Vous êtes sur la page 1sur 27

Education Economics

Vol. 14, No. 2, 155–180, June 2006

The Accumulation of Human Capital Over Time and its


Impact on Salary Growth in China

ZEYUN LIU* and JIN XIAO**


*School of Economics and Business Administration, Beijing Normal University, China **Faculty
of Education, the Chinese University of Hong Kong, China
ZeyunLiu
Education
0964-5292
Original
Taylor
202006
14
alanliu326@hotmail.com
00000June
& Article
10.1080/09645290600622913
CEDE_A_162272.sgm
andFrancis
Economics
(print)/1469-5782
2006 Ltd
Francis (online)

ABSTRACT This study compares the growth in salaries across three spatial regions in
China during the period 1993–1998, when economic reforms were implemented nation-
wide. Our study compares the impact of three forms of education and training on salary
growth, namely pre-job formal schooling, on-the-job-training provided by employers, and
adult education paid for by the employees themselves. We used a three-level hierarchical
linear model to partition variance among individual, firm, and regional characteristics.
The data were drawn from a 1998 survey of 16 485 employees from 365 firms in six
provinces (two provinces in the eastern part of the country, two in the central part, and
two in the western part). We found that: (1) regional disparities have a paramount impact
on differences in salary; (2) individual characteristics defined by firm as well as firm
characteristics are significantly related to salary decisions; (3) returns to formal schooling
increase significantly in more market-based regions; and (4) employees also benefit by
receiving on-the-job-training and by participating in adult education programs outside
their firm.

KEY WORDS: Human capital; salary; hierarchical linear model; China

Introduction
The private benefits of education are an important issue in the process of economic
development and educational expansion. A large number of empirical studies esti-
mating the rates of returns on education (RORE) have supported the view that
formal schooling is a crucial factor in differences in personal income in developed
countries (Cohn and Addison, 1998) as well as in less developed countries
(Psacharopoulos and Patrinos, 2002). However, as Mincer (1974) has claimed,
because in many cases the data used do not include information about other forms
of education and training (ET) received during the period of one’s employment,
most studies of RORE have focused on the formal schooling acquired at the initial
phase of one’s life and have not captured much of the learning acquired over a
person’s life. One survey of studies of RORE in developed countries (Cohn and

Correspondence Address: Jin Xiao, Faculty of Education, the Chinese University of Hong Kong, Sha Tin,
N.T., Hong Kong, China. Email: xiaojin@cuhk.edu.hk

0964-5292 print/1469-5782 online/06/020155-26 © 2006 Taylor & Francis


DOI: 10.1080/09645290600622913
156 Z. Liu & J. Xiao

Addison, 1998) has noted that ET beyond formal schooling is a way of accumulating
human capital in the course of production. Nevertheless, recent studies of RORE
in China (Li, 2003; Li and Ding, 2003; Lai, 1998; Johnson and Chow, 1997) also do
not include information on the participation of workers in adult education and on-
the-job training in the workplace, which have become popular during the period
of China’s economic transition. Estimations of RORE are not accurate, as participa-
tion in ET activities beyond formal schooling is neglected.
Since the early 1990s, China has experienced dramatic changes during the trans-
formation of the country’s economy from a planned one to a market-oriented one.
Many countries have likewise experienced rapid transitions in a global economy.
Some scholars (Ashton et al., 2000) have built a new framework to understand the
process of the formation of skills in different cultural contexts. They have identi-
fied four stakeholders: the state; its apparatus of formal education; employers as
users of skills from whom demands for skills arise; and workers, whose personal
capacities in the market determine the need for skills. All of these stakeholders
influence the supply of skills. Lauder (2001) further suggested that the issue of the
diffusion of skills is related to ET systems and to the structure of the labor market.
A recent survey by the Organization of Economic Cooperation for Development
on the development of adult education in six member countries has indicated that
the rise and disappearance of various types of jobs has created a demand for a
whole range of adult education programs that can help deliver skills required at
different occupational levels (Belanger and Tuijnman, 1997). A study on the train-
ing of workers in seven different occupations in Colombia, a less developed coun-
try, identified a variety of channels through which the workers have chosen to
satisfy their training needs (Ziderman and Horn, 1995). Middleton and associates
also found that employers in less developed countries provided ET to their
employees to develop their skills (Middleton et al., 1993). Surveys conducted in
Shenzhen (Xiao and Tsang, 1999) and Shanghai (Xiao, 2002a) provide further
evidence that both employers and employees have assumed responsibility for
developing skills needed in the workplace.
International experience also suggests that, in addition to ET systems, socio-
economic factors are crucial to explaining productivity. Based on 10 years of
observations in the United States, Levin and Kelley (1994) did not find either
monetary or non-monetary benefits, in terms of average income, interest in politi-
cal participation, as well as dependence in social welfare, to be positively related
to formal education. They maintained that firm profitability and employee salary
are related to multiple factors such as economic and social policies, people’s
confidence in government, educational expansion, management strategies in
workplace, and so forth.
This paper, with a survey data of employees in China’s three regions, is a study
to capture the impact of: (a) formal schooling (FS), firm-provided on-the-job
training (OJT), and adult education (AE) outside of the firm, paid for by
employees themselves, on salary growth from 1993 to 1998, a crucial period in the
transformation of China’s economy; and (b) the inequality of economic reforms in
the three regions.

Labor Markets, Education, and Training Systems in China


China is a huge country. Both economic development and educational expansion
are uneven (Hu et al., 1995; Wu, 2004; Zhang, 2003). Regional disparities are
Human Capital Impact on Salary Growth in China 157

partially, but significantly, due to government policies. The eastern coastal


regions (the East) were opened up to economic development in the 1980s, while
the central region (the Central), and the western region (the West) were opened
up in the early 1990s and the late 1990s, respectively. Hu et al. (1995, p. 210),
Zhang (2003) and Wu (2004) have argued that disparities across the three regions
can be attributed to the fact that developed regions possess most of the resources
and that opportunities in less developed regions are limited. Regional disparities
create inequalities in China so that ET may not play the same role in labor
markets. Studies conducted by Wei et al. (1999) and Chen and Chen (2002), and by
Johnston (1999), have noted the productivity of firms and inequality among
regions. Studies on RORE in China, employing the Mincerian earning equation
and using the most recent urban household survey data of 1995, have found that,
more than ever, the market economy is allowing workers to capture the benefits
gained from education (Li, 2003; Li and Ding, 2003). Li (2003) found higher
returns to schooling for those newly hired in the early 1990s, after economic
reforms had started; and both Li and Ding (2003) and Lai (1998) found that work-
ers in non-western regions can earn significantly more due to their FS. However,
both did not capture the effects of firms nor the magnitude of regional disparities.
Our study involves six provinces, namely Guangdong and Jiangsu in the East,
Hebei and Hubei in the Central, and Yunnan and Shaanxi in the West.1
Guangdong and Jiangsu are ranked as the richest, with per-capita Gross Domestic
Products (GDPs) of RMB 11 143 and RMB 10 021(in 1998 prices),2 respectively.
Hebei and Hubei are at an intermediate level, with per-capita GDPs of RMB 6525
and RMB 6300, respectively. Yunnan and Shaanxi in the West lagged far behind,
with per-capita GDPs of RMB 4355 and RMB 3834, respectively.3 An analysis of
productivity shall consider these determinants.
Educational development in China has followed a unique path. Before the
1980s, under the planned economy, ET was under the full control of the state. In
order to recruit resources to meet the huge demand for education arising during
the period of economic transition, the government employed different measures.
It devolved financing for education to local governments, to increase the provi-
sion of compulsory education for the population (China Central Committee,
1985). The government further strengthened its budgetary efforts and, since 1993,
has focused on education in urban areas and on higher education as the state
envisioned a blueprint for modernization through science and education (China
Central Committee and the State Council, 1993). The government also advocated
ET to improve the vocational skills of employees in order to adapt to changing
needs in the workplace (China Central Committee and the State Council, 1981).
But this endeavor was left entirely to work units or to the users of skills. During
the course of the economic transition, investment in ET related to workplace
needs has been influenced by government policies, by management reform at the
firm level, and by individual choice (Xiao and Tsang, 1999, 2004; Xiao, 2002a).
The restructuring of markets and the supply of ET in China are attributed to the
transformation of the economy, and thus to the management as well as the intro-
duction of technology. In the planned economy, the Party, through the govern-
ment, controlled the organization of production in the factory floor through
directives on labor and quotas on production and supplies. The market orienta-
tion of the economy has pushed the state to give up some powers so that firms
can act independently to achieve efficiency (Kang, 1999, pp. 60–82). That firms
can be formed in a variety of ownership structures and can control their own
158 Z. Liu & J. Xiao

