Abstract
Using longitudinal data for a nationally representative sample of ninth graders in South Korea, we examine socioeconomic differences in the likelihood of making transitions into different types of high school and college with a goal of testing the validity of the Effectively Maintained Inequality (EMI) hypothesis. We find significant socioeconomic disparities in the likelihood of attending an academic high school and a four-year university. However, the predicted probabilities suggest that even disadvantaged students typically choose academic high school relative to vocational high school. Further, although disadvantaged students likely end up with a two-year junior college, those disadvantaged students graduating from an academic high school typically choose a four-year university, after controlling for academic achievement and other variables. We discuss the relevance of EMI for South Korea and broad implications for elsewhere where postsecondary education becomes close to be universal.
Numerous studies across a wide range of societies have examined whether educational expansion reduces educational inequality by opening up educational opportunities to children from disadvantaged backgrounds, consistently finding persistent inequality despite educational expansion during the last decades (Blossfeld and Shavit 1993; Shavit, Arum, and Gamoran 2007). The maximally maintained inequality (MMI) hypothesis offers an explanation of the persistent inequality by proposing that socioeconomic inequality in the attainment of a specific level of education would persist unless the given level of education is saturated among high social class individuals (Raftery and Hout 1993). In other words, when privileged groups reach saturation at a given level of education, further expansion of education should lead to the decline in socioeconomic inequality (see Lucas 2009). Although MMI was supported in Ireland and Britain where the hypothesis was tested initially, studies in other countries such as Israel (Ayalon and Shavit 2004) and the United States (Lucas 2001) did not find strong support.
By contrast, the effectively maintained inequality (EMI) claims that persistent inequality exists even in the context of nearly universal enrollments at a given level of education with respect to the quality of education (Lucas 2001). Criticizing the lack of appreciation of MMI for qualitative differentials in a given level of education, EMI argues that the effect of socioeconomic background should not disappear even when a given level of education is nearly universal because socioeconomic background still affects the types of education (Lucas 2001). In his empirical analysis of U.S. high school students, Lucas showed more evident roles of family background in tracking placements (i.e., qualitative differences within high schools) than the likelihood of moving to the next grade (i.e., quantitative differences).
Lucas’ study, however, has some limitations. For example, using the data of 1980 U.S. high school sophomores, Lucas focused only on different curriculum tracks within high schools but was unable to examine qualitative differences in college matriculation. It is understandable given that transition to college in the early 1980s was far from being universal in the United States. In fact, even nowadays it is difficult to find a society where transition to college is nearly universal. Of course, socioeconomic inequality in the quality of education may emerge even before the given level of education is saturated (Ayalon and Shavit 2004). But EMI can be better tested in a context where a large, if not all, share of the socioeconomically advantaged reach saturation in college attendance.
In this respect, South Korea (Korea, hereafter) provides an excellent case for testing EMI because Korean students show nearly universal transition from middle school to high school and more than 80 percent of all Korean high school students matriculate in college (Korean Educational Development Institute [KEDI] 2010). Although saturation at the level of college entrance might not have yet occurred among the socioeconomically advantaged, Korea’s college attendance rate is much higher than that of many other countries (Organisation for Economic Co-operation and Development [OECD] 2010). Therefore, the Korean case allows us to better test the relevance of EMI as to how socioeconomic background shapes the transition to different types of colleges when college education becomes nearly universal among the socioeconomically advantaged.
Using longitudinal data that tracked a nationally representative sample of ninth-grade students through upper secondary and postsecondary education in Korea, we examine socioeconomic differences in transition into different types of educational institutions in both high school and college levels. Because we use data that followed the same students to postsecondary years, we can examine how paths into specific types of high schools affect students’ later transitions to college. We begin with a brief description of the educational setting in Korea. Next, we describe data and methods used. We then present the results of logistic and multinomial models that show the extent to which socioeconomic backgrounds influence transitions to upper levels of education. We conclude with the implications of the findings especially for the EMI hypothesis.
CONTEXTUAL BACKGROUND
Educational Transitions to High School in Korea
The Korean K-12 education system consists of (a) primary schools (Grade 1–6), (b) middle schools (Grade 7–9), and (c) high schools (Grade 10–12) (Byun, Kim, and Park 2012). Primary and middle school is compulsory education with no between-school tracking. Accordingly, almost all primary school graduates are entering middle schools and receiving uniform education (KEDI 2010). On the other hand, high school is not compulsory education and a relatively small amount of tuitions are charged. After middle schools, students are selected into basically two different types of high schools: academic high schools and vocational high schools. Academic high schools are college preparatory schools in which the majority of Korean adolescents enroll (approximately 72 percent of all high school students in 2010) (KEDI 2010). Vocational high schools serve students who want to develop vocational skills (approximately 27 percent of all high school students in 2010) (KEDI 2010). Traditionally, within-school curriculum tracking such as ability grouping was not widely practiced within each type of high school.1
Unlike primary and middle schools where school assignments are generally determined by place of residence regardless of academic achievement, admissions to different types of high schools are determined largely by middle school academic performance (Byun et al. 2012). As a result, there is a significant achievement gap between these two types of high schools. Generally speaking, academic high schools serve higher-achieving students than vocational high schools do (Kim and Byun 2006; Park 2010). Given the strong relation of socioeconomic background to academic achievement (see Park 2007a), a socioeconomic gap is also evident between these two types of high school: students attending academic high schools disproportionately come from upper middle class families, whereas students attending vocational high schools disproportionately come from working or poor families (Chang 2007; Kim and Byun 2006).
It is important to note that almost every middle school graduate moves onto high schools and the dropout rates during high school years are considerably lower compared to other countries such as the United States. In 2008, for example, the overall dropout rate of Korean high school students was approximately 1.8 percent (KEDI 2010). As a result, in this situation where a considerable proportion of youth go to college (KEDI 2010), dropping out of high schools signals a marginalized status in Korea.
