







































 

 

 

 

 

Substitution or Complementation: The 

Relationship between School Education and 

Shadow Education

Jiali Li

Shanghai Normal University, Shanghai 200234, China 

 

Abstract. In order to explore the relationship between school education 

and shadow education, this study applied two-level Bernoulli Model to 

estimate the school-level determinants of private tutoring by analyzing 
the data of PISA 2015 of four regions of China (Beijing, Shanghai, 

Jiangsu, and Guangdong). The results indicated that: (1) the four re-
gions have large scale of private tutoring ranging from 58.7% to 74.0% 

for science, mathematics, and reading; (2) the participation rate of pri-

vate tutoring in villages is much higher than that in cities; (3) the quality 
of school teachers significantly negatively affects the likelihood for re-

ceiving tutoring. Shadow education plays both substitutional and com-

plementary roles for school education, and the growth of shadow educa-
tion is in accompany with the low quality of school education in rural 

areas but with high quality of school education in urban areas. Instead 
of increasing Education resources and improving infrastructure, in-

creasing teacher quality will reduce the participation probability of pri-

vate tutoring. 

Best Evid Chin Edu 2019; 4(1):411-424. 

Doi: 10.15354/bece.20.ar015. 

Keywords: Shadow Education; Private Tutoring; Education Quality; Two-Level 

Bernoulli Model

 

 

 
 

Correspondence to: Jiali Li, Research Institute for International and Comparative Education, Shanghai Normal 

University, Shanghai 200234, China. E-mail: eduhappycharlie@163.com. 

Funding: The National Natural Science Foundation project "Family Capital, Shadow Education and Social Re-

production" (NSFC, 71774112). 

Conflict of Interests: None. 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 412 

Introduction 

HADOW education, which is also called private tutoring, has been becoming of 

great popularity in China, especially in developed regions, such as Beijing, 

Shanghai, and Guangdong. In 2014, 48.9%-58.1% of elementary students and 

66.8%-74.4% of middle school students received private tutoring in Shanghai (Zhang & 

Bray, 2016). In Beijing, 60.5% of elementary students and 58.4% of middle school stu-

dents attended private tutoring in 2015 (Lu, 2016). Early in 2012, the participation rate 

of tutoring in Guangzhou ranged from 51.7% to 73.1%. With the global expansion of 

private tutoring, a growing number of studies explored the factors underlying the de-

mand for private tutoring. Macro-factors, such as culture, education competition, and 

development of economic, have significant impact on the demand of tutoring. For ex-

ample, Confucianism, one of the cultural factors, places value on education and instills 

a hard work ethic on children, is often mentioned as the “soil” (driver) for private tutor-

ing (Zhu, 2013). The high-stakes examination exhibits great demand for tutoring as well 

as other supplementary learning activities. The growing of family income increases the 

chance to purchase tutoring service for their children, and institutionalization of public 

schooling stimulates parents and students to seek extra learning activities (Park et al., 

2016). Family background, gender, and ethnic minorities are also strongly associated 

with the development of tutoring. However, before embarking on the factors underlying 

the development of tutoring, it’s necessary to explore the definition and nature of shad-

ow education. The term of shadow education refers to various type of tutoring, and the 

metaphor of “shadow” highlights the relationship between private tutoring and regular 

schooling. Private tutoring exists only because of the existence of regular education 

though; its shape and size change following that of the regular education; the curricu-

lum in the “shadow” ones mimics the curriculum inside the regular schools and changes 

accordingly (Bray, 1999). The relationship means that the development of private tutor-

ing can be explained by the public education, and it also reflects the development of 

education quality and equity. 

Some debates exist about the relationship between regular education and shad-

ow education. First, when the mainstream education system is perceived to be of poor 

quality, parents and students may be motivated to seek extra forms of academic learning 

in private market, so some studies have attempted to find a link between quality of pub-

lic schooling and demand for private tutoring. Popa and Acedo (2006) noted that the 

backdrop of reduced funding for public schooling, increasingly poor population and 

teachers’ low salary were the reasons for the increasing of private tutoring in Eastern 

Europe. In Korea, students from schools with higher student-teacher ratios tended to 

spend more time in private tutoring (Kim & Lee, 2010). Similarly, Dang (2007) found 

that as teacher quality increased in Vietnam, the amount of money families spent on 

private tutoring decreased. This assumption postulates that the inadequacies of public 

education stimulate the high demand for private tutoring. Given the limitations of public 

schooling in meeting individualized needs for learning, private tutoring may play a role 

as a substitutional institution. However, institutional perspective on educational devel-

S 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 413 

opment offers a counter-explanation for the relationship between schooling quality and 

the demand for shadow education. Baker and LeTendre (2005) argued that growing 

demand for private tutoring reflected the institutionalization of public schooling. As the 

demand for educational opportunity and quality increases across social classes within 

and between countries, middle-class parents begin to seek private sources of additional 

learning for their children, because it becomes more difficult for them to influence regu-

lar schooling processes to maintain their children’s advantage. For example, during the 

years of 2003 to 2010, the participation rate of private tutoring was much higher in ur-

ban areas and in high-quality schools in China (Xue & Ding, 2009; Tseng, et al., 2010). 

