WPS4975
Policy Research Working Paper 4975
The Economics of Teacher Supply
in Indonesia
Dandan Chen
The World Bank
East Asia and Pacific Region
Human Development Sector Department
June 2009
Policy Research Working Paper 4975
Abstract
This paper examines the phenomenon of the over- teachers and that of alternative occupations significantly
supply of teachers but shortage of qualified teachers in influence the decision of college educated workers to
Indonesia. Using a theoretical framework of government- become teachers. It is also found that the wage rate
dominated market with government-set wage rate and set by the most recent teacher law would increase the
demand for teachers, the analysis explores how teacher share of teachers approximately from 16 to 30 percent
supply, particularly the composition of the teaching force of the college-educated labor force. This increase that is
with low or high qualification, would be determined due to the new government-set wage rate, would result
by current and future public policies. Using 2001 to in a pupil-teacher ratio of 24 to 25 pupils per teacher
2008 Indonesian Labor Force Survey data, the paper with college education, but will require a more than
further estimates the potential effect of the most recent 31 percent increase in the wage bill for teacher salaries.
teacher law, which could give college educated teachers The empirical approach of this paper is derived from a
a significant pay increase, on the composition of the structural model that takes into account the endogeneity
Indonesian teaching force with differentiated education of the wage rate and corrects for sample-selection bias
backgrounds. Using a sample of workers with college due to occupational choice.
education, the author finds that the relative wage rate of
This paper--a product of the Human Development Sector Department, East Asia and Pacific Region--is part of a larger
effort in the department to strengthen evidence-based human development policy advisory and dialogue. Policy Research
Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at dchen1@
worldbank.org.
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
Produced by the Research Support Team
The Economics of Teacher Supply in Indonesia 1
Dandan Chen
1
The author thanks Imam Setiawan for the excellent research assistance.
I. Introduction and literature review
Teachers are not only a major determinant of student learning, but also absorb a
large amount of public resources. The supply of teachers has attracted much attention in
academic research as well as in public policy debate. Dolton (2006) gave a
comprehensive review of the existing empirical work on teacher supply issues, including
entering teaching as the first job, staying in teaching, changing jobs and exiting teaching,
and re-entering teaching. A majority of the existing literature on teacher supply is on
developed countries, such as the US and the UK, where teacher supply issues have
received a high level of attention, matched with rich micro and time-series data. One
major research area is the determinants of the choice for teaching as an occupation, and
particularly the effects of the differentiated earnings of teaching and non-teacher
occupations.
Manski (1985)'s early work examined the relationship between academic ability,
earnings, and the decision to become a teacher through analysis of data from a national
sample of college graduates in the US. He found that the frequency of choice of teaching
as an occupation is inversely related to academic ability. In the meantime, the earnings of
teachers tend to rise only slightly, if at all, with academic ability. An econometric
analysis suggests that in the absence of a minimum ability standard, increases in teacher
earnings would yield substantial growth in the size of the teaching force but minimal
improvement in the average academic ability of teachers. The average ability of the
teaching force can be improved and the size of the teaching force maintained if minimum
ability standards are combined with sufficient salary increases.
2
Flyer and Rosen (1997) examined the effects of the rising labor force participation
of women on both demand for and supply of teachers. They concluded that in the US,
the rising demand for teachers during the past few decades is attributable to the growing
market opportunities for women and thus higher demand for a market substitute for home
schooling. Widened occupational choice for women has also attracted women away from
the traditionally female dominated occupations. Nonetheless, the flexibility of teaching
and less wage penalty for temporary leaves remain important attractions for women to
work in teaching jobs during their productive years.
Using the US longitudinal survey of the High School Class of 1972, Van der
Klaauw (1997) estimated a dynamic utility model of occupational choice and
occupational mobility, accounting for first and subsequent occupation choice. He found
that teacher salaries and opportunity wages are important determinants of the supply and
retention of teachers. Csellack (2002) also used a dynamic structural model and panel
data (NLSY 1979, US) looking into teacher occupational choice. The simulation results
show that a 2 percent increase in teachers' wages increases teacher supply by 2.6 percent.
Hanushek and Pace (1995), on the other hand, find that in the US, participation in
teacher training is not significantly affected by relative teacher earnings. In addition,
teacher candidates perform lower on tests than other graduates, and teacher training
completion is lowered by state requirements for courses and teacher tests.
Using UK data, Dolton (1990) and Dolton and Chung (2004) focused on looking
into the problem of recruiting graduates into the teaching profession and retaining them
in the UK. Dolton (1990) found that relative earnings in teaching and non-teaching
occupations and the corresponding growth in earnings in these two choices have a
3
marked effect on graduates' choices. In particular, the lower are the relative wages or
wage growth in teaching, the less likely is a graduate to choose that career. These
earning effects operate on initial choices and choices made later in an individual's career.
Dolton and Chung (2004) compared the earnings of qualified teachers who choose to
teach with the opportunity wage for those who do not teach. They find that the rate of
return on career choice for teachers has been declining for both men and women over the
past 25 years although teaching is still relatively well paid for women.
