Students, Classrooms, Teachers, and Schools: Competing Effects on Science Achievement
by Xin Ma, Xian Wu, Jing Yuan & Xingkai Luo - 2018
Background/Context: Students, classrooms, teachers, and schools form hierarchical units of any schooling system that compete for, say, educational resources and sometimes blame one another for the failure to meet a certain societal goal. Little educational research exists that separates competing effects on schooling outcomes across these educational levels.
Purpose/Objective/Research Question/Focus of Study: We aimed to fill in this gap in the literature by simultaneously examining in one model students, classrooms, teachers, and schools for competing effects on science achievement and identifying factors at each of these educational levels critical to science achievement.
Population/Participants/Subjects: We used data from the Program for Regional Assessment and Promotion of Basic Education Quality, a program in China. Data described learning and teaching practices of eighth-grade students and their science teachers (physics, biology, and geology) and school principals. Research Design: A four-level hierarchical linear model was developed to perform secondary analysis of science achievement data with students nested within classrooms, nested within teachers, nested within schools.
Findings/Results: Across physics, biology, and geology, the proportion of variance in science achievement attributable to students was 75%78%, 7%9% to classrooms, 1%4% to teachers, and 10%14% to schools. Across the three content areas of science, male students, younger students, and students with higher parental socioeconomic status outperformed their counterparts respectively (at the student level), and students in schools with higher student expenditure outperformed students in other schools (at the school level). Teacher effects were diverse across the three content areas of science.
Conclusions/Recommendations: The breakdown of institutional contributions to science achievement across the three content areas of science indicates a hierarchy of importance of educational units in the order of schools, classrooms, and teachers in China. From a global perspective, in China, gender differences were stronger, socioeconomic differences were weaker, age effects deviated from the international tradition, and school effects associated with student expenditure were stronger. We conclude that preparation and development are as critical for school administrators as they are for classroom teachers and that professional development organized by science content area may be more effective than pooling all science teachers together.
A rapidly changing society in the 21st century demands complex knowledge and skills of its citizens for national and personal interest (Jerald, 2009). Schools are one of the most critical social institutions for preparing students to meet this need, with school science in the center of the effort. According to Next Generation Science Standards Lead States (2013), Science is at the heart of this countrys ability to continue to innovate, lead, and create the jobs of the future. All studentswhether they become technicians in a hospital, workers in a high-tech manufacturing facility, or Ph.D. researchersmust have a solid K-12 science education (p. xiii). Before they take actions for educational change, stakeholders, policy makers, and practitioners need evidence about how our schooling system is producing schooling outcomes (e.g., science achievement). Students, classrooms, teachers, and schools are spontaneous components of any schooling system. They form hierarchical units. Students are nested within classrooms, classrooms within teachers, and teachers within schools. There are educational issues at each level. Individual differences (among students), instructional practices (among classrooms), experiences and abilities (among teachers), and context and climate (among schools) are a simplified summary. To improve educational effectiveness, it is essential to investigate how the educational units simultaneously influence schooling outcomes. These educational units do compete for, say, educational resources and blame one another for the failure to meet a certain societal goal. For any educational issue (e.g., science achievement), separating competing effects across educational levels and identifying salient factors at each level constitute an important research agenda. Very little educational research exists for these purposes; one of the reasons is a lack of empirical data that describe students, classrooms, teachers, and schools simultaneously.
The opportunity to simultaneously examine in one model students, classrooms, teachers, and schools for competing effects on academic achievement is rare and valuable. Reynolds et al. (2014) alluded to the need to link multiple levels of an educational system in order to develop complex theories on school effects. We aimed to fill in this gap in the literature by selecting educational levels sensitive to the social contexts and examining the context variables to generate more sophisticated explanations for the process variables responsible for school effects (see Reynolds et al., 2014). Our research statement was to separate competing effects on science achievement across educational levels and identify factors at each level critical to science achievement. We hypothesized that (1) students, classrooms, teachers, and schools are responsible, in this order, for science achievement, and (2) there are factors associated with the way that each educational level is responsible for science achievement.
OVERARCHING THEORETICAL FRAMEWORK FOR SCHOOL EFFECTS
The input-process-output (IPO) model is a popular overarching theoretical framework to guide the selection of variables and specification of statistical models (X. Ma, Ma, & Bradley, 2008, p. 62). Good and Brophy (1986) appreciated the IPO model for measuring the behavioral and perceptional processes that function inside the black box of schools. In the IPO model, students are the input (what students bring into their schools, such as family social values), and schools process students through educational policies and practices into different categories of schooling outcomes (the output). Although X. Ma et al. (2008) focused on schools as the process, their classification of the process variables as context and climate applies well to classrooms as another process. Contextual variables are the hardware (e.g., educational resources and characteristics of student and teacher bodies), and climatic variables are the software (e.g., academic expectation, classroom practice, and instructional leadership). X. Ma et al. (2008) emphasized the estimation of climatic effects based on the adjustment of contextual effects because climate is more likely under the control of administrators, teachers, and parents.
The IPO model can be operationalized in the analytical framework of hierarchical linear modeling (HLM) that controls background characteristics, examines the distribution of schooling outcomes across educational levels, and identifies salient contextual and climatic characteristics that process students into different categories of schooling outcomes. We adopted the IPO model because of its attention to process as the connection between input and output, its emphasis on context and climate that can bear meaningful policy implications, and its match with HLM that adds tremendous analytical power to the theoretical underpinning (see X. Ma et al., 2008). With the IPO model, we arranged our literature review by educational levels and variables that we applied (in italic) at each level to examine science achievement.
Student effects are individual differences, the input of the IPO model, that provide control of many critical student characteristics that may be confounded with classroom, teacher, and school effects (see X. Ma et al., 2008). There are gender differences in science achievement in many countries (Bedard & Cho, 2010; X. Ma, 2008; Martin, Mullis, Foy, Olson, et al., 2008). Before the year 2000, gender differences were in favor of boys and intensified as students progressed through schooling; however, these differences began to favor girls after 2000 (e.g., Bursal, 2013; Ziegler & Heller, 1997). Age predicts the rate of growth in science achievement (X. Ma & Wilkins, 2002). At the same grade, younger students outperform older students in science achievement in 14 European countries (Hampden-Thompson & Pong. 2005). In the United States, 15-year-olds attending higher grades (that is, higher than eighth grade, the normal grade for that age) fall behind in science achievement (Dalton, 2012). The effects of socioeconomic status (SES) on academic achievement are consistent (e.g., Martin, Mullis, Foy, & Stanco, 2012; National Center for Education Statistics, 2011, 2012). SES is a powerful factor influencing science achievement, explaining 14% of the variance (Organisation for Economic Co-operation and Development [OECD], 2007). Family structure refers to the makeup of a family. Living in a single-parent family is associated with lower science achievement than living in a two-parent family even after family resources are controlled (Hampden-Thompson & Pong, 2005). This achievement gap in science is greater in countries where single-parent families are prevalent (Pong, Dronkers, & Hampden-Thompson, 2003).
Individual differences associated with gender, SES, age, and single parent in academic achievement also manifest in Chinese education (e.g., Lai, 2010; Liu & Lu, 2008; Zhao, Valcke, Desoete, & Verhaeghe, 2012). Some unique characteristics of Chinese modern society have dramatically shaped individual differences in academic achievement. The decades-long one-child policy (loosened recently) is one of them. Falbo and Poston (1993) found that students without siblings (as a result of the one-child policy) outperform students with siblings in academic achievement (see also Lu, Zhang, & Xu, 1999). Massive internal migration (from rural to urban) has resulted in a critical social issue of separation of parents and children. Children show slower cognitive development if both parents migrate, and the negative impact is weaker if one parent migrates (H. Zhang, Behrman, Fan, Wei, & Zhang, 2014). Parents who migrate for a longer time have children with poorer academic achievement (M. Zhou, Murphy, & Tao, 2014).
Classroom effects are often considered as embedded in teacher effects. We believe that their integration is conceptually reasonable, but their separation is analytically informative. A few studies have employed classroom as the unit of analysis to examine classroom effects on academic achievement and claimed that classrooms are the most important educational factor that shapes academic achievement (e.g., Drent, Meelissen, & Van Der Kleij, 2013; Kaya & Rice, 2010). Physical environment of a classroom (e.g., adequately ventilated and thermal) improves science test scores (Haverinen-Shaughnessy & Shaughnessy, 2015). A randomized experiment on class size reduction in Tennessee upheld a significant long-term benefit of small class size on academic achievement (see Bosworth, 2014; Nye, Hedges, & Konstantopoulos, 2004). Peer effects in a classroom setting play an important role in motivating students to improve their academic achievement (Carmen & Zhang, 2008).
Two unique characteristics of Chinese classrooms further highlight the critical role of classrooms. Class size of a typical Chinese classroom is very large (4050), and Chinese students stay with the same classmates in the same classrooms for all school subjects and across all grades (especially in elementary and middle school). These unique characteristics do not seem to subject Chinese students to disadvantage in academic achievement, based on the Programme for International Student Assessment (PISA) and the Trends in International Mathematics and Science Study (TIMSS) (e.g., Martin, Mullis, Foy, Olson, et al., 2008; OECD, 2014).
In addition to teachers professional behavior within the classroom, the composition of the teaching force in terms of, e.g. age and educational level, their initial education and qualification, their individual beliefs and competencies, as well as professional practices on the school levelsuch as collaboration and professional developmenthave been core topics in educational policy. (OECD, 2015a, pp. 2627)
As agents for educational change, teachers are responsible directly for academic achievement (e.g., Seidel & Shavelson, 2007; Taylor et al., 2015; Yildirim & Demir, 2014). Increasing teacher effects is a sound strategy, especially in a budget-tight environment; the effects of a costly 10-student reduction in class size on academic achievement are smaller than the benefit of improving teacher quality by one standard deviation (which teacher education and professional development can readily achieve) (Rivkin, Hanushek, & Kain, 2005; see also Darling-Hammond, 2000).