From this perspective, growing demand for shadow education is complementary to the 

regular education, which is supported by the theory of Supply and Demand as well as 

the theory of Effectively Maintained Inequality (EMI). 

Based on the microeconomic theory of Supply and Demand, the demand for 

private tutoring depends on the quantity and quality of public education. When the 

quantity of public education cannot meet the demand of all the families, the families 

with high demand will take advantage of family capital to obtain more and better public 

education. While, when the supply of public schooling increases, the families with low 

demand can also receive regular education, the high-demand families will seek out-of-

school activities to keep their advantage to access to higher education (Dang & Rogers, 

2008). Xue (2015) employed EMI to indicate the relationship between regular educa-

tion and shadow education, and they mentioned that when compulsory education basi-

cally universal has not been achieved, the core of education competitive is the access 

chance to regular education for students from different social classes. When the com-

pulsory education basically universal is completed, the core of education competitive is 

the quality of school education. When the government takes measures to decrease the 

gap of educational quality between urban and rural areas, and between different schools, 

the core of education competitive will shift from educational quality to shadow educa-

tion. 

Thus, more empirical study is in need to verify whether the demand for shadow 

education is driven by the quality of public education. Is growing demand for shadow 

education due to the poor quality of school education, being substitutional to regular 

education, or due to the high quality of public schooling, being supplementary to regu-

lar education? Although previous studies indicated that students from better schools are 

more likely to receive private tutoring (Xue, 2009; Tseng et al., 2006), the dataset of 

these two studies were collected from urban cities in 2004 and 2009, when the compul-

sory education has not been completed. In order to explore the relationship between 

shadow education and regular school education, this study applies two-level Bernoulli 

Model to estimate the school-level determinants of private tutoring by analyzing the 

data of PISA 2015 of four regions of China including Beijing, Shanghai, Jiangsu, and 

Guangdong. 

Methods 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 414 

Sample 

This study uses the data of the PISA 2015 of four regions (Beijing-Shanghai-Jiangsu-

Guangdong). PISA program has been sponsored by the Organization for Economic Co-

operation and Development (OECD) every three years since 2000. It applied two-stage 

stratified sampling to randomly select 9,841 students of 286 schools in the four regions, 

and the weighted sample covered 96% of all the 15-year-old students. Apart from the 

international assessment of mathematics, reading, and science, and the survey also col-

lected extensive background information of teachers and schools. 

Variables 

Previous empirical studies have found that students’ family background, demographic 

and academic ability are related to their participation in private tutoring (Bray, et al., 

2014; Choi & Park, 2016; Li & Hu, 2017), and this study will control these factors to 

estimate the impact of school-level factors on students’ probability of receiving private 

tutoring. 

Education quality will be measured from three dimensions, i.e., education input, 

education process and education output. Items measuring education input and education 

process are usually taken as independent variables, while items measuring education 

output are used to be dependent variables (Hanushek, 1986). Education input variables 

include education resources, quality of teachers, and quality of students (Shavelson & 

Baxter, 1992). Education resources include a questionnaire with 8 items about school 

resources, the school principals’ perceptions of potential factors hindering the provision 

of instruction at school. The quality of teachers includes the index on staff shortage, 

which is derived from 4 items, as well as the proportion of teachers with bachelor or 

master degree. The quality of students is measured by the average ESCS of students’ 

family background. Education equity, which is called quality of education opportunity, 

emphasizes that students from different ESCS background should have equal access to 

education (Coleman, 1966). In China, educational equity is also called “education bal-

ance”, measured by three disparities, i.e., urban-rural disparity, region disparity and 

individual disparity (Wu, 2010). This study will analyze the relationship between edu-

cation quality and shadow education by exploring the urban-rural disparity and school-

ing disparity in receiving tutoring. For a more detailed description of the variables, refer 

to Table 1. 