There are also a few studies on the earnings effect on the aggregated teacher labor
market. Thomas (1975) uses UK time series data between 1962 and 1970, and finds
significant salary effects suggesting a 1 percent fall in relative starting salaries will
induce a 2-5 percent fall in the relative supply of male graduates entering teaching and
similar effects for average salaries of teachers. Effects for female graduates are up to
twice as big. Zabalza, Turnbull and Williams (1979) also used UK time series, and
estimated that the elasticity of graduate new entrants with respect to wage equals 2.4-3.9
for men and 0.3-1.8 for women; and that with respect to relative starting wages, it is 3.4
for men and 2.8 for women. Court et al. (1995) updated the time-series study in the UK,
extending the study period to 1986-1992. The salary effect is still strong, suggesting that
a 1 percent fall in relative starting salaries will induce a 4 percent fall in the relative
supply of graduates entering teaching.
One recent study outside the US and UK is by Ortega (2006). Using Venezuelan
data, his results suggest that the teacher wage premium and wage dispersion have little
effect on the quality mix of applicants to teaching. Most students' preference for teaching
4
is unresponsive to wage levels relative to other occupations and to wage growth prospects
within teaching.
II. Indonesian context
The overall teacher supply seems sufficient in Indonesia, with the pupil-teacher
ratio declining steadily in recent years (Figure 1). The current pupil-teacher ratio is
estimated at 19:1 for the primary level and 13:1 for the secondary level. These ratios
clearly indicate that at an aggregated level, Indonesia does not have teacher shortage
issues, unlike some other countries in the region, such as the Philippines or Cambodia.
As a matter of fact, its pupil-teacher ratios are close to those of some developed countries
such as Japan (20:1 at primary level, 13:1 at secondary level), or the US (15:1 at primary
level and secondary level) 2. Based on a school survey, the World Bank (2007) reports
that a large proportion of Indonesian schools has an over-supply of teachers even
according to the existing generous staffing norm. The report further points out that there
is some space to improve the efficiency of the staffing norm. Based on the recommended
formula, nearly 85 percent of the schools are over-staffed.
2
Edstats online database.
5
Figure 1: Pupil-Teacher Ratio in Indonesia: 2001-2006
24.0
22.2
22.0 21.0 20.7 20.7
20.0 19.5
19.0
18.0
15.7 16.0
16.0
15.3
14.0 13.9
13.1 13.5
13.5 13.7 13.7 13.4 12.5
12.0
10.0
2001 2002 2003 2004 2005 2006
primary junior secondary senior secondary
However, in recent years, concern with the composition of the teaching force has
been more prominent, particularly with the general recognition that learning outcomes are
influenced by the ability of teachers who guide the learning process. There is an often
expressed dissatisfaction with the distribution of ability within the present teaching force,
with the public sentiment that public policies should induce desirable changes. With this
line of thought, teachers' minimum qualification, in terms of educational attainment, has
been raised several times during the past decade, even though there is little consensus on
whether educational attainment is a good measure of "ability" relevant for teachers.
The most significant policy change in teacher employment is the latest Teacher
Law (UU14/2005) passed in 2005. It caught much attention with the 100 percent teacher salary
increase for certified teachers. One of the key requirements for certification is four or more years
of college education (S1 or D4). The Ministry of Education set the goal that by 2015, the whole
teaching force should be comprised of only certified teachers. The magnitude of the potential
change will be large given that a large proportion of incumbent teachers are below the minimum
qualification level (Table 1). For example, by early 2006, over 80 percent of the primary school
teachers were without college education. Given that there are nearly 2.7 million teachers in
6
Indonesia, and about half of them are primary school teachers, this translates to a large number of
teachers that need to be upgraded or replaced.
Table 1: Number and percentage of teachers below required qualification before
and after the new teacher law
Junior Senior
Primary
secondary secondary
Minimum qualification starting in 4-year degree 4-year degree or
4-year degree
2005 or diploma diploma
% of teachers below min. qualification 82% 33% 13%
Minimum qualification before 2005 2-year diploma 3-year diploma 4-year degree
% of total teachers below min.
33% 15% 13%
qualification
While the Ministry of Education is carrying out a massive upgrading program
with recognition of prior experience for incumbent teachers, whether the law will be
successful in the longer run depends foremost on how it would influence the occupational
choice decision of college-educated workers. Until now, there has been no basis for
making such forecasts. In the absence of empirical analysis, we can only guess at the
impact of changes in teacher salaries on the composition of the teaching force.
Under a theoretical framework with a government-dominated market with a
government-set wage rate and demand for teachers, this paper explores how teacher
supply, particularly the resulting composition of the teaching force with low or high
qualification, would be determined by current and future public policies. Using 2001 to
2008 Indonesian Labor Force Survey (SAKERNAS) data, this paper further estimates the
potential effect of the latest teacher law, which gives high qualification teachers a
significant pay increase, on the composition of the Indonesian teaching force with
differentiated education backgrounds.
