Vous êtes sur la page 1sur 9

Energy Economics 32 (2010) 11821190

Contents lists available at ScienceDirect

Energy Economics
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e n e c o

Technical and allocative inefciencies and factor elasticities of substitution:


An analysis of energy waste in Iran's manufacturing
Nasser Khiabani , Karim Hasani
Department of Economics, Institute for Management and Planning Studies, Mokhtar Asgari Str.10, 19395, Tehran, Iran

a r t i c l e

i n f o

Article history:
Received 21 November 2009
Received in revised form 1 May 2010
Accepted 10 May 2010
Available online 21 May 2010
JEL classication:
C33
D21
Q48

a b s t r a c t
Ignoring technical and allocative inefciencies or embedding one of them alone in a system of input
demands may result in biased elasticities. We consider a comprehensive model including technical
inefciency (in input and output forms) and allocative inefciency and apply it to panel data from Iran's
manufacturing sector. The results show that the presence of both inefciencies affects the computed
elasticities of demand and substitution. Moreover, in spite of current waste of energy in Iran's manufacturing,
the elimination of environmental constraints will prompt the manufacturing rms to increase the utilization
of energy relative to both capital and labor.
2010 Elsevier B.V. All rights reserved.

Keywords:
Technical inefciency
Allocative inefciency
Shadow cost function
Inappropriate environment
Panel data

While most literature pertaining to rm inefciency has concentrated on either technical or allocative inefciency, there are some
studies considering both of them together. A rm is output technically
inefcient if it fails to produce maximal output from a given quantity
of inputs and is input technically inefcient if it fails to use the
technically minimum level of inputs given the output and input mix
(Atkinson and Cornwell, 1994b). Furthermore, the lack of appropriate
incentives and operating environment may induce a rm to be
allocatively inefcient (Christopoulos and Tsionas, 2002; Burki and
Khan, 2004). In this case, it deviates from strict cost minimizing and
applies a combination of inputs in which the marginal rate of
substitution between any two of them is not equal to their prices ratio.
The notion of technical and allocative inefciencies was introduced
by Farrell (1957). Lau and Yotopoulos (1971) devised a parametric
approach to compare the efciency of small and large Indian farms using
the CobbDouglas prot function. Toda (1976) estimated a generalized
Leontief cost function for Soviet manufacturing when the cost is not
minimum. Atkinson and Halvorsen (1984) in their study of regulated
electric utilities, as well as Eakin and Kniesner (1988) in their study of

hospitals estimated translog non-minimum cost functions. Fare and


Lovell (1978) distinguished input and output technical inefciencies.
Atkinson and Cornwell (1994a) estimated an increase in costs due to
technical and allocative inefciencies of the U.S. airline industry. Using a
shadow cost function and measuring the allocative inefciency
component in Greek manufacturing, Christopoulos and Tsionas (2002)
tried to solve the capital-energy controversy.1 Burki and Khan (2004)
estimated the effect of allocative inefciency on resource allocation for
Pakistan's manufacturing sector.
Although recent studies performed by focusing on allocative
inefciency have shed some light on correcting the bias of estimated
elasticities and solving the capital-energy controversy, we show in
this paper that ignoring both the technical and allocative inefciencies
may produce biased parameters and elasticities. Hence, it seems that
empirical economists go wrong in interpreting computed elasticities if
they do not measure these inefciencies simultaneously.
In this paper, we considered a more generalized shadow cost
function including technical inefciency, in both input and output
forms, along with allocative inefciency to deal with the aforementioned drawback. It is important to note that the joint estimation of
technical and allocative inefciencies in a translog cost function

Corresponding author. Tel.: +98 21 22802708; fax: +98 21 22802707.


E-mail addresses: n.khiabani@imps.ac.ir (N. Khiabani), k.hasani@imps.ac.ir
(K. Hasani).

1
While studies like Hudson and Jorgenson (1974) and Berndt and Wood (1975)
estimated a complementary relationship between capital and energy, Grifn and
Gregory (1976) and Pindyck (1979) found that capital and energy are substitutes.

1. Introduction

0140-9883/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.eneco.2010.05.004

N. Khiabani, K. Hasani / Energy Economics 32 (2010) 11821190

presents a difcult problem (Greene, 1980). The difculty is that the


cost function and the deviations of optimal shares from observed
shares are complicated functions of allocative inefciency.
This problem in the literature is known as Greene's Problem (Bauer,
1990). Atkinson and Cornwell (1994a) proposed a parametric
approach to solving this problem by assuming allocative and technical
inefciencies as xed parameters in a cost system model. Kumbhakar
and Tsionas (2005) used a Bayesian approach to solving Greene's
problem by estimating technical and allocative inefciencies as
random variables in a cost system model. In this paper, we follow the
Atkinson and Cornwell (1994a) approach and leave the Kumbhakar
and Tsionas (2005) approach for future work.
We estimated the model of this paper using panel data from Iran's
large-scale manufacturing sector for the years 1984 to 2005. We think
that, as it will be discussed in the next section, Iranian manufacturing
provides a particularly interesting case study because there are many
manifestations for the presence of allocative inefciency in Iranian
manufacturing preventing rms from strict cost minimizing. On the
other hand, there are the outward signs of technical inefciency of
rms, which disrupt the neoclassical hypothesis of technical efciency. The results show that ignoring each of technical and allocative
inefciencies may result in biased estimation of elasticities of demand
and substitution. In addition, the recent efforts of Iran's government to
privatize manufacturing rms and remove their environmental
constraints without a remarkable change in energy price will lead to
more energy waste.
The remainder of this study is organized as follows. Section 2
describes the inappropriate operating environment of manufacturing
in Iran. Section 3 presents the theoretical framework we use to model
technical and allocative inefciencies. Section 4 provides a description
of the panel data from Iran's large-scale manufacturing sectors.
Section 5 pertains to statistical estimation and analyzes the empirical
results. Finally, Section 6 concludes the paper and shows some
warning ndings important for policy makers.

