青少年 + 情绪控制,” and “youth & adolescents + psychological health 青少年 + 心理健康”), the study team extracted a total of 3,253 articles from the databases from the inception of each database to November 5, 2018. Removing duplicative articles, the study team harvested a total of 3,099 studies. From a brief and focused review, studies that did not sample participants whose ages were between 6 and 12 years were removed, leaving 1,743 articles. Then, after scanning the abstracts, introductions, and conclusions of these articles, the study team further removed articles not pertaining to children’s behavior and health. This produced 493 articles. Reading these articles in detail, the study team identified 52 articles that used SIP or other cognitive–behavioral theories as a conceptual framework. Of the 52 articles, only 5 involved intervention research. All these studies underscored the importance of promoting SIP skills for Chinese children, particularly their functions in emotional regulation and aggressive behavior reduction. Of the five intervention studies, only three used SIP theory to guide the design of interventions, no study was sited in a rural area, no sample exceeded 300 children, and no study employed a control group design. Figure 1 shows the flowchart of the systematic review of the Chinese literature.

Figure 1.

A flowchart of the systematic review of China National Knowledge Infrastructure and China’s Social Sciences Citation Index.

The systematic review of the Western literature was performed on three databases: MEDLINE, PsycINFO, and SocINDEX. The review focused on peer-reviewed journals and dissertations that were published during the period of January 1, 1997, to February 28, 2019. Using four sets of key words (i.e., “social information processing,” “children + emotion,” “social work + intervention,” and “children + mental health”), the study team retrieved 268,175 articles. Removing duplicative articles and those irrelevant to children aged 6 to 12 years (i.e., the age-group on which the current research focuses), 30,980 articles were recovered. Of these articles, 24,500 were pertinent to children’s cognitive, behavioral, and mental health. Removing articles unrelated to SIP, the study team reduced the sample to 337 articles. Scanning the abstracts, introductions, and conclusions of these articles, the study team further reduced the sample to 35 articles that pertained to intervention research using the SIP theory. Of these 35 articles, only 4 took place in China, and the remaining studies took place in Western societies including the United States, the Netherlands, Greece, and Spain. These studies in general show that SIP-based interventions helped improve emotional regulation and had the potential to reduce proactive/reactive aggression and conduct problems. Of the four studies about China, no study was sited in a rural area, no study had a sample greater than 300 children, and three studies employed a control group design. Figure 2 shows the flowchart of the systematic review of this literature.

Figure 2.

A flowchart of the systematic review of United States’ MEDLINE, PsycINFO, and SocINDEX.

The Study

The Program

LBF was adapted from MC, which was originally developed in the United States (Fraser et al., 2000, 2005). Sponsored by the National Institute on Drug Abuse and the Department of Education in the United States, a team worked over 20 years, starting in 1995, to develop the MC program from research on processing social information and regulating emotions. For elementary school children, the program was tested in a variety of settings by itself (Fraser et al., 2004, 2014; Terzian et al., 2014) and in conjunction with other interventions (Fraser et al., 2004). In addition, it was adapted for and tested with preschool-age children (Fraser et al., 2011).

To address Chinese children’s social development in the context of rapid social and economic development, a research project was launched in 2005 to develop programs that strengthen the social skills of children. Jointly sponsored by the China’s National Commission on Family Planning and Population Control and the University of North Carolina at Chapel Hill, a research team translated the MC manual from English into Chinese, adapting existing program content and creating new content for Chinese children.

The cultural adaptation of MC for children in China (i.e., the translation and adaptation of the program manual and the publication of its Chinese version) involved three major efforts: (1) a research team comprised of 12 researchers in China, and the United States studied the Chinese child development literature (including SIP); read and translated the MC treatment manual and related materials; observed MC being used in U.S. elementary schools; visited Chinese primary school classrooms; and discussed the project with China’s stakeholders including government officials, primary school teachers, and school administrators to understand the needs and key issues confronting children in China. (2) The research team appraised the SIP-based interventions in both countries, translated and back-translated the MC treatment manual, and adapted the program materials and vignette-based outcome measures for assessing SIP skills in children. Finally, (3) the research team invited Chinese scholars, mainly experts on child development and child mental health, to review the translated and adapted manual. Consistent with Chinese cultural and governmental policy practices, this review was formatted as a “public hearing” in Beijing. Ten experts were invited to review the LBF program and comment in a public meeting on the program’s design, features, feasibility, relevance, and replicability. Based on written and public comments from these 10 reviewers, the LBF manual was revised, published, and prepared for testing (Wu et al., 2016).

The Chinese version of the treatment manual, entitled “Let’s Be Friends: Interpersonal Skills Training for Children” was published in 2011 by the Chinese Population Press (LBF Research Group, 2011). The adapted program LBF is comprised of eight parts and 31 lessons. This intervention and its treatment manual, at the time of its publication, was one of the few programs of its kind available in China. Major categories of the adaptation of MC for LBF are specified in the Chinese treatment manual. Because intervention research in China, especially programs designed to strengthen children’s social development, is new and rare, this program remains a pioneering effort (see, for instance, Figures 1 and 2). Tests of the LBF program have the potential to produce evidence that contributes to the knowledge base of developmental psychology and social work science, and findings from studies of LBF may help policy makers address the needs of children in China.

Pilot test in Tianjin

Sponsored by China’s Ministry of Education, a research team chose Tianjin city as an initial study site to pilot test LBF in 2011–2012. The Tianjin pilot study was a controlled trial. Children (n = 91), aged 8–10, in five after-school childcare centers in Tianjin, China, received the program. Using propensity score adjustments (Guo & Fraser, 2015), the SIP skills of children who received the program were compared to the skills of children (n = 123) recruited from neighborhood elementary schools. The Tianjin LBF study showed that the adapted program strengthened children’s encoding skills; patterns for other information processing skills were promising but mixed. As a proof-of-concept trial, the study suggested that the program had the potential to strengthen the SIP skills of children in China (Wu et al., 2016).

Shaanxi test

This study was designed to extend the findings from the Tianjin pilot test by implementing LBF in a rural area with a more rigorous design and a sample that would include children left with relatives as their parents migrate to urban areas. Jingyang County in Shaanxi Province was selected as the study site. A rural county with agriculture as the main industry, Jingyang is located in China’s Midwest. The study site contains three overlapping subpopulations representing modern Chinese children. These include (1) children who mainly grow up in the countryside and live with parents who are not rural-to-city migrants, (2) children who live with grandparents or other relatives while their parents relocate to cities or other employment centers (i.e., the so-called left-behind children), and (3) children who grow up in a single-child households without siblings. Hence, the study provides unique opportunities to estimate program effects for important subpopulations among Chinese children, to further test the effectiveness of LBF, and to develop practice principles regarding the translation and cultural adaptation of U.S. research for China. This article only reports findings pertaining to subpopulation (1) and (3) as findings about the left-behind children or subpopulation (2) are published elsewhere.

The MC and LBF programs developed a series of lessons intended to teach SIP skills. Using a variety of age-appropriate activities, such as playing games, conducting role-plays, drawing or creating art work, and discussing topics in small groups, the intervention provides children with opportunities to learn SIP skills step by step (Fraser et al., 2000; LBF Research Group, 2011).

