Introduction

It has been well disseminated that students from high socioeconomic status (SES) families outperform their counterparts from low SES families (Lee, 2005). Students from low-income parents show lower mathematical and scientific literacy than those from high-income parents (Morgan, Farkas, Hillemeier, & Maczuga, 2009). In the Program for International Student Assessment (PISA) 2015, approximately three quarters of students with limited understanding of science are from low SES family background, while 65% of students with high-level mathematics achievement are from high SES families (Organization for Economic Co-operation and Development [OECD], 2016b). Moreover, Heckman (2011) reported that high SES students are 2 years advanced in mathematics competency compared to the lower SES students, and this earlier exposure would lead to inequality in students’ learning outcomes. Equity in education becomes more critical because individual academic success is closely related to one’s quality of life (OECD, 2008).

Educational reformers have attempted to reduce the learning gap between high and low SES students. One of many reformers’ approaches to address this inequity in education is to emphasize student-centered learning environments (Acar, 2015; Mehalik, Doppelt, & Schuun, 2008). They point out that one main reason for achievement gaps in traditional classrooms is unequal access to educational opportunities in learning mathematics and science (Darling-Hammond, 1998). Reform-based instruction intends to provide equal learning opportunities in which all students actively engage in communicating their understanding of concepts. In such learning environments, students are more likely to bring their own ideas and prior understanding to learn mathematics and science, compared to traditional teacher-centered learning environments (Akkus, Gunel, & Hand, 2007; Cavagnetto & Hand, 2012; von Secker, 2002).

However, teachers sometimes implement student-centered learning approaches in ways different from the intentions of educational reformers (Lefstein, 2008). For example, secondary school mathematics and science teachers have a wide range of interpretations of student-centered learning environments (Anderson, 2002). Some teachers may not have solid understanding of what student-centered instruction is and how to implement such an approach when teaching. Moreover, Marshall (2009) reported that teachers only use 39% of instructional time to implement student-centered learning approaches throughout the school year. Wide variety in implementation may yield results different from reformers’ expectation for the positive impact of student-centered learning approaches on equity in education.

The relationship of three important issues of education—student performance, equity in education, and instructional approach—is important to investigate to promote equal opportunities to learn and advance in mathematics and science for all students. In particular, we investigate how teachers’ instructional practices moderate the relationships between students’ socioeconomic background and learning of mathematics and science in different educational contexts. Focusing on teachers’ student-centered instruction, this study aims to describe the moderated relationships between socioeconomic status and mathematical and scientific literacy in the PISA 2012 and 2015, respectively.

Theoretical Background

Equity in Learning Mathematics and Science

Definitions of equity depend on where researchers put their emphases in the process of explaining inequality (Bulkley, 2013). The American Library Association (2014) defined equity as keeping fairness and justice by removing uneven starting points or providing extra measurements to disadvantaged groups of people. In education, equity is referred to as equal access to learning resources, equality in learning outcomes, equal connections to family culture, and equal distribution of agency (Lynch, 2000). The PISA 2012 (OECD, 2013) conceptualized equity in education through the strength of the relationships between one’s SES and academic performance. These definitions of equity in education agree that all students ought to have equal access to take advantage of educational opportunities regardless of their SES.

Due to the importance of equity in education, researchers have investigated and shown the influences of learners’ background on mathematical or scientific literacy development. Socioeconomically and culturally disadvantaged students have lower mathematical and scientific literacy than advantaged students (Maerten-Rivera, Ahn, Lanier, Diaz, & Lee, 2016; Rodriguez, 1998). In particular, non-free/reduced lunch (Non-FRL) program students have higher performance than those in FRL programs; boys outperform girls; Whites and Asians are more proficient than Blacks and Hispanics; and native English-speaking students are better in mathematics and science than non-native English-speaking students. Atwater (2000) argued that underrepresented students’ frequent misplacements into lower level classes contribute to their lower level scientific literacy. Moreover, students with access to limited home educational resources are less likely to perform well in assessments when they do not get enough support from schools (Darling-Hammond, 1998). Students could further lose interest and willingness to learn mathematics when they have both a lack of school supports and limited home resources (Martin, 2009).