Educational Transitions to College in Korea
The Korean higher educational systEMIs characterized as a stratified system where a second-tier system consists of two-year junior colleges that are lower in prestige than the first-tier system, four-year universities (Park 2007b). Importantly, Korean higher education heavily relies on private institutions to accommodate college students (Park 2007b). For the 2004–2005 academic year, for example, 78 percent of full-time college students in Korea enrolled in private universities (OECD 2007, Table B5.1a). Moreover, Korean households contribute more than 80 percent of the cost of higher education, which is significantly more than in any other OECD country (OECD 2010, Table B3.2b.), suggesting a considerable financial burden for some families, especially among those with limited financial resources. Not surprisingly, studies in Korea document significant differences in socioeconomic background between no-college vs. two-year junior college vs. four-year university goers (Byun and Kim 2010b, 2012; Kim 2008; Kim and Byun 2006).
Although financial resources matter, probably the most critical factor for the transition to college is a student’s score on the College Scholastic Ability Test (CSAT) (Byun and Kim 2010b).2 A heavy reliance on the high-stake test for college admissions in Korea is contrasted with many other countries such as the United States where non-cognitive criteria such as extracurricular activities and recommendation letters in addition to cognitive test scores (e.g., SAT) play an important role in college admissions (Byun, Schofer, and Kim 2012). Because CSAT is administered nationally only once a year and because the CSAT score is so critical for college admissions, it is not uncommon that students, who failed to score high enough to be admitted to a college of choice in their high school senior year, spend one or more additional years after high school graduation to retake CSAT.
In nature, these repeating students are not necessarily low-achievers; rather in many cases these students are average above or even high achievers who believe that they could do better on CSAT if had a second chance. Furthermore, these repeating students are not necessarily disadvantaged in socioeconomic background because they already completed high school and then usually attend private institutions, which charge high tuition fees, to prepare for CSAT even after high school graduation. Indeed, Korean literature suggests that students who did not enroll in college right after high school graduation are not disadvantaged but are actually more advantaged in socioeconomic background and academic achievement compared to students who enrolled especially in two-year junior colleges (Kim 2008; Kim and Byun 2006).
It is important to note that a growing number of vocational high school graduates enroll in college. In 1995, for instance, the college enrollment rate among vocational high school graduates was 19.2 percent (vs. 72.8 percent for academic high school graduates); a decade later in 2009, the rate spiked to 73.5 percent (vs. 84.9 percent for academic high school graduates) (KEDI 2010). Yet, there is a significant gap in the type of college attended between academic and vocational high school graduates: vocational high school graduates are far less likely than academic high school students to enroll in a four-year university (Byun and Kim 2010b, 2012; Kim 2008; Kim and Byun 2006).
METHODS
Data and Sample
We used data from the Korean Education and Employment Panel (KEEP), conducted by the Korea Research Institute for Vocational Education & Training, a government-funded educational research agency. We chose KEEP over other large-scale datasets3 available in Korea because it provided the CSAT scores of the surveyed students with the assistance of the Korea Institute for Curriculum and Evaluation, which administers CSAT. As noted earlier, because the CSAT is the most important factor determining college admissions, most students keenly focus on performing their best on this test. Therefore, CSAT scores may offer the most reliable set of information on academic achievement in Korea.
In 2004, KEEP drew random samples of approximately 20 ninth graders in each of 100 randomly selected middle schools, yielding a nationally representative sample of 2,000 ninth-grade students. KEEP then followed these students every year since then. At the time when the current study was conducted, only data from the baseline (2004) through fifth follow-up survey (2009) were available. For the current investigation, therefore, we restricted to the students who participated in the initial survey and all follow-up surveys until 2009 (two years after high school) and whose information about their educational pathways were available (N = 1,519).4
Measures
Dependent variables.
We were interested in the transitions from middle school to high school and then to college. Consistent with the general trend, the number of ninth graders, who did not make transition to high school, is extremely small in our KEEP data.5 Therefore, the transition from middle school to high school was classified by whether the respondent attended an academic vs. vocational high school. On the other hand, the transition from high school to college was classified by whether the student ever enrolled in a postsecondary institution as of 2009. We also examined the types of college which the student attended (two-year junior college enrollment vs. four-year university enrollment vs. no college enrollment). As described, a not-small portion of Korean students who failed to earn admissions to college of choice tend to spend another year preparing for the next CSAT (Byun and Kim 2010b). Therefore, we measured the college enrollment status by drawing on the fifth follow-up (2009) survey. If the student did not report the college enrollment status in 2009, we used the fourth (2008) follow-up survey to determine the college enrollment status.6
Independent variables.
For socioeconomic background, we included (a) parental education which is higher one between mother’s and father’s education (middle school graduation or less [reference group] vs. a high school diploma vs. a college degree), (b) monthly family income (natural logarithm), and (c) home ownership. We controlled for gender and prior academic achievement. To predict the transition from middle school to high school, we controlled for a student’s academic achievement at ninth grade (i.e., the last year in middle school), which was measured by her/his homeroom teacher’s report on academic standing in her/his school (or classroom) on the percentile scale from 0 to 100 (higher values indicate higher academic standing). For the transition from high school to college, we included prior academic achievement using the students’ CSAT score, which was measured by the level (from 1 = lowest to 9 = highest levels) instead of the raw score metric. All independent variables were drawn from the 2004 base year survey with the exception for the CSAT level; the CSAT level was measured in either 2008 or 2009 when the student took CSAT.
Analytic Strategies
First, we descriptively presented differences in background characteristics between academic vs. vocational high school students and among students who enrolled in junior colleges vs. four-year universities vs. no college enrollment. Second, using logistic regression, we analyzed the likelihood of attending an academic high school compared to attending a vocational high school. Given the significant influence of socioeconomic background on academic achievement that determines the educational transitions to upper levels of education (Byun and Kim 2010b; Kim and Byun 2006), we estimated two logistic regression models. The first model included all independent variables but prior academic achievement; the second model added the prior academic achievement measure (academic standing in ninth grade).
