David Newhouse Daniel Suryadarma
Outline of presentation 1. Motivation Vocational education expansion 2. Data 3. Determinants of choice of type 4. Effects of high school type Entire sample Cohort vs. age effects Father s education, test score
Vocational education is declining in Indonesia Percent 40 35 30 25 20 15 10 5 0 Share of vocational high schools (SMK) students to total senior secondary Predicted 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014
New policy to reverse decline Current ratio of SMK to SMA students is 30:70 Goal is to increase ratio to 50:50 by 2010, and 70:30 by 2015 Implies moving 4.1 million students into SMK Millions 7 6 5 4 3 2 SMA & SMK Enrollment, 2015 Motivated by desire to reduce unemployment. 1 0 Weak empirical support SMA SMK SMA SMK Predicted Targeted
Planned VTE expansion will raise public costs Source: Ghazali (2006) and IFLS3
Existing research is mixed Vocational graduates earn premiums in: Egypt (El-Hamidi, 2006) Isreal (Neuman and Ziderman) Thailand (Moenjak and Worswick, 2003) Penalties in: Suriname (Horowitz and Schenzler, 1999) Tanzania for university grads (Kahyara and Teal, 2008)
Existing research is mixed No discernible differences in: South Korea (KRIVET, 2008) Singapore (Sakellariou, 2003) Romania (Malamud & Pop-Eleches, 2008) Variety of evidence on gender differences Males: SMK earn higher in Egypt but lower in Singapore. Females: SMK earn higher in Singapore but lower in Egypt.
Past research in Indonesia can be improved VTE results in neither advantage nor disadvantage with respect to earnings or employment (Chen, 2009). But Wide confidence intervals: OLS: 0 to 60% of mean earnings IV: -50 to 150% of mean earnings Sample restricted to adults under 25 30 percent of sample still in college. Selection correction excludes head education, test scores, prior household income from earnings regression Few control variables (gender, rural, age)
Our contributions Separate men and women Distinguish between public and private schools Several cohorts (born between 1940 and 1980) Rich set of predetermined controls Parental education (even if non-coresident) District of Jr. High school dummies Four main LM indicators Participation, unemployment, formal work, wage
Important to separate by gender Because of differences in both participation and vocational education majors Women tend to study business or tourism rather than technical or industrial subjects.
Main messages: Public VTE Overall, no evidence of wage effect. Confidence intervals: -10 to 12 percent for men -6 to 23 percent for women Worrisome recent fall in returns for most recent cohort of men -29 percent at age 24 (vs. 39 percent for middle cohort) -43 percent at age 31 (vs. -8 for older cohort)
Main results: Other types Large public school premium 20 percent wage premium for men Private general high schools are underperforming Outcomes similar to private VTE, despite observably more privileged students Why do men attend private general schools?
Outline 1. Data and sample 2. Determinants of School type 3. Labor Market Effects of school type Aggregate Effects Cohort and Age Effects Effects by father s education and student test score 4. Conclusion
Data Construction Four rounds of Indonesian Family Life Survey 1993, 1997, 2000, 2007 7,300 households in 1993, 12,600 by 2007 13 provinces representing 83% of Indonesia in 1993 Retrospective data on school attendance Type of school collected at all school levels Includes test scores for those born 1983 or later. We limit to high school graduates that are out of school
Data construction Men Obs Women Obs High school grad and out of school 2891 6449 2430 5662 Reported school information and satisfy common support condition 2675 6084 2260 5330 In labor force 2621 5934 1753 3456 Employed 2460 5439 1516 2875 Reported earnings 2352 5066 1427 2681 Reported school information 2675 6084 2260 5330 Old cohort (born 1940-1963) 923 2594 574 1730 Middle cohort (1964-1972) 935 2034 819 1929 Recent cohort (1973-1980) 866 1456 934 1671 Of which reported test score 737 1245 766 1366