Equity in education across different educational contexts has been investigated by international studies. Studies have revealed similar patterns in the relationships between SES and academic achievement. In addition, large-scale international comparison studies such as the PISA and the Trends in International Mathematics and Science Study (TIMSS) display achievement gaps between various student groups—social class, gender, and language across countries and educational systems. The PISA 2003 and 2009 showed that students’ SES have strong influences on students’ performance in mathematics and science (OECD, 2010; Schütz, West, & Wöbmann, 2007). The OECD (2009) suggested disparities in access to educational resources between those groups as a reason for the consistent achievement gap: Students with advantaged backgrounds are more likely to attend high-qualified schools and have sufficient interactions with teachers to verify their understanding compared to underserved students.

Classroom Instruction and Learning Opportunities

Educational reformers have pointed out possible relationships between instructional approaches and inequity in education. To discuss those relationships, two types of instructional approaches are compared: teacher-centered and student-centered instruction.

Teacher-Centered Instruction.

Teacher-centered instruction is a traditional approach to teaching characterized as a monological learning environment. The teacher is the one who speaks most of the time in front of the class, while students are expected to sit quietly, take notes, and learn from what the teachers teach. In these environments, teachers transmit information to students, whose role is one of a passive information receiver in the classroom (National Research Council [NRC], 2000). Teacher-centered instruction rarely encourages students to communicate with teachers and classmates about what students do not fully understand. As a result, students’ lack of opportunities to actively examine the given information prevents them from learning through critical thinking practices (Delpit, 1988; Townley, 1993).

In this inactive and rigid classroom atmosphere, students might need to rely on home resources when they are not able to reach understanding in class. Students may or may not have a chance to catch or make up their deficit outside the classroom, depending on their available resources: Students with sufficient home resources will receive supplementary opportunities to learn and understand whereas students without such resources may not have additional chances. Thus, home resources play a significant role in students’ learning, which leads to inequity in education (Lee & Luykx, 2007). Students without enough parental supports and home resources are likely marginalized from learning situations in the traditional classroom settings (Acar, 2015; Lee, Smith, & Croninger, 1997).

Student-Centered Instruction.

Compared to teacher-centered instruction, student-centered instruction (e.g., inquiry-based and student-oriented instruction in this study) emphasizes students’ active participation in the learning situation. Students are encouraged to communicate in classrooms by engaging in activities such as argument-based inquiry, designing an experiment (Ballenger, 1997), exploring mathematics tasks, or mathematical modeling (Artigue & Blomhøj, 2013). Student-centered instruction requires students to negotiate different ideas through verbal and written arguments and to use multiple representations to explain scientific theories or mathematical logic (Cavagnetto, 2010). In this learning environment, students are expected to have equal chances to share their ideas, ask questions, and develop understanding of new concepts. Active participation in classroom discussion invites student to establish critical thinking skills (Mistler-Jackson, & Songer, 2000; Richardson, 1996; Yore, Bisanz, & Hand, 2003).

Students who have little academic experience from out-of-school education can benefit from student-centered learning environments more than teacher-centered learning environments (Ballenger, 1997). In inquiry-based or problem-solving learning, students are encouraged to engage in making arguments, expressing their thoughts, and negotiating different opinions in the process of developing understanding. When participating in this process, students have opportunities to express misunderstandings and to catch up with what was not fully understood. This practice is particularly beneficial for students who do not have access to academic support outside of the classroom, as they have chances to learn and understand within the school, as opposed to the teacher-centered learning environments where students are less like to catch up with what they missed (Fraser, 2015; Roth, 1996). The NRC framework (2012) displayed the same notion of student-centered learning that values children’s everyday experiences as the natural process of learning. Through student-centered learning experiences, supportive student-teacher relationships and learning ownership are consequently developed. In student-centered learning environments, non-school-related factors such as socioeconomic background may not play significant roles in the learning of mathematics and science, and students can instead utilize whatever resources they bring to their learning and develop their own voice. This could be more beneficial for all students, including the ones in low SES groups.

