You are on page 1of 9

Utilities Policy 10 (2001) 75–83

www.elsevier.com/locate/utilpol

A data envelopment analysis of the efficiency of China’s thermal


power generation
Pun-Lee Lam a,∗, Alice Shiu a
a
Department of Business Studies, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong Special Administrative Region,
China

Received 11 January 2002; received in revised form 2 September 2002; accepted 8 September 2002

Abstract

This study applies the data envelopment analysis (DEA) approach to measure the technical efficiency of China’s thermal power
generation based on cross-sectional data for 1995 and 1996. Our results show that municipalities and provinces along the eastern
coast of China and those with rich supplies of coal achieved the highest levels of technical efficiency. There is no clear evidence
of excess capacity. However, the presence of labor slack in many regions indicates that labor redundancy was a serious problem.
In our second stage regression analysis, we find that fuel efficiency and the capacity factor significantly affect technical efficiency.
Provinces and autonomous regions that were not under the control of the State Power Corporation (SPC) achieved higher levels
of efficiency. The presence of foreign investment, however, did not have a significant effect on efficiency.
 2002 Elsevier Science Ltd. All rights reserved.

JEL classification: L5; Q4

Keywords: DEA approach; technical efficiency; China’s thermal power generation

1. Introduction sector are lacking, mainly due to insufficient data about


the factor inputs that are used in the power sector. In
The electric power industry in China has been under recent years, the Chinese authorities have begun to pub-
state ownership and control since 1949. At present, the lish more information about the production and perform-
state sector owns all of the transmission lines and nearly ance of the power sector over time, and across different
all of the distribution networks. It also accounts for most regions. With more information available, the way is
of the power generation. Since the 1980s and the adop- opened for efficiency analyses of China’s power sector.
tion of an open-door economic policy, structural reforms, China relies heavily on thermal power generation,
market incentives, and decentralization policies were which accounts for three-quarters of the total generating
introduced to attract investment into the power sector. capacity. In this study, we apply data envelopment
Local governments and power enterprises co-operated analysis (DEA) to measure the cross-sectional efficiency
with foreign investors to form independent power pro- of China’s thermal power generation. We also conduct a
ducers (IPPs). From that time, generating capacity in second stage regression analysis to identify the important
China has expanded rapidly. factors that affect the performance of thermal power gen-
Today, China is the world’s second largest producer eration in China. Our empirical results shed light on the
of electric power, both in terms of installed capacity and environmental factors that affect generation efficiency,
actual generation. Despite its growing importance, which will provide useful information for the restructur-
empirical studies on the productivity of China’s power ing of the generation market in China.
This paper is organized as follows. In Section 2 we
outline the current structure of the electric power indus-

Corresponding author. Tel.: 852-2766-7123; fax: 852-2765-0611. try in China. In Section 3 we review the results of the
E-mail address: bupllam@inet.polyu.edu.hk (P.-L. Lam). efficiency measurement of power industries in various

0957-1787/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 7 - 1 7 8 7 ( 0 2 ) 0 0 0 3 6 - X
76 P.-L. Lam, A. Shiu / Utilities Policy 10 (2001) 75–83

