Educational Achievement and the Evaluation of the Nine-year Compulsory Education Policy: The Case of Taiwan
II. The Empirical Model and Estimation Method
In the literature, family background and living environment are important factors affecting individual’s educational attainment and thus they were widely examined in the empirical studies.5 However, in their survey article Haveman and Wolfe (1995) propose that a more comprehensive framework for determining children’s educational attainment should contain three primary factors,
4 Chu, Tsay and Yu(2005) and Chu, Xie and Yu(2005) also used PSFD data to investigate educational achievement, however, they do not include family educational resources and personal characteristics variables.
5 See, for example, Datcher (1982), Hill and Duncan (1987), Teachman (1987), Graham, Beller, Hernandez (1994), Ermisch and Franceconi (2000), and Aakvik, Salvanes, and Vaage (2005), among others. Ashenfelter and Rouse (1998) show that in their twins sample up to 60% of the cross-sectional variation in educational attainment can be attributed to
the choices made by the government, the choices made by the parents, and the choices made by children themselves. The part of children’s decisions such as ability, motivation, and values are mostly neglected in the literature.6 In this study besides family background and living environment factors we also include personal characteristics to measure individual’s heterogeneity.
The empirical model for educational achievement is specified as:
i
Where Y is for the year of education, FAM is for family background variable, CHR is for variables of personal characteristics, GEN is a dummy variable for gender (1 for female and 0 for male), UBN is a dummy variable for region (1 for rural and 0 for urban), ETH is a dummy variable for father’s ethnicity (1 for Taiwanese and 0 for mainland Chinese), COH is a variable controlling for cohort effect, and ε is the disturbance term.
In order to evaluate the impact of the Nine-year Compulsory Education policy, this paper first tests the policy impact on the educational achievement across gender, region, and ethnic groups, respectively. The empirical model for policy evaluation can be written as
i
Where Z stands for variables of personal characteristics, family background, and Cohort year, POL is a dummy variable for compulsory education policy (1 for those receiving compulsory education and 0 for those not). The coefficientα5 in equation (2) represents the policy effect of the compulsory education. In order to take policy intensity into account, we also multiply the policy dummy by the number of junior high school. The coefficient α6 in equation (3) represents the policy impact on the differences of educational achievement between women and men. If α6 is positive (negative), it implies that compulsory education reduces (enlarges) the educational gap between women and men. By the same token, the coefficients of interaction terms in equations (4)
6 Few exceptions include prior choices made by the child are religiosity in Hill and Duncan (1987) and school
and (5) represent the policy effects on the educational gap between regions and ethnic groups, respectively.
As the educational environment encountered by gender, region, and ethnicity groups might be different due to interactions between groups, estimations in equations (3)-(5) may be biased. To avoid the problem, we further perform multiple comparisons of policy effects on gender, region, and ethnic groups to dif
7
ferentiate the policy effects on particular groups (such as Taiwanese women, rural women, and rural Taiwanese) and also allow for the possible interactions between groups.
The model for policy evaluation is further modified as
i
For example, the coefficient α16 in equation (6) evaluates the policy effect on the educational achievement of rural Taiwanese women, the group that tends to have least education.8 The estimation results of equations (6) will enable us to verify the real effects of compulsory education policy on gender, region, and ethnicity.
teristics, those studies fail to precisely estimate educational achievement (Loah (2001)). This paper adopts the Panel Study of mia Sinica. PSFD collects extensive family information including spouse, parents, children, siblings of that individual,
III. Description of Variables and Data Analysis
In the previous analysis of Taiwan’s educational achievement, the most widely used data are either from the Manpower Utilization Survey or the Survey of Family Income and Expenditure.
However, due to the lack of information in kinship and personal charac
Family Dynamics, which has been conducted yearly since 1999 by the Acade
7 Additional comparison groups will reduce the importance of biases or random variation in a single comparison group.
See Meyer (1995) for a discussion of advantages using multiple comparison groups for policy evaluation.
