Upload
kalea
View
42
Download
0
Embed Size (px)
DESCRIPTION
Mother's Labor Force Participation in Early Childhood and the Child's Educational Attainment. Miki Kohara and SunYoun Lee Osaka University. 1. Purpose. - PowerPoint PPT Presentation
Citation preview
1. PurposeExamine how maternal employment at the child’s age of three affects his/her educational outcome at age of eighteen, using Korean panel data.
2. MotivationThe long-run impact of family environment (maternal employment) on child’s development (educational outcome)?
Two opposite directed effects : maternal employment may increase money input but may decrease time input on child’s education => a total effect of maternal employment on child’s educational outcome is not decisive theoretically
Mother's Labor Force Participation in Early Childhood and the Child's Educational Attainment
Miki Kohara and SunYoun LeeOsaka University
1 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Maternal employment at age three affects test scores at age eighteen, but the effect is opposite between high (minus effect) and low educational levels (plus effect)
3. Main Result
Introduction
4. Originalities and Contributions- Consider endogeneity of maternal employment and heterogeneity of
its effect- Show a long-term effect of maternal employment on the child’s
development- Examine the Korean case
This result is obtained, after (1) controlling for mother’s educational attainments, father’s occupations, and economic conditions, (2) allowing for the existence of unobserved heterogeneity in a child’s educational outcomes and the mother’s employment, and (3) allowing for non-linearity (heterogeneity) in the effects.
2 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Introduction 2
5. Why Korea? - Important policy issue: there is a social concern about how parent’s
(mother’s) behavior affects educational outcomes. > Education obsession > Extremely low married female labor supply - Good for the analysis: we can use test score data as educational
outcomes “CSAT” College Scholastic Ability Test; once a year at the same time all over the country >Objective measure of educational outcomes >Showing a degree of educational attainments >Measuring outcomes at the end of high-school *Many students take this exam: 82% of the students in the last year of high school (04)
3 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Labor Force Participation of Women (LPR)
2001 1981
4 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
LPR of married women
LPR for the 25-29 age group were 7-27% points lower compared to those of other child bearing age groups of 20-24 and the 30-49
LPF for the age group 15-19 appears to have continuously declines since 1970.
5 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Maternal employment deteriorates the child’s educational outcomes?-Ambiguous results. Not always.
The impact of maternal employment on child’s development
Baker, Gruber and Milligan (2008)
Canada ―
Ruhm (2008) USA ― for high economic condition group, butNo effect for lower groups
Bernal (2008) USA ― if a mother started working as a full time worker within a year after giving a birthNo effect of temporary income change
Dustman and Schonberg (2008)
Germany ― Policy change of prolonged maternal leave raises child’s educational outcomes
Tanaka and Yamamoto (2009)
Japan No effect of maternal employment at the child’s age 0-3* But maternal employment after that can lower the probability of going to (probably highly ranked) private or national junior high schools.
6Literature Review 1
6 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Literature on causation behind the effect
(1 )Maternal employment ⇒ more income & more investment on education ⇒ child’s educational attainments ↑
… Ambiguous results
(2) Maternal employment ⇒ less time with children ⇒ child’s educational attainments ↓… Ambiguous results
NOTE: Maternal employment - full time worker / high-skilled labor = mother’s higher education level
⇒ child’s educational attainments↑ - part-time worker / low-skilled labor = mother’s lower education level ⇒ child’s educational attainments↓☝We need to control for mother’s educational attainment.
Maternal employment - weak preference for child’s care child’s educational attainments↓⇒ - strong preference for child’s care child’s educational ⇒attainments ↑☝We need to remove a bias raised by unobserved heterogeneity.
7Literature Review 2
7 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
-Estimations with categorical data or quantile regression can be insufficient.