Students learn more in mathematics from teachers with coursework, degrees, or certification in mathematics (Wayne & Youngs, 2003). Students whose teachers majored or minored in science outperform their peers by 40% of a grade level in science (Wenglinsky, 2000). High school teachers with an advanced degree in science have positive effects on science achievement (Rice, 2003). Science achievement increases in parallel to an increase in teacher science teaching experience (Atar & Atar, 2012). Across countries, the eighth graders whose teachers had 10 or more years of teaching experience show better science achievement than those whose teachers have less than 10 years of teaching experience (Martin et al., 2012). Chu et al. (2015) reported that having a teacher with the highest professional rank (title) is positively related to academic achievement of students (see also W. Zhang, Xin, & Kang, 2010).
Certain academic (scientific) behaviors of teachers (e.g., considering different solutions when solving a problem) improve science achievement. Wenglinsky (2000) highlighted science laboratory skills, and Yang (2012) emphasized critical thinking behaviors. A related issue is teachers professional attitude about teaching. The basis for teacher effects is teachers professional motivation (Kleickmann, Tröbst, Jonen, Vehmeyer, & Möller, 2016; Taylor et al., 2015). Professional optimism and enthusiasm improve academic achievement (Chang, 2011; Kunter et al., 2013). Teachers self-efficacy in and beliefs about science teaching are predictors of science achievement (Lumpe, Czerniak, Haney, & Beltyukova, 2012). Teacher professional development (growth) generates positive effects on science achievement by improving teachers beliefs about and skills in science education (e.g., Taylor et al., 2015). Participation in long-term and intense professional development predicts science achievement (Lumpe et al., 2012). Professional development that engages teachers in research-based hands-on instructional practice of science promotes science achievement (Silverstein, Dubner, Miller, Glied, & Loike, 2009). Teacher leaders foster positive change in their schools, improving academic achievement and organizational capacity (Nathan, 2015). Teacher leadership style functions as a key factor affecting academic achievement (Yildirim, Acar, Bull, & Sevinc, 2008). Teachers participation in distributed leadership at least indirectly affects academic achievement (Chang, 2011). Whipple, Evans, Barry, and Maxwell (2010) identified professional turnover (teacher mobility) as one of the major educational risk factors associated with academic achievement (see also A. Miller, 2013).
Common characteristics of general classroom practice are associated with academic achievement (e.g., Johnson, Kahle, & Fargo, 2007). Kyriakides, Christoforou, and Charalambous (2013) highlighted the dynamic model of eight factors that describe teachers general classroom practice (orientation, structuring, questioning, modeling, application, management, environment, and assessment). Academic achievement increases when teachers classroom practice moves from passive to active to constructive to interactive (Menekse, Stump, Krause, & Chi, 2013). Unique characteristics of science classroom practice are associated with science achievement (e.g., Jiang & McComas, 2015), such as inquiry-based heuristic science writing (Akkus, Gunel, & Hand, 2007) and hands-on science learning activities (Wenglinsky, 2000). Compared with teacher-centered instruction, the laboratory inquiry approach has positive effects on, and closes the gender gap in, science achievement, and the critical thinking approach reduces the minority disadvantage in science achievement (Von Secker & Lissitz, 1999).
School culture, or the environment in which teachers work (e.g., colleagues care for one another), matters to science achievement (X. Ma & Wilkins, 2002). The positive association is stronger in developed than developing countries (Mohammadpour, Shekarchizadeh, & Kalantarrashidi, 2015). Science achievement increases when principals and teachers have a positive view of school culture (Martin, Mullis, & Foy, 2008). One critical element of school culture, perception by teachers of principal leadership, is often singled out for examination. Science achievement is related to teachers perception of their principals trustworthiness (Smetana, Wenner, Settlage, & McCoach, 2016). Teacher job satisfaction is related to science achievement both directly (Yildirim & Demir, 2014), and indirectly through promoting teachers retention (Perrachione, Rosser, & Peterson, 2008), sense of belonging (Skaalvik & Skaalvik, 2011), engagement (Klassen et al., 2012), and self-efficacy (Caprara, Barbaranelli, Steca, & Malone, 2006). Emotionally exhausted teachers have a low class average in academic achievement (Arens & Morin, 2016). Teachers sense of belonging (to school) makes a difference in mathematics achievement (X. Ma, 2008). Building new teachers sense of belonging retains them and improves their classroom practice (Allen, 2009). Teachers sense of belonging is developmental and needs to be developed as early as student teaching (Ussher, 2010).
A unique Chinese teacher effect concerns homeroom teachers; students stay in the same homeroom for all subjects and across all grades. B. Zhang and Jiang (2009) found that a positive relationship between a homeroom teacher and students improves academic achievement (see also P. Zhou & Song, 2004). The positive effects of homeroom teachers on academic achievement extend from traditional to vocational schools (S. Zhang, 2006). With 13 million teachers (National Bureau of Statistics of China, 2014), China has one of the largest teacher labor forces in the world (Chu et al., 2015). The Chinese society demands teacher effects because competition for higher education and the labor market is stunningly strong. Throughout the years of reform, teachers have had to stretch their professional capacity in order to satisfy competing demands engendered by reform measures and educational reality (Lo, Lai, & Wang, 2013, p. 239). China has implemented a series of intensive National Plans for mandatory teacher training to increase teacher effects (Chinese Ministry of Education, 2010).
Although reanalyses of the pioneering Coleman data (Coleman et al., 1966) have largely replicated their resultsthat school variables have small effects on academic achievement as compared with family background (e.g., Konstantopoulos & Borman, 2011)large-scale studies of science achievement have revealed greater school-level variations and, therefore, greater school effects (Young, Reynolds, & Walberg, 1996, p. 277). School effects are more important in accounting for variance in science achievement than gender differences (Young & Fraser, 1994). Schools contribute to 21% of the variation in science achievement, and school predictors explain 65% of the school contribution (Konstantopoulos, 2006). Raudenbush and Willms (1995) clarified school effects as Type A (effects of school practice and contextual influences of social and economic factors outside the control of a school) and Type B (effects of school practice controlling for contextual influences), with the latter more objectively measuring school effects.
School contextual variables affect academic achievement (e.g., Uline & Tschannen-Moran, 2008). School (enrollment) size has stronger negative effects on academic achievement in schools with a higher concentration of socially disadvantaged students (V. Lee & Smith, 1997). Even when school size does not show effects on academic achievement (Barton, 2015), it can still have a strong impact on the social distribution of academic achievement (V. Lee & Bryk, 1989). Given our earlier discussion on gender differences and teacher education, the importance of percentage of girls and percentage of teachers with at least a bachelors degree to science achievement is apparent. One of the pivotal causes of inadequate academic achievement is shortage of qualified teachers, especially in mathematics and science (Ingersoll & Perda, 2009). Many teachers intend to abandon a science teaching career (e.g., 32% with 13 years of experience and 37% with 1015 years of experience; Mangrubang, 2005). Traditionally, teacher quality is simply a different expression of teaching experience and teacher education (see Zuelke, 2008). Principals often evaluate teacher quality as an aggregation of multiple variables beyond simple addition of experience and education. This makes sense in that a large age gap between teachers and students increases the difficulty of intergenerational communication and understanding about science (Wößmann, 2003). Monk (1994) also noticed that the amount of undergraduate coursework in life sciences does not matter to science achievement.
Higher educational expenditure per student is related to inferior mathematics and science achievement (Wößmann, 2003). Expenditure per student shows zero or negative correlation with science achievement in two U.S. states (Barton, 2015; Lamons, 2009). Clearly, there comes a point when the money spent on education stops being correlated with student achievement and success as much as it is initially (Anderson, 2011, p. 28). Martin et al. (2012) found that, across countries, students in schools without shortage in science resources (e.g., instructional materials) have higher science achievement than those in schools with partial or severe shortage (see also Wößmann, 2003). Chiu (2007) reasoned that the availability of public educational resources (e.g., books) enhances the value of tangible family resources for learning (e.g., cultural communication) so as to improve science achievement. Principals become effective in promoting academic achievement as they gain working experience (Clark, Martorell, & Rockoff, 2009; Louis, Dretzke, & Wahlstrom, 2010). Novice leaders often lack the time for adequate exposure to science; awareness of curricular innovations, pedagogical strategies, and equipment needs in teaching science; and experience in supervising science teachers (Barish, 2008).
School climatic variables related to the process of the IPO model usually attract more attention (e.g., Onocha & Okpala, 2001). There is an overall structural relationship among school climate, participation in learning, and science achievement (Chen, Lin, Wang, Lin, & Kao, 2012). Our data contained important variables related to the work of principals for addressing the fact that the literature on principals is sparse (Clark et al., 2009, p. 1). Principals demonstrate educational innovation leadership when they encourage teachers to engage students in debate on competing evidence-based ideas rather than simply accepting facts (Moorosi & Bush, 2011). Principals demonstrate democratic government leadership when they promote critically minded citizens who can analyze and challenge social structures with the knowledge of the natural world and the process of gaining that knowledge (Longbottom & Butler, 1999). Although doubt exists regarding whether principals school management ability can be a strong determinant of academic achievement (e.g., R. Miller & Rowan, 2006; Ogundokun, 2012), principal support for teaching is important for novice teachers to enact reform-based instruction (McGinnis, Parker, & Graeber, 2004), and principal support for professional development promotes teachers professional commitment, inquiry-based teaching practice, and investigative classroom culture (Singh & Billingsley, 1998; Supovitz & Turner, 2000). Academic achievement is higher when there is more school autonomy in choosing textbooks, hiring teachers, and allocating within-school budgets (Fuchs & Wößmann, 2007). School autonomy benefits academic achievement more than examinations for accountability do (Wößmann, Lüdemann, Schütz, & West, 2007). Researchers have found positive associations between extracurricular participation and academic achievement (Guest & Schneider, 2003, p. 89). Although boys more actively participate in extracurricular activities in science than girls (Breakwell & Beardsell, 1992), girls who do participate have higher science achievement (Chambers & Schreiber, 2004).
