1 Who Goes to Graduate School in Taiwan? Evidence from the 2005 College Graduate Survey and Follow- Up Surveys in 2006 and 2008 Ping-Yin Kuan Department of Sociology Chengchi Unviersity Taiwan Presented at International Conference on Social Inequality and Mobility in Chinese Societies: Towards a Comparative Study Dec , 2011, Hong Kong
2 Expansion of Higher Education in Taiwan Between 1997 and 2010, the number of colleges and universities in Taiwan has increased from 78 to 164 (about 1/3 are public universities or colleges). The number of graduate programs has increased from 728 to 3,360 in the same period. In 2000, the gross percentage of high school graduates entering colleges (no. of college freshmen/no. of high school graduates*100) is 118%. In 2010, about 95% of senior high graduates and about 80% of senior vocational high graduates went to colleges.
3 Net Percentage of High School Graduates Entering Colleges Senior High Senior Vocational Year %
4 Expansion of Higher Education in Taiwan In 1997, the number of doctoral students in Taiwan is 10,013 and in 2010, the number is 34,178. In 1997, the number of master s students is 38,606 and in 2010, the number is 185,000. Internationally, Taiwan is only second to South Korea in terms of the gross enrollment ratio of tertiary education (% of people between age 18 to 21 enrolled in tertiary education). Source: tent_sn=27359.
5 Research Question In light of rapid expansion of higher education in Taiwan, the competition for an even higher level of education is emerging. Hence, it is important to examine the possible inequality in access to graduate schools, which in turn may lead to inequality in life chances. Essentially, the present research is interested in possible unequal access to graduate schools in terms of family SES, gender, and ethnicity.
6 Three Theoretical Views Model of Educational Transitions (MET) Maximally Maintained Inequality (MMI) Effectively Maintained Inequality (EMI)
7 Model of Educational Transitions Robert Mare proposed this most commonly used analytical model to study educational transitions in early 1980s (1980, 1981). In the United States, levels of educational attainment increased dramatically and that their dispersion declined. In view of educational expansion, one would have expected that educational inequality between social strata declined across cohorts. Mare conceptualized educational attainment process as a sequence of transition points at which students either continue to the next level or drop out. At each transition point, social origins affect the log odds (logit) of continuing to the next level, given that the previous level had been attained.
8 Model of Educational Transitions (MET) Mare s analysis of the 1973 Occupational Change in a Generation data revealed that within cohorts of males, social background effects on educational transitions decline from earlier to later transitions. He also found that for college graduates, the influence of social origins on their decisions to pursue further schooling is almost nil. (Mare, 1980: 301) Other studies (Ethington & Smart, 1986; Stolzenberg, 1994; Mullen, Goyette, & Soares) have similarly found no direct effect of family backgrounds on enrollment of graduate programs. Based on MET, one could also expect that for the current cohort of young people in Taiwan that their parental SES would have little or no effect on their chances of getting into colleges or graduate schools.
9 Maximally Maintained Inequality (MMI) Based on the rational choice approach, Raftery & Hout (1993) proposed the model of maximally maintained inequality, which stipulated that parents and students themselves have great interests in students achievement and together they would mobilize the resources at their disposal to advance the students achievements as fully as possible. Hence, education expansion in and of itself is unlikely to reduce educational inequalities simply because those from more advantaged socioeconomic backgrounds are better placed than others to take up the new educational opportunities that expansion affords.
10 Maximally Maintained Inequality (MMI) Inequality would diminish only once the enrollment rate for the most socioeconomically advantaged group reaches saturation point. Based on MMI, one would expect that once college enrollment becomes universal, the class difference would persist at a higher level of education.
11 Maximally Maintained Inequality (MMI) Shavit, et al. (2007), using datasets from 15 countries, have shown that only when eligibility rates of higher education exceeding 80% and attendance rates over 40% would the impact of parental education and father s class started to decrease. Torche (2011) analyzing 5 longitudinal datasets in the U. S. finds that the intergenerational association is strong among those with low educational attainment; it weakens or disappears among bachelor s degree holders but reemerges among those with advanced degrees, leading to a U shaped pattern of parental influence.
12 Effectively Maintained Inequality (EMI) Samuel Lucas (2001, 2009) proposed the thesis of effectively maintained inequality, which argued that in the face of educational expansion, those from more advantaged socioeconomic backgrounds would secure for themselves a qualitatively better kind of education at any given level. Tsai & Shavit (2007) using Taiwan Social Change Survey data collected in 2000 found that as tertiary education expanded, white collar families lost interest in education offered by junior colleges and preferred universities. Hence, in the case of Taiwan, it is expected that those with advantaged family backgrounds would attain better (public) universities and would also have a better chance to access to graduate schools.
13 Gender Differences In Taiwan, the gross enrollment ratio in tertiary education for women is 85.1% and 79.5% for men in The high proportions of females in higher education is similar to many other developed countries in recent years (Buchmann, et al. 2008). There are gender differences in terms of type of institution, specialization (major), and enrollment in graduate programs: Women in junior colleges: increased from 53.7% in 2001 to 71.3% in Women in 4 year colleges: 49.9% in 2001 and 48.9% in Women in master s programs: 36% in 2001 and 43.3% in Women in doctoral programs: 22.8% in 2001 and 28.9% in 2010.
14 Gender Differences in Fields of study % OECD 100 Taiwan Education Humanities, Arts Social Science, Law, Business Sciences Engineering Agriculture Medicine, Health Services
15 Ethnic Inequalities Four major ethnic groups in Taiwan (in term of father s ethnicity, 2005): Hokkien (72%), Hakka (15%), Mainlanders (10%), Aborigines (2%). Tsai & Shavit (2007): Mainlander advantage in attending higher education in the earlier cohort ( ) disappears in recent cohorts. Aborigines disadvantage, however, increased.
