Heterogeneous relationships between family private education spending and youth academic performance in Korea

Children and Youth Services Review 69 (2016) 136–142

Contents lists available at ScienceDirect

Children and Youth Services Review journal homepage: www.elsevier.com/locate/childyouth

Heterogeneous relationships between family private education spending and youth academic performance in Korea☆ Yoonsun Han, Seonglim Lee ⁎ Sungkyunkwan University, Republic of Korea

a r t i c l e

i n f o

Article history: Received 18 March 2016 Received in revised form 1 August 2016 Accepted 1 August 2016 Available online 5 August 2016 Keywords: Private education Family spending Quantile regression South Korea Youth

a b s t r a c t Private education in Korea, a country with one of the largest private education systems in the world, is a doubleedged sword: it is culturally and socioeconomically indispensable to the youth, but financially onerous to the family. An in-depth examination of this topic is warranted due to inconsistency in the relationship between informal education and youth academic outcomes. The current exploratory study attempted to clarify this ambiguous relationship between family spending on private education and self-reported academic performance scores using quantile regression techniques and a nationally representative sample of Korean youth (N = 2,120). The methodological advantage of using quantile regression is that it allows exploration of whether the magnitude of the association between family spending on private education and academic scores differs across quantiles of academic performance. The concave representation of relationship sizes across the distribution of academic scores indicated that the magnitude is greatest around the median of the academic performance distribution, whereas the size of the coefficient is smallest at the extreme ends. In other words, reliance on mean-based results may have masked the full picture of the heterogeneous association between family spending in private education and youth academic performance. Our findings may help direct targeted strategies for improving youth academic outcomes. © 2016 Elsevier Ltd. All rights reserved.

1. Introduction Educational investment is critical, particularly in adolescence, as academic performance during this period determines youth self-concepts, educational aspirations, and occupational choices. Investment in education may occur formally through the formal school system (e.g., middle school, high school) and informally outside the school through the informal private education system (e.g., one-on-one private tutoring, private learning institutions, private test preparation services, after-school cramming schools). Although their relative importance differs across cultures and contexts, traditionally, the formal school system has been a major source of education. Recently, however, the private education system has garnered growing attention due to its increasing role in youth development. In many societies, particularly in Asia, private education is at least as important as the formal school system. With its growing salience, however, private education has been considered a double-edged sword: it is culturally and socioeconomically indispensable to the youth, but it is financially onerous to the family. Further

☆ Acknowledgements: This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2014S1A5A2A03065251). ⁎ Corresponding author at: Department of Consumer and Family Sciences, Sungkyunkwan University, 25-2 Sungkyunkwan-ro, Jongno-gu, Seoul, 110-745, Republic of Korea. E-mail address: [email protected] (S. Lee).

http://dx.doi.org/10.1016/j.childyouth.2016.08.001 0190-7409/© 2016 Elsevier Ltd. All rights reserved.

investigation is warranted to determine whether the relationship between private education and academic performance is significant and if so whether the size of this association is heterogeneous. 1.1. Formal school education vs. informal private education In most industrialized societies and a large proportion of developing countries, formal schools provide an influential social environment in which youth spend the bulk of their time. Formal school institutions are designed to perform a broad range of functions related to youth development in academic, psychological, and social domains. Thus, schools are a key arena for the education of youth in most societies (Steinberg, 2014). It is not surprising that a large corpus of research has examined the effects of various aspects of formal education on academic achievement, such as public spending (Wenglinsky, 1998), classroom size (Kwon, 2003; Levin, 2001), public vs. private (Kim, 2012; Witte, 1992), and teacher quality (Chung, Lee, & Kim, 2014; Wayne & Youngs, 2003). Investment in education may occur outside the formal school system as well. Parents and youth invest in informal education in various forms, including one-on-one private tutors, small group tutoring, internet tutoring, and private test-preparation or learning institutions, also known as cram schools (Bray, 2009; Byun, 2014). The research literature has coined the term shadow education to describe private instruction in the informal education system because it mimics and

Y. Han, S. Lee / Children and Youth Services Review 69 (2016) 136–142

supplements the formal school system (Manzon & Areepattamannil, 2014). The literature defines shadow education as “a set of educational activities that occur outside formal schooling and are designed to enhance the student's formal school career” (Stevenson & Baker, 1992 p.1639). Instead of providing opportunities for development in a broad range of areas, private informal educational institutions primarily focus on strengthening youth performance in a specific targeted area of study, usually in key academic areas. However, in East Asian societies such as Korea, Japan, China, and a growing number of societies elsewhere around the globe, private education is an integral component of the educational sector (Bray, 2009; Manzon & Areepattamannil, 2014; Stevenson & Baker, 1992; Yang, 2012). Despite the growing role of the private education system (Bray, 2009), the research literature has paid disproportionally greater attention to formal education. Few studies have shown that participation in private education is associated with modest improvements in academic performance, but others have identified no significant relationship (Choi, 2008; Dang, 2007; Lee, Lim, & Min, 2010; Liu, 2012). In some cases, study results even point to an inverse relationship between private education and academic outcomes (Cheo & Quah; Yang, 2012). To date, the mixed empirical evidence has prevented consensus about the effects of private education (Byun, 2014). Due to the relative dearth of research about family spending in private education and inconsistency in the association between private education and outcomes, an indepth examination of this relationship is warranted. The current study used quantile regression to clarify the ambiguous relationship between family spending in private education and academic performance among a nationally representative sample of Korean youth. Investment in Korea's informal education system, one of the largest in the world (Bray, 2009), is equivalent to 34% of the annual governmental budget allocated to formal education (Statistics Korea, 2014a; 2014b). Driven by cultural, social, and economic demands, approximately 70% of all school-aged children currently participate in the private educational system (Statistics Korea, 2014a). Given the scale and intensity of private education, Korea provides an ideal context for thoroughly examining the relationship between family spending in private education and youth academic outcome (Byun, 2014).