Among Chinese school effects, there may be schools in a certain region competing for competent teachers, and particularly students (i.e., competing schools). Public schools in U.S. states with large private school sectors often improve academic achievement because of school competition (Arum, 1996). Chinese research on school effects began with J. Sun and Hong (1994), who introduced the Western literature. H. Wang and Chen (2010) found only 12 empirical studies on school effects from 1994 to 2009 in China. These studies revealed the importance of school process variables to academic outcomes. The lack of empirical studies on effective schools in one of the largest education systems in the world is unfortunate in that effective Chinese schools may offer considerable insights for educational policy and practice in the developing world.
CURRENT CHINESE SCIENCE EDUCATION
As the second largest economy in the world, China is increasingly emphasizing science and technology innovation. Chinese science education can be uniquely characterized as a strong emphasis on coherent, rigorous, and in-depth curriculum; a strong focus on teacher-centered instruction, logical reasoning, and theoretical thinking; and a heavy course load centered on the highly competitive college entrance examination (Su, Goldstein, & Su, 1995; W. Wang, Wang, Zhang, Lang, & Mayer, 1996). Elementary school students study a form of integrated science; middle school students study either an integrated science or separated science curricula (physics, chemistry, biology, and geology), with the decision made locally; and high school students study separated science curricula. Each science curriculum evolves across grades and contains both compulsory and optional chapters at each grade. Classroom practice is increasingly encouraged to take on a student-centered approach rather than a teacher-centered or textbook-centered approach, with an emphasis on the adoption of curricular materials relevant to the daily life of students, the experience-based learning approach, and the constructive interaction between teachers and learners (Huang, 2004; J. Wang, 2012). Scientific inquiry becomes the major trademark of science education, in parallel to scientific content. Under the curricular reform, Chinese students from Shanghai scored the highest average in science among 65 participating education systems in PISA 2012 (OECD, 2014).
Data for the present study came from the 2014 Program for Regional Assessment and Promotion of Basic Education Quality (PRAPBEQ) (see Chinese National Innovation Center for Assessment of Basic Education Quality, 2015). This national assessment, which is curriculum based and competence oriented, collects data from regional education systems once a year and covers Chinese, mathematics, science, and English (as a foreign language). Student, teacher, and principal questionnaires are also administered. PRAPBEQ was initiated in 2003 in response to the emerging demand for monitoring and improving regional education quality and equity under the Chinese New Curriculum Reform. PRAPBEQ 2014 included a special sample of eighth graders from the city of Zhengzhou, located in the central region of China.1 All schools with eighth graders in that city were included in the assessment. In each school, all eighth graders and their teachers in Chinese, mathematics, and science participated in data collection, together with their principals. The sample therefore naturally inherited the educational hierarchy, with students nested within classrooms, nested within teachers, nested within schools.2 Sample sizes at various levels across different content areas of science are provided in Appendix A.
We obtained academic data from the science assessment and nonacademic data from the student, teacher, and principal questionnaires. PRAPBEQ employs the term science literacy for scientific knowledge and skills with three dimensions: nature and knowledge of science, competence of scientific inquiry, and attitude toward science. PRAPBEQ science achievement concerns the first two and is measured by a standardized test with 36 items (reliability = .83). The test, administered to all students (i.e., no matrix sampling or plausible values), is aligned with the curriculum. Items address cognition (knowing, understanding, and applying) and skills of scientific inquiry (questioning, seeking evidence, and explaining). We used science achievement in the subtests of physics, biology, and geology as our dependent (outcome) measures.
The physics test covers structure of substance, force and motion, and energy and energy resources. For example, students are asked to design a scientific experiment to test which material is the best for heat preservation. To perform well, students need to understand basic concepts of physics, apply the strategy of control (variables) for scientific experiments, and build proper models of physics to solve real-world problems. The biology test covers multilevel biological systems (e.g., cell, organ, and organism), energy transformation among organisms, evolution, and relationships among human activities, health, and the environment. For example, students are asked to list the essential factors for plants to grow. To perform well, students need to understand basic biological concepts as they manifest in their daily lives and apply scientific reasoning to develop evidence-based biological explanations. The geology test is about earth science and covers the position of the Earth in the solar system, crustal movement, water systems on the Earth, and weather and climate. For example, students are asked to illustrate the effects of climate on human life. To perform well, students need to have basic knowledge about the Earth and understand problems and challenges that modern human society faces (e.g., climate change).
Independent (predictor) variables came from multiple levels of students, teachers, and schools. At the student level, there were nine variables descriptive of individual (gender and age) and family characteristics (father SES, mother SES, single-parent household, one child, rural migrant child, and parental migration status with two dummy variables; see Appendix B).3 Some student-level variables are unique to Chinese education. The one-child policy was in effect in 2014, but it was not uncommon to find families with more than one child for various reasons. The rapid economic development in China has motivated millions of farmers to leave rural regions for better economic opportunities in cities, and parents who migrate to other places often have to leave their children in the care of one parent or even relatives (i.e., no parents). The variables of rural migrant child and parental migration status aimed to capture this phenomenon.
Although the Zhengzhou sample allows for the identification of classroom as an analytical unit, there were no variables descriptive of classrooms. Classroom functioned as an analytical level for the partition of variance in science achievement. At the teacher level, there were 17 variables descriptive of teacher background (teacher education with two dummy variables, teacher experience, teacher professional title, and homeroom teacher status), teacher behavior (academic behavior, professional attitude, professional growth, teacher leadership, and teacher mobility with two dummy variables), classroom practice (general classroom practice and science classroom practice), and working environment (school culture, perceived principal leadership, job satisfaction, and sense of belonging).4 Many teacher-level variables were composite variables constructed from multiple items (see Appendix B for variable descriptions, coding information, and reliability estimates).5
At the school level, there were 17 variables descriptive of school context and school climate. Contextual variables included (existence of other) schools competing for students, school (enrollment) size, percentage of girls, school mean SES, percentage of teachers with at least a bachelors degree, teacher shortage, teacher quality, per-student expenditure, educational resources, and principal working experience.6 Climatic variables from PRAPBEQ 2014 focused heavily on principal leadership, including educational innovation, democratic government, school management, support for teaching, and support for professional development. School autonomy and extracurricular activities were additional climatic variables. Many school-level variables, especially climatic variables, were composite variables constructed from multiple items (see Appendix B for variable descriptions, coding information, and reliability estimates).
For composite variables constructed from multiple items, we reversed response options for negatively worded items so that a higher value indicates a more positive response across all variables (e.g., see school culture in Appendix B). Response options were on a 5-point Likert type scale. Following the common statistical practice, we coded 1 = fully disagree, 2 = disagree, 3 = not sure, 4 = agree, and 5 = fully agree (i.e., treating a Likert scale as an interval scale).
We reiterate the two goals of the present study as separating competing effects on science achievement across students, classrooms, teachers, and schools and identifying salient factors at each level critical to science achievement. These goals were accomplished analytically through a four-level hierarchical linear model (HLM). The data hierarchy is students (at level 1) nested within classrooms (at level 2), nested within teachers (at level 3), nested within schools (at level 4).7 To separate competing effects on science achievement across students, classrooms, teachers, and schools, what is often referred to as a null model was developed. The dependent variable is the only variable in the model. A null model is the chief way to partition the variance in the dependent variable (Raudenbush & Bryk, 2002). We performed three separate analyses of a null model, one for each content area of science (physics, biology, and geology). Proportion of variance attributable to each educational level was calculated to address the first research goal. To identify salient factors at each level critical to science achievement, what is often referred to as a full model was developed. Independent variables are present at each level to predict the dependent variable. A full model is the chief way to identify statistically significant predictors of the dependent variable (Raudenbush & Bryk, 2002). We applied nine variables at the student level and 17 variables each at the teacher and school levels.8 All variables were employed for data analysis without centering. Again, we performed three separate analyses of a full model, one for each content area of science. Statistically significant student-level, teacher-level, and school-level variables were identified collectively as salient predictors of science achievement.
We took advantage of the rare opportunity of having four levels of an educational system simultaneously to partition variance in science achievement into student, classroom, teacher, and school levels and identify statistically significant predictors of science achievement at all levels (except classroom, where information was unavailable). Table 1 shows descriptive statistics on outcomes and predictors at different levels.9 Table 2 presents correlations between outcomes and predictors at different levels, providing some intuitive clues about the unadjusted relationships. The main analysis was based on HLM with four levels of data. Appendix A provides information on sample sizes for the nesting structure of the data.
Table 1. Descriptive Statistics on Science Achievement and Student, Teacher, and School Characteristics by Science Subjects
Note. School-level data are the same except for teacher shortage and teacher quality, which are different across the three content areas.
Table 2. Correlations Between Outcome Measures of Science Achievement and Student, Teacher, and School Characteristics
Note. All correlations are statistically significant at the alpha level of .05 (with statistically nonsignificant correlations omitted).