daily operations are indicators of a changed socio-economic environment. The


economic sectors have become detached from the state in terms of administration
and have to build their competence to survive in the market. At the same time,
continuing integration into the international market has prompted firms to
introduce technological innovations in the workplace. This has become another
important factor behind the demand for learning in the workplace. As govern-
ment-funded FS is mainly responsible for the initial education received by
individuals, there is a need for ET that is responsive to changes in the workplace.
Benson and Zhu (2002) recently found that firms in Beijing and Shanghai make
use of both internal and external labor markets. When the external market for
hiring cannot meet their human resource needs, they use an internal strategy of
OJT to produce the required human resources. We may assume that firms that
have gained control over production expect their employees to learn new sets of
knowledge, skills, and values to become proficient in major work roles that differ
from those in the planned economy. Due to the firm’s own mission of survival
and ideology of management, the OJT of each firm serves to better to diffuse
specific job skills to employees and socialize them into the firm’s culture
(Bowman, 1996 ; Hake, 1999, pp. 83–84).
The breaking of the state-guaranteed ‘iron rice bowl’ has opened up occupa-
tional horizons for individual workers. But they can also be dismissed if their job
performance is not up to expectations or if the firm is downsizing. As argued by
Hake (1999, pp. 85–87), individuals increasingly have to assume personal respon-
sibility for formulating their identities in life courses. Studies in other parts of
China have found that improving job-related skills is ranked as more important
than other reasons for taking AE courses (Jones and Wallis, 1992). Therefore,
paying for AE is becoming another important avenue to advance one’s career.
Individuals can realize their occupational aspirations through the external market
if they fail to do so in the internal market.
Both firms as employers and workers as employees are assuming the responsi-
bility to develop skills in workers, making both OJT and AE responsive to
changes in the workplace and to their needs. This has enabled a market for skill
formation to evolve outside the government education system. Both OJT and AE
are becoming new forms of the accumulation of human capital that are close to
the workplace. They are a crucial factor affecting returns to education as well.
Hence, we assume that, in addition to FS, OJT and AE are also rational choices of
investment in human capital and can improve individual productivity. Therefore,
we maintain that any estimation of the returns to education should include these
two kinds of investment in human capital in China’s context.
Of the three major sources for the supply of ET, the government mainly
finances pre-job FS; while firms organize up to nine types of OJT training for their
own workers in a firm-based mode.4 AE has come to involve many job-related
programs for working adults who often voluntarily pay for them. Therefore,
Chinese workers have a mixed path of ET and receive three types of education,
which are funded differently (Xiao and Tsang, 1999; Xiao, 2004).
OJT and AE are widely used in China nowadays to improve job skills. For
example, during the period 1992–1997, 59%, 60%, and 40% of the sampled
workers in three metropolitan cities received OJT (Shenzhen, Shanghai, and
Chongqing); and the percentage of workers who attended AE programs are 31%,
38%, and 36%, respectively (Xiao, 2004, p. 174). Our survey shows that 65%, 54%,
and 49% of the sampled employees received OJT in the East, the Central, and the
Human Capital Impact on Salary Growth in China 159

West, respectively, during the period from 1993 to 1998. At the same time, 21%,
25%, and 23% of the sampled employees attended AE programs, respectively
(Xiao, 2004).
We therefore hypothesize that: (1) in addition to FS, both AE and OJT have an
impact on decisions on employee salaries; (2) change in the workplace is another
important learning experience and influences decisions on salaries; (3) firm-level
factors play an important role in salary decisions; (4) all of these factors vary from
region to region because inequalities in China have an impact on the firms’
decisions on salaries. This study adds information about the ET received by
workers beyond FS. It compares the impact of different forms of human capital
development in the course of production. The analysis also reveals magnitudes of
discounted RORE when inequalities cause a region to lag behind in market
reforms.
Our two tasks are: (a) to measure the effects of individual-level, firm-level, and
regional-level characteristics on salary determinations in the period of the first
five years of reform to a market-oriented economy (1993–1998); and (b) to exam-
ine the impact of FS, OJT, and AE on salary growth. In the next section, we will
discuss a modified earning equation with a hierarchical linear model and our
data. The third section presents the empirical results. The concluding section will
provide a discussion of the implications of the findings.

Modeling Salary Growth and Data


Data in social studies are often stratified. For instance, an employee is nested in a
work-unit/firm and a work-unit/firm is nested in a region/province. These
features form a multilevel structure. Two methods are often used to deal with
multilevel data in a traditional regression analysis. One is to use higher-level
variables (i.e., the characteristics of an organization) as dummy variables to make
a regression at the lower level (i.e., the individual level). An estimation bias will
occur because it is not correct to assume that individuals within the same organi-
zation are independent and that the characteristics of a firm have the same effect
on all employees from different firms. The other method is to aggregate lower-
level variables (i.e., often individual) and integrate them in a regression at the
higher level (i.e., the level of the organization). This treatment will cause some
useful information to be lost; that is, the variance among employees within the
same firm, which accounts for a large part of the total variance of a dependent
variable.
The Mincerian earning equation (MEE) (Mincer, 1974) has been a standard
method to estimate the private benefits of education as argued by human capital
theory. In order to capture the complicated context of economic transition in
China as discussed earlier, this study will use a three-level hierarchical linear
regression model (HLM), a modification of the MEE method, to partition the total
variance into three levels: (a) differences in salary growth over the years for a
particular person; (b) differences in individual characteristics; and (c) differences
in firm characteristics. This is important in China’s context, where inequality in
educational resources and policies has been a fundamental issue.
At level one, the HLM growth model (Raudentbush and Bryk, 2002, pp. 160–204)
will be employed to capture increases in salary at three points in time over a period
of five years of the early stage of economic transition, much as with the MEE
method, but instead using only one point wage as the dependent variable. This
160 Z. Liu & J. Xiao

will portray the slope of the increases in salary over the years due to the experience
gained from work, or learning by doing. At the second and third levels of the
regressions of HLM, individual characteristic variables and firm characteristic
variables will be put in, respectively, to differentiate the variance and source of the
impacts. By using the modified models, Xiao (2001, 2002b), in her studies on the
determinants of salary growth in Shenzhen and Shanghai, found that: (1) FS affects
the initial salary but not the growth in salary; (2) participation in AE affects salary
growth in the case of Shanghai, while received OJT correlates significantly with
salary growth in Shenzhen; and (3) a firm’s characteristics, such as ownership and
economic sector, play a decisive role in salary growth.
Analysis with HLM contains variance at each level and introduces variables of
each level to explain the variance in the dependent variable at that level (Rauden-
bush and Bryk, 2002). The advantage of a modified MEE as demonstrated by
Xiao’s studies (2001, 2002b) is that it can be used to differentiate sources of impact
and their magnitude on wage growth due to maturation in work, individual
characteristics, and firm characteristics. This delineates the structure of one’s
work life as: one matures over the years; how this maturation is affected by one’s
personal characteristics; and how one’s job-related characteristics are influenced
by the characteristics of one’s work unit. This paper will employ the same
modified models to portray salary growth using data from 12 counties across
China during the early period of the economic transition.

Survey and Data


This study adopted the reverse tracer study technique (RTST) to investigate an
employee’s salary and educational experience. RTST, as illustrated by Ziderman
and Horn (1995), begins with the current employment status of each employee
and seeks to identify each major alternative ET route pursued by the employee.
Thus, RTST permits the analysis of an array of education and training options
utilized both by firms and individual employees for the accumulation of human
capital. This technique also allows a comparison of profitability among types of
ET programs. In addition, it has a lower study cost.
Two questionnaires were distributed in autumn 1998 to collect the data. The
firm questionnaire investigated firm characteristics and the employee question-
naire was designed for respondents to recall their relevant experiences over the
period 1993–1998. The employee questionnaire consisted of five groups of data:
(1) the employee’s demographic information and work experience; (2) FS before
the first job; (3) OJT provided by the firm; (4) AE attended outside of the firm; and
(5) characteristics reflecting individual productivity including job position, level
of technical proficiency, experience of changes in the workplace, and salary.5
Table 1 presents descriptions of level 1 variables reflecting changes in individ-
ual salaries over time. TIME is the number of years that have elapsed from 1993 to
the year when the data were reported. SALARY refers to the monthly salary in
1993 prices at three time points, 1993, 1995, and 1998. In this study, salary consists
of three components: basic salary, bonus, and allowance. In the analysis, Y is the
natural logarithmic value of SALARY.6
Table 2 presents descriptions of individual variables for a level 2 analysis. SEX
is a dummy variable, with 0 for female and 1 for male. EDUYEAR is coded as the
actual number of years of FS that the employee received. An employee who
completed primary school is coded as 6, lower secondary school as 9, upper
Human Capital Impact on Salary Growth in China 161