1. Choice of school type T i = β Z + β P + β P + ε z i Pi i T id = School Type Public general, public VTE, private general, private VTE Z i = Predetermined characteristics Cohort, sex, age 12 residence (city, town, village), adult height, public jr. high, worked in jr. high, worked in elem, failed grade in jr. high, failed grade in elem, province of junior high) P i = Parental education (both parents) Elementary, jr. high, high school, univ dummies Vocational dummy pd P d = District parental education shares d id
Vocational school attract children of least educated parents Pub gen VTE Men Pri gen VTE Pub gen VTE Women Father elem 4-1 4-7* -7-5 6 5 Jr. high 6-4 7-10** -6-9* 8 6 High school 6-7 13** -12** -2-9 11-1 University 19** -12*** 12* -19*** 6-11* 7-2 Mother elem 0-5* 6* -1 7 7* -5-9*** Jr. high 3-5 7-6 13** -2 2-13*** High school 5-9* 2 2 10 5-2 -13** University 19* -5-5 -9 17* 7-11 -13** Base case prob 13 30 18 38 51 20 18 12 Pri gen VTE
School type determinants: Young cohort For youngest cohort, test scores are available. T i = β Z + β P + β P + β S + ε z i Pi i S i = Score tercile (on jr. high exit exam) pd d s i id
Public schools attract high scorers Public schools attract highest scoring students Public general slightly higher scoring than public VTE Private VTE attracts lowest-scoring students Pub gen VTE Men Pri gen VTE Pub gen VTE Women Pri gen Test middle tercile 14*** 8* -3-19*** 5 9** -3-11*** Test upper tercile VTE 24*** 16*** -17*** -23*** 20*** 13*** -10** -23*** Base case prob 13 31 18 38 51 20 18 12 Observations 737 771 Pseudo-r2 0.16 0.22
Selection: Summary Rankings Ranking Parental education Test scores Top Private general Public general Second Public general Public VTE Third Public VTE Private general Bottom Private VTE Private VTE
Estimating labor market outcomes Y it = β Z + β P + β D + β D + β T + ε z i Pi i = person, t = year, d = district i d T i = dummies for pub VTE, pri gen, pri VTE. Z i = Predetermined characteristics P i = Parental education shares D d = District of jr. high dummies D t = Year dummies Y i = Four outcomes: Participation, Unemployment (conditional on participation) Formal job, Log wage/profit (conditional on working) d t t s i id
Double Robust Regression Rebalances sample Weight observations by inverse probability that the type of school attended was chosen. Reduces precision but increases robustness to nonlinear functional forms.
Rebalancing slightly helps Measure of balance: standardized difference of observable controls across school type X X X T = T 2 T S + S Mean Standardized difference PubGen 2 PubGen Unbalanced Well below rule of thumb threshold of 0.25 = Mean for type T = Variance for type T Balanced Public VTE -0.016 0.006 Private general -0.022 0.005 Private VTE -0.027 0.001 X t 2 S T
Definition of formal job Simplified national definition Status Industry Non-Agriculture Agriculture Family worker Informal Informal Self-employed alone Informal Informal Self-employed with temporary workers Formal Informal Employers or Employees Formal Formal Defined for all workers (incl. 7% unpaid family) Formal workers: appear to earn higher wages (but very hard to tell) Get more benefits and are slightly more satisfied with job
Private general lowers LFP for women Men Women LFP Unemp LFP Unemp Public VTE 0.013* -0.006 0.023-0.017 (0.007) (0.011) (0.029) (0.012) Private 0.013* -0.003-0.076** 0.016 general (0.007) (0.008) (0.032) (0.010) Private VTE 0.005 0.010-0.032 0.004 (0.008) (0.012) (0.034) (0.013) Pub gen prob 0.971 0.051 0.693 0.045 r 2 0.090 0.171 0.175 0.232 Observations 6,084 5,931 5,330 3,452