SEPARATING COMPETING EFFECTS ON SCIENCE ACHIEVEMENT
Table 3 presents the results of partition of variance in science achievement across the educational levels (students, classrooms, teachers, and schools) for each content area of science (physics, biology, and geology). Values in the table represent proportions. In physics, 75% of the variance came from students, 9% from classrooms, 3% from teachers, and 13% from schools. Stated differently, students were 75% responsible for the variation in physics achievement, classrooms were 9% responsible, teachers were 3% responsible, and schools were 13% responsible. Variance components at classroom, teacher, and school levels were all statistically significant (at the alpha level of .05). In biology, 78% of the variance came from students, 7% from classrooms, 1% from teachers, and 14% from schools. Variance components were statistically significant at classroom and school levels but not at the teacher level.10 Finally, in geology, 77% of the variance came from students, 9% from classrooms, 4% from teachers, and 10% from schools. Variance components at classroom, teacher, and school levels were all statistically significant. Overall, the breakdown of institutional contributions to science achievement across the three content areas of science clearly indicates a hierarchy of importance of educational units to science achievement in the order of schools, classrooms, and teachersthus, our first hypothesis was largely rejected.
Table 3. Partition of Variance in Outcome Measures of Science Achievement
Note. All (12) variances are statistically significant at the alpha level of .05 except one (i.e., variance in biology among teachers).
IDENTIFYING SALIENT PREDICTORS OF SCIENCE ACHIEVEMENT
Table 4 presents the estimates of effects of student, teacher, and school characteristics on science achievement across the three content areas of science (physics, biology, and geology). To measure the magnitude of each effect and compare effects between variables within and across content areas, we scaled statistically significant effects into a common metric that reports statistical results in effect size units (i.e., standard deviation [SD] units). We adopted the common practice of using the overall SD (from the null model) as the denominator to obtain effect size.
Table 4. Results of (Four-Level) HLM Models Estimating Effects of Student, Teacher, and School Characteristics on Outcome Measures of Science Achievement
Note. *p < .05. < .07. Because there is no statistically significant variance in biology achievement at the teacher level, there is no modeling effort at that level.
At the student level, male students outperformed female students (statistically) significantly across all three content areas. Effect size ranged from .11 SD (in physics) to .24 SD (in geology). There is a global effect size of .07 SD for gender differences in science achievement (Chiu, 2007).11 Chinese students demonstrated stronger gender differences in science achievement from the global perspective. Younger students outperformed older students significantly across all three content areas in the same grade cohort (i.e., Grade 8). Effect size ranged from .09 SD (in geology) to .14 SD (in biology). Bedard and Dhuey (2006) reported an international range of age differences in science achievement in Grade 8 from 2 percentile to 9 percentile (in favor of older students). This range translates into effect size from .05 SD to .25 SD. Chinese age differences fell within this range but differed in direction (in favor of younger students). Students whose fathers had higher SES significantly outperformed students whose fathers had lower SES in physics and geology. In biology, socioeconomic differences in favor of mothers with high SES were significant. Effect size ranged from .001 SD (in physics and biology) to .003 SD (in geology). Compared with the global effect size of .09 SD for (main) socioeconomic effects on science achievement (Chiu, 2007), Chinese effect size showed much weaker socioeconomic differences in science achievement.
Statistically significant student effects that scattered across the three content areas of science were also present. In physics, students from two-parent families outperformed students from single-parent families (effect size = .12 SD); in physics and biology, students who were single children outperformed students who had siblings (effect size = .09 SD and .08 SD); and in physics, students who were migrants from rural areas outperformed students who were not (effect size = .08 SD). Although these Chinese individual differences were isolated, they were stronger than relevant global measures of individual difference in science achievement, with .03 SD in favor of two-parent families (main effects), .04 SD in favor of few siblings (main effects), and .06 SD in favor of students who are (first generation) immigrants (Chiu, 2007).
At the teacher level, statistically significant results were scarce across the three content areas of science, due partially to the lack of significant variance in biology among teachers. The only significant finding was that in physics, students of teachers with a bachelors degree outperformed students of teachers with a degree lower than a bachelors (effect size = 1.43 SD). Kaya and Rice (2010) reported a relevant teacher effect of .56 SD on science achievement associated with teachers completion of a science major in the United States. The isolated Chinese teacher effect might imply stronger effects of teacher preparation on science achievement.
At the school level, statistically significant results were all contextual. We found only one school-level variable with consistent results across all three content areas of science. Students in schools with a higher per-student expenditure outperformed students in schools with a lower per-student expenditure (effect size = .12 SD in physics, .22 SD in biology, and .37 SD in geology, using 10,000 Chinese Yuan as one measurement unit). Across all (U.S.) states from 1972 to 2012, the overall relationship between education spending and academic achievement across school subjects was weak, with a correlation of .075, which translates into effect size as .15 SD (Coulson, 2014). Chinese school effects associated with student expenditure were stronger.
Statistically significant school effects also scattered across the three content areas of science. In geology, students in larger schools outperformed students in smaller schools (effect size = .03 SD using 100 students as one measurement unit). L. Sun, Bradley, and Akers (2012) reported .23 SD as the effect size (using 100 students as one measurement unit) of school size on science achievement in Hong Kong. In physics and biology, students in schools with higher mean SES outperformed students in schools with lower mean SES (effect size = .07 SD in both cases). L. Sun et al. (2012) reported .28 SD as the effect size of school SES composition on science achievement in Hong Kong. In biology, students in schools with a smaller percentage of female students outperformed students in schools with a larger percentage of female students (effect size = .51 SD using 10% as one measurement unit). Effect size was .14 SD, for an overall increase in academic achievement across school subjects when male students were moved from a coed school to a single-sex school, based on Korean national data (S. Lee, Turner, Woo, & Kim, 2015). In geology, students in schools where a larger percentage of teachers had at least a bachelors degree outperformed students in schools where a smaller percentage of teachers had at least a bachelors degree (effect size = .43 SD using 10% as one measurement unit). Based on U.S. national surveys, Darling-Hammond (2000) reported a correlation from .67 to .75 between (Grade 8) mathematics achievement and the percentage of certified teachers with a major in their field, which translates into an effect size from .18 SD to .23 SD (using 10% as one measurement unit). In biology, students in schools where principals perceived a lower level of teacher quality outperformed students in schools where principals perceived a higher level of teacher quality (effect size = .51 SD). L. Sun et al. (2012) reported .45 SD as the effect size of quality instruction on science achievement in Hong Kong. In biology, students in schools where principals had more working experience outperformed students in schools where principals had less working experience (effect size = .05 SD). RAND (2014) reported that three years of working experience for principals improves academic achievement across school subjects, from .7 percentile to 1.3 percentile in the United States (from .02 SD to .04 SD as effect size).
Our second hypothesis states that there are salient factors at each educational level responsible for science achievement. This hypothesis was essentially sustained given that there were statistically significant variables for science achievement at student, teacher, and school levels, although not as many and not as consistently as we would expect.
Finally, we provided a measure on the performance of the predictors in accounting for variance in science achievement. Following the common statistical practice, we calculated the proportion of variance in science achievement that was explained by the predictors. Overall, the predictors accounted for 9% of the total variance in physics, 10% in biology, and 12% in geology. These percentages were acceptable in social sciences as evidence that the predictors did explain the practically important amount of variance in science achievement (see Gaur & Gaur, 2006). The predictors performed particularly well at the school level, accounting for 53%, 69%, and 98% of the variance in physics, biology, and geology, respectively, among schools. Analysis of the U.S. national data indicated that 65% of the variance in science achievement was explained at the school level (Konstantopoulos, 2006).
SUMMARY OF PRINCIPAL FINDINGS
We separated the competing effects on science achievement among four educational unitsstudents, classrooms, teachers, and schoolsand identified factors at each level critical to science achievement. In physics, 75% of the variance came from students, 9% from classrooms, 3% from teachers, and 13% from schools. In biology, 78% of the variance came from students, 7% from classrooms, 1% from teachers, and 14% from schools. In geology, 77% of the variance came from students, 9% from classrooms, 4% from teachers, and 10% from schools. Variance components at classroom, teacher, and school levels were statistically significant across all three content areas (except for variance in biology at the teacher level).
Salient variables were sought among students, teachers, and schools (classroom variables were unavailable). At the student level, across all three content areas, males outperformed females, and younger students outperformed older students (in the same grade cohort). If we consider father and mother SES together, students whose parents had higher SES outperformed students whose parents had lower SES. In physics, at the teacher level, students of teachers with a bachelors degree outperformed students of teachers with lower than a bachelors degree. At the school level, across all three content areas, students in schools with a higher per-student expenditure outperformed students in schools with a lower per-student expenditure.
RELATIVE IMPORTANCE OF EDUCATIONAL UNITS
The vast majority of the variance in science achievement resided among students across the three content areas of science. The proportion of variance appears to be quite consistent at the student level (from 75% to 78%). It was not surprising that students themselves are mostly responsible for the variation in science achievement. Nonetheless, beyond student effects, we have found consistent sizeable institutional effects related to classrooms, teachers, and schools across the three content areas. With the three educational units together, institutional effects contributed between 22% and 25% to science achievement. This range comes close to the findings of Konstantopoulos (2006), who reported school contribution of 21% to science achievement in the United States. This is nearly one quarter of the total variance in science achievement, signaling that educational units (other than students) in the schooling system are substantially responsible for the variation in science achievement.
In previous studies, the breakdown of institutional effects becomes impossible from this point on. We are able to further break down institutional effects into three educational unitsclassrooms, teachers, and schoolsthus capturing institutional contributions to science achievement of all essential educational units in any schooling system. We found that school contribution to science achievement is quite consistent across the three content areas of science (from 10% to 14%). The same can be said about both classroom contribution (from 7% to 9%) and teacher contribution (from 1% to 4%) to science achievement across the three content areas. Of course, to many researchers, such consistencies come only as expected when examining content areas within science.