Table 1. Descriptions of level 1 variables


Variable Definition Year Code Overall East Central West

TIME Years elapsed 1993 0 0 0 0 0


from 1993
1995 2 2 2 2 2
1998 5 5 5 5 5
SALARY Actual monthly 1993 Mean (standard 253(251) 327(275) 215(251) 156(92)
salary in 1993 deviation)
prices (RMB)
1995 Mean (standard 350(318) 445(339) 297(333) 238(115)
deviation)
1998 Mean (standard 389(310) 482(323) 322(313) 305(190)
deviation)
N 16 485 7181 6151 3153

secondary school (including high school and upper secondary vocational school)
as 12, junior college as 14, and university as 16 and graduate study as 19. EXP
refers to the total work experience in years up to 1993, which is the sum of both
EXPFIRM, the work experience in the current firm, and EXPOTHER, the work
experience in other firms. JOB refers to the employee’s job position in 1993 and
the codes reflect a hierarchical order. A front-line worker is coded as 0, a member
of staff such as salesperson or clerical personnel is coded as 1, and a member of
managerial or professional staff is coded as 2. RANK refers to the employee’s level
of proficiency in the JOB position in 1993. We use three levels to define the indi-
vidual’s technical proficiency: entry, intermediate, and senior, which are coded as
0, 1, and 2, respectively. TRAIN is a dummy variable, with 1 referring to having
received OJT during the period 1993–1998 and 0 to having received no such train-
ing. AE is a dummy variable with 1 indicating attendance in AE courses from
1993 to 1998 and 0 for not having attended such courses. CHANGE refers to the
amount of changes the employee experienced in the workplace from 1993 to 1998.
In the survey, the employees were asked whether they had experienced four
kinds of changes: the introduction of new products or services, new equipment,
new production technology, and new job positions. If they experienced no
changes they were coded as 0, if they had experienced one kind of change they
were coded as 1, two kinds as 2, three kinds as 3, and four or more kinds as 4.
IN_GAIN refers to advancements in an employee’s proficiency ranking in the
same job position during the five years. A gain from entry-level to intermediate-
level or from intermediate-level to senior-level is coded as 1, a gain from entry-
level to senior-level is coded as 2, and no gain in the five years is coded as 0.
BE_GAIN refers to promotion in an employee’s job position in the five-year
period from 1993 to 1998. Promotion from worker to salesperson/clerical person-
nel or from salesperson/clerical personnel to managerial/professional staff is
coded as 1, promotion from worker to managerial/professional staff is coded as 2,
and no promotion in the five years is coded as 0.
Firm-level variables are presented in Table 3. REGION refers to the geographic
localities of the East, the Central, or the West of China. PROVINCE refers to the
two provinces in each region; namely, Jiangsu and Guangdong in the East, Hubei
and Hebei in the Central, and Shaanxi and Yunnan in the West. Ownership refers
to the type of investment ownership. PRI_COLL is a dummy variable with 1 for a
162 Z. Liu & J. Xiao

Table 2. Descriptions of level 2 variables


Variable Definition Code Overall East Central West

SEX Female, male (%) 0 42.3 45.8 39.9 39.0


1 57.7 54.2 60.1 61.0
EDUYEAR Years of formal schooling Mean 11.1 11.0 11.3 10.9
EXP Years of work experience up to 1993 Mean 12.4 13.3 11.7 11.7
EXP_SQ The square of work experience Mean 228.4 257.1 205.8 207.0
EXPFIRM Years of work experience in the firm up Mean 9.2 9.2 9.0 9.2
to 1993
EXPOTHER Years of work experience in other firms Mean 3.2 4.1 2.6 2.5
up to 1993
JOB Job position in 1993 (%)
Front-line worker 0 66.1 67.1 63.0 69.8
Salesperson/clerical personnel 1 9.1 9.2 9.0 9.2
Managerial/professional staff 2 24.8 23.8 28.0 21.0
RANK Proficiency level in job position in
1993 (%)
Entry-level 0 77.1 76.8 75.9 79.9
Intermediate-level 1 19.5 19.6 20.6 17.1
Senior-level 2 3.4 3.6 3.4 3.0
CHANGE Experience of changes in workplace (%)
None 0 37.4 36.3 36.3 42.1
One kind of change 1 25.8 24.9 26.4 26.6
Two kinds of changes 2 13.3 14.0 13.5 11.6
Three kinds of changes 3 12.4 12.6 12.8 10.9
Four kinds of changes 4 11.1 12.3 10.9 8.7
TRAIN Received OJT during 1993–1998 (%) 1, 0 63.9 70.5 60.4 55.9
AE Attended AE courses during 1, 0 21.7 18.7 24.6 22.8
1993–1998 (%)
IN_GAIN Gain in proficiency level within the
same job position during 1993–1998 (%)
None 0 86.6 88.1 84.5 87.0
One gain in proficiency level 1 12.1 11.0 13.8 11.2
Two gains in proficiency level 2 1.4 0.9 1.6 1.8
BE_GAIN Promotion between job positions
during 1993 to 1998 (%)
None 0 90.1 91.9 88.0 90.0
One-level promotion 1 3.7 3.0 4.2 4.1
Two-level promotion 2 6.3 5.1 7.8 5.9
N 16 485 7181 6151 3153

local private or collective firm. STA_CORP is dummy variable with 1 referring to


a state-owned or corporate firm. FOREIGN is a dummy variable with 1 for a firm,
which had investment from Hong Kong, Macao, Taiwan, or other countries in
either mode of sole investment or joint venture. SECTOR is a dummy variable,
with 1 referring to firms in the service sector and 0 referring to those in the indu-
sial sector. SIZE is coded in an ordinal manner, with 0 for small firms (those with
fewer than 300 employees), 1 for medium firms (301–800 employees), and 2 for
large firms (those with more than 800 employees). AV_EDU refers to the average
number of years of schooling of the sampled employees within a firm. TREX-
TENT refers to the extent of OJT in a firm, measured by the proportion of all
Human Capital Impact on Salary Growth in China 163

Table 3. Descriptions of level 3 variables


Variable Definition Code Overall East Central West

REGIONa The East, the Central, the West (%) 2, 1, 0 37.0 38.6 24.4
PROVINCEb Jiangsu, Guangdong (%) 0, 1 37.1
Hubei, Hebei 0, 1 49.6
Shaanxi, Yunnan 0, 1 50.6
PRI_COLL Private or collective (%) 0, 1 34.5 32.6 29.1 46.1
STA_CORP State-owned or corporate (%) 0, 1 51.5 51.1 53.2 49.4
FOREIGN HK, Macao, Taiwan, or foreign (%) 0, 1 14.0 16.3 17.7 4.5
SECTOR Economic sector (%)
Industrial 0 69.9 67.4 73.8 67.4
Service 1 30.1 32.6 26.2 32.6
SIZE Size of firm (%)
Small (fewer than 300 employees) 0 44.9 37.8 39.7 64.0
Medium (301–800 employees) 1 26.0 34.1 24.8 15.7
Large (more than 800 employees) 2 29.0 28.2 35.5 20.2
TREXTENT Extent of training during 1993–1998
Employees who received OJT (%) 60.2 68.5 56.7 52.9
AEEXTENT AE extent during 1993–1998
Employees who attended AE (%) 22.6 19.7 25.7 21.9
AV_EDU Average years of formal schooling Mean 11.2 11.2 11.5 10.9
(year)
N 365 135 141 89

a
The Central is the reference group. b Jiangsu is the reference group, Hubei is the reference group, and
Shaanxi is the reference group.

employees who received OJT during 1993–1998. Finally, AEEXTENT refers to the
proportion of employees who participated in AE within the firm. The latter three
variables measure the average human capital stock accumulated in the firm.

Measuring Salary Determination in three Regions


The analysis consists of three levels of models: (a) the level 1 model contains
temporal variables relating to the individual, taking employee salary at three time
points as the dependent variable; (b) the level 2 model contains individual vari-
ables including sex, FS, OJT, AE, work experience, job position, proficiency level,
advancement in proficiency level, job promotion, and experience of changes in
the firm; and (c) the level 3 model contains firm-level variables including region,
size, sector, ownership, average years of schooling, extent of OJT provided by the
firm, and extent of AE pursued by employees. The level-1 model is:7

Ytij = π0 ij + π1ij (Time)tij + π 2 ij (Time 2 )tij + etij (1)

where t = 0, 2, 5 refers to the years 1993, 1995, and 1998, respectively; i = 1, 2, …, I


(I is the number of employees in firm j used for the analysis); j = 1, 2, …, J (J is the
number of firms used for the analysis); Ytij is the outcome variable (i.e., the log
monthly salary at time t for employee i nested in firm j); π0ij is the parameter of
the initial salary for employee ij at the beginning of 1993; π1ij is the parameter of
the average annual salary growth rate for employee ij during the five years; π2ij is
164 Z. Liu & J. Xiao

the parameter of the accelerated rate of salary growth based on a squared term for
employee ij; and etij is the error item. It is assumed that each etij is independently
and normally distributed with a mean of zero and a constant variance, σ2.
The variable, schooling in the MEE (Mincer, 1974), is treated as an individual
characteristic and placed in level 2. This modified level 1 growth model, retaining
only temporal variables, captures salary growth and one’s maturation or learning
by doing on the job.
Then, the parameters in the level 1 model become the outcome variables of the
level 2 model; and the variability of one’s initial salary and growth in salary will
be predicted by the variables for employee characteristics. The level 2 model is as
follows:

Q
π pij = β p 0 j + ∑ β pqj X qij + rpij ( 2)
q =1

where p = 0, 1, 2 are the first-level parameters; q = 1, 2, …, Q (Q is the number of


individual-level predictors); βp0j is the intercept term for πpij; Xqij is a characteristic
of employee i in firm j (e.g., sex, FS, OJT, AE, etc.); βpqj is the effect of Xqij on the pth
growth parameter; and rpij is the error item. The set of P + 1 random effects are
assumed to be multivariates, normally distributed with a full covariance matrix,
T, dimensioned (P + 1) × (P + 1). Each has a mean of zero.
The parameters in the level 2 model then become outcome variables in the level
3 model, and their variability will be predicted by the variables for firm character-
istics to determine how firms shape the behavior of their employees. The level 3
model is as follows:

S
β pqj = γ pq 0 + ∑ γ pqsWsj + upqj ( 3)
s =1

where s = 1, 2, …, S (S is the number of firm-level predictors); γpq0 is the intercept


term for βpqj; Wsj is a firm characteristic used as a predictor for the firm’s effect
(γpqs) on βpqj; and upqj is a level 3 random effect that represents the deviation of firm
j’s parameter, βpqj, from its predicted value based on a firm-level model. For each
firm, there are Q + 1 equations in the level 3 model. The residuals are assumed to
be multivariates normally distributed. Each is assumed to have a mean of zero,
some variance, and covariance among all pairs of elements. This three-level
model presents the determinants of salary, including effects due to time, individ-
ual characteristics as well as firm characteristics in an organizational setting.