7
III. Theoretical framework: Aggregate teacher labor market
Following Zabalza, Turnbull, and Williams (1979), the labor market for teachers
can be thought of within a traditional supply and demand framework, with the additional
complication that the government is virtually the sole hirer of labor. The demand for
teachers can be determined by the number of children in the country of school age, and
the government's desired pupilteacher ratio. For a given such ratio, the demand for
teachers is therefore a constant, denoted by q in Figure 2. Under the reasonable
assumption that the supply of teachers with college education is a positive function of
average teacher earnings, an upward-sloping labor supply schedule can be drawn as S. In
a perfectly competitive market, a teacher wage of w would therefore clear this labor
market.
Figure 2: The labor market for teachers
S
w*
W
D
Q1 q* Q2
8
However, the teachers' labor market is of course not competitive, and the
government, in its role as (almost) exclusive purchaser of teaching labor, has other
considerations, prime among which is the level of expenditure on teachers' salaries in
total. For a given level of such expenditure, an inverse relationship can be plotted
between teachers' earnings and the number of teachers hired, labeled D in Figure 2. If the
government wants to raise the salaries of teachers, it can afford to hire fewer of them,
given a fixed budget.
Another complication added to the model is that teacher salaries do not adjust
freely, as a majority of the teachers are civil servants, and therefore in most cases, their
salaries follow a country's civil servant remuneration scale.
Figure 2 illustrates a general case of how an aggregated teacher labor market
would work. With a mostly fixed wage rate W, the government can afford to hire Q2
teachers at this salary level. However, the market supply of teachers with college
education at this salary level could be only Q1. Therefore the government can hire (Q2-
Q1) low qualification teachers. Q2 can be higher than what is the actual need for teachers
q*, as in this case, and therefore over-supply of total number of teachers, measured as
(Q2-q*), can co-exist with the shortage of teachers with college education, measured as
(q*-Q1). This seems to explain the situation in Indonesia. Among all teachers Q2, the
proportion of qualified teachers is thus (Q1/Q2), while that of under-qualified teachers is
(Q2-Q1)/Q2.
On the other hand, if the government wants to hire all needed teachers (q*) with
college education, the wage rate then needs to be set at w*. However (q*, w*) could be
above the fiscal capacity. This case is illustrated in Figure 3: the government would have
9
the choice of either hiring only q0 college-educated teachers, or lower the wage rate to W'
to meet the demand for teacher numbers but with lowered average qualifications, with
Q1' teachers with college education, and (q*-Q1') teachers under qualified.
Figure 3: Labor market for teachers: low public budget
S
w*
w'
W
D
Q1 q* Q2
q0 Q'1
The other scenario is that (q*, w*) is below the fiscal constraint line D, as
illustrated in Figure 4. In this case, increasing the teacher salary level to w* would
reduce the total excess of teachers, retain the right number of teachers with college
education, as well as save public resources. However, in reality, if this is the case, there
would usually be an upward pressure to further raise teacher salaries beyond the
necessary level, or hiring more teachers, until the total budget allocation is spent. These
can be illustrated by the move upward of teacher salary from w* to w'', or hiring q*'
teachers rather than q*.
10
Figure 4: Labor market for teachers: high public budget
S
w''
w*
W
D
Q1 q* q*' Q2
For public policies that aim at inducing changes in aggregated teacher profile by
changing teacher remuneration, key information is needed on the slope of S, and the level
of w* that can induce the desired level q*.
IV. Estimating teacher supply
The econometric model that this paper uses is based on Dolton (1990), which
modifies that of Zabalza et al. (1979) and Willis and Rosen (1979), and considers that
there are two possible outcomes to a college graduate's decision. The graduate can either
decide to become a teacher (a) or not (na). It is assumed that the earnings streams in
these two regimes may be parameterized by a simple geometric process. In this model,
decisions are considered from the perspective of life-cycle earnings.
11
Consider an individual chooses to enter teaching at time T then the present value
of his expected earnings stream is:
a
V a = W (t ) exp(- rt ) dt. (1)
T 1-
The maximum present value of earnings, chosen over alternative occupations is:
1 -
V na = Max W na (t ; s ) exp(- rt ) dt , (2)
T
s
Where:
W (t ) is earnings at time t;
is "propensity to teach";
s is the choice set of other occupations.
We further define the earnings profile for teachers from time T and onwards is:
[
W a (t ) = WTa exp g a (t - T ) ] if T < t . (3)
Similarly, define earnings profile for non-teachers as:
[ ]
W na (t , s ) = WTna (s ) exp g na (s )(t - T ) if T < t , (4)
where:
WT is the earnings at period T, and g is earnings growth rate.