137

In addition to the cultural significance of education within the family, driven by Confucian norms, features of Korea's current formal education system at the structural level reinforce youth participation in the private education system. Some of those conditions are centrally administered formal examinations for college entry; intense competition to enter prestigious universities; and clear links between exam scores, university acceptance, and future educational, occupational, and status opportunities (Stevenson & Baker, 1992). These characteristics have led Korean families to firmly believe that the college their child attends is directly linked to future employment opportunities and, more broadly, social and economic success. Given the cultural emphasis and the enormous lifetime payoffs of educational outcomes, it is unsurprising that Korea has one of the largest private education industries in the world (Bray, 2009). Despite its importance, spending in private education has surfaced as a great financial burden to families. In Korea, the average monthly family spending on private education is approximately 350,000 KW (US$ 350) among 8- to 19-year-olds participating in private education (Statistics Korea, 2014a). This amount of spending is not small, as reflected in a national survey of Korean youth and their primary caregivers, in which nearly 70% of parents reported that the costs for private education are a burden to the family budget (Ministry of Gender Equality and Family, 2014). In sum, private education in Korea is imperative to youth academic development, but also burdensome to the family. The magnitude of the financial stress imposed by private education is so large that this family-level issue has been treated as a major national economic concern in the past several presidential campaigns (Yang, 2013). At the same time, however, families view private education as not a supplement or “shadow” to formal education, but rather as a necessity for academic advancement (Ihm, Woo, & Chae, 2008; Yang, 2006). Some youth rely more on after-school cram classes for academic advancement than on the formal school system (Kim & Kim, 2002). Clearly, family spending in private education is a crucial part of the lives of youth in Korea, with implications that expand beyond families to the national level in such areas as socioeconomic stratification and inequality (Yang, 2006). Hence, furthering the overall understanding of the role of private education is critical. 1.3. Mixed evidence on the role of education

1.2. Investment in private education: Korean context At this time, private education is an indispensable part of the daily lives of children and parents, nearly equal with the formal education system in Korea (Yang, 2012). According to a recent government report on education spending (Statistics Korea, 2014a), approximately 70% of school-aged children and youth participate in private education for an average of 5.8 h per week. The primary purpose of participating in the private education system is to increase academic scores in a specific area of study on core subjects, such as mathematics and English (Statistics Korea, 2014a). In this sense, private education in Korea practices a narrow concept of education—to enhance academic performance in basic subjects over a short period of time using test-prep techniques. This is in contrast to the holistic definition of education—to advance intellectual knowledge and analytic skills, as well as encourage growth in personality, social capacity, critical thinking, citizenship, and culture—that is often adopted in the formal school system. Cultural values for educational success, intertwined with socioeconomic conditions in Korean society may account for this phenomenon. Traditional Confucian values that exalt literary scholarship and its central role in social mobility (Seong, 1993) have bred a culture in which families possess high educational aspirations and are committed to prioritizing their children's education (Zhou & Kim, 2006). As educational advancement is considered to be the most effective means of getting ahead in society (Zhou & Kim, 2006), Korean families do not hesitate to invest their resources in the private educational system (Lee & Kwon, 2011).

Subjective evaluations of the benefits of private education suggest that, in general, both Korean parents and youth hold a favorable view of the effect of private education. Among primary and secondary school students, 70–80% reported experiencing better school grades with family spending (Kim, 2001; Kim, 2003). Although there is a significant relationship between private spending and its perceived benefits, whether investment in private education is actually linked with those anticipated effects is unclear. Despite large amounts of spending in youth education with the consequent burden to families, no strong empirical evidence confirms the anticipated academic benefits of private education (Yang, 2012). A careful review of the literature suggests that the association between private education and academic performance is mixed and not well-established (Byun, 2014; Kim, 2003; Kim & Kim, 2002). Some research has identified a positive relationship between private education and academic attainment (Yang, 2013) and achievement (Kim & Lee, 2011; Lee & Lim, 2009; Sang & Baek, 2005) among youth. For example, in a nationally representative sample of first-year middle school students, family investment in education was associated with high math and English test scores even after controlling for a host of individual background characteristics, teaching quality, and individual learning activities (Lee & Lim, 2009). Conversely, another branch of research has failed to identify a strong link between private education and academic performance (Choi, 2008; Lee, Kim, & Yoon, 2004; Lee et al., 2010). For instance, in a national study of high school students, the statistical significance between investment in private education and entrance to