Two difficulties in the estimation
Mother’s employment at age 3
Educational Outcomes at 18 years old
Environmental effect or peer effect of female labor supply at that time, but not to affect child’s educational outcome later at age of 18
Mother’s unobserved ability+/- Over/Under estimateMother’s preference for child’s education-/+ Under/Over estimate(1) Problem of Endogeneity
-Linear estimation with endogenous variables can be biased.
(2) Heterogeneous Effects
Empirical Framework 1
82013/3/3
Female Labor ForceParticipation Rates
Estimation for different thresholds
Ti is a dummy variable indicating 1 if a child’s test score is in a “higher” group Mi is a dummy variable indicating 1 if his or her mother was working
Empirical Framework 2
Data : Korean Labor & Income Panel Study ( KLIPS)
Method: Probit Model with Endogenous Treatment
Non-linear simultaneous decisions
Note.Heterogeneous effect of maternal employment on child’s test score: the effect could be different among the levels of test-scores. Conduct the estimation, changing the threshold!
9 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
,1,1,0,0~,|,)0(1
)0(1
2121
*
*
222*
111*
NxxMM
TT
xM
xMT
Empirical Framework 3 Non-linear effect of maternal employment on child’s test score
Idea! Rank 12
Rank 1
Rank 9
Rank 7=1 (if test score rank>=7)
=0 (if test score rank<7)=1 (if test score==7) otherwise 0 (if test score≠7)
Rank 10
Rank 8
Rank 11
10 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
,1,1,0,0~,|,)0(1
)0(1
2121
*
*
222*
111*
NxxMM
TT
xM
xMT
Korean Labor and Income Panel Study;KLIPS(1998-2008), conducted by the Korea Labor Institute, a government-sponsored research organization. Household Survey : Householder or spouse ( nationally representative sample of
5,000 households ) Individual Survey : Each person in the household aged 15 and over ( approx.12,000
persons) Additional Survey ( 2001 ~ 2008 ): yearly special modules for restricted sample,
such as the youth, the old or the employed. ( approx.4500 persons ) Survey method : Household Interview
Data used for this study Additional Survey (2006): contains detailed information on education-related
records of the youth aged 15 to 35 at the time of survey(4,389 persons) Maternal employment
Individual and Household Surveys (1998-2006): individual characteristics Individual (2002): Test score for the College Scholastic Ability Test (CSAT) Individual (1998-2006): Parental and child’s educational attainment, birth year and place to
grow up, socio-economic status at age 14, father’s occupation at age 14 Household (1998-2006): demographic information: family composition, marital status
11
Data 1
11 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Important variables: (1) Educational Outcomes at age 18 Test score for the university entrance
exam (CSAT) Historical changes in university entrance examination test1. 1981-1993: Achievement Test started (Highest = 340) 2. 1994-1996: College Scholastic Ability Test 1 started (Highest = 200)3. 1997- : CSAT 2 started (Highest = 400) 4. 2002- : Diversification of entrance exam. : University can use other
types of exams. Our sample took either (1), (2) or (3). Test scores are answered as 12 ranks We need to control for the difficulties in each
year.
12
0.1
.2.3
.4D
ensi
ty
1 2 3 4 5 6 7 8 9 10 11 12 13Testscore at age 18
Data 2
Advantages of this indicator1) Achievement test and college scholastic ability test were the most important single
determinant for the university admissions in Korea.2) Most of Korean high school students with the same academic attainment take the same
test on the same day (advancement rate to university: 82%, test takers were 572,218 and third grade of high school students 582,216 in 2004)
3) Because the answers are formatted in score ranges with the interval of 10 to 40 scores, potential measurement errors may be small.