2. Overall structure of manufacturing in Iran


Iran's industrialization started in the early 1960s, with a rapid
increase in manufacturing output. Many new industries were
established between 1962 and 1972. Higher oil revenues in the
1970s accelerated economic development. A number of large-scale
industrial projects with government investments were undertaken
during the period of the Fifth Development Plan (1973 to 1978). These
investments were mostly concentrated in petrochemicals and basic
metal industries as well as crude oil production. Industrial production
grew at close to 20% per year, and a diversied industrial base was
established. Until 1975, manufacturing and mining (excluding electric
power and construction) contributed to the GDP by about 10%.

1183

Business failures and a generally declining economy led to strikes


and political instability in 1978 and 1979. The ight of capital and
factory owners after the 1979 Revolution led to the nationalization of
industries in the summer of 1979. The decline of the industrial sector
was hastened by the war with Iraq. Since 1980, Iranian manufacturing
has been characterized by prevalence of highly subsidized and
overstaffed state-owned enterprises having completely changed the
manufacturing structure in comparison with that prior to 1978. After
1980, the government's control over the economy increased. Statecontrolled prices dissuaded foreign investment and the government
began regulating all economic activities through an unwieldy permit
system. Import Substituting Industrialization (ISI) was the hallmark of
policy, and large-scale nationalization cemented the ascendancy of the
state in manufacturing. With the implementation of ISI, the performance growth of Iran's manufacturing was highly dependent on the
domestic demand, import substitution policies and availability of
foreign exchange to procure imported inputs. The ofcial exchange
rate was not adjusted despite soaring prices, the price of the dollar on
the black market increased to more than ten times its ofcial rate. This,
in turn generated massive implicit subsidies for state-owned enterprises in manufacturing sectors which were also among importers of
essential goods and capital goods for their productive activities.
In general, changes in the allocation policy and the existence of
massive explicit and implicit subsidies in Iran's manufacturing caused
state-owned enterprises to play a dominant role and inhibited the
growth of the private sectors in manufacturing. For instance, on
average about 60% of sales, 63% of value added and 32% of
employment in the sector belonged to just the largest 200 stateowned rms in the period of 1996 to 2005. The state owns and
administers 3 large industries, including motor vehicles, basic metal
and steel products, and petroleum renery productions that contributed about 40% of total manufacturing sales in 2002 (KhodadadKhashi, 2007).
Additionally, energy consumption has been heavily subsidized by
state since 1982. The government, as the sole supplier, froze energy
prices from 1978 to 1996 primarily in support of the poor and
manufacturing and transportation sectors of economy. This policy has
induced energy to be consumed inefciently in whole sectors of
economy, especially in manufacturing. As can be seen in Fig. 1, energy
intensity, the ratio of energy consumption to output, has had an
increasing trend since 1980. Energy waste has made Iran's government worried such that National Iranian Oil Company (NIOC), under
the direction of the Ministry of Petroleum of Iran, in 2000 established
the Iranian fuel conservation company (IFCO) with the mission to
regiment the energy consumption in different sections of Iran's
economy including the transportation system, industry, and the
building sector.
Just after the end of the 8-year war with Iraq, in 1989, the
government's strategies in favor of a liberalized economy were

Fig. 1. Energy Intensity of Iran.


Source: US Energy Information Adminstration (EIA).

1184

N. Khiabani, K. Hasani / Energy Economics 32 (2010) 11821190

formed in a series of new ve-year development plans (FYDP).


Economic reforms through the rst FYDP began with the privatization
of state-owned enterprises, the reduction of the system of subsides
and protections, and the unication of multiple exchange rates.
However, the rst and second FYDPs were not completely successful
in achieving their goals. The unication of multiple exchange rates in
1993 was broken down in 1995 with the emerging of balance
payment crises and a fall in oil revenues. The government also started
a gradual price reform on energy reners in 1996 in order to reduce
the energy consumption in the economy and improve the energy
efciency of manufacturing industries. Nevertheless, because of the
gradual and insufcient increase in the price of energy, this policy has
not been effective in reducing the energy intensity of economy (see
Fig. 1). After 1999 with a permanent increase in oil revenues, the
government succeeded in establishing a unied exchange rate system,
eliminating non-tariff barriers and reducing tariff rates through the
third and fourth plans (19992003 and 20032008). However, some
fundamental economic problems pertaining to manufacturing still
remain: rstly, the pervasive role of state and low participation of
private sector in the economy and particularly in manufacturing
sectors and secondly, the high energy intensity and consumption of
energy manufacturing sectors.
According to the information above, it appears that there are many
manifestations of allocative inefciency in Iranian manufacturing that
have prevented rms from strict cost minimizing. Shortages of foreign
exchange, particularly during the 1982 to 1999 period, caused the
allocation of foreign exchange based on ofcial exchange rates to
become completely administrative. Furthermore, the import of raw
materials was limited by internal barriers and international sanctions
and their allocation was based on a discriminative behavior between
state-owned and private rms.
In spite of a gradual increase in the price of energy, the price
policies were not successful in adjusting energy prices with respect to
the long-term (1978 to 2005) price index. The resulting distortion in
energy prices has had a two-fold effect in Iran: rstly, energy prices
are generally far below the competitive global market prices;
secondly, the distortion in the relative prices of energy has led to
more incentive to utilize it. Although it has been possible for
manufacturing rms to use more capital-intensive technologies,
most machinery and plants used in industry are still energy-intensive.
The nal explanation for the existence of allocative inefciency in
Iran's manufacturing is the separation of management and ownership
in the public sector. Hence, managers in the public sector, which
includes the majority of manufacturing rms, do not have enough
incentive for strict cost minimizing. Overstafng of most manufacturing rms can be attributed to the fact that managers aim to
maximize their own utility rather than the prot of public rms.
Similarly, the neoclassical hypothesis of technical efciency of
rms also seems to be violated in Iranian manufacturing. Because of
an insufcient expert labor force, inadequate investment in R&D
(Research and Development), and worn out machinery and plants, it
appears that manufacturing rms in Iran have deviated from their
potential capability to produce on the frontier of their production
function.
All in all, it seems that manufacturing rms in Iran have both
technical and allocative inefciencies. Furthermore, notwithstanding
the possibility of technological change toward conserving energy, it
appears that they do not have enough incentive for this change.