The Empirical Results


The empirical results are presented in three groups. First, the level 1 model estab-
lishes a baseline model to capture the unconditional growth curve of the increase
in salary over the period 1993–1998 for each individual workers. Then variables of
individual characteristics are put into the conditional level 2 model to predict the
initial salary and salary growth. Finally, variables of firm characteristics and
regions are put in the full level 3 model to predict variations in individual
employees across firms. This delineates an employee’s work life as follows: a
person matures over the years; how a person’s maturation in work is affected by
Human Capital Impact on Salary Growth in China 165

his/her personal characteristics; and a person’s job-related characteristics are


influenced by the characteristics of his/her work unit.

Unconditional Level 1 Model


Using equation (1) without introducing any individual-level variables to predict
any πpij and without firm-level variables to predict βpqj, we obtain the following
unconditional level 1 model:

Ytij = π0 ij + π1ij (Time)tij + π2 ij (Time 2 )tij + etij ( 4)

π0 ij = β 00 j + r0 ij ( 5)

π1ij = β 10 j + r1ij (6)

π 2 ij = β 20 j (7 )

β00 j = y000 + u00 (8)

β10 j = y100 + u10 j (9)

β20 j = y200 (10)

The result of unconditional level 1 model consists of two components: fixed


effects and random effects. Fixed effects offer estimated parameters for temporal
variables only, and capture salary growth and one’s maturation or learning by
doing on the job. Because there are only three occasions in the observations,
variance at the individual level can only allow t – 1 random variance for freedom
in calculation, in this case, only two level 2 random effects, r0ij and r1ij with
variances τ00j and τ11j, respectively, and with a covariance of τ01j.
Table 4 presents both fixed and random effects for the overall sample as well as
for each of the three regions. The estimated grand-mean of the monthly salary at
the initial point in 1993 for employees in the East was RMB 291.82 (= e5.676153), for
employees in the Central was RMB 182.27 (= e5.205472), and for employees in the
West was RMB 139.49 (= e4.937981), about US$39.3, US$24.6 and US$18.6, respec-
tively. The estimated grand-mean of the initial salary in the West was not yet half
of that in the East, and the initial salary in the Central was not yet two-thirds of
that in the East. Over the five-year period, the estimated average annual growth
rates of salaries were 22.1% in the East, 24.4% in the Central, and 29.6% in the
West. The estimated growth rates in a squared acceleration were all negative in
each region, indicating that salary growth decelerated during the five years. The
fixed effects indicate a concave downward trajectory of salary growth the longer
an employee stayed in a firm.
The lower panel of Table 4 presents the random effects. With regard to the
employee’s initial salary, the variance within firms (46.2%) was smaller than vari-
ance among firms (53.8%). There fore, we decided to run a test of the variance for
the sample for each region; the variance within the firm (59.23% for East, 71.84%
for Central, and 61.61% for West) turned out greater than the variance across
firms (40.77% for East, 28.16% for Central, and 38.39% for West) within each of the
166 Z. Liu & J. Xiao

Table 4. Fixed and random effects of unconditional level 1 model


Overall East Central West

Fixed effects: estimated parameters


For initial salary π0ij
For intercept β00ϕ
For intercept γ000 5.3227*** 5.6762*** 5.2055*** 4.9380***
For salary growth rate π1ij
For intercept β10j
For intercept γ100 0.2457*** 0.2210*** 0.2438*** 0.2956***
For accelerated rate of salary growth π2ij
For intercept β20j
For intercept γ200 −0.0288*** −0.0268*** −0.0300*** −0.0313***
Random effects: variance partitioning
Initial salary
Variance within the firm (%) 46.2 59.2 71.8 61.6
Variance among firms (%) 53.8 40.8 28.2 38.4
Salary growth rate
Variance within the firm (%) 46.9 51.5 52.8 46.9
Variance among firms (%) 53.1 48.5 47.3 53.1

***p < 0.001.

three regions. This indicates that the difference in initial salary was greater among
regions than within each region. For salary growth rate, again, the variance
among firms (53.1%) was greater than the variance within firms (46.9%) for the
overall sample. For each region, the variance was more due to difference within
the firm in the East and Central, but the situation was the reverse in the West.8 All
of this indicates that (a) regional inequalities explain most of the differences in
salary and in salary growth, but (b) within each region individual differences
within a firm explain more than half of the differences in salary. This means that if
the same person moved from West to Central, or from Central to East, he/she
would automatically get a higher salary, while within the same region individual
difference mattered more than firm factors with regard to decisions relating to
salary.
Nevertheless, the proportion of between-firm variance in the total variance
(28–53%) was bigger than that noted in Shanghai and Shenzhen, which was 16–
26% (See: Xiao, 2001, Table 7; Xiao, 2002b, Table 6).9 Shanghai and Shenzhen rank
as the most developed cities in China. Our sample consists of firms in counties,
which are much less developed. This comparison of Shanghai and Shenzhen with
our analysis may indicate that firm-level factors have a greater effect on salary
decisions in county-level firms than those in a metropolis. We suppose that such
a difference may be due to inequalities among segregated labor markets in China.
This is a matter that has not been given enough attention in previous studies on
labor markets as the use of the MEE method and urban survey data.

Conditional Level 2 Model


At level 2, individual variables are introduced to predict πpij. We propose that
level 1 coefficients, which are dependent variables at a level 2 estimation, are
Human Capital Impact on Salary Growth in China 167

associated with individual predictors. Fixed effects will present the impact of
these individual predictors on salary and salary growth. Random effects allow us
to find how much of the variance in initial salary and in salary growth rate the
individual predictors would explain. Firm-level predictors are not introduced for
βpqj at this step, so level 3 equations are not presented here with the conditional
level 2 model.

2
Ytij = π0 ij + π1ij (Time) tij + π 2 ij (Time )tij + etij (11)

π0 ij = β 00 j + β 01 j (Sex)ij + β 02 j (Eduyear )ij + β 03 j (Exp 2 )ij + β 04 j (Expfirm)ij


+ β 05 j (Expother )ij + β 06 j ( Jobrank )ij + β 07 j ( Rank )ij + r0 ij (12)

π1ij = β 10 j + β 11 j (Sex)ij + β 12 j (Eduyear )ij + β 13 j (Expfirm)ij + β 14 j (Expother )ij


+ β 15 j (Change)ij + β 16 j (Train)ij + β 17 j ( Ae)ij + β 18 j ( In _ Gain)ij + β 19 j (Be _ Gain)ij
+ β 110 j ( Jobrank )ij + r1ij (13)

π 2 ij = β 20 j + β 21 j (Eduyear )ij + β 22 j (Train)ij + β 23 j ( Ae)ij (14)

Table 5 presents the fixed effects. For SEX, the initial salary of male employees
was significantly higher than that of female employees. The coefficient showing
gender differences in the West (0.105) was higher than that in the Central (0.074),
and that in the Central was higher than that in the East (0.071). The fact that
gender differences in salary were greater in poorer regions is accordance with
Meng’s finding (Meng, 1998). The rate of salary growth, however, did not differ
significantly between male and female employees.
For EDUYEAR, years of FS had a significantly positive effect on initial salary in
the three regions. For every one more year of FS they had received, employees
would have a salary premium of 1.7% at the initial point in the East, 1.1% in the
Central, and 1.5% in the West.10 In regard to salary growth (β12j), the coefficients
for FS were significantly positive in the East and the Central; that is, for every one
more year of FS, an employee would receive 0.05% and 1% more of an increase in
salary each year, respectively, in the East and Central. But FS was not significantly
associated with salary growth in the West. Considering that the West is the poor-
est region in China, we argue that because it had been opened up for fewer years
and a market-based economy had not yet been established, firms in the West had
yet to associate FS with individual productivity.
If we assume that the positive impact of FS on initial-point salary reflects the
signaling function of education (Spence, 1973) and that the positive impact of FS
on salary growth reflects the function of education in improving one’s productiv-
ity and/or ability to deal with disequilibria (Shultz, 1961, 1975), results from this
study might indicate that both of these two functions exist at the county level in
East and Central China. The results also support the weak version of the screen-
ing hypothesis put forward by Psacharopoulos (1979). Nevertheless, in the
studies of Shanghai and Shenzhen, the most developed cities in China, Xiao (2001,
2002b) found FS to be irrelevant to the growth in employee salaries. Therefore, we
cannot yet conclude that FS can improve individual productivity in the overall
168 Z. Liu & J. Xiao