8 In the samples of PSFD, the average years of education for rural Taiwanese women are 3.99, whereas those for urban
and
family socioeconomic status. Though family income refle
the parents/siblings of the individual's spouse.9 The 1999 pilot survey contained 1000 random samples aged from 36 to 45 (born between 1953-1964) in Taiwan, according to the household registration provided by the Ministry of the Interior. This paper uses a mixed sample of 4,110 individuals from the 1999, 2000, and 2003 PSFD surveys ((RI1999, RI2000, and RI2003). The 1999 survey contains 999 samples aged 35-46, 2000 contains 1959 samples aged 46-65, and 2003 survey contains 1152 samples aged 27-39.
In current literature, family background factors influencing an individual’s educational achievements include socioeconomic status, numbers of siblings, birth order, and location of childhood (Butcher and Case (1994); Greenhalgh (1985); Hauser and Kuo (1998); Haveman and Wofe (1995); Lillard and Willis (1994) and Huang (2000), among others). Among them, family socioeconomic status has been highly informative, and parents’ education, parents’ occupation, and family income have been used as proxies for
cts family resources and economic status, people are usually reluctant to report their true family income (it may be underreported or overreported). Thus, the data on family income are either unavailable, especially during their children’s years of study, or might be subject to bias or error.10 In contrast, the parents’ occupations are usually available and more credible in terms of stable occupational trend in observed worker’s life cycle. Moreover, PSFD contains data on father’s occupation during their children schooling years. Because women in the labor market consist a small proportion of the total female population during the early stages of Taiwan’s economic development, we use the father’s occupation to proxy the family socioeconomic status.
We also include variables like talent training, after-school supplementary education, family transfer to better school district, and incentive offer for academic performance by parents to capture extra educational resources provided by family.11 Another advantage to include family
mainland Chinese men are 13.36.
9 For a detailed description and the design of the PSFD, see the official website at http://psfd.sinica.edu.tw .
10 Another problem of using family income is that the short-run aspect of family income may not reflect the truth e the most permanent income of family. Carneiro and Heckman (2003) indicate that long-term income effects ar
important. Using data from Norway, Aakvik, Salvanes, and Vaage (2005) also find that permanent family income is much more important for educational attainment than is short-term credit constraint.
11 Teachman (1987) finds evidence that parents use resources to create a home environment like having an individual
educational resources is that family educational resources are more important than family income in determining children’s educational achievement and the control of family educational resources will also
filial duty and glory for the family in Chinese society.
Variable name Description
abstract the possible resources effect contained in family socioeconomic variables such as parents’ education and family income.
You can lead horses to river but you cannot force them to drink. Besides family background, individual’s heterogeneity represented by personal characteristics, like ability, attitude, and values of life, are also considered to be important factors driving individual’s educational achievement.12 Our variables include scholarship for academic performance, doing part time job during schooling, and two influential traditional values for
As the sample aged between 27 to 65 years old, additional cohort dummies are added to control for possible cohort effects derived from different macro environment.
Table 1 presents the description of all variables used in the paper, while Table 2 shows the summary of basic statistics of the variables.
Table1. Variable Description
Education level Eight levels: Illiterate and self-study, primary school, junior high school, senior high and vocational school, Junior college, University, Master’s degree, Ph.D. degree.
Years of education Years of education for vocational school, Juni
primary school, junior high school, senior high and or college, University, master degree, Ph.D. degree are specified as 6, 9, 12, 14, 16, 18, 20, and 24 years.
Personal Characteristics:
Dummy variable: 1 for female, 0
Scholarship for performance
for obtaining scholarships based on academic
Scholarship for family
sed on low family income ith certain academic performance requirement during school years, 0 otherwise.
Gender for male.
Part-time job Dummy variable: 1 for having part-time job to earn money to support oneself or family during schooling years, 0 otherwise.
Dummy variable: 1
academic performance during school years, 0 otherwise.
Dummy variable: 1 for obtaining scholarships ba
poor w
Filial duty One’s opinion on the idea “one must sacrifice personal interests in order to
their children. ? also find that easy access to college and university has a positive effect on educational achievement.
12 Heckman (2001) finds that more able people also tend to be more efficient in schooling and learning. More
accomplish one’s parents’ wishes”: scale from 1 to 5, 1=not important, up to the
family lutely important.
Fam :
E ludes aborigines, Fujian, and
Hakka), 0 for mainland Chinese.
, 9, 12, 14, 16, 18, 20, and 24 years.
education
Father’s ations including professionals, administrators and workers is e reference group.
ks in Public sector
ining
Supplementary edial or supplementary
wise.