Data Analysis 

A two-level Bernoulli model that is also called two-level hierarchical generalized linear 

model (HGLM) was used to estimate the effects of various student- and school-level 

variables on students’ participation in private tutoring. The two-level Bernoulli model 

was used because student-level variables nested in the school-level, and the dependent 

variable is a dummy variable indicating whether or not the student participates in pri-

vate tutoring. The level-1 and level-2 models are as follows: 

 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 415 

Level-1 model: 

        
   

     
                 

 

Level-2 model: 

                   

 

        

 

where Xi is a vector of level-1 control variables, βij refers to regression coefficient of 

level-1, εij is the error term for level-1; Kj is a vector of level-2 control variables, γ0s 

refers to regression coefficient of level-2, μ0j is the error term for level-2. (Only inter-

cept term is random, and the coefficients of level-2 are fixed) 

Results 

Descriptive Statistics 

Table 2 shows the results of the descriptive statistics. In general, the participation rate 

of science, mathematics and reading tutoring is 58.7%, 74%, and 63.4%, respectively in 

the four regions. The academic achievement of students from cities, high quality 

schools and high ESCS families is better than that of the students from villages, low 

quality schools, and low ESCS families. But the scale of tutoring is larger in villages, 

low quality schools, and low ESCS families than in their counterparts. Comparing with 

the high school students, middle school students received more tutoring, but the aca-

demic achievement of high school students is much higher than middle school students. 

There are significant intergroup disparities in the distribution of participation in private 

tutoring and students’ performance regarding urban-rural areas, institutional, family 

backgrounds, and individual ability. The results, that the participation rate of private 

tutoring in villages is much higher than that in towns and cities, and the lowest ESCS 

families have the highest participation rates, are totally different from the results of 

Xue’s report, from which the participation rate of tutoring is much higher in undevel-

oped county-level towns, poor quality schools than that in capital cities, high quality 

schools in 2009. The results based on PISA2015 indicated that the prevalence of tutor-

ing might gradually transfer from urban areas to rural areas, and the tutoring market has 

also begun to attract students from poor families. 

Due to the availability of data, most of previous studies mainly focused on the 

prevalence of tutoring, as well as the determinants and impacts of tutoring, few studies 

explored what kind of tutors the students preferred. Our study estimates the selection 

and distribution of tutors to explain the relationship between school education and 

shadow education. Table 3 indicates that regular teachers are most frequently used tu-

tors, with over 50% of students indicated tutoring by their own teachers. Around 29%- 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 416 

Table 1. Variable Descriptions. 

Variable Variable Descriptions 

Dependent Variable 

Private Tutoring  0=no, 1=yes; three subjects: science, mathematics, reading. 

Independent Variables 

Student-Level 

Female 0=male, 1=female. 

Study Program  0=middle school, 1=high school. 

Students’ Ability The mean score of science, mathematics, and reading. 

Students’ Educa-
tional Expectation 

1= middle school, 2=two-year high school, 3= high school, 4=vocational high 
school, 5=three-year college, 6=four-year college and above. 

Family Economic, 
Social And Cul-
tural Status 
(ESCS) 

Parents’ highest education level, parents’ highest occupation status, and home 
possessions compute the index of family ESCS. More the scores are, higher the 
family ESCS status is. 

School-Level 

School Location 
A village, hamlet or rural area (fewer than 3000 people)(reference group), a small 
town (3,000 to about 15,000 people), a town (15,000 to about 100,000 people),a 
city (100,000 to about 1,000,000 people), a large city (with over 1,000,000 people). 

Quality of 
Education 
Resources 

The shortage of school Education resources is scaled using four items: a lack of 
educational material (e.g. textbooks, IT equipment, library or laboratory material), 
inadequate or poor quality educational material (e.g. textbooks, IT equipment, li-
brary or laboratory material), a lack of physical infrastructure (e.g. building, 
grounds, heating/cooling, lighting and acoustic systems), inadequate or poor quality 
physical infrastructure (e.g. building, grounds, heating/cooling, lighting and acoustic 
systems). More the scores are, lower quality of the schooling resources is. 

Quality of 
Teachers 

(1) Staff shortage is derived from four items: A lack of teaching staff, inadequate or 
poorly qualified teaching staff, and a lack of assisting staff, inadequate or poorly 
qualified assisting staff. More the scores are, lower quality of teachers is. 
(2) The proportion of teachers with bachelor qualification (PROAT5AB), the propor-
tion of teachers with master qualification (PROAT5AM), the proportion of fully certi-
fied teachers (PROATCE). 

Quality of 
Students 

The mean scores of students’ family ESCS aggregated in school level. More the 
scores are, higher the family ESCS status of school is. 

Note: 1. PISA 2015, additional instruction is translated to 课外辅导 in the mainland of China, Which is 

also called 课外补习. 2. Family economic, social and cultural status is ESCS. 

 

 

 

40% of the students choose specialized tutors working for a business or organization. 

Non-specialized tutors are the least used. The academic achievement of the students 

choosing specialized tutors is the highest, whereas the students choosing non-

specialized tutors are the lowest. 