138

Y. Han, S. Lee / Children and Youth Services Review 69 (2016) 136–142

college disappeared after controlling for family background characteristics such as household income and parental education (Choi, 2008). Some researchers reported inconsistent results within a single study. In a nationally representative panel study of urban Korean families, Yang (2012) found that the probability of attending a four-year college versus a two-year college increased with the number of years youth participated in private education. That same study, however, found no effect on the likelihood of attending a four-year college versus no college and a lower probability of attending a two-year college versus no college with an increasing number of years in private education (Yang, 2012). In some cases, private education was only effective for educational attainment or achievement and was restricted to specific academic subjects. Private education for the Korean and English languages showed no significant link with attending a high-ranked college or college entrance exam performance, but private education in mathematics was significantly associated with college entrance exam scores (Kim & Kim, 2009). Findings in Korea concerning the ambiguous role of family spending in education are consistent with those in other populations around the globe (Byun, 2014). Participation in shadow education increased the chance of attending university in Japan (Stevenson & Baker, 1992), cram school participation improved students' analytic ability and mathematics scores in Taiwan (Liu, 2012), and spending on private tutoring had a positive effect on academic ranking in Vietnam (Dang, 2007). On the other hand, the average effect of private tutoring was not significant in China (Zhang, 2013). Moreover, private tutoring negatively affected school grades in Singapore, possibly through mechanisms of over-investment and excessive studying in concentrated academic subjects (Cheo & Quah, 2005). 1.4. Limitation of “average” information The inconclusive results on the relationship between private education and academic performance across multiple studies may result from the variety of statistical models used across studies (Byun, 2014). In identifying the associations between various predictors of academic outcomes, including private education and academic performance, empirical studies have primarily used traditional regression methods such as ordinary least squares (Lee & Lim, 2009), logistic analysis (Stevenson & Baker, 1992; Yang, 2012), and instrumental variables models (Dang, 2007; Zhang, 2013). The issue of using relationship parameters estimated from those analytic methods, however, is that they all represent mean relationships (Koenker & Bassett, 1978). That is, parameter estimates derived from those models reflect the association between private education and the average academic performance level in the analytic sample. Although information identified by a conditional mean of academic performance is important and meaningful, it is limited in providing the full picture (Haile & Nguyen, 2008). If heterogeneous or asymmetric patterns exist, a single estimation of the relationship size at the mean can be misleading. In an attempt to explain the inconsistent findings on the relationship between private education and academic performance, we used a statistical model—quantile regression—that allows the estimation of asymmetric relationship sizes across the distribution of academic performance. To date, very few studies have used quantile regression to examine the relationship between family spending in private education and academic outcomes. Empirical evidence supports the idea that the marginal returns on investments in private education to academic performance may differ at various points in the distribution of academic performance. In China, for instance, there was a positive effect only at the lower end of the English test score distribution among urban students (Zhang, 2013). Similarly, Lee and Kwon (2011) found that returns from private education were greatest among the low-performing group, and the relationship size gradually decreased at higher scores. Furthermore, studies examining the link between health and academic achievement (Eide, Showalter, & Goldhaber, 2010) and family

background and academic attainment (Haile & Nguyen, 2008) indicate the importance of exploring different relationship sizes at multiple points in the distribution of academic outcomes. In sum, the current study attempts to examine the relationship between family spending in private education and youth academic performance using conventional ordinary least squares regression and further explores the possibility of differential relationships between private expenditure and the distribution of youth academic scores using quantile regression methods drawing on a nationally representative sample of Korean youth. The current study may provide support for the idea that it is not only important to identify whether an increase in private spending is associated with increased academic performance, but also important to understand whether this relationship is different across the distribution of academic performance and how large the different relationship sizes might be. Thus, our findings may provide important implications for Korea and possibly other contexts in which education is highly valued and the culture of excessive family spending on private educational is prevalent. 2. Methods 2.1. Data and procedure The analysis was based on the Korean Children and Youth Panel Survey (KCYPS), which was designed to track psychological and social development among a nationally representative sample of Korean youth. The National Youth Policy Institute (NYPI) gathered information from 2351 first-year middle school students at wave 1 (2010) and followed them for five consecutive years (follow-up rate was 89% in 2014). The students were selected using multilevel stratified sampling methods (National Youth Policy Institute, 2015). Specifically, after 16 metropolitan cities and provinces were identified, schools were selected through proportionate probability sampling. Then, classes were randomly chosen from each of those schools. Only wave 3 of the KCYPS contained information on all of our key variables of interest and was used for our analysis. Youth provided self-reported information about their demographic and school characteristics, and parents provided self-reported data on measures of family socioeconomic status. The NYPI's Institutional Review Board takes oversight of the entire data collection and management process (National Youth Policy Institute, personal communication, July 8, 2016). At the beginning of the study, all targeted students and their parents were first contacted via telephone, and those who orally agreed to participate in the panel study were asked to submit a signed consent form. Students were informed about the intention of the panel study in writing on the questionnaire and in person by the NYPI data administrator. NYPI ensures that personal information and responses to questionnaires are protected, and that the publicly available dataset does not contain information that allows identification of participants. 3. Measures 3.1. Dependent variable The academic performance variable was an equally weighted average of self-reported scores for three core subjects—Korean, math, and English. For each of the three subjects, scores were self-reported on a Likert scale ranging from 1 to 8 (1 ≥ “96”, 2 = “90–95”, 3 = “85–89”, 4 = “80– 84”, 5 = “75–79”, 6 = “70–74”, 7 = “65–69”, 8 = “≤64”). Scales were reverse coded so that a higher score would indicate better academic performance. 3.2. Independent variable The key variable of interest, monthly spending on private education, was originally measured in 10,000 Korean won, but was rescaled to

