Are Private Schools More Effective than PublicSchools?
Mohamad Fahmi∗
Department of Economics, Universitas Padjadjaran, Indonesia
July 27, 2010
Abstract
I attempt to replicate the research carried out in the paper entitled "The Effectiveness ofPrivate Versus Public Schools: Case of Indonesia" the work of Bedi and Garg (2000). Bediand Garg (2000) find that selectivity bias in the earnings estimation reverses the superiorityof public schools over private schools. To confirm the findings, I use the same sample dataused by Bedi and Garg (2000) and a sample data that created from the Indonesia FamilyLife Survey 1. I also discuss the use of school quality proxies by Bedi and Garg (2000) thatmay bias their estimates of the earnings differential. My findings show that the surprisingfindings of Bedi and Garg (2000) are not robust and suggest that public school graduatesearn significantly higher than private non-religious school graduates and imply that thequality of public schools are better than private non-religious schools.
JEL classification: J31Keywords: School effectiveness; Earnings; Indonesia
1. Introduction
While it is generally accepted that public secondary schools are at the highest quality in In-
donesia (Strauss et al., 2004; Newhouse and Beegle, 2006), Bedi and Garg (2000) find that
workers who attended private non religious schools experience a 75 per cent earnings advan-
tage over graduates of public schools. Without correcting for selectivity bias, public school
∗e-mail: [email protected].
1
graduates are found to earn 31 per cent more over private non religious graduates. However,
a negative selection effect for private non religious schools is identified and correcting for this
selection effect results in a large earnings premium of private non religious schools over public
schools. Despite private non religious school graduates generally having lower academic qual-
ifications than public school graduates, it is argued that the private non-religious schools are
more effective than the public schools. Bedi and Garg’s (2000) finding supports Hannaway’s
(1991) claim, “that private schools perform better due to greater school level autonomy and
their responsiveness to the needs of students and parents”. The policy implication of these
findings is to encourage a greater private sector role in Indonesian education since the results
suggest the private sector is more efficient and effective in delivering education.
Along with Bedi and Garg (2000), Newhouse and Beegle (2006) investigate the effective-
ness of private and public lower secondary schools in the Indonesian context, focus on the
relationship between school choice and academic performance rather than school choice and
future earnings. Newhouse and Beegle (2006) found the academic performance of public lower
secondary school students was superior to private school students as measured by national fi-
nal test exam scores (UN1) upon completion of lower secondary school. An important aspect
of the findings by Newhouse and Beegle (2006) is that they are difficult to reconcile with the
earlier findings by Bedi and Garg (2000). In particular, Newhouse and Beegle (2006) suggest
that the resource advantages of public schools are unlikely to be outweighed by any efficiency
benefits of private schools.
In this paper, I re-examine the earnings differential between public and private lower sec-
ondary school students, originally studied by Bedi and Garg (2000). Despite careful attempt to
replicate these earlier results, I am unable to do so. While Bedi and Garg (2000) use the first
wave of the Indonesia Family Life Survey (IFLS1) that issued in 1996 (DRU-1195-CD), my
sample data are obtained from the re-release version of IFLS1 (IFLS1-RR).
In order to explain the differences, I discuss the use of school quality proxies by Bedi
1UN or Ujian Nasional is the newest system of national centralized final examination. In 2008, IndonesianGovernment increased the standard average passing grade from minimum 5.25 to 5.50 for 6 subjects.
2
and Garg (2000). Several variables which identify the condition of the last school attended
confound lower and upper secondary schools. The use of such proxies may bias their estimates
of the earnings differential.
Using the re-release sample data of IFLS1 and with an absence of school quality indicators,
my findings suggest that the surprising findings of Bedi and Garg (2000) are not robust. My
findings suggest that public school graduates earn significantly higher than private non reli-
gious school graduates and imply that the quality of public schools are better than private non
religious schools.
In the next section, I try to replicate Bedi and Garg (2000) data sample. The following
section introduces the empirical strategy to re-estimate the effect of lower secondary schools
quality on earnings differential. Section 4. provides the results of school choice and earnings
decomposition estimates, while Section 5. concludes the paper.
2. Sample Replication
In order to replicate the result of Bedi and Garg (2000), I first try to create an identical dataset,
using the Indonesia Family Life Survey 1 (IFLS1) 1993. The IFLS1 is a large-scale lon-
gitudinal survey of socio economic and health status of the individual and household level.
The IFLS1 sampling scheme was based on provinces, then the sample was randomly selected
within these provinces. For cost-effectiveness reasons the survey focused on only 13 out of
26 provinces on the Island of Java, Sumatra, Bali, West Nusa Tenggara, Kalimantan, and Su-
lawesi. These were selected to represent approximately 83 per cent of the Indonesian popu-
lation. While waves 2, 3, and 4 of the IFLS have been released between 1997 and 2009, the
analysis in Bedi and Garg (2000) focuses on IFLS1. I assume that their study commenced
when only IFLS1 was available, and while IFLS2 was released in 1997, it did not contain
employment data required to extend the analysis.
I created a sample data based on Bedi and Garg’s (2000) guidance (pages 467-468). Fol-
3
lowing the guidance in pages 467-468 of Bedi and Garg (2000), I attempt to replicate their
sample. However, I was unable to exactly reproduce their sample. My initial sample data set
consisted of 7220 respondents who have earnings and are no longer students. The size of the
initial data was almost twice the size of Bedi and Garg’s (2000) initial sample data of 4900
observations. Missing and miscoded data and also sample restrictions then reduced the data set
by 6170 (more than 85 per cent) to 1050 observations. Most of the observations, 5448, were
dropped as respondents had not proceeded beyond primary school, while 274 observations
were dropped since respondents had more than 12 years of education. Moreover, I dropped 13
respondents due to missing information on school type and 9 observations as they had either
99997 or 999997 on total monthly earnings. Finally, I exclude a further 389 observations as
they had some missing information, miscoded class size (41 observations), number of months
in school period per year (45), failed in primary school (1), parents’ education (294), province
where school is located (6) and religion (2). Table 1 presents the full comparison of the exclu-
sion process with results of Bedi and Garg (2000).