countries. Section 4 describes our methodology, and we such as housing, education, and medical care (Zhu,
report our results in Section 5. The last section provides 1999). As a result of this social burden, SOEs in China’s
conclusions and discusses the implications of our empiri- electric power industry are grossly unproductive. At
cal results for China’s thermal power generation. present, China’s power sector employs more than 2
million workers, and the problem of redundancy is
very serious.
2. China’s power industry In recent years, the power industry’s SOEs have been
corporatized and converted into “independent holding
China’s electric power industry came into existence companies”, which are responsible for the economic
in 1882, with the establishment of the country’s first gen- management and planning of their respective regions. At
eration plant in Shanghai (Ministry of Energy, 1989; the provincial level, power corporations are separated
Contemporary China Press, 1994; Economic Manage- from the power bureaus. Since restructuring, all power
ment Press, 1997). In the early years of the industry’s corporations, be they at the regional or provincial level,
development, electric power systems expanded very operate under commercial principles and no longer
slowly. Most power generating units were of small receive large government subsidies. To obtain new capi-
capacity with low thermal efficiency, and were concen- tal for expansion, some power companies have issued
trated in a few large cities in the coastal regions. Power shares to private shareholders, and publicly traded shares
supply to the interior provinces was very limited are available to foreign investors. It is hoped that corpor-
(Ministry of Energy, 1989). atization and listing on the stock market will create
Since 1949, electric power in China has been under greater pressure on SOEs in the power sector to
state ownership and control. The state sector presently improve efficiency.
owns all transmission networks. From the 1950s to In January 1997, the State Council formed the State
1970s, the transmission networks expanded from the Power Corporation of China (SPC), which is the largest
coastal regions to the interior areas, and from the cities power corporation in China. The functions of running
to the extensive countryside. In the 1980s and 1990s, state-owned assets and enterprise management that were
after China adopted an open-door economic policy, formerly undertaken by the Ministry of Electric Power
structural reforms, market incentives, and decentraliz- (MEP) were transferred to the SPC upon its formation.
ation policies were introduced to attract investment into The SPC is now responsible for the construction of
the power sector. Since then, the country’s installed gen- power networks nationwide, as well as the operation and
erating capacity and total annual generation have management of inter-regional and inter-provincial power
increased rapidly. During the past decade, the Chinese grids and large-scale power plants. Most of the regional
government has allowed foreign companies to participate and provincial power corporations are under the control
in building up the country’s generating capacity. Despite of the SPC, except those in Guangdong, Hainan, Inner
the remarkable growth in electricity generation that has Mongolia, and Xizang (Tibet), which are controlled by
been experienced over the last two decades, the per cap- their provincial governments.
ita consumption of electricity in China is still low when Thermal power and hydropower have long been Chi-
compared to consumption in developed countries. Fur- na’s dominant generating capacities. In 1999, they
thermore, distribution is uneven; while many areas of accounted for 75.7% and 23.5% of the total capacity
the country continue to suffer from power shortages, a respectively (China Electric Power Information Center,
few areas are oversupplied. 2000). The share of nuclear power in total capacity was
To date, there are seven regional power networks and only 0.8%. Although China has a large potential for hyd-
five independent provincial power grids in China. The ropower, less than 20% of it is exploited. Of the thermal
generation, transmission, and distribution of electricity generating capacity, approximately 90% is coal-fired,
in different areas are vertically integrated and under the and the combined share of oil and gas fired capacity is
management of a public monopoly. Regional and prov- only 10%.
incial state-owned power enterprises are responsible for In the 1980s and the early 1990s, the local govern-
the construction, management, and operation of electric ments in China constructed a lot of small power plants
power systems. with a capacity less than 50 MW, which did not require
The system of state-owned enterprises (SOEs) was approval from the central government (Wirtshafter,
established after the Communist Party took power in 1990; Yang and Yu, 1996). This policy of decentraliz-
1949 (Chow, 1997). Since then, the state sector has ation provided incentives and flexibility for local govern-
dominated the economy and suffered chronic problems ments to expand generating capacity quickly. However,
of mounting losses, huge borrowing, and low pro- the wide use of low quality fuel for power generation
ductivity. SOEs have had to shoulder the social responsi- and the proliferation of small generating units reduced
bility of creating employment, and providing employees the thermal efficiency of China’s energy sector. The
with various kinds of fringe benefits and service facilities average thermal efficiency of electricity generation in
P.-L. Lam, A. Shiu / Utilities Policy 10 (2001) 75–83 77