Third, using logistic regression, we analyzed the likelihood of enrolling in any kind of college. And then, using multinomial logistic regression, we examined socioeconomic differences in reaching one of the three different destinations after high school graduation: no college attendance, two-year junior colleges, and four-year universities.7 For these logistic and multinomial logistic regression analyses for college transition, we estimated three models. The first model included all independent variables except for the prior academic achievement and high school track variables. Subsequently, we entered the prior academic achievement (Model 2) and high school track variables (Model 3), respectively. The aim was to determine whether there were significant socioeconomic differences in the transition to college after controlling for academic achievement and the high school track respectively.
Finally, based on the logistic and multinomial coefficients estimated above, we calculated the predicted probabilities of reaching particular destinations for students from “advantaged” families and their counterparts from “disadvantaged” families because “regression-type coefficients by themselves cannot reveal whether social background moves people over thresholds” (Lucas 2001:1671) (see Lucas 2001 for more information). Here, we defined a student’s family background as “advantaged” when (a) parents had a college degree, (b) logged family income was 5.87 (i.e., the average logged family income among families whose parents had a college degree), and (c) families owned their homes. In contrast, we defined a student’s family background as “disadvantaged” when (a) parental education was middle school or less, (b) logged family income was 5.05 (i.e., the average logged family income among families whose parental education was middle school or less), and (c) families did not own their homes. We fixed values of the other covariates to be the same between students from advantaged families and their counterparts from disadvantaged families.
To address missing data for the independent variables (but not for the dependent variables), we employed a multiple imputation technique using the Stata ICE module (see Table 1 below for the percentage of missing data). Following recommendations set forth by Johnson and Young (2011), we included all of the dependent and independent variables in the imputed model to predict missing values and generated five imputed datasets. However, we decided to create one single, complete data by averaging the five imputed values and then replacing the missing values with these average values, because the statistical literature has no consensus on the best way to combine predicted probabilities across imputed datasets. To address the nested nature of the KEEP data (i.e., students were randomly selected within the sampled schools), we used robust standard errors which adjust for the inflated standard errors resulting from the violation of the independent errors (Rogers 1993).
Table 1.
Descriptive Statistics for the Independent Variables Included in the Analyses
| Type of high school | Type of college enrollment | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Vocational (n = 376) |
Academic (n = 1,143) |
No college (n = 203) |
two-year junior college (n = 468) |
four-year university (n = 848) |
Total (N = 1,519) |
% imputed | ||||||||
| Variable | M | (SD) | M | (SD) | M | (SD) | M | (SD) | M | (SD) | M | (SD) | Range | |
| Parental education | ||||||||||||||
| Middle school or less | 0.29 | (0.45) | 0.13 | (0.33) | 0.26 | (0.44) | 0.24 | (0.42) | 0.11 | (0.31) | 0.17 | (0.37) | 0–1 | — |
| High school | 0.58 | (0.49) | 0.52 | (0.50) | 0.53 | (0.50) | 0.60 | (0.49) | 0.50 | (0.50) | 0.53 | (0.50) | 0–1 | |
| College | 0.13 | (0.34) | 0.36 | (0.48) | 0.21 | (0.41) | 0.17 | (0.37) | 0.40 | (0.49) | 0.30 | (0.46) | 0–1 | |
| Family income (logged) | 5.26 | (0.67) | 5.61 | (0.64) | 5.31 | (0.83) | 5.39 | (0.71) | 5.65 | (0.56) | 5.52 | (0.66) | 0–8.01 | 1.9 |
| Owned homes | 0.57 | (0.50) | 0.73 | (0.45) | 0.55 | (0.50) | 0.63 | (0.48) | 0.75 | (0.43) | 0.69 | (0.46) | 0–1 | — |
| Female | 0.51 | (0.50) | 0.53 | (0.50) | 0.53 | (0.50) | 0.55 | (0.50) | 0.51 | (0.50) | 0.52 | (0.50) | 0–1 | — |
| Academic achievement at ninth grade | 29.84 | (19.81) | 60.85 | (23.69) | 36.03 | (25.30) | 41.37 | (22.89) | 63.79 | (23.45) | 53.17 | (26.43) | 0–100 | 10.8 |
| Academic achievement at 12th grade | 2.98 | (1.23) | 4.85 | (1.84) | 3.42 | (1.68) | 3.45 | (1.42) | 5.13 | (1.83) | 4.38 | (1.89) | 1–9 | 29.9 |
Data source: Korea Education and Employment Panel of 2004–2009
RESULTS
Descriptive Findings
Figure 1 shows the educational transitions among these KEEP students whose educational pathways were identified. Of 1,519 middle school graduates, 75 percent went to academic high schools, while 25 percent to vocational high schools. In addition, 56 percent attended a four-year university as of two years after high school, 31 percent attended a two-year junior college, and only the remaining 13 percent did not enroll in any kind of college. In other words, almost nine out of ten high school graduates in our sample made transition to postsecondary education. Among those students enrolling in academic high schools (N = 1,143), 67 percent enrolled in a four-year university, while 25 percent in a two-year junior college. Only 13 percent did not enroll in a college. On the other hand, among those students enrolled in vocation high schools (N = 376), only 22 percent enrolled in a four-year university, 50 percent in a two-year junior college, and 29 percent did not enroll in a college.
Figure 1.
Flow Chart Showing Educational Pathways
(with Percentages in the Korean School System: Transitions Completed in 2009 for Cohorts Who Attended Middle School in 2005)
Note. Transitions: t1 = transition from lower secondary to upper secondary school (age 15); t2 = transition to tertiary education.