To explain why the participation rate of tutoring in village and small town is 

much higher than that of in city, we explored whether the choice of tutors varies across 

the groups of students. The choice of tutors varies significantly across students’ school- 

and family- backgrounds. Regular teachers are mostly used tutors for the students from 

village schools, but low quality schools with smallest number of teachers with master 

qualification and lowest ESCS families. Although special tutors working for a business  



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 417 

Table 2. The Distribution of Private Tutoring in the Four Regions. 

 

Science Mathematics Reading 

% 
Mean 
Scores 

% 
Mean 
Scores 

% 
Mean 
Scores 

Total Sample 58.7 518 74.0 531 63.4 494 

School-Level 

Regions 

A village 65.0 488 80.3 506 74.4 466 

A small town 61.6 495 76.9 509 70.2 471 

A town 59.2 527 74.7 541 66.0 502 

A city 52.0 544 65.0 555 56.0 524 

A large city 55.0 587 71.2 593 48.4 566 

 
x

2
=54.1 

p<0.001 

F =392.8 

p<0.001 

x
2
=88.4 

p<0.001 

F =307.6 

p<0.001 

x
2
=304.7 

p<0.001 

F =379.5 

p<0.001 

The proportion of Teachers with Master Qualification 

The Lowest 
1/4 

62.4 469 75.6 483 73.2 441 

The Lower 1/4 57.7 505 72.2 517 65.6 4801 

The Upper 1/4 56.4 528 73.7 539 61.9 501 

The Top 1/4 58.3 561 74.4 572 52.0 541 

 
 

x
2
=16.4 

p<0.001 

F =598.2 

p<0.001 

x
2
=6.9 

p<0.1 

F =541.2 

p<0.001 

x
2
=205.1 

p<0.001 

F =657.8 

p<0.001 

Student-Level 

ESCS 

The Lowest 
1/4 

62.4 475 75.6 488 73.2 444 

The Lower 1/4 57.7 516 72.2 530 65.6 496 

The Upper 1/4 56.4 534 73.7 547 61.9 515 

The Top 1/4 58.3 589 74.4 597 52.0 567 

 
x

2
=16.4 

p<0.001 

F =674.5 

p<0.001 

x
2
=6.9 

p<0.1 

F =575.4 

p<0.001 

x
2
=205.1 

p<0.001 

F =714.0 

p<0.001 

Study Program 

Middle school 64.4 498 79.6 510 71.3 473 

 High school 51.0 567 66.4 579 52.4 547 

 
x

2
=144.2 

p<0.001 

t =-36.5 

p<0.001 

x
2
=189.4 

p<0.001 

t =-36.6 

p<0.001 

x
2
=309.1 

p<0.001 

t =-37.3 

p<0.001 

 

 

 

are the mostly used for the students from lager cities, high quality schools with largest 

number of teachers with master qualification and highest ESCS families. Although 

more and more disadvantage students start to access to tutoring market, the quality of 

tutoring they receive might be not as good as the advantage students. We can also spec- 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 418 

Table 3. Type of Tutors Used by the Students Receiving Tutor-
ing. 
 Science Mathematics Reading 

% 
Mean 
Scores 

% 
Mean 
Scores 

% 
Mean 
Scores 

1. One of My Regular Teach-
ers 

53.4 496 52.3 510 56.4 468 

2. Other Regular Teachers in 
School 

36.7 505 33.2 515 29.6 479 

3. Specialized Tutors Works 
for A Business or Organiza-
tion. 

39.9 510 34.4 524 29.6 486 

4. Non-Specialized Tutors 16.2 470 13.6 488 14.4 439 

Note: this is a multiple-choice question. 1. The teacher is one of my regular teachers in this year’s school 
courses; 2. The teacher regularly teaches students my age in school but is not my teacher in any of my 
regular school courses; 3. The teacher mainly works for a business or organization specialized in addi-
tional instruction; 4. The teacher is not specialized teaching personnel (e.g. a student). 

 

 

 

Table 4. Type of Tutors Used Across Groups of Students. 