fossil fuel plants ranges between 27% and 29%, com- the SFA is a parametric approach. The DEA approach
pared to around 38% in Organization for Economic does not require the arbitrary assumptions regarding the
Cooperation and Development (OECD) countries functional form or the distributional form of the error
(Economic Management Press, 1997; Financial Times terms that the SFA approach uses,2 and it is more suit-
Energy, 1999). The development of hydropower and the able for handling the multiple inputs and outputs of the
improvement in the energy efficiency of coal-fired plants power industry. Both approaches are susceptible to data
will help to improve the economic and energy efficiency measurement errors, but the SFA approach has
of China. additional problems of specification errors and omitted
variables.
Since the early 1980s, economists have applied the
3. Previous productivity studies of power DEA approach to measure the productivity performance
generation of electric utilities under different ownership structures.
Instead of using cost and production functions, they
Since the early 1960s, there has been a growing inter- specify and construct a piecewise linear technology from
est in measuring the productivity and efficiency of the the observed input and output data. Productivity
power sector. Different methods have been used to mea- measurement is used to evaluate the performance of the
sure technological change and economies of scale. In the power industry under different ownership structures and
early period, the technological change of the power regulatory regimes.
industry was measured by studying shifts in production The use of the DEA approach not only allows us to
and cost functions (e.g. Komiya (1962); Nerlove (1963); compare individual firms to best practice firms, but also
Barzel (1963, 1964); Dhrymes and Kurz (1964); to identify sources of inefficiency. Such benefits allow
McFadden (1964); Galatin (1968); Courville (1974); regulators to formulate policies on deregulation and priv-
Cowing (1974); Boyes (1976); Christensen and Greene atization, and to determine the appropriate productivity
(1976).1 In the 1970s, empirical studies of productivity factor when imposing price-cap regulation or yardstick
tended to focus on the effect of ownership structure on competition on electric utilities. The DEA-like linear
efficiency. De Alessi (1974) explored the economic programming methods have been applied to measure the
consequences of government ownership and regulation productivity of the power sector in the U.S. and in other
in the electric power industry of the U.S. Meyer (1975) countries. Unlike parametric analyses, which focus on
investigated the comparative efficiency of private and generation, both generation and transmission and distri-
public electric utilities in the U.S. He used a dummy bution have been covered in studies using the DEA
approach to pool both private and public firms and meas- approach.
ure the cost difference due to ownership structure, with- Fare et al. (1985) were the first to use the DEA
out taking factors such as input prices and technology approach to compare the efficiency of public and private
into account. electric utilities. They used the data from Atkinson and
Pescatrice and Trapani (1980) also used a dummy Halvorsen’s study, and found that public utilities were
variable to measure the productivity difference between more efficient than private utilities, and that inefficiency
private and public electric utilities in the U.S. Unlike was largely due to the lack of allocative efficiency. Cote
Meyer’s study, however, they used a translog function (1989) applied the stochastic frontier cost function to
that controlled the effects due to output, input prices, estimate the technical efficiency of electric utilities under
and technology. Their results suggested that the unit cost different ownership structures. His results suggested that
of public firms was 24–33% lower than that of private co-operatives were the most efficient type of ownership
firms. The source of inefficiency was due to the rate-of- structure, while small private and public electric utilities
return constraints that were imposed on privately-owned had similar levels of technical inefficiency.
electric utilities. The study that was conducted by Dilor- Fare et al. (1990) used the Malmquist productivity
enzo and Robinson (1982) indicated similar results, index to study the productivity growth of 19 coal-fired
while that of Atkinson and Halvosen (1986) indicated generating plants in Illinois from 1975 to 1981. They
that price inefficiency existed in both public and private found that the average rates of productivity growth were
electric utilities. relatively stable, except during a productivity slowdown
More recently, frontier methods such as data envelop- from 1976 to 1977. Pollitt (1995) found no significant
ment analysis (DEA) and stochastic frontier analysis differences in efficiency between different types of own-
(SFA) have been used to measure efficiency and pro- ership or economic organization in a study of generation
ductivity. The DEA approach is non-parametric, whereas utilities in OECD countries. Coelli (1997) applied both

1 2
Cowing and Smith (1978) survey econometric analyses of steam- Refer to Lovell (1996) for a review of the ability of the SFA and
electric generation based on production and cost functions. DEA approaches to measure efficiency and productivity change.
78 P.-L. Lam, A. Shiu / Utilities Policy 10 (2001) 75–83

the DEA and SFA approaches to measure the total factor The above model imposes constant returns to scale
productivity (TFP) change of 13 base-load, coal-fired

K

plants in Australia from 1981–1982 to 1990–1991. The (CRS). If the condition of zkt ⫽ 1 is added, then
empirical results suggested a TFP growth over the 10- k⫽1

year period of up to 16%. Olatubi and Dismukes (2000) variable returns to scale (VRS) are imposed (see Fig. 1).
used the DEA approach to measure cost efficiency In Fig. 1, DMUs A, C, and D are considered relatively
opportunities for coal-fired generation facilities in the efficient under VRS, whereas only DMU C is considered
U.S., and found that allocative inefficiency was the most relatively efficient under CRS. They all have technical
important source of inefficiency in 1996. efficiency scores of 1, which implies that they are
The DEA approach has also been applied to measure operating on the best practice frontier. DMU F is con-
efficiency in electricity transmission and distribution sidered as relatively inefficient under both kinds of
(e.g. Weyman-Jones (1991); Miliotis (1992); Hjalmars- returns to scale. Under CRS, the output-oriented techni-
son and Veiderpass (1992a, 1992b); Bagdadioglu et al. cal efficiency score for DMU F is given by TEo,crs =
(1996); Kumbhakar and Hjalmarsson (1998)). ZF/ZK, whereas under VRS, the output-oriented techni-
cal efficiency score for DMU F is given by TEo,vrs =
ZF/ZS. In our study, we assume VRS in power gener-
4. Methodology ation.
Consider the VRS frontier in Fig. 1. One section runs
In this study, our focus is on technical efficiency. The parallel to the horizontal axis (i.e. DE), and this feature
non-parametric DEA approach is used to compute the could lead to the presence of input slack, which is the
technical efficiency of China’s thermal power gener- extent to which inputs could be further reduced even
ation. The use of the DEA does not require any specifi- when a DMU is projected onto the production frontier.
cation of the functional form of the production relation- Consider inefficient DMU G. Even when G is projected
ship. Given the inputs that are used and the output that onto the frontier, one can further reduce the amount of
is produced, the prior weighting of the relative impor- input (i.e. DH) and produce the same level of output.
tance of outputs and inputs is not required. DEA ana- This amount of DH is known as input slack.3
lyzes each decision-making unit (DMU) separately and
identifies those which exhibit best practice. A frontier of
these units is then constructed, and the efficiency meas- 5. Data and results
ure of each DMU is determined relative to this best prac-
tice frontier. Those DMUs that lie on the frontier are Detailed information about the power generation of
regarded as relatively efficient and have scores of 1, thermal power plants in China, like total output, generat-
whereas those that lie inside the frontier are regarded as ing capacity, and fuel consumption can be obtained from
relatively inefficient and have scores of less than 1 but
greater than 0.
The DEA model can be applied in an input or output
orientation. In this study, we use an output-orientation
for measuring the provincial efficiency for electricity
generation. Suppose we have k = 1,…, K number of
firms which produce M outputs, i.e. ym, kt, m = 1,…, M,
using N inputs, xn, kt, n = 1,…, N, at each time period t,
where t = 1,…, T. zkt represents the intensity levels,
which make the activity of each observation expand or
contract to construct a piecewise linear technology.
Ft0(xkt,ykt) ⫽ maxd represents the output-oriented Farrell
efficiency measure that indicates the maximum possible
expansion of yt for firm k at period t.
Ft0(xkt,ykt) ⫽ maxd,