The mentioned breakdown of institutional contributions highlights a hierarchy of importance of educational units to science achievement in the order of schools, classrooms, and teachers. This finding can be rather provocative in that leadership [administrators] is second only to classroom instruction [teachers] as an influence on student learning (Louis, Leithwood, Wahlstrom, & Anderson, 2010, p. 9). Teachers are expected to be the most important educational unit (agent) to influence the learning (outcomes) of students, followed by administrators. Our finding challenges this notion even if one combines classroom effects with teacher effects as a way to consolidate teacher effects. This combination of classrooms and teachers shows a (combined) contribution to science achievement from 8% to 13% across the three content areas. The consolidated teacher effects are at optimal strength or at most as important as school effects.
Some researchers may believe that this finding is specific to the Chinese context. But, given the vigorous training of Chinese teachers in both content and pedagogy, the finding that teacher effects (even consolidated teacher effects) do not exceed school effects opens doors for serious investigations. China is one of the few countries in the world where entire universities are devoted to teacher training (i.e., normal universities). At any Chinese normal university, faculties in academic departments pursue advanced research as vigorously as faculties of the same fields at any other universities, making advanced (content) knowledge available to their students (preservice teachers). It is also common among Chinese normal universities to have departments entirely devoted to advanced research in pedagogy, making the latest pedagogical knowledge available to their students. Studies showing superior preparation of Chinese teachers are not rare (e.g., L. Ma, 1999). All this information suggests, at very least, that the educational unit of schools has great potential to impact student learning, implying that preparation and development are as critical for school administrators as they are for classroom teachers.
INDIVIDUAL DIFFERENCES IN SCIENCE ACHIEVEMENT IN CHINA
Consistent across the three content areas of science, individual differences in science achievement concern mainly students gender, age, and SES in China. Gender differences in favor of males in mathematics and science are evident in China, given that mathematics and science are considered male domains (J. Wang & Staver, 1997). Such a gender stereotype is rather common in Asia (Cheng & Seng, 2001), and with the outcomes of regional, national, and international studies (see Neuschmidt & Hastedt, 2008; OECD, 2015b), it functions as a self-fulfilling prophecy. Age differences in favor of younger students make sense in China because the country practices educational retention. Chinese parents in general do not disagree with schools that retain their children for academic catchup (L. Li, Deng, Liu, & Yan, 2013). Some other Asian countries (e.g., Singapore) also practice educational retention. Socioeconomic differences in favor of high (father and mother) SES are a product of modern Chinese economic development (Liu & Lu, 2008). Before the Chinese economic reform in the 1970s, there were hardly any academic-achievement-related socioeconomic differences among Chinese students (X. Zhou, Moen, & Tuma, 1998) because their parents were at a very similar socioeconomic level under the socialist system. What is striking about our finding is how fast economic development can spur socioeconomic differences in educational attainment; it took at most three decades in China.
From the global perspective (see the Results section), Chinese individual differences in science achievement imply two educational endeavors for China. Efforts are needed to narrow the large gender differences and maintain the small socioeconomic differences in science achievement. We emphasize that Chinese age differences (in favor of younger students) in science achievement break the international tradition (in favor of older students; see Bedard & Dhuey, 2006). We suspect that this breakaway relates strongly to the Chinese retention policy.
Current hot social issues in China do not bear substantial implications for Chinese science education. The (long-standing) one-child policy did not produce consistent positive effects across the three content areas, and the massive internal migration that separates parents and children did not show any negative effects. We expected consistent concerns across the three content areas about students with siblings and rural migrant students, especially those with migrant parents. Because these expectations did not occur, our findings challenge earlier studies in favor of one child and against migration (e.g., Falbo & Poston, 1993; H. Zhang et al., 2014).
TEACHER EFFECTS ON SCIENCE ACHIEVEMENT IN CHINA
Chinese teacher effects on science achievement can be understood from both between and within content areas of science. The pattern of teacher effects is quite different between content areas: a single predictor in physics, no variance in biology, and no predictor in geology. Apart from the classroom level (without variables), the teacher level was the only one for which there were no consistent effects across content areas even though they were in the same school subject. Teachers across content areas may apply rather different ways to promote science achievement. From a different perspective, (selected) variables at the teacher level together create a working environment that allows teachers to give and take (e.g., give as holding leadership positions, and take as being influenced by school culture). Such a working environment does not engage physics, biology, and geology teachers in the same way. This situation may have implications for professional development in science education. If the ways for teachers to affect science achievement are different across content areas, professional development organized by content areas may be more effective than pooling all science teachers together.
Within content areas of science, we observed scarce Chinese teacher effects (see Table 4). This situation makes sense in China, because teacher education and professional development are usually organized around unified national emphases and efforts that aim to bring all teachers up to a set of national expectations (e.g., professional commitment and classroom practice). The result is that Chinese teachers often behave in a similar way in their profession. Chinese teachers can all be effective (e.g., L. Ma, 1999), but few stand out with unique individual (professional) qualities that pull them ahead of the pack. Chinese teacher effects therefore propose a critical conceptual issue about what exactly teacher effectiveness is. Can we define effectiveness as all teachers modeling after (and thus becoming) one exemplary teacher so that there may be very few teacher effects? Or should we encourage teachers to be different by pursuing individual creativity and innovation that uniquely reform and shape their teaching practice so that there may be a harvest of teacher effects? Answers to these questions can be culturally quite diverse. If China leans more toward the all after one approach, the scarcity of teacher effects may actually be a sign of teacher effectiveness.
SCHOOL EFFECTS IN CONTEXT OF NATIONAL CURRICULUM
In China, what matters consistently at the school level to science achievement is student expenditure. School effects associated with student expenditure are stronger on science achievement in China than in the United States (see Results). What is unique about Chinese school effects is not the identification of student expenditure as their hallmark of school effects. This school-level variable has shown its importance to academic achievement. Lower student expenditure is associated with across-the-board lower achievement in reading, writing, mathematics, science, and social studies (Jones & Slate, 2010). Student expenditure is an issue that has a strong global implication for academic achievement (Vegas & Coffin, 2015).
What is truly unique about Chinese school effects is the lack of effects associated with variables descriptive of school climate at both teacher and school levels. We included a large number of school characteristics related to school climate at both teacher and school levels (see Table 4). With such a comprehensive inclusion and such a large sample, we expected a number of school climate effects, yet school characteristics critical to science achievement turned out to be all contextual. We believe that this situation has much to do with the Chinese practice of national curriculum. We do not have data to establish a credible link between the two, but we suspect that a national curriculum brings curricular and instructional practices to a similar level among teachers and schools. The Chinese national educational standards touch all aspects of schooling concerning students, teachers, and school administrators. The implementation of these national standards is effectively facilitated by uniform professional development, monitoring, and evaluation designed for both teachers and school administrators. As a result, teachers and schools become more similar than different in terms of school curriculum and instruction as well as school management and operation (we have alluded to a similar point from the perspective of teacher effects). This situation would leave only school context to make a difference in academic achievement. Given the regional socioeconomic diversity of Chinese economic reforms, some schools can afford to hire more educated teachers and invest more in school operation. We have likely started to observe the impact. Nonetheless, Chinese school contextual effects do not appear to be mature for two reasons. One is that most school contextual variables have not established a consistent pattern of effects across the three content areas, and the other is that the scattered contextual effects do not stand out from a global perspective (see Results).
The exclusive presence of school contextual effects on science achievement lends support to Reynolds et al. (2014), who emphasized the importance of the context variables, the study of which they believe can generate more sophisticated understandings about the process variables responsible for school effectiveness. In countries like China that practice uniform teacher education and professional development to fulfill a national curriculum, school contextual effects may well be the source of explanations for why students are still different from school to school in knowledge and skills. For example, the consistent effects of (school-level) student expenditure on science achievement across the three content areas of science imply that, because teachers are similarly effective, what matters to science achievement may become the addition of teachers, or the hiring of teaching assistants, or the addition of support staff, or the creation of climate-controlled classrooms. Rich schools can afford to pursue all these options. In other words, when schools are standardized for educational policy and practice, school climate varies little across schools. Some schools may have the means to change school context, which researchers traditionally believe is difficult to do, especially in Western and European countries.
Negative effects associated with principals perceived teacher quality occurred, although only in biology. With similar (absolute) effect size, L. Sun et al. (2012) found positive effects of quality instruction on science achievement in Hong Kong. This comparison implies that Chinese principals in high-achieving schools actually tend to push their teachers harder to become more effective. This situation makes sense in cultures such as Chinas, where effort often comes from an emphasis on the negatives. Chinese principals in high-achieving schools likely emphasize more with their teachers the difficulties and challenges they are facing (so that they need to work harder), as opposed to the accomplishments and excellence they have achieved.
LIMITATIONS AND ALTERNATIVES
Despite of the effort to construct the SES variable within the Chinese context, it may not be the best proxy for SES without the incorporation of parent education and family income. The relatively small Chinese socioeconomic differences in science achievement (see Results) may be considered tentative. That prior science achievement cannot be controlled for is another limitation. This variable has the potential to make some effects, particularly some school contextual effects (e.g., effects of school mean SES), disappear once it is controlled for in a model (e.g., Marks, 2015).