4
Table 1: Comparison of Exclusion Process
Item Bedi and Fahmi*
Garg (2000)
Initial income information 4900 7220
Had not proceeded beyond primary education 3391 5448
Had more than 12 years of education 291 274
Lack of information on hours of work 33 37
Missing information on school type 10 13
Reported incomes seemed implausibly high 3 9
Missing information on class size - 41
Attend(ed) school more than 12 months (miscoded) - 45
Missing information on failed in primary school - 1
Missing information on father’s education - 214
Missing information on mother’s education - 80
Missing information on school location - 6
Missing information on religion - 2
Number of remaining observation 1194 1050
5
In order to understand the differences between my sample and the sample extracted by Bedi
and Garg (2000) it is important to note that Bedi and Garg (2000) used the IFLS1 issued by
RAND in 1996 (DRU-1195-CD), while I used the IFLS1 data set called IFLS1-RR (re-release
in 2000) that updates the original IFLS1. Peterson (2000) explains that IFLS1-RR revises and
restructures the original IFLS1 to accompany with IFLS2. The different between IFLS1 DRU-
1195-CD and IFLS1-RR maybe the source of the discrepancy between my sample and that
of Bedi and Garg (2000). Arjun Bedi kindly sent the sample data set, PUBPRIV.DTA2. Bedi
and Garg (2000)create the file on 7 February 1998 which consists of 1527 observations and
231 variables. However, they were unable to provide the code used to construct the sample,
making it impossible to clearly identify the sources of the discrepancies.
2The file PUBPRIV.dta originally consists of 1,527 observations, 231 variables. It is created on 7 February1998. The exclusion of some observations of missing data on earnings information drops the final sample data setto 1194.
6
Table 2: Tracking Process of Mismatch Sample Data
No. Note
Obs.
745 Identical
17 Unidentified
152 Had more than 12 years’ education.
34 - Missing information on period in school in months.
- Bedi and Garg (2000) substitute the missing data by sample mean.
32 - Missing information on class size.
- Bedi and Garg (2000) substitute the missing data by sample mean.
154 - Missing information on father’s education.
- Bedi and Garg (2000) put "0" instead of missing values in three dummy
variables of father’s of education.
- Three variables of father’s education are FATH_PRI, FATH_JH and
FATH_SH.
60 - Missing information on mother’s education.
- Bedi and Garg (2000) put "0" instead of missing values in two dummy vari-
ables of mother’s education.
- Two variables of mother’s education are MOTH_PRI and MOTH_SEC.
I was able to compare my sample (1050) with Bedi and Garg’s (2000) sample (1194) using
survey identification code. It is possible to match 745 respondents precisely. Of the remaining
449 observations, 17 observations are unidentified and 432 are considered as missing informa-
tion. Conversely, my sample contained 305 observations that were not in the sample provided
by Bedi and Garg (2000).
Of the 305 observations with missing data, 34 observations have no information on the
number of months per year of attending school and 32 observations have no information on
7
class size. Bedi and Garg (2000) substitute for the missing data on those observations by using a
sample mean instead of dropping the number of observations. The remaining 214 observations
have no information on either father’s or mother’s education. Bedi and Garg (2000) put "0"
value on those observations rather than dropping them. While the exclusion process is clearly
documented, the substitution process on the 305 observations is not explained in the paper. I
provide the details of my comparison in Table 2. I also present a full complete comparison of
summary statistics between the two sample in Table 3 and the description of all variables in
Table 9 in Appendix 5.. Given the differences in the sample data, I proceed to replicate their
methodology using the sample that I have extracted.
Table 3: Comparison of Descriptive Statistics
Variable Bedi and Garg (2000) Fahmi
Mean Std. Dev Mean Std. Dev
LOGEARN -0.202 1.079 -0.290 1.063
EARN 1.492 2.567 2.030 17.655
AGE 34.66 7.502 34.264 7.321
JUNIOR 0.307 0.462 0.415 0.493
SENIOR 0.521 0.499 0.527 0.500
MALE 0.672 0.469 0.689 0.463
BAHASA 0.404 0.491 0.370 0.483
HIN_BUD 0.066 0.248 0.074 0.262
CHRIST 0.091 0.289 0.092 0.290
PRI_FAIL 0.204 0.403 0.208 0.406
SCHOLAR 0.048 0.215 0.040 0.196
FATH_PRI 0.422 0.494 0.521 0.500
FATH_JH 0.101 0.302 0.113 0.317
FATH_SH 0.085 0.279 0.084 0.277
MOTH_PRI 0.380 0.485 0.470 0.499
Continued on Next Page. . .
8
Table 3 – Continued
Variable Bedi and Garg (2000) Fahmi
Mean Std. Dev Mean Std. Dev
MOTH_SEC 0.109 0.312 0.094 0.292
DIRT FLOOR 0.067 0.251 0.044 0.205
CLASS SIZE 36.47 9.301 36.651 8.884
MONTHS 9.459 1.849 9.638 1.710
OTH_PR 0.023 0.148 0.031 0.175
SKALI_ED 0.043 0.204 0.036 0.187
NSUMA_ED 0.106 0.308 0.097 0.296
WSUMA_ED 0.068 0.253 0.049 0.215
SSUMA_ED 0.051 0.220 0.052 0.223
LAMP_ED 0.023 0.151 0.027 0.161
EJAVA_ED 0.120 0.325 0.135 0.342
WJAVA_ED 0.139 0.346 0.131 0.338
CJAVA_ED 0.141 0.348 0.155 0.362
BALI_ED 0.048 0.215 0.058 0.234
NTB_ED 0.042 0.200 0.056 0.230
YOGYA_ED 0.067 0.251 0.065 0.246
SSULA_ED 0.042 0.202 0.038 0.192
JAKAR_ED 0.079 0.270 0.069 0.253
URBAN 0.708 0.455 0.670 0.470
SKALMNT 0.043 0.204 0.050 0.219
NSUMATRA 0.098 0.297 0.084 0.277
WSUMATRA 0.066 0.250 0.045 0.207
SSUMATRA 0.053 0.225 0.057 0.232
EJAVA 0.103 0.304 0.117 0.322
WJAVA 0.131 0.338 0.125 0.331
CJAVA 0.088 0.284 0.098 0.298
BALI 0.054 0.226 0.068 0.251
Continued on Next Page. . .