R Reward offered by parents for academic performance, increasing scale from 0
Family transfer
mily to a district with better education environment or N Number of siblings in the family.
Birth order brothers or sisters Coh
O Young
Dummy variable: Old = born before 1950, Middle-aged = 5=absolutely important.
Glory for Ones opinion on the idea “one must do something to glorify ones family”:
scale from 1 to 5, 1=not important, up to 5=abso ily background
thnicity Father’s ethnicity: 1 for Taiwanese (which inc Father’s
education
Years of education for primary school, junior high school, senior high and vocational school, Junior college, University, master degree, Ph.D. degree are specified as 6
Mother’s Same as above.
Dummy variables: occup
occupation executives, clerks, sales workers, services workers, agriculture related workers, production operators and laborers, and agriculture related
th
Father wor Dummy variable: 1 if the father works in the public sector, 0 otherwise.
Talent tra Dummy variable: 1 for attending talent training (such as studying piano, painting, calligraphy, dance, etc.) during school years, 0 otherwise.
Dummy variable: 1 for having after-school rem education
eward for
education, 0 other school
performance
to 3 (0=none, 1=seldom, 2=some times, 3=often)
Dummy variable: 1 for having the experience before age of 16 that parent deliberately moved fa
easy access to school, 0 otherwise.
umber of siblings
Ranking in birth order.
Number of Number of older brothers, older sisters, younger brothers, and younger sisters in the family.
Region
ort:
ld
Dummy variable: 1 for residency in an urban area before age 16, 0 otherwise.
Middle-age
born between 1951
Education policy ine years of compulsory education, 0
and 1960, and Young = born after 1961 (Old is the reference group).
Dummy variable: 1 for receiving n otherwise.
Table 2. Summery of Basic Statistics Samp
Variable le size Mean Variance
Educational attainment:
Illiterate and
self-study 3636 0.0842 0.2781
Primary school 3636 0.2770 0.4476
Junior high school 3636 0.1337 0.3403
Senior high or
vocational school 3636 0.2643 0.4410
Junior college 3636 0.1194 0.3243
University 3636 0.1004 0.3006
Master’s degree 3636 0.0182 0.1335
Ph.D. degree 3636 0.0028 0.0524
Years of education 3636 9.7063 4.6054
Personal
performance 3633 0.1214 0.3266
Scholarship for poor
Aborigine 3629 0.0215 0.1450
Fujian 3629 0.7812 0.4135
Hakka 3629 0.1144 0.3183
Mainland Chinese 3629 0.0829 0.2758
Father’s education 3623 4.9095 4.8118
Mother’s education 3630 2.9645 3.8363
Father’s occupation
Professionals 3598 0.0439 0.2049 Administrators
and executives 3598 0.0503 0.2186
Clerks 3598 0.0675 0.2510
Sales workers 3598 0.1226 0.3280
Services workers 3598 0.0584 0.2345
Agriculture
related workers 3598 0.4341 0.4957
Production
operators and
laborers 3598 0.2232 0.4164
Father in public
Talent training 3635 0.1356 0.3424
Supplementary
education 3633 0.2188 0.4135
Reward to school
performance 3635 0.0646 0.2459
Family transfer 3628 0.0510 0.2200
Birth order 2.0023
Num er
Cohort:
3635 3.0564
ber of old
brothers
ung 3635 1.0094 1.2149
Number of yo
brothers 3635 1.1568 1.1575
Number of older
sisters
nger 3635 1.0470 1.2680
Number of you
sisters 3635 1.0798 1.2510
region 3633 0.4646 0.4988
Old 3636 0.3650 0.4815
Middle-aged 3636 0.2998 0.4582
Young
riable 3636
3636 0.3353
0.4450 0.4721
0.4970 Policy va
Source: RI1999, RI2000, and RI20 3 surveys, Panel St0 udy of Family Dynamics, Academia Sinica.
From Table 2, regarding personal characteristics, 28% of people had part-time jobs to earn money to support oneself or one’ ily during their school years, 14% (0.485 for male and 0.515 for scholarships for either school performance or financial need. The average sca pt of filial duty and glory for the family are 3.42 and 3.81 (in a full scale of 5), respectively, implying that th traditional va widely rooted in Chinese society.
ily background verage the f ducation is years higher than years for the f and 3 years mother). T ively few years of
fathers’ occupations are work in the agriculture sector, consisting of 43.41% of the sample, followed by production operators or laborers (22.32%). In the early agriculture-oriented economy, most people were engaging in agricultural activities or jobs with low skill content. About 16.1% had fathers working in public sector and 45.88% whose mothers are also working. About 7.84%
children live in a single family house.