We applied HLM in an exploratory manner with four levels of data hierarchy (students, classrooms, teachers, and schools) because conceptually these levels were in the structure of the data. To use conceptual levels as random in the model, we tested variances at each level. Only was variance in biology at the teacher level not statistically significant, and normally this level should be omitted. However, we kept four levels for biology to make data analysis consistent across the three content areas of science for comparison (see Note 10). Although a reasonable justification, there is a broader issue of which data hierarchy (three or four levels) is more appropriate, given that typically, few classrooms are nested within teachers (e.g., see Appendix A). Statistically this is not a problem for HLM (see Barnett, Raudenbush, Brennan, Pleck, & Marshall, 1995), but one does face alternatives. The choice may come from the main purpose of an empirical study. We were strongly motivated to separate competing effects on science achievement among educational units to fill in a gap in the research literature, and as such, it appears appropriate to adopt four levels conceptually. If the identification of salient predictors of science achievement for educational policy and practice is the solo motivation, three levels may offer a better chance of finding strong predictors at the teacher level because of the consolidation of variances from classrooms and teachers.
1. Located on the southern bank of the Yellow River, this capital city of the Henan province is one of the eight ancient capitals of China (i.e., the national capital of the Shang dynasty). The city is one of the largest industrial centers in China, with a GDP per capita of about $6,200 in 2008. The population is about 9.38 million. Nearly 3,000 schools have approximately 2.86 million students and around 189,000 teachers. About 40% of the student population come from migrant families.
2. Although the way that the data were collected makes the Zhengzhou sample a (local) population, Zhengzhou is considered in PRAPBEQ as a typical metropolitan city in China, representing a large number of cities with rapid economic growth. These cities share many common characteristics. They are large industrial centers; are important strategic transportation hubs (rail or sea); attract large domestic and international investments; have strong political, economic, technological, and educational establishments; and attract migrants in huge numbers from all over the country. Zhengzhou is considered a sample from or representative of those highly industrialized cities (i.e., simple random samples of students, teachers, and principals). For this reason, we used inference models (with error terms assumed at student, classroom, teacher, and school levels). The strategy to use the city to generate a population rather than a sample simplifies complicated random sampling and data collection procedures and offers more reliable data for more credible generalization. Although research on many educational issues (including those in the present study) is impossible with the regular PRAPBEQ sample, this strategy makes it possible with the special Zhengzhou sample.
3. The conceptual clarity about the essential nature of social stratification is important in measuring SES (Oakes & Rossi, 2003). To clarify the Chinese social structure, C. Li (2005) analyzed Chinese national census data to establish a SES regression equation (see Duncan, Featherman, & Duncan, 1972) that allows for the construction of a standardized social stratification index based on occupation (e.g., 31.82 for farmers and 73.37 for people working in the business or management sector). We did not produce any SES measure (i.e., no indexing based on coding); instead, we matched occupational categories between PRAPBEQ (see Appendix B) and C. Li (2005) to derive SES scores for fathers and mothers. Parental education was available but correlated highly with the resulting SES measure. As a result, we retained C. Lis (2005) SES measure and employed parental education to estimate parental occupation when missing (e.g., primary schooling corresponds to workers). Because PRAPBEQ 2014 is not longitudinal (i.e., focusing on the eighth grade), prior achievement was not available. Other student-level variables often considered in the Western literature but absent in the present study were race-ethnicity and language spoken at home. The overwhelming majority of Chinese population is Han and speaks Mandarin. Language spoken at home was not measured in PRAPBEQ, and minorities other than Han were less than 1% in our data (and thus not used).
4. This classification of teacher-level variables is heuristic rather than theoretical or analytical, grouped on the basis of measurement orientations. Analytically, they are treated as equal and do not necessarily constitute any test of importance among those categories.
5. Appendix B presents many scales that PRAPBEQ has developed to measure critical constructs at the teacher and school levels. Many items were adopted from different large-scale national and international student assessments, such as the National Assessment of Educational Progress (NEAP) and PISA. Some items were modified to accommodate Chinese conditions, and some items unique to Chinese education were added. Student, teacher, and principal questionnaires were all piloted to ensure validity and reliability.
6. At the school level, data were available that indicated, within each school, the number of teachers across the four categories of education (see Appendix B). The percentage of teachers with at least a bachelors degree was then calculated from those data, representing the overall education of teachers in a school. Separately, at the teacher level, there were two dummy variables descriptive of the educational level of each specific teacher (teacher characteristics). School contextual variables often considered in the Western literature but absent in the present study were school location and school sector. Because our data concern a Chinese metropolitan city and private schools are rather rare in China, those two variables were irrelevant.
7. The four levels of data hierarchy were inherited (because of the use of all eighth graders in a city) rather than (cluster) sampled. We applied HLM in an exploratory way because conceptually there were multiple levels in the structure of the data (i.e., students, classrooms, teachers, and schools). To use conceptual levels as random in the model, one typically tests variances at each level for statistical significance. Statistically significant variances should then be used in the estimation of the model. Although a common statistical practice, this is a data-driven decision rather than a sampling-driven decision.
8. Because the research goal was to identify statistically significant predictors of science achievement, student-level variables were fixed at the next (classroom) level, and teacher-level variables were fixed at the next (school) level, as recommended in Thum and Bryk (1997).
9. We omitted a number of large tables of correlations among predictors at various levels. We provided here brief summaries (in absolute values). Across physics, biology, and geology, correlations ranged from .00 to .33, from .00 to .35, and from .00 to .33, respectively, at the student level; and correlations ranged from .00 to .81, from .00 to .80, and from .00 to .83, respectively, at the teacher level. Correlations ranged from .00 to .79 at the school level. We note that the vast majority of correlations were well below .50 at both teacher and school levels.
10. Only was variance in biology at the teacher level not statistically significant in Table 3. In this case, we could have combined classrooms and teachers into one level for biology (so as to have three levels with all variances statistically significant). We kept four levels for biology to make data analysis consistent across the three content areas of science for comparison. This is a reasonable decision, given also that we did not perform any data analysis at the teacher level for biology (because of the lack of statistically significant variance).
11. Although there are general categories about small, moderate, and large effect sizes, we adopted for two reasons a different strategy for interpreting effect sizes. We realize that context matters the most when one categorizes effect sizes (e.g., a fifth of a SD is meaningful in education sciences). We also note that those general categories are usually established from Western and European research which may not precisely describe effects in the developing world. Our strategy, thus, is to interpret our effect sizes by comparing them to estimates in previous work with the same or similar research purposes. When possible, we select global or international average measures to make our interpretation even more meaningful.
The authors are grateful to the principal investigators of the Program for Regional Assessment and Promotion of Basic Education Quality for providing data for the present study.
Akkus, R., Gunel, M., & Hand, B. (2007). Comparing an inquiry-based approach known as the science writing heuristic to traditional science teaching practices: Are there differences? International Journal of Science Education, 29(14), 17451765.
Allen, J. (2009). A sense of belonging: Sustaining and retaining new teachers. Portland, OR: Stenhouse.
Arens, A. K., & Morin, A. J. S. (2016). Relations between teachers emotional exhaustion and students educational outcomes. Journal of Educational Psychology, 108(6), 800813.
Arum, R. (1996). Do private schools force public schools to compete? American Sociological Review, 61(1), 2946.
Atar, H. Y., & Atar, B. (2012). Examining the effects of Turkish education reform on students TIMSS 2007 science achievement. Educational Sciences: Theory and Practice, 12(4), 26322636.
Barish, L. (2008). Leading in science: Teachers perceptions of their principals supervisory effectiveness (Unpublished doctoral dissertation). Fordham University, New York, NY.
Barnett, R. C., Raudenbush, S. W., Brennan, R. T., Pleck, J. H., & Marshall, N. L. (1995). Change in marital experiences and change in psychological distress: A longitudinal study of dual-earner couples. Journal of Personality and Social Psychology, 69, 839850.
Barton, J. D. (2015). An investigation of the relationship between school size, socio-economic status, expenditure-per-student, mobility rate, and percentage of non-white secondary students taking state science exams (Unpublished doctoral dissertation). Texas A&M University, College Station.
Bedard, K., & Dhuey, E. (2006). The persistence of early childhood maturity: International evidence of long-run age effects. The Quarterly Journal of Economics, 121, 14371472.
Bosworth, R. (2014). Class size, class composition, and the distribution of student achievement. Education Economics, 22(2), 141165.
Breakwell, G. M., & Beardsell, S. (1992). Gender, parental and peer influences upon science attitudes and activities. Public Understanding of Science, 1(2), 183197.
Bursal, M. (2013). Longitudinal investigation of elementary students science academic achievement in 4-8th grades: Grade level and gender differences. Educational Sciences: Theory and Practice, 13(2), 11511156.
Caprara, G. V., Barbaranelli, C., Steca, P., & Malone, P. S. (2006). Teachers self-efficacy beliefs as determinants of job satisfaction and students academic achievement: A study at the school level. Journal of School Psychology, 44, 473490.
Chambers, E. A., & Schreiber, J. B. (2004). Girls academic achievement: Varying associations of extracurricular activities. Gender and Education, 16, 327346.
Chang, I. H. (2011). A study of the relationships between distributed leadership, teacher academic optimism and student achievement in Taiwanese elementary schools. School Leadership and Management, 31(5), 491515.
Chen, S. F., Lin, C. Y., Wang, J. R., Lin, S. W., & Kao, H. L. (2012). A cross-grade comparison to examine the context effect on the relationships among family resources, school climate, learning participation, science attitude, and science achievement based on TIMSS 2003 in Taiwan. International Journal of Science Education, 34(14), 20892106.
Cheng, S. K., & Seng, Q. K. (2001). . Studies in Educational Evaluation, 27(4), 331340.
Chinese Ministry of Education. (2010). Implementation of national plans [Chinese]. Beijing, China: Author.
