ARTICLE IN PRESS
Journal of Development Economics xx (2003) xxx – xxx
www.elsevier.com/locate/econbase
Skill differentials, return to schooling, and market
segmentation in a transition economy:
the case of Mainland China
Belton M. Fleishera,*, Xiaojun Wangb,1
a Department of Economics, The Ohio State University, 1945 N. High St., Columbus, OH 43210, USA
b Department of Economics, University of Hawaii, Honolulu, HI 96822, USA
Received 1 February 2001; accepted 1 December 2002
Abstract
A body of existing research attributes evident underpayment of workers and low private returns to
schooling in China through the mid-1990s to the persistence of labor-market monopsony. We find
that rural enterprises overpay production workers relative to a monopsony profit-maximizing
benchmark, while there is extreme underpayment of skilled workers relative to the monopsony
profit-maximizing amount. This relatively large ‘‘exploitation’’ of skilled workers explains, in a
proximate sense, low private returns to schooling.
D 2003 Elsevier B.V. All rights reserved.
JEL classification: P23; J24; J31; D33; O15
Keywords: Wages; Return to schooling; Market segmentation; Productivity; Transition economies; Chinese
economy
1. Introduction
A persistent puzzle in China’s economic evolution from reform through at least the
mid-1990s is that wage differences by level of skill, occupation, and/or schooling
remained very narrow and returns to higher education remained low in comparison with
those in other countries,2 both industrialized and industrializing, and when compared to
* Corresponding author. Tel.: +1-614-292-6429.
E-mail addresses: fleisher.1@osu.edu (B.M. Fleisher), xiaojun@hawaii.edu (X. Wang).
1 Tel.: +1-808-956-7721.
2 See, for example, Jamison and Van Der Gaag (1987), Dessi (1991), Byron and Manaloto (1990), Fleisher
et al. (1996), Gregory and Xin (1995), Maurer-Fazio (1997), Maurer-Fazio et al. (1999), Psacharopoulos (1985),
Wang et al. (1995), Knight and Li (1996), Li and Zhang (1998), Zax (1994), and Yang and An (1997).
0304-3878/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.jdeveco.2002.12.002
DEVEC-00987

---

ARTICLE IN PRESS
2
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
those in some smaller transition economies, including, for example, the Czech Republic
(Munich et al., 2000), Slovenia (Orazem and Vodopivec, 1995), and Bulgaria (Jones and
Ilayperuma, 1994). Although returns to higher education in the Russian Republic are
among the lowest in the world, this can in large part be attributed to the extraordinarily
high proportion of college graduates in Russia (over 20% of individuals aged 25 – 64 in
1995), which is nearly equal to that in the United States and higher than the average for
OECD countries. (Sheidvasser and Benı´tez-Silva, 2000). It is difficult to attribute the low
return to higher education in China to a super-abundance of college graduates, because the
proportion of graduates of 4-year universities in China in the population 16 years of age
and older was less than 1% in 1997 (Statistical Yearbook of China, 1998).
It is extremely unlikely that low private returns to education in China have reflected a
low marginal product of labor. Although the return to schooling in the agricultural sector
when measured in terms of productivity or the profit of family enterprises does appear to
be relatively low (Yang and An, 1997; Yang, 2000), this is not the case in Chinese
industry, and low private returns to schooling are found in both sectors. Not only have
significant gaps between wages and the estimated marginal product of labor been reported
in a number of studies3, but also, and of critical importance to the main focus in this paper,
the ratio of the marginal product of highly educated workers to that of other workers
appears to be much higher than the ratio of their rates of pay or earnings. (Previous
research showing this inequality is reported in Fleisher et al., 1996; Fleisher and Chen,
1997.)
In this paper, we address the puzzle of this differential wage gap and attempt to learn
why, in a society where educated labor is relatively very scarce, its remuneration
remained far below its contribution to productivity, more than a decade and a half after
reforms began. The paradigm of labor-market monopsony provides a useful organizing
framework for our research, because in rural China, it may exist in small labor markets
dominated by one or a very few rural enterprises (Dong and Putterman, 1996). We
extend previous research by tying the monopsony approach to skill- or schooling-based
earnings differentials.
In Section 2 we provide evidence on the gap between the marginal product of labor and
wages. We first estimate marginal products of two classes of workers: (1) production
workers and (2) technical, administrative, and staff (TAS) workers in the context of a
production function that allows for effort-enhancing wage payments. We then calculate
wage gaps (marginal product – wage ratios) for both groups of workers. The larger wage
gap for TAS workers ‘‘explains’’ in a proximate sense observed low returns to schooling.
Next, we explore alternative explanations of the estimated wage gaps. We find that the
wage gaps cannot be simply explained in terms of profit maximization under monopsony.
We then explore the relationship between the wage gaps and variables likely to be
3
Examples include Dong and Putterman (1996), Xu (1991, 1995), Dong and Putterman (2000a,b), Parker
(1999), Pitt and Putterman (1999), and Yang and Zhou (1999). See also Svejnar (1990), Hay et al. (1994), and
Jefferson and Rawski (1994). Jefferson et al. (1992) report estimated nominal marginal products of labor in 1988
to be 2974 and 1648 yuan for enterprises and collectives (urban and TVE’s), respectively. The China Statistical
Yearbook 1991 reports average annual wages of staff and workers in state-owned enterprises in 1988 to be 1853
yuan and in urban collectives to be 1426 yuan.

