Student participation in private supplementary education: A comparative analysis of Japan,Korea,Shanghai,and the USA

Abstract

Private supplementary education, which refers to private lessons and learning outside of formal schooling purchased by families, has been widely practiced in East Asia. Its demand has grown even beyond East Asia, however, as educational competition for social mobility has intensified in many parts of the world. This global trend makes it important to determine who has greater access to private supplementary education and address the implications of the differential access for educational inequality. The current study compares how family socioeconomic status (SES) and students’ prior academic performance are related to their participation in private supplementary education in three East Asian societies—Japan, Korea, and Shanghai (China)—and the USA. Private supplementary education has existed on a substantial scale in Japan and Korea. The public and scholars have increased their concerns regarding the growing prevalence of private supplementary education in China. Although it has not been a major educational strategy, the demand for private supplementary education is rising in the USA as well, which can offer a useful insight into the global application of private supplementary education in contemporary educational systems. The focused comparisons across a small number of societies allow analyses of detailed patterns in each society in comparative perspective, thus moving beyond both single-country research and large-scale cross-national studies. The data for the current study come from 15-year-old students who participated in an international survey of student achievement, Programme for International Student Assessment (PISA) 2012. With a clearly defined measure of private supplementary education, PISA 2012 provides an exceptional opportunity for comparison. Drawing on data for 15-year-old students in four societies from PISA 2012, the linear probability models highlight heterogeneity among three East Asian societies. Japan is distinctive from Korea and Shanghai in terms of the influences of family SES on student participation in private supplementary education. Once school-fixed effects are taken into account, the strength of the relationship between family SES and private supplementary education in Japan is similar to the strength in the USA, which is weaker than the strength in Korea and Shanghai. Japan and the USA are also similar in that family structure is not significantly associated with private supplementary education, while students in non-two-parent families are significantly less likely to receive private supplementary education in Korea and Shanghai. The school-fixed-effects models also show that the within-school relationship between students’ prior academic performance and their participation in private supplementary education is generally negative. Japan, Shanghai, and the USA show a significantly negative relationship, while only Korea shows no significant relationship. Future research directions are suggested in the conclusion.

Keywords

Children education inequality stratification East Asia

Introduction

One of the main characteristics of East Asian education is the considerable level of student participation in private supplementary education (PSE)¹ activities after regular school hours (Kuan, 2011; Park, 2013; Stevenson and Baker, 1992). A significant share of K-12 students in Japan, South Korea (Korea, hereafter), and Taiwan (China) attend a private institute, called a juku, hakwon, or buxiban, which offers lectures by private instructors on major school subjects such as language arts, mathematics, and English. One-to-one or group private tutoring is another form of PSE activity. Given the nature of private investment and its potential consequences for students’ academic performance, student participation in PSE has important implications for educational inequality (Dang and Rogers, 2008; Matsuoka, 2018; Park et al., 2011). Therefore, in seeking to understand sources of educational inequality among students in East Asia, it is critical to examine the extent to which students’ participation in PSE varies across their socioeconomic backgrounds.

It is notable that PSE has recently grown in various parts of the world. In China, where the ‘private tutoring industry is booming’ (Chan, 2019), for instance, growing demand for additional lessons and learning after formal schooling has received attention from both media and academics, raising concerns about educational inequality (Chen, 2018; Zhang and Bray, 2017). Studies in the USA have increasingly paid attention to the roles of various forms of PSE activities, including commercial SAT/ACT test prep courses and private tutoring, in explaining social class and/or racial/ethnic differences in student outcomes (Buchmann et al., 2010; Byun and Park, 2012; Ho et al., 2019). Indeed, PSE has increasingly become a global trend beyond East Asia, although its prevalence varies considerably across countries. As a review highlights, in many parts of the world, increased competition for educational success, along with rising economic inequality and globalization, may have increased demand for students and families to gain a competitive edge through additional learning after regular school hours (Park et al., 2016). This global trend makes it important to address socioeconomic differentials in student participation in PSE from a cross-national, comparative perspective.

In this study, we compare participation in PSE among 15-year-old students in three East Asian societies—Japan, Korea, and Shanghai (China)—and in the USA. The comparative data from an international survey of student performance, Programme for International Student Assessment (PISA) 2012, offer an excellent opportunity for comparing students’ participation in PSE. In PISA 2012, only students in Shanghai participated within China, making it infeasible to address the issue at the national level. Despite this limitation on national representation in China, PISA 2012 is an exceptional source in that it is the only international dataset offering internationally comparable measures of student participation in PSE and other characteristics (see Park et al., 2016). A fundamental challenge in studying PSE is the problem associated with measurement: in most cases it is not feasible to identify whether supplementary education activities are private or provided by schools or communities for free (i.e., basically public in nature).²

Fortunately, PISA 2012 asked students to respond to indicate hours per week they ‘attend out of school classes organized by a commercial company, and paid for by [your] parents.’ In other words, the question clearly refers to ‘private’ supplementary education activities. Comparing the PISA 2012 question to the following question asked in another international student achievement survey, 2003 TIMSS (Trends in International Mathematics and Science Study) makes our point clear: ‘During this school year, how often have you had extra lessons or tutoring in mathematics that is not part of your regular class?’ This question may include extra lessons or tutoring offered by school teachers after regular school hours or even free tutoring services provided in a community.