 
Science Mathematics Reading 

1 2 3 4 1 2 3 4 1 2 3 4 

Region 

A Village 58.9 37.0 41.5 19.3 55.1 32.2 37.0 17.3 57.9 30.3 33.2 15.6 

A Small 
Town 

61.2 36.9 37.5 17.8 59.1 33.3 31.9 14.7 62.5 30.8 29.6 16.1 

A Town 56.6 36.5 34.6 16.2 58.2 31.2 29.9 13.2 59.3 27.9 25.6 14.2 

A City 51.6 35.1 36.6 15.9 53.0 31.6 31.3 14.2 53.5 27.4 26.6 14.8 

A Large 
City 

36.1 37.1 48.7 11.8 34.7 36.0 42.7 9.4 40.9 29.0 34.0 9.1 

The Proportion of Teachers with Master Qualification 

Low 1/3 62.3 38.6 37.5 20.3 61.7 32.7 31.1 17.1 61.9 31.6 29.7 16.8 

Middle 
1/3  

52.8 35.9 37.6 14.8 53.0 33.2 34.1 13.2 57.0 28.7 28.3 14.0 

Top 1/3  43.2 35.6 45.8 12.8 40.4 34.0 39.0 10.3 47.2 27.8 31.5 11.4 

ESCS 

The 
Lowest 
1/4 

64.9 33.7 33.3 21.8 63.5 31.5 27.6 19.0 63.5 29.8 27.0 18.4 

The 
Lower 
1/4 

55.0 37.1 36.5 16.6 57.1 30.5 32.1 13.7 59.2 28.5 26.2 14.7 

The 
Upper 
1/4 

53.9 37.5 40.0 15.1 50.8 33.4 36.4 11.7 55.2 30.4 32.5 12.8 

The Top 
1/4 

39.4 38.8 50.4 11.1 37.9 37.3 41.7 10.2 43.8 29.6 34.4 10.1 

Note: 1= one of my regular teachers; 2=other regular teachers in school; 3=specialized tutors working for 
a business or organization; 4=non-specialized tutors. Data are percentages. 

 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 419 

ulate that most of regular teachers still provide paid tutoring, especially in village and 

small town (Table 4). 

The Determinants of Private Tutoring 

Given participation rate of private tutoring varies across groups of students, we ana-

lyzed the determinants of receiving private tutoring by employing a two-level Bernoulli 

model. Table 5 displays the results. 

For the individual student-level factors, students’ gender, education expectation, 

and ability, are significantly related to the probability of receiving private tutoring. Fe-

male students are more likely to receive science and reading tutoring than male students. 

Education expectation can significantly positively predict the probability of receiving 

science and mathematics tutoring. Students’ academic achievement has significantly 

negatively associated with the demand of private tutoring, which means Chinese stu-

dents may seek tutoring for remedial purposes. It is obvious that students from prosper-

ous families (high ESCS) have more chance to receive private tutoring than those from 

low economical families (low ESCS). 

For the school-level factors, we constructed three models to estimate the impact 

of schooling quality on private tutoring. In model 1, we explored the relationship be-

tween the locations of schools with the demand of tutoring. In model 2, we estimated 

the impact of Education resources and teachers’ quality on the demand of tutoring. In 

model 3, we discussed whether the quality of students can predict the probability of 

receiving tutoring. The results showed that: (1) Students from village schools are more 

likely to receive private tutoring than students from (large) city schools, and the quality 

of education resources, teachers, and students drives this result. With that when we put 

the variables of the quality of education resources, teachers, and students into the model 

3, students from large city schools become less probability to receive private tutoring 

than students from village schools. (2) The quality of school teachers negatively pre-

dicts the possibility of receiving tutoring. The lower proportion of teachers with master 

qualification the schools have, the students are more likely to receive science, 

mathmatics and reading tutoring. Students from schools with lower proportion of fully 

certified teachers have more possibility to take reading tutoring. 

The urban-rural disparity does appear to affect the likelihood of participating in 

private tutoring, and students of village schools are more likely to receive private tutor-

ing. The shortage of high-quality teachers in the village schools stimulates parents to 

purchase tutoring service to substitute for regular school education. 

Conclusion and Discussion 

This study applies two-level Bernoulli model to estimate the school-level determinants 

of private tutoring by analyzing the data of PISA 2015 conducting in four regions of 

China (Beijing, Shanghai, Jiangsu and Guangdong). The major findings are summarized 

below: 

 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 420 

Table 5. Results of the Two-Level Bernoulli Model. 

 Science Mathematics Reading 

M1 M2 M3 M1 M2 M3 M1 M2 M3 

Fixed Effects 

Small 
Town 

0.04 

(0.34) 

0.01 

(0.27) 

-0.18 

(0.20) 

0.08 

(0.27) 

0.12 

(0.23) 

0.10 

(0.24) 

-0.09 

(0.39) 

0.01 

(0.32) 

0.08 

(0.29) 

Town 
-0.17 

(0.36) 

-0.24 

(0.32) 

-0.51
† 

(0.21) 

-0.19 

(0.28) 

-0.25 

(0.26) 

-0.28 

(0.29) 

-0.37 

(0.41) 

-0.27 

(0.37) 

-0.09 

(0.35) 

City 
-0.46 

(0.34) 

-0.56
† 

(0.28) 

-0.53
† 

(0.24) 

-0.67
† 

(0.27) 

-0.71
‡ 

(0.23) 

-0.75
‡ 

(0.26) 

-0.81
† 

(0.41) 

-0.75
†
 

(0.32) 

-0.56
* 

(0.32) 

Large 
City 

-0.44 

(0.33) 

-0.26 

(0.29) 