Fig. 2. A rm with input technical inefciency.

input and output. Output technical inefciency is the failure of a rm


to produce maximal output given inputs, and its failure to produce a
given quantity of output using minimal inputs is its input technical
inefciency (Atkinson and Cornwell, 1994b).
P
Consider a rm using a vector of inputs x = (x1,,xn) to produce
output y and assume the rm's production function is f, a neoclassical
production function. Figs. 2 and 3 respectively depict the input and
output technical inefciencies of this rm.
In Fig. 2, it is assumed that our rm intends to produce yA, a given
P
quantity of output, and to do this, it uses xA input vector. In this gure,
P
xE is the minimal level of inputs which can be technically used to
P
P
produce yA, so xE = bxA (0 b b1). This rm has input technical
inefciency, for it fails to produce a given quantity of output using
minimal inputs, and b, the input technical efciency parameter, would
be one if the rm was technically efcient.
P
In Fig. 3, it is assumed that our rm using xA, a given vector of
inputs, intends to produce maximal output, but it produces yA. In this
gure, yE is the maximal level of output which technically can be
P
produced using xA, so yA = ayE (0 b a 1). This rm has output
technical inefciency, for it fails to produce the maximal output
given inputs, and a, the output technical efciency parameter, would
be one if the rm was technically efcient.
In Fig. 2, the cost minimization problem facing the rm can be
written as:
w 
w 
1
n
bx1 + +
bxn
b
b
s:t: f bx1 ; ; bxn = yA

Min

where w = (w1,,wn) is a vector of input prices.

3. Theoretical framework
3.1. Input and output technical inefciency
In general, a rm's failure to produce on the frontier of its
production function can be dened as its technical inefciency. As
shown by Fare and Lovell (1978), it can be interpreted in two forms:

Fig. 3. A rm with output technical inefciency.

N. Khiabani, K. Hasani / Energy Economics 32 (2010) 11821190

At point A:

In this case, the general form of the cost function is:



C = C


P
w
1 P
; y = C w; y
b
b

0 b b 1:

The last equality follows from the fact that a cost function is
linearly homogeneous in wi.
The share equations for all inputs corresponding to Eq. (2) are
derived as

Si

w 
i
xi
wx
b
=
= i i = Si
P
1
C
C

w;y

i = 1; ; n:

In Fig. 3, the cost minimization problem of rm can be stated as


Min w1 x1 + + wn xn
y
s:t: f x1 ; ; xn = A :
a

= C w; y = a

0ba1

wi xi
P
C w; y = a

i = 1; ; n:

  
 
k1
w1
w1
=k
k2
w2
w2

0 b k1 ; k2 ; k N 1

where ki,i = 1,2 are the allocative efciency parameters. The rm is


allocatively efcient providing that k1 = k2. It can be concluded from
Eq. (7) that at point A (A) that parameter k is greater (less) than one, the
rm must increase (decrease) x1 relative to x2 in order to be efcient.
If we call w si shadow price (wis = kiwi), the cost minimization
problem facing the rm can be stated as:
Min C s = k1 w1 x1 + + kn wn xn
s:t: f x1 ; ; xn = y

where Cs is the shadow cost function.


The shadow cost share of the j th input for the i th rm is dened as

s

Sji =


kj wji xji

Cis

where Cis is the shadow cost function of the i th rm, wji is the price of
the j th input for the i th rm, and kj is the j th input-specic allocative
efciency parameter.
From Eq. (9) it is implied that

xji =

and the general form of share equations is


Si =

w1 f1
k w
b
= 1 1 =
w2 f2
k2 w2

Here is the general form of cost function:


C

1185

Therefore, in contrast to input technical inefciency in which the


efciency parameter (b) only appears in the cost function, in the
output technical inefciency, the efciency parameter (a) appears in
both cost and share equations. As shown by Fare and Lovell (1978)
and Atkinson and Cornwell (1994b), for a rm only under constant
return to scale a is equal to b, under increasing return to scale b is less
than a, and under decreasing return to scale b is greater than a.