As for father’s ethnicity, the majority ethnic group is Fujianese (78.12%), followed by Hakka (11.44%), mainland Chinese (8.29%), and aborigines (2.15%). This ethnicity composition of the sample is fairly representative to that of Taiwan total population. In this paper, following conventional classifications, we define the Taiwanese as the summation of Fujianese, Hakka, and aboriginal ethnic groups.
The average number of children in a family is 4.77 people and the number for older brothers, older sisters, younger brothers, and younger sisters are 0.89, 1.00, 0.96, and 0.92. Among them, the proportion of first-born children is 5.25% and that of last-born children is 19.53%.
About 13.56% and 21.88% received latent training and after school supplementary education, respectively; while 6.46% and 5.10% experienced rewards for academic performance offered by parents and family deliberately transfer to a better school district, respectively.
The sample used in this paper consists of people born between 1934 and 1976. During this period of more than forty years, the economic situation and educational environment have changed tremendously. Thus, we also include a cohort dummy variable to control for the effects of age.
s fam female) received
le value for the conce
e two lues are
As for fam , on a ather’s e about 2
mother’s (4.91 ather for the he relat
parental education is expected as the age of the sample’s parents is above 45. The majority of
Acco
und.
On a
rding to birth year, we classify the sample into three cohorts: old (born before 1950), middle-aged (born 1951-1960), and young (born after 1961). The sizes of the samples are 36.5%, 30% and 33.5%, respectively. For robustness test, we also use birth year dummies to control for year effect.
Table 3 presents educational achievement by personal characteristics and family backgro verage, the duration of men’s’ education is 10.62 years, while women’s’ is 8.83 years. Except for primary school, the proportion of men is higher than that of women at all educational levels.
Table 3. Basic Data Analysis
Educational Achievement
Years of Primary Junior
school
Senior vocational
Junior University Master Ph.D.
Gender
education school high high and school Private sector 9.155 0.310 0.144 0.256 0.105 0.074 0.012 0.003 Father
Middle education 12.877 0.069 0.083 0.335 0.237 0.229 0.036 0.003 Higher education 14.144 0.038 0.019 0.260 0.202 0.351 0.115 0.005 Ethnicity
Aborigine 0.4827.376 0.176 0.200 0.035 0.000 0.000 0.000
(3.622)
ddle-age 0.30610.047 0.157 0.294 0.104 0.095 0.012 0.004 (3.934)
(2.705) Father’s workplace
Public sector 12.502 0.109 0.080 0.308 0.192 0.236 (3.776)
(4.537)
’s occupation
Professionals 0.18911.520 0.051 0.214 0.209 0.219 0.051 0.000 (4.783)
executives 12.297 0.146 0.059 0.274 0.210 0.164 0.096 0.014 (4.399)
Clerks 0.16711.548 0.103 0.267 0.160 0.221 0.036 0.000 (4.351)
(4.148)
Services workers 10.782 0.222 0.129 0.238 0.173 0.165 0.020 0.000 (4.399)
Agriculture related 7.311 0.427 0.150 0.180 0.049 0.029 0.002 0.003 Production
Sales workers 10.735 0.217 0.125 0.322 0.136 0.140 0.015 0.002
workers (4.342)
orers 10.568 0.222 0.143 0.344 0.151 0.090 0.012 0.002 (3.887)
Years of education for children whose father is low (primary school and below), l (junior high en h , and vocational school), and high (junior college
. This im hat ig e
one’ r is rea s e na vem ill
, if r la nes hil edu tta a s
am et ou pared with 10.58 years for children whose father 7 years for n w at
Furth th ort hig cat also r fo c
ther is mainl es t hild os r i ane T ,
ducational achievement between ethnic groups is considered to be evident.
s the economy d , t ng erat de ceiv ed
l on. The years of education fo d, m ec
r worke lic r, his ch ren tha e c
in secto The average years of education for the former are 12.5 and
’s education mid leve school, s ior hig school
and above) are 8.68, 12.88, and 14.14 years, respectively plies t the h her th educational level of s fathe ; the g ter one’ ducatio l achie ents w be.