Table 5. Fixed effects of conditional level 2 model


Overall East Central West

For individual initial salary π0ij


For intercept β00j
For intercept γ000 4.859974*** 5.205833*** 4.777103*** 4.440681***
SEX β01j 0.077261*** 0.070718*** 0.074378*** 0.104565***
EDUYEAR β02j 0.014153*** 0.016463*** 0.010846** 0.014926**
EXP_SQ β03j −0.000059* −0.000061 −0.000062 −0.000218
EXPFIRM β04j 0.019697*** 0.017219*** 0.019255*** 0.023579***
EXPOTHER β05j 0.015568*** 0.014294*** 0.013685*** 0.020528***
JOB β06j 0.051451*** 0.051560*** 0.050706*** 0.052943***
RANK β07j 0.067764*** 0.077697*** 0.053175*** 0.079474***
For individual salary growth
rate π1ij
For intercept β10j
For intercept γ100 0.186014*** 0.166818*** 0.142519*** 0.310376***
SEX β11j −0.001498 −0.000429 −0.003235 −0.003218
EDUYEAR β12j 0.005872*** 0.005079*** 0.010031*** −0.001484
EXPFIRM β13j −0.001489*** −0.001235*** −0.001852*** −0.001261***
EXPOTHER β14j −0.000934*** −0.000851*** −0.001261*** −0.000576*
CHANGE β15j 0.001724*** 0.000753 0.002343*** 0.003018***
TRAIN β16j 0.004968 0.004867 −0.000350 0.025635***
AE β17j 0.011337** 0.015909** 0.015919** −0.013449
IN_GAIN β18j 0.008966*** 0.014681*** 0.005927** 0.004077
BE_GAIN β19j 0.008440*** 0.009607*** 0.006320*** 0.010055***
JOB β110j −0.000029 0.002039 −0.001335 −0.002119
For individual accelerate rate of
salary growth π2ij
For intercept β20j
For intercept γ200 −0.017104*** −0.016675*** −0.010348*** −0.031836

***p < 0.001, **p < 0.01, *p < 0.05.

context of China, even though we can accept that FS is a useful signal in labor
markets. Employee salary growth is related to some other firm-level characteris-
tics. We will illustrate this shortly.
With regard to work experience counted in years, we identified three types: the
total worker experience before 1993 (EXP_SQ),11 which includes work experience
in the current firm (EXPFIRM) and work experience in other firms (EXPOTHER).
For salary at the initial point, we have three major findings. First, work experience
squared to 1993 does not have an impact. Second, both EXFIRM and EXPOTHER
significantly increase one’s salary at the initial point, but employees get more
benefits as they acquire experience working in the current firm than from
working in other firms. This is illustrated by coefficients of 0.0246 for EXPFIRM
being greater than 0.0205 for EXPOTHER in the West; 0.0192 for EXPFIRM being
greater than 0.0137 for EXPOTHER in the Central; and 0.0172 for EXPFIRM being
greater than 0.0143 for EXPOTHER in the East. This finding is consistent with the
results form Meng’s study of Chinese firms (Meng, 1998) and Huang’s study of
Taiwanese firms (Huang, 2001). If we regard work experience as a type of job-
specific training (Becker, 1964) through which employees accumulate human
Human Capital Impact on Salary Growth in China 169

capital, we can say that firms value it and will pay more for the more relevant
experience gain from working in the current firms. Firms in poorer regions put
more weight on work experience in determining salaries. For salary growth rate,
work experience of any type before 1993 was negatively associated with the
decision of a firm to increase an employee’s salary. This indicates that for those
who had more years of work experience before 1993, the rate of increase would be
smaller than that for those who had fewer years of work experience before 1993,
maybe more years to come.
Experience of change (CHANGE) in the workplace predicts salary growth in the
West and the Central, not in the East. For example, for employees overall, the
premium in the salary growth rate for those who experienced four kinds of
changes than those who experienced no change during the period 1993–1998 was
0.9% (0.009372 = 0.002343 × 4) in the Central and 1.2% (0.012072 = 0.003018 × 4) in
the West. That is, employees who underwent more changes in the workplace had
a steeper salary growth rate. This finding allies with findings in Shanghai and
Shenzhen (Xiao, 2001, 2002b). Changes in the workplace offer an opportunity for
employees to learn and develop one’s ability while the firm is transforming.12
Firms in the West and Central will acknowledge such experience in due course,
but firms in the East will not.
With regard to OJT and AE, their effects on salary growth differ among regions.
For instance, receiving OJT only positively affects salary growth in the West (β16j
= 0.025635, p < 0.001), while experience of AE might increase employee salaries in
the East (β17j = 0.015909, p < 0.01) and the Central (β17j = 0.015919, p < 0.01).13 OJT
has been a common strategy in the East and Central for developing human
resources. Since most of the employees received OJT,14 OJT cannot make a
remarkable difference in the level of technical proficiency between employees
and has no significant effect on salary increase decisions. Firms in the East and the
Central, therefore, recognize employees who pursue learning opportunities
outside firms. On the other hand, employees in the West have fewer opportuni-
ties to gain access to OJT, so employees who have received OJT are more likely to
get an increase in salary. That AE is not associated with salary growth in the West
may be partly because firms in the West do not recognize its relevance to the job
performance of employees. These findings may indicate that firms use different
strategies to develop their human resources.
The productive function of AE and OJT may vary in different localities. Xiao
found that the decision made by employees to participate in AE will affect salary
growth in Shanghai, while the OJT provided by employers is significantly corre-
lated with the growth in salaries in Shenzhen (Xiao, 2001, 2002b). Considering all
of these findings, we argue that locality forms a complicated context and that each
Chinese firm seeks a unique way to cope with it. They may find alternatives to
accumulating human capital and increasing productivity. So it is impossible to
come to a universal conclusion on the role of education (FS, OJT, or AE) in
improving productivity without taking into account the context of the decisions
of firms.
Let us look at the impact of job position and level of proficiency. In order to
take into account the complexity of competence in the workplace, we define four
variables. Job position (JOB) is a workplace hierarchical factor and reflects an
employee’s work task assigned by the firm. Proficiency level (RANK) is the rank
of technical efficiency, indicating an employee’s competence in a particular
position or productivity acknowledged by the firm. We also define a promotion
170 Z. Liu & J. Xiao

in job position (BE_GAIN) and a promotion in improved proficiency (IN_GAIN)


for the five-year period.
We find that, in all three regions, both JOB and RANK were positively related to
salary at the initial point. However, job position (JOB) in 1993 was not associated
with salary growth during the next five years. Nevertheless, promotion in job
position (BE_GAIN) significantly affects salary growth: the estimated coefficients
were 0.0096 for the East, 0.0063 for the Central, and 0.01 for the West. In regard to
one’s level of proficiency (IN_GAIN), firms in the East (0.015) gave more weight to
improvements to one’s proficiency than firms in the Central (0.006). But in the
economically lagging West, the link between improvements in job proficiency
and salary growth has not yet been established. These findings reflect a firm-level
or micro-level difference in regional development. Another important finding of
this study is that estimates from the overall sample do not reflect different
patterns for firms in different spatial localities.
The random effects of the conditional level 2 model (see column 2 in Table 6)
indicate that individual predictors explained substantial variance in the initial
salaries for employees (19% for overall sample; 18%, 19%, and 22% for the East,
the Central, and the West, respectively). But the explanatory power for the vari-
ance in the growth rate of salaries for employees is smaller (5% for overall sample;
3%, 7%, and 3% for the East, the Central, and the West, respectively). The results
are acceptable compared with those from other studies using a three-level growth
model (for example, Boyle and Willms, 2001; Raudenbush and Chan, 1992;
Raudenbush, 2001).

Conditional Level 3 Model


We put firm-level variables in the level 3 model to predict βpqj, outcome variables
in the level 2 model. Theoretically, each βpqj parameter has a regression function.
For instance, we add region, ownership, sector, size, average FS years, and OJT
extent to explain variability of the firm’s mean initial salary (β00j) and firm mean
salary growth (β10j).15 The estimation takes three steps. First, each firm-level
variable indicating location (region or province), ownership, sector, or size will be
put in level 3 models respectively to discern the effect of each variable. Second,
based on the results from the former step, we select one or several firm-level
variables to explore the explanatory power of these conventional characteristics.16
Third, based on the former step, variables representing the average stock of
human capital in a firm (e.g., average years of schooling, extent of OJT, and extent
of AE) will be added to form the final model.