Y. Han, S. Lee / Children and Youth Services Review 69 (2016) 136–142

100,000 KW to improve the interpretability of the coefficients in the quantile regression. Youth who did not receive any private education were assigned “0” spending. 3.2.1. Control variables Given the centrality of school adjustment (Hong & Cho, 2014) and academic motivation, interest, and engagement (Singh, Granville, & Dika, 2002) in predicting academic performance among youth, measures of school adjustment and academic attitude were included as covariates. The KCYPS used the short version of Min's (1991) school adjustment measure modified by NYPI. School adjustment was a composite of 20 items representing aspects of learning activities, school rules, peer relationships, and teacher relationships. The academic attitude variable, an 18-item construct originally designed by Yang (2000) and modified by KYPI, measured youths' academic values, proficiency goals, self-control, and time management skills. For both measures, youth were asked to respond to questions on a Likert scale ranging from 1 to 4 (1 = “very true” and 4 = “not very true”). The scale was reversed, such that a higher score indicated better youth adjustment in school or pro-academic attitudes. Numerous studies have shown that family socioeconomic status, such as income and parental education (Lee, 2006; Sirin, 2005), and family structure (Jeynes, 2005) are strongly linked with youth academic attainment and achievement. Therefore, these family characteristics were controlled for in the analytic model. Monthly family income was originally measured in 10,000 Korean won but was rescaled to 100,000 KW for easier interpretability in the quantile regression analysis. Parental education was measured as the highest level of completed education between the mother and father (or primary caregiver for youth without any parents). Family structure was a dummy variable, measured as “1” for youth from two-parent families (both biological and non-biological parents) and “0” for youth from a single-parent family. 4. Analysis A summary of the demographic, psychological, and socioeconomic characteristics of the sample (Table 1), a distribution of the academic scores by spending on private education (Table 2), and academic motivation by academic performance (Table 3) are provided. Ordinary least squares (OLS) regression was used to study the average relationship between family spending in private education and youth academic performance, conditioned on youth demographics, family, and school factors. Quantile regression was also used to explore whether the magnitude of the association between family spending on private education and academic scores differed across quantiles of academic performance, while controlling for individual and family characteristics (Table 4). Table 1 Descriptive summary of variables (N = 2120). Variables

M (%)

SD

Min

Max

Youth variables Male School grade School adjustment Academic habits

50% 4.03 2.89 2.68

2.17 0.40 0.47

1 1.36 1

8 4 4

25.29 226.86

0 0

300 3333

Family variables Private education spending (per month) Income (per month) Two-parent family Parental education Less than middle school High school graduate 2-year college graduate 4-year college graduate At least a Master's degree

27.10 392.06 88% 3% 41% 11% 40% 5%

Note: Monetary unit for private education spending and income is 10,000KW.

139

Table 2 Academic performance by private education spending (N = 2120). Distribution

Mean Quantiles

Spending on private education

10% 25% 50% 75% 90%

0–20%

21–40%

41–60%

61–80%

81–100%

2.89 1 1.33 2.33 4 5.67

3.77 1 1.67 3.67 5.67 6.67

4.16 1.33 2.33 4 6 7

4.63 1.67 3 4.67 6.33 7.33

5.53 2.33 4.33 6 7.33 7.67

Theoretically, relationship parameters can be examined at any point along the distribution, but for parsimony, we focused on examining the importance of private expenditure in predicting academic performance on the 10th, 25th, 50th, 75th, and 90th percentiles of the conditional academic performance distribution (Bassett, Tam, & Knight, 2002; Koenker & Bassett, 1978). Bootstrapped (1000 replications) standard errors were computed for all variables of interest (Buchinsky, 1998). 5. Results 5.1. Descriptive summary All youth, evenly distributed in terms of gender, were in their third year of middle school. The average school grade was 4.03, which falls in the range of “75–79” points on a 100 point scale. School adjustment and academic habit scores were 2.89 and 2.68, respectively. In terms of family characteristics, average monthly family income was 3,920,000 KW (equivalent to 3920 US$), and per month spending on private education was 271,000 KW. Most youth were from two-parent families (88%) in which at least one parent was at least a high school graduate (97%). There were no group mean differences (p-value ranging from 0.48 to 0.96) in descriptive statistics between the analytic sample (N = 2120) and the original wave 3 sample (N = 2259) on key variables of interest, indicating that results derived from the analytic sample are representative of the national youth population. Additionally, when decomposing academic scores at the mean and various quantiles by spending on private education, it became clear that that the relationship between spending on education and academic performance differed along the distribution of academic scores. The average academic performance score was 2.89 for youth from families whose investment in private education was in the lowest quintile of the sample; this mean score gradually increased to 5.53 among youth from families at the highest quintile of the sample. When examining the 10th, 25th, 50th, 75th, 90th percentiles of the performance distribution, a gradual increase in academic scores by family spending on private education was still observed. The degree of change, however, differed by quantile of the academic score distribution. Variation in academic scores was much greater around the 50th percentile of the academic performance distribution (ranging from 2.33 to 6) than in the 10th or 90th percentiles (ranging from 1 to 2.33 and 5.67 to 7.67, respectively). Furthermore, study time management skills, academic self-regulation ability, intrinsic value of learning, and educational aspirations also differed by academic score distribution, with the lowest