Ci Sji
:
k j wji

The total actual cost function of the ith rm is


Ci = wji xj

According to the rationality hypothesis, a rm among all


combinations of inputs from which a given quantity of output can
be produced must select the optimal combination having the
minimum cost; otherwise, the rm is allocatively inefcient. In
Fig. 4, the rm produces y using two inputs x1 and x2, L(y) is the isoquant curve for producing y, and point E where the marginalrate
 of
w1
substitution between x1 and x2 is equal to their relative price
is
w2
the optimal combination of inputs. Nonetheless, because of environmental constraints, point E is out of the rm's feasible space. Hence,
the next point on curve L(y) which has the minimum cost to produce
y after point E, like point A, is chosen by the rm.

j = 1; ; n i = 1; ; m:

11

Substituting Eq. (10) in Eq. (11), the total actual cost function
becomes
s

3.2. Allocative inefciency

10

s 1

Ci = Sji Ci kj :

12

Taking the logarithm of both sides of Eq. (12) yields


!
 s
s 1
log Ci = log Ci + log Sji kj
:
j

13

The actual cost share of the jth input for the ith rm is
wji xji
:
Ci
Substituting Eqs. (10) and (12) into Eq. (14) yields

Sji =

Sji =

Sjis k1
j
j Sjis k1
j

14

15

We assumed translog functional form to represent the shadow


cost function of the ith rm by incorporating technological change:
1
1
2
2
s
log Ci = 0 + y logyi + yy log yi + t t + tt t
2
2




+ j log kj wji + jt log kj wji t
j

16





1
+ jl log kj wji logkl wli + jy logyi log kj wji
2 j l
j
+ y logyi t
Fig. 4. A rm with allocative inefciency.

where t denotes time.

1186

N. Khiabani, K. Hasani / Energy Economics 32 (2010) 11821190

By applying Shepard's Lemma to Eq. (16), the shadow cost share


equations are obtained:
s

Sji =

 
log Cis

 = j + j t + jl log kl wli + jy log yi :
l
log kj wj

17

(PIM). Moreover, due to the eight-year war between Iran and Iraq
(1980 to 1988), we calculated the rate of depreciation for the war and
post-war years separately.
The price series of labor was accounted as PL = WL where W is the
total payment to L number of workers per year. The price series of
energy (PE), was computed as Pi Si where Pi is the price of one

3.3. Incorporating technical and allocative inefciencies in one equation


system
Modeling technical inefciency in input form and indicating the
technical efciency parameter with bi, we can derive the actual cost
function of the ith rm from Eqs. (2), (16), and (17) as:


1
2
log Ci = 0 + y logyi + yy yi + jy logyi log kj wji
2
j




1
+ j log kj wji + jl log kj wji logkl wli
2 j l
j


1
2
+ j log kj wji + t t + tt t
2
(
j

)
1
+ y logyi tlog kj j + j t + jl logkl wli + jy log yi
j
l
 
1
:
+ log
bi

18

By substituting Eq. (17) into Eq. (15), the actual share equation of
the jth input for the ith rm is derived as:


j + j t + jl logkl wli + jy log yi k1
j
l


:
Sji =
j + j t + jl logkl wli + jy log yi k1
j
j

19

Symmetry on the cross price effect implies lj = jl,j,l j l and


because C is is linearly homogeneous in shadow prices, the following
restrictions must be imposed on the parameters:
j = 1; jy = 0; jt = 0 ; jl = jl = jl = 0:
j

In this study, the system of equations including Eqs. (18) and (19)
in which input technical and allocative inefciencies have
 been

1
in
modeled is called equation system (System I). If term log
bi
equation system (System I) is dropped from Eq. (18) and yi is replaced
y
with i in both Eqs. (18) and (19), equation system (System II), in
a
which i output technical and allocative inefciencies have been
modeled, is obtained. By and large, in both equation systems, the
existence of allocative efciency parameters (kj, j = 1,,n) captures
the effect of environmental constraints on the cost minimizing
behavior of the rm, whereas the existence of technical efciency
parameters (bi and ai, i = 1,,m) captures the ability of the rm to
produce on the frontier of its production function.
4. Data and construction of variables
According to the second revision of International Standard
Industrial Classication (ISIC Rev 2), the manufacturing sector of an
economy is divided into nine general sectors associated with codes 31
through 39 (refer to Table A in Appendix A).
The data used in this analysis consists of annual panel data for nine
sectors of Iran's manufacturing employing more than ten workers
which has been published by Statistical Center of Iran (S.C.I) for the
years 1984 to 2005.
The price series of capital was computed as PK = NP
CS where NP is
the net prot, and CS is the capital stock. Capital stock was calculated
as Kt = (1 )Kt 1 + It where It is the gross investment and is the
rate of depreciation calculated by the perpetual inventory method

equivalent ton of crude oil, and Si is the share of the ith type of energy,
including electricity, natural gas, gas oil, gasoline, liquid gas, kerosene
oil, and fuel oil, from one equivalent ton of crude oil consumed by
every sector of manufacturing. Output (Y) was measured in value
added terms. Total actual cost (C) was obtained by adding the value of
capital, cost of consumed energy, and total payment to the labor force.
Furthermore, the total actual cost and output were expressed in 1997
constant prices.
Here we also investigate the integration properties of the data to
deal with the possibility of spurious relationships that may emerge in
the estimation of a system of demand equations. Recently, some
studies have pointed out that the variables of a demand system may
not have the same order of integration (e.g., Lewbel and Ng, 2005;
Karagiannis and Mergos, 2002; Lim and Shumway, 1997). For
instance, if in our system the three cost shares, total actual cost and
output level are thought of as stationary variables and the three
nominal prices as non-stationary variables, there cannot exist a
demand system with such data structures. However, if the three
nominal prices are cointegrated, they will form a linear combination
which is stationary. This makes an equation balanced and we are able
to use classical econometrics methods to estimate demand system.
We began by employing the unit-root panel test suggested by Im
et al. (2003) to investigate the integration properties of the variables.
As shown in Table 1, the cost shares, total actual cost and output level
are stationary, whereas the three nominal prices are non-stationary,
therefore in order to reliably estimate system equations (System I)
and (System II), nominal prices are required to be cointegrated. It is
necessary to note that because of adding up condition, there are only
two independent factor demand equations. Hence, we expect to nd
two cointegration vectors when investigating the relationship
between the factor prices in our demand systems. In Table 2, we
present the results of the Fisher panel cointegration test (Maddala and
Wu, 1999) applied to the three nominal prices. The test cannot reject
the existence of two long-run relationships among the prices. These
results show that the estimation of our equation systems will not
make up spurious relationships, while the relative factor prices and
the other variables in the demand systems are stationary.