Individual chooses not to go to teaching if Vna>Va. Defining I=ln(Vna/Va), and
plugging equation (3) and (4) into (1) and (2) would give:
1-
I 1i = ln + ln WT - ln 1 - - ln WT - ln r - g + ln r - g .
na
a na a
( ) ( ) (5)
A linear approximation of equation (5) can be written as:
12
( )
I = 0 + 1 ln WTna - ln WTa + 2 g a + 3 g na + 4 + X , (6)
which is able to be estimated empirically.
This teacher supply framework has various drawbacks when applied to empirical
work. One obvious limitation is that earnings data are rarely available across any
appreciable time span in the life cycle. This poses problems for meaningful econometric
analysis. This means that little can be said about life cycle earnings without some
assumption concerning earnings growth. In addition, even though panel data can provide
a few observations per individual at different points of time, the existing panel data sets
in Indonesia (i.e. three waves of "Indonesia Family Life Survey") do not have large
enough samples for teachers, particularly of teachers with college education. In this
paper, we use labor force survey data for a few consecutive years and control for year-
specific fixed-effects in earnings estimation for teachers and non-teachers.
The second limitation of the model is that the non-pecuniary rewards of teaching,
or the individual's "propensity to teach," cannot be directly measured. Females with
children may have higher propensity to become a teacher given that teaching is usually a
more flexible job, and has less wage penalty when temporary leave happens (Flyer and
Sherwin 1997). Dolton (1990) used the probability of having the first job as a teacher as
a proxy for propensity to teaching when estimating occupation change between teacher
and non-teacher at later stages of one's career. However, this requires panel or historical
data that is not easily available for the intended analysis. Even under this approach, equal
non-pecuniary rewards to jobs were assumed for prior entry into any job any individual
has. In this paper we include variables such as dummy for female, marital status,
13
household size in an attempt to capture the variations in individual's preference for
teaching.
The third key empirical difficulty is obviously the problem of how to estimate the
foregone earnings in other occupations that influence an individual's occupational
decision, as one can only observe the earnings in the occupation that has been chosen. To
solve this issue, we follow Dolton (1990)'s empirical approach involving 3-stages of
estimation.
We need to estimate two earnings functions for teachers and non-teachers:
ln W a = X1 + u1 , and (7)
ln W na = X 2 + u 2 . (8)
Obviously, people are not randomly selected into teachers and non-teachers. OLS
estimates would be biased. Sample selection bias can be corrected by starting with
estimating a probit model of being a teacher:
I (a ) = Z + , (9)
where I equals 1 if an individual is a teacher, and 0 otherwise. Z includes all exogenous
variables.
The second stage estimates the log earnings function for teachers and non-
teachers (eq. 7 and 8), by inserting Mill's ratio ( ) on the right-hand side. For teacher
and non-teacher sub-samples, the consistent estimates can be obtained by OLS estimates
of the following:
ln W a = X1 + 1 11 + 1 , and
ln W na = X 2 + 2 2 2 + 2 .
where
14
(Z / )
1 = - ,
(Z / )
(Z / )
2 = ,
1 - (Z / )
1 = var(u1 ) ,
2 = var(u 2 ) ,
1 = corr ( ,u1 ) ,
2 = corr ( ,u 2 ) .
These first and second stages can also be estimated using maximum likelihood
based on conditional distributions of earnings for teachers and non-teachers. The
likelihood function of teacher earnings is
(ln W a - X1 )
f (u 1 | Z + > 0 ) = f (u1 | > - Z ) = f (u1 | < Z ) = .
(Z )
The likelihood function for non-teacher earnings is:
(ln W na - X 2 )
f (u 2 | Z + < 0 ) = f (u 2 | < - Z ) =
1 - (Z )
With consistent estimates of 1 and 2 , we use predicted values of Wa and Wna
for each individual to estimate the structural model of the following:
( )
I = 0 + 1 ln WTna - ln WTa + 2 Z +
^ ^ (10)
V. Data and results
We use the 2001-2008 Indonesia Labor Force Survey (SAKERNAS) to look into
the occupational choice of working cohorts with college education. Our sample includes
40,019 workers with college education from year 2001 to 2008. Overall, only around 3%
15
of the 20-year old and above population are able to attain this level of education. The
proportion increases overtime, but at very slow pace between 2001 and 2008 (Table 2).
Table 2 also shows that teaching job is a prominent choice for college graduates.
Between 2001 and 2008, around one-fifth to a quarter of the college graduates are
teachers.
Table 2: Composition of labor force with college education
% with college
education
non- among population
Year teacher teacher Total age 20 and above
2001 2,414 333 2,747 3.0%
2002 3,729 687 4,416 2.7%
2003 4,750 846 5,596 3.1%
2004 4,979 982 5,961 3.3%
2005 3,411 911 4,322 2.8%
2006 4,363 972 5,335 3.2%
2007 4,683 954 5,637 3.3%
2008 5,160 845 6,005 3.6%
Total 33,489 6,530 40,019 3.1%
Figure 5 depicts the trend of relative earnings for teacher and non-teacher college
graduates by age group. Earnings of teachers have been below that of non-teachers in for
the past few years. However, the real earnings gap is narrowing. Teacher's real earnings
growth has been faster than that of non-teachers in recent years. A closer look reveals
that teacher's real earnings has been mostly constant over the years, while it is non-
teacher's earnings that has actually been eroded by inflation over time.