Table 3 Academic motivation by academic performance level (N = 2120). Academic motivation

Total Self-regulation Time management Intrinsic value Aspiration

Academic performance 0–20%

21–40%

41–60%

61–80%

81–100%

2.42 2.23 2.27 2.65 2.41

2.55 2.33 2.39 2.79 2.57

2.69 2.44 2.55 2.93 2.77

2.80 2.55 2.67 3.04 2.83

2.97 2.73 2.85 3.17 3.08

140

Y. Han, S. Lee / Children and Youth Services Review 69 (2016) 136–142

Table 4 Ordinary least squares and quantile regression results (N = 2120). OLS

Quantile regression 10%

25%

50%

75%

90%

0.206⁎⁎⁎ (0.018)

0.122⁎⁎⁎ (0.021)

0.246⁎⁎⁎ (0.033)

0.265⁎⁎⁎ (0.026)

0.212⁎⁎⁎ (0.030)

0.131⁎⁎⁎ (0.032)

School adjustment

−0.169⁎ (0.079) 0.839⁎⁎⁎

−0.061 (0.085) 0.678⁎⁎⁎

−0.257⁎⁎ (0.089) 0.731⁎⁎⁎

−0.165 (0.114) 0.763⁎⁎⁎

−0.077 (0.124) 1.141⁎⁎⁎

−0.205 (0.115) 0.916⁎⁎⁎

Academic habits

(0.131) 1.131⁎⁎⁎ (0.109)

(0.144) 0.394⁎⁎ (0.134)

(0.173) 1.056⁎⁎⁎ (0.162)

(0.173) 1.438⁎⁎⁎ (0.162)

(0.183) 1.209⁎⁎⁎ (0.138)

(0.208) 0.758⁎⁎⁎ (0.207)

Parental education

0.006⁎⁎ (0.002) 0.247⁎⁎⁎

0.005 (0.003) 0.146⁎⁎

0.007⁎ (0.003) 0.244⁎⁎⁎

0.006⁎⁎ (0.003) 0.276⁎⁎⁎

0.004 (0.003) 0.239⁎⁎⁎

0.005 (0.004) 0.160⁎

Two-parent family

(0.041) 0.287⁎

(0.047) 0.038 (0.101) −2.354⁎⁎⁎ (0.500)

(0.059) 0.167 (0.115) −3.956⁎⁎⁎ (0.407)

(0.067) 0.293 (0.203) −4.104⁎⁎⁎ (0.380)

(0.064) 0.354 (0.233) −2.942⁎⁎⁎ (0.451)

(0.064) 0.152 (0.220) 0.848 (0.772)

Private education spending Youth covariates Male

Family covariates Family income

Constant

(0.130) −3.116⁎⁎⁎ (0.317)

⁎ p b 0.05. ⁎⁎ p b 0.01. ⁎⁎⁎ p b 0.001.

6. Discussion Despite the growing centrality of the private education system in many countries around the globe, no clear empirical consensus has emerged on the association between investment in private education and youth academic performance. Prior studies have identified positive, negative, and

0.40

In the conditional mean OLS model, an increase of 100,000 KW toward private education spending was associated with a 0.206 point increase in academic performance (p b 0.001). Consistent with the OLS results, quantile regression results also indicated a positive association between family spending on private education and youths' academic scores across the conditional distribution of academic performance. The size of the relationship coefficient was largest around the 50% quantile, whereas the magnitude was smaller at the lowest and highest deciles of the conditional academic performance distribution. Every 100,000 KW difference in investment in private education was associated with an increase of 0.122 points on the 10th percentile, 0.246 points on the 25th percentile, 0.265 points on the 50th percentile, 0.212 points on the 75th percentile, and 0.131 points on the 90th percentile of the academic performance distribution, conditioned on youth and family attributes (all p b 0.001). To reiterate, for youth on the 10% quantile, given the same individual and family characteristics so the only difference is family spending in private education, for every 100,000 KW spent monthly on private education, the difference in academic scores across individuals was 0.122 points on a 8 point scale. The same interpretation held for relationship sizes at other points of the conditional distribution of academic scores. Simultaneous test results indicated that the size of coefficients in the lowest (10%) and highest (90%) deciles were not significantly different, but those coefficients were different (p b 0.001) from that on the median (50%). Fig. 1 is a graphical representation of the single coefficient estimated from the OLS model (medium-dotted line) and the heterogeneous relationship sizes at various percentiles as estimated from the quantile regression model (long-dotted line).