5. Estimation and empirical results


For the purpose of estimation, we normalize technical efciency
parameters (ai or bi, i = 31,,39) to unity for a sector considered as
the most technically efcient sector, called *. Then, relative output and
input technical efciencies can be dened as ai = aai and bi = bbi ,
*
*
respectively. In this study, we selected the sector associated
with code
2
36 as the *, for only in this case were estimated technical efciency
parameters of other sectors signicant and more than one. Similarly,
allocative efciency parameters (kj, j = K,L,E) can be dened relatively
k
as kj = k j , where k* is a coefcient correlated with one of the inputs
*
and normalized
to unity. It is worth knowing that the estimated
relative value of kj's and all other parameters are invariant both to the
choice of which kj to normalize and its assigned value (Atkinson and
Halvorsen, 1986). In this study, we normalized the coefcient of
capital (kk) to unity in the estimation of equation systems (System I)
and (System II). To avoid the problem of singularity, the share
2
Manufacture of Non-Metallic Mineral Products, except Products of Petroleum and
Coal.

N. Khiabani, K. Hasani / Energy Economics 32 (2010) 11821190


Table 1
ImPesaranShin (IPS) unit root test.
Variable

IPS (F), level

IPS (F),
rst difference

IPS (F, Tr), level

IPS (F, Tr),


rst difference

SK
SL
SE
log(C)
log(Y)
log(PL)
log(PK)
log(PE)

2.6
3.7
1.7
0.2
1.6
4.4
0.78
2.6

14
14
11
13
9.4

1.9
2.4
1.8
4.9
4.5
1.2
0.8
0.001

9.0
8.0
6.7

Indicates signicance at 5%. IPS statistics has a null hypothesis of non-stationary. F and
Tr denote individual effects and individual linear trends, respectively. Optimal lags
were selected based on Schwarz information criterion (SIC).

equation for capital was dropped and its parameters were recovered
with the help of adding up restrictions.
While the system of equations in Eqs. (18) and (19) is a nonlinear
seemingly unrelated regression model with xed technical and
allocative inefciency parameters, the estimated parameters can be
obtained by using the Iterative Seemingly Unrelated Regression
Method (ISURE). Since this is equivalent to the maximum likelihood
method, the results are invariant to which share equation is dropped.
We estimated the system of equations by the ISURE method and to
make sure that our likelihood function has achieved a global
maximum, the estimation was repeated three times by dropping a
different share equation each time. We found that the estimated
parameters and values of the likelihood function remain the same no
matter which equation is dropped.
The estimated parameters and the asymptotic standard errors are
presented in Table 3, and as seen, most of the parameters are
statistically signicant at conventional levels. The shadow cost
function is associated with a constant return to scale production
function in the case that y = 1 & jy = yy = yt = 0, j = K,L,E. This
hypothesis was rejected by Wald test (the computed 2 test statistics
for equation systems (System I) and (System II), 33.93 and 25.74 in
turn, are greater than critical value at 0.01 level of signicance).
Therefore, according to Fare and Lovell (1978) as mentioned in
Section 3.1, this result implies that input and output technical
efciencies are not equal. To investigate whether the cost function
is Hicks-neutral or technological change existed in the span of 1984 to
2005, we tested the hypothesis that: kt = lt = et = 0. The result of
Wald test strongly rejected this hypothesis (the computed 2 test
statistics for equation systems (System I) and (System II), 109.81 and
109.55 in turn, are greater than the critical value at the 0.01 level of
signicance). According to the positive and signicant estimation of
kt and et, as well as the negative and signicant estimation of lt, it
seems that technological change has been energy-using, capital-using,
and labor-saving. These results can be attributed to a new generation
of machinery and plants used in manufacturing in developing
countries like Iran; similar technological changes have also been
found in Pakistan (Burki and Khan, 2004).
In both equation systems, kK was normalized to one and kL and kE in
turn were estimated less and greater than one. That kL b 1 can be
interpreted to mean that manufacturing rms over-utilize labor relative
to capital. This result is consistent with extra employment in public

Table 2
Fisher panel cointegration test on prices.
H0: Rank = p

Trace statistics

P-value

p=0
p1
p2

109.400
33.220

0.000
0.015
0.950

8.600

Indicates signicance at 5%.


Probabilities are computed using asymptotic Chi-square distribution.