12 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
KLIPS answer sheet
13
Data 3
13 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Name of Variable Mean S.D.Educational Outcomes
Test score attained at the university entrance examination 7.856 2.420
Family backgroundMaternal Employment at child's age 3 (=1)
0.390 0.488
Female labor force participation rage at child's age 3 38.062 7.564
Mother's employment rateof the child's birth place
0.378 0.108
Mother's educational attainment 3.028 1.100
Father's educational attainment 3.739 1.330
Living with parents at age 14 (=1) 0.025 0.157
0.049 0.2170.230 0.4210.568 0.4960.154 0.361
Household economic statusat age 14
Name of Variable Mean S.D.Child's characteristicsFirst child (=1) 0.401 0.491Number of siblings 1.885 1.093
Residential area at age 140.211 0.4080.442 0.497
Sex 0.370 0.483
Kinds of university entrance exam 0.252 0.4350.276 0.4470.472 0.500
Degree of interaction with parents at age 14 16.788 5.725
Private education before school entry 0.276 0.447
Data 4
14 Miki Kohara and SunYoun Lee; Osaka University
Father’s occupation= 1 if a father works in agricultural industry
15
Data 5
Notes: Average test scores are controlled for in several ways. - dependent variable: “respondent’s own score - average test score” whether this difference is positive or not / more than 30 or less than 30 / …. - controlling for year dummies and/or average test score
LPR of women when our sample was 2 to 5 years old ( % ) Source: National Statistical Office, Annual Report on the Economically Active Population Survey Economically Active Population: The employed and those are who are currently looking for a job LPR: (Economically Active Population/labor force aged 15 and over)*100
1 ) By age range Imputed from i) mother’s age when the child was 3 years old and ii) LPR of women in Korea at the time at
the child’s age 3 => LPR by age range of mothers at child’s age 3 2) By educational level
Imputed from i) mother’s educational attainment ii) LPR of women in Korea at the time at the child’s age 3 3 => LPR by educational attainment of mothers at child’s age 3
Test score ranks, 7 and 10 are focused.
15 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Estimation Results 1 Dependent variable:12 test score ranks
Considering endogeneity but not heterogeneity in the effect
(linear IV)
Eq(1)(2):2SLS 2SLS EstimationEndogenous variableMaternal employment at child's age 3 -0.3438
(2.666)Instrumental variableFemale Labor Force Participation Rate (By age) 0.005 *
(0.003)Father's educational attainment 0.3233* -0.053 **
(0.186) (0.024)Mother's educational attainment 0.2160 0.027
(0.167) (0.033)Household economic status at age 14 (2) 1.2143** 0.058
(0.573) (0.093)Household economic status at age 14 (3) 0.6414 0.095
(0.591) (0.091)Household economic status at age 14 (4) 0.9177 0.097
(0.656) (0.106)Father's occupation at age 14 0.3422 0.392 ***
(1.101) (0.073)Testscore_mean 15.7105*** -0.122
(4.224) (0.860)Observations 483F Test of all explanatory variables 5.14*** 4.39 ***F Test of excluded instruments 2.94 *
First-stage regression
Estimation Results 2 Dependent variable:Probability of being classified to each test score rank (12 ranks)
Considering heterogeneity in the effect but not endogeneity
(ordered probit)
Variable
Maternal employment at child's age 3 -0.0494 Test Score Ranks(0.086)
Father's educational attainment 0.1104** Rank 1 0.0001(0.047) (0.000)
Mother's educational attainment 0.1425** Rank 2 0.001(0.058) (0.001)
Household economic status at age 14 (2) 0.5252*** Rank 3 0.002(0.198) (0.004)
Household economic status at age 14 (3) 0.3052 Rank 4 0.003(0.189) (0.006)
Household economic status at age 14 (4) 0.3343 Rank 5 0.005(0.211) (0.009)
Father's occupation at age 14 0.1011 Rank 6 0.005(0.140) (0.009)