Spending on Private Education 0.10 0.20 0.30

5.2. OLS and quantile regression

null relationships (Byun, 2014; Kim & Kim, 2002). Motivated by the ambiguous messages in the literature, the current study explored whether the inconsistent findings are tied to the use of general regression models, which rely on a single parameter estimated on the conditional mean. Only a few researchers (Lee & Kwon, 2011; Zhang, 2013) have thoroughly examined this possibility in non-Western youth populations. The current exploratory study is one of the first to use quantile regression—a methodological tool that identifies the full nature of relationship sizes across a distribution rather than estimating the average effect size (Bassett et al., 2002; Koenker & Bassett, 1978)—to examine whether family spending in private education is associated with academic performance among a nationally representative sample of Korean youth. The unique Korean context, in which private education is no

0.00

levels of academic habits and motivation among youth with low academic scores and the highest levels among youth with high academic scores (Table 3).

.05 .1 .15 .2 .25 .3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95 Academic Perfomance Quantile

Fig. 1. Graphical representation of heterogeneous relationship sizes from OLS and QR Models (N = 2120). Note: The medium-dotted horizontal line represents the OLS regression coefficient (B = 0.206, p b 0.001), with short dotted lines representing the 95% confidence interval. The long-dotted line represents quantile regression coefficients at multiple quantiles across the conditional distribution of academic performance, with the gray shaded area representing the 95% confidence interval (B = 0.122 on the 10th percentile, B = 0.246 on the 25th percentile, 0.265 points on the 50th percentile, B = 0.212 on the 75th percentile, B = 0.131 on the 90th percentile of the academic performance distribution, conditional on youth and family attributes; all significant at p b 0.001).

Y. Han, S. Lee / Children and Youth Services Review 69 (2016) 136–142

longer considered a shadow or supplement to the formal educational system (Ihm et al., 2008; Kim & Kim, 2002; Yang, 2006), presents an ideal setting to explore those differential relationship sizes. Furthermore, the repercussions of private education, a double-edged sword, go beyond individuals and families to the national level in Korea (Kim & Lee, 2011; Yang, 2012). Quantile regression results supported the plausibility of heterogeneous patterns in the link between family spending in private education and youth academic scores across the conditional distribution of academic performance. Generally, there was a positive association between spending in private education and academic performance. The size of that positive link, however, was heterogeneous, with the magnitude of the relationship reaching its peak around the median of the performance distribution. In other words, the size of the relationship between family spending on private education and youth academic performance varies across the distribution of academic scores. Given those results, the conditional mean-based coefficient found using the OLS regression overestimated the relationship size near the lowest and highest ends of the conditional academic performance distribution. At the same time, the average coefficient underestimated the relationship size near the median interval of the distribution. In other words, the difference in academic scores associated with difference in family spending was smaller among the very high and very low scores than among the middle performing group, such that a single relationship size estimate would have provided a limited picture of the entire story of family expenditure on private education and youth academic performance. Together, the OLS regression and quantile regression results show that the size of the relationship between investment in private education and academic performance depends heavily on the conditional distribution of academic scores. Thus, the inconclusive findings in the literature might result from different sample compositions across studies; study samples composed primarily of low, medium, or high performers would produce different relationship sizes between investment in private education and academic performance. Although the exact mechanisms driving the heterogeneous pattern of results remain to be determined, one possible explanation can be speculated as follows. Because students and parents who turn to cram schools and private tutoring demand rapid academic enhancement, private education providers tend to focus on improving student performance in a single area of study during a short period. This objective could affect the quality of instruction and the content of material learned in private education institutions. Specifically, unlike the formal school system, the private system might lack detailed explanations of the principles or definitions of core concepts and provide limited hands-on experience or experiments to aid learning (Kim & Kim, 2002). Instead, private education instruction focuses on quickly teaching test-prep techniques or repeatedly injecting information certain to appear on school exams or standardized tests, rather than fostering the ability to think critically (Byun, 2014; Kim & Kim, 2002). The instruction styles offered at private education institutions, injecting techniques for fast score improvement, can be effective for some youth but not for others. Relating the specific instruction styles of private institutions with the concave-shaped pattern of the family spending coefficient derived in the analytic results, the following ideas may be elaborated. For middle-performing students, acquiring quick score-enhancing tips and techniques may be a key to higher academic performance, particularly among high-aspiring youth who adopt private education as an “enrichment” strategy to “stay ahead of their peers” (Byun & Park, 2012, p.42). The same might not apply to low-performing youth, however, particularly those who lack the motivation or aspirations to do well in school. In fact, descriptive analysis showed that indicators of academic habits and motivation (e.g., study time management skills, academic self-regulation ability, and intrinsic value of learning) are lowest among students at the bottom of the score distribution and increase gradually with higher academic scores. This indicates that the teaching