K

st dym,ktⱕ zktym,kt m ⫽ 1,…,M, Fig. 1. Graphical illustration of the DEA approach and the presence
k⫽1
of input slacks.
(1)

K

zktxn,ktⱕxn,kt n ⫽ 1,…,N, 3
We used the DEAP software to compute the DEA models and the
k⫽1
labor slack. The slack was calculated using the multi-stage method that
zktⱖ0,k ⫽ 1,…,K is discussed by Coelli et al. (1998).
P.-L. Lam, A. Shiu / Utilities Policy 10 (2001) 75–83 79

Table 1
Summary statistics for 30 DMUs in China’s generation

Electricity Electricity generated Generating capacity Total fuel used Labor


generation (GWh)b (MW)b (TJ)b (Number of persons)a

1995

Mean 26911.47 5431.35 456012.21 25676.67


S. D. 21231.48 4291.36 335372.61 18897.22
Minimum 165.00 47.10 79.55 315.00
Maximum 73676.00 16260.40 1153665.52 71280.00
1996

Mean 29270.03 5962.12 487508.78 23691.30


S. D. 22798.49 4702.97 353979.89 19471.14
Minimum 133.00 55.70 293.08 198.00
Maximum 79223.00 19723.30 1239205.20 75330.00

a
Labor refers to the number of workers engaged in thermal power generation in the power sector.
b
GWh refers to gigawatt hours, MW refers to megawatt and TJ refers to terajoule.

the Annual Report of Electricity Industry that is pub- of technical efficiency. Provinces with rich supplies of
lished by the China Electricity Press. However, the coal, such as Shanxi, Shandong, Hebei, and Henan, also
report only provides the total labor input for the power achieved higher levels of technical efficiency. The tech-
industry, and does not provide separate figures of the nical efficiency for Yunnan was the lowest among all of
labor input for thermal power generation. Fortunately, the DMUs.
we were able to locate the figures for 1995 and 1996 in The input slack results are presented in Table 3. The
China’s Industrial Markets Yearbook (Li and Tse, 1997 presence of slack reveals the scope for further non-radial
and 1999), and our study covers those two years. reduction in inputs once an inefficient DMU is projected
Our cross-sectional study covers 30 provinces, auton- onto the production frontier. The excessive use of inputs
omous regions, and municipalities, each of which is con- accounts for about 7.85–8.54% of total fuel and 11.10–
sidered as a Decision Making Unit (DMU). The data that 26.13% of total labor used in those DMUs with slacks.
is available allows us to measure the technical efficiency Capital slack is insignificant, which implies that excess
scores of the different DMUs in each of the two years. capacity is not a serious problem in China’s power sec-
Electricity that is generated from thermal power plants tor. Instead, the problem of power shortage continues to
in each DMU is used as the output variable, while capi- exist in many provinces.
tal, fuel, and labor are the three inputs. Capital is meas- The amount of labor slack increased substantially
ured in terms of installed thermal generating capacity from 1995 to 1996. As compared to the efficient pro-
in megawatt (MW). Fuel consumption is measured in duction frontier, some DMUs faced serious problems of
terajoules (TJs). The different fuels (such as coal, oil, labor redundancy. In 1996, the labor input for each DMU
and gas) that are used for power generation are converted with slack could have been reduced by an average of
into TJ equivalents, and then aggregated to obtain the 6,312 without affecting output. The sharp increase in
amount of fuel input. Labor input is the number of work- labor slack might have been due to the restructuring of
ers in thermal power generation. the power sector in 1996 and the subsequent formation
Table 1 shows the summary statistics for the output of the SPC in 1997. In the face of restructuring, state-
and inputs that were used to construct the DEA models. owned power enterprises in some provinces were quick
The construction of the models allows the investigation to shed labor to improve their efficiency. As a result,
of relative efficiency scores and the presence of slack in labor productivity became more diverse among DMUs.
each year for power generation. The results for each However, it should be emphasized that the problem of
cross-sectional DEA model for 1995 and 1996 are shown labor redundancy would become much more serious if
in Table 2. The average efficiency score increased from we were to compare China’s power sector to the best
0.888 in 1995 to 0.903 in 1996. The technical efficiency performance that was achieved by power companies in
scores show great variations among DMUs. Eight prov- other developed countries.
inces (Beijing, Hebei, Shanghai, Jiangsu, Shandong, Once we computed the cross-sectional efficiencies, we
Guangdong, Xizang, and Ningxia) operated along the performed the second stage (Tobit) regression analysis
production frontier in both years. Municipalities and to identify factors that might influence technical
provinces along the eastern coast achieved higher levels efficiency. In our regression model, the estimated cross-
80 P.-L. Lam, A. Shiu / Utilities Policy 10 (2001) 75–83