---

ARTICLE IN PRESS
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
3
associated with the degree of monopsony power and estimated firm returns to scale. The
variables associated with monopsony power do not satisfactorily explain variation in the
wage gaps as well as does the estimate of returns to scale. We infer that the monopsony-
profit maximization hypothesis alone cannot explain the wage gaps but that some form of
worker-mobility restrictions must be invoked to explain the continued ‘‘exploitation’’ of
both production and TAS workers.
Section 3 concludes and derives implications for economic reforms and the effects of
institutional rigidities on economic growth and the distribution of wages and incomes.
2. Production, labor, schooling, and wages
In order to derive estimates of the marginal product of two classes of workers, we
estimate the following augmented Cobb – Douglas production function:
Y ¼ KaðjðejLjÞbjÞexpð/Z þ eÞ
ð1Þ
j
where: Y = value added; K = net capital stock; Lj = labor of the jth group; j = 1, 2; ej = effort
function for the jth group of employees; Z = vector of dummy variables; e = an iid
disturbance, and a and b are elasticities of output with respect to capital and (various
types of) labor.
The effort function is defined as


W
gj
j
ej ¼ Bj
;
ð2Þ
Waj
where B and g are parameters controlling for baseline effort and responsiveness of effort
with respect to the wage premium, respectively. Wj is the observed wage of the jth group of
employees, and Waj the estimated spot-market competitive wage. Details on how to obtain
Waj is explained later.
Augmentation of the conventional production function with an effort function serves to
capture one of the key elements of Chinese labor market reform on production, namely, to
break with egalitarianism and introduce material incentives. This idea is conformable with
the basic notion of an efficiency wage. Our specification reflects the reality that production
workers and technical workers perform different tasks in a firm. Since we distinguish these
two types of workers, different efficiency wage structures can be recovered for each group.
The data set is a panel survey of 200 large rural enterprises (mostly TVEs) for the years
1984 – 1990.4 The survey covers 20 enterprises in each of 10 provinces. The 10 provinces
are Anhui, Hubei, Guangdong, Jiangsu, Zhejiang, Sichuan, Hebei, Liaoning, Shanxi, and
Gansu. We define Hebei, Guangdong, Liaoning, Jiangsu, and Zhejiang as coastal
provinces and the rest as non-coastal provinces. The survey not only includes quantitative
statistics about the individual firms, but also provides important environmental statistics
4
We are grateful to Dennis Yang, Yaohui Zhao, Xiao-yuan Dong, Isabelle Perrigne, and Gary Jefferson for
their help in obtaining and using these data.