141

methods and materials designed to reap rapid improvements in the private education system might not be effective for students who lack the desire to do well in the first place and thus do not serve as an effective remedial (Byun & Park, 2012) tool for low-performing students. For those youth, alternative methods—such as offering mentoring programs that heighten educational aspirations, improving attachment to teachers and school, or stimulating academic curiosity by establishing a rigid knowledge foundation—could be more helpful in improving academic performance than simply attempting to inject technical material and skills. Finally, youth at the higher deciles of the academic distribution have difficulty displaying a drastic enhancement in academic scores due to the ceiling effect—there is little room for variation because these youth already have very high academic scores (Lee & Kwon, 2011). As shown in Table 2, youth in the 80th–100th percentile had an average score of 7.57, which approximates the highest possible score of 8. Alternatively, it is also plausible that private education does not help high-performing youth to advance further, but it does help them maintain their position at the top of the rank by preventing them from falling behind (Yang, 2012). A deeper understanding of the specific associations between the content and quality offered in private education and individual differences in academic motivation and aspirations in addition to consideration of maintaining the status quo as a real effect are needed in future research. Interpretation of these study findings must consider the following limitations. First, the cross-sectional nature of the data set does not allow any causal interpretations. The results do not indicate that additional family spending on private education will increase youth academic scores. The results do indicate that, given identical individual and family characteristics, family spending in private education is linked with youth academic scores, and that the size of this relationship differs across low, medium, and high scores. Second, the current study considered only the quantity of investment—specifically, a family's monthly average spending—in private education. Empirical work, however, suggests that the type and quality of private education matter in determining youth academic performance (Byun, 2014; Lee, 2006). The size of the relationship between family spending and youth academic performance may be biased if the type or quality of private education is systematically different across the distribution of academic performance. Although the quality of private education can be proportional to its cost (Byun, 2014), the ability to control for the type and duration of participation in private education may help derive more accurate results. Notwithstanding those limitations, the findings from the current exploratory study may shed light on the inconsistency in the literature concerning the effect of private education in youth academic achievement or attainment. In concordance with previous work, OLS results confirmed that family investment in private education was associated with higher academic performance. The current study results using quantile regression add to the literature base by providing a detailed account of where this link between family spending on private education and youth academic performance is greatest or smallest and how large is the connection. The concave representation of the relationship between family spending in private education and youth academic performance across the distribution of academic scores indicates that the magnitude is greatest at the median of the academic performance distribution, whereas the size of the coefficient is smallest at the extreme ends of the academic performance distribution. Taken together, our results highlight important implications for parents, educators, and policy makers who are interested in implementing targeted strategies for improving youth academic outcomes: mean-based results would have masked the entire picture of the heterogeneous association between family spending in private education and youth academic performance.

References Bassett, G. W., Jr., Tam, M. -Y. S., & Knight, K. (2002). Quantile models and estimators for data analysis. Metrika, 55(1), 17–26. http://dx.doi.org/10.1007/s001840200183.

142

Y. Han, S. Lee / Children and Youth Services Review 69 (2016) 136–142

Bray, T. M. (2009). Confronting the shadow education system: what government policies for what private tutoring? United Nations Educational, Scientific and Cultural Organization; International Institute for Educational Planning. Buchinsky, M. (1998). Recent advances in quantile regression models: A practical guideline for empirical research. The Journal of Human Resources, 33(1), 88–126. Byun, S. (2014). Shadow education and academic success in Republic of Korea. In H., & K. (Eds.), Korean education in changing economic and demographic contexts (pp. 39–58). Springer. Byun, S., & Park, H. (2012). The academic success of East Asian American youth the role of shadow education. Sociology of Education, 85(1), 40–60. Cheo, R., & Quah, E. (2005). Mothers, maids and tutors: An empirical evaluation of their effect on children's academic grades in Singapore. Education Economics, 13(3), 269–285. Choi, H. (2008). The effect of private tutoring on college entrance. Kukje Kyungje Yongu, 14(1), 73–110. Chung, J., Lee, H., & Kim, S. (2014). A hierarchical analysis of the factors influencing on student achievement: Using the teacher and student factors of TIMSS 2011. J. Korean Teach. Educ., 31(2), 53–75. Dang, H. -A. (2007). The determinants and impact of private tutoring classes in Vietnam. Economics of Education Review, 26(6), 683–698. Eide, E. R., Showalter, M. H., & Goldhaber, D. D. (2010). The relation between children's health and academic achievement. Children and Youth Services Review, 32(2), 231–238. Haile, G. A., & Nguyen, A. N. (2008). Determinants of academic attainment in the United States: A quantile regression analysis of test scores. Education Economics, 16(1), 29–57. Hong, A., & Cho, G. (2014). The effect of social support, academic self-efficacy and school life adjustment of elementary school students on their school happiness and academic achievement. J. Learn.-Cent. Curric. Instr., 14(1), 45–68. Ihm, C. S., Woo, M. S., & Chae, J. E. (2008). An analysis of private education demand in Korea: Learning supplement and strategic investment. J. Econ. Finance Educ., 17(2), 1–27. Jeynes, W. H. (2005). Effects of parental involvement and family structure on the academic achievement of adolescents. Marriage & Family Review, 37(3), 99–116. Kim, C. (2001). The effects of out-of-school learning on classroom learning. Korean J. Youth Stud., 8(2), 61–89. Kim, K. (2003). The effect of private lessons on academic achievement. Korean J. Sociol. Educ., 13(3), 45–65. Kim, K. (2012). The effects of school sectors on student academic achievement: Causal analysis using social experiment. Korean J. of Sociol. Educ., 22(2), 1–29. Kim, Y., & Kim, M. (2002). Analysis of educational situation of cram schools designed to prepare students for college entrance examination. Research report 2002–1Korean Educational Development Institute. Kim, J., & Kim, J. (2009). Analysis on the effect of private tutoring: Propensity score matching model. The Journal of Economics and Finance of Education, 18(3), 63–87 Retrieved from http://scholar.dkyobobook.co.kr/searchDetail.laf?barcode=4010022839915. Kim, K., & Lee, H. (2011). The effectiveness of shadow education on CSAT. The Journal of Educational Research, 32(2), 35–49. Koenker, R., & Bassett, G., Jr. (1978). Regression quantiles. Econometrica, 46(1), 33–50. Statistics Korea (2014a). 2014 report of spending on private education. Kwon, K. (2003). A study on reduction strategies of the class size for improvement of education achievement in schools. Korean J. Educ. Methodol. Stud., 15(2), 23–55. Lee, E. (2006). The impact of household income and private education on academic performance of middle school students. Korean J. Youth Stud., 13(6), 247–274. Lee, K., & Kwon, Y. (2011). The effect of private tutoring expenses and hours on academic achievement: Quantile regression approach. J. Econ. Finance Educ., 20(3), 99–133. Lee, S., & Lim, H. (2009). An analysis of the effect of private tutoring expenditure on middle school Students' achievement. J. Econ. Finance Educ., 18(1), 141–166.

