See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/46553790

Private Tutoring and Demand for Education in South Korea

Article in Economic Development and Cultural Change · January 2010

DOI: 10.1086/648186 · Source: RePEc

CITATIONS READS
213 5,386

2 authors, including:

Ju Ho Lee
KDI School of Public Policy and Management
43 PUBLICATIONS 899 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ju Ho Lee on 16 May 2014.

The user has requested enhancement of the downloaded file.

 Private Tutoring and Demand for Education in South Korea*
(Revised)

by

Sunwoong Kim
University of Wisconsin - Milwaukee
Milwaukee, WI, USA
+1-414-229-6924
[email protected]

and

Ju-Ho Lee
KDI School of Public Policy and Management
Seoul, Korea
+82-2-3299-1016
[email protected]

March 2002

* The authors would like to thank Keith Bender, Eric Hanushek, John Heywood, John
Riew, and anonymous referees for their helpful comments and suggestions. And we also
thank Kwang Jun Lee for his research assistance in the data work.
 Private Tutoring and Demand for Education in South Korea

Abstract

South Korea’s education sector has expanded dramatically as her economy grew
over the last several decades. In 2000, the government spent more than 3.5% of GDP on
primary and secondary schooling, comparable to the level in the other OECD countries.
Despite the substantial government expenditure, households additionally spent about the
same amount on private tutoring. We argue that the prevalent private tutoring is a market
response to the government’s rigid and uniform education policy. In order to achieve
rapid economic growth, the government pushed hard for universal primary schooling and
the equalization of secondary schools. Unsatisfied demand for education by parents and
students in a highly regulated educational environment has resulted in an enormous
increase in private tutoring despite government's strong policy measures to reduce it.

Key words: private tutoring, demand for education, South Korea.

 Private Tutoring and Demand for Education in South Korea

1. Introduction

Scholars and policy makers have grappled with the issue of the proper role of the

government in the education sector in developing countries.1 When many households are poor

and cannot afford schooling even when the private return in education is higher, Schultz (1961)

argues that investment in human capital through public schooling promotes economic growth.

Empirical evidence on this subject seems to corroborate his thesis.2 Moreover, many studies on

the returns to education (e.g., Psacharopoulos 1994), report that the returns are higher in low-

income countries and that they are higher in primary school level.3 Consequently, the prevailing

wisdom among education planners has been to provide universal public education (at least in the

primary school level) as quickly as possible.

However, as the governments in those countries typically have limited fiscal resources,

rapid expansion of public education often create severe fiscal stress in the public school system.

Consequently, the public school system in developing countries may result in low teacher

salaries, poor school facilities, and low quality education in general. In this environment,

households typically supplement their children’s education with private tutoring.

Private tutoring is widely practiced all over the world, though it is more common in Asia,

Africa, and Latin America.4 In some poor countries in which teachers are paid poorly and the

1
See Lott (1987) or Trosel (1996) for a review of various arguments for the public provision of schooling.
2
For example, by cross country regression on 35 East Asian countries, McMahon (1998) finds that more schooling,
particularly the expansion of secondary schooling, is an important determinant of the growth rates for the last thirty
years. Similarly, Birdsall et al. (1995) presents cross-country regressions showing that higher growth observed in
East Asia can be attributed to the investment in education. They argue that the supply of high-quality education
reduces income inequality, which in turn promotes growth.
3
However, McEwan (1999) argues that the rates of return in developing countries may have been overestimated
since the calculation typically ignores the educational costs paid by households directly.
4
In the monograph on private tutoring, Bray (1999) reports the staggering percentages of students receive private

1
government’s monitoring is not stringent, such as Indonesia, Lebanon, Nigeria, Cambodia, India

and Tanzania, private tutoring is the main source of the teachers’ extra income. Under this

situation, teachers may deliberately shirk in order to create the demand for private tutoring.

Students often take private tutoring from the very teacher after school hours, sometimes at the

same school facility.5 In other countries where teachers are paid relatively well and/or teachers

are prohibited to provide private tutoring, such as South Korea and Japan, private tutoring is

mostly provided by university students and by professional private tutors. Consequently, there

exists a well-developed market for private tutoring out of the formal school system.

Despite its importance, there have been not many studies on private tutoring. This paper

is an attempt to fill the void by exploiting rich datasets on private tutoring from South Korea

(Korea hereafter). We argue that the rampant private tutoring in Korea is a natural market

response to the increase in educational demands that is not satisfied by the uniform public

provision of schooling. In particular, the households with higher demand for education are more

likely to seek for private tutoring, as the gap between the demand and public supply is greater.

Also, the competition for a more prestigious university exacerbates the situation. As private

tutoring is a relatively expensive way to provide educational service, and it is only available to

households who are willing and able to pay for it, the hasty state-led expansion of secondary

education may result in an inequitable as well as inefficient school system.

The organization of the paper is as follows. In the next section, we describe the rapid

expansion of primary and secondary education led by government policies and mushrooming

tutoring around the world: nearly 70% of students had received tutoring by the time they finished high school in
Japan in 1993; over 50% of students received tutoring in Rio de Janeiro public schools in 1997; 74% of Grade 8 in
Egypt in 1991; 45% of primary school and 36% of middle school in Hong Kong; About 83% by the high school in
Malaysia in 1990; more than half of high school students in Morocco; 70% of Grade 6 in Tanzania; 81% of
secondary schools in Taiwan; and 61% of Grade 6 students in Zimbabwe.
5
Biswal (1999) analyzes the incentive structure and optimal educational system in this situation.

2
private tutoring in Korea for the last several decades. In section 3, we develop a theoretical

model in which private tutoring and public education are perfect substitutes. This model is used

to derive empirically testable hypotheses. In section 4, using the household survey data on

private tutoring practices, we estimate the empirical model of the demand for private tutoring.

Conclusions and policy implications are offered in the last section.

2. Expansion of primary and secondary schooling and private tutoring in Korea

It is well known that Korea has transformed herself from a backward agrarian economy

to a fledgling advanced economy during the last four decades by successfully implementing

government-led industrialization policies. During the same time, the government has put major

emphasis on the expansion of public education system (Adams and Gottlieb, 1993 and McGinn

et al., 1980). In 1945, when Korea was liberated from Japanese colonial rule, only 65% of

primary school aged children and less than 20% of the secondary school aged children were

enrolled in schools. Moreover, the Japanese teachers, who consist of 40% of primary school

teachers and 70% of secondary school teachers returned to Japan soon after the liberation, and

much of the school facilities were destroyed during the Korean War (1950-1953).

Despite the inadequate educational resources, the Korean government wanted to establish

universal primary school (Grades 1-6) education as soon as possible. The strong commitment

for the expansion of primary education by Rhee's government (1948-1960) has resulted in a

remarkable quantity expansion. Several teachers' colleges were established in order to produce

primary school teachers quickly. The government started an aggressive construction campaign

by building more than 5,000 classrooms per year starting in 1954. Consequently, primary school

enrollment increased from 1.37 million in 1945 to 2.27 million in 1947 to 4.94 million in 1965.

3
The number of teachers increased from 20,000 in 1945 to 79,000 in 1965. By 1965, the goal of

universal primary school education had been more or less achieved (see Table 1).6

Table 1. Enrollment Rates and Advancement Rates

When General Park took over the power by a bloodless coup d'état in 1961, he started

successful economic growth plans with strong government initiatives. The rising income and the

expansion of elementary school graduates created a strong surge for the demand for secondary

education. Up until then, each middle school and high school, regardless of pubic or private, was

allowed to choose students through competitive entrance examinations. Competition for better

schools became fierce as more students were graduating from elementary schools, and rankings

among middle schools and high schools had been well established. The situation was commonly

called as ipsi-jiok (literally, entrance examination hell). Education policy makers recognized

several problems of such fierce competition including rising private tutoring.7 The government's

answer to these problems was the school equalization policy.