Difficulty of Implementing Student-Centered Instruction

Despite the potential benefits of student-centered instructional practices, it is difficult to say that teacher implementation of such practices is carried out with fidelity, which is related to the quality of instruction (Kirschner, Sweller, & Clark, 2006). This is because theoretical constructs and classroom materials for student-centered instruction struggle to capture the complexity of the realm of schools (Maaß & Artigue, 2013). Teachers shared their concerns about student-centered instruction, such as class time limitations and bigger efforts required compared to lecture-based instruction (Loughran, 1994; Mercer, 2008). Implementing student-centered instruction requires both teachers’ and students’ cognitive engagement more so than in teacher-centered instruction (Magnusso & Palincsar, 2005). Due to these difficulties, many science teachers still heavily rely on their “cook-book hands-on activities” for inquiry-based learning approaches (Akkus et al., 2007). They utilize traditional methods of classroom instruction when implementing inquiry-based learning by practicing learn to inquiry-structure activities, which follow the sequence of inquiry, but in a teacher-directed way and with limited student engagement. This indicates that students might not experience the core of student-centered instruction in their classrooms, especially when teachers’ understanding is not aligned to theoretical constructs of student-centered instruction (Akkus et al., 2007).

In addition to teachers’ practical concerns, implementing student-centered instruction is difficult because teachers need to provide learning materials in accordance with individual students’ learning styles and culture: For example, some students prefer using technology when they learn new skills or ideas, while others prefer to learn through talking with peers. Teachers’ awareness of student diversity influences students’ learning through this type of instruction. This is because proper learning materials based on their understanding encourage students’ multicultural and harmonious interactions and collaboration in the classroom (Atwater, 1996). Therefore, teachers’ implementation of student-centered instruction requires understanding of sociocultural perspectives of learning (Banks, 1997; Barba, 1995). However, it could be very demanding to understand every single student’s characteristics well. It might be further burdensome for teachers to select and provide distinctive and proper educational intervention matching a students’ diverse background when implementing quality student-centered instruction (Oakes, 1990).

A review of literature on instructional approaches in relation to equity in education helps us hypothesize that student-centered instruction contributes to greater equity in learning mathematics and science. However, we also recognize that those contributions could be obscured depending on how teachers understand and implement such instruction. As discussed before, different educational contexts could further diversify the patterns of interactions between student-centered instruction and equity in education. Therefore, it is necessary to scrutinize how student-centered instruction moderates the relationships between students’ SES and their learning outcomes in various educational contexts.

Using the PISA data, the purpose of this study is to investigate how mathematics and science classrooms embrace equity in education. The specific questions guiding this research are: (a) Does student-centered instruction broaden or narrow the achievement gaps between low and high SES students across different educational contexts? And (b) are students from high or low SES expected to have better achievement through student-centered learning environments across different educational contexts?

Method

Country Selection

Many countries have participated in the PISA: 65 countries participated in the PISA 2012 and 72 countries did in the PISA 2015. Scrutinizing all countries would be difficult and interpreting results would be even more complex in a single study. Thus, we purposively select ten countries participating in both PISA 2012 and 2015 data—Brazil, Chinese Taipei, Finland, France, Korea, Norway, Peru, Qatar, Singapore, and the United States of America (USA; alphabetical order) for the following reasons: to reduce such complexity in analysis, to cover a wide range of achievement and equity measured in the PISA (OECD, 2013, 2016b), and to find patterns by international comparison.

