Table 5 presents the results of the empirical models on educational attainment. All the regression models in Table 5 have a F value which passed the 99% statistical significance level and the adjusted F2 is over 0.5, which implies excellent explanatory potential and that the models are appropriate.
As in the literature, family background variables have significant effects on individual’s educational achievement.15 The effect of the father’s education has a significant positive effect, which implies that the higher the education level of a father is, the higher the education level of his children will be. The effect of the mother’s education is also positive and significant. These results are consistent with the literature that suggests that genetic predisposition matters or that better-educated parents tend to offer better study environment for their children. It is worth pointing out that the estimated coefficient for the father’s education is actually greater than that of the mother’s, implying that its effect on educational achievement is greater from the father than from the mother.16
. Estim resul
After trollin or va us famil kground v bles a person l charac policy e ulsory ducat on a posit nd sig ficant fect on a
15 See, for example, Haveman and Wolfe (1995) and Card (1999) for detailed discussion and references therein.
16 We also consider parents’ educational attainment instead of years of education to measure the effect of the parents’
education; the results are similar to what we report here. Using Taiwan’s data, Liu, Hammitt and Lin (2000) also find that the father’s education has a greater effect than the mother’s education on their children’s return to education.
Aakvik, Salvanes, and Vaage (2005) also find robust result showing that father’s education has a greater effect on children education than mother’s in Norway. Card (1999, Table 2) show that each additional year of schooling of either parent raises completed schooling by about 0.2 years, and roughly 30% of the observed variation in schooling among US adults is explained by parental education.
The father’s occupation also affects his children’s’ education. A father who works as an adm
achievement.17 Living in a single family also has a negative effect but o low income family and tends to encounter financial diffi
arents education and occupations.19
inistrator has the largest effect. Children whose father’s occupation was administrators or executives receive an additional 2.12 years of education over children whose father worked in the agricultural sector and additional 1.63 years for children whose father worked as a sales person.
Children who father worked in public sector also receive additional 0.6 years of education mainly because of education subsidies provided by government.
However, if mother is also working, the effect is negative and significant implying the sacrifice of mother’s home caring and nurturing due to time constraint has rendered an adverse effect on children’s educational
significant. Single family usually belongs t
culties, which has a unfavorable effect on human capital investment.18 The possible reasons for the insignificance here may be due to the proportion of single family is relatively small (about 7.84% of our sample) and the effect may be abstracted after controlling for family socioeconomic status such as p
As for family structure, the number of siblings has a significant negative effect, implying that holding family resources constant, the larger the number of siblings, the lower the educational achievement, because family resources spent on additional children decline as the number of children increase.20 The coefficient of birth order is positive and significant, while that of its square is also positive but insignificant. It implies that holding the number of siblings constant, the latter the birth order, the better the educational achievement. The may be due to the latter born children usually receive more family resources as the family income becomes more stable at the
17 See, for example, Chin and Newman (2002) and McLanahan and Sandefur (1994) for the discussion of working mother’s negative effects
18
on children’s educational achievement.
See, for example, Elliott and Rechirds 91991), Mclanahan and Sandefur (1994), Haveman and Wolfe (1995), and Ermisch and Francesconi (2001).
19 Gregg and Machin (1998) also find that after controlling for family financial constraints the effect of single family variable becomes insignificant.
20 For the sibling effect we also consider both the order and number of siblings, the estimated results show that children whose have older sisters tend to receive more education and those who have younger sisters or brothers tend to have less education. The main reason is that in Taiwan’s early economy due to family budget constraints, older sisters, especially first born girls, usually sacrificed their education or marriage to go to work early and married later to support the family and save resources for the education of their younger sisters or brothers. Parish and Wills (1993) also find
later
in rural area are the least educated group.
It is
children born at different times may have diffe
positive and significant. Attending talent or skill training shows a gain of 1.21 years on plying that talent training will enhance academic ability. Having pplementary education will add an additional 1.44 years of education, stage of their parents’ career.
After controlling for family background, the estimated coefficients of dummy variables gender, region, and ethnicity are all positive and significant. Significant educational divergence exists between gender, region, and ethnic groups. On average, women had 1.30 years less than men;
rural areas had 0.53 years less than urban areas; and Taiwanese had 0.61 year less than mainland Chinese. This implies that Taiwanese females who live
obvious that children lived in rural areas receive less education because of the lack of educational resources. Due to their experience of fleeing from the civil war in China, mainland Chinese values more on human capital, which is moveable, than physical capital; whereas Taiwan is a traditional agriculture-oriented economy where education is not so important for a family and thus Taiwanese tends to underestimate in education.21 Affected by deep-rooted traditional family thinking that prefer male than female, hence females tend to be discriminated and underinvested in their education.