The middle school (Grades 7-9) equalization policy was first implemented for Seoul (the

capital of and the largest city in Korea) in 1969, for major cities in 1970, and throughout the

6
However, in order to achieve the early universal primary school education, the government was forced to sacrifice
the quality of schooling. The average student teacher ratio for elementary school was over 60 during this period, an
d class sizes often exceeded 80. Also, the government asked parents to share a significant portion of educational exp
enses such as textbooks, supplies, activity fees and so on.
7
First, the heavy stress of preparing for the entrance exam hinders the healthy (physical as well as psychological)
growth of eleven-year old children. Second, schooling in the elementary schools, particularly in the grade 6, was
geared too much for the preparation of the exam. Therefore, teaching “normal” curriculum was difficult. Third, a
substantial amount of household expenditure was spent for private tutoring to prepare their children for the exam.
Fourth, the quality of middle school education differed very much from school to school, and students and their
parents were obsessed with the most prestigious schools, commonly known as, illyubyung (the disease obsessed
with the first class). Many students who failed to get in to the their desired schools repeated the sixth grade in order
to prepare for the next year's entrance exam. Also, many parents desired to send their children to an elementary
school that is more successful in sending its graduates to more prestigious middle schools. This created unbalanced
demand for elementary school student allocations across school districts.

4
country in the following year. The high school (Grades 10-12) equalization policy was first

adopted for Seoul and Pusan (the two largest cities in Korea) in 1974 and was gradually

expanded to several major cities until 1980. However, during the 1980s the government slowed

down the implementation because of growing opposition, and some small and medium sized

cities and rural school districts are allowed to keep the old system. Also, several types of high

schools for gifted children in arts, foreign languages, music, sports, and sciences had been

established through the country.8

The equalization policy replaced the individually administered entrance examinations

with a random allocation of students within separate school districts. Students were randomly

assigned to different schools in the district, regardless of public or private, by lottery as long as

the student passes a nation-wide qualification examination. Under the new policy, private

schools had to give up the rights to select new students and take all students assigned by the

Ministry of Education. It also made levels of tuition, salaries of teachers, and the curriculums of

private schools equal to those of the public schools through strong regulations and necessary

financial assistance to private schools. Accordingly, private schools became almost identical to

public schools in terms of the accessibility to the students, contents of the learning, and the

quality of teachers.

At the same time, the government maintained a quota for the number of university

admissions, and mandated strict student selection procedures for both public and private

8
The establishment of special high schools for foreign languages and for sciences in 1980s was clearly the single
most important deviation from the equalization policy, as these schools attracted most academically gifted students
who had to pass the entrance examination in order to be admitted. It is widely recognized that they are the best
preparatory schools that sent their graduates to the most prestigious universities. As the competition to enter these
schools was intensified, the government prohibited universities to take into account the quality differences among
high schools in their evaluation of students' high school records for the admission purpose. Consequently, many
bright students dropped out of the schools.

5
universities.9 The uniform student selection procedures imposed by the government created

inevitable consequences on secondary education, as the universities have well established

rankings and high school students wanted to enter more prestigious universities.10

Strong regulations and controls by the government have almost eliminated the

competition among secondary schools, but definitely not among students. On the contrary, they

produced runaway escalation in the expenditures on private tutoring. In 1980, the military junta

led by General Chun, who took the power through a military coup in a power vacuum created by

the unexpected assassination of President Park, took a very dramatic measure toward private

tutoring, by banning all forms of private tutoring except a few exceptions.11 However, such a

ban was difficult to enforce, and the private tutoring did not stop.

Even after the apparent failure of the outright ban on the private tutoring in 1980, the

government has maintained the strict legal regulations on private tutoring. Until the banning was

found unconstitutional in 1999, the government has allowed only two types of the private

tutoring: private instruction by university students and hakwon. The government has prohibited

all the other forms of the private tutoring including the private instruction by schoolteachers

outside the school, the private instruction by hakwon instructors outside hakwon, and the private

instruction through the mail, phones, and TVs. Moreover, the Korean government has

9
Due to political pressure, the government was forced to increase the number of university students in early 1990s.
Also, the government started to relax the regulations for admission procedures since 1995.
10
In line with equalization policy, government prohibited universities from weighing the differences in high school
quality in the student selection procedure. As the relative position of student in a school without any information on
the quality of the school does not convey much about the true academic capability of the student, universities rely
more heavily on entrance examinations in the student selection process.
11
There are three major forms of private tutoring catering for diverse market demand. First, individual tutoring for
one or a group of students is provided typically at student’s home. Second, instruction is provided by a private for-
profit learning institution, called hakwon. It has classrooms and instructors, and the instruction is given in a
classroom-like setting. Third, there are self-study sheets delivered and graded by mail or the internet. The
individual tutoring is the most expensive type, and the study sheet is the cheapest.

6
maintained strong controls over hakwon.12 Under this regulatory environment, the number of

hakwons has increased tremendously from 381 in 1980 to 14,043 in 2000, and the number of

students enrolled at hakwons has increased from 118,000 in 1980 to 1,388,000 in 2000.13

Table 2 shows the public and private expenditure in primary and secondary education in

Korea. The out-of-pocket payment by parents for the schooling of their children has been

gradually reduced from 1.6 % of GDP in 1977 to 0.8 % of GDP in 1998. However, the

escalating expenditure on private tutoring far more offset the reduction of out-of-pocket payment

for schooling.14 Since the education reform in 1995, there has been a drastic increase in

government spending on schools, reflected in the jump in government spending from 2.7% of

GDP in 1994 to 3.4% in 1998.15 However, escalating expenditure on private tutoring despite the

big increase in government spending indicates that the mere increase in public spending on

education might not be sufficient to reduce households’ expenditure on private tutoring.

Table 2. Primary and Secondary Education Expenditure to GDP

In 1997, more than 70 percent of elementary school students and about half of middle and

high school students are reported to take private tutoring. Many studies show that private

tutoring in Korea is not limited to the wealthy population, and wide spread across groups with

12
Those who want to establish a hakwon should acquire a permit from the government. Instructors at hakwon have
to have the certain required academic qualifications, and lecture rooms should satisfy certain physical requirements.
Also instruction fees should be kept below the guidelines set by the committee headed by the superintendents at the
local education authority.
13
The strong regulation on hakwon by government is a sharp contrast to the laissez-faire approaches of Japan on
juku (the Japanese counterpart to hakwon). The Japanese government has treated jukus like the other small
businesses, does not regulate them. Accordingly, compared to jukus that encompasses a variety of forms of private
instruction and meets the educational demands with flexibility, hakwon is a more narrowly defined form and is made
to be more like schools. In effect, the Korean government has been trying to confine the private tutoring to hakwon,
which is easily put under the tight control of the government.
14
It is highly probable that the expenditure on private tutoring had been underreported in early 1980s because of the

7
different incomes or consumption levels. KEDI (1999) showed that private tutoring expenses

were about nine percent of incomes of the households for all income groups except for the fifth

(highest income) group with the spending of 7.4 percent of their incomes on private tutoring.

Lee and Woo (1998) estimates that Korean households spent 12.4% of GDP per capita per

elementary student on private tutoring in 1997.16

3. Theoretical Model

In this section, we present a theoretical model that provides not only a theoretical

framework in which private tutoring can be discussed but also testable hypotheses that can be

estimated in the next section. Consider a household i’s maximization problem with respect to

education (e) and a numeraire good (x). Education can be obtained either by formal schooling

(publicly supported and publicly provided education), e1 or by privately paid private tutoring,

e2.17 We assume that formal schooling and private tutoring are perfectly substitutable.18 As the

education is supplied privately as well as publicly, the structure of the model is similar to the

model of public provision of private goods in which the household is allowed to supplement the

public provision with additional purchase in the private market.19 Since we are mainly interested

in the household’s demand for tutoring, our model will be a partial equilibrium model.

ban on most of the private tutoring.