Testscore_mean 6.8340*** Rank 7 0.003(1.672) (0.005)
Observations 636 Rank 8 -0.001LR chi2(14) 98.20*** (0.001)
Rank 9 -0.003(0.005)
Rank 10 -0.005(0.008)
Rank 11 -0.005(0.008)
Rank 12 -0.007(0.011)
Maternal employmentMarginal Effect
Estimation Results 3 Dependent variable:Probability of being classified to each test score rank (3 ranks)
Considering heterogeneity in the effect but not endogeneity (ordered probit)
Variable
Maternal employment at child's age 3 -0.0205 Test Score Ranks(0.096)
Father's educational attainment 0.1269** Rank 1 0.0068(0.052) (0.032)
Mother's educational attainment 0.1141* Rank 2 -0.00013(0.065) (0.001)
Household economic status at age 14 (2) 0.3883* Rank 3 -0.0067(0.221) (0.031)
Household economic status at age 14 (3) 0.2442(0.210)
Household economic status at age 14 (4) 0.2733(0.235)
Father's occupation at age 14 0.0781(0.156)
Testscore_mean 6.6600***(1.866)
Observations 636LR chi2(14) 87.79***
Maternal employmentMarginal Effect
18
Rank 1= test ranks 1-4 2= 7-9 3= 10-12
Estimation Results 4 Dependent variable:12 test score ranks
Considering heterogeneous effects but not endogeneity
(quantile regression)
Variable
Maternal employment at child's age 3 -0.1565(0.293)
Father's educational attainment 0.4139***(0.160)
Mother's educational attainment 0.0065(0.196)
Household economic status at age 14 (2) 0.6729(0.666)
Household economic status at age 14 (3) 0.1862(0.634)
Household economic status at age 14 (4) 0.5193(0.712)
Father's occupation at age 14 0.0263(0.470)-5.6133(4.006)
Observations 636Pseudo R2 0.0702
Quantile Regression (Dependent Variable: Test score ofuniversity entrance exam)
19
These results suggest: We cannot find a robust unambiguous effect of maternal employment: either
positive, negative, insignificant??? The effect of maternal employment may be heterogeneous: positive in lower
groups, and negative in higher groups
Estimation Results 5 Testscore_mean, included
Maternal employment at child's age 3 -1.5935*** 1.5216***(0.091) (0.088)
Female Labor Force Participation Rate (By age) 0.0151*** 0.0172***
(0.005) (0.005)Father's educational attainment 0.0070 -0.1813** 0.2048*** -0.1701**
(0.066) (0.072) (0.065) (0.069)Mother's educational attainment 0.1021 0.0991 -0.0089 0.0994
(0.080) (0.087) (0.083) (0.089)Household economic status at age 14 (2) 0.4126 0.2350 0.0844 0.1212
(0.257) (0.272) (0.240) (0.269)Household economic status at age 14 (3) 0.3485 0.3635 -0.0480 0.2586
(0.245) (0.260) (0.233) (0.262)Household economic status at age 14 (4) 0.3825 0.3171 0.0350 0.1713
(0.274) (0.285) (0.275) (0.303)Living with parents at age 14 (=1) -0.3341 -0.3107 0.2274 -0.1904
(0.298) (0.345) (0.306) (0.403)First child (=1) 0.1986* 0.0910 0.1275 0.0788
(0.116) (0.126) (0.119) (0.131)Number of siblings 0.0047 0.0381 -0.0397 -0.0173
(0.069) (0.078) (0.065) (0.074)Male (=1) 0.0695 0.1447 -0.2727** 0.1976
(0.121) (0.133) (0.121) (0.133)Residential Area at age 14 (6 metropolital cities) -0.2395* -0.1885 0.0352 -0.2911*
(0.145) (0.155) (0.152) (0.171)Residential Area at age 14 (9 provinces and Jeju) -0.1008 -0.0023 -0.0572 -0.0548
(0.151) (0.163) (0.155) (0.175)CSAT_1 -0.0122 -0.0205 -0.0050 0.0140
(0.165) (0.180) (0.168) (0.182)CSAT_2 -0.0140 -0.0011 -0.3358** -0.0121
(0.161) (0.178) (0.165) (0.171)Father's occupation at age 14 0.8035*** 0.8262*** -0.5900***1.2082***
(0.166) (0.185) (0.213) (0.239)Testscore (mean value by year) 5.6941*** -0.2346 2.3708 -1.1752
(2.143) (2.392) (2.024) (2.304)Constants -4.3390***-0.8206 -2.0998 -0.1002
(1.504) (1.677) (1.438) (1.655)Observations 483 483 483 483rho:correlation between errors 13.7974 -11.6281
(9.489) (19.522)
Higher than 10 Higher than 7
Controlling for ave test scores
Probit with endogenous decisions Considering