Particularly, the combination of students’ performance and equity in education in Fig. 1 is used for this purposive selection of the countries (circled in Fig. 1). First, we consider countries with above-average achievement but below-average equity in either the PISA 2012 or 2015—this result in the selection of Chinese Taipei (Mathematics) and Singapore (Science). Second, we pay attention to above-average-achieving countries with above-average equity. Among such countries, Korea and Finland are selected. We then shift our focus to countries with average achievement in three levels of equity: France (below-average equity), USA (average equity), and Norway (above-average equity). Lastly, we select below-average-achieving countries with different degrees of equity—Peru (below-average equity), Brazil (average equity), and Qatar (above-average equity). Table 1 shows all selected countries with their equity and achievement levels. Although we have a sizable number of countries and students, we acknowledge potential limitations of generalizability due to the uniqueness of individual educational contexts and purposive selection.

Fig. 1
figure 1

Strength of the socioeconomic gradient mean performance in mathematics (PISA 2012, left; OECD, 2013, p. 27) and science (PISA 2015, right; OECD, 2016c, p. 218). The ten selected countries are circled

Full size image
Table 1 Ten selected countries for data analysis
Full size table

Variables

For students’ mathematical and scientific literacy, we utilize all sets of plausible values: five sets of plausible values for mathematics literacy in the PISA 2012 and ten sets of plausible values for scientific literacy in the PISA 2015 as prior studies recommended (e.g., Laukaitye & Wiberg, 2017). Additionally, “using [multiple] plausible values will provide an accurate representation of [the] underlying relationships” of background information (Foy, Brossman, & Galia, 2012, p. 3). For socioeconomic status, we use the variable called the index of economic, social, and cultural status (ESCS) in both the PISA 2012 and 2015 database, which “provides a comprehensive measure of student socio-economic background, based on parents’ education levels and occupational status and possessions in the home” (OECD, 2009, p. 32).

We select two different variables to represent student-centered instructional approaches: “teacher behavior: student orientation” (SO) for mathematics in the PISA 2012 and “inquiry-based science teaching and learning practices” (IB) for science in the PISA 2015. The selection of two different variables of the PISA is due to different focal disciplines in the PISA 2012 (Mathematics) and 2015 (Science). Detailed information about mathematics instruction is only available in the PISA 2012, while information about science instruction is only available in the PISA 2015.

Mathematical and Scientific Literacy.

Through the PISA 2012 to 2015, the OECD (2016a) defined mathematical literacy as an individual’s capacity to formulate, employ, and interpret mathematics in a variety of contexts. The capacity of formulate, employ and interpret mathematics is suggested by integrating mathematical modeling into the prior PISA frameworks. A half of the PISA items aim to measure students’ employing mathematical concepts, procedures, and reasoning. The two quarters of them focus on formulating stations mathematically and interpreting mathematical outcomes, respectively.

The meaning of scientific literacy is explained as the ability to conduct scientific investigation and engage in a decision-making process through argumentation (Cavagnetto, 2010; OECD, 2013). Students, who inquire about the natural phenomena, construct a valid experimental design to test a hypothesis, and make claims by providing evidence, are regarded as “a reflective citizen” (OECD, 2016a, p. 20) who fostered scientific literacy. In accordance with the definition of scientific literacy, the PISA measured students’ scientific literacy through problems that include analyzing socio-scientific contexts, understanding of science concepts, and evaluating scientific ideas using rigid scientific design.

Index of Economic, Social, and Cultural Status (ESCS).

To represent students’ socioeconomic status, the OECD (2017) constructs a composite score called the ESCS using principal component analysis (PCA). The ESCS is grounded on “the indicators parental education (PARED), highest parental occupation (HISEI), and home possessions (HOMEPOS) including books,” which are constructed from responses in the student questionnaire (OECD, 2017, p. 339). All participating countries equally contributed to the PCA for the estimation of the ESCS scores. However, to report the ESCS scores, this scale is “transformed with zero being the score of an average OECD student and one being the standard deviation across equally weighted OECD countries” (OECD, 2017, p. 340). Because the ESCS is considered as a representation of SES, we will use only the term SES in the rest of this article.

Teacher Behavior: Student Orientation (SO).