Random effects. Let us begin by looking at random effects to get a sense of the
explanatory power of level 3 variables. In Table 6, column 1 presents variance
partitioning between individual-level and firm-level in the baseline model as
illustrated in Table 4; column 2 presents variance explained by employee charac-
teristics; column 3-1a presents variance explained by the variable of region (in the
overall model) or province (in the East, the Central, or the West model); columns
3-1b–3-1d present variance explained by ownership, size, and sector, respectively;
column 3-2 (conventional characteristics model) presents variance explained by
all of the variables of region, province, ownership, size, and sector together; and
column 3-3 presents variance explained with a full model by adding the average
stock of human capital over the all of the characteristics of firms together.
Table 6. Random effects (%)
Level 1 Level 2 Level 3 model
Variance Variance
partitioning explained Additional variance explained
Unconditional With employee
model characteristics With firm characteristics
Region and Conventional Firm’s average
province Ownership Sector Size characteristics human capital
Column Column Column Column Column Column
Column 1 Column 2 3-1a 3-1b 3-1c 3-1d 3-2 3-3

Overall samplea
Level 2 (within a firm)
Individual initial salary τπ0 46.2 19
Individual growth rate τπ1 46.9 5
Level 3 (among firms)
Firm mean initial salary τβ0 53.8 3 49 3 0 2 51 57
Firm mean growth rate τβ1 53.1 6 15 1 −1 0 17 19
The East sampleb
Level 2 (within firm)
Individual initial salary τπ0 59.3 18
Individual growth rate τπ1 51.5 3
Level 3 (among firms)
Firm mean initial salary τβ0 40.8 −11 34 0 3 1 38 41
Firm mean growth rate τβ1 48.5 5 0 5 4 0 9 12
The Central samplec
Level 2 (within a firm)
Individual initial salary τπ0 71.8 19
Individual growth rate τπ1 52.8 7
Level 3 (among firms)
Firm mean initial salary τβ0 28.2 13 2 5 6 1 10 17
Firm mean growth rate τβ1 47.3 14 0 5 −1 2 4 7
Human Capital Impact on Salary Growth in China 171
Table 6. (continued)
172

Level 1 Level 2 Level 3 model


Variance Variance
partitioning explained Additional variance explained
Unconditional With employee
model characteristics With firm characteristics
Region and Conventional Firm’s average
Z. Liu & J. Xiao

province Ownership Sector Size characteristics human capital


Column Column Column Column Column Column
Column 1 Column 2 3-1a 3-1b 3-1c 3-1d 3-2 3-3

The West sampled


Level 2 (within a firm)
Individual initial salary τπ0 61.6 22
Individual growth rate τπ1 46.9 3
Level 3 (among firms)
Firm mean initial salary τβ0 38.4 3 6 9 2 6 14 16
Firm mean growth rate τβ1 53.1 5 18 0 2 1 21 26

a
The Central is the reference group. b Jiangsu is the reference group. c Hubei is the reference group. d Shaanxi is the reference group.
Human Capital Impact on Salary Growth in China 173

First, let us concentrate on the overall model. If differences among regions are
not taken into account, the firm characteristic variable ownership (FOREIGN)
accounts for at most an additional 3% of the variance in firm mean salary at the
initial point, and SECTOR counts for 0 and SIZE counts for 2% rate (see columns
3-1b–3-1d, for variance, τβ0 explained by level 3 variables in the overall sample).
These three firm variables at most explain an additional 1% of the variance in firm
mean salary growth (τβ1). However, when we put in the locality dummy variable
(the Central is the reference group), an amazing 49% of the variance in firm mean
salary at the initial point and 15% of the variance in firm mean salary growth rate
will be predicted (column 3-1a). Therefore, the conclusion is that the differences
in both the initial salary and in the salary growth among firms are mainly due to
differences between the three regions. Compared with localities, the contribution
of ownership, sector, or size can almost be ignored in this context.17 This finding
has a significant implication: as far as an income distribution system within the
firm is concerned, it is impossible for firms and their employees in less developed
regions to catch up with firms in more developed regions (e.g., the Central with
the East and the West with the Central) by management reforms to increase effi-
ciency at the firm level (such as changing ownership, altering economic sector,
and increasing economies of scale). In order to equalize the regional disparities,
some macro-level policies are necessary.18
As the variable REGION would explain most of the variance, we ran the model
with the sample of each region to detect whether firm characteristics would have
greater impact within a region. The lower panel provides information on random
effects. Localities (provinces) within the same region have significant explanatory
power in the East and West. In the East, the region dummy variable (Jiangsu
province is the reference group) can explain 34% of the variance between firm
mean salaries at the initial point. In the West, the region dummy variable
(Shaanxi province is the reference group) can explain 6% of the variance in firm
mean salary at the initial point and 18% of the variance in the rate of increase in
firm mean salary. In the Central, the region dummy variable (Hubei province is
the reference group) has little explanatory power. Variance explained by conven-
tional model, including all the firm variables (e.g., REGION or PROVINCE,
ownership, section and size), is about the sum of variance explained by these
variable entered one at a time (see column 3-2).
Finally, after the average stock of human capital variable of each firm is put into
the equation over the firm characteristic variables (column 3-3), the variances
explained an increase of 6% in mean salary at the initial point and an increase of
2% in the mean growth rate for the overall sample, 3% and 3% for the East, 7%
and 3% for the Central, and 3% and 5% for the West, respectively. This compari-
son shows that the average human stock has more or less the same explanatory
power as the three variables of firm characteristics; namely, ownership, sector
and size together.

Fixed effects. We shall sum up our major findings here.19 First, for provincial
differences within a region, as compared with Jiangsu Province in the East, firms
in Guangdong province offered a significantly higher mean initial salary (γ001 =
0.280587, p = 0.001), but gave a significantly lower rate of increase in salary (γ101 =
−0.075574, p < 0.0001). In the East, firm size has a positive effect on gender in initial
salary (γ013 = 0.019119, p = 0.024) and salary growth rate for those employees who
experienced a gain in their level of proficiency (γ182 = 0.004682, p = 0.019). In the
174 Z. Liu & J. Xiao

Central, there are no insignificant differences in firm mean salary at the initial point
neither in the growth rate of the firm mean salary between the two provinces, Hebei
and Hubei. Firms in Yunnan and Shaanxi have almost the same initial salary;
however, Yunnan has a faster growth rate (γ101 = 0.045026, p < 0.001).
With regard to ownership, in the East firms with foreign investment offered a
lower initial salary than state-owned or corporate firms (γ004 = −0.243182, p =
0.026), but a very higher rate of growth in salary (γ104 = 0.045143, p = 0.020).
Private/collective firms and state-owned or corporate firms offered almost the
same initial salary, but the former had a slower rate of salary growth (γ103 = −
0.033406, p = 0.034). In firms in the East with foreign investment, employees
earned more at the initial point (γ023 = 0.020247, p = 0.005) than those in state-
owned or corporate firms with the same level of FS and work experience. In the
West, job position and level of proficiency may have had a greater effect on salary
at the initial point in private or collective firms than in state-owned or corporate
firms (γ061 = 0.026671, p = 0.031; γ071 = 0.051122, p = 0.014).
For sector, in the East firms in the service sector gave a higher initial salary than
that of firms in the industrial sector (γ002 = 0.113114, p = 0.024), but a lower growth
rate in salary (γ102 = −0.020669, p = 0.019). Employees in the service sector had a
lower rate of increase in salary for their improved job proficiency (γ181 = −
0.007123, p = 0.035). In the Central, firms in the service sector also gave a higher
salary at the initial point than those firms in the industrial sector (γ002 = 0.592553,
p < 0.001), but their salary growth rates were not different.
For the stock of human capital in a firm, the average FS was positively related
to the firm mean salary at the initial point in the East (γ005 = 0.062332, p = 0.007)
and in the West (γ004 = 0.075348, p < 0.001). In the Central and the West, the overall
proportion of OJT received among employees in a firm had a positive effect on
the firm mean rate of salary growth (γ104 = 0.038230, p = 0.019 and γ103 = 0.045493,
p = 0.028), indicating that firm’s value OJT for its role in improving the job compe-
tence of their employees. But the overall proportion of employees attending AE
did not have much effect on salary decisions in all three regions.

Concluding Discussion
This study has focused on how education and training are contributing to the
growth in employee salaries in the transition from a planned economy to a
market-based economy. Economic returns to human capital are achieved in the
complicated context of regional disparities. Thus, we placed the analysis of deter-
minants of the salary of employees in a multi-level framework that considers
regional factors, the features of firms, and individual characteristics, as well as
personal maturation. Methodologically, we set up models both for the overall
sampled employees and for employees in each of the three regions for compari-
son. The findings give strong support to our assumptions.
The main findings of this study reveal that: (1) regional disparities have a para-
mount impact on differences in salary and, in comparison, firm characteristics play
very limited roles; (2) individual characteristics matched with job characteristics
(e.g., JOB, RANK, work experience, improved proficiency and promotion) are signif-
icantly related to firms’ salary decisions; (3) formal schooling has significant impact
on salary decisions, too; and (4) finally, employees benefit by alternatives of continu-
ing learning, including being involved in workplace change, receiving on-the-job-
training and in participating in adult education programs outside their firm.
Human Capital Impact on Salary Growth in China 175