endogeneity and heterogeneous effects
Dependent variable:
Pr(T* 10)≧Higher threshold
Dependent variable:
Pr(T* 7)≧Lower threshold
Pr(M=1) Pr(M=1)
Maternal employment at child's age 3 -1.5935*** 1.5216***(0.091) (0.088)
Female Labor Force Participation Rate (By age) 0.0151*** 0.0172***(0.005) (0.005)
Father's educational attainment 0.0070 -0.1813** 0.2048*** -0.1701**(0.066) (0.072) (0.065) (0.069)
Mother's educational attainment 0.1021 0.0991 -0.0089 0.0994(0.080) (0.087) (0.083) (0.089)
Household economic status at age 14 (2) 0.4126 0.2350 0.0844 0.1212(0.257) (0.272) (0.240) (0.269)
Household economic status at age 14 (3) 0.3485 0.3635 -0.0480 0.2586(0.245) (0.260) (0.233) (0.262)
Household economic status at age 14 (4) 0.3825 0.3171 0.0350 0.1713(0.274) (0.285) (0.275) (0.303)
Father's occupation at age 14 0.8035*** 0.8262*** -0.5900***1.2082***(0.166) (0.185) (0.213) (0.239)
Testscore (mean value by year) 5.6941*** -0.2346 2.3708 -1.1752(2.143) (2.392) (2.024) (2.304)
Observations 483 483 483 483rho:correlation between errors 13.7974 -11.6281
(9.489) (19.522)
Higher than 10 Higher than 7
Estimation Results 6
21
Probit with endogenous decisions Dependent variable: Pr(T*>=μ) Higher threshold Lower threshold
Controlling for ave test scores
Maternal employment at child's age 3 -1.5967*** 1.5475***(0.092) (0.092)
Female Labor Force Participation Rate (By age) 0.0192*** 0.0172**(0.007) (0.007)
Father's educational attainment 0.0323 -0.1905*** 0.2189*** -0.1618**(0.066) (0.072) (0.067) (0.073)
Mother's educational attainment 0.1080 0.1146 -0.0252 0.1031(0.081) (0.087) (0.081) (0.086)
Household economic status at age 14 (2) 0.4367 0.2908 0.0245 0.1739(0.278) (0.322) (0.257) (0.278)
Household economic status at age 14 (3) 0.3344 0.4089 -0.1037 0.2883(0.267) (0.315) (0.250) (0.275)
Household economic status at age 14 (4) 0.3597 0.3608 0.0257 0.2118(0.295) (0.347) (0.288) (0.317)
Father's occupation at age 14 0.7786*** 0.8733*** -0.5786***1.2090***(0.173) (0.201) (0.214) (0.227)
Testscore (year dummies) Controlled Controlled
Higher than 10 Higher than 7
Estimation Results 7
22
Probit with endogenous decisions Dependent variable: Pr(T*>=μ)
Higher threshold Lower threshold
Controlling for test years
Estimation Results 8
23
Probit with endogenous decisions Dependent variable: Pr(T* -AveT*>=μ)
Higher threshold Lower threshold
Maternal employment at child's age 3 -1.0580** 1.3462***(0.523) (0.077)
Female Labor Force Participation Rate (By age) 0.0173** 0.0101*(0.007) (0.006)
Father's educational attainment 0.0361 -0.1663** 0.1615*** -0.1612**(0.085) (0.073) (0.061) (0.067)
Mother's educational attainment 0.0964 0.0888 -0.0103 0.1015(0.085) (0.087) (0.079) (0.090)
Household economic status at age 14 (2) 0.4418 0.1793 -0.0390 0.2599(0.301) (0.292) (0.249) (0.281)
Household economic status at age 14 (3) 0.3841 0.2866 -0.2075 0.3229(0.290) (0.283) (0.245) (0.280)
Household economic status at age 14 (4) 0.6204** 0.2863 -0.0697 0.3034(0.315) (0.319) (0.277) (0.317)
Father's occupation at age 14 0.3116 1.0131*** -0.7989*** 1.1700***(0.328) (0.243) (0.209) (0.250)
Observations 483 533 483rho:correlation between errors 0.7469 -4.5578
(0.566) (5.395)Marginal EffectMaternal employment at child's age 3 -.4657106 .5297742
Testscore Difference >=30 Testscore Difference >=0
= 1 if a father works in agricultural industry
Summary of the results
Maternal employment at age 3 lowers the probability that a child is ranked in higher than or equal to 10 (whether or not a child can go to so-called good universities)
Maternal employment at age 3 raises the probability that a child is ranked in higher than or equal to 7 (whether or not a child can go to universities or colleges)
• Unobservables are controlled for by a simultaneous estimation mechanism.