The variable labeled TCHBEHSO in the PISA 2012 database (SO in this article) is used for data analysis to investigate how student-centered mathematics instruction moderates the relationships between SES and mathematical literacy. The OECD (2013) built the SO from the questions in the students’ survey in Table 2. The SO variable represents students’ perceptions of teachers’ behaviors related to student-oriented instruction. Response categories are “Every lesson,” “Most lessons,” “Some lessons,” and “Never or hardly ever.” Responses to the survey questions are scored so that higher values of the SO indicate that students perceive such teachers’ behaviors more frequently. The SO is scaled to have an international metric with the mean of zero and the standard deviation of one. The range is from − 1.60 to 3.31.

Table 2 Item and statements for “Teacher Behavior: Student Orientation” in the PISA 2012 (cite) and “Inquiry-Based Science Teaching and Learning Practices” in the PISA 2015 (cite)
Full size table

Inquiry-Based Science Teaching and Learning Practices (IB).

To examine how student-centered science instruction moderated the relationships between SES and scientific literacy, the variable labeled IBTEACH in the PISA 2015 database (IB in this article) is used for data analysis. As seen in the description of the IB, this variable denotes students’ perceptions of how often teachers enact actions relating to inquiry-based instruction. The IB is constructed using the eight questions in Table 3. Higher values of the IB indicate teachers’ behaviors more frequently perceived by students. This variable also has an international metric with the mean of zero and the standard deviation of one. The range of this variable is from − 3.341 to 3.183.

Table 3 Descriptive statistics of the variables by country
Full size table

Data Description

The PISA target population is 15-year-old students enrolled in educational institutions in each country. The OECD (2014, 2017) collected the data for all participating countries (except for the Russian Federation) using a two-stage stratified sample design. Because our research interests are at the student level, this design requires the incorporation of student weights in data analysis. This application helps to ensure the representative of sampled students (OECD, 2017) and statistical results such as means and regression outputs in this research are weighted by “final trimmed nonresponse adjusted student weight” in the database. Descriptive statistics by country are reported in Table 3. Furthermore, we utilize Fay’s method of balanced repeated replication suggested in the PISA to estimate sampling variances (OECD, 2017). Using variables “final student replicate BRR-Fay weights” in the databases, we could overcome the issue that “the variance estimator can be unstable” relying on the sample design (OECD, 2017, p. 123).

Data Analysis

We conduct two steps of data analysis. At the first step, we apply a linear regression analysis separately to each country to describe how teachers’ student-centered instruction moderated the relationships between mathematical or scientific literacy and socioeconomic status. The linear model for each country includes mathematics or science literacy scores as a dependent variable, an intercept, single variables—SES and IB or SO—, and the interaction between the variables as independent variables in Eqs. 1 and 2. At the second step, we construct the specific linear models by substituting the SO or IB with − 2 (two-standard-deviation-below-mean), − 1, 0 (the international mean), 1, and 2 (two-standard-deviation-above mean) into the first-step outcomes. We exclude the case that SO is − 2 because this is out of the range of SO. Finally, we construct graphical representations of the linear models to show how the SO or IB moderated the relationships between SES and achievement in Fig. 2.

Fig. 2
figure 2

The relationships between SES and Achievement moderated by student-centered instruction in PISA 2012 and 2015

Full size image

The linear regression models of the PISA 2012 and 2015 data are evaluated based on the following equations:

$$ MATH=\left({\beta}_{\mathrm{m}0}+{\beta}_{\mathrm{m}1}\times \mathrm{SO}\right)+\left({\beta}_{\mathrm{m}2}+{\beta}_{\mathrm{m}3}\times \mathrm{SO}\right)\times \mathrm{SES}+{e}_m $$
(1)
$$ SCIE=\left({\beta}_{\mathrm{s}0}+{\beta}_{\mathrm{s}1}\times \mathrm{IB}\right)+\left({\beta}_{\mathrm{s}2}+{\beta}_{\mathrm{s}3}\times \mathrm{IB}\right)\times \mathrm{SES}+{e}_s $$
(2)

where MATH and SCIE represent the plausible values of mathematical and scientific literacy, respectively; βm0, βm1, βm2, βm3, βs0, βs1, βs2, and βs3 are regression coefficients; and em and es represent error terms.