-0.41
* 

(0.22) 

-0.30 

(0.26) 

-0.11 

(0.25) 

-0.18 

(0.29) 

-1.10
‡ 

(0.39) 

-0.66
† 

(0.33) 

-0.31 

(0.36) 

Resources 
Shortage 

 
0.08 

(0.08) 

0.09 

(0.08) 
 

0.13 

(0.07) 

0.13
* 

(0.07) 
 

0.04 

(0.09) 

0.07 

(0.08) 

Staff 
Shortage 

 
-0.08 

(0.07) 

-0.09 

(0.07) 
 

-0.10 

(0.06) 

-0.09 

(0.07) 
 

-0.03 

(0.07) 

-0.07 

(0.06) 

PROAT5AB  
0.35 

(0.37) 

0.49 

(0.41) 
 

0.36 

(0.38) 

0.28 

(0.41) 
 

-0.06 

(0.44) 

0.36 

(0.46) 

PROAT5 
AM 

 
-2.38

‡ 

(0.65) 

-1.75
† 

(0.81) 
 

-1.73
‡ 

(0.61)
 

-2.13
‡ 

(0.76) 
 

-3.88
‡ 

(0.72) 

-1.88
† 

(0.83) 

PROATCE  
-2.17 

(1.41) 

-2.08 

(1.39) 
 

-1.36 

(1.22) 

-1.42 

(1.22) 
 

-2.09
* 

(1.15) 

-1.81
* 

(1.04) 

Average 
ESCS 

  
-0.14 

(0.12) 
  

0.09 

(0.11) 
  

0.45
‡ 

(0.03) 

High 
School  

0.01 

(0.16) 

0.01 

(0.16) 

0.01 

(0.16) 

-0.33
* 

(0.19) 

-0.33
* 

(0.19) 

-0.33
* 

(0.19) 

-0.20 

(0.18) 

-0.21 

(0.18) 

-0.21 

(0.18) 

Girls  
-0.21

‡ 

(0.07) 

-0.21
‡ 

(0.08) 

-0.21
‡ 

(0.08) 

-0.06 

(0.08) 

-0.06 

(0.08) 

-0.06 

(0.08) 

-0.32
‡ 

(0.08) 

-0.33
‡ 

(0.08) 

-0.33
‡ 

(0.08) 

Ability  
-0.67

‡ 

(0.07) 

-0.68
‡ 

(0.07) 

-0.68
‡ 

(0.07) 

-0.82
‡ 

(0.07) 

-0.82
‡ 

(0.07) 

-0.82
‡ 

(0.07) 

-0.89
‡ 

(0.07) 

-0.91
‡ 

(0.07) 

-0.91
‡ 

(0.07) 

Education 
expectation 

0.09
‡ 

(0.03) 

0.10
‡ 

(0.03) 

0.10
‡ 

(0.03) 

0.12
‡ 

( 0.03) 

0.12
‡ 

( 0.03) 

0.12
‡ 

(0.03) 

0.04 

(0.03) 

0.05 

(0.03) 

0.05 

(0.03) 

ESCS 
0.14

‡ 

(0.05) 

0.14
‡ 

(0.05) 

0.14
‡ 

(0.05) 

0.22
‡ 

(0.05) 

0.22
‡ 

(0.05) 

0.22
‡ 

(0.05) 

0.09
† 

(0.04) 

0.09
† 

(0.04) 

0.09
† 

(0.04) 

Intercept 
0.53

‡ 

(0.07) 

0.50
‡ 

(0.06) 

0.48
‡ 

(0.06) 

1.22
‡ 

(0.06) 

1.19
‡ 

(0.06) 

1.21
‡ 

(0.06) 

0.81
‡ 

(0.08) 

0.75
‡ 

(0.06) 

0.70
‡ 

(0.06) 

Random Effects 

Variance 
Component  

0.37 0.34 0.34 0.35 0.33 0.33 0.53 0.46 0.41 

X
2
 847.3

‡
 793.4

‡
 787.5

‡
 750.6

‡
 704.1

‡
 702.8

‡
 1030.0

‡
 906.8

‡
 838.4

‡
 

Note: Standard Error of Mean in parentheses; *<0.1, †<0.05, ‡<0.01. 

 

 

 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 421 

Similar to the previous studies (Zhang & Bray, 2016; Lu, 2016; Wu, 2014), 

shadow education is prevalent in developed areas of China. The four regions we studied 

have a large scale of private tutoring, and the participation rate of private tutoring on 

Science, Mathematics and Chinese is 58.7%, 74.0%, 63.4%, respectively. 