9
Table 3 – Continued
Variable Bedi and Garg (2000) Fahmi
Mean Std. Dev Mean Std. Dev
NTB 0.042 0.202 0.057 0.232
LAMPUNG 0.029 0.168 0.034 0.182
YOGKARTA 0.067 0.251 0.065 0.246
SSULAWES 0.042 0.202 0.040 0.196
JAKARTA 0.176 0.381 0.160 0.367
Number of Sample 1194 1050
3. Estimation Procedure and Sample
I use a two-step earnings estimate with selection bias correction and Blinder-Oaxaca Decom-
positions to determine the earnings differential between public school group and private school
groups. The two step earnings estimates that is corrected for selection bias problem is based
on the technique that developed by Lee (1983). I follow Lee (1983) to employ an unordered
multinomial logit (MNL) model in obtaining the selection correction terms and estimating the
lower secondary school choice.
To determine the effect of school quality on earnings differential, I follow Bedi and Garg
(2000) to estimate four separate earnings estimates: public, private non religious, private Is-
lamic and private Christian. According to Kingdon (1996) this method is able to avoid endo-
geneity problem. The earnings determination of individual i who attended school type j may
be written as
10
Yij = βjXij + eij (1)
where Y is earnings, β is parameters of exogenous variable X that consists of personal and
family characteristics, and u is the error terms.
The inclusion of selection correction terms to overcome the selection bias problem modifies
the equation 1 to
Yij = βjXij + θjλij + ηij (2)
where
λij =φ(Hij)
Φ(Hij
(3)
and
Hij = Φ−1(Pij) (4)
θ is the coefficient on inverse Mills ratio λij , ηij is error terms, φ(Hij is the standard normal
density function, Φ(Hij is the normal distribution and Pij is probability of individual i chooses
the type school j.
I follow Bedi and Garg (2000) using the Blinder-Oaxaca decomposition to estimate earn-
ings differential between public school and private school graduates. Bedi and Garg (2000)
use the two-fold decomposition that included some non-discriminatory coefficient vectors to
determine the contribution of the gap in the predictors. According to Reimers (1983), the two
fold decomposition can be written as
lnYk − lnYm = (X̄j − X̄m)[β̂jD + β̂m(I −D)] + (β̂j − β̂m)[X̄j(I −D) + X̄mD] (5)
11
where the subscript k refers to the public schools group and the subscriptm refers to private
schools groups. lnY is the natural logarithm of individual earnings. X is a vector of observed
characteristics and β is a vector of coefficients on observed characteristics. I is the identity
matrix and D is a diagonal matrix of weights.
Now the two-fold decomposition is
R = Q+ U (6)
where R is the earnings difference. The first component, Q, is the earnings differential
that is "explained" by group differences in the predictors. The first difference is also known as
quantity effect. The second part, U is the "unexplained" part. U is the differences caused by
discrimination and unobserved variables.
I also follow Bedi and Garg (2000) to use the mean coefficients between the low and the
high model or D = 0.5 proposed by Reimers (1983). Reimers (1983) believes that the dis-
crimination in labour market could affect the earnings of either the majority or minority group.
Therefore, Reimers (1983) suggests that the diagonal of D (matrix of weights) should equal
0.5 to avoid the inconsistency in decomposition result.
I use several STATA modules to estimate school choice and earning differential. To obtain
the marginal effect of the explanatory of the multinomial logit estimates, I use the STATA
module mfx2 developed by Williams (2006). Williams (2006) claims that mefx2 enhances
the marginal effect computation and post-estimation table formatting. I use the STATA module
selmlog developed by Jann (2008a) to estimate the earnings equation with the selection
correction terms. Using selmlog, I obtain the mean and the coefficient on λ in the first step
and estimate the earnings equation with selection bias corrected in the second step.
To estimate the Blinder-Oaxaca earnings decompositions, I employ the STATA module
oaxaca3 developed by Jann (2008b). The oaxaca’s command is not only capable to calculate
the earnings decompositions but it is also capable to obtain the standard errors and covariances.
3See Jann (2008c) for methods and formulas.
12
The sample data for this study that is obtained from the first wave of the Indonesia family
Life Survey (IFLS1) consists of 1,526 observations. The sample of this study consists of in-
dividuals with minimum lower secondary education and their primary activity during the past
week of the interview is working, trying to work or helping to earn income.
The sample contains 966, 295, 155 and 110 observations from public school, private non
religious, private Islamic and private Christian school groups. The size of initial sample with
earnings and lower secondary education information is 2,104. Dropping observations with
missing values on one ore more variables reduces the sample size to 1,526. I present the sample
means of all variables by school-type in Table 4. I also present the summary of statistics of all
sample in 10 in Appendix 5..
13
Tabl
e4:
Sam
ple
Mea
nsby
Scho
ol-T
ype
Vari
able
Publ
icPr
ivat
eN
RPr
ivat
eIs
Priv
ate
Ch
Mea
nSt
d.D
ev.
Mea
nSt
d.D
ev.
Mea
nSt
d.D
ev.
LO
GE
AR
NIN
GSM
-0.1
131.
077
-0.3
451.
043
-0.2
921.
195
-0.1
261.
13
EA
RN
ING
SM2.
481
19.3
771.
196
1.99
81.
857
4.30
41.
546
1.77
4
AG
E34
.907
7.16
334
.325
7.47
84.
729
7.52
937
.209
6.65
4
SEX
0.73
20.
443
0.69
80.
460.
697
0.46
10.
627
0.48
6
LA
NG
IND
O0.
414
0.49
30.
393
0.48
90.