Chinese National Innovation Center for Assessment of Basic Education Quality. (2015). Science report for Zhengzhou: The 2014 Program for Regional Assessment and Promotion of Basic Education Quality [Chinese]. Beijing, China: Author.
Chiu, M. M. (2007). Families, economies, cultures, and science achievement in 41 countries: Country-, school-, and student-level analyses. Journal of Family Psychology, 21(3), 510519.
Clark, D., Martorell, P., & Rockoff, J. (2009). School principals and school performance. Washington, DC: The Urban Institute.
Coleman, J. S., Campbell, E. Q., Hobson, C. J., McPartland, F., Mood, A. M., Weinfeld, F. D., & York, R. L. (1966). Equality of educational opportunity. Washington, DC: U.S. Government Printing Office.
Coulson, A. J. (2014). State education trends: Academic performance and spending over the past 40 years. Washington, DC: Cato Institute.
Dalton, B. (2012). Grade level and science achievement: U.S. performance in cross-national perspective. Comparative Education Review, 56(1), 125154.
Darling-Hammond, L. (2000). Teacher quality and student achievement: A review of state policy evidence. Education Policy Analysis Archives, 8(1), 144.
Drent, M., Meelissen, M. R. M., & Van Der Kleij, F. M. (2013). The contribution of TIMSS to the link between school and classroom factors and student achievement. Journal of Curriculum Studies, 45(2), 198224.
Duncan, O. D., Featherman, D. L., & Duncan, B. (1972). Socioeconomic background and achievement. New York, NY: Seminar.
Gaur, A. S., & Gaur, S. S. (2006). Statistical methods for practice and research: A guide to data analysis using SPSS. Thousand Oaks, CA: SAGE.
Good, T. L., & Brophy, J. E. (1986). School effects. In M. Wittrock (Ed.), Third handbook of research on teaching (pp. 570602). New York, NY: Macmillan.
Guest, A., & Schneider, B. (2003). Adolescents extracurricular participation in context: The mediating effects of schools, communities, and identity. Sociology of Education, 76, 89105.
Hampden-Thompson, G., & Pong, S. (2005). Does family policy environment moderate the effect of single-parenthood on childrens academic achievement? A study of 14 European countries. Journal of Comparative Family Studies, 36(2), 227248.
Huang, F. (2004). Curriculum reform in contemporary China: Seven goals and six strategies. Journal of Curriculum Study, 36(1), 101115.
Ingersoll, R., & Perda, D. (2009). The mathematics and science teacher shortage: Fact and myth. Philadelphia: Consortium for Policy Research in Education, University of Pennsylvania.
Jiang, F., & McComas, W. F. (2015). The effects of inquiry teaching on student science achievement and attitudes: Evidence from propensity score analysis of PISA data. International Journal of Science Education, 37(3), 554576.
Johnson, C. C., Kahle, J. B., & Fargo, J. D. (2007). Effective teaching results in increased science achievement for all students. Science Education, 91(3), 371383.
Jones, T. B., & Slate, J. R. (2010). The 65% instructional expenditure ratio and student achievement: Does money matter? Current Issues in Education, 13(4).
Klassen, R. M., Aldhafri, S., Mansfield, C. F., Purwanto, E., Siu, A. F. Y., Wong, M. W., & Woods-McConney, A. (2012). Teachers engagement at work: An international validation study. Journal of Experimental Education, 80(4), 317337.
Kleickmann, T., Tröbst, S., Jonen, A., Vehmeyer, J., & Möller, K. (2016). The effects of expert scaffolding in elementary science professional development on teachers beliefs and motivations, instructional practices, and student achievement. Journal of Educational Psychology, 108(1), 2142.
Konstantopoulos, S. (2006). Trends of school effects on student achievement: Evidence from NLS:72, HSB:82, and NELS:92. Teachers College Record, 108(12), 25502581.
Konstantopoulos, S., & Borman, G. (2011). Family background and school effects on student achievement: A multilevel analysis of the Coleman data. Teachers College Record, 113(1), 97132.
Kunter, M., Klusmann, U., Baumert, J., Richter, D., Voss, T., & Hachfeld, A. (2013). Professional competence of teachers: Effects on instructional quality and student development. Journal of Educational Psychology, 105(3), 805820.
Kyriakides, L., Christoforou, C., & Charalambous, C. Y. (2013). What matters for student learning outcomes: A meta-analysis of studies exploring factors of effective teaching. Teaching and Teacher Education, 36, 143152.
Lai, F. (2010). Are boys left behind? The evolution of the gender achievement gap in Beijings middle schools. Economics of Education Review, 29, 383399.
Lamons J. (2009). An analysis of data collected from the 2007-2008 Tennessee state report card and the variables related to science test results (Unpublished doctoral dissertation). East Tennessee State University, Johnson City.
Lee, S., Turner, L. J., Woo, S., & Kim, K. (2015). The impact of school and classroom gender composition on educational achievement (Unpublished manuscript). Department of Economics, University of Maryland, College Park.
Lee, V. E., & Bryk, A. S. (1989). A multilevel model of the social distribution of high school achievement. Sociology of Education, 62(3), 172192.
Lee, V. E., & Smith, J. B. (1997). High school size: Which works best and for whom? Educational Evaluation and Policy Analysis, 19(3), 205227.
Li, C. (2005). Prestige stratification in the contemporary China: Occupational prestige measures and socio-economic index. Sociological Studies, 30, 74102.
Li, L., Deng, J., Liu L., & Yan, G. (2013). An empirical study on the policy for educational retention in Shanghai [Chinese]. Retrieved from http://www.edum.org.mo/download/177.pdf
Liu, X., & Lu, K. (2008). Student performance and family socioeconomic status: Results from a survey of compulsory education in western China. Chinese Education and Society, 41(5), 7083.
Longbottom, J. E., & Butler, P. H. (1999). Why teach science? Setting rational goals for science education. Science Education, 83, 473492.
Louis, K. S., Dretzke, B., & Wahlstrom, K. (2010). How does leadership affect student achievement? Results from a national U.S. survey. School Effectiveness and School Improvement, 21(3), 315336.
Louis, K. S., Leithwood, K., Wahlstrom, K., & Anderson, S. (2010). Investigating the links to improved student learning: Final report of research findings. New York, NY: Wallace Foundation.
Lu, Z., Zhang, G., & Xu, X. (1999). Investigating school record on children of single-parent and only-child families [Chinese]. Journal of Chinese School Health, 20(6), 416.
Lumpe, A., Czerniak, C., Haney, J., & Beltyukova, S. (2012). Beliefs about teaching science: The relationship between elementary teachers participation in professional development and student achievement. International Journal of Science Education, 34(2), 153166.
Ma, L. (1999). Knowing and teaching elementary mathematics: Teachers understanding of fundamental mathematics in China and the United States. Mahwah, NJ: Erlbaum.
Ma, X. (2008). Within-school gender gaps in reading, mathematics, and science literacy. Comparative Education Review, 52(3), 437460.
Ma. X., Ma, L., & Bradley, K. (2008). Using multilevel modeling to investigate school effects. In A. A. O Connell & D. B. McCoach (Eds.), Multilevel modeling of educational data (pp. 59110). Charlotte, NC: Information Age.
Mangrubang, F. R. (2005). Issues and trends in science education: The shortage of qualified science teachers. American Annals of the Deaf, 150(1), 4246.
Marks, G. N. (2015). Are school-SES effects statistical artefacts? Evidence from longitudinal population data. Oxford Review of Education, 41, 122144.
Martin, M. O., Mullis, I. V. S., & Foy, P. (2008). TIMSS 2007 international science report. Washington, DC: International Association for the Evaluation of Educational Achievement.
Martin, M. O., Mullis, I. V. S., Foy, P., Olson, J. F., Preuschoff, C., Erberber, E., . . . Galia, J. (2008). TIMSS 2007 international science report: Findings from IEAs Trends in International Mathematics and Science Study at the eighth grade. Chestnut Hill, MA: TIMSS & PIRLS International Study Center, Boston College.
Martin, M. O., Mullis, I. V. S., Foy, P., & Stanco, G. M. (2012). TIMSS 2011 international science report. Chestnut Hill, MA: TIMSS & PIRLS International Study Center, Boston College.
McGinnis, J. R., Parker, C., & Graeber, A. O. (2004). A cultural perspective of the induction of five reform-minded beginning mathematics and science teachers. Journal of Research in Science Teaching, 41(7), 720747.
Menekse, M., Stump, S .G., Krause, S., & Chi, M. T. H. (2013). Differentiated overt learning activities for effective instruction in engineering classrooms. Journal of Engineering Education, 102(3), 346374.
Miller, A. (2013). Principal turnover and student achievement. Economics of Education Review, 36, 6072.
Miller, R. J., & Rowan, B. (2006). Effects of organic management on student achievement. American Educational Research Journal, 43(2), 219253.
Mohammadpour, E., Shekarchizadeh, A., & Kalantarrashidi, S. A. (2015). Multilevel modeling of science achievement in the TIMSS participating countries. The Journal of Educational Research, 108, 449464.
Monk, D. H. (1994). Subject area preparation of secondary mathematics and science teachers and student achievement. Economics of Education Review, 13, 125145.
Moorosi, P., & Bush, T. (2011). School leadership development in Commonwealth countries: Learning across the boundaries. International Studies in Educational Administration, 39(3), 5975.
Nathan, B. (2015). The power of teacher leaders: Their roles, influence, and impact. New York, NY: Routledge.
National Bureau of Statistics of China. (2014). China statistical yearbook [Chinese]. Retrieved from http://www.stats.gov.cn/tjsj/ndsj/2014/indexch.htm
National Center for Education Statistics. (2011). The nations report card: Science 2009 (NCES 2011-451). Washington, DC: Institute of Education Sciences, U.S. Department of Education.