---

ARTICLE IN PRESS
4
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
Table 1
Production function (dependent variable: ln Y a)
Variable
Description
(1) OLS
(2) OLS
(3) IV
(4) IV
Constant
À 1.33 ( À 6.35) À 0.79 ( À 3.96) À 0.99 ( À 4.20) À 0.58 ( À 2.55)
K
ln net capital stock
0.42 (13.73)
0.40 (12.68)
0.50 (15.96)
0.48 (14.69)
L
ln total employment
0.74 (16.65)
–
0.68 (13.81)
–
PW
ln production
–
0.49 (9.89)
–
0.46 (7.93)
workers
TAS
ln technical/
–
0.30 (5.91)
–
0.31 (5.06)
administrative staff
MT
efficiency wageb
0.36 (6.21)
–
0.34 (5.53)
–
MPW
efficiency wageb
–
0.20 (3.17)
–
0.17 (2.50)
for production
workers
MTAS
efficiency wageb
–
0.17 (2.53)
–
0.20 (2.74)
for technical/
administrative staff
YR85
year 1985 dummy
0.24 (2.06)
0.25 (2.18)
–
–
YR86
year 1986 dummy
0.03 (0.25)
0.05 (0.47)
À 0.31 ( À 2.75) À 0.28 ( À 2.51)
YR88
year 1988 dummy
0.49 (4.26)
0.49 (4.27)
0.02 (0.23)
0.03 (0.28)
YR89
year 1989 dummy
0.52 (4.46)
0.51 (4.36)
0.08 (0.72)
0.07 (0.61)
YR90
year 1990 dummy
0.63 (5.48)
0.63 (5.43)
0.17 (1.56)
0.17 (1.49)
Number of
988
979
831
822
observations
Adjusted R2
0.60
0.60
0.61
0.61
T-statistics are in parentheses. The 1987 observations are dropped due to inadequate data. So the nominal
number of observations should be 1200, the discrepancies reflect missing values. In the IV estimation lagged
labor (1 year), for each group, respectively, is used as instruments. Thus, the 1984 observations are dropped.
That explains the further loss of observations in the IV results.
a Y is value added.
b See footnote 4.
describing the markets in which the firms operates. For example, not only is there data on
an enterprise’s employment, but there is also information on total employment in the
village where the firm is located. This allows us to test hypotheses on the impact of market
structure on the behavior of the firm. Table 1 shows the results of estimating Eq. (1).
Column (1) is based on all labor aggregated into one category, while column (2) is based
on two categories of labor, production workers, and technical/administrative staff. The
estimated labor and capital elasticities reported in Table 1 are close to the ordinary least
squares (OLS) results reported by Pitt and Putterman (1999) and to the GLS estimates
reported by Dong and Putterman (1996) using the same data set. There is evidence of
unexploited scale economies. The estimated coefficient of the efficiency-wage variable is
highly significant for all workers and for production and TAS workers separately.5 Since
5
We have defined the efficiency wage to be the deviation (in ratio form) of the worker’s actual wage to the
average wage paid to that type of worker (production worker or technical and administrative staff worker) in the
same province. The ratio of the estimated efficiency-wage coefficients to their respective production – labor
elasticities is much smaller than unity, and this is true for all workers taken together as well as for production and
TAS workers separately. Taken at face value, these coefficients suggests that wage rates are set higher than their
profit-maximizing level under an efficiency-wage scenario.