16
Figure 5: Log real earnings of teachers and non-teachers with college
education in Indonesia, by age group, 2002-2008
2002 2004
14.5
14
13.5
13
2006 2008
14.5
14
13.5
13
20-29 30-39 40-49 50-5920-29 30-39 40-49 50-59
age group
log real earnings of teachers log real earnings of non-teachers
Graphs by year
Figure 6 shows the trend of relative earnings (ratio) of teachers, and the share of
labor force with college education that are teachers. Between 2001 and 2005, the share
of college graduates shows an growing trend, reaching over a quarter of total workers
with college education. Between 2005 and 2008, this trend seems to experience a reverse.
By 2008, the share of college graduates on the labor market and who are teaching came
back down to 19 percent.
For college graduates, even though the premium of non-teaching job is eroding
over time, there were periods that show a reversed trend, such as between 2005 and 2008.
17
The earnings ratio of non-teachers to teachers increased slightly from 1.1 to 1.3. More of
interest is the negative correlation between the pay-off of non-teaching job and the share
of college graduates in teaching (Figure 7). The correlation is -0.83, and highly
significant (P=0.01).
Figure 6: Trend of relative earnings and share of college graduates in
teaching, 2001-2008
1.7 0.28
0.26
% college graduates in
1.5
0.24
1.3 0.22
teaching
Wnt/Wt
0.2
1.1
0.18
0.9 0.16
0.14
0.7
0.12
0.5 0.1
2001 2002 2003 2004 2005 2006 2007 2008
Wnt/Wt % of college graduates in teaching
Source: SAKERNAS.
Figure 7: Relative earnings vs. share of college graduates in teaching, 2001-2008
0.28
% of college grad in teaching
0.26
0.24
0.22
0.2
0.18
0.16
0.14
0.12
0.1
1 1.1 1.2 1.3 1.4 1.5 1.6
Wnt/Wt
18
Now we turn to individual level data and estimate the effect of earnings
differentials on the occupation choice of teachers versus non teachers, using the empirical
framework laid out in the previous section. Table 3 is the descriptive sample statistics.
Table 3: Sample summary statistics:
Variable Standard
Definition Mean
name error
Dependant
variables:
Dteacher 0-1 dummy variable, =1 if individual is a teacher 0.163 0.370
Earnings Monthly earnings in Rupiah (Rp) 1,262,190 2,038,909
lnW Log earnings, =ln(Earnings) 14.106 0.757
Explanator
y variables:
Age age 37.173 10.183
Dfemale 0-1 dummy variable, =1 if individual is female 0.396 0.489
Drural 0-1 dummy variable, =1 if individual lives in rural area 0.153 0.360
Dmarried 0-1 dummy variable, =1 if individual is married 0.725 0.447
D2001 0-1 dummy variable, =1 if individual is from 2001 sample 0.069 0.253
D2002 0-1 dummy variable, =1 if individual is from 2002 sample 0.110 0.313
D2003 0-1 dummy variable, =1 if individual is from 2003 sample 0.140 0.347
D2004 0-1 dummy variable, =1 if individual is from 2004 sample 0.149 0.356
D2005 0-1 dummy variable, =1 if individual is from 2005 sample 0.108 0.310
D2006 0-1 dummy variable, =1 if individual is from 2006 sample 0.133 0.340
D2007 0-1 dummy variable, =1 if individual is from 2007 sample 0.141 0.348
D2008 0-1 dummy variable, =1 if individual is from 2008 sample 0.150 0.357
0-1 dummy variable, =1 if individual lives in Nanggroe Aceh
D11 Darusalam 0.026 0.158
D12 0-1 dummy variable, =1 if individual lives in Sumatera Utara 0.037 0.190
D13 0-1 dummy variable, =1 if individual lives in Sumatera barat 0.028 0.165
D14 0-1 dummy variable, =1 if individual lives in Riau 0.017 0.128
D15 0-1 dummy variable, =1 if individual lives in Jambi 0.011 0.103
D16 0-1 dummy variable, =1 if individual lives in Sumatera selatan 0.017 0.131
D17 0-1 dummy variable, =1 if individual lives in Bengkulu 0.014 0.116
D18 0-1 dummy variable, =1 if individual lives in Lampung 0.014 0.116
D19 0-1 dummy variable, =1 if individual lives in Bangka belitung 0.005 0.070
D21 0-1 dummy variable, =1 if individual lives in Kepulauan Riau 0.006 0.079
D31 0-1 dummy variable, =1 if individual lives in DKI Jakarta 0.228 0.419
D32 0-1 dummy variable, =1 if individual lives in Jawa Barat 0.073 0.261
D33 0-1 dummy variable, =1 if individual lives in Jawa Tengah 0.066 0.248
D34 0-1 dummy variable, =1 if individual lives in DI Yogyakarta 0.060 0.237
D35 0-1 dummy variable, =1 if individual lives in Jawa Timur 0.113 0.317
19
Variable Standard
Definition Mean
name error
D36 0-1 dummy variable, =1 if individual lives in Banten 0.024 0.152
D51 0-1 dummy variable, =1 if individual lives in Bali 0.039 0.193
0-1 dummy variable, =1 if individual lives in Nusa Tenggara
D52 Barat 0.020 0.140
0-1 dummy variable, =1 if individual lives in Nusa Tenggara