16 Research Design Data The college graduate survey gathered by Taiwan Integrated Postsecondary Education Database (TIPED) in 2005 and the data of the follow up surveys gathered in 2006 and In 2005, TIPED surveyed 183,564 college graduates of the year. In 2008, the sample size of the web survey is 68,294. The sample analyzed: 70% of linked samples. Significance test: p <.0001
17 Research Design Analysis Due to severe sample attrition, there is a possibility of selection bias. Heckman selection model is used to adjust for possible bias. Stata: binary and multinomial logistic regression
18 Research Design Outcome variables 4 types of undergraduate institution: Public University/College, Public Technological University/College, Private University/College, Private Technological University/College (reference group) Graduate school attendance since 2005: Yes (1) vs. No (0)
19 Research Design Explanatory variables Family SES: Parents level of education: primary or below, junior high, senior high (ref.), junior college, 4 year college, graduate school Parents occupation: coded into 5 skill levels Parents work sector: public (ref.), private, not employed. Student loan: Yes (1) vs. No (0)
20 Research Design Gender: Female (1) vs. Male (0) Ethnicity: Hokkien (ref.), Hakka, Mainlander, Aborigines, Other Variables related to learning experiences: fields of study (sciences, social sciences, humanities and arts), having part time job, double major, minor, interdisciplinary programs, as research assistant, student status (general vs. others), grade point average (4 levels) Institutional factors: type of institution, daytime division, technological or vocational programs.
21 Research Design Variables in the selection equation Number of pages completed in 2005 survey 2005 (0-4; dummy coded with 0 as the reference) Number of pages completed in 2006 survey 2005 (0-5; dummy coded with 0 as the reference) Gender, Field of Study, Institutional factors
22 Findings Selection bias in completing 2008 survey Public or private university graduates have a lower completion rate Graduates of 4-year technological or 2-year technological programs ( 四 技 or 二 技 ) have a higher completion rate Graduates of the daytime division ( 日 間 部 ) have a higher completion rate
23 Findings Outcome: type of undergraduate institution Higher odds of entering either public/private universities: Male, Hokkien, higher level of father s and mother s education, mother works for the public sector, no student loan Higher odds of entering public tech: Better family backgrounds. However, the size of effects tends to be smaller than the above outcomes. Supports EMI thesis.
24 Findings Outcome: graduate-school attendance Initial model: Only family SES, gender, and ethnicity are included Better family backgrounds have higher odds Male or Hokkien has a higher odds Supports MMI thesis The coefficient size of Heckman selection model is smaller than that of OLS analysis; by 27.5%~64%
25 Findings Full model: further includes variables related to college learning experiences and institutional factors Family SES: Only father s educational level and student loan remain significant. Male No difference among ethnic groups College learning experiences Lower odds: those who took interdisciplinary programs, those who worked during college years Higher odds: as a research assistant, major in science related fields, better GPA
26 Findings Institutional factors Those who entered public universities, daytime division, general (non technological or vocational) program have higher odds.
27 Conclusions In the face of expansion of higher education in Taiwan, those from more advantaged socioeconomic backgrounds secure for themselves a qualitatively better kind of college education (supporting EMI thesis) and are now advancing themselves to a higher level of education (supporting MMI thesis). Despite near gender equality in entering colleges, women still tend to enter private technological colleges and are less likely to pursue graduate degrees. Hokkien group seems to gain advantage in entering better institutions of higher education, which is quite different from what researchers had observed in the past.
28 Conclusions The sample bias caused by serious sample attrition may not be fully accounted for in the present study. The alternative: TEPS B (Taiwan Education Panel Survey and Beyond) data
29 Supplementary Analysis with TEPS-B Data TEPS B Data A continuation of TEPS which surveyed nearly 20,000 senior high, senior vocational high and five year junior college students in 2001 (11th grade) and 2003 (12th grade). Currently, TEPS B has completed 3,888 follow up face toface interviews of this cohort. In term of the distribution of predicted IRT scores (achievement scores) obtained in 2003, the follow up survey sample is a reasonably representative subsample of the 2001/2003 TEPS sample.
30 Preliminary Findings of TEPS-B TEPS-B Educational Attainment % Senior high Senior vocational 5 year junior coll 2 year junior coll Tech. univ/coll 4 year coll Master Doctoral
31 Preliminary Findings of TEPS-B With the attainment of high school diploma as the basis for comparison, the multinomial logistic regression model that includes both parents educational levels and ethnicities, respondent's sex and the predicted IRT scores, the analysis found that there is a higher relative risk ratio (rrr) of entering Junior college: father s education level is higher, predicted IRT is higher Tech univ/coll: both parents education level are higher, female, predicted IRT is higher (if mother is Hakka, the rrr is lower) 4 year univ/coll: both parents education level are higher, female, and predicted IRT is higher Graduate school: both parents education level are higher, and predicted IRT is higher (if mother is Hakka, the rrr is lower)
32 Preliminary Findings of TEPS-B Including only those who entered 4-year colleges in the analysis and with entering public 4-year univ/coll as the basis of comparison, there is a higher relative risk ratio of entering Private 4 year univ/coll: lower predicted IRT Public tech univ/coll: female, mother is Hakka Private tech univ/coll: female, lower predicted IRT
33 Preliminary Findings of TEPS-B Who goes to graduate school? Both parents levels of education are higher Male Attended public univ/coll (>private univ/coll > private tech univ/coll > public tech univ/coll) Better GPA in college Predicted IRT (in 12th grade) is higher
34 Preliminary Findings of TEPS-B A similar U shaped pattern of the influence of parental level of education. Consistent findings about gender differences.
35 Thanks for your attention! Q & A