316
0.46
60.
445
0.49
9
ISL
AM
0.85
70.
350.
780.
415
0.99
40.
080.
473
0.50
2
CIT
Y0.
212
0.40
90.
261
0.44
0.13
50.
343
0.28
20.
452
TOW
N0.
264
0.44
10.
224
0.41
70.
168
0.37
50.
373
0.48
6
VIL
LA
GE
0.52
40.
500
0.51
50.
501
0.69
70.
461
0.34
50.
478
PRIF
AIL
0.19
80.
398
0.24
10.
428
0.22
60.
419
0.19
10.
395
JUN
IOR
0.33
10.
471
0.39
30.
489
0.45
20.
499
0.31
80.
468
SEN
IOR
0.48
30.
500
0.47
50.
500
0.44
50.
499
0.45
50.
500
Con
tinue
don
Nex
tPag
e...
14
Tabl
e4
–C
ontin
ued
Vari
able
Publ
icPr
ivat
eN
RPr
ivat
eIs
Priv
ate
Ch
Mea
nSt
d.D
ev.
Mea
nSt
d.D
ev.
Mea
nSt
d.D
ev.
Mea
nSt
d.D
ev.
WO
RK
SD0.
068
0.25
20.
078
0.26
90.
097
0.29
70.
082
0.27
5
FAT
HPR
I0.
759
0.42
80.
817
0.38
70.
852
0.35
70.
745
0.43
8
FAT
HJH
0.12
80.
335
0.09
20.
289
0.08
40.
278
0.14
50.
354
FAT
HSH
HE
0.11
30.
317
0.09
20.
289
0.06
50.
246
0.10
90.
313
MO
TH
PRI
0.87
10.
336
0.89
50.
307
0.89
70.
305
0.8
0.40
2
MO
TH
JH0.
084
0.27
70.
061
0.24
0.07
10.
258
0.07
30.
261
MO
TH
SHH
E0.
046
0.20
90.
044
0.20
60.
032
0.17
70.
127
0.33
5
Obs
erva
tions
966
295
155
110
15
4. Empirical Results
4.1 Lower Secondary School Choice
For estimation of the lower secondary school choice, I use a multinomial logit (MNL) model.
Following Bedi and Garg (2000), I assume individuals and their parents choose a set of lower
secondary school types that consists of a public school category, private non religious category,
private Islamic category, and private Christian category. I present the results of the school
choice multinomial logit estimates in Table 6 and I also provide marginal effects of explanatory
variables in Table 5.
The results show that individuals with more educated parents are more likely to attend pub-
lic schools. There is evidence that individuals with more educated parents also prefer to study
in private Christian school. However, this finding may be ambiguous since there are contradic-
tory signs on father’s upper secondary and higher education (FATHSHHE) and mother’s upper
secondary and higher education (MOTHSHHE) variables. Individuals with less educated par-
ents have higher probability of attending private non religious schools. There is no evidence
that parents’ education influence the decision of attending private Islamic schools.
As expected, non-Islamic background has a positive effect on enrolment of both private non
religious and private Christian schools. On the other hand, individuals with Islamic background
are more likely attending public and private Islamic schools. Males are more likely to enrol
in public schools, while females prefer to attend private Christian. There is no evidence that
sex affect the probability of access in private non religious and private Islamic schools. These
results are corresponding to Bedi and Garg’s 2000 findings.
Furthermore, individuals who lived in urban area when they were 12 years old have higher
probability of attending private non religious and private Christian schools. On the contrary,
public and private Islamic schools are more likely attended by individuals from rural area.
16
Table 6: Lower Secondary School Choice Estimations
Variable Private NR Private Is Private Ch
Coeff/(SE) Coeff/(SE) Coeff/(SE)
CITY 0.365** -0.671** 0.937***
(0.174) (0.268) (0.279)TOWN -0.095 -0.713*** 0.815***
(0.168) (0.235) (0.255)SEX -0.228 -0.295 -0.431*
(0.149) (0.193) (0.230)LANGINDO -0.124 -0.217 -0.151
(0.147) (0.196) (0.223)ISLAM -0.571*** 3.266*** -2.015***
(0.170) (1.008) (0.219)PRIFAIL 0.237 0.086 0.037
(0.162) (0.213) (0.276)WORKSD 0.115 0.253 0.514
(0.264) (0.307) (0.398)FATHJH -0.410* -0.497 -0.100
(0.239) (0.326) (0.310)FATHSHHE -0.343 -0.549 -0.949**
(0.256) (0.400) (0.421)MOTHJH -0.231 0.204 -0.149
(0.292) (0.379) (0.418)MOTHSHHE 0.165 0.077 1.484***
(0.362) (0.508) (0.404)CONSTANT -0.539*** -4.346*** -0.962***
(0.203) (1.020) (0.300)Pseudo R-Square 0.066N 1526Chi2 157.603***
Note: Standard errors are in parenthesis and heteroscedasticity consistent. * Significance
at 10% level, ** Significance at 5% level and *** Significance at 1% level.
17
Table 5: Marginal Effects of School Choice Estimations
Variable Public Private NR Private Is Private Ch
Coeff/(SE) Coeff/(SE) Coeff/(SE) Coeff/(SE)
CITY -0.069* 0.056 -0.042*** 0.055**
(0.035) (0.030) (0.013) (0.020)
TOWN 0.005 -0.017 -0.040*** 0.052**
(0.030) (0.026) (0.012) (0.018)
SEX 0.061* -0.028 -0.014 -0.019(0.027) (0.024) (0.013) (0.012)
LANGINDO 0.032 -0.016 -0.011 -0.005(0.026) (0.023) (0.012) (0.010)
ISLAM 0.138*** -0.073* 0.111*** -0.176***
(0.034) (0.030) (0.010) (0.026)
PRIFAIL -0.039 0.038 0.002 -0.001(0.030) (0.028) (0.013) (0.013)
WORKSD -0.049 0.008 0.013 0.027(0.049) (0.042) (0.022) (0.027)
FATHJH 0.077* -0.055 -0.022 0.000(0.036) (0.032) (0.015) (0.015)
FATHSHHE 0.094* -0.039 -0.024 -0.031**
(0.040) (0.036) (0.018) (0.011)
MOTHJH 0.025 -0.037 0.018 -0.006(0.049) (0.041) (0.028) (0.018)
MOTHSHHE -0.116 -0.004 -0.007 0.127*
(0.069) (0.055) (0.028) (0.052)
N 1526
Pseudo R2 0.066
Log likelihood -1466.449
Note: Standard errors are in parenthesis and heteroscedasticity consistent. * Significance at
10% level, ** Significance at 5% level and *** Significance at 1% level.