Male public school wage premium Men Women Formal Wage Formal Wage Public VTE 0.036** 0.009 0.032 0.087 (0.017) (0.056) (0.025) (0.075) Private -0.042* -0.171*** -0.052* -0.047 general (0.025) (0.062) (0.028) (0.076) Private VTE 0.019-0.203*** 0.007-0.014 (0.020) (0.064) (0.029) (0.081) Pub gen prob 0.575 0.566 r 2 0.559 0.230 0.584 0.314 Observations 5,642 5,065 3,288 2,681
Public wage premium robust Positive public VTE wage premium for median woman Men Women OLS LAD OLS LAD Public VTE 0.009 0.042 0.087 0.133*** (0.056) (0.032) (0.075) (0.049) Private -0.171*** -0.254*** -0.047-0.202*** general (0.062) (0.032) (0.076) (0.064) Private VTE -0.203*** -0.162*** -0.014-0.048 (0.064) (0.034) (0.081) (0.058) Observations 5,065 2,681 2,681
Robust to including test scores (on young cohort of men) Men, Young Cohort Formal Wage Test Scores included? Yes No Yes No Public VTE 0.039 0.039-0.328*** -0.322*** (0.040) (0.040) (0.101) (0.101) Private 0.023 0.023-0.180-0.205* general (0.047) (0.046) (0.114) (0.114) Private VTE 0.067 0.068-0.134-0.153 (0.051) (0.045) (0.106) (0.104) Observations 979 979 803 803
Robust to including test scores (on young cohort of women, conditional on school type) Women, Young Cohort Formal Wage Test Scores included? Yes No Yes No Public VTE -0.042-0.037-0.175-0.122 (0.053) (0.053) (0.142) (0.159) Private -0.122** -0.109** -0.063-0.049 general (0.054) (0.052) (0.137) (0.149) Private VTE -0.098** -0.077** -0.222-0.145 (0.042) (0.039) (0.164) (0.157) Observations 752 752 578 578
Age and cohort effects Interact type with year, separately by cohort Y i = Three Cohorts β Z + β P + β P + β T * D + ε z i Pi i Old (1940-1963); Middle (1963-1972); Recent (1973-1980) Key term is interactions of school type and year dummies All four types included, main effects omitted Graph estimated effect vs. mean age for each cohort and year For public VTE only pd d T i t id
Age and cohort effects: LFP Men Effect of public VTE on LFP Men Old cohort Middle cohort Young cohort Women Effect of public VTE on LFP Women Old cohort Middle cohort Young cohort 0.07 0.20 0.06 0.05 0.04 0.15 0.10 Estimated effect on LFP 0.03 0.02 0.01 Estimated effect on LFP 0.05 0.00 0.00-0.01-0.02-0.05-0.10-0.03 20 30 40 50 60-0.15 20 30 40 50 60 Average age in years Average age in years
Age and cohort effects: Unemp Estimated effect of public VTE on unemp 0.06 0.04 0.02 0.00-0.02-0.04-0.06-0.08-0.10 Men Effect of public VTE on Unemployment Men Old cohort Middle cohort Young cohort 20 30 40 50 60 Average age in year Estimated effect of public VTE on unemp 0.06 0.04 0.02 0.00-0.02-0.04-0.06-0.08 Women Effect of public VTE on Unemployment Women Old cohort Middle cohort Young cohort 20 30 40 50 60 Average age in year
Age and cohort effects: Formality Men Effect of public VTE on formality Men Old cohort Middle cohort Young cohort Women Effect of public VTE on formality Women Old cohort Middle cohort Young cohort 0.25 0.30 Estimated effect of public VTE on formality 0.20 0.15 0.10 0.05 0.00-0.05 Estimated effect of public VTE on formality 0.20 0.10 0.00-0.10-0.20-0.30-0.40-0.10 20 30 40 50 60-0.50 20 30 40 50 60 Average age in year Average age in year
Age and cohort effects: Wages Men Effect of public VTE on wages Men Old cohort Middle cohort Young cohort Women Effect of public VTE on wages Women Old cohort Middle cohort Young cohort 0.60 0.60 0.50 Estimated effect of public VTE on formality 0.40 0.20 0.00-0.20-0.40 Estimated effect of public VTE on formality 0.40 0.30 0.20 0.10 0.00-0.10-0.20-0.30-0.60 20 30 40 50 60 Average age in years -0.40-0.50 20 30 40 50 60 Average age in years
Age and Cohort Effects Male Wages Public VTE Wage Cohort Old Middle Young 2007 0.045 0.080-0.431** (0.178) (0.161) (0.189) 2000-0.004-0.084-0.291** (0.106) (0.128) (0.132) 1997 0.232* 0.126 0.110 (0.137) (0.120) (0.262) 1993 0.160 0.394* (0.131) (0.234)
Age and Cohort Effects Female Wages Public VTE Wage Cohort Old Middle Young 2007 0.290-0.105 0.516 (0.192) (0.280) (0.396) 2000-0.170 0.026-0.078 (0.226) (0.182) (0.144) 1997-0.024-0.391* -0.252 (0.179) (0.214) (0.183) 1993 0.130 0.102 (0.170) (0.225)