Study Approval and Human Subject Protections

A group of peer-based professionals of Xi’an Jiaotong University and administrators of Jinyang County including the County governor reviewed and approved the study’s measures for human subject protections on May 2018. This procedure is equivalent to the Institutional Review Board known in the United States.

Design

A matched clusters design with random sampling was used to form the treatment and control groups. Specifically, the design employed a blocked or matched clusters design with random selection of intervention schools. Schools, rather than students, were randomly selected into a treatment condition under the Stable Unit Treatment Value Assumption or SUTVA (Rubin, 1986), which is a fundamental assumption embedded in all program evaluation projects. The assumption states that potential outcomes for any unit do not vary with the treatment assigned to any other units, and there are no different versions of the treatment. The matched clusters design first randomly selected the treatment schools from the sampling frame and then employed the Mahalanobis matching to find a control school for each treatment school. Using matching or nonrandom method to form the control group reduces selection bias and help increase the study’s internal validity.

Mahalanobis matching

Treatment schools were selected randomly from the 67 eligible county schools, and control schools were selected from the remaining 53 schools. After randomly sampling treatment schools from the sampling frame, control schools were selected through pairwise matching.

Based on administrative data available at the school level, the study team identified a matched school for each treated school using Mahalanobis metric distances. The Mahalanobis metric distance is computed by using the following formula: $d (i, j) = {(u - v)}^{T} C^{- 1} (u - v)$ , where u and v denote values of the vector of matching variables for the treated school i and the control school j, respectively, and C denotes the sample’s variance matrix. In the above formula, d(i, j) is the Mahalanobis metric distance. For every treated school i, the study computed its d(i, j) distances with all potential controls remaining in the matching pool, where j = 1, 2,…n, and n is the total number of potential controls. The study defined j₁ as the matched school for i, if and only if d(i, j₁) is the smallest among all Mahalanobis distances associated with i. Eight school-level variables were used to form the u and v vectors: total number of students in the school, percentage of school’s registered left-behind children, number of computers per 100 students, number of library books per student, teacher–student ratio, percentage of teachers with a college education or higher, percentage of teachers owning a middle-level certificate or higher, and percentage of teachers owning a high-level certificate.

Using Mahalanobis distances, 14 treatment schools were matched—or blocked—with 14 control schools. Matching was done without replacement. At the time when the LBF intervention began, one randomly selected treatment school was shut down for remodeling, and all students from this school were transferred to a neighboring elementary school that happened to be a treated school. As a result, the treatment condition was comprised of 13 schools (355 students), and the control condition was comprised of 14 schools (341 students). The final data set used in the analysis further removed 15 students because these students were in special needs classrooms, and the LBF program is designed for regular education classrooms. In the final data set as well, pairwise matching was ignored. Described below, clustering and selection effects were controlled statistically. The total number of participants analyzed is 681, with 343 students (50.4%) receiving treatment and 338 students (49.6%) serving as controls.

Providing the LBF program in treatment schools, the intervention agents—or trainers—were Master of Social Work and Master of Arts students from a neighboring research university located in Xi’an. Trainers received 7 days (i.e., a total of 49 hr) of professional training before the intervention. A supporting team of administrators and classroom-teacher representatives from Jingyang County participated also in the training. The training focused on topics pertaining to intervention research, SIP theory, and the LBF program. Through lectures and presentations made by key developers and researchers from the MC program and by the developers of the LBF treatment manual, all participants received systematic training on the guiding principles and main content of the intervention. Training included demonstration sessions enacted by the intervention supervisors, participatory role-playing of treatment sessions, and group-based rehearsals of selected sessions.

Implementation and supervision

The LBF intervention was comprised of 14 sessions, implemented from September to December in 2018, on a weekly basis. Based on experiences from the Tianjin pilot test, the Shaanxi project condensed the 31 lessons of LBF into 14 sessions and designed the length of each session to be approximately 60 min. All 31 lessons from LBF were completed. Because some of the 13 schools were large, the intervention was provided in 15 groups, with each group having two–three trainers. In total, 29 trainers provided the intervention across the 13 schools. Trainers received university credit for their participation in the project. In addition, they received modest compensation for expenses (e.g., travel). In a typical intervention session, there was a lead trainer moderating the intervention, while one or two additional trainers provided assistance, observed, or kept records.

Project supervisors observed in all treatment sessions. They monitored the provision of intervention content and served as a resource to solve unexpected problems. In addition, elementary school teachers in the intervention condition observed all sessions in which children from their classes participated, although they did not interfere with any treatment activities moderated by the trainers.

Fidelity to the LBF program was maintained and monitored in three ways. First, a detailed treatment manual was used in all sessions (LBF Research Group, 2011). Second, across all sites, two training documents were used: a PowerPoint file of intervention content for each lesson and a trainer’s treatment guideline for each session. Both documents were prepared by site trainers and then were consolidated and finalized by the project lead investigator in charge of the intervention. Before each session, the trainers used the finalized documents to rehearse. Each week, rehearsal was intensive and lasted for 5–6 hr. In total, the intervention trainers conducted 14 rehearsals. And finally, during each treatment session, the project designated a trainer to observe and to complete a questionnaire entitled the “Lesson Process Report.” The report was comprised of 9 items, and each item was evaluated on a 0–10 scale with 0 indicating never and 10 indicating always. Sample questions from the report include “uses nonjudgmental language when describing the lesson and responding to students” and “encourages students to discuss their own experiences.”

Sampling

Participants were third-grade students in primary schools in Jingyang County of Shaanxi Province. The third grade is the main age-group LBF aims to treat and was the group treated in the Tianjin study. Chosen purposively, the county is 42.5 km from Xi’an city where the study team was located. In 2016, the county had 153,695 households with a total population of 542,140, accounting for 1.20% of total households and 1.37% of the total population, respectively, in the Province. Agriculture is the main industry in the county. The county’s gross domestic product per capita in 2016 was 33,024 RMB yuan, while in Shaanxi Province overall, it was 49,000 RMB yuan (Jingyang County Bureau of Statistics, 2017).

Because of the clustering of third-grade students in schools, schools in the county were randomly sampled for participation in the LBF program. That is, because students were embedded in classrooms and schools, students could not be assigned individually to treatment and control conditions, and so, schools were randomly selected for participation in the LBF program. Administrative data from the Jingyang’s Bureau of Education showed that in April, 2018, the county had 73 elementary schools. Removing one school located in the county’s only urban area (i.e., a city school), 72 schools in rural areas formed the sampling frame. Of these 72 schools, 5 were recently established private schools—all were small in size and lacked school-level data—and hence, they were removed from the sampling frame. Each of the remaining 67 schools had, on average, 25 third-grade students.