Due at least partially to the lack of cross-nationally comparable measures of PSE activities, many existing studies of PSE focus on single-country cases that are well summarized in previous reviews on the topic (Bray, 2009; Dang and Rogers, 2008; Park et al., 2016). When scholars do compare PSE across countries, they tend to include many countries from large-scale international surveys of student achievement (e.g., Baker et al., 2001; Byun et al., 2018). These large-scale comparative studies highlight how student participation in PSE varies across countries according to some country-level characteristics. In this type of research, each country is included as a case for the regression analysis to establish cross-national relations, and therefore detailed analysis for each country, combined with comprehensive information on its educational contexts, is usually absent. In order to examine specific patterns of student participation in PSE within a country from a comparative perspective, we need a middle-level analysis with a limited number of countries for which we have thorough understanding of educational contexts. Studying the influences of family background on students’ academic test scores in China in comparison with those in the USA and Germany, Lyu et al. (2019) demonstrate the advantages of focused comparisons across three countries.

Given the comparably high prevalences of student participation in PSE in East Asian societies, it makes sense to examine similarities and differences in patterns of student participation across three of these: Japan, Korea, and Shanghai. The educational system in the USA is apparently different from East Asian education in many respects. However, given the rising demand for PSE in the USA, identifying US students’ patterns of participation, compared to their counterparts in East Asia, will provide a useful insight into the global application of PSE in contemporary educational systems. The manageably small number of societies involved can facilitate detailed comparisons of specific student characteristics that are associated with student participation in PSE.

In studying student participation in PSE, we focus on the extent to which family socioeconomic status (SES) is related to the increased likelihood of students taking advantage of PSE. As a composite of parental education, parental occupation, and home possessions, family SES measured in PISA 2012 indicates the overall socioeconomic and cultural status of a student’s family (OECD, 2014a).³ Considering that parents have to pay for the extra lessons that their children receive in a private institute, it would not be so surprising to see a positive relationship between family SES and student participation in PSE. However, it is not yet clear to what extent this relationship between family SES and student participation varies within East Asian societies as well as between East Asia and the USA. If PSE enhances students’ academic performance, a stronger relationship between family SES and student participation in PSE would suggest a stronger role of such private activity in educational inequality by family SES.

In addition to family SES, we are interested in assessing how students’ prior academic performance is related to participation in PSE. Previous studies distinguished two different purposes of PSE (Baker et al., 2001; Byun and Park, 2012; Ho et al., 2019). In some contexts, low-achieving students are more likely to take PSE to try to do as well as their high-achieving counterparts, while in other contexts, it is high-achieving students who are more likely to take advantage of PSE to advance even further. As this debate on PSE as a remedial or excelling strategy has been of central interest, we pay attention to the relationship between students’ academic performance and private supplementary education, and to cross-national variation therein. However, it is notable that investigation of this issue in the current study is fundamentally limited, given that in our cross-sectional data, student’s academic performance is measured concomitantly with participation in PSE. It is not straightforward to determine whether academic performance facilitates student participation in PSE or taking PSE influences students’ academic performance. This issue could be better addressed with longitudinal data that would follow the same students over time. Although we treat our measure of academic performance as prior academic performance in a manner similar to that of previous studies dealing with cross-section data (e.g., Baker et al., 2001), findings should be interpreted with caution.

Family SES, academic performance, and private supplementary education (PSE)

Given that families must pay for students to take PSE, a significant relation between family SES and students’ likelihood of participating in PSE is easily expected. Numerous studies across a wide range of contexts consistently show such a positive relationship (Dang and Rogers, 2008; Park et al., 2016). The positive association of family SES, however, is due not only to money but also to parental involvement. As some studies show, student participation in PSE requires extensive parental involvement: parents need to gather information on various options for PSE, carefully select a specific class that may fit their child’s needs, and constantly check how their child is doing in the class (Park et al., 2011). High-SES parents are likely to be more involved in children’s education than their low-SES counterparts, which may explain why high-SES students are more likely to use PSE than their low-SES counterparts.