1187

rms, especially in ofcial, non-producing parts of rms. As mentioned


in Section 2, since the ownership and management are separated in the
public sector comprising the majority of manufacturing rms in Iran, the
managers try to maximize their utility instead of rm's prot. For
instance, they employ an extra labor force made up of their own friends,
acquaintances, and relatives who are not necessary or benecial for their
rms. Overstafng of most public rms has also been intensied by the
government's administrative support of employment to alleviate social
problems caused by unemployment.
That kE N 1 means manufacturing rms under-utilize energy
relative to capital, i.e. if they want to be allocatively efcient and
use the least cost combination of input factors, they should utilize
more energy. This result is very notable, for as discussed in Section 2,
energy intensity in Iran has increased over the 19842005 and the
administrative policies of the government have not been successful in
controlling the waste of energy. There is an answer to this dilemma:
according to the result of this study, as soon as constraints vanish and
rms are enabled to strictly minimize their costs, they will utilize
more energy relative to capital in comparison to the present time.
Thus, it can be concluded that however energy is being wasted even
now; manufacturing will also boost its utilization of energy if an
appropriate environment without any constraints is provided. The
low price of energy is the logical explanation for this phenomenon.
This result is in contrast with administrative policies of the
government to persuade manufacturing rms to substitute energy
with other factors of production without a remarkable increase in
energy price. This outcome will intensify more, for the government
has simultaneously tended toward privatization of rms. In this case,
manufacturing rms will have more incentive to maximize their
prot, and consequently they will consume energy more and more at
this low level of price. In fact, it can be taken as a warning to policy
makers who support such privatization policies without insisting on
changes in former energy price policies. To gure out how the energy
labor ratio has been distorted from that in the efcient allocation of
resources, equation systems (System I) and (System II) were
estimated again assuming kL = 1, dropping the share equation of
labor. The estimated parameters (kj,j = K,L,E) are presented in Table 4.
Normalized relative to labor, kk (in fact) was estimated to be greater
than one. This result is consistent with that of previous estimation in
which the normalization had been done relative to capital and kL (in
kL
fact ) had been estimated to be less than one. All in all, both results
kK
show over-utilization (under-utilization) of labor (capital) relative to
kE
capital (labor). Furthermore, kE (in fact ) in Table 4 was estimated to
kL
be greater than one and this result shows over-utilization of labor
relative to energy. So it can be deduced that the removal of rms'
constraints, mentioned in Section 2, will prompt them to augment
their consumption of energy relative to both capital and labor.
The estimation of relative input and output technical efciency
parameters for nine sectors of manufacturing associated with codes
31 through 39 is reported in Table 5. As mentioned before, we
normalized both input and output technical efciency parameters of
sector 36 to unity. Results show that at rst sector 393 and after that,
sectors 33 and 34 with a little difference are in turn the most
technically inefcient sectors. In light of the fact that the structure of
sector 39 is indenite, it is hard to interpret the results for this sector,
but that sectors 33 and 34 are the most technically inefcient ones
was predictable. This outcome can be attributed to the fact that they
almost have the most poor and archaic infrastructures among other
sectors. On the other hand, sectors 36 and 37 were estimated as the
most technically efcient sectors in both equation systems and this
result can be attributed to the large amount of investment by
the government in these sectors. Although measuring technical

Other Manufacturing Industries.

1188

N. Khiabani, K. Hasani / Energy Economics 32 (2010) 11821190

Table 3
Parameter estimates.
Equation system (System II)
Parameter
ly
ey
e
l
el
ee
ll
lt
et
kt
0
y
l
e
k
t
tt
yt

Equation system (System I)

Estimate

Asymptotic t-statistics

Estimate

Asymptotic t-statistics

0.0209
0.0106
0.2421
1.248
0.0710
0.0474
0.0564
0.0104
0.00746
0.00294
0.6219
0.8168
0.8937
2.16
1
0.0972
0.00301
0.00328

4.569
1.817
2.672
17.798
7.946
2.957
5.154

0.02156
0.0174
0.2455
1.2002
0.0665
0.0482
0.0601
0.0101
0.00715
0.00295
0.3169
0.9895
0.8870
2.2560
1
0.0896
0.0026
0.00323

5.348
4.020
2.864
17.996
7.424
3.558
5.464

10.409
5.896

0.501
3.253
10.974
8.55

3.038
2.883
0.997

10.404
5.599

0.3071
4.634
10.928
8.635

3.248
2.562
1.1034

Indicates signicance at 1%.


Indicates signicance at 5%.

efciency in input or output form may cause cardinal and ordinal


differences for sectors; sectors 33 and 34 in both equation systems are
the least technically efcient sectors and sectors 36 and 37 are the
most.
In terms of generalized cost functions in equation systems (System
I) and (System II), returns to scale can be expressed as:
"
#1
 s #1


log C
RTS =
= y + yy logy + jy log kj wj + yt t
logy
j
"

potential percent cost reduction for achieving both allocative and


technical efciencies as SI.

AI =



CC AE
AE
; C = C kj = 1
AE
C

TI =

CC
C TE

SI =



CC
E
; C = C kj = 1&bi = 1 or ai = 1 :
CE

3
"
#1
 s  1


 y
6 log C 7
+ jy log kj wj + yt t :
RTS = 4
y 5 = y + yy log
a
j
log
a
System II

In both cases, RTS b 1 implies a decreasing return, RTS N 1 implies


an increasing return, and RTS = 1 implies a constant return to scale.
The estimated returns to scale are reported in Table 5, and as shown,
for all sectors in both equation systems RTS b 1. In this case, as
mentioned in Section 3.1 about the relationship between return to
scale and input (output) technical efciency, in both equation
systems we expect bi to be less than ai,(bi b ai,i = 31,39). As seen
in all rows of Table 5, input efciency (inefciency) is less (greater)
than output efciency (inefciency).
Since the presence of any kind of inefciency results in extra
cost for rms, allocative inefciency (AI) can be dened as
potential percent cost reduction for achieving allocative efciency
(see Eakin and Kniesner (1988)). Similarly, technical inefciency
(TI) can be dened as potential percent cost reduction for
achieving technical efciency, and eventually we can dene the

Table 4
Estimated allocative efciency parameters with normalization relative to labor (L).
Equation system (System II)

Equation system (System I)

Parameter

Estimate

Asymptotic t-statistics

Estimate

Asymptotic t-statistics

kl
ke
kk

1
1.762
1.248

1
1.148
1.199

9.29764
12.2324

Indicates signicance at 1%.