Table 1 presents descriptive statistics for the independent variables included in our analyses by the types of high school and college. The first panel shows socioeconomic disparities between students attending academic versus vocational high schools. For example, 36 percent of the academic high school students had parents who had a college degree, whereas only 13 percent of the vocational high school students did so. In addition, a percentage of students whose families owned their homes were 73 for academic high school students and 57 for vocational high school students. Furthermore, there was a substantial achievement gap between academic and vocational high school students (e.g., 60.9 vs. 29.8 for academic achievement at ninth grade) favoring academic high school students.
The second panel of Table 1 shows significant socioeconomic differences in the transition to college especially favoring those students who enrolled in a four-year university. Noteworthy is that students who did not enroll in a college were not necessarily disadvantaged in socioeconomic background compared to students who enrolled in a two-year junior college. For example, 40 percent of students enrolling in a four-year university had parents who earned a college degree, while the corresponding percentage was 2 1 for students who did not enroll in a college and 17 percent for students enrolling in a two-year junior college.
Socioeconomic Differences in the Transition to High School
Table 2 presents results of the logistic regression models predicting the likelihood of attending academic vs. vocational high schools. Model 1, which included all independent variables but academic achievement, shows significant socioeconomic differences in the odds of attending academic vs. vocational high schools. With respect to parental education, for example, the odds of attending an academic high school among students whose parents had a college degree are four times the odds among their counterparts whose parents had a middle school or less education (e1.37 = 3.94). Concerning other variables, higher family income and owning homes were also associated with the higher likelihood of attending an academic high school. However, there was no significant gender difference in the likelihood.
Table 2.
Logistic Models Predicting Transition from Middle School to High
| Model 1 (baseline) |
Model 2 (+ academic achievement) |
|||
|---|---|---|---|---|
| Variable | Coef. | (SE) | Coef. | (SE) |
| Parental educationa | ||||
| High school | 0.43* | (0.18) | 0.15 | (0.20) |
| College | 1.37*** | (0.24) | 0.88** | (0.28) |
| Family income (logged) | 0.52*** | (0.14) | 0.49** | (0.17) |
| Owned homes | 0.47*** | (0.12) | 0.26 | (0.14) |
| Female | 0.25 | (0.14) | 0.04 | (0.17) |
| Academic achievement at ninth grade | 0.06*** | (0.00) | ||
| Constant | −2 72*** | (0.73) | −4.49*** | (0.92) |
| Log-likelihood | −778.22 | −599.73 | ||
| Pseudo (McFadden’s) R2 | 0.08 | 0.29 | ||
| N | 1,519 | 1,519 | ||
Data source: Korea Education and Employment Panel of 2004–2009
Note. Standard errors are corrected for clustering within schools.
the reference group was middle school or less.
p<.001,
p<.01,
p<.05 (two-tailed tests)
When academic achievement was taken into account (Model 2), the impact of having parents with a college degree was reduced but still significant and the effect of family income also remained significant. In contrast, the effect of having parents with a high school diploma and the effect of home ownership became no longer significant. As expected, academic achievement was significantly associated with the higher odds of attending an academic high school. In sum, socioeconomic background had a significant direct effect on the likelihood of attending an academic high school relative to a vocational high school as well as an indirect effect via its influence on academic achievement.
Socioeconomic Differences in the Transition to College
College enrollment.
Table 3 presents logistic regression results for any kind of college enrollment. Model 1, which included all socioeconomic variables but academic achievement (measured at 12th grade) and the type of high school attended, showed significant socioeconomic differences in the likelihood of enrolling in college relative to no college enrollment. With respect to parental education, for example, students who had parents with a college degree showed a significantly higher likelihood of attending a college relative to no college than their counterparts whose parents had a middle school or less education. Family income and owning homes were also significantly associated with the likelihood of attending a college.
Table 3.
Logistic Model Predicting Transition from High School to College: College Enrollment
| Model 1 (baseline) | Model 2 (+ academic achievement) | Model 3 (+ type of high school attended) | ||||
|---|---|---|---|---|---|---|
| Variable | Coef. | (SE) | Coef. | (SE) | Coef. | (SE) |
| Parental educationa | ||||||
| High school | 0.32 | (0.19) | 0.21 | (0.20) | 0.15 | (0.20) |
| College | 0.59* | (0.23) | 0.21 | (0.23) | 0.03 | (0.24) |
| Family income (logged) | 0.30** | (0.10) | 0.25* | (0.10) | 0.19 | (0.12) |
| Owned homes | 0.56*** | (0.15) | 0.50** | (0.15) | 0.43** | (0.15) |
| Female | 0.03 | (0.16) | −0.12 | (0.16) | −0.13 | (0.16) |
| Academic achievement at 12th grade | — | (0.05) | 0.22*** | (0.05) | ||
| Academic high school track | — | — | 1.02*** | (0.17) | ||
| Constant | −0.48 | (0.55) | −1.23* | (0.57) | −0.97 | (0.62) |
| Log-likelihood | −577.24 | −552.93 | −536.44 | |||
| Pseudo (McFadden’s) R2 | 0.03 | 0.07 | 0.10 | |||
| N | 1,519 | 1,519 | 1,519 | |||
Data source: Korea Education and Employment Panel of 2004–2009
Note. Standard errors are corrected for clustering within schools.
the reference group was middle school or less.
p<.001,
p<.01,
p<.05 (two-tailed tests)
When academic achievement was taken into account (Model 2), the effects of socioeconomic background on college enrollment were substantially reduced. Although family income and owning homes remained significant predictors of attending a college, the effect of having parents with a college degree was no longer significant. As expected, given the role of CSAT scores for college admission, academic achievement measured by a student’s CSAT score was significantly related to the higher odds of attending a college.
The effect of socioeconomic background on college attendance was further reduced, when the type of high school attended was additionally added (Model 3). In fact, family income was no longer a significant predictor of college enrollment after controlling for the high school track. Yet home ownership remained a significant predictor. The effect of academic achievement was also somewhat reduced but remained significant after controlling for the type of high school attended. Finally, attending an academic high school was important to predict college enrollment.