18
4.2 Earnings Equations
In his college choice and earnings model, Strayer (2002) believes that there are two sources of
earnings determination that is influenced by school quality. First, a different type of high school
will generate a different level of earnings. The quality of a type of high school influences the
probability of attending a type college, then the earnings determination from different type of
college could be considered comes from the quality of the high school. Second, a high quality
school directly increases skills of students to gain higher earnings in labour market.
In my model, I assume that earnings differential mainly comes from the indirect effect
of school quality which is determined by different types of lower secondary school. I use
this assumption based on two reasons. First, the proxies of school quality indicators from
IFLS1 that used by Bedi and Garg (2000) may bias the earnings determination. Bedi and Garg
(2000) use three proxy variables: a dummy variable of whether the school has a dirt floor
(DIRT FLOOR), the length of the school term (MONTHS), and the number of students in the
class (CLASS SIZE). In IFLS2 and the later waves, the questions regarding length of school
term and class size are categorized on basis school level. IFLS2 do not provide information
about type of school’s floor in education section (Book 3A). According to the manual book
of IFLS1, type of floor, length of school term, and class size4 provide information about the
school characteristics last attended by respondents. Since my sample consists of individuals
whose range of level education are between lower secondary education and higher education,
information on these proxy variables may bias the estimations. Second, the availability of
standard variables for school quality indicators such as teacher-student ratio in IFLS1 is poor.
I only find 286 of 1530 observations that have teacher-student ratio.
I estimate separate earnings equations for each type of school. The earnings estimates are
estimated by including the selection correction variables that obtained from the multinomial
logit school choice estimates. I focus on the influence of school quality to earnings formation
4The information regarding type of floor, length of school term and class size are provided in the book ofBUK3DL3
19
and I also investigate the effect of academic achievement and attainment, personal and family
characteristics, and parents’ education on earnings determination. The estimated coefficients
on those variables measure the direct effect of these variables to earnings formation.
Table 7 shows the earnings estimates based on type of lower secondary school attended.
The estimated coefficients on personal and family characteristics measure the direct contribu-
tion of these factors to earnings determination. In all four school groups, the education attain-
ment variables (HE and SENIOR) are the most important factors for earnings determination.
Individuals with upper secondary and higher education earn higher earnings than their counter-
part who only graduate from lower secondary schools. The 0.522 log point difference on HE
implies that an individual from the public school group who is educated in a higher education
institution has probability to earn about 68 per cent higher than those with upper secondary
education. Similar condition also occurs in private non religious (46% higher), private Islamic
(51% higher) and private Christian (15% higher) school groups. An individual with upper sec-
ondary education earns about 57% (in public the school group), 46%(private non religious),
38% (private Islamic), and 35% (private Christian) higher than an individual who only attend a
lower secondary school. The significant coefficient on dummy variable PRIFAIL also inform
there is a direct return of early education achievement on earnings. Individuals who failed a
grade in primary education earn about 16 per cent lower if they attend public schools, while
the negative gap increases to about 115 per cent if they attend private Christian schools.
As expected, personal and family characteristics also have a direct effect to earnings for-
mation. Being male increases an individual’s earnings for about 23 per cent and 37 per cent
in public school and private non religious school groups respectively. Living in urban areas
increases individuals’ income for about 28 per cent and 95 per cent in private non religious
and private Islamic school groups respectively. The positive and significant coefficient on
LANGINDO in public (0.177) and private Christian (0.782) school groups implies that indi-
viduals who speak Indonesian language in daily life as they more likely lives in urban areas
enjoy a higher income by 19 per cent and 118 per cent respectively. It is also not surprising that
20
Individuals who live in urban areas and attend private Islamic schools earn 94 percent higher
than their counterpart within the sector.
Table 7: OLS Estimates of Earnings with Selection Correc-
tion
Variable Public Private NR Private Is Private Ch
Coeff/(SE) Coeff/(SE) Coeff/(SE) Coeff/(SE)
CONSTANT -2.623*** -3.611*** -2.036 -2.103
(0.880) (1.355) (2.844) (3.038)
AGE 0.087* 0.145* -0.088 0.107
(0.040) (0.057) (0.113) (0.152)
AGE2 -0.001 -0.002* 0.002 -0.001
(0.001) (0.001) (0.002) (0.002)
URBAN 0.115 0.249 0.663** -0.064
(0.076) (0.142) (0.224) (0.286)
SEX 0.203* 0.316* -0.081 0.369
(0.084) (0.137) (0.229) (0.213)
LANGINDO 0.177* -0.064 0.341 0.782**
(0.084) (0.154) (0.299) (0.276)
ISLAM 0.179 0.080 1.136 0.093
(0.169) (0.223) (1.411) (0.562)
PRIFAIL -0.152 -0.131 -0.254 -0.731**
(0.084) (0.150) (0.213) (0.241)
SENIOR 0.456*** 0.378** 0.324 0.303
(0.071) (0.124) (0.193) (0.230)
HE 0.978*** 0.761*** 0.737* 0.439
(0.096) (0.190) (0.339) (0.273)
FATHJH 0.082 -0.164 0.462 0.366
Continued on Next Page. . .