15
See Park (2000) for a detail on the education reform of 1995.
16
UNESCO(2000) reports that Korea’s public expenditure per pupil as a percentage of GNP per capita is 16% for
pre-primary and primary education combined and 12% for secondary education in 1995.
17
As described in the previous section, formal schooling includes education provided by private schools as well as
by public schools, as private schools in Korea are effectively identical to public schools. Hence, the only possible
private education would be private tutoring.
18
When the major purpose of private tutoring is to prepare students for major examinations such as college entrance
examinations, private tutoring may emphasize test-taking skill while formal school system may focus on broader
educational objectives. Hence, it can be argued that private tutoring and formal schooling are imperfect substitutes.
Nonetheless, we take the perfect substitution assumption for analytical convenience.
19
When the private good is indivisible so that the household cannot supplement its consumption with the private
market purchase, Besley and Coate (1991) argued that universal provision of public good as an income
redistribution mechanism is not efficient, as both the rich and the poor would prefer direct income transfer.

8
 For the analytical simplicity, we assume that each household has only one child.20 The

child’s scholarly ability at the time is given by ai. In order to capture the heterogeneous

preferences over education across households, we parameterize the preference over education

with parameter θi. The higher the value of θi, the greater its preference for education is.21 We

also recognize the consumption externality of education. The household cares not only about the

amount of education that its child receives but the amount of education of all other children

receives, E-i.22 The tournament aspect of education can be manifested in the labor market.

Despite the fact that college premium in the Korean labor market shows a decreasing trend in

most of 1980s and 1990s as a result of rapid expansion of college education (Kim and Lee, 2000),

there are plentiful reasons for fierce competition to enter a few elite universities in Korea. Kim,

Lee, and Kim (2002) shows that graduates of top six universities consist of 59 percent in the list

of Who’s Who in four major newspapers and 85 percent of who passed exams for high-rank

public officials. Chang (2000) shows the pronounced wage premium of about 40 percent for the

graduates of top five universities in Korea. And 57 percent of members of National Assembly in

2000 turned out to be graduates from top three universities.

However, several authors argued that public provision of private goods is a second best Pareto efficient if there
exists information constraints in which the government cannot use the optimal nonlinear income tax or if there is a
time inconsistency problem (e.g., Nichols and Zeckhauser 1982; Boadway and Marchand 1995; and Blomquist and
Christiansen 1995). In the model in which the public provision can be supplemented with private purchase and the
level of public provision is determined by the majority voting and financed by proportional income tax, Epple and
Romano (1996) showed that total expenditure on the good is higher in the dual supply case than either in the private
supply only case or public supply only. Blomquist and Christiansen (1999) argued that if the government can
exercise a non-linear income tax system, efficient level of public provision can be obtained with majority voting or
two party political system under certain conditions.
20
Allowing more than one children for a household will create a question of how to allocate private tutoring
expenditure among children in addition to the question of how much to spend on private tutoring.
21
Specifically we assume that as θ increases, the marginal utility of education increases and the marginal utility of
other goods decreases.
22
As the level of the education of other children increases, the probability of the household’s child decreases.
Therefore we assume that the marginal utility of own education increases as the other’s education increases.
However, as there is no obvious relationship between the marginal utility of the numeraire good (x) and the other

9
 Since there are many households, we shall ignore the strategic interaction between the

households regarding the choice level of education, i.e., we shall treat the other children’s

education (E-i) exogeneous. The household wants to maximize its utility function ui (x, e; ai, θi,

E-i) subject to

(1) e = e1 + e2

(2) e1 = ē1 + α (ā – ai)

(3) x + p e2 = y - η ē1

(4) x > 0, e1 > 0, e2 > 0.

We assume in equation (1) that total educational services obtained by the household is the sum of

formal schooling (e1) and private tutoring (e2). The second equation says that the education

service received through schooling depends not only on level of education provided by the public

school (ē1) but also on the child's ability (ai) and his peer student. Specifically, there is a positive

peer effect to the student proportional to the difference between his ability and the average

ability (ā) of the children in the classroom whose ability is greater than his.23 In regions under

the equalization policy, where school choices are not allowed, average ability of children in the

classrooms can be regarded as exogeneous. The parameter α represents the degree of this peer

effect. The level of schooling ē1 is determined by the government, and it is provided free of

children’s education (E), we assume the cross partial derivative of the utility with respect to the two variables is zero.
23
We are assuming a type of “baseline model” of peer effects where peer effects have distributional consequences
but no efficiency consequences. Under the model, in order to give one student a better peer, one must take that peer
away from another student; the two effects exactly cancel and total societal achievement remains the same (Hoxby,
2000b).

10
charge.24 However, in order to provide schooling, the government collects the lump-sum tax of

the amount of η ē1. The parameter η (> 1) represents the efficiency of the formal school system.

The lower the value of η is, the more efficient the school system is. The price of x is normalized

to one, and the price of private tutoring is p. Private tutoring is available with continuous

amount e2.

Since the price of schooling is zero, every household will choose to attend schools as

long as the educational service the household receives (e1) is positive. If not, the household will

pull the student out of the formal school system. However, there are virtually no alternatives for

such students within Korea. Some households, therefore, send their children to other countries

for education.25 However, since we are going to examine households who are in Korea, we shall

assume that e1 is positive. Thus, the household's utility maximization problem is reduced to

choose x and e2 subject to the budget constraint (3). The Lagrangian function is written as:

(4) L = u(x, ē1 - α b + e2) + λ (y - x - p e2 - η ē1),

where b = (ai - ā).26 We have dropped the subscript i for convenience.

24
The household has to pay nominal amount of tuition for secondary schools in Korea. Since formal schooling is
much cheaper than private tutoring the assumption of free schooling is mostly for analytical convenience. Moreover,
the government recently announced that middle schooling education is free of charge starting in the 2002 school
year.
25
In August 2000, the government started to regulate the studying abroad by primary school and middle school
students. It is estimated that more than 13,000 primary and secondary school students are studying abroad in 1999,
and the number is growing rapidly (Joong-Ang Daily, August 4, 2000). The U.S., Canada, and Australia are the
most favorite destinations.
26
Under the assumption of the uniform distribution of the ability b = (ai – a*)/2, where a* is the highest level of
ability. In general, b will be positively related to ai. If the ability has a bell-shape distribution, the peer effect would
be greater for a high ability student than an average ability student.

11
 Assuming that the marginal utility of income (or the Lagrangian multiplier λ, associated

with the income constraint) and the consumption of x is positive, the Kuhn-Tucker conditions for

the problem are:27

(5) Lx = ux - λ = 0

(6) e2 Le = e2 (ue - λp) = 0, e2 > 0, Le < 0

(7) Lλ = y - x - p e2 - η ē1 = 0

The household may choose no private tutoring (e2 = 0). In this case, Le< 0. Therefore,

the marginal rate of substitution between education and other goods (ue/ux) is less than the price

ratio (p). On the other hand, the household may choose positive amount of private tutoring. In

this case, ue/ux = p. The idea of the separation between the no-tutoring vs. positive tutoring

households can be illustrated by the Figure 1. In the Figure, indifference curves for the two

households are drawn in the space of (x, e). As households are taxed by the amount of ηē1, the

disposable income is y- ηē1. Since ē1 is provided freely, every household will consume at least ē1.

However, some households will choose point A at which private tutoring is zero, whereas others

will choose point B with some positive amount of private tutoring. For the former households

whose preference for education is lower, the marginal rate of substitution between education and

all other goods is lower than the price ratio between the private tutoring and all other goods.

Alternatively, their marginal utility of education is lower than the price of private tutoring, but is

certainly higher than the price of school-provided education, which is zero. For the latter

27
The second order condition is easily satisfied if each of the marginal utility for x and e are diminishing as the level
of consumption for that good rises and if the cross marginal utility(uxe) is positive. These assumptions seem very
reasonable as education and the numeraire good would be substitutes.