• Parent’s education, occupation, and economic conditions are also controlled for.
• We attempted the other splits based on parent’s educational levels, occupation, and economic conditions, but we could not find any difference between the groups.
Estimation Results 9
Maternal employment deteriorates the child’s educational outcomes for students in high ranks. But maternal employment raises the child’s educational outcomes for students in low ranks.
24 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
25
Why opposite effects?
Maternal employment may decrease an interaction time with children, which discourages test score for children in high test-score groups. This deteriorating effect may be offset by the positive effect of money inputs increased by maternal employment in low test-score groups.
Estimation Results 10
Maternal employment Child’s education
High-ranked groups: time (-) >>> money (+)Low-ranked groups: time (-) <<< money (+)
26 Miki Kohara and SunYoun Lee; Osaka University2013/3/3
Maternal employment at age three affects test scores at age eighteen, but the effect is opposite between high (minus effect) and low educational levels (plus effect)
Results and Implications
Conclusion
Explanation other than parent’s education, occupation and economic conditions should be given.One possibility is that maternal employment may decrease an interaction time with children, which discourages test score for children in high test-score groups. This deteriorating effect may be offset by the positive effect of money inputs increased by maternal employment in low test-score groups.
1. Not only discouraging effect but also encouraging effect of maternal employment on child’s educational attainments is found at least for those who were born between 80s & 90s in Korea.2. We need to consider both endogeneity of maternal employment and heterogeneity in its effect when discussing maternal effect on the child’s educational outcomes.
c.f.
Maternal employment at child's age 3 -1.5030*** 1.3649***(0.088) (0.078)
Female Labor Force Participation Rate (By age) 0.0158*** 0.0092(0.005) (0.006)
Father's educational attainment -0.0453 -0.1514** 0.1372** -0.1521**(0.062) (0.073) (0.062) (0.069)
Mother's educational attainment 0.0877 0.1054 -0.0259 0.1221(0.078) (0.085) (0.080) (0.093)
Household economic status at age 14 (2) 0.3768 0.2074 -0.0837 0.2859(0.258) (0.309) (0.257) (0.286)
Household economic status at age 14 (3) 0.3350 0.3789 -0.3003 0.3611(0.249) (0.307) (0.255) (0.286)
Household economic status at age 14 (4) 0.5356* 0.3135 -0.1695 0.3466(0.278) (0.333) (0.285) (0.324)
Degree of interaction with parents at age 14 0.1891 -0.2777** 0.3876*** -0.1956(0.122) (0.135) (0.115) (0.129)
Observations 482 533 482rho:correlation between errors 12.1083 -11.0742
(12.100) (7.871)Marginal EffectMaternal employment at child's age 3 -.7800035 .5375795
Testscore Difference >=30 Testscore Difference >=0