Although family SES is expected to be positively associated with PSE, however, the strength of the relationship may vary across societies. On the one hand, a cultural explanation argues that educational inequality by family SES is lower in China (perhaps in East Asia in general) than in Western societies due to the influence of Confucian culture, which emphasizes the value of education and of effort over ability for educational success (Lyu et al., 2019; see also Liu and Xie, 2016). Examining students’ test scores across China, the USA, and Germany, Lyu et al. (2019) present empirical support for the cultural argument. If PSE mediates the relationship between family SES and students’ academic performance, the weaker relationship between family SES and students’ test scores may suggest a similarly weaker relationship between family SES and PSE in China (and other East Asian societies) than in Western societies.

On the other hand, the strength of the relationship between family SES and PSE may depend on how substantial and important PSE is as a family strategy for children’s educational success in a society. When PSE is not a critical activity for educational success or its relevance for educational success is declining, family SES may not make a big difference in student participation in PSE. In contrast, when PSE plays (or at least is perceived to play) an important role in educational success, gaps by family SES in student participation in PSE may be substantial. This leads to an expectation that the relationship between family SES and PSE should be weaker in the USA and Japan than in Korea and Shanghai. Although its relevance has increased in US education, the role of PSE as an educational strategy among middle-class families is still limited. Recent studies of Japanese education have shown a significant decline over time of in-school and out-of-school learning times, and perceived pressure to study among Japanese students, which suggests declining relevance of PSE in Japan (Komatsu and Rappleye, 2018; Rappleye and Komatsu, 2018).

The situation is quite different in Korea and Shanghai. As will be seen below, almost half of 15-year-old students participate in a class outside of the regular classroom run by a commercial company, suggesting the significant role of PSE in educational competition in the two societies. There is abundant evidence of significant associations of various socioeconomic backgrounds with family expenditures on or student participation in PSE in Korea (Byun, 2014; Jung and Lee, 2010; Kim and Park, 2010). A growing literature in China also reveals significant associations of household income and parental education with student participation in PSE (Liu and Bray, 2017; Zhang and Bray, 2017; Zhang and Xie, 2016).

It is notable, however, that previous studies rarely assessed the relationship between family SES and student participation in PSE within schools. Previous studies show that some school characteristics, such as school’s socioeconomic composition, are associated with the increased likelihood of students participating in extra lessons and learning outside of schools. Further developing work by Smyth (2009) identifying associations of schools’ socioeconomic composition and school types with individual students’ likelihood of participating in PSE in Ireland, Matsuoka (2015 and 2018) hypothesized that schools (and neighborhoods) being clustered with students from parents of high SES and high educational expectations in Japan should increase individual students’ participation in extra lessons and learning outside of formal schooling, net of student-level characteristics. In schools composed of students whose parents tend to have high SES and high educational expectations, educational competition, pressure, and norms may facilitate student’s participation in PSE.

This association of school characteristics, particularly socioeconomic composition, suggests that students from high-SES families are more likely to participate in PSE because they are more likely to attend a school primarily composed of students from high-SES families, which in turn increases student participation in PSE. Although two studies by Matsuoka (2015 and 2018) show that the individual-level relationship between family SES and PSE is still significant after including school (and neighborhood) SES, it is not yet clear whether the relationship between family SES and student participation in PSE within schools still exists. A better modelling strategy to assess the within-school relationship between family SES and student participation in PSE is to study the relationship with a school-fixed-effects model. This strategy is not appropriate for identifying specific school characteristics associated with student participation in PSE, in which most existing studies on school contextual effects on PSE have been interested. However, if the research interest is to assess the relationship between family SES and student participation in PSE within schools, controlling for any potential (either observed or unobserved) between-school differences with a school-fixed-effects model is a better strategy.

Turning to the relationship between students’ prior academic performance and their participation in PSE, Baker and his colleagues (2001) found that in 31 out of 41 countries, academically struggling students were more likely to take extra lessons/cram schooling in math than high-achieving students (‘remedial’ strategy). In only three countries high-achieving students were more likely to take such extra lessons than their low-achieving counterparts (‘enrichment’ strategy). Korea was one of those three countries. In a study of US students, Buchmann and her colleagues (2010) found a significantly negative relationship between students’ prior academic achievement and their attendance at commercial SAT courses. However, Byun and Park (2012) revealed a more complicated pattern by showing the positive relationship among East Asian and other Asian Americans.

It is notable that as in studies of family SES and PSE participation, most previous studies did not examine within-school relationships between prior academic performance and participation in PSE. It is possible that high-achieving students are more likely to attend a school that has more high-achieving students, which can increase a student’s participation in PSE, similar to the way in which schools clustered with students of high-SES and high-expectation parents are conducive to student participation in PSE. In other words, it may not be students’ prior academic performance per se but schools’ composition of high-achieving students that directly facilitates student participation in PSE. Therefore, it is critical to examine the relationship between prior academic performance and PSE within schools.