12
households, interior solution obtains. That is to say, the marginal rate of substitution is equal to

the price ratio, or marginal utility of education is equal to the price of tutoring.

Figure 1: Household's Choice over Education and Other Goods

The comparative static results indicate the following predictions of the model.28 First, as

the income rises, the demand for private tutoring rises. As long as education is a normal good,

higher income households will have higher demand for education. As the public education

supply is fixed at the level of ē1, they will want to supplement their child’s education with private

tutoring. Second, the higher the child's ability, the greater the demand for private tutoring will be,

as the higher ability child will have the more negative externality due to the equalization policy.

Third, as the price of the private tutoring rises, the demand decreases.29 Fourth, as the level of

education collectively provided by the government rises, the demand for private tutoring

diminishes.30 Fifth, if the public school system is less efficient in providing education, private

28
For the maximization problem has the unique interior solution (e2 > 0), the determinant of the bordered Hessian
matrix (H) of the constrained maximization problem must be positive, that is to say,
u xx u xe − 1
|H| = u ex u ee − p = 2puxe - uee - p2 uxx > 0. Assuming this condition is satisfied, it is straightforward to perform
−1 −p 0
the comparative static exercise for parameters (y, b, α, ē1, p, η, θi, E-i). It can be shown that:
∂e 2 u xe − pu xx ∂e 2 αη(u xe − pu xx ) ∂e 2 bη(u xe − pu xx ) ∂e 2 −η(u xe − pu xx )
= > 0, = >0, = > 0, = < 0,
∂y H ∂b H ∂α H ∂ e1 H
∂e 2 −e 2 (u xe − pu xx ) − λ ∂e 2 − e1 (u xe − pu xx ) ∂e 2 −u xθ + pu eθ ∂e 2 pu
= < 0, = < 0, = > 0, = eE > 0.
∂p H ∂η H ∂θ i H ∂E −i H
29
However, it is not clear whether the total expenditure on tutoring decreases as well, since it will depend on the
price elasticity of private tutoring. If the private tutoring is price inelastic (elastic), higher price will result in higher
(lower) expenses.
30
As this model is for an individual household, and it is assumed that the individual tax contribution to the
improvement of the overall level of public school is ignored, the result follows from the income effect. That is, the
higher taxes for education reduces the household’s disposable income. Higher taxes on education are likely to
increase the level of public schooling and to decrease the level of tutoring. However, the amount of the reduction in
private tutoring depends on the efficiency of the school system.

13
tutoring will increase. Sixth, the higher the preference on education, the greater the demand for

private tutoring will be. Finally, as the education level of other children rises, demand for

private tutoring increases.

The model presented here suggests that the prevalent existence of private tutoring in

Korea may be the result of the various following factors. First, as income rises due to the

tremendous success of economic growth, the demand for education rises as well. However, the

high demand for education for some households is not be satisfied by the formal schooling since

the government uniformly controls formal schooling even in private schools.31 Second, the

equalization policy can be directly blamed for high demand for private tutoring. This demand

would be particularly acute for academically strong students who have good chances of being

admitted to prestigious universities. Under the equalization policy, they do not benefit from peer

group effect in the school they attend. If the school had admitted other academically strong

students, the students would have gotten better education and would have been prepared better

for the university entrance examinations. Third, the high demand for private tutoring may be due

to the ineffectiveness of the public school education provision, as the public education system is

completely insulated from the market forces and local parents’ demand. Finally, the tournament

aspect of entering into better universities brings about an even greater demand for private

tutoring. In short, private tutoring in Korea is the market response to the unsatisfied demand for

education under heavily regulated educational environment.

4. Data and Empirical Analysis

31
If the private schools were allowed to operate more independently, private tutoring would diminish as more
households with high education demand send their children to those schools.

14
 Based on the theoretical discussion in the previous section, we shall estimate the

following censored regression model:

(8) yi = Xiβ + εi, if Xiβ + εi > 0

= 0, if Xiβ + εi < 0,

where yi is private tutoring expenditure for household i, Xi is a vector of independent

variables that affect the level of private tutoring such as student characteristics, household

characteristics and environmental variables, and εi is the error term.

The obvious choice is to estimate the equation (8) with Tobit estimator. However, as

Deaton (1997) pointed out, household consumption survey data is likely to be heteroskedastic

across income level. Moreover, Tobit estimator is inconsistent when the error term is either

heteroskedastic or non-normal (Arabmazar and Schmidt, 1981, 1982). It is known that censored

least absolute deviation (CLAD) estimator by Powell(1984) is robust to heteroskedasticity and

consistent and asymptotically normal for a wide class of error distributions. Therefore, we shall

adopt CLAD estimator.32 However, for comparison, we also present results based on ordinary

least square (OLS) and Tobit estimator.

This paper utilizes two data sets. The first set is the Survey on Private Tutoring (SOPT)

conducted by Korea Institute for Consumer Protection in 1997. SOPT surveyed households with

at least one child who is attending elementary or secondary school or a repeater who is preparing

for the next year’s university entrance exam after graduating from high school. Besides the

general household characteristics, it has detailed information on each child including the

performance in school and the private tutoring expenditure for each child. In order to take

advantage of the detailed information on individual child, we transformed the household data of

32
A public domain STATA routine, written by Dean Jolliffe, Bohdan Krushelnytskyy, and Anastassia Semykina was

15
SOPT into the individual child data. However, it does not contain the information on the

location of the household so that we cannot identify whether the student is in the school district

in which equalization policy is adopted. The second data set we used is the 1998 Annual Urban

Household Expenditure Survey (UHES) by the National Statistical Office of Korea. Although

UHES contains the household location data, it does not contain many variables on individual

student. The SOPT data has 6,804 observations and UHES has 19,389 observations.33 Besides

the common independent variables, we are particularly interested in finding out whether students

with higher abilities spend more in private tutoring in the former data set and whether the

equalization policy decreases private tutoring as the policy makers have hoped in the latter

dataset.

The independent variables include student characteristics (academic achievement, gender,

level of attending school), household characteristics (household income, education level of

parents, home ownership, computer ownership, number of children in the household, whether

mother works outside of home, and father’s occupation), and environmental characteristics (size

of the city that the household resides, density of the neighborhood, and whether the school

district is under the equalization policy).

Appendix Table A1 shows the definitions, means, and standard deviations of the

variables in the two data sets. About 75 percent of households in both datasets have positive

private tutoring expenditure. The average expenditure on private tutoring amounts to 4.7 percent

of household income in UHES, and 5.8 percent for the households with positive expenditure. In

general, households with positive tutoring expenditure have higher socio-economic status than

those with no expenditure though the difference is not dramatic.

used in the estimation.

33
Although UHES has more observations, observations with no income data was dropped. Our sample includes

16
 The Table 3 includes all independent variables that are available in SOPT (Model 1). In

these tables, Model A uses OLS, Model B uses Tobit, and Model C and D use CLAD. As the

equalization policy mainly targets big cities, leaving small cities and rural areas allow to choose

between the equalization and no equalization policy, we run regressions not only on the whole

sample (Model A through C) but also on Seoul and five next largest cities (Metro Cities) 34 that

are all under equalization policy (Model D).

Table 3. Per Child Expenditure on Private Tutoring in 1997

Results by different estimators show similar pattern. For many variables, however,

coefficient estimates in Tobit are greater in absolute value and more significant than those in

OLS. Also, some coefficient estimates in CLAD models are substantially different from those

from OLS or Tobit, which indicates that the bias due to censoring, the non-normality and

heteroskedacity of the error term may be non-trivial. Overall, our findings in empirical models

from 1A to 1D strongly support the major predictions of the theoretical model.