As the larger environment changes over time,
rent educational achievements. The cohort dummies, proxies for the change of macro environment over time are positive, i.e., the more recently born the student, the higher their educational level. Compared to the older population, middle-aged people had an additional 2.28 years of education while young people had nearly 3.56 years more.
As some of family background variables may also contain resources effect, such as family socioeconomic status (parents’ education and occupations), column 2 in Table 5 further includes variables family educational resources to single out the likely resources effect. The estimated coefficients for after-school talent or skill training and remedial or supplementary education are
educational achievement, im after-school remedial or su
implying that supplementary after-school education is very useful under the examination-oriented
21 Besides the army of the Nationalist retreated from China to Taiwan in 1949, majority of mainland Chinese immigrant
educational system in Taiwan.22 Attending talent training and/or supplementary education may also
ally be harmful for individual educational achievement. Theoretically, traditional fami
esults seem to support the latter espe
signal that those children are more likely from wealthy families which tend to offer better resources and environments for academic development.
Personal character and motivation important factor driving individual’s educational achievement. According to human capital theory, other thing being equal, more able people should receive more education. Therefore, the lack of controlling for personal characteristics may bias the estimated coefficients of family background upwards. Column 3 in Table 5 further includes variables of personal characteristics. The results show that having a part-time job or receiving scholarships or awards during the period of study had a positive and significant effect on educational achievement. This result, as we expected, implies that people of more independent character or more capable people tend to have higher levels of education. Variables of filial duty and glorifying one’s family have a significant negative effect, implying that traditional values of family may actu
ly values may be a driving force to push descendants to work hard and pursue a higher educational level, as having a high level of education is the most important channel to win one fame and wealth in traditional Chinese society. However, obedience to traditional family values also implies a dependent character willing to sacrifice personal individuality and self-interest, which may jeopardize an individual’s pursuit of higher education. The r
cially holding family’s socioeconomic status constant.
As the response for educational investment by gender may be different, we further divide the sample in to male and female subsamples and run the regressions respectively. The results are shown in Columns 4 and 5 of Table 5. The gender effects of family background, family educational resources, and personal characteristics on educational achievement are all significant and share same sign for the coefficients but they differ in magnitude. On personal characteristics, female has a greater effect of gaining academic performance and having part-time job on
22 The effectiveness of cram schools (Busiban) for supplementary education is apparent in Taiwan as the number of
educational achievement.23 Family educational resources also have stronger effect for female.
As traditional values favor educational investment in male, shifting additional resources to female is likely to have greater effect as the marginal return is higher for female than for male. As for fami
round, family educational resources, and perso
Table 5. Regression results for educational achievement
ly background, parents’ education have significant positive effect on both son and daughter.
However, father’s education has a larger effect on son, while mother’s education has a larger effect on daughter. These results are consistent with the theory of observational learning that parents can be role models for their children.24 Furthermore, father’s occupation has a stronger effect on daughter than on son, whereas, father worked in public sector has a stronger effect on son than on daughter. Working mother has a significant negative effect on both son and daughter’s educational achievement; however, the negative effect is greater on daughter than on son.
As for siblings effect, the number of siblings has a negative effect for both male and female, however, the effect is greater for female than for male. The birth order effect is positive but significant for male and insignificant for female. The latter born son tends to receive more education. Other thing equal, educational gap between ethnic groups is greater for male than for female, while educational gap between rural and urban areas is greater for female than for male.
The results from Table 5 suggest that family background has an important effect on individual’s educational achievement; however, without controlling for family educational resources and personal characteristics, the estimated effects of family background are likely bias upwards. The overall explanatory power for family backg
nal characteristics are 82.1%, 6.99%, and 10.81%, respectively.
Dependent Whole Whole Whole Personal
Characteristics
Variables sample sample sample Male Female Part-time job 1.199*** 0.921*** 1.370***
23
performance.
24 See, for example, Bandura (1977). Using household survey data from Mozambique, Heltberg and Johammesen (2002) also find similar result that mothers’ schooling has a stronger effect on girls’ education and fathers’ schooling has a larger impact on the education of boys.