Results

Linear model estimation output for mathematics and science are reported in Table 4 in the “Results” section. As seen in Eqs. 1 and 2, the constant terms, βm0 and βs0 represent expected scores of students with the average SES, which means SES = 0. The regression coefficients βm1 and βs1 indicate the degree of changes in students’ expected achievement scores with SES of zero as they perceived student-centered instruction more frequently (see Eqs. 1 and 2). In addition, βm2 and βs2 show the degree of changes in expected achievement scores as SES changes for students with the average SO or IB.

Table 4 Estimation results of mathematics and science linear models
Full size table

The values and significance of the regression coefficients βm3 and βs3 need to be interpreted with other regression coefficients collectively because those coefficients are related to the interaction terms of SO × SES and IB × SES. Positive values of βm3 and βs3 respectively indicate that the slope coefficients of SES of Eq. 1 and SES of Eq. 2 would become greater as SO or IB increases. If βm3 or βs3 is statistically significant, the relationships between SES and mathematical or scientific literacy depend on students’ perception of student-centered instruction. Brazil, Chinese Taipei, and Qatar show significant interaction terms in the results about mathematical literacy (p < 0.1 for all; see Table 4): the relationships between SES and mathematical literacy were moderated by the perceived frequency of student-oriented instruction in mathematics classrooms. The science literacy result of Peru and USA shows significant interactions between IB and SES with p values less than 0.1 (see Table 4). The relationships between performance and socioeconomic status of these two countries are contingent on students’ responses about the frequency of inquiry-based instructions.

As mentioned in the method section, we substitute IB or SO in the regression equations with the specific values of − 2, − 1, 0, 1, and 2 to reflect the perceived frequencies of IB or SO implementation in classroom, and so that the results would be more interpretable with graphical representations. Calculated linear graphs are displayed in Fig. 2, and all slopes of SES and intercepts are reported in Tables 5 and 6. In Fig. 3, nine generalized patterns are presented and ten countries’ mathematics and science patterns are categorized into each generalized pattern.

Fig. 3
figure 3

Summary of the findings in the ten countries. The nine patterns in this figure are generalized and specific slope coefficients and achievement levels for each country could differ from the generalized graphs. a Classifications heavily rely on visual representations in Fig. 2 than statistical significances in Table 4 while considering the interaction coefficients

Full size image

Mathematical literacy scores in all ten countries are likely to decrease as students perceive frequent student-oriented instructions in mathematics classrooms. However, the mathematical literacy gap between high and low SES students is likely to be reduced in two countries, Brazil and Qatar, which have significant interactions between SO and SES (pattern 7 in Fig. 3). Norway also shows this pattern throughout the graphs although such a significant interaction is not found. In these three countries, as students perceived student-oriented instruction more frequently, students with high SES are likely to decrease mathematics literacy scores and, in turn, the literacy gap between low SES and high SES narrow. Students across all SES levels in Finland, France, Korea, Peru, and USA (pattern 6) tend to have lower scores of mathematical literacy as the frequency of student-oriented instruction increased. On the other hand, the achievement gap is likely to be broaden in Chinese Taipei and Singapore (pattern 8): Students from low SES are expected to have lower achievement with more student-oriented instruction.

The results about scientific literacy in the PISA 2015 also show the two main patterns, patterns 6 and 7, compared to the results about mathematical literacy. Overall, students’ scientific literacy scores are generally expected to decrease as inquiry-based instruction is given more frequently. For the one unique case of Singapore (pattern 9), the relationships between students’ achievement in scientific literacy and SES are expected to remain in constant with more inquiry-based instruction. In terms of equity in education, we find that the achievement gap in scientific literacy between high and low SES levels decreases in Peru, Qatar, and USA (pattern 7). In these three countries, as inquiry-based instruction is offered more frequently, expected scores of students from high SES decrease while those from low SES seem to be remain unchanged. In the other six countries (Brazil, Chinese Taipei, Finland, France, Korea, and Norway for pattern 6), inquiry-based science instruction does not change the achievement gap in scientific literacy between high and low SES students as the expected scores of all SES groups decrease.