21
Table 7 – Continued
Variable Public Private NR Private Is Private Ch
Coeff/(SE) Coeff/(SE) Coeff/(SE) Coeff/(SE)
(0.121) (0.230) (0.369) (0.317)
FATHSHHE 0.238 -0.074 0.503 0.725
(0.150) (0.246) (0.453) (0.501)
MOTHJH -0.184 0.544 0.188 -0.422
(0.121) (0.281) (0.375) (0.459)
MOTHSHHE -0.042 0.276 -0.732 -0.041
(0.182) (0.311) (0.561) (0.593)
LAMBDA 0.072 -0.0715 -0.067 0.523
(0.594) (0.720) (0.666) (0.636)
Adj R2 0.255 0.243 0.339 0.390
N 966 295 155 110
Note:
- The dependent variables is the logarithm of hourly earnings.
- Standard Errors are in parenthesis and heteroscedasticity consistent.
- * Significance at 10% level, ** Significance at 5% level and *** Significance at 1% level.
- Dummy variables of regions of residence and school are included in estimation but not reported.
4.3 The Blinder-Oaxaca Earnings Decompositions
Table 8 highlights the effect of school quality on the earnings decompositions. I use the
Blinder-Oaxaca decomposition to determine the earnings gaps between public and private
schools. The base model is estimated from the earnings estimates in 2. The results on Table
8 suggest there are positive earnings gaps between public and private school groups. With-
out including the selection correction terms, the public school group earn about 26%, 19%
and 1% higher than private non religious, private Islamic and private Christian school group
22
respectively. However, only the public-private non religious gap that is statistically significant.
Using the two-step earnings estimate and including the selection correction term to the
estimates, the positive earnings gap between public and private non religious school group is
corrected or increased to about 45 per cent. This result is contradictory with Bedi and Garg’s
(2000) finding when the selection correction terms reversed the superiority of individuals form
the public school group over the private non religious group.
The positive earnings gap between public and private Islamic school group is also increased
to about 39 per cent. On the contrary, the inclusion of selection bias correction term reverses
the advantage of the public school group over the private Christian school group as the gap is
reversed to about -123 per cent. However, all these gaps are not statistically significant and
provide the weak evidence of the effect of selection bias to earnings decomposition of public
and private school groups.
The academic achievement gap in public-private non religious school and public-private
Islamic decompositions are positive and significant. These results suggest that academic at-
tainments have more effect on earnings for individuals from public school group than those
private school groups. Furthermore, the experiences in labour market are more important for
the private Christian school group than the public school groups since it contributes to a -6
percent of the total earnings gap.
23
Table 8: Two-Folds Decomposition
Private NR Private Is Private Ch
Difference 0.232** 0.178** 0.013
(0.078) (0.088) (0.099)
Adjusted 0.374 0.330 -0.803
(0.971) (1.388) (1.144)
Total Explained: 0.115** 0.066 -0.080
(0.046) (0.113) (0.177)
1. Experience 0.024 0.004 -0.064*
(0.016) (0.025) (0.036)
2. Personal/Familiy 0.015 -0.013 0.065
(0.017) (0.96) (0.142)
3. Academic 0.056** 0.091** -0.022
(0.022) (0.033) (0.028)
4. Parents’ Education 0.005 0.025 -0.002
(0.010) (0.023) (0.032)
5. Other Variables 0.016 -0.041 -0.057(0.030) (0.049) (0.077)
Total Unexplained 0.259 0.264 -0.723
(0.983) (1.467) (1.270)
Note: 50 replication of bootstrap standard errors are in parenthesis. * Significance at 10% level,
** Significance at 5% level and *** Significance at 1% level. Experience: age, age2, Personal and
Family Characteristics: Urban, Sex, Langindo and Islam; Academic achievement and attainment:
Prifail, Senior and HE; Parents’ Education: Fathjh, Fathshhe, Mothjh and Mothshhe.
24
The results that I have presented imply that the lower secondary school quality has an
indirect effect on earnings of individuals. The results show public school graduates earn sig-
nificantly higher than private non religious school graduates and these imply that the quality of
public schools are better than private non religious schools.
5. Conclusion
In this paper, I have tried to replicate the results of Bedi and Garg (2000). Given their use of
an early release of the IFLS data and my use of a re-released IFLS dataset, I was unable to
reproduce the sample they had used. I attempt to replicate their results using the sample that
obtained form the re-release version of IFLS1 (IFLS1-RR). While, I unable to replicate many
of their results, I am also unable to reproduce the large earnings differential of private non
religious school graduates relative to public school graduates.
The use of some proxies as school quality indicators in Bedi and Garg (2000) earnings
model may also bias the results. IFLS1 provides the information of school quality based on
school characteristics last attended by respondents, rather than information of school quality
by level of education. Since some of the respondents attended senior or higher education,
therefore, it may bias the validity of the model. I also obtain insufficient observations from
IFLS1 that have the standard school quality indicator, teacher-students ratio. With the absence
of school quality variables, I assume that different of school quality only affects indirectly to
earnings decomposition.
Using the re-release sample data of IFLS1 and with an absence of school quality indicators,
my findings suggest that public school graduates earn significantly higher than private non
religious school graduates and these imply that the quality of public schools are better than
private non religious schools.
Education attainments and academic achievement are the most important factors for earn-
ings determination. The graduates of public, private non-religious and private Islamic schools
25
who attend higher education institution earn about 50 per cent higher than their counterparts
with upper secondary education. Attending an upper secondary school increases the earn-
ings of graduates of all types of school by about 40 per cent. On the other hand, failing a
grade in primary education reduces the earnings of graduates by 115 and 16 per cent in private
Christian school and public school groups, respectively. These results support the productivity
argument for investing in young children (Heckman et al., n.d.; Heckman, 2006) and suggests
early childhood academic performance increases the individual’s earnings in the future.
26
References
Bedi, A. S. and Garg, A. (2000), ‘The effectiveness of private versus public schools: The case
of indonesia’, Journal of Development Economics 61, issue 2, 463–494.
Hannaway, J. (1991), ‘The organization and management of public and catholic schools: Look-
ing inside the black box’, International Journal of Educational Research 15, 463–481.