Dynarski (2005), Angrist and Levy (2002), and Angrist, Lang, and Oreopoulos (2006) all find that female has a greater response and thus a significant effect to academic scholarship or educational subsidies schemes on school
(0.115) (0.160) (0.165)
(0.240) (0.227) (0.310) (0.328)
(0.017) (0.017) (0.016) (0.023) (0.021)
5***
(0.020) (0.019) (0.018) (0.026) (0.025)
Fatocc prof
and executives
(0.282) (0.273) (0.259) (0.372) (0.356)
Service workers 0.579*** 0.542** 0.479** -0.170 1.026***
ope lab
Single family -0.154 -0.123 -0.104 -0.275 -0.089
Number of *** *** *** **
-0.119***
(0.041)
Birth order 0.217 0.142 0.163 0.182 0.121
(0.083) (0.080) (0.076) (0.08 (0.111)
7.254 6.870 7.754 8.017 6.135
5.79 209.75 70.03 145.87
(0.162) (0.155) (0.238) (0.202)
Supplementary
education 1.439*** 1.300*** 1.271*** 1.332***
(0.132) (0.126) (0.169) (0.185)
Rewards for
academic 1 236*** 1.203*** 0.842*** 1.355***
(0.217) (0.206) (0.306) (0.276)
Family transfer -0.111 -0.116 -0.035 -0.132
(0.304) (0.295) (0.281) (0.400) (0.389)
Administrators 2.117*** 1.873*** 1.864*** 1.732*** 2.058***
(0.271) (0.263) (0.250) (0.346) (0.356)
Clerks 0.955*** 0.802*** 0.936*** 0.664* 1.306***
Sales workers 1.630*** 1.416*** 1.446*** 1.239*** 1.628***
(0.182) (0.177) (0.168) (0.242) (0.230)
(0.276) (0.268) (0.254) (0.370) (0.344)
Production rators and
orers 0.847*** 0.772*** 0.705*** 0.506** 0.886***
(0.155) (0.150) (0.142) (0.203) (0.196)
Father in public
sector 0.596*** 0.566*** 0.485*** 0.920*** 0.079
(0.196) (0.190) (0.180) (0.255) (0.252)
Mother at work -0.258*** -0.242** -0.225** -0.171** -0.279***
(0.106) (0.105) (0.105) (0.085) (0.144) (0.184) (0.155) (0.168) (0.242) (0.263)
Siblings structure
siblings -0.136 -0.109 -0.110 -0.094
(0.034) (0.033) (0.032) (0.049)
*** ** ** **
-0.007*** -0.001 -0.000 -0.006 -0.009
(0.000) (0.000) (0.000) (0.011) (0.013)
Gender
(0.137) (0.133) (0.127) (0.188) (0.170)
Young 3.558*** 3.441*** 2.828*** 2.184*** 3.486***
(0.158) (0.154) (0.151) (0.215) (0.210)
Constant *** *** *** *** ***
(0.330) (0.321) (0.363) (0.523) (0.491)
Observations 3585 3575 3570 1725 1845 Adj-F2 0.5277 0.5584 0.6033 0.5003 0.6626 F value 236.54 21
E
Personal factors 10.81% 8.53% 11.71%
Notes: Figures in the parenthe s are standard viation. * *** sta atistical nt level at 10%, 5%, respectively.
ysis
to test fo the robu of est ent, a
ysis is fu ther cond We f nsider on-line ffect of parents’
g a term for parents’ ed . Co of Ta
e emains a positive and s nt effe its square terms are e but significant fo d insignificant for mothe This imp fect of father’s education is not constant but diminishing. Using parents’ educ l level d of y f schooling, the results show that then higher the parents’ edu l level great t ct on children’s
ievem
’ effect, in order to further exam e sibli ucture ender preference, w es variables as well as the numbe der br ounge r, old sister, and
o e re ws th g old has a cant p effect. Adding
x r will edu achie by 0. s. This is consistent with the ills (1993) that at early stage iwan’s ic development under limited dget constr ly bo ales u sacrif r edu nd marriage by orking early and pos g marr o earn to sup their fa nd their younger r sisters’ education. Having old brot a positive but insignificant effect, implying th nce for ma cially n) tha le in traditional Chinese society.
er to actual lect the econom ’s occupations
fr -digit seven ation -digit -six c tions. esults are shown
r stness imation results for educational achievem s nsitivity anal r ucted. irst co the n arity e
education by addin square ucation lumn 1 ble 6 shows that parents’
ducation r ignifica ct and negativ
r father an r. lies that the ef
ationa instea ears o cationa , the he effe educational ach ent.25
For siblings ine th ngs str and g
e use dummi r for or other, y r brothe
y unger sister, th sult sho at havin er sister signifi ositive e tra older siste increase cational vement 12 year
find of Parish and W of Ta econom
family bu aint e ra rn fem sually ice thei cation a
w tponin iage t money port mily a
brothers o her has
e prefere le (espe elder so n fema
In ord ly ref socio ic status we further expand father om one classific s to two seventy lassifica The r
25 om the P bour Fo ey (PL 1992 , Beblo er (2002) also find .
data fr
Using olish La rce Surv FS) from to 2000 and Lau similar result for Poland