Type of college enrollment.
Table 4 presents multinomial regression results for the type of college enrollment. Model 1 showed significant socioeconomic differences in the likelihood of enrolling in a four-year university, relative to no college enrollment. With respect to parental education, for example, students who had parents with a high school diploma as well as parents with a college degree showed a significantly higher likelihood of attending a four-year university relative to no college than their counterparts whose parents had middle school or less education. Family income and home ownership were also significantly associated with the likelihood of attend a four-year university. Compared to the significant effects of family background variables for a four-year university attendance relative to no college, socioeconomic differences in the comparison between attendance at a two-year junior college and no college were much less: only home ownership was associated with the likelihood of attending a two-year junior college relative to no college.
Table 4.
Multinomial Logistic Model Predicting Transition from High School to College: Type of College Enrollment
| Model lb (baseline) | Model 2b (+ academic achievement) | Model 3b (+ type of high school attended) | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| two-year college | four-year university | two-year college | four-year university | two-year college | four-year university | |||||||
| Variable | Coef. | (SE) | Coef. | (SE) | Coef. | (SE) | Coef. | (SE) | Coef. | (SE) | Coef. | (SE) |
| Parental educationa | ||||||||||||
| High school | 0.12 | (0.23) | 0.53** | (0.19) | 0.12 | (0.23) | 0.40* | (0.20) | 0.08 | (0.23) | 0.33 | (0.19) |
| College | −0.26 | (0.26) | 1.04*** | (0.25) | −0.27 | (0.26) | 0.60* | (0.26) | −0.38 | (0.27) | 0.40 | (0.27) |
| Family income (logged) | 0.14 | (0.11) | 0.49*** | (0.12) | 0.14 | (0.11) | 0.37** | (0.11) | 0.11 | (0.12) | 0.27* | (0.13) |
| Owned homes | 0.32* | (0.16) | 0.71*** | (0.16) | 0.32* | (0.16) | 0.69*** | (0.17) | 0.28 | (0.16) | 0.62*** | (0.17) |
| Female | 0.10 | (0.18) | −0.01 | (0.17) | 0.09 | (0.18) | −0.29 | (0.17) | 0.08 | (0.18) | −0.31 | (0.18) |
| Academic achievement at 12th grade | — | — | 0.01 | (0.06) | 0.56*** | (0.06) | −0.05 | (0.06) | 0.42*** | (0.06) | ||
| Academic high school track | — | — | — | — | 0.58** | (0.19) | (0.21) | |||||
| Constant | −0.17 | (0.59) | −2.32*** | (0.62) | −0.22 | (0.64) | −3.63*** | (0.63) | −0.07 | (0.68) | −3.49*** | (0.66) |
| Log likelihood | −1376.74 | −1248.36 | −1212.58 | |||||||||
| Pseudo (McFadden’s) R2 | 0.05 | 0.14 | 0.17 | |||||||||
| N | 1,519 | 1,519 | 1,519 | |||||||||
Data source: Korea Education and Employment Panel of 2004–2009
Note. Standard errors are corrected for clustering within schools.
the reference group was middle school or less
the base category was no college enrollment
p<.001,
p<.01,
p<.05 (two-tailed tests)
In Model 2, the effect of socioeconomic background on four-year university enrollment was much reduced. For example, the coefficient of having parents with a college degree was reduced from 1.04 to .60 (approximately 42 percent reduction). Yet parental education, family income, and home ownership were all significant predictors of attending a four-year university even after controlling for academic achievement. Academic achievement was significantly related to the higher odds of attending a four-year university. However, academic achievement rarely mediated the effects of other variables for the likelihood of attending a two-year junior college compared with no college enrollment. Academic achievement was not related to the odds of attending a two-year junior college relative to no college enrollment.
In Model 3, the effect of socioeconomic background on attendance at a four-year university was further reduced. In fact, parental education was no longer a significant predictor of attending a four-year university. Yet family income and home ownership remained significant predictors. The effect of academic achievement was also somewhat reduced from .56 to .42 but remained significant after controlling for the type of high school attended. Attending an academic high school was important to predict four-year university enrollment. Specifically, the odds of attending a four-year university relative to no college were about five times greater for students who attended an academic high school than their counterparts from a vocational high school (e1.59 =4.90). Academic high school graduates also showed the higher likelihood of attend a two-year junior college relative to no college enrollment.
Assessing EMI in Korea: Predicted probabilities
The first panel of Table 5 presents estimated probabilities of attending an academic vs. vocational high school between advantaged and disadvantaged students on the basis of two models estimated in Table 2. The predicted probabilities of attending an academic high school estimated from Model 1 (including all other variables except for academic achievement) showed 92 percent vs. 54 percent for the advantaged and disadvantaged students, respectively (38 percentage point difference). When academic achievement was taken into account (Model 2), the prediction for disadvantaged students approached 73 percent but still lagged behind the advantaged students (93 percent).
Table 5.