Data and methods

Data and samples

PISA surveyed 15-year-old students who are enrolled in schools, both private and public, regardless of grades. Therefore, depending on the timing of school enrollment and grade progression policy, PISA countries may differ in the grade distribution of 15-year-old students. Fifteen-year-old Japanese students in PISA 2012 were all in 10th grade. Ninety-four percent of 15-year-old Korean students in PISA were in 10th grade, while 6% and 0.2% were in 9th and 11th grade, respectively. Since only a handful of 15-year-old students in Korea were not in 10th grade, we excluded those in 9th or 11th grade from the analysis. However, the situation for Shanghai and the USA is different. PISA includes 40% and 54% of 15-year-old students who were in 9th and 10th grade, respectively, in Shanghai. A remaining 6% of students in Shanghai were distributed across 7th, 8th, 11th, and 12th grades. Those outside of 9th and 10th grades were excluded from the analysis. Finally, the US data in PISA include students across a wide range of grades. Eleven percent, 73%, and 16% of 15-year-old students were in 9th, 10th, and 11th grade, respectively. Students across those three grades were included. Only 0.2% of 15-year-old students who were in 8th or 12th grade were excluded in the USA. In short, the grade distribution of 15-year-old students is much narrower in Japan (with no variation at all) and Korea than in Shanghai and the USA. The cross-national differences in the grade distributions may reflect various factors, including the timing of the PISA survey conducted, the month of a new academic year, grade retention, and, possibly, student population heterogeneity by race/ethnicity and migrant status (OECD, 2014b, 2014c).

The analytic samples for the current study are further reduced. PISA asked the question about PSE to only a portion of students participating in the survey. By randomly distributing different survey questionnaires within each country, PISA includes about 66% of the total sampled students in each country who received the questionnaire with the item on PSE. Therefore, the analytic samples consist of 4222 and 3138 10th graders in Japan and Korea, respectively, as well as 3224 9th and 10th graders in Shanghai and 3306 from 9th through 11th graders in the USA. Among the total samples for the analysis, 212, 104, 23, and 116 students in Japan, Korea, Shanghai, and the USA, respectively, did not answer the question about private supplementary education even though they were asked for it. We exclude those with missing information on private supplementary education. Finally, a handful of students (42, 2, 3, and 13 in Japan, Korea, Shanghai, and the USA, respectively) who had missing information on family SES, which will be described below in detail, are excluded, In the end, final samples result in 3968 Japanese, 3032 Korean, 3198 Shanghai, and 3177 US students. Except for those cases with missing information on family SES, no case was excluded due to missing data.

Measures

As described above, the focal variable of the current study is whether the respondent attended ‘out-of-school classes organized by a commercial company, and paid for by parents’ for any school subject. Although the original question asked respondents to indicate hours of attendance each week, we distinguish respondents who spent any hour per week attending such out-of-school classes from their counterparts who did not attend these at all. In other words, we transform the original variable into a dichotomous variable indicating participation in PSE or not. The key independent variable predicting participation in PSE is family SES. Combining parents’ highest occupational status, parents’ highest level of education, and home possessions (wealth-related items, cultural items, and educational resources), PISA created an index of economic, social, and cultural status for each student (OECD, 2014a). We use the PISA index as a measure of family SES. Although the original PISA index was scaled to have a mean of 0 and a standard deviation of 1 across OECD students, we standardize the index to have a mean of 0 and a standard deviation of 1 within each society. With this standardization, a coefficient of the index indicates the change in the outcome per 1 standard deviation increase in family SES.

In estimating the relationship between family SES and student participation in PSE, we control for some students’ demographic characteristics, including gender, grade (in Shanghai and the USA), and family structure. Students are classified into two groups with respect to family structure: (a) two-parent families; and (b) non-two-parent families. The latter includes students with a single parent and others who do not live with any parent but someone else. We also take into account students’ academic performance. PISA administered tests on reading, mathematics, and science to measure students’ literacy levels (OECD, 2013). Instead of a single score, PISA provides five plausible values for each student to be used to estimate her/his latent score on each test. We average three scores for each student to measure her/his overall academic performance. As we already mentioned, our measure of academic performance is not ideal, as it was measured simultaneously with the outcome variable, PSE.

Table 1 presents descriptive statistics for all the variables used in analyses. We will discuss cross-national differences in the average percentage of students participating in PSE below in the result section. Although a standardized version of family SES is used in the analysis, in Table 1 the mean values in the original scale are presented. The value 0 indicates the OECD average, and the higher value indicates a higher level of family SES. On average, students in the USA have a higher level of family SES than students in the other three societies. Students in Shanghai have the lowest level of SES on average among our samples. However, they show the highest average test score, while students in the USA, who have the highest level of SES on average, display the lowest score.