---

ARTICLE IN PRESS
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
5
the employment decision is endogenous with output, we also instrument labor with its own
value lagged one period; column (3) reports the IV results for aggregated labor, while
column (4) reports the IV results for disaggregated labor. The estimation results are robust
to the form of estimation used.
When the OLS elasticity estimates are used to calculate marginal products of
labor, we find that for an intermediate year of the sample period, 1988, the average
marginal product of all workers taken together was 5846 yuan, while the average rate
of pay was only 1670 yuan annually, indicating a wage gap ratio of 3.50. When
workers are disaggregated into production and TAS categories, the wage gaps in
1988 are 3.09 and 9.18, respectively. We proceed to investigate the reasons for these
gaps.
2.1. Does monopsony power affect wage gaps?
An important hypothesis that has been proposed to explain the positive MPL-wage
gap in Chinese rural industry is monopsony power. Dong and Putterman (1996) argue
that monopsony power of rural employers in the presence of restrictions on intercom-
munity migration can explain the wage gap in rural industry. They ask how labor-
supply curves facing Chinese rural enterprises can be upward-sloping when it is said
that there is pervasive unemployment in rural China. They provide the following
answers:
1. Wu (1993) reports that about 80% of TVEs were located in villages rather than in
the town centers or municipal areas of their counties. This location strategy
coupled with primitive communications and transportation infrastructure create
sufficient conditions for monopsony under the assumption that there are economic
and/or political limitations on the number of enterprises that locate in smaller
villages.
2. The hukou system makes it difficult to change permanent residence, even among rural
villages. Rural – urban migration is even more costly and disruptive, leaving a
substantial rural – urban income gap, even for industrial workers who have left farm
employment (also, see Zhao, 1995).
3. Travel from one village to another is expensive, and disruptions of family lives are
psychologically costly.
4. They report that during 1986 – 1990, only 28.3% of workers employed in TVEs came
from other villages.
5. Skilled workers are in short supply, even if ‘‘common’’ labor experiences considerable
underemployment.
Zhang et al. (2002) note that Mallee (2000) and Yang and Zhou (1999) demonstrate that
a number of barriers, such as land tenure arrangements and mandatory marketing
delivery quotas, have increased the cost of out-migration and dampen off-farm labor
market participation. It appears plausible to us that the supply of TAS workers is less
elastic than that of production workers, because there is surely a smaller pool of
educated workers from which to attract new hires. Moreover, the hukou system, when

---

ARTICLE IN PRESS
6
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
Table 2
Elasticity of labor supply
Elasticity
Profit-maximizing gap
Production Workers
0.19 (4.10)
6.26
TAS
0.30 (6.67)
4.33
T-statistic in parenthesis. Labor supply function is estimated by regressing log employment on log wage, year
dummies, worker type dummies, interaction terms between wage type. On the basis of an F-test, we cannot
reject the hypothesis that labor-supplied elasticities are constant over time for both production TAS workers.
However, we can reject the hypothesis that the elasticity for production TAS workers are equal with p-value
equal to about 0.1.
applied to educated workers, imposes relatively more severe limits on the portability of
their human capital.6
In order to establish the role of monopsony in explaining observed wage gaps, we use
as a benchmark the degree of monopsonistic ‘‘exploitation’’ that would occur under profit-
maximizing behavior. An obvious direct test of the joint hypotheses that the wage gaps we
and others have estimated reflect profit-maximizing monopsony requires knowledge of
labor-supply elasticities from which profit-maximizing wage gaps can be derived.7
Estimated labor-supply elasticities are reported in Table 2. We assume that any exogenous
shifts in communities’ labor-supply functions are captured by year dummy variables and
their interaction with the wage regressor. A single estimated elasticity for each class of
workers is reported in Table 2. The highly significant elasticities imply profit-maximizing
wage-gap ratios of approximately 6.3 and 4.3 for production workers and TAS workers,
respectively. Our estimation results with the complete set of year-interaction terms imply
that there was no statistically significant trend in the labor-supply elasticities over time, but
that there is a statistically significant difference between the labor-supply elasticity of TAS
workers and that of production workers.
Comparisons of profit-maximizing with estimated wage gaps are shown in Table 3 for
each year, 1985 – 1990, except 1987. [The magnitudes of the wage gaps reported in Table 3
are very close in magnitude to those reported by Xu (1995, p. 36), where the marginal
product estimates are based on provincial aggregate data for TVEs during about the same
period as the data in our sample.] In 5 of the 6 years for production workers, the estimated
profit-maximizing wage gap is larger than the observed wage gap, implying that enter-
prises place a positive value on employment in addition to profit. Thus, the conventional
6
Zhao (1999) reports that in a 1995 household survey conducted in Sichuan province, rural nonfarm
nonmigrant workers had higher schooling levels than did those who had migrated out of their local areas, even
though they contributed less to household incomes than comparable workers who migrated, implying that private
benefits, at least, are greater for migrants. Zhao infers that the uncertainty of employment without hukou,
transportation, lodging, and psychic costs of migration outweighed immediate economic gains. Zhao’s results
may not to apply to all of China, though, as she (see Zhao, 1995, 1999) does report that higher levels of schooling
significantly increased the probability of obtaining a permanent hukou for migrants to Beijing. Consistent with
Zhao’s research, Zhu (2002) finds, based on a 1993 survey in Hubei province, that the effect of education on the
income level of migrants was less than for nonmigrants. Even as late as 1995, Cai et al. (2002) view rural
outmigration as a three-step process, and one that does not transcend regional boundaries.
7
The profit-maximizing wage gap is 1 + 1/g, where g is the elasticity of supply.