sectional efficiency score (EFF-SCORE) for 1996 is Table 3


regressed on a number of potentially influencing exogen- Summary statistics of slack variables
ous factors. The specific vector of variables that are used
Input No. of DMUs Mean (amount % of slack in
include capacity factor (CAPACITY), fuel use per kWh with slacks of slack per total input
of electricity (FUEL), the average size of generating DMU)
units (SIZE), the average age of generating units (AGE),
and two dummy variables. The two dummy variables are Capacity
used to indicate whether the region is under the control 1995 1 21.9 2.17
of the SPC (0 = not under SPC control and 1 = under 1996 0 0 0
Fuel
SPC control) and the presence of foreign investment in 1995 8 41017 7.85
the power sector (FOREIGN) (0 = without foreign 1996 3 39000 8.54
investment and 1 = with foreign investment). Labor
Because of the lack of information about the load fac- 1995 9 2923 11.10
tors (i.e. average load divided by peak load) of all prov- 1996 17 6312 26.13
inces, we used capacity factor (i.e. average load divided
by installed capacity) to measure the impact of load
characteristics on generating efficiency. If the installed
capacity is planned to meet the peak load, then the
capacity factor will be highly correlated with the load

Table 2
Cross-sectional variable returns to scale (VRS) DEA efficiency scores

DMU Efficiency scores for 1995 Efficiency scores for 1996

Northern Region
Beijing 1.000 1.000
Tianjin 0.890 0.845
Hebei 1.000 1.000
Shanxi 0.971 0.973
Inner Mongolia 0.825 0.883
North Eastern Region
Liaoning 0.910 0.895
Jilin 0.832 0.950
Heilongjia 0.852 0.838
Eastern Region
Shanghai 1.000 1.000
Jiangsu 1.000 1.000
Zhejiang 0.856 0.891
Anhui 1.000 0.969
Fujian 0.940 0.956
Jiangxi 0.743 0.758
Shandong 1.000 1.000
Southern Region
Henan 0.970 0.968
Hubei 0.836 0.749
Hunan 0.822 0.801
Guangdong 1.000 1.000
Guangxi 0.685 0.713
Hainan 0.719 1.000
South Western Region
Sichuan 0.848 0.841
Guizhou 0.834 0.896
Yunnan 0.661 0.647
North Western Region
Xizang (Tibet) 1.000 1.000
Shaanxi 0.936 0.940
Gansu 0.991 0.989
Qinghai 0.694 0.767
Ningxia 1.000 1.000
Xinjiang 0.813 0.825
Average 0.888 0.903
P.-L. Lam, A. Shiu / Utilities Policy 10 (2001) 75–83 81