Sample size and power

Using the Optimal Design software program Version 3.01, the sampling of schools was based on a statistical power analysis. Results of the analysis showed that in order to achieve a minimum statistical power of .80 and a minimum statistical significance of .05, 14 treatment schools and 14 control schools (i.e., to have approximately 350 treated students and 350 control students) were needed. Assuming an intraclass correlation coefficient (ICC) of .05 and .10, respectively, the study was powered to detect medium effect sizes. For an ICC = .05, the study had the capacity to detect an effect size of .34 that is slightly higher than the small effect size defined by Cohen; for an ICC = .10, the study had the capacity to detect an effect size of .44 that is slightly greater than the medium effect size defined by Cohen.

The study sample of 700 participants approximately represents 260,000 third-grade children living in the rural area of Shaanxi Province, China. Statistical significant findings of the study can also be inferred to all rural areas in China’s Midwest region that share basic socioeconomic characteristics as Jingyang County.

Measures

Following the U.S.-based MC program, the LBF—Shaanxi program employed two sets of instruments to measure students’ change on SIP skills. Both sets of measures are evidence based with proven psychometric validity and reliability. The first set is the skill-level activity (SLA) test (Dodge, 1980). The purpose of the SLA is to measure how children interpret and respond to specific interpersonal vignettes. Through pictures and stories, children are asked to imagine that they are in five different story situations involving watching a “dodgeball game,” attending a “mathematics class,” wearing “new pants,” eating “lunch,” and losing “new magazine.” Based on the responses of children in each hypothetical situation, competence in four SIP skill areas is assessed with a grading rubric. These four skills include encoding (α = .78), hostile attribution (α = .52), goal formulation (α = .76), and response decision (α = .80).

The second set of measures is drawn from the Carolina Child Checklist—Teacher Form (CCC). The CCC was completed by the elementary school teacher who worked most closely with each study child. Teachers were asked to rate children on the basis of their behavior over the past 1 month. The CCC questionnaire is comprised of 50 items; each item is assessed on a 6-point Likert-type scale ranging from 0 indicating never to 5 indicating always. Using items from the CCC, 10 defined scales were calculated: Cognitive Concentration, Authority Acceptance, Authority Acceptance subscale—Overt Aggression, Authority Acceptance subscale—Oppositional, Authority Acceptance subscale—Covert Antisocial, Social Contact, Social Competence, Social Competence subscale—Prosocial, Social Competence subscale—Emotional Regulation, and Relational Aggression. The CCC and these scales have been shown in previous work to have construct validity and test–retest reliability (Macgowan et al., 2002).

In the postintervention data collection on December 2018, the study team distributed a Family Survey to each child’s primary caretaker. The survey contained items on the number of people in the child’s family, number of siblings, family income, with whom the child stayed for most of the time, main workplace of the child’s father, main workplace of the child’s mother, length of time the child stayed with their parents or one of them in 2018 (in a metric of month), and paternal and maternal education (years of schooling). These demographic and socioeconomic data provided covariates for the statistical analysis, and they also provided a basis to identify “left-behind children.”

Hypotheses

The LBF—Shaanxi study attempts to test the following two research hypotheses:

Hypothesis 1: The LBF—Shaanxi intervention exerts positive impacts on children’s changes on the SLA and CCC measures.

Hypothesis 2: The change scores on the SLA and CCC measures for children growing up in single-child families are worse than children living with siblings.

Hypothesis 1 was derived from the design of the LBF intervention and the SIP theoretical perspective. Hypothesis 2 was derived from prior studies stating that children growing up from single-child households exhibit more behavioral problems than children living with siblings (J. Liu, 1999).

Statistical Analysis

Although the study employed a controlled experiment, whether the cluster matching produced sufficient balance to remove selection biases was not known. Prior studies have shown that due to unexpected and unobserved factors, controlled studies are often compromised and may fail (Barnard et al., 2003). Therefore, we conducted balance checks on observed covariates.

Two statistical tests were performed to check the balance of covariates. The first was a series of bivariate tests. The study used independent sample t tests for continuous covariates and χ² tests for categorical covariates. In all these bivariate tests, a p value from the test that is less than .05 would suggest imbalance on the tested covariate.

The second test for balance was the normalized difference, ND _x (Imbens & Woodridge, 2009). The formula of this test is $N D_{x} = \frac{|{\overset{¯¯}{X}}_{1} - {\overset{¯¯}{X}}_{0}|}{\sqrt{S_{1}^{2} + S_{0}^{2}}},$ where ${\overset{¯¯}{X}}_{1}$ and ${\overset{¯¯}{X}}_{0}$ represent the mean of x for the treated and control groups, respectively, and $S_{1}^{2}$ $S_{0}^{2}$ represent the variance of x for the two groups, respectively. ND _x > .25 suggests imbalance of x between the two treatment conditions. Normalized differences were estimated for all covariates.

The statistical analysis for continuous outcome variables (i.e., the SLA encoding variable and all CCC scales) employed a multilevel analytic model also known as a hierarchical linear model. The model is shown by the following equation:

Y_{i j} = γ_{00} + \sum_{p = 1}^{P} γ_{p 0} {(X)}_{p i j} + γ_{01} {(Intervention)}_{j} + \sum_{q = 2}^{Q} γ_{0 q} {(W)}_{q j} + r_{0 j} + e_{i j},

where $Y_{i j}$ is the outcome variable for the ith student from the jth school, ${(X)}_{p i j}$ is the pth individual-level covariate for that student, and there are a total of P individual-level covariates, (W) _qj is the qth school-level covariate for the jth school, and there are a total of Q school-level covariates, (Intervention) _j is a dichotomous variable indicating treatment condition—(Intervention) _j = 1 indicates treated and (Intervention) _j = 0 indicates control. $γ_{00}$ , $γ_{p 0}$ , $γ_{01}$ , and $γ_{0 q}$ are model’s regression coefficients in which $γ_{01}$ indicates the treatment effect. e_ij is the residual term of the model. r_0j is the random intercept effect—the inclusion of this coefficient helps control for the clustering effect of children in schools. All outcome variables used in the analysis are difference scores; that is, $Y_{i j}$ is a difference score that uses the postintervention outcome minus its preintervention outcome.

The change scores of three SLA scales (i.e., Hostile Attribution, Goal Formulation, and Response Decision) are ordinal variables in nature. There are three values on each change score which denote three situations regarding a child’s change: improvement, worsening, and remaining unchanged at either positive or negative status. Depending on the coding rubric, there are different meanings for these values. The interpretation of the hostile attribution is −1 indicates that the study child changes from “hostile attribution” in the preintervention test to “nonhostile attribution” in the postintervention test (i.e., improved), 0 indicates that the study child remained unchanged on a hostile attribution (i.e., unchanged on a negative outcome), and 1 indicates that the study child changes from “nonhostile attribution” in the preintervention test to “hostile attribution” in the postintervention test (i.e., worsened). The interpretation for both goal formulation and decision response is that −1 indicates that the study child changes from “nonaggressive” in the pretest to “aggressive” in the posttest (i.e., worsened), 0 indicates that the study child remained unchanged on “nonaggressive” (i.e., unchanged on a positive outcome), and 1 indicates that the study child changes from “aggressive” in the pretest to “nonaggressive” in the posttest (i.e., improved). To discern factors affecting these changes, the study employed an ordered logistic regression model that analyzes the probability of falling into any one of the three changes. The ordered logistic regression is expressed by the following equation:

P (y_{i} = m | x_{i}) = Λ (τ_{m} - x_{i} β) - Λ (τ_{m - 1} - x_{i} ββ),

where $P (y_{i} = m | x_{i})$ is the ith child’s probability of falling into one of the three conditions (i.e., −1, 0, and 1), conditional upon the observed covariates; Λ(.) denotes the cumulative distribution function of the logistic distribution; x_i is vector of the ith child’s independent variables; $τ_{m}$ is the cutoff value of the ordered logistic regression, and there are two cutoff values in our study because there are three values of the outcome change score; and β is the vector of regression coefficients. The study employed a robust estimator of standard errors to control for the clustering effect in which school is used as a cluster. Results of the ordered logistic regression are presented by model-based predicted probabilities computed as average marginal effects. The statistical software package Stata Version 15 was used in all statistical analyses.