The distribution of demand for private tutoring in the four regions of China has 

the following characteristics: village-city gap, interschool disparity, and regular teachers 

as paid tutors. In 2015, students from village schools with low quality are more likely to 

receive private tutoring than the students from city schools that have high quality. This 

finding is different from the previous studies, for example, according to Lei (2005), 

both the participation rates of private tutoring and tutoring expenditure in urban areas 

are much higher than that in rural areas in Beijing and Jiangsu. The following discus-

sion is mainly concerned with the reason why the private tutoring is more prevalence in 

village than in city in 2015. First of all, when the goal in the effort to make nine-year 

compulsory education basically universal has been attained, the demand for school edu-

cation increases in undeveloped areas, but the poor quality of regular education cannot 

meet students’ demand, so parents have to send their children to tutoring centers. Se-

cond, the results showed that regular school teachers are mostly used tutors for the dis-

advantage students, while specialized tutors are most likely used for the advantage stu-

dents. Perhaps regular teachers might be the best choice for disadvantage students in 

villages and small towns, where do not have any specialized tutoring organizations, 

because the fee is low and it’s convenient to study in teachers’ home. However, it is the 

low quality of public schools in villages and ineffective teaching practices in school 

education stimulates parents seeking private tutoring service as a “substitution” for reg-

ular school education. In fact, the phenomenon of regular teachers as paid tutors is 

common in China, especially in villages and small towns. Although school teachers are 

strictly prohibited from providing tutoring in China, it seems that this official policy has 

not been effectively implemented. Tutoring provided by mainstream teachers is a dou-

ble-edged sword, the dark side is that we cannot prohibit the school teachers from re-

ducing the coverage of regular lessons or revealing examination answers in tutoring 

class to foster demand for tutoring. What’s worse, teachers might trade off their own 

students with other teachers to avoid being penalized by the policy.  

The village-city gap and interschool disparity on the quality of regular educa-

tion have significant association with the prevalence of shadow education. The results 

indicated that students from public schools in villages have more probability to receive 

private tutoring and poor quality of teachers in school lead to high demand of private 

tutoring in 2015. During the past ten years, the scale of shadow education in villages 

began to expand. We attempt to explain this change based on the theory of Supply and 

Demand: when compulsory education basically universally has not been attained, shad-

ow education is more common in city. Because the quantity of public education cannot 

meet the demand of disadvantage families in village, they have to fight for more public 

education. However, advantage families in city who enjoy the quantity and quality of 

public education purchased tutoring service as a “complementation” to make sure they 

will success in education competition. When compulsory education basically universal-



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 422 

ly has been completed, shadow education becomes more popular in village and small 

town. For the quality of education cannot meet the demand of disadvantage families in 

village and small town, they have to purchase tutoring service as a “substitution” to ob-

tain more entrance opportunity. Although families in large cities still purchase tutoring 

service to keep their advantage in university opportunities, the participation rate of pri-

vate tutoring increased slower than that in village and small town because the quality 

and equity of public education has been further developed in these areas. 

In overall, shadow education plays both substitutional and complementary role 

in school education, that the growth of shadow education is in accompany with the low 

quality of school education in village and small town, where families purchase shadow 

education as a “substitution” to obtain more entrance opportunity when the quality of 

school education cannot meet their demand. The growth of shadow education develops 

with the high quality of school education in city, where families purchase shadow edu-

cation as a “complementation” to make sure they keep their advantage in education 

competition. However, shadow education is mostly provided by regular teachers of pub-

lic schools in village and small town, and in city, most of the students choose special-

ized tutors. 

It’s necessary to re-examine the problem of school education according to the 

development of shadow education. First, there is still urban-rural disparity in education-

al quality. To decrease the participation probability of private tutoring, the government 

should increase teacher quality of regular education, with the increasing of the quality 

of education resources and improving of the infrastructure building. Second, the phe-

nomenon of regular teachers as tutors is common, especially in undeveloped areas, such 

as villages and small towns. “Why do school teachers provide paid tutoring? Do the 

students participate in tutoring provided by teachers on their own? Are there any “cor-

ruption” behavior for school teachers to attract more students to participate in their tu-

toring class?” Schools and educational institution should pay more attention to these 

problems. Third, no matter the “substitutional” role of shadow education in village, or 

the “complementary” role in city, shadow education seems to replicate educational ine-

quality outside school, which has negative impact on the quality and equity of regular 

education. Of course, it remains to be verified whether the results of the relationship 

between regular education and shadow education based on the data of these four regions 

are applicable to other provinces and cities.  

 

 

 

 

 

 

 

 

 

 



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 423 

References

Baker, D., LeTendre, G. K. (2005) National 

differences, global similarities: World cul-

ture and the future of schooling. Stanford 

University Press. 

Bray, M. (1999) The shadow education system: 

private tutoring and its implications for 

planners. Fundamentals of Educational 

Planning Series; Number 61. 