Predicted Probabilities for Advantaged (High) and Disadvantaged (Low) Students
| Transition to High school | Transition to College | ||||||
|---|---|---|---|---|---|---|---|
| Logistic Models | Logistic Models | Multinomial Logistic Models | |||||
| Vocational High School | Academic High School | No College Enrollment | Any College | No College Enrollment | Two-Year College | Four-Year University | |
| Model 1 | |||||||
| Advantaged | 0.08 | 0.92 | 0.08 | 0.92 | 0.08 | 0.17 | 0.76 |
| Disadvantaged | 0.46 | 0.54 | 0.25 | 0.75 | 0.25 | 0.46 | 0.29 |
| Model 2 (M1 + academic achievement = average) | |||||||
| Advantaged | 0.07 | 0.93 | 0.09 | 0.91 | 0.10 | 0.22 | 0.69 |
| Disadvantaged | 0.27 | 0.73 | 0.20 | 0.80 | 0.23 | 0.44 | 0.33 |
| Model 3 (M2 + high school track) | |||||||
| Academic High School | |||||||
| Advantaged | — | — | 0.08 | 0.92 | 0.08 | 0.20 | 0.72 |
| Disadvantaged | — | — | 0.14 | 0.86 | 0.16 | 0.41 | 0.43 |
| Vocational High School | |||||||
| Advantaged | — | — | 0.19 | 0.81 | 0.24 | 0.33 | 0.43 |
| Disadvantaged | — | — | 0.31 | 0.69 | 0.34 | 0.48 | 0.18 |
Note. Advantaged students were average achieving females whose parents had a college degree; whose logged family income was the average logged family income among families whose parents had a college degree; and whose families owned their own homes. In contrast, disadvantaged students were average achieving females whose parents had a middle school or less education; whose logged family income was the average logged family income among families whose parents had a middle school or less education; and whose families did not own their homes.
The second panel of Table 5 presents estimated probabilities of attending any college vs. no college between socioeconomically advantaged and disadvantaged students on the basis of the three models estimated in Table 3. The predicted probabilities estimated from Model 1 showed that 92 percent and 75 percent of the advantaged and disadvantaged students, respectively, were predicted to attend any college (the difference of 18 percentage points). Socioeconomic differences in the likelihood of enrolling in a college were reduced to 11 percentage points when academic achievement was taken into account in Model 2 (91 percent vs. 80 percent for the advantaged and disadvantaged students, respectively). When the type of high school attended was included (Model 3), 92 percent of the advantaged students, who attended an academic high school, were predicted to attend a college; the corresponding percentage was 86 percent for the disadvantaged students who attended an academic high school. On the other hand, the prediction for the advantaged students, who attended a vocational high school, showed that 81 percent should attend a college, whereas the corresponding percentage was 69 percent for the disadvantaged students who attended the same high school track (11 percentage point difference).
Finally, the last panel of Table 5 presents estimated probabilities of attending a different type of college between socioeconomically advantaged and disadvantaged students on the basis of the three models estimated in Table 4. With respect to four-year university enrollment, the predicted probabilities estimated from Model 1 showed the difference of 47 percentage points in the probability of enrolling in a four-year university between advantaged and disadvantaged students (76 percent vs. 29 percent). The difference in the probability between advantaged and disadvantaged students was not only substantial but it indicated that qualitatively different destinations were predicted for the two groups of students. That is, it was a four-year university that the advantaged students mostly likely ended up with as of two years after high school, whereas it was a two-year junior college that the disadvantaged students mostly likely did so.
When academic achievement was taken into account (Model 2), this socioeconomic difference in the likelihood of enrolling in a four-year university was reduced to 35 percentage points (69 percent and 33 percent for the advantaged and disadvantaged students, respectively). However, the most likely destination for the disadvantaged students was still a two-year junior college, whereas it was still a four-year university for the advantaged students. In other words, even after controlling for students’ academic achievement, socioeconomic background factors still mattered to make the modal category of destination qualitatively different between students from advantaged and disadvantaged families.
When the type of high school attended was additionally controlled for (Model 3), the prediction for the disadvantaged students who attended an academic high school approached 43 percent for four-year university enrollment but still lagged behind for the advantaged students who attended the same high school track (72 percent), showing a 29 percentage point difference. Compared to the previous two models, in this model the modal category for destination became a four-year university (43 percent vs. 41 percent for a two-year junior college) even for the disadvantaged students. However, when looking at students who attended vocational high schools, the typical destinations for the advantaged and disadvantaged students were different. The modal category for destination was a four-year university for the advantaged students who attended a vocational high school, whereas the modal category for the disadvantaged students who attended a vocational high school was a two-year junior college.
DISCUSSION
Using data for a nationally representative sample of ninth graders in Korea, our study showed that socioeconomic background played an important role in determining the likelihood of attending an academic high school. Without controlling for prior achievement, almost all of the advantaged students (92 percent) attended an academic high school after middle school but only half of the disadvantaged students did so. When prior achievement was included in the model, the difference in the probability of attending an academic high school was considerably reduced but still remained substantial (20 percentage point difference). The predicted probabilities, however, suggest that like their counterpart advantaged students, disadvantaged students typically choose academic high school, rather than vocational high school. In other words, although there are significant differences in the likelihood of attending an academic high school relative to a vocational high school between advantaged and disadvantaged students, even disadvantaged students typically go to academic high school, rather than vocational high school.
When it came to the transition to college, we also found significant socioeconomic disparities with advantaged students to be far more likely than disadvantaged students to go to college. The estimated probabilities based on the logistic regression model predicting college enrollment confirm significant socioeconomic differences in the likelihood of college enrollment. However, they also suggest that even disadvantaged students typically go to college regardless of the type of high school attended. In other words, although there are significant differences in the likelihood of college enrollment between advantaged and disadvantaged students, it does not necessarily mean that disadvantaged students typically do not go to college.
Importantly, as for the transition to different types of college, we found significant socioeconomic differences in the likelihood of attending a four-year university. The predicted probabilities based on the multinomial logistic model also confirm these socioeconomic disparities. Before we controlled for the high school track, we found that advantaged students likely ended up with a four-year university, whereas disadvantaged students typically ended up with a two-year college, supporting EMI. However, when we examined these socioeconomic differences by the type of high school attended, we found somewhat different patterns. Among academic high school graduates, even disadvantaged students typically attended a four-year university, questioning EMI. By contrast, among vocational high school graduates, disadvantaged students typically ended up with a two-year junior college, supporting EMI. In sum, evidence is mixed in Korea, depending on the type of high school attended.