First, as predicted by the theoretical model, a household with higher income spends more

on private tutoring after controlling for other variables regardless of different specifications. The

estimated income elasticity is around .5 for the average household. Second, students with the

highest academic performance (upper 10 percentile) spend significantly more on private tutoring

compared to students with around the average academic performance (between 30 to 70

percentile) after controlling for other variables regardless of different specifications. Furthermore,

the spending on private tutoring increases in proportion to the students’ academic performance;

only the household that has children in kindergarten, primary schools, middle school, or high schools.
34
Metro Cities are large cities (between 2 to 5 million in populations) that are authorized to have autonomous

17
highest spending for students with the highest academic performance, the second higher

spending for the students with the next the academic performance ladder (10 to 30 percentile),

and so forth. Moreover, we found larger coefficient estimates on upper 10 percentile students in

Seoul and metropolitan cities (Model 1D) compared to results over all regions (Model 1C),

though the there is no difference for the students in 10-30 percentile. The above findings are

consistent with the predictions of our theoretic model that the demand for private tutoring is

higher for students with higher ability and this tendency becomes more acute in the region under

the equalization policy. It should be noted that our data on the academic performance of students

is based on the survey to their parents and is significantly skewed toward the higher performance

(Table A1). The resulting paucity of observations on the students with lower academic

performance might lead to the less significant coefficient estimates of dummy variables of lower

academic performance students. However, the result that the student in higher academic

performance might be due to the possibility that more private tutoring improves academic

standing.

Third, households with higher preference for education spend more on private tutoring.

Coefficient estimates on most of the variables concerning preference for education turn out to be

significant with an anticipated sign. Years of educational attainment of parents significantly

increase private tutoring expenditure. Also, households that own houses, maybe due to the

higher level of wealth, spend more on private tutoring. The household with larger number of

children is likely to spend less on private tutoring per child. This is consistent with the

development literature stressing that a smaller size of family increases per capita spending on

human capital accumulation. Father’s occupation does not turn out to be significant in all

specifications. However, we find some evidence that when mother works out of home, the

governments, equivalent to provincial governments.

18
household spends less on private tutoring. This result may be due to the possibility that wife is

likely to stay home when the household has high preference over education.

We also found the evidence consistent with the idea that the desire to out-compete others

in the university entrance exam increases the spending on private tutoring. Those living in a

high density residential development area are found to spend more on private tutoring probably

because their intimacy with neighbors might provide higher competitive pressures to spend more

on private tutoring for their children. We can also interpret coefficient estimates on dummy

variables of the levels of schools that a child is attending (middle school, high school, and

repeater, compared to elementary school) in a similar line. As a child advances from elementary

school to middle school and to high school, and approaches university entrance exam, the

demand for private tutoring could increase because the latest academic performance weighs more

in university entrance exam. However, in the Korean education system, high school education is

divided into two tracks: academic and vocational. If a middle school student advanced to a

vocational school, the demand for private tutoring for the student is significantly reduced

because of the little chance to take an entrance exam for 4-year universities. Accordingly, we

found large and significant coefficient estimates for dummy variables for middle school students

and negative and significant coefficients for high school students.35 On the other hand, repeaters

are found to spend the highest amount on private tutoring because of the strong need to show

better performance in entrance exam and also because of the lack of proper formal educational

institutions, like community colleges in U.S.A., to accommodate those who failed the university

entrance exam.

35
Unfortunately, we do not have data regarding whether the high school student is attending a vocational school or a
college preparatory school.

19
 Interestingly, education of the mother is found to affect private tutoring expenditure more

strongly than that of the father, though this effect seems to disappear in the large cities sub-

sample (Models 1D). This appears to be consistent with the typical division of labor within

family, still prevalent in Asian countries including Korea, where father works outside home for

income and mother takes care of family including education for children. Also, households

with a computer are found to spend more on private tutoring, suggesting that computer cannot be

a good substitute for private tutoring.36 It is also interesting but not surprising to find that the

gender dummy indicates more private tutoring spending for female students. We suspect it is

due to the fact that the female students are more encouraged to take lessons in music and arts,

which tend to be more expensive.

We need further discussion on the reasons why spending on private tutoring is higher in

larger cities compared to rural areas, after controlling for other factors. One possible answer in

line with our theoretical model is that larger cities are under the stricter control of the state

because the major target areas of the equalization policy is Seoul and the other metropolitan

cities, allowing about half of the medium and small cities and all the rural areas uncovered by

equalization policy. Another explanation could be that the cultural aspect or the life style of the

big city could be related to higher concerns of parents on the relative position of the their

children in the competition to enter better universities.

In addition to the problems with unobservables, there may be a sample selection problem.

Many families devoted to educate their children may from rural area to urban area for the better

education of their children. Therefore, rural areas could be left with families that have lower

preference on education. Likewise, one might move to Seoul to find better private tutoring for

36
Again, this is consistent with an alternative view that the household with higher preference over education is more
likely to buy a home computer.

20
their children, which provides potential explanation for the positive coefficient estimates on the

dummy variables of Seoul and metropolitan areas. A higher cost of living in big cities could

attribute to a larger expense as well.

Some of the variables included in the models reported in Table 3 may be endogeneously

determined with the level of private tutoring. In particular, the student’s school ranking may be

influenced by the amount of private tutoring, that is, a student who spends more on private

tutoring may have a higher academic achievement. Also, the decision to allow a child to repeat

for the next year’s college entrance examination would be simultaneously determined by the

level of private tutoring expenditure for the student. The decision to work outside of home may

be correlated with the household’s preference over education and the level of private tutoring.

The decision to purchase a computer and the number of children in the household maybe

endogenously determined as well. In Table 4, we report the results of the regression excluding

the variables that are possibly endogeneously determined. The coefficient estimates are very

similar to those reported in Table 3.

Table 4. Per Child Expenditure on Private Tutoring in 1997

(Models without possible endogeneous variables)

Next, we turn to the estimation on per household expenditure on private tutoring using

UHES data. The focus of this empirical work is to see whether the equalization policy leads to

lower spending on private tutoring as many education policy makers hoped. In the regions not

covered by the equalization policy (school choice regions), high schools (private or public) can

21
choose students, and students can also choose high schools.37 However, the selection process

must be based on region-wide entrance examinations. Accordingly, high schools in these

regions have well-established rankings in terms of the minimum test scores of the incoming

students. Therefore, students with similar academic capabilities sort themselves and end up with

the same high school as a result of school choices based on the test.

It should be emphasized that high schools in these regions do not have full-fledged

autonomy. Strict regulation over curriculum, textbooks, tuitions, teachers, principals, etc. applies

to both public and private schools regardless of whether the school is covered by equalization

policy or not. Though there are differences among high schools in these regions, the differences

are relatively small as the government tried to equalize virtually all school inputs. The most

important difference between schools covered and uncovered by equalization policy is the peer

group effect due to sorting and competition among students.38 Self-sorting of students according

to test scores (and academic ability) might enable schools to cater to the demands of parents and

students more easily. Also, higher competitive pressures among high schools, particularly to

attract better students, in the regions uncovered by the equalization policy might lead to higher

productivity of schools. Although our theory predicts that higher productivity of school system

reduces private tutoring, it is purely an empirical question whether the potential increase in the

productivity of schooling due to the school choices in the regions uncovered by the equalization

policy might reduce the demand for private tutoring.

One popular argument is that in school choice regions one would expect more

competition among middle school students or even among elementary students to enter better

37
As was indicated earlier, middle schools (Grades 7-9) are fully equalized throughout the nation.
38
Hoxby (2000a) examines the effects of the greater Tiebout choice on the productivity of public schools in U.S. In
Korea, although regional educational authorities in Korea have very limited financial independence, public schools
in regions uncovered by equalization policies enjoy the same extent of school choices as private schools.