Results

To facilitate openness, transparency, and reproducibility, the study data will be submitted to a publicly available repository in future (i.e., approximately 3 years after the study concluded). Except noted specifically, findings of all analyses were reported fully in this section.

Results from the Lesson Process Report indicated that the mean value of all measures of treatment fidelity ranged from 8.1 to 9.5, indicating that the observers were very satisfied with the performance of trainers. The optimal length for each session was 60 min, and in practice, the shortest session was Session 11 with an average of 57.3 min, and the longest session was Session 1 with an average of 67 min. The implementation data suggest that LBF was delivered with fidelity.

Table 1 reports the sample descriptive statistics and results of the balance checks. The balance checks show that the p values for all bivariate differences between intervention and control groups were greater than .05, indicating that the two conditions did not differ significantly. All ND _x values were less than .25. Similarly, these tests suggest that the blocked cluster design with random sampling balanced the two treatment groups on observed variables. Differences observed from the change scores on outcome measures are unlikely to be due to selection biases related to observed variables.

Table 1.

Descriptive Statistics of Study Children and Balance Check.

Variable	Sample Descriptive Statistics				Balance Check
Variable	N	%	Mean	Standard Deviation	% Treated (Mean Treated, Mean Control)	P value From Bivariate Test	Normalized Difference
Treatment condition
Control	338	49.6
LBF treated	343	51.0
Total study children	681	100.0
Gender
Boy	329	48.3			50.2	.914	.006
Girl	352	51.7			50.6
Age			8.689	0.471	(8.72, 8.65)	.062	.101
Number of people in the family			5.091	1.215	(5.11, 5.07)	.716	.020
Number of siblings			0.874	0.897	(0.897, 0.849)	.478	.039
Grow from a single-child family
Yes	197	28.9			52.3	.523	.035
No	484	71.1			49.6
Per capita family income in 2018
3,000 or below	306	44.9			51.3	.196	.020
3,001–4,000	147	21.6			46.9
4,001–5,000	74	10.9			59.5
5,001–6,000	56	8.2			44.6
6,001–7,000	24	3.5			33.3
7,001–8,000	13	1.9			38.5
8,001–9,000	12	1.8			41.7
9,001–10,000	19	2.8			52.6
10,001 or above	30	4.4			66.7
Father’s number of schooling years			8.796	2.191	(8.71, 8.88)	.326	.053
Mother’s number of schooling years			9.006	2.409	(8.92, 9.10)	.334	.053

Shown in Table 1, the sample was comprised of slightly more girls (51.7%) than boys (48.3%), the average age of children at the time of intervention was 8.689 years, the average family size was 5.091, the average number of siblings was 0.874, and 28.9% of the children lived in single-child households. The distribution of family income per capita was as follows: 306 children (44.9%) whose family income per capita was 3,000 yuan or below, 147 children (21.6%) whose family income per capita was 3,001–4,000 yuan, 74 children (10.9%) whose family income per capita was 4,001–5,000 yuan, and these three groups of children accounted for the majority of the study sample (77.4%). The highest family income per capita was 10,001 yuan or above (30 children or 4.4%). The average number of schooling years for children’s fathers was 8.796 years, and the average number of schooling years for children’s mothers was 9.006 years.

To assess the effectiveness of the intervention, Table 2 focuses on the following question: In what direction did observed changes go? If the direction is positive and the mean change of the LBF group is greater than that of the control group, the results suggest that the LBF intervention had a positive effect and strengthened SIP skills or behavioral outcomes. If the direction is negative and the mean change of the LBF group is smaller than that of the control group, the results suggest that the intervention was not beneficial. If the direction is positive/negative and the regression coefficient shows the same direction, then the results suggest that the intervention is beneficial.

Table 2.

Evaluation of Treatment Effects: Results of Bivariate and Multivariate Analysis.

Outcome	Direction of Beneficial Effect	Mean Change of LBF Group	Mean Change of Control Group	Estimated Regression Coefficient of LBF
Outcome	Direction of Beneficial Effect	Mean Change of LBF Group	Mean Change of Control Group	Estimated Regression Coefficient	95% Confidence Interval
SLA encoding
Dodgeball	+	2.915	0.757	2.101***	[1.76, 2.44]
Math class	+	2.574	0.757	1.773***	[1.53, 2.01]
New pants	+	2.764	0.793	1.971***	[1.55, 2.39]
Lunch	+	2.822	1.157	1.688***	[1.31, 2.06]
New magazine	+	1.359	0.462	0.913***	[0.59, 1.23]
SLA hostile attribution
Dodgeball	−	−0.058	0.056	−0.294	[−0.85, 0.27]
Math class	−	−0.155	−0.083	−0.277	[−0.79, 0.24]
New pants	−	−0.117	−0.065	−0.064	[−0.38, 0.25]
Lunch	−	−0.169	0.059	−0.722**	[−1.25, −0.20]
New magazine	−	−0.201	−0.027	−0.518*	[−0.98, −0.06]
SLA goal formulation
Dodgeball	+	0.257	0.080	0.542*	[−0.05, 1.12]
Math class	+	0.058	0.065	−0.061	[−0.38, 0.26]
New pants	+	0.085	−0.009	1.019**	[0.44, 1.60]
Lunch	+	0.233	0.065	0.622***	[0.27, 0.97]
New magazine	+	0.117	0.068	0.210	[−0.14, 0.56]
SLA response decision
Dodgeball	+	0.175	0.098	0.384*	[0.02, 0.75]
Math class	+	0.102	−0.015	1.001***	[0.49, 1.51]
New pants	+	0.120	0.018	0.756***	[0.39, 1.13]
Lunch	+	0.172	−0.033	0.782***	[0.46, 1.10]
New magazine	+	0.131	0.038	0.514*	[0.12, 0.90]
CCC Cognitive Concentration	+	4.181	1.299	3.341*	[0.50, 6.17]
CCC Authority Acceptance	+	0.907	0.550	1.152	[−0.73, 3.03]
CCC Authority Acceptance subscale: Overt Aggression	+	0.513	0.012	0.757*	[−0.10, 1.61]
CCC Authority Acceptance subscale: Oppositional	+	0.347	0.485	0.263	[−0.70, 1.23]
CCC Authority Acceptance subscale: Covert Antisocial	+	0.047	0.053	0.086	[−0.25, 0.42]
CCC Social Contact	+	2.318	0.749	1.458	[−0.36, 3.27]
CCC Social Competence	+	1.644	1.178	0.888	[−2.25, 4.02]
CCC Social Competence subscale: Prosocial	+	0.813	0.639	0.273	[−1.49, 2.03]
CCC Social Competence subscale: Emotional Regulation	+	0.831	0.538	0.608	[−0.91, 2.13]
CCC Relational Aggression	−	−0.653	−0.536	−0.803	[−2.59, 0.99]

Note. All p values come from one-tailed test. The estimated regression coefficients and their 95% confidence intervals were derived by the multilevel analysis with random effects and the ordered logistic regression. LBF = Let’s Be Friends; SLA = skill-level activity; CCC = Carolina Child Checklist.