factor of a region. It is expected that a higher capacity the effect of small units on efficiency scores, we add a
factor (CAPACITY) will better utilize the existing gen- variable SMALL, which is defined as the proportion of
erating capacity, and this will increase the efficiency generating units smaller than 100 MW, in our second
score. stage (Tobit) regression analysis.
The quality of fuel that is used for power generation The results of our second stage regression are shown
in different power plants will definitely affect the in Table 4. As expected, the coefficients for CAPACITY
efficiency of thermal power generation. Although infor- and FUEL have significant effects on the technical
mation about the heat content of fuel (e.g. in terms of efficiency of DMUs. A higher capacity factor means a
joules per kilogram of coal) is not available, the power higher utilization rate of generating capacity, which
enterprises in different provinces have published data on could save capital investment and help to achieve a
fuel use (in terms of standard coal in grams) per kWh higher level of technical efficiency. The use of high qual-
of electricity. If the fuel that is used for power generation ity fuel (which implies a smaller amount of fuel use) has
is of lower quality (in terms of heat content), then the also raised the technical efficiency. Our results show that
fuel use per kWh of electricity will be higher. It is DMUs that were independent of the SPC achieved
expected that the relationship between the efficiency higher levels of technical efficiency, while the presence
score and the fuel use (FUEL) is negative. Higher fuel of foreign investment (FOREIGN) did not have any sig-
use implies a lower quality of fuel, which will lower the nificant effect on technical efficiency. Based on the sam-
efficiency score. ple data of generating units with a capacity of 100 MW
As mentioned earlier, the thermal efficiency of power or above, the size and age of generating units are not
plants in China is low because of the proliferation of significant factors in the technical efficiency of thermal
small generating units constructed in the 1980s and the power generation. A surprising result is that the variable
early 1990s. Power enterprises in China only provide SMALL is positive and significant, which implies that
information about the operating characteristics of gener- a wider use of small generating units in a province has
ating units with a capacity of 100 MW or above, infor- increased efficiency. These small units are less efficient
mation about smaller units is not available. Therefore, than large units in terms of size and fuel use, but since
our data on the average size (SIZE) and age (AGE) of they are not subject to tight control by provincial and
generating units are based on information from those central governments, their flexibility in operations and
units with a capacity of 100 MW or above, which labor recruitment might have increased their overall
accounted for almost 60% of the total capacity. Data for efficiency.
three DMUs (Hainan, Qinghai and Xizang) are not avail-
able, hence we only have 27 observations when the vari- 6. Conclusions and implications
ables SIZE and AGE are included. In the light of this
data limitation, the estimated coefficients for these two This paper examines the technical efficiency of Chi-
variables should be interpreted with caution. To measure na’s thermal power generation industry. We apply the

Table 4
Regression analysis of technical efficiency

Variable Coefficient Coefficient Coefficient Coefficient

CONSTANT 1.5626∗∗∗ (0.3009)h 1.4843∗∗∗ (0.3085)h 1.0307∗∗∗ (0.3927)h 1.6591∗∗∗ (0.3066)h


CAPACITYa 0.3646∗ (0.2069)h 0.5554∗∗∗ (0.1946)h 0.5457∗∗∗ (0.1822)h 0.6090∗∗ (0.2656)h
FUELb ⫺0.0017∗∗∗ (0.0006)h ⫺0.0022∗∗∗ (0.0006)h ⫺0.0013∗ (0.0008)h ⫺0.0027∗∗∗ 0.0008h
SPCc ⫺0.2015∗∗∗ (0.0722)h ⫺0.0621 (0.0657) ⫺0.0787 (0.0636) ⫺0.1632∗∗ (0.0732)h
FOREIGNd 0.0138 (0.0416) 0.0119 (0.0347) 0.0142 (0.0332) ⫺0.0147 (0.0444)
SIZEe 0.0001 (0.0004) 0.0004 (0.0003)
AGEf 0.0107 (0.0065)
SMALLg 0.2624∗ (0.1421)h
R2 0.5186 0.5561 0.5942 0.5193
No. of observations 30 27 27 30

a
CAPACITY = capacity factor (average load divided by installed capacity).
b
FUEL = fuel use (in terms of standard coal) per unit of electricity generated.
c
SPC = 1 for provinces under SPC control, it is equal to 0 otherwise.
d
FOREIGN = 1 for provinces with foreign investment, it is equal to 0 otherwise.
e
SIZE = average size of thermal generating units with capacity 100 MW or above.
f
AGE = average age of thermal generating units with capacity 100 MW or above.
g
SMALL = proportion of generating units smaller than 100 MW.
h
∗∗∗, ∗∗, ∗ Significant at 1%, 5% and 10%, respectively; standard errors are in parentheses.
82 P.-L. Lam, A. Shiu / Utilities Policy 10 (2001) 75–83