---

ARTICLE IN PRESS
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
7
Table 3
Observed profit-maximizing wage gap
MPL/wage
Year
Observeda (MPL/wage)
Profit
maximizingb
All workers
1984
2.56 (2553/997)
5.75
1985
2.51 (3681/1466)
7.25
1986
2.72 (3554/1307)
3.04
1988
3.50 (5846/1670)
3.86
1989
3.78 (6577/1740)
3.08
1990
4.84 (7277/1504)
11.00
Production workers
1984
2.10 (1855/883)
5.76
1985
1.86 (2701/1452)
8.14
1986
2.27 (2716/1196)
3.44
1988
3.09 (4520/1463)
4.13
1989
3.31 (5108/1543)
4.57
1990
3.98 (5765/1448)
11.00
TAS
1984
8.44 (12042/1427)
6.88
1985
10.08 (17176/1704)
4.13
1986
7.24 (16073/2220)
4.03
1988
9.18 (23872/2600)
4.13
1989
9.44 (25103/2659)
4.13
1990
12.77 (27246/2134)
6.55
a The observed gap is the calculated marginal product of labor using the estimates from the OLS production
function divided by monthly wages (shown in parentheses).
b The profit-maximizing gap is 1 plus the inverse of the elasticity of labor supply. This is the ‘‘hypothetical’’
gap if the enterprise is indeed a profit-maximizing monopsony.
wisdom that overemployment is the rule in Chinese enterprises is supported against the
standard implied by the joint hypotheses of monopsony and profit maximization.8 On the
other hand, the estimated profit-maximizing wage gap is uniformly lower than the
observed wage gap for TAS, implying underemployment of this class of workers. Bearing
in mind that the annual differences in the estimated profit-maximizing gaps are highly
insignificant, it is interesting to note that the observed gaps tend to drift upward, implying
perhaps a reduced tendency over time to ‘‘overemploy’’ production workers and an
increased tendency to ‘‘underemploy’’ TAS workers.
2.2. Scale economies and the division of output
The difference between the wage gaps of production and TAS workers cannot be
explained in terms of simple profit maximization under monopsony. Perhaps production
workers tend to benefit from political favoritism, but we have no independent evidence
that this is true. A possibly important additional consideration is suggested by the
estimated production functions reported in Table 1, which indicate that the typical rural
collective operates in the range of increasing returns to scale. An implication is that in the
absence of a subsidy or an entrepreneur with deep pockets, it is impossible for all factors to
8
For another approach to analyzing Chinese enterprises’ goals in choosing between profits, wages, and
employment, see Svejnar (1990) and Pitt and Putterman (1999).