7.47708
11.8545

; C

TE

= C bi = 1 or ai = 1 i = 31; ; 39 j = K; L; E

System I
2

TE

The computed AI, TI, and SI are presented in Table 6. Sectors 324
and 385, which have the most AI among all sectors, are very far from
free enterprise. Specically sector 38 has always been inuenced by
political decisions of governmental managers rather than economic
decisions based on prot maximizing. Sector 32 has often faced many
constraints on import of raw material that needs to produce its
output. Except sector 39 whose structure is not denite, the least AI
belongs to sectors 36 and 31 whose environments in comparison with
other sector's have less constraints and governmental interventions.
Consistent with the estimation of technical efciency parameters,
sectors 33 and 34 have the most TI, and sectors 36 and 37 have the
least TI. On average in Iran's manufacturing, increase of cost due to
technical inefciency is greater than that due to allocative inefciency.
In light of the fact that real demand of rm for input j (xj) is derived
C
 s ,
from the shadow cost function applying Shepard's Lemma, xj =
log wsj

partial elasticities of factor substitution (ij), the own-price (Eii) and


cross-price elasticities (Eij) are calculated as follows:
ii
s
+ si 1
ssi
ij
s
Eij = s + sj 1
si

Eii =

ij =

ij + ssi
ssi ssi

i = K; L; E
i; j = K; L; E
i; j = K; L; E ij:

To demonstrate the impact of neglecting each kind of inefciencies


on the biased estimated elasticities, we estimated them in six diverse

4
5

Textile, Wearing Apparel and Leather Industries.


Manufacture of Fabricated Metal Products, Machinery and Equipment.

N. Khiabani, K. Hasani / Energy Economics 32 (2010) 11821190


Table 5
Estimated input and output technical efciencies and computed return to scale for each
sector.
Equation system
(System II)

Equation system
(System I)

Sector code

ai

RTS

bi

RTS

31
32
33
34
35
36
37
38
39

0.688
0.608
0.510
0.484
0.593

0.483
0.482
0.493
0.488
0.480
0.486
0.481
0.479
0.489

0.641
0.539
0.461
0.443
0.552

0.485
0.482
0.492
0.489
0.484
0.486
0.482
0.481
0.489

1.000
0.773
0.56
0.417

1.000
0.756
0.492
0.371

Indicates signicance at 1%.


ai = output technical efciency.
bi = input technical efciency.

states, reported in Table 7. The elasticities shown in every row were


estimated by a specic assumption: in the rst row with the assumption
of allocative and technical efciencies (AE and TE), in the second row
with the assumption of allocative efciency but output technical
inefciency (AE and OTI), in the third row with the assumption of
allocative efciency but input technical inefciency (AE and ITI), in the
fourth row with the assumption of allocative inefciency but technical
efciency (AI and TE), in the fth row with the assumption of allocative
and output technical inefciencies (AI and OTI), and in the sixth row
with the assumption of allocative and input technical inefciencies (AI
and ITI). Comparing the rst row with the forth row, we see that
considering only allocative inefciency causes the elasticity of substitution between labor and energy (LE) and capital and energy (KE) to
increase (LE from 2.316 to 1.899 and KE from 0.459 to 0.168), but
between capital and labor (KL) to decrease (from 0.254 to 0.157).
However, when we consider technical inefciency too, LE and KE
decrease and KL increases. LE increases to 2.15 and 2.013, KE increases
to 0.614 and 0.611, and KL decreases to 0.289 and 0.276 in the fth
and sixth rows, associated with output and input forms of technical
inefciency. It shows that if we only consider allocative inefciency, we
may go wrong in not only the amount but also the direction of change of
estimated elasticities of substitution.
The absolute value of price elasticity of demand for energy and
capital decrease, and for labor increases, when we consider allocative
inefciency alone (From 0.855 to 0.628 for capital, 0.084 to 0.02 for
energy, and 0.287 to 0.452 for labor. See the rst and fourth rows of
Table 7.). Although after considering technical inefciency, the change
of these elasticities continues with the same trend, it can be stated that
we mistake the amount of estimated elasticities, if we only consider

Table 6
Increase of cost due to technical and allocative inefciencies and both of them for each
sector.
Sector
code
31
32
33
34
35
36
37
38
39

Equation system (System II)

Equation system (System I)

TI

AI

SI

TI

AI

SI

53.266
76.372
101.167
119.106
84.252
0.000
34.522
97.895
149.804

2.441
4.148
3.798
2.995
2.059
2.819
2.398
3.852
0.646

57.135
83.867
109.102
126.023
88.260
2.819
37.825
105.740
151.975

21.28
30.754
39.874
42.336
29.369
0.000
12.863
36.05
53.807

2.564
4.686
3.486
3.011
2.151
2.950
2.625
4.388
0.255

24.391
45.246
45.246
46.623
32.153
2.950
15.826
42.021
54.169

TI = increase of cost due to technical inefciency.


AI = increase of cost due to allocative inefciency.
SI = increase of cost due to both technical and allocative inefciencies.