DEA approach to measure the technical efficiency of dif- would also like to thank Harry Wu at the Hong Kong
ferent regions in China. Our study involves 60 obser- Polytechnic University, Rolf Fare at Oregon State Uni-
vations from 30 provinces, autonomous regions, and versity, Ziping Wu at Queen’s University, and parti-
municipalities in 1995 and 1996. Apart from comparing cipants at the international workshop on the Chinese
the efficiency among different regions, our study exam- economy for their valuable comments on earlier drafts.
ines the presence of input slack and the possible factors Thanks are also due to Pat Lam and Alice Ng for their
that affect efficiency. assistance throughout the study. The work that is
Our results show that there was a slight increase in the described in this paper was supported by research grants
average efficiency scores in the two years under study. In from the Hong Kong Polytechnic University (Project
general, municipalities and provinces along the eastern A/C Code: G-T196) and the Research Grants Council of
coast and those with rich supplies of coal achieved Hong Kong (Project A/C PolyU 5213/01H).
higher levels of technical efficiency. Consistent with pre-
vious studies, our results indicate that environmental fac-
References
tors affect efficiency scores, and any benchmarking
method should take these into consideration. There is no Atkinson, S.E., Halvorsen, R., 1986. The relative efficiency of public
clear evidence of excess capacity. However, the presence and private firms in a regulated environment: The case of U.S.
of labor slack in many regions indicates that labor redun- electric utilities. Journal of Public Economics 29 (3), 281–294.
dancy is a serious problem. Bagdadioglu, N., Price, C.M.W., Weyman-Jones, T.G., 1996.
Efficiency and ownership in electricity distribution: A non-para-
In our second stage (Tobit) regression analysis, we metric model of the Turkish experience. Energy Economics 18 (1-
find that capacity factor and fuel efficiency are signifi- 2), 1–23.
cant factors affecting technical efficiency. Government Barzel, Y., 1963. Productivity in the electric power industry, 1929-
policies such as peak load pricing and demand-side man- 1955. The Review of Economics and Statistics 45 ((Nov.)), 395–
agement would help to improve the load curve and the 408.
Barzel, Y., 1964. The production function and technical change in the
utilization rate of generating capacity. These policies steam power industry. Journal of Political Economy 72 (2), 133–
would help increase the technical efficiency of thermal 150.
power generation. The policy of mandated closures of Boyes, W.J., 1976. An empirical examination of the Averch-Johnson
inefficient diesel generating units would help to improve effect. Economic Inquiry 14 (1), 25–35.
the fuel efficiency of power generation. The presence of China Electric Power Information Center 2000. Electric Power Indus-
try in China. China Electric Power Information Center, Beijing.
foreign investment did not have any significant effect on China Electricity Press, 1995 and 1996–1997. Annual Report of Elec-
the efficiency scores. Hence, the inflow of foreign capital tricity Industry. China Electricity Press, Beijing.
into China’s power market seems to help overcome capi- Chow, D.C.K., 1997. An analysis of the political economy of China’s
tal constraints rather than to transfer new technologies enterprise conglomerates: A study of the reform of the electric
and improve technical efficiencies. power industry in China. Law and Policy in International Business
28 (2), 383–433.
The Chinese government has recently announced the Christensen, L.R., Greene, W.H., 1976. Economies of scale in U.S.
restructuring of the SPC, which will be divided into sep- electric power generation. Journal of Political Economy 84 (4),
arate regional transmission companies and required to 655–676.
divest its generating facilities. The generating assets of Coelli, T., 1997. Total factor productivity growth in Australian coal-
the SPC will be divested to four or five IPPs, with the fired electricity generation: A Malmquist index approach. Paper
presented at the international conference on public sector
aim of reducing the dominant power of the SPC. Our efficiency, UNSW, Sydney, 27–18 November, 1997.
results indicate that provincial power enterprises that are Coelli, T., Prasada Rao, D.S., Battese, G.E., 1998. An Introduction to
independent of the SPC have achieved higher levels of Efficiency and Productivity Analysis. Kluwer Academic Pub-
efficiency. Provinces with relatively more small generat- lishers, Boston.
ing units have also achieved higher levels of efficiency. Contemporary China Press 1994. Electric Power Industry in Contem-
porary China. Contemporary China Press, Beijing.
Most of these small units were constructed in the 1980s Cote, D.O., 1989. Firm efficiency and ownership structure - The case
and the early 1990s, without the approval of the central of U.S. electric utilities using panel data. Annals of Public and
government. Our results provide some support for Cooperative Economics 60 (4), 431–450.
divestiture of the generating assets of the SPC and Courville, L., 1974. Regulation and efficiency in the electric utility
relaxing the control of the central government on the industry. The Bell Journal of Economics and Management Science
5 (1), 53–74.
power sector. The policies of decentralization and Cowing, T.G., 1974. Technical change and scale economies in an
divestiture would help improve the performance of engineering production function: The case of steam electric power.
power sector in different regions of China. Journal of Industrial Economics 33 (1), 135–152.
Cowing, T.G., Smith, V.K., 1978. The estimation of a production tech-
nology: A survey of econometric analyses of steam electric gener-
Acknowledgements ation. Land Economics 54 (2), 156–185.
De Alessi, L., 1974. An economic analysis of government ownership
The authors would like to thank Guangdong Electric and regulation: Theory and the evidence from the electric power
Power for arranging a study visit in early 1999. We industry. Public Choice 19 (Fall), 1–42.
P.-L. Lam, A. Shiu / Utilities Policy 10 (2001) 75–83 83