Heckman, J. (2006), ‘Skill formation and the economics of investing in disadvantaged chil-
dren’, Science 312(5782), 1900.
Heckman, J., Arbor, A. and Masterov, D. (n.d.), The productivity argument for investing in
young children.
Jann, B. (2008a), ‘The blinder-oaxaca decomposition for linear regression models’, The Stata
Journal 8(4), 453–479.
Jann, B. (2008b), ‘Oaxaca: Stata module to compute the blinder-oaxaca decomposition’, Sta-
tistical Software Components S 456936.
Jann, B. (2008c), A stata implementation of the blinder-oaxaca decomposition, ETH Zurich
Sociology Working Papers 5, ETH Zurich, Chair of Sociology.
Kingdon, G. (1996), ‘The quality and efficiency of private and public education: A case-study
of urban india’, Oxford Bulletin of Economics and Statistics 58(1), 57–82.
Lee, L. F. (1983), ‘Generalized econometric models with selectivity’, Econometrica 51, 507.
Newhouse, D. and Beegle, K. (2006), ‘The effect of school type on academic achievement:
Evidence from indonesia’, Journal of Human Resources 41(3), 529–557.
Peterson, C. E. (2000), Documentation for ifls1-rr: Revised and restructured 1993 indonesian
family life survey data, wave 1, Technical report, RAND.
27
Reimers, C. W. (1983), ‘Labor market discrimination against hispanic and black men’, The
Review of Economics and Statistics Vol. 65(No. 4), pp. 570–579.
Strauss, J., Beegle, K., Dwiyanto, A., Herawati, Y., Pattinasarany, D., Satriawan, E., Sikoki,
B., Sukamdi and Witoelar, F. (2004), Indonesian Living Standards Before and After the
Financial Crisis: Evidence from Indonesia Family Life Survey, Rand Corporation, USA and
Institute of Southeast Asian Studies.
Strayer, W. (2002), ‘The returns to school quality: College choice and earnings’, Journal of
Labor Economics 20(3), 475–503.
Williams, R. (2006), ‘Mfx2: Stata module to enhance mfx command for obtaining marginal
effects or elasticities after estimation’, Statistical Software Components.
28
Table 9: Definitions Variable Used in Bedi and Garg (2000)
Variable Description
LOGEARN Log hourly earnings
EARN Hourly earnings in thousands of rupiahs
AGE Age in years
JUNIOR Completed junior secondary education
SENIOR Completed senior secondary education
MALE Male
BAHASA Indonesian language spoken at home
HIN_BUD Religion Hindu or Buddhist
CHRIST Religion Christian
PRI_FAIL Failed a primary school grade
SCHOLAR Received scholarship at secondary school
FATH_PRI Father has primary education
FATH_JH Father has junior secondary education
FATH_SH Father has senior secondary education
MOTH_PRI Mother has primary education
MOTH_SEC Mother has secondary education
DIRT FLOOR School has dirt floorsa
CLASS SIZE Number of students in class a
MONTHS Length of school terma
OTH_PR Educated in other provincesb
SKALI_ED Educated in South Kalimantan
NSUMA_ED Educated in North Sumatra
WSUMA_ED Educated in West Sumatra
SSUMA_ED Educated in South Sumatra
LAMP_ED Educated in Lampung
EJAVA_ED Educated in East Java
Continued on Next Page. . .
29
Table 9 – Continued
Variable Description
WJAVA_ED Educated in West Java
CJAVA_ED Educated in Central Java
BALI_ED Educated in Bali
NTB_ED Educated in Nusa Tenggarra Barat
YOGYA_ED Educated in Yogyakarta
SSULA_ED Educated in South Sulawesi
JAKAR_ED Educated in Jakarta
URBAN Resides in an urban area
SKALMNT Resides in South Kalimantan
NSUMATRA Resides in North Sumatra
WSUMATRA Resides in West Sumatra
SSUMATRA Resides in South Sumatra
EJAVA Resides in East Java
WJAVA Resides in West Java
CJAVA Resides in Central Java
BALI Resides in Bali
NTB Resides in Nusa Tengarra Barat
LAMPUNG Resides in Lampung
YOGKARTA Resides in Yogyakarta
SSULAWES Resides in South Sulawesi
JAKARTA Resides in Jakarta
30
Tabl
e10
:Var
iabl
esD
escr
iptio
nan
dSa
mpl
eM
eans
Vari
able
Des
crip
tion
Mea
nSt
d.D
ev.
LO
GE
AR
NIN
GSM
Log
hour
lyea
rnin
gs-0
.177
1.09
EA
RN
ING
SMH
ourl
yea
rnin
gsin
thou
sand
sof
rupi
ah2.
102
15.5
16
Pers
onal
and
Fam
ilyC
hara
cter
istic
s
AG
EA
gein
year
s34
.942
7.25
2
SEX
Mal
e=1,
Fem
ale=
00.
714
0.45
2
LA
NG
IND
OIn
done
sian
lang
uage
spok
enat
hom
e0.
402
0.49
1
ISL
AM
Rel
igio
nIs
lam
?ye
s=1,
no=0
0.82
80.
377
CIT
YR
esid
esin
anci
tyar
eaw
hen
12ye
ars
old
0.21
90.
414
TOW
NR
esid
esin
anto
wn
area
whe
n12
year
sol
d0.
254
0.43
6
VIL
LA
GE
Res
ides
inan
villa
gear
eaw
hen
12ye
ars
old
0.52
70.
499
UR
BA
NR
esid
esin
anur
ban
area
0.70
580.
456
WO
RK
SDw
orki
ngw
hile
inpr
imar
ysc
hool
0.07
40.
262
Aca
dem
icA
chie
vem
ent
and
Att
ain-
men
t
PRIF
AIL
Faile
da
prim
ary
scho
olgr
ade
0.20
80.
406
JUN
IOR
Atte
ndlo
wer
seco
ndar
yed
ucat
ion
0.35
50.
479
Con
tinue
don
Nex
tPag
e...