D53 Timur 0.014 0.119
0-1 dummy variable, =1 if individual lives in Kalimantan
D61 Barat 0.015 0.120
0-1 dummy variable, =1 if individual lives in kalimantan
D62 Tengah 0.011 0.105
0-1 dummy variable, =1 if individual lives in kalimantan
D63 Selatan 0.019 0.138
0-1 dummy variable, =1 if individual lives in Kalimantan
D64 Timur 0.017 0.130
D71 0-1 dummy variable, =1 if individual lives in Sulawesi Utara 0.015 0.121
D72 0-1 dummy variable, =1 if individual lives in Sulawesi tengah 0.015 0.120
D73 0-1 dummy variable, =1 if individual lives in Sulawesi Selatan 0.035 0.184
0-1 dummy variable, =1 if individual lives in Sulawesi
D74 tenggara 0.017 0.127
D75 0-1 dummy variable, =1 if individual lives in Gorontalo 0.007 0.082
D76 0-1 dummy variable, =1 if individual lives in Sulawesi barat 0.003 0.058
D81 0-1 dummy variable, =1 if individual lives in Maluku 0.013 0.112
D82 0-1 dummy variable, =1 if individual lives in Maluku Utara 0.007 0.081
D91 0-1 dummy variable, =1 if individual lives in Papua barat 0.004 0.065
D94 0-1 dummy variable, =1 if individual lives in Papua' 0.011 0.106
The sample includes the labor force with college education between 2001 and
2008. There are 40,019 observations in our sample, among which about 40 percent are
female, and 16.3 percent are primary or secondary school teachers.
The results of the maximum likelihood estimates for college-educated teacher and
non-teacher log earnings functions, corrected for self-selection, are presented in Table 4.
The result shows that for teachers, controlled for age, there are no significant earnings
differentials between urban and rural areas, or between male and female teachers. As
mentioned in the previous section, this may reflect the fact that in Indonesia, a majority
of teachers are civil servants, and follow standard pay scales.
20
Table 4: Maximum likelihood estimates of Log Earnings function for
teachers and non-teachers
Teachers log earnings Non-teachers log earnings
LnWt LnWnt
Independent Standard Standard
Coefficient Coefficient
variable error error
Constant 9.516 *** (0.141) 12.114 *** (0.071)
Age 0.153 *** (0.006) 0.078 *** (0.003)
2
Age -0.001 *** (0.000) -0.001 *** (0.000)
Dfemale -0.010 (0.016) -0.023 * (0.012)
Drural 0.024 (0.021) -0.119 *** (0.017)
D2002 0.021 (0.039) 0.066 ** (0.025)
D2003 0.225 *** (0.037) 0.265 *** (0.023)
D2004 0.162 *** (0.037) 0.331 *** (0.023)
D2005 0.229 *** (0.037) 0.291 *** (0.025)
D2006 0.300 *** (0.037) 0.416 *** (0.024)
D2007 0.368 *** (0.037) 0.511 *** (0.024)
D2008 0.405 *** (0.037) 0.538 *** (0.024)
Censored obs = 33,489 Censored obs = 6,530
Uncensored Uncensored
obs = 6,530 obs = 20,934
Wald 2 (11) = 3,228.82 Wald 2 (11) = 3,937.5
Prob > 2
= 0.000 Prob > 2
= 0.000
*** p<0.01, ** p<0.05, * p<0.1.
For workers in non-teaching jobs, however, there are significant earnings
differentials. Equally college-educated, a worker in rural areas earns 12 percent less.
Woman also earns less than man. Even though the earnings differential between man and
woman is small (2 percent), it is statistically significant.
21
In addition, the earnings growth rate, as measured by the coefficient estimates on
Age, appears to be higher for teachers than for non-teachers: 15 percent per year for
teachers as compared with 7.8 percent for non-teachers. However, a teacher's earnings
peak at age 54, while a non-teacher's at age 62. This is possibly due to the mandatory
retirement age at 55 for civil servant teachers.
Table 5 presents the structural and reduced form estimates of occupational choice,
equation (9) and (10) respectively. To avoid relying solely on the non-linearity of the
functional forms for identification, excluded variables need to be identified for the
earnings function, and the structural function of the occupational choice. The province
dummy variables are used for estimating the occupational choice function, but excluded
from the earnings function. The argument could be that localized demand for teachers,
due to various enrollment rates influence by local household socioeconomic conditions,
can affect the likelihood of entering teaching profession, but not teacher earnings. On the
other hand, the dummy variables for various years are included in the earnings function
estimation to capture overall labor market shifts in labor costs from year to year, but are
excluded from the structural estimates for occupational choice assuming the year-to-year
labor market change only affects individual's choice through changing earnings
differentials.