Neuschmidt, O., & Hastedt, J. B. D. (2008).Trends in gender differences in mathematics and science (TIMSS 19952003). Studies in Educational Evaluation, 34(1), 5672.
Next Generation Science Standards Lead States. (2013). Next generation science standards: For states, by states. Washington, DC: National Academies Press.
Nye, B., Hedges, L. V., & Konstantopoulos, S. (2004). Do minorities experience larger lasting benefits from small classes? The Journal of Educational Research, 98(2), 94100.
Oakes, J. M., & Rossi, P. H. (2003). The measurement of SES in health research: Current practice and steps toward a new approach. Social Science and Medicine, 56(4), 769784.
Ogundokun, O. K. (2012). The impact of school management strategies on academic achievement in Texas schools (Unpublished doctoral dissertation). Trident University, Cypress, CA.
Onocha, C., & Okpala, P. (2001). Family and school environmental correlates of integrated science achievement. Journal of Psychology, 121(3), 281286.
Organisation for Economic Co-operation and Development. (2015a). PISA 2015 draft questionnaire framework. Paris, France: Author.
Organisation for Economic Co-operation and Development. (2015b). What lies behind gender inequality in education? Paris, France: Author.
Perrachione, B. A., Rosser, V. J., & Peterson, G. J. (2008). Why do they stay? Elementary teachers perceptions of job satisfaction and retention. Professional Educator, 32(2), 117.
Pong, S., Dronkers, J., & Hampden-Thompson, G. (2003). Family policies and childrens school achievement in single- versus two-parent families. Journal of Marriage and Family, 65(3), 681699.
RAND. (2014). Principal preparation matters: How leadership affects student achievement. Santa Monica, CA: Author.
Raudenbush, S. W., & Bryk, A. B. (2002). Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: SAGE.
Raudenbush, S. W., &Willms, J. (1995). The estimation of school effects. Journal of Educational and Behavioral Statistics, 20(4), 307335.
Reynolds, D., Creemers, B., Nesselrodt, P. S., Shaffer, E. C., Stringfield, S., &Teddlie, C. (Eds.). (2014). Advances in school effectiveness research and practice. New York, NY: Elsevier.
Rice, J. K. (2003). Understanding the effectiveness of teacher attributes. Washington, DC: Economic Policy Institute.
Rivkin, S. G., Hanushek, E. A., & Kain, J. F. (2005). Teachers, schools, and academic achievement. Econometrica, 73(2), 417458.
Seidel, T., & Shavelson, R. J. (2007). Teaching effectiveness research in the past decade: The role of theory and research design in disentangling meta-analysis results. Review of Educational Research, 77(4), 454499.
Silverstein, S. C., Dubner, J., Miller, J., Glied, S., & Loike, J. D. (2009). Teachers participation in research programs improves their students achievement in science. Science, 326, 440442.
Singh, K., & Billingsley, B. S. (1998). Professional support and its effects on teachers commitment. The Journal of Educational Research, 91(4), 229239.
Skaalvik, E. M., & Skaalvik, S. (2011). Teacher job satisfaction and motivation to leave the teaching profession: Relations with school context, feeling of belonging, and emotional exhaustion. Teaching and Teacher Education, 27, 10291038.
Smetana, L. K., Wenner, J., Settlage, J., & McCoach, D. B. (2016). Clarifying and capturing trust in relation to science education: Dimensions of trustworthiness within schools and associations with equitable student achievement. Science Education, 100(1), 7895.
Su, Z., Goldstein, S., & Su, J. (1995). Science education goals and curriculum designs in American and Chinese high schools. International Review of Education, 41(5), 371388.
Sun, J., & Hong, Z. (1994). Preliminary study on school effectiveness. Education and Economy, 3, 15.
Sun L., Bradley, K. D., & Akers, K. (2012). A multilevel modeling approach to investigating factors impacting science achievement for secondary school students: PISA Hong Kong sample. International Journal of Science Education, 34(14), 21072125.
Supovitz, J. A., & Turner, H. M. (2000). The effects of professional development on science teaching practices and classroom culture. Journal of Research in Science Teaching, 37(9), 963980.
Taylor, J. A., Getty, S. R., Kowalski, S. M., Wilson, C. D., Carlson, J., & Scotter, P. V. (2015). An efficacy trail of research-based curriculum materials with curriculum-based professional development. American Educational Research Journal, 52(5), 9841017.
Thum, Y. M., & Bryk, A. S. (1997). Value-added productivity indicators: The Dallas system. In J. Millman (Ed.), Grading teachers, grading schools: Is student achievement a valid evaluation measure? (pp. 100109). Thousand Oaks, CA: Corwin Press.
Ussher, B. (2010). Involving a village: Student teachers sense of belonging in their school-based placement. Asia Pacific Journal of Teacher Education, 38(2), 103116.
Vegas, E., & Coffin, C. (2015). When education expenditure matters: An empirical analysis of recent international data. Comparative Education Review, 59(2), 289304.
Von Secker, C. E., & Lissitz, R. W. (1999). Estimating the impact of instructional practices on student achievement in science. Journal of Research in Science Teaching, 36(10), 11101126.
Wang, J. (2012). Curriculum reform in mainland China, 19782008: Change, maintenance, and conflicts. Chinese Education and Society, 45(1), 5968.
Wang, J., & Staver, J. R. (1997). An empirical study of gender differences in Chinese students science achievement. The Journal of Educational Research, 90(4), 252255.
Wang, W., Wang, J., Zhang, G., Lang, Y., & Mayer, V. J. (1996). Science education in the Peoples Republic of China. Science Education, 80(2), 203222.
Wayne, A. J., & Youngs, P. (2003). Teacher characteristics and student achievement gains: A review. Review of Educational Research, 73(1), 89122.
Wenglinsky, H. (2000). How teaching matters: Bringing the classroom back into discussions of teacher quality. Princeton, NJ: Educational Testing Service.
Whipple, S. S., Evans, G. W., Barry, R. L., & Maxwell, L. E. (2010). An ecological perspective on cumulative school and neighborhood risk factors related to achievement. Journal of Applied Developmental Psychology, 31(6), 422427.
Wößmann, L. (2003). Schooling resources, educational institutions and student performance: The international evidence. Oxford Bulletin of Economics and Statistics, 65(2), 117170.
Wößmann, L., Lüdemann, E., Schütz, G., & West, M. R. (2007). School accountability, autonomy, choice, and the level of student achievement (OECD Education Working Papers, No. 13). Paris, France: OECD Publishing.
Yang, Y. T. C. (2012). Cultivating critical thinkers: Exploring transfer of learning from pre-service teacher training to classroom practice. Teaching and Teacher Education, 28(8), 11161130.
Yildirim, O., Acar, A. C., Bull, S., & Sevinc, L. (2008). Relationships between teachers perceived leadership style, students learning style, and academic achievement: A study on high school students. Educational Psychology, 28(1), 7381.
Yildirim, O., & Demir, S. B. (2014). The examination of teacher and student effectiveness at TIMSS 2011 science and math scores using multilevel models. Pakistan Journal of Statistics, 30(6), 12111218.
Young, D. J., & Fraser, B. J. (1994). Gender differences in science achievement: Do school effects make a difference? Journal of Research in Science Teaching, 31, 857871.
Young, D. J., Reynolds, A. J., & Walberg, H. J. (1996). Science achievement and educational productivity: A hierarchical linear model. The Journal of Educational Research, 89, 272278.
Zhang, B., & Jiang, T. (2009). A correlation analysis of teacher-student relationship and academic achievement in junior high schools [Chinese]. Psychological Science, 32(4), 10151017.
Zhang, H., Behrman, J. R., Fan, S. C., Wei, X., & Zhang, J. (2014). Does parental absence reduce cognitive achievements? Evidence from rural China. Journal of Development Economics, 111, 181195.
Zhang, S. (2006). A correlation analysis on student and teacher factors related to academic achievement in vocational schools [Chinese]. Neijiang Science and Technology, 26(1), 6162.
Zhang, W., Xin, T., & Kang, C. (2010). Effects of teacher characteristics on 4th grade students math achievement: A value-added study [Chinese]. Journal of Education Studies, 6(2), 6976.
Zhao, N., Valcke, M., Desoete, A., & Verhaeghe, J. P. (2012). The quadratic relationship between socioeconomic status and learning performance in China by multilevel analysis: Implications for policies to foster education equity. International Journal of Educational Development, 32, 412422.
Zhou, M., Murphy, R., & Tao, R. (2014). Effects of parents migration on the education of children left behind in rural China. Population and Development Review, 40(2), 273292.
Zhou, P., & Song, H. (2004). Effects of encouragement from teachers on student achievement. Contemporary Educational Science, 18(1), 57.
Zhou, X., Moen, P., & Tuma, N. B. (1998). Educational stratification in urban China: 194994. Sociology of Education, 71(3), 199222.
Ziegler, A., & Heller, K. (1997). Gifted females: A cross-cultural study. In J. Chan, R. Li, & J. Spinks (Eds.), Maximizing potential: Lengthening and strengthening stride (pp. 242247). Hong Kong, China: World Council for Gifted and Talented.
Zuelke, L. A. (2008). Relationships among science teacher qualifications, instructional practices, and student science achievement (Unpublished doctoral dissertation). University of Florida, Gainesville.
Sample Sizes for Nesting Structure of Data Across Science Content Areas
Note. For each content area, average numbers indicate how many students within a classroom, how many classrooms within a teacher, and how many teachers within a school.
Description of Student, Teacher, and School Characteristics