*p < .05. **p < .01. ***p < .001.

The bivariate analysis shows the mean change scores for the treated and control groups on each outcome. This analysis does not control for covariates and does not use a multilevel model to correct for the clustering effects. The bivariate analysis suggests that of 30 change scores (i.e., change scores of four SLA skills under five story situations, and change scores of the 10 CCC scales), the LBF-treated children show positive and beneficial changes on 26 of them.

Results from the multivariate analysis are presented in the last two columns in Table 2 with the heading of “Estimated Regression Coefficient of LBF.” These are the major results testing treatment effectiveness that controlled for covariates. The estimated regression coefficients for SLA encoding and CCC scales can be interpreted as unstandardized effect sizes (i.e., differences on the outcome between treated and control groups). Results suggest that the LBF children had greater skill in encoding social and environmental cues than the control children. All these differences were statistically significant. Specifically, controlling for all other variables, the LBF children scored higher than the control children on the change of “dodgeball” encoding by 2.101 units (p < .001), on the change of “math class” encoding by 1.773 units (p < .001), on the change of “new pants” encoding by 1.971 units (p < .001), on the change of “lunch” encoding by 1.688 units (p < .001), and on the change of “new magazine” encoding by 0.913 units (p < .001). The program appears to enhance children’s encoding competence.

Of the 10 change scores from CCC, two showed beneficial and positive effects for the LBF intervention: controlling for covariates, the LBF children were rated higher than the control children by their teachers on “cognitive concentration” by 3.341 units (p < .05) and on the “authority acceptance subscale: overt aggression” by 0.757 units (p < .05). Teacher ratings suggest that the LBF intervention strengthens children’s cognitive concentration and reduces children’s overt aggression. The remaining eight change scores of CCC did not show statistically significant differences, but patterns of change suggest that children in the LBF condition were rated by teachers more positively.

To help interpret results generated by the ordered logistic regression, Tables 3 –5 present the model-predicted probabilities based on average treatment effects. These analyses focus on patterns. First, the predicted probabilities on change in hostile attribution based on the “lunch” vignette (Table 3) indicate that a typical child (i.e., labeled “all” in the table, meaning when all independent variables used in the model are fixed at the sample mean level) has 21.3% probability of change from “hostile attribution” in the pretest to “nonhostile attribution” in the posttest, or making an improved change; the typical child has 63.2% probability of remaining unchanged on being “hostile” between both tests; and the typical child has 15.5% probability of changing from “nonhostile attribution” in the pretest to “hostile attribution” in the posttest or making an undesirable and negative change. The interpretation of predicted probabilities for “all” bear the same meaning for other outcomes as those described above.

Table 3.

Differences of Predicted Probabilities on the Change Score of SLA Hostile Attribution Between LBF and Control Children.^a

SLA Scale & Scenario	Changed From Hostile Attribution at Pretest to Nonhostile Attribution at Posttest	Hostile Attribution at Both Pre- and Posttests	Changed From Nonhostile Attribution at Pretest to Hostile Attribution at Posttest
SLA hostile attribution
Lunch
All ^b	.213	.632	.155
LBF versus control	.116*	−.025	−.09**
New magazine
All ^b	.251	.614	.135
LBF versus control	.093*	−.035*	−.058*

Note. LBF = Let’s Be Friends; SLA = skill-level activity.

^a Predictions are based on the estimated coefficients of the ordered logistic regression model and are computed as the average marginal effects.

^b Average predicted probabilities of all sample participants.

*p < .05. **p < .01. ***p < .001—one-tailed test.

Table 4.

Differences of Predicted Probabilities on the Change Score of SLA Goal Formulation Between LBF and Control Children.^a

SLA Scale & Scenario	Changed From Nonaggressive at Pretest to Aggressive at Posttest	Nonaggressive at Both Pre- and Posttests	Changed From Aggressive at Pretest to Nonaggressive at Posttest
SLA goal formulation
Dodgeball
All ^b	.126	.579	.296
LBF versus control	−.057*	−.051*	.108*
New pants
All ^b	.035	.891	.073
LBF versus control	−.033**	−.032*	.065***
Lunch
All ^b	.071	.706	.223
LBF versus control	−.039**	−.064***	.103***

Note. LBF = Let’s Be Friends; SLA = skill-level activity.

^a Predictions are based on the estimated coefficients of the ordered logistic regression model and are computed as the average marginal effects.

^b Average predicted probabilities of all sample participants.

*p < .05. **p < .01. ***p < .001—one-tailed test.

Table 5.

Differences of Predicted Probabilities on the Change Score of SLA Response Decision Between LBF and Control Children.^a

SLA Scale & Scenario	Changed From Nonaggressive at Pretest to Aggressive at Posttest	Nonaggressive at Both Pre- and Posttests	Changed From Aggressive at Pretest to Nonaggressive at Posttest
SLA response decision
Dodgeball
All ^b	.060	.743	.196
LBF versus control	−.021*	−.037*	.059*
Math class
All ^b	.044	.868	.088
LBF versus control	−.039***	−.035***	.074***
New pants
All ^b	.049	.834	.117
LBF versus control	−.034***	−.042***	.076***
Lunch
All ^b	.138	.654	.208
LBF versus control	−.089***	−.034***	.123***
New magazine
All ^b	.063	.789	.148
LBF versus control	−.03**	−.033**	.063**

Note. LBF = Let’s Be Friends; SLA = skill-level activity.

^a Predictions are based on the estimated coefficients of the ordered logistic regression model and are computed as the average marginal effects.

^b Average predicted probabilities of all sample participants.

*p < .05. **p < .01. ***p < .001—one-tailed test.

Second, the pattern of results regarding change of hostile attribution in the “lunch” scenario (Table 3) indicates that the probability of improvement (i.e., the probability of changing from “hostile attribution” in the pretest to “nonhostile attribution” in the posttest) for the LBF children is 11.6 percentage points higher than that for the control children (p < .05), and the probability of worsening (i.e., the probability of changing from “nonhostile attribution” in the pretest to “hostile attribution” in the posttest) for the LBF children is 9 percentage points lower than that for the control children (p < .01). The change pattern for hostile attribution in the “new magazine” scenario (Table 3) indicates that the probability of improvement for the LBF children is 9.3 percentage points higher than that for the control children (p < .05), the probability of remaining unchanged for the LBF children is 3.5 percentage points lower than that for the control children (p < .05), and the probability of worsening for the LBF children is 5.8 percentage points lower than that for the control children (p < .01). These pattern analyses suggest that the LBF children benefited from intervention.