22
high schools, which will fuel the competition for more spending on private tutoring. However,

increase in private tutoring in middle school or even in elementary school to enter better high

school may be outweighed by decrease in private tutoring in high school due to the higher

productivity of high schools and/or improved peer group effect.

Therefore, we ran OLS, Tobit, and CLAD on the samples of households with at least one

student in UHES data. In order to test the argument that school choice in high schools may

decrease private tutoring for high school students but increase for elementary and middle school

students, we added interaction variables between school choice and number of children in

various schools. Our findings are reported in Table 5. We found that the results are quite

consistent with the major predictions of the theoretical model and the previous empirical findings.

Table 5. Per Household Expenditure on Private Tutoring in 1998

First, we found that households with higher incomes spend more on private tutoring after

controlling for other variables regardless of different specifications. Moreover, the estimated

coefficients are comparable to those reported in Tables 3 and 4, again suggesting that income

elasticity of private tutoring is around 0.5. Years of educational attainments of the head of

households again exert a significant effect on the spending on private tutoring, and the estimated

coefficients are comparable to those in the previous tables. Households with own houses spend

more on private tutoring, and again the estimates are comparable. Whether mother works

outside of home is not statistically significant for all model specifications, and that result is

consistent with CLAD estimates in the previous tables. As was in the previous tables, the

number of children attending different levels of schooling increases the amount of private

tutoring. The private tutoring expenditure is substantially higher in Seoul compared to small and

23
medium sized cities in all model specifications. However, coefficient for Metro Cities are only

significant in CLAD. It is quite remarkable that all these results are robust to different

specifications with two completely different data sets.

Second, the results on the effect of school choice are not as strong as the other

independent variables described above. However, considering the possibility that the effect of

school choice in Korea could be limited because school is not different from each other due to

the lack of school autonomy, the results are consistent with the theoretical discussions above.

Though most of the estimated coefficients involving school choice are not statistically significant,

they have predicted signs.39 That is, under equalization policy there seems to be less private

tutoring for elementary and middle school students and more for high school students. Overall,

equalization policy seems to increase private tutoring. Certainly, it does not decrease private

tutoring expenditure as policy-makers hoped.

5. Conclusions

The prevalent practice of private tutoring in Korea can be traced to the paradigm of state-

led development policy that pursued rapid economic growth through industrialization and export

promotion. Following the universal primary school education, the military government has

equalized secondary education so that opportunities for secondary education greatly expanded.

Concerns over the excessive wasteful competition among students to enter better schools during

the period of rapid expansion of school system made the public more receptive to the

government’s policy of equalization policy toward secondary schools. However, the virtual

socialization of private secondary schools accompanied with a heavy-handed regulation on

39
Only exception is the interaction term between the school choice and the number of middle school students in
OLS and Tobit.

24
university admission procedures has resulted in an ever-increasing demand for private tutoring.

Currently, Koreans spend as much money on private tutoring as the government expenditure on

primary and secondary education.

The theory and empirical evidence provided in this paper strongly suggest that rampant

private tutoring is a market response to the under provision of public education and the heavy

regulation and strict controls of the government. It is predicted by our model and confirmed by

our empirical finding that students with high academic ability, high family income, and whose

parents are highly educated, spend more on private tutoring because their educational demands

are not properly met by the formal school system that is provided by the government. Also,

students in regions uncovered by equalization policy and therefore with school choices spend

less on private tutoring, indicating that private tutoring is related to the strict government

regulation on schools.

Under the current education system in which consumption externality in secondary

education is important because of the competition to enter a more prestigious university, private

tutoring practice is expected to flourish. However, it should be recognized that the rampant

private tutoring might be inefficient as well as inequitable. Given that financial market is not

perfect, poor parents who have high demands for private tutoring because of high ability of their

children might end up with lower private tutoring because they are not able to foot the tuition

bills for expensive private tutoring. That inequitable distribution of private tutoring would also

lead to inefficient allocation because marginal rate of substitution between education and a

numeraire good becomes not the same across individuals.40

40
One of possible policy to address the problem is to provide financial assistance for private tutoring to low income
groups. In 2000, one Education Minister actually tried to introduce the policy, which brought about strong
opposition by teachers’ unions and resulted in the resign of the Minister.

25
 Although many reform initiatives pushing for deregulation in the educational sector were

proposed by the government-sponsored Education Reform Committee of 1995, the

implementation of the proposal has been slow due to severe opposition by interest groups as well

as the inertia associated with the legacy of the developmental state. The Korean experience

clearly shows that the runaway escalation of expenditures on private tutoring is strongly related

to strict regulation and controls over schools put in place during the rapid expansion of school

system.

26
 References

Adams, Donald and Ester Gottlieb (1993), Education and Social changes in Korea, Garland

Publishing Company: New York, NY.

Arbamazar, Abbas, and Peter Schmidt (1981), “Further Evidence on the robustness of Tobit

Estimator to Heteroskedasticity,” Journal of Econometrics 17, 253-58.

________ (1982), “An Investigation of the Robustness of the Tobit Estimator to Non-normality,”

Econometrica 50, 1055-63.

Besley, Timothy and Stephen Coate (1991), “Public Provision of Private Goods and the

Redistribution of Income,” American Economic Review 81(4), 979-984.

Birdsall, Nancy, David Ross, and Richard Sabot (1995), “Inequality and Growth Reconsidered:

Lessons from East Asia,” The World Bank Economic Review 9(3), 481-482.

Biswal, Bagala P. (1999), “Private Tutoring and Public Corruption: A Cost-Effective Education

for Developing Countries,” Developing Economies 37(2), 222-240.

Blomquist, Sören and Vidar Christiansen (1999), “The Political Economy of Publicly Provided

Private Goods,” Journal of Public Economics 73, 31-54.

Blomquist, Sören and Vidar Christiansen (1995), “Public Provision of Private Goods as a

Redistributive Device in an Optimal Income Tax Model,” Scandinavian Journal of

Economics 97 (4), 547-567.

Boadway, R. and M. Marchand (1995), “The Use of Public Expenditures for Redistributive

Purposes,” Oxford Economic Papers 47(1), 45-59.

Bray, Mark (1999), The Shadow Education System: Private Tutoring and Its Implications for

Planners. International Institute for Educational Planning: Paris, France.

27
Chang, Su Myung (2002), “Daehak Gyoyuk Uie Gyungje Hak [Economics of Higher

Education],” Rodong Jungchaek Yunkoo [Journal of Labor Poliy] 2(1). [in Korean]

Deaton, Angus (1997), The Analysis of Household Surveys: A Microeconometric Approach to

Development Policy, Baltimore, MD: The Johns Hopkins University Press.

Epple, Dennis and Richard E. Romano (1996), “Public Provision of Private Goods,” Journal of

Political Economy 104(1), 57-84.

Frank, Robert (1996), “Consumption Externalities and the Financing of Social Services,” in

Victor R. Fuchs ed. Individual and Social Responsibility, 175-190, The University of

Chicago Press: Chicago, IL.

Hoxby, Caroline (2000a), “Does Competition among Public Schools Benefit Students and

Taxpayers?” American Economic Review 90(5), 1209-1238.

Hoxby, Caroline (2000b), “Peer Effects in the Classroom: Learnig From Gender and Race

Variations,” NBER Working Paper 7867.

Joong-Ang Daily, August 4, 2000, Seoul, Korea. (http://www.joins.com)

Kim, Dae-Il and Ju-Ho Lee (2000), “Changes in the Korean Labor Market and Future Prospect,”

in Kenneth Judd and Young Ki Lee (eds.), An Agenda for Economic Reform in Korea:

International Perspectives, Hoover Institution Press, 341-373.per 7867.

Kim, Sunwoong, Ju-Ho Lee, and Sung Bo Kim (2002), “Hierarchies and Competition among

Korean Universities,” mimeo (forthcoming).