Lee, J., Kim, Y., & Yoon, C. (2004). The effects of pre-class tutoring on student achievement: Challenges and implications for public education in Korea. KEDI J. Educ. Policy, 1(1). Lee, H., Lim, C., & Min, K. (2010). Effects of coaching on college admissions. J. Econ. Finance Educ., 19(3), 151–175. Levin, J. (2001). For whom the reductions count: A quantile regression analysis of class size and peer effects on scholastic achievement. Empirical Economics, 26(1), 221–246. Liu, J. (2012). Does cram schooling matter? Who goes to cram schools? Evidence from Taiwan. International Journal of Educational Development, 32(1), 46–52. Manzon, M., & Areepattamannil, S. (2014). Shadow educations: Mapping the global discourse. Asia Pac. J. Educ., 34(4), 389–402. Min, B. (1991). The effect of school adjustment and self-concept on academic achievement. Hongik University. Ministry of Gender Equality and Family (2014). Comprehensive report on adolescents. Reserach Report 2014-53Ministry of Gender Equality and Family. National Youth Policy Institute (2015). User Guide for Korea Children and Youth Panel Survey. Retrieved from http://archive.nypi.re.kr/brdartcl/boardarticleView.do?menu_nix= qZc474Ak&brd_id=BDIDX_k9Fd9oFi29nooCcJ7kX1I4&srch_ctgry_idx=CTIDX00043. Sang, K., & Baek, S. (2005). The effects of private tutoring on high school student's learning. J. Educ. Eval., 18(3), 39–57. Seong, T. J. (1993). Education for college preparation and excessive private education. The Journal of Educational Research, 31(2), 67–85. Singh, K., Granville, M., & Dika, S. (2002). Mathematics and science achievement: Effects of motivation, interest, and academic engagement. The Journal of Educational Research, 95(6), 323–332. Sirin, S. R. (2005). Socioeconomic status and academic achievement: A meta-analytic review of research. Review of Educational Research, 75(3), 417–453. Statistics Korea (2014b). Public finance statistics. Retrieved from http://www.index.go. kr/potal/main/EachDtlPageDetail.do?idx_cd=2739 Steinberg, L. D. (2014). Adolescence (10 ed.). New York, NY: McGraw-Hills. Stevenson, D. L., & Baker, D. P. (1992). Shadow education and allocation in formal schooling: Transition to university in Japan. The American Journal of Sociology, 1639–1657. Wayne, A. J., & Youngs, P. (2003). Teacher characteristics and student achievement gains: A review. Review of Educational Research, 73(1), 89–122. Wenglinsky, H. (1998). Finance equalization and within-school equity: The relationship between education spending and the social distribution of achievement. Educ, Eval. Policy Analysis, 20(4), 269–283. Witte, J. F. (1992). Private school versus public school achievement: Are there findings that should affect the educational choice debate? Economics of Education Review, 11(4), 371–394. Yang, M. (2000). The study on the development and validation of self-regulated learning model. Seoul National University. Yang, J. (2006). A longitudinal analysis of private tutoring expenditure: A polarization index and Gini coefficient analysis of KLIPS. J.Econ. Finance Educ., 15(2), 199–219. Yang, J. (2012). Effects of private tutoring on college entrance. J.Econ. Finance Educ., 21(2), 27–52. Yang, J. (2013). Analyzing the trends of private tutoring expenditure from 1982 to 2017. Journal of Educational Administration, 31(4), 421–448. Zhang, Y. (2013). Does private tutoring improve students' National College Entrance Exam performance?—A case study from Jinan, China. Economics of Education Review, 32, 1–28. Zhou, M., & Kim, S. (2006). Community forces, social capital, and educational achievement: The case of supplementary education in the Chinese and Korean immigrant communities. Harvard Educational Review, 76(1), 1–29. http://dx.doi.org/10.17763/ haer.76.1.u08t548554882477.