---

ARTICLE IN PRESS
8
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
be paid the value of their marginal products. ‘‘Underpayment’’ of at least some factors is a
mathematical necessity. It is intriguing, therefore, to explore the extent to which the wage
gaps we have estimated are associated with the severity of this ‘‘adding-up’’ problem.
To do this, we respecify the rural-enterprise production function in terms of gross
output, with intermediate inputs included among the right-hand variables, as follows:
lnGY ¼ Constant þ gKlnK þ gPlnPW þ gT lnTAS þ gRlnRM
ð3Þ
where GY is gross output, K is net value of fixed capital, PW is number of production
workers, TAS is number of technical workers, and RM is raw materials. The g’s are
corresponding parameters. Respecification of the production function in terms of gross
output with intermediate inputs included among the regressors is appropriate because we
are looking for the possible role of increasing returns to scale in explaining wage gaps, and
we should not assume constant returns to raw materials. Moreover, we are searching for
the ultimate ‘‘residual claimants,’’ and the intermediate-input provider is certainly a
legitimate candidate.
We then derive the ratios of marginal products to factor payments as
MPi
g
¼ i
ð4Þ
WAGEi
Si
where i indexes each of the four groups of factors of production, Si is the share of payment
to this group in GY.9 (The return to capital is defined to be the gross value of output less
payments for intermediate inputs and wages. Thus, return to capital includes all reported
accounting profits, taxes, and interest.) The estimated gaps are then regressed on estimated
returns to scale ( d
RTS ).10 In order to obtain large enough samples to estimate reliable
production-function parameters, we group the data by province and year, obtaining 60
samples within which Eq. (3) is estimated, yielding 60 estimates of each production-
function parameter, which are the basis for estimating second-stage equations in which
factor-payment-marginal product gaps are regressed on estimated returns to scale.
The results of the second-stage estimates are reported in Table 4. They are consistent
with the following interpretation. The payment gap for intermediate inputs is not
associated with estimated returns to scale. Intermediate-input providers must be paid
market prices and do not receive lower payments from unprofitable enterprises. The
payment gap we attribute to ‘‘capital’’ is weakly and negatively correlated with returns to
scale. Thus, the hypothesis that providers of nonlabor inputs act as residual claimants as a
group can be rejected. In contrast, the payment gaps for production and technical workers
9
This can be verified as follows. Take production workers as example. Multiply the numerator and
denominator of the right-hand side of Eq. (4) by GY/PW, then it becomes
GY
g PW
GY
S PW
The numerator now has the interpretation of the marginal product of production workers, while the denominator is
average wage paid to them.
10
Estimated returns to scale are the predicted values from regression of returns to scale to a vector of
instrument variables. These variables include capital, two types of labor, raw material, 5-year dummies and nine-
province dummies.

---

ARTICLE IN PRESS
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
9
Table 4
Marginal product – wage gap and returns to scale (N = 60)
Marginal product – wage gap
K
PW
TAS
RM
Constant
1.154 (3.19)
À 2.728 ( À 1.48)
À 31.054 ( À 2.03)
0.684 (1.61)
d
RTS
À 0.553 ( À 1.66)
5.103 (2.99)
45.050 (3.19)
0.100 (0.26)
Adjusted R2
0.03
0.12
0.13
À 0.02
T-statistics in parentheses. d
RTS is the predicted returns to scale from regression of returns to scale on a vector of
instrument variables. See footnote 9 for details.
are both positively and significantly correlated with returns to scale, with the regression
coefficient for technical workers being about eight times larger than that for production
workers. This is consistent with the hypothesis that both groups, particularly technical
workers, are de facto residual claimants in the presence of unexploited scale economies.
The sociopolitical forces that lead to this division of output are not obvious, but it is clear
that this ‘‘exploitation’’ of labor would be impossible in the absence of restrictions on
worker mobility. In other words, it is consistent with a form of monopsony wage setting.
2.3. Monopsony, scale economics, and wages
To gain further insight into determinants of the wage gaps for production and TAS
workers, we regress the mean wage gap for production and TAS workers, respectively, on
the following variables: estimated returns to scale, local employer-concentration ratios and
available land per worker; provincial measures of foreign direct investment per worker,
and unemployment.11 We hypothesize that under monopsony, the estimated coefficients of
estimated returns to scale and employer concentration will be positive and that of
unemployment will be negative. The rationale is that increasing returns to scale preclude
‘‘full’’ payment to all factors, with labor being ‘‘exploited’’ under monopsony; employer
concentration is an indirect measure of monopsony power; while higher unemployment
will increase the elasticity of labor supply. The estimated coefficient of the land – labor
ratio is uncertain under the profit-maximization-monopsony joint hypothesis, because,
while more land per person should increase agricultural labor productivity, the effect on
the elasticity of marginal product with respect to labor (and, hence, on the elasticity of
labor supply) is ambiguous.12 The foreign-investment variable is included to represent
funds available in an environment of very imperfect financial markets. Given that
estimated returns to scale is included in the regression, the estimated net relationship
between FDI and the wage gap may be interpreted as the effect of ‘‘ability to pay’’ on
wages. We take a negative coefficient for the foreign-investment variable to be consistent
11
Land per worker is available in the local community data. Foreign investment per worker is obtained from
the Statistical Yearbook of China. Unemployment estimates are reported in Liu (1997) and are based on an
estimate of available labor force minus the sum of workers required to operate family farms and nonagricultural
employment.
12
Under the assumption of a Cobb – Douglas production function, an increase in the labor – land ratio shifts
the labor supply function to the nonagricultural sector upward, but leaves the elasticity unchanged.