Korea Education Development Institute [KEDI] (1999), Sakyoyukui Siltae Chosa [Survey on

Private Tutoring], KEDI: Seoul, Korea. [in Korean]

Korean Ministry of Education (1998), Kyoyuk 50 Nyunsa [Fifty Years of History of Korean

Education]. [in Korean]

28
________, Statistical Yearbook of Education, various years, Seoul, Korea.

(http://www.moe.go.kr)

Lee, Ju-Ho and Cheonsik Woo (1998), “Hankoog Kyoyukui Silpaewa Kyehyuk [Failures and

Reforms in Korean Education],” KDI Jungchaek Yunkoo [KDI Journal of Public Policy]

20(1). [in Korean]

Lee, Maw-Lin, Ben-Chieh Liu, and Ping Wang (1994), “Education, Human Capital

Enhancement and Economic Development: Comparison Between Korea and Taiwan,”

Economics of Education Review 13(4), 275-288.

Lott, John R., Jr. (1987), “Why Is Education Publicly Provided? A Critical Survey,” Cato Journal

7(2), 475-501.

McEwan, Patrick J. (1999), “Private Costs and the Rate of Return to Primary Education,”

Applied Economics Letters 6(11), 759-80.

McGinn, Noel F., Donald R. Snodgrass, Yung-Bong Kim, Shin-Bok Kim, and Quee-Young Kim

(1980), Education and Development in Korea, Harvard University Press: Cambridge, MA.

McMahon, Walter (1998), “Education and Growth in East Asia,” Economics of Education

Review 17(2), 159-172.

Nichols, A.L. and R. J. Zeckhauser (1982), “Targeting Transfers through Restricts on

Recipients,” American Economic Review: Papers and Proceedings 72(1), 372-377.

Park, Kang H. (1996), “Education Expansion and Educational Inequality on Income

Distribution,” Economics of Education Review 15(1), 51-58.

Park, Se-Il (2000), Managing Education Reform: Lessons from the Korean Experience, 1995-97,

Korea Development Institute, Seoul.

Powell, James (1994), “Least Absolute Deviations Estimation for the Censored Regression,”

29
 Econometrica 54, 303-25.

Psacharopoulos, George (1994), “Returns to Investment in Education: A Global Update,” World

Development 22(9), 1325-1343.

Schultz, T. W. (1961), “Investment in Human Capital,” American Economic Review 51, 1-17.

Trostel, Philip A. (1996), “Should Education be Subsidized?” Public Finance Quarterly 24(1), 3-

24.

UNESCO (2000), World Education Report 2000: The Right to Education, Paris: UNESCO.

30
 Figure 1. Household's Choice over Education and Other Goods

all other
goods (x)

A
y- ηē1

ē1 ē1+e2 education (e)

31
 Table 1. Enrollment Rates (ER) and Advancement Rates (AR)

Kinder Elementary School Middle School High School

garten
ER ER AR ER AR ER AR- AR-
Academic Vocational
1953 - 59.6 - 21.1 - 12.4 - -
1955 - 77.4 44.8 1 30.9 1 64.6 1 17.8 - -
1960 - 86.2 39.7 2 33.3 2 73.3 2 19.9 - -
1965 - 91.6 45.4 3 39.4 3 75.1 3 27.0 - -
1970 1.3 100.7 66.1 51.2 70.1 28.1 40.2 9.6
1975 1.7 105.0 77.2 71.9 74.7 41.0 41.5 8.8
1980 4.1 102.9 95.8 95.1 84.5 63.5 39.2 11.4
1985 18.9 99.9 99.2 100.1 90.7 79.5 53.8 13.3
1990 31.6 101.7 99.8 98.2 95.7 88.0 47.2 8.3
1995 39.9 100.1 99.9 101.6 98.5 91.8 72.8 19.2
1999 37.3 98.6 99.9 98.8 99.4 97.3 84.5 38.5

Source: Data before 1970 are from McGinn (1980), and other data are from MOE (1998).

Notes:
ER = percentage of students enrolled out of corresponding school-aged children
AR = percentage of the students who advance to the next level school
1
1956-57
2
1959-60
3
1954-65

32
 Table 2. Primary and Secondary Education Expenditure in Percentage of GDP

Total Publicly Paid Privately Paid Privately Paid

on Schooling on Schooling on Private
Tutoring
1977 4.6 (100) 2.3 (50.5) 1.6 (34.4) 0.7 (15.1)
1982 4.8 (100) 2.7 (56.5) 1.7 (34.7) 0.4 (8.8)
1985 4.9 (100) 2.6 (53.2) 1.4 (28.5) 0.9 (18.3)
1990 4.7 (100) 2.5 (52.9) 1.0 (20.9) 1.2 (26.2)
1994 5.2 (100) 2.7 (52.1) 0.7 (14.4) 1.8 (33.6)
1998 7.1 (100) 3.4 (47.3) 0.8 (11.8) 2.9 (40.9)

Source: Survey on Educational Expenditures, KEDI, various years.

Note: The numbers in the parentheses are the ratios to the total expenditure.

33
 Table 3. Per Child Expenditure on Private Tutoring in 1997

Model 1A - OLS Model 1B - Tobit Model 1C - CLAD Model 1D - CLAD

(All Regions) (All Regions) (All Regions) (Seoul & Metro Cities)

-50836*** -139891*** -34793*** -18123

Constant
(-3.918) (-8.386) (-2.745) (-1.268)
0.02718*** 0.03372*** 0.02055*** 0.02291***
Household Income
(14.312) (14.3232) (8.526) (36.213)
12407*** 20392*** 16531*** 22122***
Upper 10%
(2.707) (3.550) (4.844) (5.620)
4968 11302* 10836*** 10196***
10-30%
(1.079) (1.943) (3.805) (2.772)
-16933 -40754*** -12129 -2875
70-90%
(-1.578) (-2.786) (-1.019) (-0.217)
-21983 -51573* -25687 -29036
Lower 10%
(-1.077) (-1.854) (-1.344) (-1.102)
5311*** 8001*** 3876*** 2607***
Mother Education
(5.058) (6.028) (4.415) (2.292)
1942** 3491*** 2310*** 2793**
Father Education
(2.064) (2.937) (3.189) (2.456)
-9082** -17105*** -12603*** -8735
Mother Working
(-2.218) (-3.147) (-3.937) (-1.598)
1732 -7038 -2265 3430
Father Professional
(0.266) (-0.873) (-0.391) (0.462)
8117* 11528** 11457** 11616***
Own House
(1.835) (2.066) (2.450) (3.333)
22196*** 28056*** 18391*** 18309***
High Density
(5.062) (5.147) (5.271) (4.087)
13052*** 19225*** 6483** 7226**
Computer
(3.249) (3.791) (2.112) (2.200)
-15264*** -26687*** -20439*** -16965***
No. Children
(-5.121) (-6.869) (-8.620) (-5.800)
3397*** 22429*** 30221*** 47093***
Middle School
(7.347) (3.864) (7.278) (9.242)
2422 -36555*** -55475*** -33338***
High School
(0.496) (-5.810) (-8.936) (-4.687)
164633*** 164683*** 171264*** 181705***
Repeater
(10.271) (8.456) (7.616) (7.914)
-9736** -12733*** -6889** -9725***
Gender
(-2.606) (-2.703) (-2.190) (-2.992)
48722*** 78926*** 47274*** 21965***
Seoul
(7.890) (10.075) (9.491) (6.242)
21460*** 48514*** 25092***
Metro City ---
(3.686) (6.502) (5.610)

9118* 31355*** 20060***

S&M City ---
(1.931) (4.244) (4.859)

Initial size 6718 Initial size 3317

Sample Size 6555 6555
Final size 5779 Final size 3162

Adj. R2 = 0.1625 Psd. R2 = .0109 Psd. R2 = .1271 Psd. R2 = .1224

Joint significance F(20, 6555) = 64.77 χ2 (20) = 1498.70

(p-value) (0.0000) (0.0000) --- ---

Notes: t-values are in parentheses below coefficient estimates

*Statistically significant at the 10% level; **at the 5%level; ***at the 1% level