Table 1.

Descriptive statistics of all the variables by country.

	Japan	Korea	Shanghai	USA
Participation on shadow education (%)	17.6	46.9	47.6	10.2
Family SES^a,b	−0.05	0.04	−0.28	0.20
Family SES^a,b	(0.71)	(0.74)	(0.93)	(0.97)
Test score^a	547.9	545.9	594.6	494.6
Test score^a	(84.7)	(83.0)	(78.4)	(85.5)
Grade (%)
9th grade	0.0	0.0	41.8	11.4
10th grade	100.0	100.0	58.2	71.6
11th grade	0.0	0.0	0.0	17.0
Gender: Female (%)	48.8	46.1	52.1	49.8
Family structure: Non-two-parent (%)	14.9	16.2	15.4	29.3
Sample size	3968	3032	3198	3177

^aFor this variable, means and standard deviations (in parenthesis) are presented.

^bThe standardized version of this variable is used in multivariate analyses.

Methods

Considering the dichotomous nature of the dependent variable—participation in PSE, we apply a linear probability model represented by the following equation:

\Pr (y = 1| X) = β_{0} + β_{1} F amilySES + β_{2} F emale + β_{3} T estScore + β_{4} T woParent + β_{5} G rade

(1)

An advantage of this linear probability model is that it can be easily estimated as the ordinary least squares (OLS), and interpretation of the coefficients is straightforward. For instance, in Equation (1), $β_{1}$ indicates the change in the probability of participating in PSE per one standard deviation increase in family SES. In addition to this simple model, we estimate the linear probability model with school-fixed effects. Instead of identifying specific school characteristics that are associated with participation in PSE, we estimate the relationships of individual and familial characteristics with private supplementary education within schools by applying school-fixed effects models. The coefficients of family SES and test scores, derived from school-fixed-effects models, indicate within-school relationships with participation in PSE once unspecified school characteristics as a whole are taken into account. The differences in the coefficients between Equation (1) and the school-fixed-effects model, moreover, indirectly reveal relevance of school characteristics (whatever those characteristics are) for students’ likelihood of using PSE.

Results

Prevalence of private supplementary education (PSE)

Figure 1 shows the overall percentage of 15-year-old students in PISA who attended a PSE class in each society. As is well documented in the media and literature, the share is comparably large in Korea (47%), which is similar to the level in Shanghai. The share of Japanese students who attended such a class is relatively small (18%). It is notable that despite a fairly specific definition of PSE as education organized by a commercial company and paid for by parents, 10% of US students attended a PSE class. In other words, although the share of 15-year-old students who engage in PSE in the USA is comparably small, it is by no means negligible. Although the timing of participation and the exact measure of the activity slightly differed, the estimate of about 10% is generally consistent with the finding that about 12% of 11th graders in the USA indicated participation in ‘academic instruction outside of school such as a Saturday Academy, learning center, personal tutor or summer school program’ (ever since 9th grade) (Ho et al., 2019).

Figure 1.

Percentage of 15-year-old students attending a private supplementary education (PSE) class.

Family SES, test scores, and private supplementary education (PSE)

Before turning to the results of multivariate analysis, we present bivariate correlations among family SES, test scores, and private supplementary education to see the overall relationships. Matsuoka (2015) provides a similar correlation matrix of PSE participation among Japanese students.⁴ Note that the variable of private supplementary education is binary. Table 2 shows that in Japan, Korea, and Shanghai, there is a positive relationship between family SES and PSE. The relationship is positive in three East Asian societies, while the corresponding relationship in the USA is tenuous but negative. Another interesting difference between the three East Asian societies and the USA is the relationship between test scores and PSE. In the three East Asian societies, high test scores are associated with the increased likelihood of attending a PSE class, while the corresponding relationship is negative in the USA. In short, in the three East Asian societies, more affluent and academically stronger students are more likely to take a PSE class than their counterparts, but in the USA the opposite pattern emerges.

Table 2.

Correlations among private supplementary education (PSE), SES, and test score.

Japan				Korea
	PSE	SES	Test score		PSE	SES	Test score
PSE	1			PSE	1
SES	0.153	1		SES	0.233	1
Test score	0.063	0.305	1	Test score	0.160	0.314	1
Shanghai				USA
PSE	1			PSE	1
SES	0.171	1		SES	-0.054	1
Test score	0.069	0.386	1	Test score	-0.301	0.389	1

Note: all the correlation coefficients are significantly different from 0 at 0.01 level.