Our study has implications for the relevance of EMI in Korea as well as elsewhere where postsecondary education becomes close to be universal. Our finding suggests that for the transition from middle school to high school, even disadvantaged students typically choose academic high school, rather than vocational high school. Furthermore, our finding suggests that although disadvantaged students attending a vocational high school typically move to a two-year junior college, similarly disadvantaged students who attended an academic high school typically go to a four-year college, rather than a two-year junior college. Note that when we did not control for the high school type in Model 2 in Table 5, the result supported EMI showing that a typical destination was a two-year junior college for disadvantaged students but a four-year university for advantaged students. In other words, our mixed results reveal an important point in evaluating EMI. Depending on educational systems, students diverge into different types of schools not only in the level of college education but even in the secondary school level, like in Korea where students are sorted into academic vs. vocational high schools. Therefore, in examining how socioeconomic background is related to the types of college, it is important to recognize that students were already sorted into different types of high school, which have a quite different prospect for college attendance. In a situation where college education becomes increasingly common but qualitative differentials do exist among higher education institutions, a four-year university may become a typical destination even to disadvantaged students once they attend academic high schools. In short, taking into account students’ educational trajectories before higher education is important in assessing EMI at the stage of higher education.
However, we acknowledge several limitations that may preclude us from drawing a definitive conclusion about the relevance of EMI for Korea. As described earlier, the Korean higher education system is highly stratified along with the selectivity of four-year universities. Therefore, a simple distinction between two-year junior colleges and four-year universities that the current study relied on may not capture nuanced qualitative differentials among higher education institutions in Korea. Indeed, a recent Korean study found more salient socioeconomic gaps in college enrollment between highly selective four-year universities and non-selective four-year universities (Byun and Kim 2010b). Future research should use more diversified and stratified educational destinations of advantaged and disadvantaged students to better test EMI in Korea.
Moreover, educational pathways through which students take to postsecondary education are far more complex in Korea. As described above, many Korean students who failed to earn admission to a college of their choice decide to repeat CSAT until they earn a high score enough to be admitted into that college. Even some students who are officially admitted to less desirable institutions often do not attend colleges, but rather prepare to repeat CSAT (Byun and Kim 2012). Therefore, although we attempted to take into account this complex transition pattern by using an extended period of time for college enrollment (i.e., two years after high school graduation), we might not be able to fully address the complexity of educational pathways among Korean students. A longitudinal study with a longer time span may increase our understanding of the complex patterns of college enrollment among Korea students.
Despite these limitations, our study has broader implications for testing EMI beyond Korea. Secondary education now become universal in almost all developed countries as well as in many developing countries (OECD 2013). In addition, over the past few decades, postsecondary education has dramatically expanded worldwide (Schofer and Meyer 2005). The Korean case suggests that the expansion of secondary and postsecondary education opportunities and the increasing importance of academic credentials in future opportunities may change the contour of qualitative differentials such that the high school track or/and the higher education system is further differentiated within each of the systems. In addition, socioeconomic differentials would appear in form of different pathways to college with more complex enrollment patterns. Indeed, recent research in the United States suggests that college pathways significantly differ by students’ socioeconomic background (Goldrick-Rab 2006). Future research should take into account increasingly diversified qualitative differentials when testing EMI.
Footnotes
In recent years, however, an increasing number of high schools adopted within-school tracking practices (Byun and Kim 2010a; Park 2013). Unfortunately, data that we used did not allow us to examine within-school tracking practices and we acknowledge this as a limitation of our study. Meanwhile, beginning in 11th grade, academic high school education offers two different curriculum tracks: liberal arts (emphasizing on humanities and social sciences) and natural sciences (emphasizing on math and science) (Byun et al. 2012). However, choosing these two different curriculum tracks is determined largely by the student preference rather than by student performance.
Another important factor for the college admissions is high school GPA (Kim and Byun 2006). Yet, generally speaking, CSAT is more important than high school GPA for college admissions even for recent years, although the relative importance between CSAT and high school GPA for college admission varies by year as well as by college.
We also analyzed data from Korea Youth Panel Survey that followed a nationally representative panel of eighth graders (in 2003) to high school and college years (until 2008) and found similar results we reported here. These results are available from the authors on request.
In order to examine how students who were followed up until 2009 and those who were lost before 2009 differed in their characteristics, we conducted supplementary analysis of comparing socioeconomic and academic background characteristics available in the baseline survey (when they were in the ninth grade) between the two groups of samples. Our logistic regression predicting attrition status (as of the 2009 survey) by socioeconomic and academic background variables in the 2004 baseline survey showed no consistent pattern in predicting attrition by socioeconomic background variables. For instance, the relationship with attrition varied depending on parental education, homeownership, and family incomes. In other words, at least with respect to socioeconomic background, there seems no systematic pattern of attrition, which suggests that our results might not have been strongly affected by attrition according to socioeconomic background. However, we did find that lower achieving students in the ninth grade, as reported by their homeroom teachers, were less likely to be followed up until the 2009 survey than higher achieving students in the ninth grade. Therefore, the effect of academic achievement in the ninth grade might have been underestimated in our models, which require some caution in drawing conclusions from our results.
Among respondents who were successfully resurveyed in the first follow up, only 11 respondents (0.6 percent) were found not to go to high school. We excluded these 11 cases from our analysis for transition to high school because of the small sample. Our supplementary analyses including these 11 respondents did not affect the conclusion we reported here. These results are available from the authors on request.
Some students who failed to earn admissions to a college of choice in the first time continue to prepare CSAT more than one year. Therefore, if we had a longer time span for data collection, we could have more respondents who eventually enrolled in a college. 7 We tested the assumption of the independence of irrelevant alternatives (or the IIA assumption) using a Hausman test and a Small-Hsiao test, and found that both supported the IIA assumption, offering warrant for the use of multinomial regression.
Earlier versions of this manuscript were presented at the summer meeting of Research Committee on Social Stratification and Mobility (RC28), August 9th–12th, 2011, Iowa City, IW and at the at the symposium for the Study of Inequality, July 4th, 2014, Seoul, South Korea.