As far as research hypotheses 1 and 2 are concerned, we found that both FS and
learning experiences in the workplace including OJT, AE, and changes experi-
enced in the job (learning-by-doing) are determinants of salary growth in China.
However, they may have different roles in human resource development and
their roles may vary in different localities. FS has a significant impact in the East
and Central. In the period under study, for every additional year of FS received,
each employee in the East received a salary increase of an average annual rate of
0.5%; while a person who worked in the Central received an increase of an annual
rate of 1%. Unfortunately, employers in the West do not take FS into much
account when making decisions on increasing salaries. If one received AE during
the five-year period, he/she would get a salary increase at the annul rate of 1.6%
in either the East or Central. In the West, receiving OJT during the five-year
period is likely to get an employee an annual increase in salary of 2.6%. Learning-
by-doing (CHANGE) has an impact on increases in salary of annual rate of 0.2% in
the Central and 0.3% in the West. These are alternatives to accumulate human
capital through continued learning while working.
In a market-based economy, firms introduce new technology or new products
and services and to transform themselves to meet the needs of the market. Firms
are also transforming. Such change in the workplace will impel employees to
continue to learn and thus increase their competence. It is rational for employees
to acquire job-related abilities by way of OJT, AE, and so forth. Learning is there-
fore resulting from change. Our findings indicate that employers in firms at the
county level are aware of such issue and pay attention to the learning experiences
of employees in the workplace. Our findings suggest a transformation in the
human resource development strategies of firms during economic transition. FS
as an educational credential may be used as only as a signal of potential produc-
tivity when employers make hiring decisions. An analysis of the returns to
human capital should include the role of human capital accumulated in the
course of production.
Hypothesis 3 is focused on the role of firm characteristics in salary decisions.
Firm characteristics affect employee salaries in two aspects, at the individual level
as well at the firm level. At the individual level, firms assign employees to a job
position and determine their level of technical proficiency, something that has a
paramount impact on a person’s career. We found that a person’s job position and
level of technical proficiency have a greater impact on the level of his/her salary.
For instance, for the period under study, the coefficients for JOB and RANK on
salary at the initial point are 5.1% and 7.8%, respectively, for the East, 5.1% and
5.3% for the Central, and 5.3% and 7.9% for the West. During the five-year period,
recognition of improved proficiency (IN_GAIN) or a promotion (BE_GAIN)
would get a salary increase at an annual rate of 1% and 1.5%, respectively, for five
years for the employee in the East and the figures were 0.6% and 0.5% for the
Central. In the West, where the economy was lagged far behind, the figures were
1% and 0%. An increase in salary will occur when employee characteristics match
job requirements.
In the overall sample, the variable of region signifies inequality of as much as
49% of the variance for salary at the initial point and 15% of the variance for
salary growth. In comparison, the conventional characteristics of firm (owner-
ship, size, and sector) as well average firm human capital play a very limited role
in determining salaries between regions. However, firm characteristics and
average human capital accumulated may play varied roles within each region.
176 Z. Liu & J. Xiao

The story is that taking ‘China’ as a whole for analysis or policy-making is meth-
odologically erroneous, and will be misleading when interpreting achievements
and issues in the transitional period of development.
Hypothesis 4 concerns varied roles of ET variables (e.g., SF, OJT or AT), indi-
vidual characteristics, and firm characteristic in different regions. Our analysis
demonstrates that regional disparities due to differences in regional development
policies (as discussed in the second section) have the most impact on employee
salaries: that is, the economic zones where firms are located is the major determi-
nant of employee salaries. Altogether, ET variables, the individual variables as
well as firm variables cannot change the distribution of wages to counterbalance
the effects of salary inequalities due to regional disparities in the labor markets.
Within each region, firms may have adopted different strategies to keep
employee competent. The results show that the pattern of salary determination in
each region is different from each other.
Using this multilevel model makes it possible to examine the role of regional
factors, firm-level features, OJT, AE, and job-related individual characteristics in
determining salaries. The multilevel analysis technique accords with social
structures, and the partitioning of random effects is informative in determining
the explanatory powers of variables at different levels, while the reverse trace
study technique is suitable for making comparisons between educational
programs. The findings from our study have some important implications for
future research and policy-making in China.
First, disparities across the three regions in China are overwhelming. This can
be attributed to preferable economic policies for the East and coastal region,
which receive more resources and opportunities for development than the other
regions (Hu et al., 1995, p. 210; Zhang, 2003; Wu, 2004). In the twenty-first century,
the Chinese government should make great efforts to narrow regional gaps by
allocating more government investment to the central and western regions.
Resources can go to infrastructure and social welfare to alleviation of poverty.
That will allow individual workers to capture benefits from economic develop-
ment. On the other hand, we shall not overstress the role of FS in capture private
returns; our study has illustrated that private RORE is dramatically higher in the
regions in which government has poured more public resources and to which it
has given preferable policies.
Second, our analysis provides an empirical case for the argument that human
capital continues to accumulate in the course of production when rapid transition
is undertaken in China. This study suggests that analysis and policy-making on
FS, without referring to other educational programs, would lead to a neglect of
the effects of other forms of ET on the accumulation of human capital. In policy-
making, the provision of ET by local institutions, firms, non-profit organizations,
and so forth should receive more favorable policy guidance. For instance, OJT
and AE for adults, which are flexible alternatives to FS, should be integrated into
socioeconomic policies, along with an expansion and consolidation of compul-
sory basic education. Firms that provide a large amount of OJT or give subsidies
to employees who participate in AE should be accorded a preferential rate of
taxation.
Third, the government should continue to decentralize its administration
especially in the central and western regions. This would allow these laggard
regions to avail themselves of the resources and opportunities of a market-based
economy. A firm’s autonomy in management and production should be
Human Capital Impact on Salary Growth in China 177

respected and guaranteed so that salaries can reflect marginal productivity of


employees, and an optimal allocation of resources in the labor market be
achieved.
In regard to education and training, a high quality of basic education matters.
The government should allocate more resources on compulsory education at the
county level in the Central and the West so that the population in rural will be
able to have an adequate initial education. A working adult without sufficient FS
is at a disadvantage in gaining access to OJT and AE as well as in many other
aspects of work life (Xiao and Tsang, 2004).

Acknowledgements
Both authors contributed equally to this work and their names are stated in alpha-
betical order. This paper is part of a research project entitled ‘Education and
Work: The Efficacy of Schooling in Human Resource Development in Three
Regions in China’, which is partially funded by the Research Grants Council of
Hong Kong (CUHK 4379/00H). Jin Xiao is the principal investigator for this
research project.

Notes
1. The classification of provinces is in accordance with Hu et al. (1995, pp. 18–50).
2. At an exchange rate of 7.4, per-capita GDPs are corresponding to about US$842.6 and US$741.5,
respectively, in the East; US$482.9 and US$466.2 in the Central; and US$322.3 and US$283.7 in the
West.
3. Another indicator also shows clear regional inequality in China. In 1998, per-capita foreign direct
investment in Guangdong was RMB 168, and was RMB 92 in Jiangsu, RMB 22 in Hebei, RMB 16 in
Hubei, RMB 4 in Yunnan, and RMB 8 in Shaanxi (National Bureau of Statistics of China, 1999).
4. Please see Xiao (2004) for specific types of OJT organized by firm.
5. Stratified random sampling was used to select the participating firms and employees. First, we
selected two provinces from each of the three regions, the East, Central, and West as classified by
Hu et al. (1995), using major economic indicators (see discussion about labor market in the second
section). They are: Guangdong and Jiangsu in the East, Hebei and Hubei in the Central, and
Yunnan and Shaanxi in the West. Second, two counties with a medium level of economic develop-
ment in each of the six provinces were selected. They were Jiangmen and Heshan in Guangdong
province, Wuxi and Jiangyin in Jiangsu province, Handan and Cixian in Hebei province, Yichang
and Zhijiang in Hubei province, Mile and Luxi in Yunnan province, and Baoji and Fengxiang in
Shaanxi province. Third, in each county, firms were randomly selected from among those of
different ownership types, sizes, and industrial sectors. Finally, in each firm, one or two intact
workgroups or production lines on the floor were selected to obtain a sample, which consisted of
all of the personnel at a work site, from managerial staff, frontline workers to supporting staff.
Altogether, information was collected from 37 316 employees in 401 firms, corresponding to a
92.0% response rate. To analyze the three-level model, 16 485 individual cases and 365 firms were
used, after deleting cases that contained missing values in any of the variables. Some firms did not
want provide information about the productivity of either the firm or their employees. Therefore,
we lost cases largely due to unwillingness to reveal salary data. Therefore, the data may underesti-
mate the salary of employees during the transitional period.
6. The salary is recalled data and bias can exist. Salaries were universally very low in an egalitarian
planned economy and at the beginning of the economic reforms in 1993. As part of the economic
reforms, increases in salary came to be determined according to proficiency of performance and abil-
ity. This was a big event for everyone in China, as such an approach was novel. Therefore, we assumed
that our sampled employees would best be able to recall their salary in 1993, the baseline point, and
in 1995 when they had their first raise. We also compared our data with the national census data of
1995 as used by Li (2003) and Li and Ding (2003). Our data are very much along the same pattern.
7. The MEE (Mincer, 1974) is: . ln Yi = β 0 + β 1S i + β 2 X i + β 3 X i2 + ε i .
178 Z. Liu & J. Xiao

8. In the West, the variance within the firm was smaller (46.9%) than the variance (53.1%) among
firms for salary growth rate.
9. A study on school culture in the United States showed that between-school variance for leadership
accounted for 23% of the total variance (Rowan et al., 1991, p. 254).
10. The coefficients β02j for initial-point salary in 1993 accompanied with one more year of FS are
0.0165, 0.0108, and 0.0149 in the East, the Central, and the West, respectively.
11. As EXP is the sum of EXPFIRM and EXPOTHER, our model did not include EXP to avoid
multicollinearity.
12. Xiao and Tsang (2004) found that technological change in the workplace is the most important
factor prompting employees to engage in continuous learning on the job, and that such learning
effectively improves individual productivity.
13. In our trial analysis, we put types of OJT as predictors in models, but the coefficients had the similar
results.
14. More than 70% and 60% of employees in the sampled firms in the East and the Central,
respectively, received at least one kind of OJT during the five years, while the figure in the West
was less than 56% (see Table 2).
15. β00j = γ000 + γ001(Region)j+ γ002(Ownership)j + γ003(Section)j + γ004(Size)j + γ005(Av_edu)j + u00j; β10j =
γ100 + γ101(Region)j + γ102(Ownership)j + γ103(Section)j + γ104(Size)j + γ105(Trextent)j + u10j.
16. Conventional characteristics refer to those characteristics that are usually used to describe a firm’s
features, such as geographic location, ownership, size, and sector (Naderi and Mace, 2003 ).
17. Of the three firm characteristics—namely, ownership, sector, and size—ownership is the most
powerful explanatory variable (see columns 3–1b, 3–1c and 3–1d). It can predict an additional 5%
and 9% of the variance in the firm mean salary at the initial point in the Central and the West,
respectively, and 5% of the variance in the firm mean salary growth rate in the Central region.
18. This finding helps to explain why the phenomena of an internal brain drain and floating popula-
tion continue, because moving across regions to the east will automatically result in at least a 50%
increase in salary, according to Table 4. Education might have another screening function as a
ticket to get into more developed labor markets.
19. As the full model is too long, we do not list out all the result. But statistics are available from the
authors.