Now we move to the results of multivariate analyses. Table 3 presents the results of linear probability models without (M1) and with school-fixed effects (M2). First, in Model 1, family SES is significantly associated with the increased probability of taking a PSE class in all four societies, after taking into account other student characteristics. However, the strength of the relationship varies across societies. In the USA, a one-standard-deviation increase in family SES leads to a 2.3% change in the probability of attending a PSE class, while the corresponding changes in Korea and Shanghai are almost 10% and 8%, respectively. The strength in Japan is between these two (5.5%). In other words, family SES is likely to be more strongly related to PSE in Korea and Shanghai than in the USA and (to a lesser extent) Japan.

Table 3.

The linear probability model of private supplementary education participation (PSE).

	Japan		Korea		Shanghai		USA
	M1	M2	M1	M2	M1	M2	M1	M2
		School-fixed		School-fixed		School-fixed		School-fixed
SES	0.055***	0.029***	0.097***	0.066***	0.084***	0.067***	0.023***	0.029***
Female (vs. male)	0.004	0.004	0.012	0.017	0.075***	0.071***	−0.006	−0.008
Test score (in 100 points)	0.007	−0.068***	0.055***	0.003	0.014	−0.068***	−0.108***	−0.103***
Non-two-parent family	−0.025	−0.013	−0.081***	−0.097***	−0.052**	−0.046*	−0.016	−0.020
(vs. two-parent family)
Grade (vs. 10th grade)
9th grade					0.173***	0.111**	0.063***	0.068**
11th grade							0.028*	0.008
Intercept	0.137**	0.680***	0.172**	0.386***	0.288***	1.025***	0.625***	0.505***
R-squared	0.024	0.174	0.066	0.198	0.064	0.151	0.104	0.171

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

Once family SES and other student characteristics are controlled for, students’ (prior) academic performance is not significantly related to the probability of attending a PSE class in Japan and Shanghai. Consistent with previous literature, the relationship is positive in Korea, suggesting that high-achieving students are more likely to use private supplementary education to further excel (Baker et al., 2001). Also, as in previous studies, the relationship is significantly negative in the USA, (Buchmann et al., 2010) where, in contrast with Korea, it is low-achieving students who are more likely to use PSE.

How do these relationships of family SES and prior academic performance with PSE change once school-fixed effects are taken care of? After we control for potential between-school differences in socioeconomic and academic composition and other unobserved school characteristics, are more affluent and higher-achieving students still more likely to attend a PSE class than their less affluent and lower-achieving counterparts? Model 2 in Table 3 shows that the within-school relationship between family SES and PSE is still significantly positive in all four societies, although the strength is substantially reduced from the strength in Model 1 in each of the three East Asian societies. In Japan, the coefficient of family SES, for instance, decreases from 0.055 in Model 1 to 0.029 in Model 2. This reduction in the coefficient indicates that taking into account school effects is important for understanding the positive relationship between family SES and individual students’ probability of attending a PSE class. However, the still significant relationship of family SES in Model 2 suggests that even within schools, more affluent students are more likely to attend a PSE class than their less affluent peers.

With substantial reduction of family SES coefficients with school-fixed effects in the three East Asian societies but no change in the USA, cross-national differences in the strength of the relationship between family SES and private supplementary education become relatively less than corresponding differences in Model 1. The within-school relationship in Japan becomes similar in strength to that in the USA. However, Korea and Shanghai still show a stronger relationship than the USA and Japan. In both Korea and Shanghai, the probability of attending a PSE class increases about 7% per one standard deviation increase in family SES within schools. In Japan and the USA, the corresponding increase is about 3%.

In school-fixed-effects models, the relationship between prior academic performance and PSE appears different from the relationship without school-fixed effects being controlled for. In Japan, the tenuous relationship between prior academic performance and private supplementary education in Model 1 now turns out to be significantly negative in Model 2. Shanghai shows the same change between Models 1 and 2. The significantly positive relationship in Model 1 in Korea now becomes insignificant in Model 2. The finding suggests that the positive relationship between prior academic performance and PSE, often found in previous studies (e.g., Baker et al., 2001), may exists because they failed to control for school effects. High-achieving students can be more likely to attend schools with a high concentration of high-achieving students, which may lead to students’ increased use of PSE. However, once such differences across schools are taken into account, prior academic performance may be no longer associated with private supplementary education within schools in Korea. As with family SES, the result for the USA does not change substantially after school-fixed effects are controlled for.

As our focus is on family SES and prior academic performance, we have hardly discussed other student-level characteristics. Gender differences in participation in PSE are minimal except in Shanghai, where female students are significantly more likely to attend a PSE class than their male counterparts, once other student-level variables are taken into account. Along with a relatively stronger relationship between family SES and PSE in Korea and Shanghai, living in a non-two-parent family is also significantly related to the diminished probability of attending a PSE class in Korea and Shanghai. But in Japan and the USA, family structure is not significantly associated with student participation in PSE. In both Shanghai and the USA, 9th graders are more likely to use private supplementary education than 10th graders.