34
 Table 4. Per Child Expenditure on Private Tutoring in 1997
(Models without possible endogeneous variables)

Model 2A - OLS Model 2B - Tobit Model 2C - CLAD Model 2D - CLAD

(All Regions) (All Regions) (All Regions) (Seoul & Metro Cities)

-96743*** -223280*** -96650 -53054***

Constant
(-9.636) (-17.030) (-10.80) (-8.253)
0.02888*** 0.03541*** 0.02016*** 0.02352***
Household Income
(15.445) (15.268) (10.437) (11.601)
5251*** 8391*** 4406*** 2746***
Mother Education
(4.979) (6.279) (5.454) (2.906)
2832*** 4957*** 3330*** 2940***
Father Education
(3.005) (4.156) (5.222) (3.107)
3359 -5266 -3688 1354
Father Professional
(0.512) (-0.646) (-.645) (0.452)
11150** 14684** 10925*** 14652***
Own House
(2.522) (2.630) (3.576) (2.703)
25998*** 33348*** 19365*** 20173***
High Density
(5.899) (6.072) (4.555) (5.072)
30098*** 18467*** 30000*** 45124***
Middle School
(6.564) (3.207) (6.611) (3.574)
2357 -34564*** -51947*** -31588*
High School
(0.499) (-5.658) (-9.881) (-1.715)
-5394 -6645 -4818* -6303**
Gender
(-1.457) (-1.419) (-1.922) (-2.285)
59148*** 95029*** 58923*** 26007***
Seoul
(9.756) (12.311) (13.09) (5.682)
27695 58156*** 34444***
Metro City ---
(4.797) (7.834) (8.11)
14951 41137*** 28027***
S&M City ---
(2.626) (5.599) (6.953)
Initial size 6718 Initial size 3317
Sample size 6576 6576
Final size 5849 Final size 3182

Adj. R2 = 0.1414 Psd. R2 = .0095 Psd. R2 = .1087 Psd. R2 = .0986

Joint significance F(12,6563)=91.25 χ2 (12)=1313.12

--- ---
(p-value) (0.0000) (0.0000)

Notes: t-values are in parentheses below coefficient estimates

*Statistically significant at the 10% level; **at the 5%level; ***at the 1% level

35
 Table 5. Household Expenditure on Private Tutoring in 1998

Model 3A - OLS Model 3B - Tobit Model 3C - CLAD

-127992*** -210662*** -170567***

Constant
(-27.64) (-35.15) (-18.43)
-4530 -8763* -4239
No Equalization
(-1.26) (-1.91) (-0.74)
0.0227*** 0.0264*** .02271***
Household Income
(34.85) (32.52) (13.44)
8172*** 10962*** 9021***
Head Education
(29.39) (30.60) (16.55)
14545*** 19628*** 116910***
Own House
(9.04) (9.65) (5.17)
-786 112 -1624
Mother Working
(-0.47) (0.05) (-0.81)
18796*** 24377*** 20990***
No. Kindergarten
(9.85) (10.13) (9.07)
47168*** 59502*** 56658***
No. Elementary School
(30.03) (29.95) (9.61)
50002*** 57038*** 42167***
No. Middle School
(22.76) (20.55) (12.14)
46348*** 57290*** 35111***
No. High School
(17.12) (16.71) (5.89)
27044*** 26308*** 14780***
Seoul
(9.36) (7.22) (3.15)
1285 2411 10009***
Metro City
(0.52) (0.77) (2.74)
No Equalization * 4915** 6321** 4101
No. Elementary School (1.99) (2.03) (0.69)
No Equalization * -6208* -3643 3513
No. Middle School (-1.85) (-0.86) (0.80)
No Equalization * -8327* -4617 -1135
No. High School (1.77) (-0.77) (0.13)
Initial size 19389
Sample size 19389 19389
Final size 17730

Adj. R2 = 0.2257 Psd. R2 = .0118 Psd. R2 = .1175

Joint significance F(14,19374)=403.55 χ2 (14)=4755.06

(p-value) (0.0000) (0.0000)

Notes: t-values are in parentheses below coefficient estimates

*Statistically significant at the 10% level; **at the 5%level; ***at the 1% level

36
 Appendix Table A 1. Means and Standard Deviations of Variables
Variables Definition Mean (Standard Deviation)
All Positive No Tutoring
Observations Tutoring Expenditure
Expenditure Only
Only
< SOPT Data: 1997 >
Per Child Expenditure on Monthly spending on private tutoring per child in won 108,177 143,954 0
Private Tutoring (200,731) (220,160)
Household Income Monthly income of the household in won 2,843,705 2,832,930 2,876,324
(8,617,947) (7,808,844) (1,070,000)
Upper 10% 1 if student is above top 10 % of the class; 0 otherwise 0.30 0.33 0.21
10-30% 1 if student is between 10 to 30 % of the class; 0 otherwise 0.27 0.28 0.24
70-90% 1 if student is between 70 to 90 % of the class; 0 otherwise 0.03 0.02 0.07
Lower 10% 1 if student below bottom 10 % of the class; 0 otherwise 0.01 0.01 0.02
Mother Education Years of mother’s education 11.39 (2.81) 11.89 (2.62) 9.86 (2.80)
Father Education Years of father’s education 12.52 (3.11) 13.04 (2.94) 10.94 (3.08)
Mother Working 1 if mother has a job; 0 otherwise 0.30 0.27 0.41
Father Professional 1 if father has a professional job; 0 otherwise 0.10 0.11 0.07
Own House 1 if the household owns one or more houses; 0 otherwise 0.73 0.73 0.73
High Density 1 if high density residential development area; 0 otherwise 0.35 0.40 0.21
Computer 1 if the household owns computers; 0 otherwise 0.49 0.52 0.42
No. Children Number of children in the household 2.07 (0.65) 2.03 (0.59) 2.19 (0.78)
Middle School 1 if middle school student; 0 otherwise 0.26 0.25 0.28
High School 1 if high school student; 0 otherwise 0.25 0.19 0.44
Repeater 1 if repeater; 0 otherwise 0.01 0.02 0.004
Gender 1 if male; 0 otherwise 0.52 0.52 0.51
Seoul 1 if Seoul; 0 otherwise 0.22 0.25 0.14
Metro City 1 if metropolitan city; 0 otherwise 0.27 0.29 0.21
S&M City 1 if small and medium sized city; 0 otherwise 0.28 0.29 0.25
(numberofobservations) 6,804 5,113 (75%) 1,691 (25%)
<UHES Data: 1998>
Per Household Expenditure Monthly spending on private tutoring per household in won 104,247 136,784 0
on PrivateTutoring (119,852) (119,990)
No Equalization 1 if not covered by equalization policy; 0 otherwise 0.29 0.29 0.30
Monthly income of the household in won 2,232,886 2,348,539 1,862,330
Household Income
(1,289,364) (1,309,111) (1,148,125)
Head Education Years of education of the head of the household 12.64 (3.05) 12.96(2.92) 11.63 (3.24)
Mother Working 1 if mother has a job; 0 otherwise 0.33 0.34 0.31
Own House 1 if the household owns one or more houses; 0 otherwise 0.58 0.60 0.50
No. Kindergarten Number of kindergarten students in the household 0.37(0.55) 0.37 (0.54) 0.38 (0.57)
No. Elementary Number of elementary school students in the household 0.63(0.74) 0.68(0.74) 0.44(0.69)
No. Middle Number of middle school students in the household 0.44(.54) 0.43(.55) 0.47(.54)
No. High Number of high school students in the household 0.18(0.39) 0.18 (0.38) 0.18(0.38)
(numberofobservations) 19,389 14,777 (76%) 4,612 (26%)

37

View publication stats