Third, Table 4 reports the predicted probabilities for the goal formulation scenarios in the SLA. The change in goal formulation skill from the “dodgeball” scenario indicates that the probability of worsening for the LBF children is 5.7 percentage points lower than that for the control children (p < .05), and the probability of improvement (i.e., the probability of changing from “aggressive” in the pretest to “nonaggressive” in the posttest) for the LBF children is 10.8 percentage points higher than that for the control children (p < .05). The change in goal formulation skill from the “new pants” scenario indicates that the probability of worsening for the LBF children is 3.3 percentage points lower than that for the control children (p < .05), and the probability of improvement for the LBF children is 6.5 percentage points higher than that for the control children (p < .05). The change in goal formulation from “lunch” scenario indicates that the probability of worsening for the LBF children is 3.9 percentage points lower than that for the control children (p < .05), and the probability of improvement for the LBF children is 10.3 percentage points higher than that for the control children (p < .05). Pattern analyses suggest that the LBF intervention is effective in strengthening the goal formulation skills of children.

Fourth, Table 5 reports the predicted probabilities for skill level changes regarding making response decisions. The change in response decision skill based on the “dodgeball” scenario indicates that the probability of worsening for the LBF children is 2.1 percentage points lower than that for the control children (p < .05), and the probability of improvement for the LBF children is 5.9 percentage points higher than that for the control children (p < .05). The change in response decision skill for the “math class” scenario indicates that the probability of worsening for the LBF children is 3.9 percentage points lower than that for the control children (p < .05), and the probability of improvement for the LBF children is 7.4 percentage points higher than that for the control children (p < .05). The change in response decision skills for the “new pants” scenario indicates that the probability of worsening for the LBF children is 3.4 percentage points lower than that for the control children (p < .05), and the probability of improvement for the LBF children is 7.6 percentage points higher than that for the control children (p < .05). The change in response decision skill for the “lunch” scenario indicates that the probability of worsening for the LBF children is 8.9 percentage points lower than that for the control children (p < .05), and the probability of improvement for the LBF children is 12.3 percentage points higher than that for the control children (p < .05). The change in response decision skill for the “new magazine” skill indicates that the probability of worsening for the LBF children is 3 percentage points lower than that for the control children (p < .05), and the probability of improvement for the LBF children is 6.3 percentage points higher than that for the control children (p < .05). On balance, pattern analyses suggest that the LBF intervention is effective in strengthening the response decision skills of children.

Overall, findings regarding the hypotheses were mixed. Bivariate, multivariate, and pattern-finding analyses from change scores on the SLA and the CCC support Hypothesis 1. That is, the LBF—Shaanxi intervention had beneficial and positive effects on children’s SIP skills. Not presented in tables, the analyses regarding change differences across all 30 outcome variables between children from single-child families and children with siblings produced no statistically significant differences. That is, on all SLA outcomes and the 10 CCC scales, no statistically significant differences between these two groups of children were observed. Hypothesis 2 cannot be supported.

Limitations

The findings are conditioned on important limitations. First, the results are based on a pretest–posttest, control group design. It is unknown whether the observed effects endure beyond the end of the intervention period. Second, one of the outcome measures—the CCC—was a teaching rating of child behavior. Teachers were not masked to the assignment condition of children. Indeed, teachers in the intervention condition observed the LBF trainers. Compared to teachers in the control condition, teachers in intervention schools may have been primed to expect changes in the behavior of their children. For this reason, the SLA is important. SLA findings were based on a direct test of children’s SIP skills (across multiple scenarios or vignettes) rather than the ratings of teachers or others such as the trainers. SLA findings suggest strongly that the SIP skills of children in the LBF condition improved significantly in comparison to children in control group schools. Third, although the two conditions were balanced on covariates and matched on eight school-level variables that could influence outcomes, it is possible that unmeasured variables affected the findings. That is, unobserved heterogeneity cannot be ruled out. Fourth, the LBF program focused on children, and it was delivered by trainers who were graduate students. The program did not attempt to change the behavior of teachers or parents, and it was not delivered by professional social workers, school counselors, or others who might bring more experience to bear in training sessions. To strengthen the effect and ensure that skills continue to be reinforced in the classroom and the home, a multielement intervention that is delivered by in vivo professional staff and that more directly involves teachers and parents warrants investigation. And finally, although measures from the Lesson Process Report were all high, the study’s fidelity needs to be further enhanced. The ICC check on all outcome variables suggests the existence of nonignorable differences on intervention skills among trainer groups. It is important to take additional measures in future research to ensure that the intervention is indeed implemented as designed. It is important to provide further training to trainers with low treatment skills.

Discussion and Applications to Practice

The LBF program is the first intervention designed to strengthen the SIP skills of Chinese children. From the Tianjin pilot test to the Shaanxi trial, results suggest that the program strengthens children’s SIP skills and reduces aggressive behavior while promoting cognitive concentration. Cognitive concentration—the ability to focus and to avoid disturbing others—is a behavioral predictor of academic achievement.

The intervention research based on scientific theories and methods is one of the most important and compelling tasks for social work research and practice, and this is particularly true in China where projects of intervention research remain to be new and rare. The most important implication for social work practice is that evidence-based intervention with rigorous methods is feasible in China, and social work practice could be cross-cultural that borrows successful experiences gained from other countries. This is important in today’s setting of globalization.

Conclusion

Using scientific methods with a sound theoretical framework, this study shows that SIP is an important factor affecting children’s development, and the SIP skills are malleable. Providing children with interventions at their key stages of socialization and development will exert important and beneficial effects on their life-course outcomes. Children growing up from the single-child households may not have more severe problems in behavior than children with siblings.

Footnotes

Acknowledgments

We thank Mary McKay, Weifeng Tuo, Wei Lu, Tiange Fan, Xiaoting Lv, Zhen Zhang, Liming Li, Yanjie Bian, Shun Zhang, Chuntian Lu, Xinya Xue, Na Li, Qi Chen, Yihua Fang, Xulei Ma, and Jiyue Li for their support to the project and superb research assistance. We also thank 29 graduate students of Xi’an Jiaotong University who served as trainers of the intervention and principals, teachers, and third-grade students from 27 elementary schools in Jingyang County, China, who participated in the project. We thank two anonymous reviewers for comments and suggestions for revision of the article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received the following financial support for the research, authorship, and/or publication of this article: This study was sponsored by China’s Rici Philanthropic Foundation, Yangtze-River Scholarship of Xi’an Jiaotong University, and Frank J Bruno Endowment Funds of Washington University in St. Louis.