Conclusion

As discussed in the introduction, the demand for private supplementary education (PSE) is rising in many parts of the world as educational competition for social mobility is intensified. The changing educational contexts across the globe make it important to understand which groups are more likely to use PSE and what the implications of such differential access to PSE are for educational inequality. The current study has taken a cross-national comparative approach in examining how more affluent and higher-achieving students differ from their less affluent and lower-achieving counterparts in access to PSE by comparing the patterns in three East Asian societies and the USA. With focused comparisons across a small number of societies, our study goes beyond single-country studies by offering a comparative perspective without losing detailed understanding of each society, as often happens in large-scale international comparative studies. Of course, our comparative strategy has its own limitations. With only four societies, we are not in a position to systematically link between-society differences to any country-level characteristics, which is the advantage of large-scale comparative research.

However, our study reveals interesting similarities and differences across Japan, Korea, Shanghai and the USA, which can be useful to formulate potential hypotheses that can be better tested later by large-scale comparative studies. First of all, although the high prevalence and long-established practice of PSE in East Asia is well known, our three societies within East Asia show different patterns of student participation in PSE. In particular, Japan is distinctive from Korea and Shanghai. Once school-fixed effects are taken into account, the strength of the relationship between family SES and student participation in PSE in Japan is similar to the strength in the USA, which is weaker than the strength in Korea and Shanghai. Even without school-fixed effects, Korea and Shanghai show a relatively stronger relationship than the USA and Japan. Japan and the USA are also similar in that family structure is not significantly associated with PSE, while students in non-two-parent families are significantly less likely to take a PSE class in Korea and Shanghai.

In the hypothesis section, we indeed expected such cross-national differences in the strength of the influence of family SES based on the argument that the strength of the relationship between family SES and PSE should depend on how important PSE is for educational competition in each society. When PSE is or is perceived to be a key educational strategy to gain a competitive edge, middle-class parents will utilize it as much as they can, and thus there should be a substantial difference in access to PSE by family SES. When PSE is still not a major strategy for middle-class parents or its relevance for educational competition is declining, we expected a somewhat limited differential in access to PSE by family SES. With only four countries, we are not able to assess whether the importance and relevance of PSE in an educational system shapes the way in which family SES influences students’ participation in it. Future research may be able to test our hypothesis based on large-scale, cross-national data, following the work by Baker et al. (2001) and Byun et al. (2018).

Our school-fixed-effects models show that the within-school relationship between students’ prior academic performance and their participation in PSE is generally negative. Japan, Shanghai, and the USA show a significantly negative relationship, while only Korea shows no significant relationship. This result is consistent with Baker et al. (2001), who showed that the remedial strategy for low-achieving students is the dominant purpose of PSE in the majority of countries analyzed. Throughout the paper, we have emphasized the limitation of the academic performance measure due to the nature of cross-sectional data. Future research should assess a within-school relationship between prior academic performance and PSE using better cross-national data that include straightforward measures of students’ prior academic performance.

Notably, we have not addressed other dimensions of student background, such as race/ethnicity, immigrant status, and rural vs. urban division in differential access to PSE. In other words, our investigation of family SES and prior academic performance in relation to PSE may not well reflect varying patterns for different socioeconomic and demographic groups within societies. Internal differences in family resources, and thus possibly the likelihood of taking PSE by race/ethnicity, immigrant status, and urbanity, deserve more systematic investigation (Ho et al., 2018; Gao, 2014). PSE can be an important mechanism through which educational disparities by race/ethnicity, immigrant status, and urbanity are reproduced.

Footnotes

Acknowledgement

We thank James M. Raymo and Hideo Akabayashi for their helpful comments and suggestions.

Declaration of conflicting interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Hyunjoon Park acknowledges support from the Laboratory Program for Korean Studies through the Ministry of Education of the Republic of Korea and Korean Studies Promotion Service of the Academy of Korean Studies (AKS-2016-LAB-2250002).

Notes

ORCID iD

Hyunjoon Park

References

Baker

Akiba

LeTendre

Wiseman

(2001) Worldwide shadow education: Outside-school learning, institutional quality of schooling, and cross-national mathematics achievement. Educational Evaluation and Policy Analysis 23(1): 1–17.

Bray

(2009) Confronting the Shadow Education System: What Government Policies for What Private Tutoring? Paris: International Institute for Educational Planning.

Buchmann

Condron

Roscigno

(2010) Shadow education, American style: Test preparation, the SAT and college enrollment. Social Forces 89(2): 435–462.