References
Ashton, D., Sung, J. and Turbin, J. (2000) Towards a framework for the comparative analysis of
national systems of skill formation, International Journal of Training and Development, 4(1), pp. 8–25.
Becker, G. S. (1964) Human Capital (New York: National Bureau of Economic Research).
Belanger, P. and Tuijnman, A. (1997) The ‘silent explosion’ of adult learning, in: P. Belanger and A.
Tuijnman (Eds) New Pattern of Adult Learning: A Six-Country Comparative Study, pp. 1–16 (Oxford:
Pergamon).
Benson, J. and Zhu, Y. (2002) The emerging external labor market and the impact on
enterprise’s human resource development in China, Human Resource Development Quarterly, 13(4),
pp. 449–466.
Bowman, M. J. (1996) Training on the job, in: A. C. Tuijnman (Ed.) International Encyclopedia of Adult
Education and Training (2nd edn), pp. 48–53 (Tarrytown, NY: Pergamon).
Boyle, M. H. and Willms J. D. (2001) Multilevel modeling of hierarchical data in developmental
studies, Journal of Child Psychology and Psychiatry, 42(1), pp. 141–162.
Chen, X. and Chen, L. (2002) A study of rates of return to education in urban China, in: W. Mei (Ed) A
Study on the Operation Mechanism of Higher Education, pp. 473–499 (Beijing: Higher Education Press)
[in Chinese].
China Central Committee (1985) The Chinese Communist Party Central Committee’s decisions on
reforms in the education system, in: Administration Office, State Education Commission (Ed) Educa-
tion in Reforms: Practice and Experience of Educational Reforms in China, pp. 186–198 (Beijing: Higher
Education Press) [in Chinese].
China Central Committee and the State Council (1981) Decisions on strengthening employee training,
CC Document No. 8, 20 February, State Council, Beijing [in Chinese].
China Central Committee and State Council (1993) Guideline on China’s Education Reform and Develop-
ment Decisions on Strengthening Employee Training (Beijing: State Council) [in Chinese].
Cohn, E. and Addison, J. T. (1998) The economic returns to lifelong learning in OECD countries, Educa-
tion Economics, 6(3), pp. 253–307.
Human Capital Impact on Salary Growth in China 179

Hake, B. J. (1999) Lifelong learning in late modernity: the challenges to society, organisations, and
individuals, Adult Education Quarterly, 49(2), pp. 79–90.
Hu, A. et al. (1995) Report on China Regional Disparities (Shenyang: Liaoning People Press) [in Chinese].
Huang, L. H. (2001) Industry-specific human capital and wage determinant in Taiwan, Asian Economic
Journal, 15(1), pp. 19–37.
Johnson, E. N. and Chow, G. C. (1997) Rates of return to schooling in China, Pacific Economic Review,
2(2), pp. 101–113.
Johnston, M. F. (1999) Beyond regional analysis: manufacturing zones, urban employment and spatial
inequality in China, China Quarterly, 157, pp. 1–21.
Jones, D. and Wallis, J. (1992) Adult education in China: an inquiry into reasons for attending, Interna-
tional Journal of Lifelong Education, 111(1), pp. 25–40.
Kang, X. (1999) The Shift in Power: The Reconfiguration of Power in Transition in China, pp. 60-82
(Hangzhou: Zhejiang People’s Publishing House) [in Chinese].
Lai, D. (1998) Education, labor market and earning, Research of Economics, 5, pp. 42–49 [in Chinese].
Lauder, H. (2001) Innovation, skill diffusion, and social exclusion, in: P. Brown, A. Green and H.
Lauder (Eds) High Skills: Globalization, Competitiveness, and Skill Formation, pp. 161–203 (New York:
Oxford University Press).
Levin, H. M. and Kelley, C. (1994) Can education do it alone?, Economics of Education Review, 13(2),
pp. 97–108.
Li, F. and Ding, X. (2003) Empirical study of China’s labor market about using education to improve
productivity in market China’s, Education Research of Qinghua University, 5, pp. 41–45.
Li, H. (2003) Economic transition and returns to education in China, Economics of Education Review,
22(3), pp. 317–328
Meng, X. (1998) Male–female wage determination and gender wage discrimination in China’s rural
industrial sector, Labour Economics, 5(1), pp. 67–89.
Middleton, J., Ziderman, A. and Adams, A. V. (1993) Skills for Productivity: Vocational Education and
Training in Developing Countries (New York: Oxford University Press).
Mincer, J. (1974) Schooling, Experience, and Earning (New York: Columbia University Press).
Naderi, A. and Mace, J. (2003) Education and earnings: a multilevel analysis. A case study of the
manufacturing sector in Iran, Economics of Education Review, 22(2), pp. 143–156.
National Bureau of Statistics of China (1999) China Statistical Yearbook 1999 (Beijing: China Statistic
Press).
Psacharopoulos, G. (1979) On the weak version versus the strong version of the screening hypothesis,
Economics Letters, 4, pp. 181–185.
Psacharopoulos, G. and Patrinos, H. A. (2002) Returns to investment in education: a further update,
Policy Research Working Paper No. 2881, September, World Bank, Washington, DC.
Raudenbush, S. W. (2001) Comparing personal trajectories and drawing causal inferences from
longitudinal data, Annual Review of Psychology, 52, pp. 501–525.
Raudenbush, S. W. and Bryk, A. S. (2002) Hierarchical Linear Model: Application and Data Analysis
Methods (2nd edn) (Thousand Oaks, CA: Sage Publications).
Raudenbush, S. W. and Chan, W. S. (1992) Growth curve analysis accelerated longitudinal designs,
Journal of Research in Crime and Delinquency, 29(4), pp. 387–411.
Rowan, B., Raudenbush, S. W. and Kang, J. (1991) Organizational design in high schools: a multilevel
analysis, American Journal of Education, 99, pp. 238–266.
Schultz, T. W. (1961) Investment in human capital, American Economic Review, 51(1), pp. 1–17.
Schultz, T. W. (1975) The value of the ability to deal with disequilibria, Journal of Economic Literature,
13(3), pp. 827–846.
Spence, M. (1973) Job market signaling, Quarterly Journal of Economics, 87(3), pp. 355–375.
Wei, X. et al. (1999) Education and earning in rural China, Education Economics, 7(2), pp. 167–187.
Wu, D. (2004) Reports on disparities in education in China, Universities Service Centre, the Chinese
University of Hong Kong. Available at www.usc.cuhk.hk [in Chinese].
Xiao, J. (2001) Determinants of employee salary growth in Shanghai: an analysis of formal education,
on-the-job training, and adult education with a three-level model, The China Review, 1(1), pp. 73–100.
Xiao, J. (2002a) Human capital development in a transforming economy: the experience of Shanghai,
China, Current Politics and Economics of China, 3(2), pp. 345–391.
Xiao, J. (2002b) Determinants of salary growth in Shenzhen, China: an analysis of formal education,
on-the-job training, and adult education with a three-level model, Economics of Education Review,
21(6), pp. 557–577.
Xiao, J. (2004) Human Capital in the Developmental Economy: Firms’ Strategies (Beijing: Beijing Normal
University Press).
180 Z. Liu & J. Xiao

Xiao, J. and Tsang, M. C. (1999) Human capital development in an emerging economy: the experience
of Shenzhen, China, The China Quarterly, 157, pp. 72–114.
Xiao, J. and Tsang, M. C. (2004) Determinants of participation and non-participation in job-related
education and training in Shenzhen, China, Human Resource Development Quarterly, 15(4), pp. 389–420.
Zhang, Y. (2003) China rural research, Universities Service Centre, the Chinese University of Hong
Kong. Available at http://www.ccrs.org.cn; paper section, www.usc.cuhk.hk [in Chinese].
Ziderman, A. and Horn, R. (1995) Many paths to skilled employment: a reverse tracer study of seven
occupations in Colombia, Education Economics, 3(1), pp. 61–79.

Vous aimerez peut-être aussi