Recent fall in premiums partly robust to median regression Men Effect of public VTE on wages Old cohort Middle cohort Young cohort Women Effect of public VTE on wages Old cohort Middle cohort Young cohort 0.40 Estimated effect of public VTE on log wage 0.40 0.30 0.20 0.10 0.00-0.10-0.20-0.30-0.40-0.50-0.60 20 30 40 50 60 Average age in years Estimated effect of public VTE on log wage 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00-0.05-0.10-0.15 20 30 40 50 60 Average age in years
Outcomes by father s education Men Formal Wage Father s education <Sr high >=Sr High <Sr high >=Sr High Public VTE 0.060*** 0.036-0.001-0.34 (0.021) (0.037) (0.062) (0.142) Private -0.047* 0.015-0.223*** -0.129 general (0.027) (0.054) (0.061) (0.164) Private VTE 0.012 0.009-0.287*** -0.167 (0.024) (0.064) (0.076) (0.137) Observations 4,106 917 3,698 799
Outcomes by father s education Women Formal Wage Father s education <Sr high >=Sr High <Sr high >=Sr High Public VTE 0.022 0.005 0.125-0.053 (0.034) (0.048) (0.102) (0.146) Private -0.089** -0.037-0.200 0.174 general (0.037) (0.045) (0.146) (0.140) Private VTE -0.027-0.002-0.019 0.170 (0.041) (0.047) (0.099) (0.165) Observations 2142 875 1,713 739
Summary of outcomes by father s education Effect of school type strongest for children of lesseducated parents For these men, private school has strong negative effect on male wage For these women, private general is associated with reduced formality.
Outcomes by test score Men Formal Wage Test score Low High Low High Public VTE 0.089 0.026-0.235-0.409*** (0.098) (0.066) (0.191) (0.157) Private 0.010-0.018-0.163 0.330** general (0.082) (0.093) (0.150) (0.153) Private VTE 0.062 0.067-0.127 0.492*** (0.097) (0.099) (0.150) (0.187) Observations 570 484 581 477
Outcomes by test score Women Formal Wage Test score Low High Low High Public VTE 0.017-0.040 0.213-0.155 (0.104) (0.086) (0.321) (0.183) Private -0.132* -0.086 0.154 0.093 general (0.069) (0.084) (0.259) (0.232) Private VTE -0.193** -0.023 0.167-0.328* (0.079) (0.074) (0.160) (0.193) Observations 394 430 287 350
Outcomes by father s education Effect of school type strongest for children of lesseducated parents For these men, private school has strong negative effect on male wage For these women, private general is associated with reduced formality.
Summary Selection Rank on test score: Public general, Public VTE, private general, private VTE Rank by parent education: Private general, public general, Public VTE, private VTE
Summary Outcomes At most, small benefit from Public VTE Boosts formality by 3.6 pp for men Large (20%) wage penalty for privately schooled men Suggests that test scores and peer effects are important determinants of future earnings for men
Summary Private general graduates are underperforming Perform as poorly or worse than private VTE grads despite more educated parents. VTE penalty increased dramatically for recent male graduates Could be related to shift towards service sector Annual growth by sector Value Added Employment 90-97 99-07 90-97 03-07 Industry 9.0% 4.3% 6.9% 2.5% Services 7.0% 6.3% 6.0% 4.0%
Policy implications Expansion should proceed with caution, given recent fall in male graduates earnings Ensure students with highest test scores can attend public school Re-examine industrial public VTE curriculum Private VTE appears to be more effective then private general at preparing the weakest students
The private general school puzzle Why is there a private school wage penalty for men but not women? Patterns of participation? Why do well-educated parents send their children to private general schools? Poorly functioning market for information on quality of private general schools? Non-academic factors relatively important in determining preferences?