Bray, M., Zhan, S., et al. (2014) Differentiated 

demand for private supplementary tutoring: 

Patterns and implications in Hong Kong 

secondary education. Econ Edu Rev, 38: 24-

37. 

Byun, S. Y., and Park, H. (2012) The academic 

success of East Asian American youth the 

role of shadow education. Sociol Edu, 

85(1):40-60.  

Choi, Y., & Park, H. (2016) shadow education 

and educational inequality in south korea: 

examining effect heterogeneity of shadow 

education on middle school seniors’ 

achievement test scores. Res Soc Stratif Mob, 

44:22-32. 

Coleman, J. S., et al. (1996) Equality of educa-

tional opportunity. 

Dang H. A., Rogers F. H. (2008) The growing 

phenomenon of private tutoring: Does it 

deepen human capital, widen inequalities, or 

waste resources? World Bank Res Obs, 

23(2):161-200. 

Dang, H. A. (2007) The determinants and impact 

of private tutoring classes in Vietnam. Econ 

Edu Rev, 26(6): 683-698. 

Hanushek, E. A. (1986) The economics of 

schooling: Production and efficiency in pub-

lic schools. J Econ Lit, 24(3):1141-1177. 

Ho, E. S. C. (2011). Shadow education in Asian 

society from an international perspective. 

Analysing the quality and equality of educa-

tional systems from n international perspec-

tive insights from PISA (pp: 271-

296).Educational Science Publishing House.  

Kim, S., Lee, J. H. (2010) Private tutoring and 

demand for education in South Korea. Econ 

Dev Cult Chang, 58(2):259-296. 

Lei, W. P. (2005) Expenditure on private tutor-

ing for senior secondary students: determi-

nants and policy implications. Edu Econ, 

(1):39-42. 

Li, J.L., Hu, Y. M. (2017) Who can benefit from 

shadow education and its implication for ed-

ucation inequality. Edu Econ, (2):87-101. 

Lu, K. (2016) Investigation on private tutoring 

of primary and secondary school students in 

Beijing. Shanghai Res Edu, (5):37-40. 

Park, H., et al. (2016) Learning beyond the 

school walls: trends and implications. Ann 

Rev Sociol, 42:231-252. 

Popa, S., Acedo, C. (2006) Redefining profes-

sionalism: Romanian secondary education 

teachers and the private tutoring system. Int 

J Edu Dev, 26(1):98-110. 

Shavelson, R., Baxter, G. (1992) On measuring 

the quality of mathematics and science edu-

cation: achievement indicators. Measuring 

the Quality of Education; 35-54. 

Southgate, D. E. (2013) Family capital: A de-

terminant of supplementary education in 17 

nations. Out of the Shadows. Glob Intensif 

Suppl Edu, 22:245-258. 

Stevenson, D., and Baker, D. (1992) Shadow 

Education and Allocation in Regular school-

ing: Transition to University in Japan. Am J 

Sociol, 97(6):1639-1657. 

Tsang, M. C., Ding, X. H., & Shen, H. (2010). 

Urban- rural disparities in private tutoring of 

lower-secondary students. Edu Econ, (2):7-

11. 

Wu, K. (2010) Three levels of educational op-

portunity equity. Chin Edu Newspap, 2010-

05-04(004). 

Wu, Y. (2014) The private tutoring in middle 

school schools from the perspective of edu-

cation equality: A case study in Guangzhou. 

Edu Res, (8):75-84. 

Xue, H. P. (2015). From school education to 

shadow education: Education competition 

and social reproduction. Peking Univ 

EduRev, 13(3):47-69. 

Xue, H. P., & Ding, X. H. (2008). An empirical 

study on private tutoring for students in ur-

ban China. Econ Edu Res (Beida), 6(1):1-14. 

Zhang, W., Bray, M. (2016) Micro-

neoliberalism in China: public-private inter-



Li. Relationship between School Education and Shadow Education 

Vol.4, No. 1, 2020 424 

actions at the confluence of mainstream and 

shadow education. J Edu Pol, 32(1):63-81. 

Zhu, X. (2013) Shadow education cultural capi-

tal theory under globalization of education. 

Tsinghua J Edu, 4:51-55. 

Received: 22 October 2019 

Revised: 11 December 2019 

Accepted: 20 December 2019 

 

The Chinese version of this article has been published in Beijing Social Sciences, 2019; 5: 57-68. The 

English version has been authorized for being publication in BECE by the author(s) and the Chinese 

journal. 

李佳丽. “替代”还是“补充”: 从影子教育发展审视学校教育质量-基于 PISA2015 中国四省市
数据的分析. 北京社会科学, 2019; 5: 57-68. 

 


	Article-JialiLi-BECE_title_30Jan2020
	Article-JialiLi-BECE_Maintext30Jan2020