Dhrymes, P.J., Kurz, M., 1964. Technology and scale in electricity McFadden, D., 1964. Notes on the estimation of the elasticity of substi-
generation. Econometrica 32 (3), 287–315. tution, Working paper no. 57. Institute of Business and Economic
Dilorenzo, T.J., Robinson, R., 1982. Managerial objectives subject to Research, University of California, Berkeley.
political market constraints: Electric utilities in the U.S. Quarterly Meyer, R.A., 1975. Publicly owned versus privately owned utilities:
Review of Business and Economics 22 (2), 113–125. A policy choice. The Review of Economics and Statistics 62 (4),
Economic Management Press 1997. China’s Energy Development 391–399.
Report. Economic Management Press, Beijing. Miliotis, P.A., 1992. Data envelopment analysis applied to electricity
Fare, R., Grosskopf, S., Logan, J., 1985. The relative performance of distribution districts. Journal of Operational Research Society 43
publicly-owned and privately-owned electric utilities. Journal of (5), 549–555.
Public Economics 26 (1), 89–106. Ministry of Energy 1989. China’s Construction in Four Decades (1949-
Fare, R., Grosskopf, S., Yaisawarng, S., Li, S.K., Wang, Z., 1990. 1989): Electric Power Industry in China. Coastal International
Productivity growth in Illinois electric utilities. Resources and Investment Consultant Co. Ltd., Hong Kong.
Energy 12, 383–398. Nerlove, M., 1963. Returns to scale in electricity supply. In: Zellner,
Financial Times Energy, 1999. Power in Asia: The Asian electricity A. (Ed.), Readings in Economic Statistics and Econometrics. Little
market. Financial Times Energy, London, March 8, 1999, issue Brown Co., Boston, pp. 409–439.
No. 272. Olatubi, W.O., Dismukes, D.E., 2000. A data envelopment analysis of
Galatin, M., 1968. Economies of Scale and Technological Change in the levels and determinants of coal-fired electric power generation
Thermal Power Generation. North Holland Publishing Company, performance. Utilities Policy 9 (2), 47–59.
Amsterdam. Pescatrice, D.R., Trapani III, J.M., 1980. The performance and objec-
Hjalmarsson, L., Veiderpass, A., 1992a. Productivity in Swedish elec- tives of public and private utilities operating in the United States.
tricity retail distribution. Scandinavian Journal of Economics 94 Journal of Public Economics 13 (2), 259–276.
(suppl.), 193–205. Pollitt, M.G., 1995. Ownership and Performance in Electric Utilities.
Hjalmarsson, L., Veiderpass, A., 1992b. Efficiency and ownership in Oxford University Press/Oxford Institute for Energy Studies,
Swedish electricity retail distribution. The Journal of Productivity Oxford.
Analysis 3 (1-2), 7–23. Weyman-Jones, T.G., 1991. Productivity efficiency in a regulated
Komiya, R., 1962. Technological progress and the production function industry - The area electricity boards of England and Wales. Energy
in the United States steam power industry. Review of Economics Economics 13 (2), 116–122.
and Statistics 44 (2), 156–166. Wirtshafter, R.M., 1990. Decentralization of China’s electricity sector:
Kumbhakar, S.C., Hjalmarsson, L., 1998. Relative performance of pub- Is small beautiful? World Development 18 (4), 505–512.
lic and private ownership under yardstick competition: Electricity Yang, M., Yu, X., 1996. China’s power management. Energy Policy
retail distribution. European Economic Review 42 (1), 97–122. 24 (8), 735–757.
Li, S., Tse, D.K., 1997/1999. China’s Industrial Markets Yearbook. Zhu, T., 1999. China’s corporatization drive: An evaluation and policy
City University of Hong Kong Press, Hong Kong. implications. Contemporary Economic Policy 17 (4), 530–539.
Lovell, C.A.K., 1996. Applying efficiency measurement techniques to
the measurement of productivity change. Journal of Productivity
Analysis 7 (2-3), 329–340.

You might also like