ORCID iD

Shenyang Guo

References

Arsenio

W. F.

Adams

Gold

(2009). Social information processing, moral reasoning, and emotion attributions: Relations with adolescents’ reactive and proactive aggression. Child Development, 80(6), 1739–1755. https://doi.org/10.1111/j.1467-8624.2009.01365.x

Barnard

Frangakis

C. E.

Hill

J. L.

Rubin

D. B.

(2003). Principal stratification approach to broken randomized experiments. Journal of the American Statistical Association, 98(462), 299–323. https://doi.org/10.1198/016214503000071

Biao

(2007). How far are the left-behind left behind? A preliminary study in rural China. Population, Space and Place, 13(3), 179–191. https://doi.org/10.1002/psp.437

Carneiro

Crawford

Goodman

(2007). The impact of early cognitive and non-cognitive skills on later outcomes. http://eprints.lse.ac.uk/19375/1/The_Impact_of_Early_Cognitive_and_Non-Cognitive_Skills_on_Later_Outcomes.pdf

Chen

Lin

(2017). Suiqian zinv lixiang ziwobiaozheng, xinlirenxing dui jiudushiying de yingxiang [Migrant children’s self-expression of aspiration, resilience, and their impact on studying in new environment]. Zhongguo jiankang xinlixue zazhi, 25(5), 758–763. https://doi.org/10.13342/j.cnki.cjhp.2017.05.032

Chen

(2016). Lixing qingxuliaofa shijiao xia shehuigongzuo jieru qingshaonian xidu jiaozheng: Jiyu Guangzhoushi D jiedusuo de shiwuyanjiu [Social work intervention in youth and adolescents’ correction of substance abuse based on the perspective of rational-emotive therapy: A study of D Detoxification Center from Guangzhou]. Qingshaonian xuekan, 4, 39–42. https://doi.org/10.3969/j.issn.1673-8950.2016.04.009

Crawley

R. D.

Becan

J. E.

Knight

D. K.

Joe

G. W.

Flynn

P. M.

(2015). Predictors of physical altercation among adolescents in residential substance abuse treatment. Deviant Behavior, 36(12), 996–1018. https://doi.org/10.1080/01639625.2014.982780

Crick

N. R.

Dodge

K. A.

(1994). A review and reformulation of social information-processing mechanisms in children’s social adjustment. Psychological Bulletin, 115(1), 74–101. https://doi.org/10.1037/0033-2909.115.1.74

Denham

S. A.

Way

Kalb

S. C.

Warren-Khot

H. K.

Bassett

H. H.

(2013). Preschoolers’ social information processing and early school success: The challenging situations task. British Journal of Developmental Psychology, 31(2), 180–197. https://doi.org/10.1111/j.2044-835x.2012.02085.x

10.

Dodge

K. A.

(1980). Social cognition and children’s aggressive behavior. Child Development, 51(1), 162–170. https://doi.org/10.2307/1129603

11.

Dodge

K. A.

Laird

Lochman

J. E.

Zelli

(2002). Multidimensional latent-construct analysis of children’s social information processing patterns: Correlations with aggressive behavior problems. Psychological Assessment, 14(1), 60–73. https://doi.org/10.1037/1040-3590.14.1.60

12.

Duan

Zhou

(2005). Woguo liushouertong zhuangkuang yanjiu [A study on children left behind]. Population Research, 29(1), 29–36. https://doi.org/10.3969/j.issn.1000-6087.2005.01.004

13.

Dykas

M. J.

Cassidy

(2011). Attachment and the processing of social information across the life span: Theory and evidence. Psychological Bulletin, 137(1), 19–46. https://doi.org/10.1037/a0021367

14.

Espelage

D. L.

Rose

C. A.

Polanin

J. R.

(2015). Social-emotional learning program to reduce bullying, fighting, and victimization among middle school students with disabilities. Remedial and Special Education, 36(5), 299–311. https://doi.org/10.1177/0741932514564564

15.

Fan

(2005). Woguo qingshaonian zishayanjiu ji yufangduice [Research in China about youth’s suicidal ideation and preventive strategies]. Linchuang jingshen yixue zazhi, 15(4), 241–242. https://doi.org/10.3969/j.issn.1005-3220.2005.04.041

16.

Fang

(2006). Shehui renzhi xinxi jiagong moxing zai ertong gongjixing xingwei yanjiu zhong de yingyong [The application of the social information-processing model in the study of children’s aggressive behavior]. Dangdai qingnian yanjiu, 3, 23–25. https://doi.org/10.3969/j.issn.1006-1789.2006.03.005

17.

Fontaine

R. G.

(2006). Evaluative behavioral judgments and instrumental antisocial behaviors in children and adolescents. Clinical Psychology Review, 26(8), 956–967. https://doi.org/10.1016/j.cpr.2006.07.003

18.

Fraser

M. W.

Day

S. H.

Galinsky

M. J.

Hodges

V. G.

Smokowski

P. R.

(2004). Conduct problems and peer rejection in childhood: A randomized trial of the making choices and strong families programs. Research on Social Work Practice, 14(5), 313–324. https://doi.org/10.1177/1049731503257884

19.

Fraser

M. W.

Galinsky

M. J.

Smokowski

P. R.

Day

S. H.

Terzian

M. A.

Rose

R. A.

Guo

(2005). Social information-processing skills training to promote social competence and prevent aggressive behavior in the third grades. Journal of Consulting and Clinical Psychology, 73(6), 1045–1055. https://doi.org/10.1037/0022-006x.73.6.1045

20.

Fraser

M. W.

Lee

J. S.

Kupper

L. L.

Day

S. H.

(2011). A controlled trial of the making choices program: Six-month follow-up. Research on Social Work Practice, 21(2), 165–176. https://doi.org/10.1177/1049731510386626

21.

Fraser

M. W.

Nash

J. K.

Galinsky

M. J.

Darwin

K. M.

(2000). Making choices: Social problem-solving skills for children. National Association of Social Workers Press.

22.

Fraser

M. W.

Thompson

A. M.

Day

S. H.

Macy

R. J.

(2014). The making choices program: Impact of social-emotional skills training on the risk status of third graders. The Elementary School Journal, 114(3), 354–379. https://doi.org/10.1086/674055

23.

Guo

Fraser

M. W.

(2015). Propensity score analysis: Statistical methods and applications (2nd ed.). Sage.

24.

Heckman

Stixrud

Urzua

(2006). The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior. Journal of Labor Economics, 24(3), 411–482. http://doi.org/10.1086/504455

25.

Imbens

Wooldridge

(2009). Recent developments in the econometrics of program evaluation. Journal of Economic Literature, 47(1), 5–86. https://doi.org/10.1257/jel.47.1.5

26.

Jiang

Bai

X., A, B.

Liu

. (2016). Qingshaonian wentixing shejiaowangluo shiyong [Problematic internet use among youth and adolescents]. Xinli kexue jinzhan, 24(9), 1435–1447. https://doi.org/10.3724/SP.J.1042.2016.01435

27.

Jingyang County Bureau of Statistics. (2017, December 28). Jingyangxian guominjingji he shehuifazhan tongjigongbao. Jingyang County Government. https://www.snjingyang.gov.cn/gk/tjgb/53732.htm

28.

Kupersmidt

Stelter

(2018). Social information processing. In Bornstein

(Ed.), The SAGE encyclopedia of lifespan human development (pp. 2064–2066). Sage. http://doi.org/10.4135/9781506307633.n766

29.