22
Table 5: Probit of choosing teaching by college graduates
Reduced form Structural form
Independent Marginal Standard Marginal Standard
variable effect error effect error
Age 0.002 *** (0.000)
lnWnt_lnWt -0.295 *** (0.013)
Dmarried* 0.054 *** (0.004) 0.023 *** (0.005)
Dfemale* 0.084 *** (0.004) 0.083 *** (0.004)
Drural* 0.175 *** (0.006) 0.117 *** (0.006)
D11* 0.023 * (0.013) 0.024 * (0.013)
D12* 0.001 (0.011) 0.000 (0.011)
D13* -0.014 (0.011) -0.015 (0.011)
D14* -0.014 (0.014) -0.013 (0.014)
D15* 0.046 ** (0.020) 0.047 ** (0.020)
D16* 0.008 (0.015) 0.008 (0.015)
D17* 0.043 ** (0.018) 0.044 *** (0.018)
D18* 0.030 * (0.017) 0.027 * (0.017)
D19* 0.047 * (0.029) 0.055 ** (0.030)
D21* 0.012 (0.025) 0.011 (0.025)
D31* -0.081 *** (0.006) -0.075 *** (0.006)
D33* 0.094 *** (0.012) 0.094 *** (0.012)
D34* 0.000 (0.010) 0.004 (0.010)
D35* 0.052 *** (0.009) 0.051 *** (0.009)
D36* 0.009 (0.014) 0.004 (0.013)
D51* -0.012 (0.010) -0.013 (0.010)
D52* 0.114 *** (0.018) 0.113 *** (0.018)
D53* 0.035 ** (0.017) 0.027 * (0.017)
D61* 0.022 (0.017) 0.024 (0.017)
D62* -0.021 (0.017) -0.020 (0.017)
D63* -0.013 (0.013) -0.008 (0.014)
D64* -0.012 (0.014) -0.012 (0.014)
D71* -0.030 ** (0.013) -0.033 ** (0.013)
D72* 0.020 (0.016) 0.022 (0.016)
D73* -0.018 (0.010) -0.016 (0.010)
D74* -0.005 (0.014) -0.002 (0.014)
D75* 0.008 (0.022) 0.007 (0.029)
D76* 0.016 (0.030) 0.007 (0.022)
D81* -0.006 (0.016) -0.008 (0.016)
D82* 0.013 (0.024) 0.011 (0.023)
D91* -0.039 (0.024) -0.050 * (0.022)
D94* -0.049 *** (0.014) -0.050 *** (0.014)
23
Reduced form Structural form
Independent Marginal Standard Marginal Standard
variable effect error effect error
D2002* 0.016 * (0.010)
D2003* 0.043 *** (0.010)
D2004* 0.042 *** (0.010)
D2005* 0.054 *** (0.011)
D2006* 0.023 ** (0.009)
D2007* 0.008 (0.009)
D2008* -0.019 ** (0.008)
Number of obs = 40,019 Number of obs = 40,019
LR: (43) = LR: (36)
2 2
3,103.83 = 3,329.76
Prob > Prob >
2 2
= 0 = 0
*** p<0.01, ** p<0.05, * p<0.10.
The primary focus for the occupational choice model is the coefficient on the
earnings differential variable (lnWnt - lnWt), holding constant the other explanatory
variables relating to various background and personal characteristics. A negative and
significant coefficient on earnings differential variable would indicate that a college-
educated worker is less likely to choose teaching as profession if other occupations pay
better. Our estimate result shows the "right" sign and a high significance level. Several
other clear effects are notable in predicting whether in teaching profession or not. Being
a woman or being married is significantly correlated with being in teaching force. It also
appears that the predominant job for a college-educated worker in rural area is teaching.
The likelihood of being a teacher is 11 percentage points higher for a college graduate in
a rural area than that in an urban area.
Based on the empirical results, double the salary of teachers with college
education would result in an increase in the probability of college graduates choice of
entering teaching force. The marginal effect would be:
24
dP dP
= = 0.295 * 0.5 0.15
Wnt Wnt Wnt
d ln d
Wt Wt Wt
This would increase the proportion of college graduates choosing teaching
profession to about one-third. Using the population projection by BPS (Bureau of
statistics), and assuming constant proportion of population above age twenty with college
education, this would lead to about 24-25 pupils per teacher with college education. This
level of pupil-teacher ratio is higher than the current level, but still falls within the
adequate range compared with other countries in the region.