---

ARTICLE IN PRESS
10
B.M. Fleisher, X. Wang / Journal of Development Economics xx (2003) xxx–xxx
Table 5
Augmented monopsony regression
Variable
Description
Production workers
Technical/
administrative staff
Constant
À 0.062 ( À 0.027)
À 27.804 ( À 1.451)
MCR
ln concentration ratioa, group median
À 2.099 ( À 0.283)
À 47.131 ( À 0.758)
MMPC
ln acreage per person, group median
À 0.405 ( À 1.510)
2.409 (1.071)
MFDL/LF
foreign direct investment
À 0.025 ( À 1.385)
À 0.077 ( À 0.500)
per labor force, group median
MUNEM
unemployment rate, group median
À 4.8167 ( À 0.806)
73.476 (1.464)
d
RTS
predicted returns to scale
4.977 (2.719)
35.079 (2.282)
Number of
60
60
observations
Adjusted R2
0.13
0.13
The dependent variable is the gap between marginal product of labor – wage rates. T-statistics in the parentheses.
a This is the employment share of this enterprise among all industrial enterprises in the township or village.
with the hypothesis that capital constraints increase wage pressure in the presence of
excess labor supply and monopsony power.
Table 5 presents the results, which are based on the same 60 observations as used for the
regression reported in Table 4. The adjusted R2 equals 0.13 in both the production workers
and TAS regressions. The estimated regression coefficient of the concentration ratio is
statistically insignificant for both production and TAS workers, which is inconsistent with
the joint hypotheses of monopsony and profit maximization. The coefficient of land per
worker is negative and marginally significant for production workers, but positive with
little significance for TAS workers. Unless increased land per worker lowers the elasticity
of supply of production workers, it is difficult to see how the estimated coefficients of the
land – labor variable support the joint hypotheses of monopsony and profit maximization.
The coefficient of foreign direct investment is negative, which is consistent with the
hypothesis that a higher level of FDI allows firms to pay wages more closely approximating
marginal product, although its t value is not large for production workers and is very low
indeed for TAS. The coefficient of unemployment is negative and insignificant for
production workers and positive and marginally significant for TAS.
Table 6
Sample statistics for enterprises (N = 200)
Gross output
Net capital
Employment
Unit
10,000 yuan
10,000 yuan
person
1984
336.139 (593.218)
65.183 (88.725)
298 (309)
1985
486.576 (849.214)
96.718 (132.202)
369 (354)
1986
532.570 (1025.870)
162.001 (289.581)
419 (472)
1987
624.188 (1231.110)
237.213a (509.598)
409 (521)
1988
631.756 (1018.180)
171.432 (277.068)
387 (439)
1989
709.366 (1262.700)
209.628 (358.718)
382 (438)
1990
757.204 (1317.700)
214.514 (353.842)
374 (481)
Standard deviation in the parentheses.
a Original price of fixed capital.

---

Document Outline

• Skill differentials, return to schooling, and market segmentation in a transition economy: the case of Mainland China
  • Introduction
  • Production, labor, schooling, and wages
    • Does monopsony power affect wage gaps?
    • Scale economies and the division of output
    • Monopsony, scale economics, and wages
  • Conclusion and outline for further work
  • Uncited references
  • Acknowledgements
  • References

---