In revising and completing the draft, Hyunjoon Park was supported by the National Research Foundation of Korea Grant (NRF-2013S1A3A2055251), funded by the Korean Government. Soo-yong Byun acknowledges support from the Penn State Population Research Institute of the National Institutes of Health (R24HD041025). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.
Contributor Information
Soo-yong Byun, Associate Professor of Educational Theory and Policy, Department of Education Policy Studies, The Pennsylvania State University.
Hyunjoon Park, Korea Foundation Associate Professor of Sociology, Department of Sociology, University of Pennsylvania.
REFERENCES
- Aud Susan., Hussar Willima, Kena Grace, Bianco Kevin, Frohlich Lauren, Kemp Jana, and Tahan Kim 2011. The Condition of Education 2011 (NCES 2011–033). U.S. Department of Education, National Center for Education Statistics. Washington, DC: U.S. Government Printing Office. [Google Scholar]
- Ayalon Hanna, and Shavit Yossi. 2004. “Educational Reforms and Inequalities in Israel: The MMI Hypothesis Revisited.” Sociology of Education 77: 103–20. [Google Scholar]
- Byun Soo-yong and Kim Kyung-keun. 2010a. “Educational Inequality in South Korea: The W idening Socioeconomic Gap in Student Achievement.” Research in Sociology of Education 17:155–82. [Google Scholar]
- Byun Soo-yong and Kim Kyung-keun. 2010b. “Stratification in Korean Higher Education: Determinants of the College Destinations of General High School Graduates” (in Korean). Korean Journal of Sociology of Education 20(1):73–102. [Google Scholar]
- Byun Soo-yong, Schofer Evan, and Kyung-keun Kim. 2012. “Revisiting the Role of Cultural Capital in East Asian Educational Systems: The Case of South Korea.” Sociology of Education 85(3):219–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Byun Soo-yong and Kim Kyung-keun. 2012. “Determinants of College Attendance among Vocational High School Graduates in South Korea” (in Korean). Journal of Korean Education 39(2):79–107. [Google Scholar]
- Byun Soo-yong, Kim Kyung-keun, and Park Hyunjoon. 2012. “School Choice and Educational Inequality in South Korea.” Journal of School Choice 6(2):158–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chang Sang-soo. 2007. “Family Background and the Choice of High School Sector in Korea” (in Korean). Korean Journal of Sociology 42(2):118–46. [Google Scholar]
- Goldrick-Rab Sara. 2006. “Following Their Every Move: An Investigation of Social-Class Differences in College Pathways.” Sociology of Education 79:61–79. [Google Scholar]
- Johnson David R. and Young Rebekah. 2011. “Toward Best Practices in Analyzing Datasets with Missing Data: Comparisons and Recommendations.” Journal of Marriage and Family 73:926–45. [Google Scholar]
- Kim kyung-keun and Soo-yong Byun. 2006. “Determinants of Children’s Educational Transition in South Korea” (in Korean). Korean Journal of Sociology of Education 16(4):1–27. [Google Scholar]
- Kim Sung Sik. 2008. “A Study on the Influence of Student Backgrounds on Opportunity of Tertiary Education in South Korea” (in Korean). Asian Journal of Education 9(2):27–47. [Google Scholar]
- KEDI. 2010. 2009 Education Statistics (in Korean). Seoul, Korea: Author. [Google Scholar]
- Lucas Samuel R. 2001. “Effectively Maintained Inequality: Education Transitions, Track Mobility, and Social Background Effects.” American Journal of Sociology 106(6):1642–90. [Google Scholar]
- Lucas Samuel R. 2009. “Stratification Theory, Socioeconomic Background, and Educational Attainment: A Formal Analysis.” Rationality & Society 21: 459–511. [Google Scholar]
- OECD. 2007. Education at Glance 2007. Paris: Author. [Google Scholar]
- OECD. 2010. Education at Glance 2010. Paris: Author. [Google Scholar]
- OECD. 2013. Education at a Glance 2013. Paris: Author. [Google Scholar]
- Park Hyunjoon. 2007b.. “South Korea: Educational Expansion and Inequality of Opportunity for Higher Education.” Pp. 87–112 in Shavit Y, Arum R, and Gamoran A (eds.), Stratification in Higher Education: A Comparative Study. Stanford University Press. [Google Scholar]
- Park Hyunjoon. 2007a. “Inequality of Educational Opportunity in Korea: Gender, Socioeconomic Background, and Family Structure.” International Journal of Human Right 11:179–197. [Google Scholar]
- Park Hyunjoon. 2010. Japanese and Korean High Schools and Students in Comparative Perspective.” Pp.255–273 in Dronkers J (eds.), Quality and Inequality of Education: Cross-National Perspectives. Netherlands, Dordrecht: Springer. [Google Scholar]
- Park Hyunjoon. 2013. Re-Evaluating Education in Japan and Korea: De-Mystifying Stereotypes. New York, NY: Routledge. [Google Scholar]
- Raftery Adrian E. and Hout Michael. 1993. “Maximally Maintained Inequality: Expansion, Reform, and Opportunity in Irish Education, 1921–75.” Sociology of Education 66(1):41–62. [Google Scholar]
- Rogers William. H. 1993. “Regression Standard errors in Clustered Samples.” Stata Technical Bulletin 13:19–23. [Google Scholar]
- Schofer Evan and Meyer John W.. 2005. “The Worldwide Expansion of Higher Education in the Twentieth Century.” American Sociological Review 70(6): 898–920. [Google Scholar]
- Shavit Yossi, and Blossfeld Hans-Peter. 1993. Persistent Inequality: Changing Educational Attainment in Thirteen Countries. Boulder, CO: Westview Press. [Google Scholar]
- Yossi Shavit, Arum Richard, and Gamoran Adam (eds.). 2007. Stratification in Higher Education: A Comparative Study. Stanford, CA: Stanford University Press. [Google Scholar]