Byun

(2014) Shadow education and academic success in Republic of Korea. In: Park

Kim

(eds) Korean Education in Changing Economic and Demographic Contexts. Singapore: Springer, pp. 39–58.

Byun

Chung

Baker

(2018) Global patterns of the use of shadow education: Student, family, and national influences. Research in the Sociology of Education 20: 71–105.

Byun

Park

(2012) The academic success of East Asian American youth: The role of shadow education. Sociology of Education 85(1): 40–60.

Chan

(2019) China’s private tutoring industry is booming despite economic slowdown. South China Morning Post 25 March.

Chen

(2018) An exam today determines 10 million Chinese students’ futures. It’s fueling a more than $120 billon market. CNBC 7 June.

Dang

Rogers

(2008) The growing phenomenon of private tutoring: Does it deepen human capital, widen inequalities, or waste resources? The World Bank Research Observer 23(2): 161–200.

10.

Gao

(2014) China’s education gap. New York Times 4 September.

11.

Park

Kao

(2019) Racial and ethnic differences in student participation in private supplementary education activities. Research in Social Stratification and Mobility 59: 46–59.

12.

Jung

Lee

(2010) The determinants of private tutoring participation and attendant expenditures in Korea. Asia Pacific Education Review 11(2): 159–68.

13.

Kim

Park

(2010) The determinants of demand for private tutoring in South Korea. Asia Pacific Education Review 11(3): 411–21.

14.

Komatsu

Rappleye

(2018) Is exam hell the cause of high academic achievement in East Asia? The case of Japan and the case for transcending stereotypes. British Educational Research Journal 44(5): 802–26.

15.

Kuan

P-Y

(2011) Effects of cram schooling on mathematics performance: Evidence from junior high students in Taiwan. Comparative Education Review 55(3): 342–368.

16.

Liu

Xie

(2016) Why do Asian Americans academically outperform whites? The cultural explanation revisited. Social Science Research 58: 210–26.

17.

Liu

Bray

(2017) Determinants of demand for private supplementary tutoring in China: Findings from a national survey. Education Economics 25(2): 205–18.

18.

Lyu

Xie

(2019) The influences of family background and structural factors on children’s academic performances: A cross-country comparative study. Chinese Journal of Sociology 5(2): 173–92.

19.

Matsuoka

(2015) School socioeconomic compositional effect on shadow education participation: Evidence from Japan. British Journal of Sociology of Education 36(2): 270–90.

20.

Matsuoka

(2018) Inequality in shadow education participation in an egalitarian compulsory education system. Comparative Education Review 62(4): 565–86.

21.

Mori

Baker

(2010) The origin of universal shadow education: What the supplemental education phenomenon tells us about the postmodern institution of education. Asia Pacific Education Review 11(1): 36–48.

22.

Organisation for Economic Co-operation and Develpopment (OECD) (2013) PISA 2012 Assessment and Analytical Framework: Mathematics, Reading, Science, Problem Solving and Financial Literacy. Paris: OECD.

23.

OECD (2014a) PISA 2012 Results: Excellent through Equity - Giving Every Student the Chance to Succeed. Paris: OECD.

24.

OECD (2014b) Are disadvantaged students more likely to repeat grades? PISA in Focus, No. 43. Paris: OECD Publishing. Available at: https://doi.org/10.1787/5jxwwfp1ngr7-en (accessed April 24, 2018).

25.

OECD (2014c) PISA 2012 Technical Report. Paris: OECD.

26.

Park

(2013) Re-Evaluating Education in Japan and Korea: Demystifying Stereotypes. New York: Routledge.

27.

Park

Buchmann

Choi

Merry

(2016) Learning beyond the school walls: Trends and implications. Annual Review of Sociology 42: 231–52.

28.

Park

Byun

Kim

(2011) Parental involvement and students’ cognitive outcomes in Korea: Focusing on private tutoring. Sociology of Education 84(1): 3–22.

29.

Rappleye

Komatsu

(2018) Stereotypes as Anglo-American exam ritual? Comparisons of students’ exam anxiety in East Asia, America, Australia, and the United Kingdom. Oxford Review of Education 44(6): 730–54.

30.

Smyth

(2009) Buying your way into college? Private tuition and the transition to higher education in Ireland. Oxford Review of Education 35(1): 1–22.

31.

Stevenson

Baker

(1992) Shadow education and allocation in formal schooling: Transition to university in Japan. American Journal of Sociology 97(6): 1639–1657.

32.

Zhang

Bray

(2017) Micro-neoliberalism in China: Public-private interactions at the confluence of mainstream and shadow education. Journal of Education Policy 32(1): 63–81.

33.

Zhang

Xie

(2016) Family background, private tutoring, and children’s educational performance in contemporary China. Chinese Sociological Review 48(1): 64–82.