As for ethnicity the fath is main nd Chie e, his c dren’s cation a inment verage 13 years, the highest ong all hnic gr ps, com
is Hakka, 9.2 childre hose f her is Fujianese, and 7.38 years for children whose father is an aborigine. ermore, e prop ion in her edu ion is highe r those hildren whose fa and Chin e anth hos ce ren wh e f thea s T iwa se. In aiwan e
A eveloped he ouy er eng ion tens d t reo e more ucation than the o der generati r ol iddle-age, and young people are 6.59, 10.05, and 12.78 years, resp tive .ly13
If the fathe d in pub s toec ild received more education n t osh hil ren d whose father worked pri ate v r.
13 ed reg ethn ws educational
a Further examination of thr, region, ity is igher for d than for ung peop . p between gende e years of
and ethnic ucation for each age by gender, ion, and icity sho that the
g h ol yo le
9. the latter. T ainly cau ople who worked in the public sector receiv
or ild d us r th of on.14 e s
s con n
d highest (12.3) years of education compared to the
lo i ho her in ultu tor e e
education em te am i r
The proportion of higher education is r, above 50%, for children whose father and exe tive or p ofession l.
ference in educ ch en en t. s
f education in urban a erage 10.8 yea s, while rural areas average 8.4 years. As for the impl
d policy for cross
ity, especially gender.
Amo
nd urban areas and its magnitude was greater for mainland Chinese than
policy effect reduces the educational gap between mainland Chinese and Taiwanese and its
16 for his is m be se pe ed
education subsidies f their ch ren an this th educed e cost educati Th father’
occupation, a proxy for family oc eio omic status, shows that when the father works as a a ministrator and executive, his children have the
west (7.31) years for those ch ldren w se fat worked agric ral sec . How ver, th deviations of ac ievh ent are modera ong children whose father wo edrk n o het
occupations. greate
is an administrator cu r a
Dif ational a ievem t betwe rural and urban areas is also apparen Year
o reas av r
ementation of the Nine-year Compulsory Education policy, before the policy, average years of education were 7.6 years and after the policy the average years of education were 12.4. The Nine-year Compulsory Education policy seems to improve educational achievement in Taiwan.
Table 4 further classifies educational achievement by gender, region, ethnicity, an
-table analysis. Before the compulsory education policy, the educational gap was the largest (5.1 years) between ethnic groups, then 4.44 years between gender, and 2.13 years between rural and urban region. After the compulsory education policy, the corresponding figures for the educational gap are 1.36, 0.4, and 1.45 years, respectively. Looking at the data, basically, the implementation of the Nine-year Compulsory Education policy in 1968 seems to have in some ways reduced the educational gap between gender, region, and ethnic
ng them, the educational gap for Taiwanese (vs. mainland Chinese) between rural and urban areas dropped from 1.73 (3.27) to 1.29 (1.68), which implies that the policy in effect shrunk the educational gap between rural a
for Taiwanese. Conversely, the educational gap between mainland Chinese and Taiwanese in urban (vs. rural) areas was reduced from 4.94 (2.40) to 0.99 (0.60) years, which implies that the
14 The educational subsidies in public sector are actually progressive along the educational levels, thus it provides extra
magnitude is greater for urban than for rural areas.
In term of gender differences in education, after the policy the reduction in the education gap between women and men was greater between mainland Chinese than between Taiwanese, and the scale was greater for mainland Chinese men than for mainland Chinese women. Regardless of gender, after the policy the reduction in the educational gap in urban areas was greater than that in rural areas. However, the largest increase in educational achievement before and after the compulsory education policy falls to women in rural areas, whose average years of education increased from 3.98 to 11.02 years.
In term of gender differences in education, after the policy the reduction in the education gap between women and men was greater between mainland Chinese than between Taiwanese, and the scale was greater for mainland Chinese men than for mainland Chinese women. Regardless of gender, after the policy the reduction in the educational gap in urban areas was greater than that in rural areas. However, the largest increase in educational achievement before and after the compulsory education policy falls to women in rural areas, whose average years of education increased from 3.98 to 11.02 years.