Discussion

This study aims to investigate the relationships between socioeconomic status and mathematical or scientific literacy of ten participating countries in the PISA 2012 and 2015. We explore whether inquiry-based or student-oriented instruction moderates those relationships by analyzing student data of ten countries covering wide ranges of achievement and equity levels in education. The literature review indicated that, in theory, student-centered instruction has the potential to improve equity in education, particularly by providing students from low SES equal opportunities to learn and understand mathematics and science within the classroom. However, it is also possible that those contributions vary depending on teachers’ understanding and implementation of student-centered instruction as well as on different educational contexts.

Some countries aligned with the theoretical assumption that achievement gaps become narrower when student-centered instruction was offered more frequently. This assumption was empirically shown with one pattern for mathematical and scientific literacy across different educational contexts—pattern 7 in Fig. 3. However, the reduced gap only comes from the decreasing scores of students from high SES: Brazil, Norway, and Qatar, for mathematical literacy; and in Peru, Qatar, and USA for scientific literacy. Equity in mathematical literacy is worsened in Chinese Taipei and Singapore (pattern 8): Student-centered instruction increased the mathematics achievement gap between groups of high and low SES students.

The results in Fig. 3 indicate that our theoretical hypotheses do not always reflect a move in a positive direction based on our findings. There is no country that reduced the gap by low SES students’ taking advantages of student-centered instruction while high SES students’ scores remained the same (pattern 1 in Fig. 3). Furthermore, in many countries, the more frequently student-centered instruction is implemented, the lower scores students across all SES levels are expected to achieve in both mathematical and scientific literacy.

The findings of this research suggest that doing more student-centered instruction might not be a panacea for relieving inequality in academic achievement. Although the findings do not always support the positive impacts of student-centered instruction, we do not attribute this failure to the problem of student-centered instruction itself. There are certainly successful cases in other studies as we hypothesize (e.g., von Secker, 2002) and we do not examine all countries in the PISA 2012 and 2015. Instead, we would like to suggest a few possible explanations for the results based on several issues in relation to the survey, implementation, and educational policies in different educational contexts.

First, in the PISA 2012 and 2015, student-centered instruction is examined through the student survey, not the teacher survey. The questions in the student survey (see Tables 2 and 3) could only explain a partial dimension of student-centered instruction in classrooms. It is possible to find varied results using other measurements for student-centered instruction, such as teachers’ evaluations of their own teaching practices. Student-centered instruction may be perceived differently depending on who responds to questions about the instruction. For example, with teachers’ self-evaluations on their instructional practices, von Secker’s (2002) reported a positive impact of inquiry-based instruction on students’ achievement. This suggests considerable cautions in interpretation of our findings.

Moreover, the survey questions in Tables 2 and 3 may not assess the quality of the instruction (Kirschner et al., 2006). These questions are rather about how often students experienced teachers’ behaviors related to student-centered instruction. Although frequency of student-centered instructional practice is certainly one indicator of high fidelity of implementation, the student survey does not address other important aspects of instruction: For instance, how students make decisions in their experimental design in science classrooms or how well mathematics tasks with high cognitive demands are implemented. Results might be different with qualitative information reflecting more dimensions of student-centered instruction.

Second, as discussed in the literature review, the benefits of student-centered instruction might not be observed due to teachers’ difficulties in understanding and implementing student-centered instruction. Student-centered instruction places a higher demand on teachers’ cognitive process, in that teachers must turn on their intellectual vigilance to provide active scaffolding and challenging questions to students. Such interactions in classrooms necessitate that teachers instantly modify knowledge about students as they make sense of what has happened and what is expected during lessons (Simon, 1995). Because of this complexity in student-centered instruction, “teachers ignore, resist, subvert, misinterpret, selectively adopt, or otherwise distort reformers’ intentions. Changes tend to be superficial, seldom penetrating the core of instructional practice” (Lefstein, 2008, p. 701). For example, if teachers misunderstand inquiry-based instruction in science classrooms as a “minimal guidance,” it is more likely to generate “abandoned students” or “behavioral freedom” in science class. In this misunderstood student-centered instruction, students are less likely to be engaged in conceptual understanding and argumentation even though students experience classroom discussions and hands-on activities.