Lemerise

Maulden

(2010). Emotions and social information processing: Implication for understanding aggressive (and nonaggressive) children. In Arsenio

W. F.

Lemerise

E. A.

(Eds.), Emotions, aggression, and morality in children: Bridging development and psychopathology (1st ed., pp. 157–176). American Psychological Association.

30.

Let’s Be Friends Research Group. (2011). Let’s be friends: The interpersonal skills training manual for children. China Population Press.

31.

Song

Chang

Wang

(2016). Jiji xinlixue shijiao xia yufang qingshaonian yiyu de yanjiujinzhan [Research findings of prevention of youth’s depression based on the positive psychological perspective]. Yixue yu zhexue, 37(2), 37–40. https://doi.org/10.12014/j.issn.1002-0772.2016.02a.10

32.

Liu

(2015). Daodeng tuituo yu qingshaonian gongjixingwei de guanxi ji qishi [Moral excuse and its relation with youth’s aggressive behavior]. Xue lilun, 36, 53–54. https://doi.org/10.3969/j.issn.1002-2589.2015.36.025

33.

Liu

(1999). Characteristics of education in one-children family and effects on social nature development of children. Journal of Nanjing Normal University (Social Science Edition), 5, 64–69. http://en.cnki.com.cn/Article_en/CJFDTotal-NJSS199905010.htm

34.

Liu

Zhang

Cai

(2016). Qingshaonian xuesheng jiaolv yiyu qingxu tezheng yu yanjiuxianzhuang [Review of the current research on characteristics of youth’s anxiety and depression]. Zhongguo yiyao daobao, 13(12), 53–56. https://kns-cnki-net.web.bisu.edu.cn/KXReader/Detail?TIMESTAMP=637189821864615000&DBCODE=CJFQ&TABLEName=CJFDLAST2016&FileName=YYCY201612013&RESULT=1&SIGN=zc12fzBtZnLsrzLJeiDZCx4ajks%3d

35.

Luo

(2010). Xiaoyuan baoli chengyin zhi lilun shuping yu qishi [A theoretical review of the formation of bullying]. Zhangjiakou zhiye jishu xueyuan xuebao, 23(3), 32–34. https://doi.org/10.3969/j.issn.1008-8156.2010.03.012

36.

Macgowan

M. J.

Nash

J. K.

Fraser

M. W.

(2002). The Carolina Child Checklist of risk and protective factors for aggression. Research on Social Work Practice, 12(2), 253–276. https://doi.org/10.1177/104973150201200204

37.

Nas

C. N.

Orobio De Castro

Koops

(2005). Social information processing in delinquent adolescents. Psychology, Crime & Law, 11(4), 363–375. https://doi.org/10.1080/10683160500255307

38.

Nickoletti

Taussig

H. N.

(2006). Outcome expectancies and risk behaviors in maltreated adolescents. Journal of Research on Adolescence, 16(2), 217–228. https://doi.org/10.1111/j.1532-7795.2006.00129.x

39.

Nielsen

Nyland

Smyth

Zhang

Zhu

C. J.

(2005). Which rural migrants receive social insurance in Chinese cities? Global Social Policy: An Interdisciplinary Journal of Public Policy and Social Development, 5(3), 353–381. https://doi.org/10.1177/1468018105057416

40.

Ogle

R. L.

Miller

W. R.

(2004). The effects of alcohol intoxication and gender on the social information processing of hostile provocations involving male and female provocateurs. Journal of Studies on Alcohol, 65(1), 54–62. https://doi.org/10.15288/jsa.2004.65.54

41.

Pössel

Seemann

Ahrens

Hautzinger

(2006). Testing the causal mediation component of Dodge’s social information processing model of social competence and depression. Journal of Youth and Adolescence, 35(5), 842–851. https://doi.org/10.1007/s10964-006-9089-7

42.

Rubin

D. B.

(1986). Statistics and causal inference: Comment: Which ifs have causal answers. Journal of the American Statistical Association, 81(396), 961–962. https://doi.org/10.2307/2289065

43.

Sayette

M. A.

Wilson

G. T.

Elias

M. J.

(1993). Alcohol and aggression: A social information processing analysis. Journal of Studies on Alcohol, 54(4), 399–407. https://doi.org/10.15288/jsa.1993.54.399

44.

Smith

D. G.

Xiao

Bechara

(2012). Decision making in children and adolescents: Impaired Iowa Gambling Task performance in early adolescence. Developmental Psychology, 48(4), 1180–1187. https://doi.org/10.1037/a0026342

45.

Terzian

M. A.

Fraser

M. W.

Day

S. H.

Rose

R. A.

(2014). Social information-processing skills and aggression. Research on Social Work Practice, 25(3), 358–369. https://doi.org/10.1177/1049731514534898

46.

Venta

Mellick

Schatte

Sharp

(2014). Preliminary evidence that thoughts of thwarted belongingness mediate the relations between level of attachment insecurity and depression and suicide-related thoughts in inpatient adolescents. Journal of Social and Clinical Psychology, 33(5), 428–447. https://doi.org/10.1521/jscp.2014.33.5.428

47.

Fraser

M. W.

Guo

Day

S. H.

Galinsky

M. J.

(2016). Strengthening the social information–processing skills of children. Research on Social Work Practice, 26(5), 525–538. https://doi.org/10.1177/1049731514556995

48.

Yao

(2015). Taiwan qingshaonian zishafangzhi zhi jierucelue yanjiu [Studies on prevention and intervention strategies of youth’s suicide in Taiwan]. Qingnian tansuo, 6, 41–49. https://doi.org/10.13583/j.cnki.issn1004-3780.2015.06.006

49.

(1986). Changes in family structure in China: A simulation study. Population and Development Review, 12(4), 675–703. https://doi.org/10.2307/1973431

50.

Zhang

(2018). Shehui qingxu xuexi dui ertong qingshaonian qingxu zhili de yingxiang tanxi: Jiyu xinlijiankang jiaoyu shijiao [The impact of social emotional control training on adolescents’ emotional quotient: A study based on the perspective of psychological health education]. Journal of Kaifeng Institute of Education, 38(3), 179–180. https://doi.org/10.3969/j.issn.1008-9640.2018.03.082

51.

Zhang

(2000). Transforming China: Economic reform and its political implications. Palgrave Macmillan.

52.

Zhao

(2010). Nongcun liushouertong xinlijiankang zhuangkuang ji jiaoyu duice [Mental health status and educational measures of rural left-behind children]. Shoudu shifandaxue xuebao (Shehui kexue ban), 3, 128–130. https://doi.org/10.3969/j.issn.1004-9142.2010.03.022

53.

Zhong

Zhang

(2015). Shehui xinxi jiagong zai ertong qingxu tiaojie yu gongjixingwei jian de zhongjiexiaoying jianyan [A test of the mediating effect of social information-processing in children’s emotional regulation and aggressive behavior]. Zhongguo linchuang xinlixue zazhi, 1, 108–114. https://doi.org/10.16128/j.cnki.1005-3611.2015.01.025