31
Tabl
e10
–C
ontin
ued
Vari
able
Des
crip
tion
Mea
nSt
d.D
ev.
SEN
IOR
Atte
ndup
pers
econ
dary
educ
atio
n0.
476
0.50
0
HE
Atte
ndH
ighe
rEdu
catio
n0.
170
0.37
6
Pare
ntsE
duca
tion
FAT
HPR
IFa
ther
has
prim
ary
educ
atio
n0.
779
0.41
5
FAT
HJH
Fath
erha
slo
wer
seco
ndar
yed
ucat
ion
0.11
80.
323
FAT
HSH
HE
Fath
erha
sup
pers
econ
dary
educ
atio
n0.
104
0.30
5
MO
TH
PRI
Mot
herh
aspr
imar
yed
ucat
ion
0.87
30.
333
MO
TH
JHM
othe
rhas
low
erse
cond
ary
educ
atio
n0.
077
0.26
7
MO
TH
SHH
EM
othe
rhas
uppe
rsec
onda
ryed
ucat
ion
0.05
0.21
8
Reg
ions
ofSc
hool
OT
HPR
Edu
cate
din
othe
rpro
vinc
esb
0.13
40.
34
SKA
LIE
DE
duca
ted
inSo
uth
Kal
iman
tan
0.03
50.
185
NSU
MA
ED
Edu
cate
din
Nor
thSu
mat
ra0.
10.
3
WSU
MA
ED
Edu
cate
din
Wes
tSum
atra
0.05
40.
227
SSU
MA
ED
Edu
cate
din
Sout
hSu
mat
ra0.
050.
219
LA
MPE
DE
duca
ted
inL
ampu
ng0.
021
0.14
3
EJA
VAE
DE
duca
ted
inE
astJ
ava
0.14
20.
349
WJA
VAE
DE
duca
ted
inW
estJ
ava
0.13
80.
345
Con
tinue
don
Nex
tPag
e...
32
Tabl
e10
–C
ontin
ued
Vari
able
Des
crip
tion
Mea
nSt
d.D
ev.
CJA
VAE
DE
duca
ted
inC
entr
alJa
va0.
145
0.35
3
BA
LIE
DE
duca
ted
inB
ali
0.05
10.
22
NT
BE
DE
duca
ted
inN
usa
Teng
garr
aB
arat
0.04
90.
216
YO
GYA
ED
Edu
cate
din
Yog
yaka
rta
0.06
60.
249
SSU
LA
ED
Edu
cate
din
Sout
hSu
law
esi
0.04
10.
199
JAK
AR
ED
Edu
cate
din
Jaka
rta
0.07
10.
258
Reg
ions
ofR
esid
ence
SKA
LM
NT
Res
ides
inSo
uth
Kal
iman
tan
0.05
0.21
9
NSU
MA
TR
AR
esid
esin
Nor
thSu
mat
ra0.
086
0.28
WSU
MA
TR
AR
esid
esin
Wes
tSum
atra
0.04
80.
215
SSU
MA
TR
AR
esid
esin
Sout
hSu
mat
ra0.
054
0.22
7
WJA
VAR
esid
esin
Wes
tJav
a0.
132
0.33
9
CJA
VAR
esid
esin
Cen
tral
Java
0.09
10.
288
EJA
VAR
esid
esin
Eas
tJav
a0.
121
0.32
6
BA
LI
Res
ides
inB
ali
0.06
0.23
8
NT
BR
esid
esin
Nus
aTe
ngar
raB
arat
0.05
10.
22
LA
MPU
NG
Res
ides
inL
ampu
ng0.
026
0.16
YO
GK
AR
TAR
esid
esin
Yog
yaka
rta
0.07
10.
257
Con
tinue
don
Nex
tPag
e...
33
Tabl
e10
–C
ontin
ued
Vari
able
Des
crip
tion
Mea
nSt
d.D
ev.
SSU
LA
WE
SR
esid
esin
Sout
hSu
law
esi
0.04
10.
199
JAK
AR
TAR
esid
esin
Jaka
rta
0.16
70.
373
N15
26
34
Table 11: OLS Estimates of Earnings (No Selection Correction)
Variable Public Private NR Private Is Private Ch
Coeff/(SE) Coeff/(SE) Coeff/(SE) Coeff/(SE)
CONSTANT -2.686*** -3.528*** -1.855 -3.129(0.687) (0.918) (2.056) (2.099)
AGE 0.087** 0.145*** -0.088 0.123(0.040) (0.049) (0.113) (0.116)
AGE2 -0.001 -0.002** 0.002 -0.001(0.001) (0.001) (0.002) (0.002)
URBAN 0.115 0.248* 0.671*** 0.000(0.082) (0.143) (0.222) (0.273)
SEX 0.208*** 0.321** -0.073 0.330(0.075) (0.136) (0.206) (0.208)
LANGINDO 0.180** -0.064 0.347 0.776***(0.080) (0.172) (0.270) (0.293)
ISLAM 0.193 0.093 1.041** -0.338(0.122) (0.181) (0.525) (0.204)
PRIFAIL -0.156* -0.138 -0.255 -0.701***(0.085) (0.134) (0.187) (0.231)
SENIOR 0.456*** 0.379*** 0.326* 0.302(0.075) (0.121) (0.193) (0.229)
HE 0.978*** 0.762*** 0.742** 0.445*(0.088) (0.182) (0.323) (0.235)
FATHJH 0.091 -0.155 0.476 0.366(0.089) (0.227) (0.419) (0.299)
FATHSHHE 0.250** -0.070 0.521 0.500*(0.124) (0.263) (0.356) (0.282)
MOTHJH -0.184 0.553** 0.182 -0.362(0.118) (0.250) (0.369) (0.387)
MOTHSHHE -0.052 0.279 -0.734** 0.340(0.176) (0.309) (0.325) (0.277)
Note: Standard Errors are in parenthesis and heteroscedasticity consistent. * Significance at 10%
level, ** Significance at 5% level, and *** Significance at 1% level. Dummy variables of regions
of residence and school are included in estimation but no reported.
35