Third, we pay attentions to similarities and differences among educational contexts at the national level, namely educational policies discussed in the PISA reports. For example, we find a similarity between the findings about mathematical and scientific literacy, and teacher responsibility for curriculum measured by the three dimensions as shown in the PISA report (OECD, 2016d): choosing textbooks, deciding which courses are offered, and determining course content.

In general, students in countries where teachers have more responsibilities for their curriculum (more than 50% of responsibility is on teachers: Finland, France, and Korea) tend to show lower performance across all SES levels in both mathematics and science when student-centered instruction is frequently implemented (see pattern 6 in Fig. 3). On the other hand, in the educational systems that allow relatively low autonomy for teachers in curriculum (about less than 40% of responsibility is on teachers: Brazil, Norway, Peru, Qatar and USA), more frequent student-centered instruction negatively influence on achievement of high SES students in mathematics or science while students with low SES are affected minimally. Singapore has a distinctive system as the school board has 40% of the responsibility for curriculum, which is unusually high compared to other participating countries in the PISA. And Singapore is the country in which more inquiry-based instruction in mathematics and science classrooms have minimal effects on equity in mathematical and scientific literacy. However, it should be noted that this parallel between our findings and curriculum policy is not perfect: there are simultaneously countries like Chinese Taipei in discord with this argument for mathematical and scientific literacy.

Discussions about the responsibility for curriculum encouraged us to reconsider teachers’ difficulties in instruction to speculate possible explanations for our findings of somewhat negative effects of student-centered instruction on students’ mathematical and scientific literacy in the PISA 2012 and 2015. The demanding process for teachers to prepare and implement this type of instruction could be exacerbated when they take more responsibility for curriculum. Unlike teacher-centered instruction, student-centered instruction has variations of implementation among teachers since it requires teachers’ active, simultaneous decision making through cognitive and epistemic negotiation (Akkus et al., 2007; Luft, 2001). Students are expected to do more than reproduce and repeat what they learned from teachers in student-centered instructional approaches. This means that teachers need to be experts of creating learning environments based on their understanding of national standards, school curriculums, students’ learning levels, and the instructional approaches (Luft, 2001). Furthermore, inquiry-based/student-oriented instruction is difficult to standardize as it entails dynamic modes and patterns based on teachers’ understanding on learners and school cultures (Keys & Bryan, 2001). There is no specific guide for student-centered instruction, which results in science teachers’ struggling to implement such instruction appropriately (Newman , Abell, Hubbard, McDonald, Otaala, & Martini, 2004). Given these challenging environments for teachers, heavy responsibility for curriculum might increase the probability that teachers misguide students’ learning through student-centered instruction.

This study suggests several paths for future research. The findings provide general ideas about the relationships among instruction, socioeconomic status, and achievement. To have a better understanding of the findings, additional research emphasizing a specific educational context or geographically adjacent countries in detail will be beneficial (e.g., Bos & Kuiper, 1999). Moreover, we are only able to speculate possible factors from similarities and differences among educational systems. Further investigations on possible influential factors (e.g., quality of student-centered instructions) would help better understand the influences of instructional behaviors on the relationships between socioeconomic status and achievement.

This study confirms suggestions provided by previous international comparison studies that policymakers and administrators should not blindly mimic others’ educational contexts that have shown positive outcomes (Hatano & Inagaki, 1998). The findings in this research demonstrate that understanding the relationships between socioeconomic status and achievement are not straightforward, but potentially related to sociological, pedagogical, and even psychological factors. Educational context should be accounted for by more international comparison studies before adopting educational practices in different countries. Then better understanding from those studies can help each country to improve students’ learning science and mathematics through student-centered instruction.