Assuming this can be an acceptable pupil-teacher ratio such as q* in Figure 3, it
still requires increased amount of public resources. With increased pupil-teacher ratio
from 16:1 to 24:1, but doubled teacher remuneration, the per-pupil cost would increase
33 percent. Without this commitment, there will be either teacher shortage, or a mix of
teachers with high and low education background will continue to exist.
VI. Conclusions
Aiming at attracting high caliber human resources into teaching, the latest
Teacher Law (UU14/2005) in Indonesia promises a 100 percent teacher salary increase
for certified teachers with a minimum 4-year college education or above. In the long run,
whether the law will be successful in attracting the needed college-educated labor force
into teaching depends foremost on how it would influence their occupational choice
decision. Until now, there has been no basis for making such forecasts. The findings of
this paper provide some empirical foundations for the latest teacher law in Indonesia.
25
This paper has analyzed Indonesian teachers' labor supply under a theoretical
framework that is based on a government-dominated market with government-set wage
rate and demand for teachers. This framework could explain the phenomenon of the
overall over-supply of teachers but shortage of qualified teachers in Indonesia. The
results from the structural estimates constructed under the framework are particularly
useful for looking into the impact of the latest teacher law on the future education profile
of Indonesian teaching force.
Using a sample of workers with college education in the Indonesian Labor Force
Survey, this paper has found that the relative wage rate of teachers and that of the
alternative occupations significantly influence college educated workers' decision of
becoming teachers in Indonesia. The large-scale pay increase promised by the law for
teachers with college education will have a significant effect on attracting a college-
educated labor force to join the teaching force. It is estimated that the wage rate set in the
latest teacher law will be able to increase the share of teachers approximately from 16
percent to 30 percent of the college-educated labor force. In addition, the new
government-set wage rate can sustain a pupil-teacher ratio of 24-25 pupils per teacher
with college education, but will require a more than 30 percent increase in the teacher
salary bill.
Finally, we conclude by highlighting that attracting a high caliber labor force into
teaching is from the quality concern in the first place. Even though there is general
recognition that learning outcomes and education quality are influenced by the ability of
teachers who guide the learning process, whether qualification in terms of educational
attainment is an appropriate measure for teaching ability is debatable. Even though there
26
is established literature on the positive sorting of ability and educational attainment
(Willis and Rosen 1979), the ability to be a good teacher may also be different from
general academic ability. Furthermore, getting the right people to teach is only the vey
first step in improving educational quality and learning outcomes. How to make these
right people perform well and achieve results is an even a bigger challenge.
27
References
Court, Gillian, S. Morris, B. Reilly, and M. Williams. (1995). "Teacher's Recruitment
and the Labour Market". Institute for Employment Studies.
Csellack, J. (2002). "The Effects of Wages on Teacher Labor Supply". University of
North Carolina, Chapel Hill.
Darling-Hammond, L. (2000). "Solving the Dilemmas of Teacher Supply, Demand, and
Standards: How We Can Ensure a Competent, Caring and Qualified teacher for
Every Child". National Commission on Teaching and America's Future.
Dolton, Peter. (1990). "The Economics of UK Teacher Supply: The Graduate's Decision".
The Economic Journal (100): 91-104.
Dolton, Peter and Tsung-Ping Chung. (2004). "The Rate of Return to Teaching: How
Does it Compare to Other Graduate Jobs?". National Institute Economic Review
(190): 89-103.
Dolton, Peter. (2006). "Teacher Supply". In: Hanushek, E and Finis Welsh (Ed.),
Handbook of the Economics of Education. Elsevier B.V.:1081-1161.
Flyer, Fredrick and Sherwin Rosen. (1997). "The New Economics of Teachers and
Education". Journal of Labor Economics 15(1): 104-139.
Hanushek, Eric and R.R. Pace. (1995). "Who Chooses to Teach (and Why)?". Economics
of Education Review 14(2): 101-117.
Manski, F. Charles. (1985). "Academic Ability, Earnings, and the Decision to Become A
Teacher: Evidence from the National Longitudinal Study of the High School
Class of 1972". NBER Working Paper No. 1539.
Ortega, Daniel. (2006). "The Effect of Wage Compression and Alternative Labor Market
Opportunities on Teacher Quality in Venezuela". Working paper, The World
Bank.
Thomas, R.B. (1975). "The Supply of Graduates to School Teaching". British Journal of
Industrial Relations 13(1): 101-112.
Van der Klaauw, Wilbert. (1997). "The Supply and Early Careers of Teachers".
Department of Economics, New York University.
Willis, Robert and Sherwin Rosen. (1979). "Education and Self-selection". Journal of
Political Economy 87(5): 8-36.
28
The World Bank. (2007). "Teacher Employment and Deployment in Indonesia".
Zabalza, Antoni, Philip Turnbull and Gareth Williams. (1979). The Economics of
Teacher Supply. London: Cambridge University Press.
wb190969
C:\Documents and Settings\wb190969\My Documents\Research\Z_Teacher supply\0final data and
draft\Drafts\Chen.sg1_reviewed_final.doc
6/20/2009 7:37:00 AM
29