1189

Table 7
Estimates of elasticities of demand and substitution evaluated at the means of the data.
Assumption of estimation

LE

KL

KE

EEE

ELL

EKK

AE and TE
AE and OTI
AE and ITI
AI and TE
AI and OTI
AI and ITI

2.316
2.287
2.165
1.899
2.150
2.013

0.254
0.417
0.394
0.157
0.289
0.273

0.459
0.847
0.811
0.187
0.614
0.611

0.855
0.634
0.600
0.628
0.601
0.541

0.287
0.381
0.365
0.452
0.455
0.448

0.084
0.118
0.109
0.020
0.024
0.009

OTI and ITI denote the existence of output and input technical ineffeciencies
respectively. AI denotes the existence of allocative inefciency. AE and TE denote the
existence of allocative and technical efciencies respectively.

allocative inefciency. From the rows of Table 7, it is understood that


considering either technical or allocative inefciency alone is not
adequate to estimate the unbiased elasticities. Even that the technical
inefciency is considered in input or output form can change the
estimation of elasticities, too. On the other hand, as shown by Atkinson
and Cornwell (1994a), technical and allocative inefciencies are
correlated, in the sense that ignoring technical (allocative) inefciency
may seriously bias estimates of allocative (technical) inefciency. As a
result, we recommend that both allocative and technical inefciencies
should be considered in order to estimate the elasticities.
The negative sign of elasticity of substitution between energy and
capital can be ascribed to the low price of energy, in the sense that, the
low price of energy could never have compelled manufacturing rms
to think about substituting energy for capital in the long-run. In
addition, the low price elasticity of energy whose absolute value
decreases after considering either technical or allocative inefciencies
shows that only a large increase in energy price can change the energy
demand of manufacturing.

6. Summary and conclusions


Using a comprehensive model and panel data from Iran's manufacturing sector, this study demonstrated the importance of joint
estimation of technical and allocative inefciencies in decreasing the
biasness of estimated elasticities of demand and substitution. Since this
model also considers the technical inefciency of rms, in both input
and output forms, it is a more generalized model than that of past
studies like Christopoulos and Tsionas's (2002) and Burki and Khan's
(2004) in which only allocative inefciency has been considered.
The most important nding of this paper is that computed elasticities
of demand and substitution are affected by the presence of both allocative
and technical inefciencies. Hence, it seems that empirical economists for
more precise interpretation of estimated elasticities must take into
account the technical weakness of rms along with mis-allocation of
resources. In addition, using different forms of technical inefciency,
input or output form, can relatively affect the estimated elasticities.
The empirical results show some important ndings related to
allocation of resources in Iran's manufacturing. First, we found a negative
elasticity of substitution between capital and energy. This phenomenon
can be interpreted as an outcome of the low price of energy caused by
government's intervening policies. Although, regarding hi-tech equipment, it was possible for manufacturing rms to substitute energy and
capital, but they have never had enough incentive to do that seriously.
Second, from the estimated allocative efciency parameters, it can be
concluded that if operating environment does not impose any constraints
on the rms and the managers of rms have enough incentive to strictly
minimize their costs, the utilization of labor relative to capital will
decrease, and the utilization of energy relative to both capital and labor
will increase. With regard to the government's tendencies in recent years
to remove the environmental constraints and privatize the manufacturing rms, the latter result is very noticeable for policy makers. They
should notice that these policies, without a remarkable change in energy

1190

N. Khiabani, K. Hasani / Energy Economics 32 (2010) 11821190

price, seem to end in more wasting of energy and less incentive to


substitute energy with other factors of production.

Appendix A
According to the second revision of the International Standard
Industrial Classication (ISIC 2), manufacturing rms are divided into
nine general divisions associated with codes 31 through 39. They are
presented in Table A.
Table A
Nine sectors of manufacturing according to ISIC 2.
Code Sector
31
32
33
34
35
36
37
38
39

Manufacture of Food, Beverages and Tobacco


Textile, Wearing Apparel and Leather Industries
Manufacture of Wood and Wood Products, Including Furniture
Manufacture of Paper and Paper Products, Printing and Publishing
Manufacture of Chemicals and Chemical, Petroleum, Coal, Rubber and
Plastic Products
Manufacture of Non-Metallic Mineral Products, except Products of
Petroleum and Coal
Basic Metal Industries
Manufacture of Fabricated Metal Products, Machinery and Equipment
Other Manufacturing Industries

References
Atkinson, S.E., Cornwell, C., 1994a. Parametric estimation of technical and allocative
inefciency with panel data. International Economic Review 35, 231243.
Atkinson, S.E., Cornwell, C., 1994b. Estimation of output and input technical efciency
using a exible functional form and panel data. International Economic Review 35,
245255.
Atkinson, S.E., Halvorsen, R., 1984. Parametric efciency tests, economies of scale, and
input demand in U.S. electric power generation. International Economic Review 25,
647662.
Atkinson, S.E., Halvorsen, R., 1986. The relative efciency of public and private rms in a
regulated environment: the case of US electric utilities. Journal of Public Economics
29, 281294.

Bauer, P.W., 1990. Recent development in the econometric estimation of stochastic


frontier. Journal of Econometrics 46, 3956.
Berndt, R.E., Wood, O.D., 1975. Technology prices and derived demand for energy. The
Review of Economics and Statistics 56, 259268.
Burki, A.A., Khan, M.-H., 2004. Effects of allocative inefciency on resource allocation
and energy substitution in Pakistan's manufacturing. Energy Economics 26,
371388.
Christopoulos, D., Tsionas, E., 2002. Allocative inefciency and the capital-energy
controversy. Energy Economics 24, 305318.
Eakin, K.B., Kniesner, J.T., 1988. Estimating a non-minimum cost function for hospitals.
Southern Economic Journal 54, 583597.
Fare, R., Lovell, C.A.K., 1978. Measuring the technical efciency of production. Journal of
Economic Theory 19, 150162.
Farrell, M.J., 1957. The measurement of production efciency. Journal of the Royal
Statistical Society 120, 253290.
Greene, W.H., 1980. On the estimation of a exible frontier production model. Journal of
Econometrics 13, 101115.
Grifn, J.M., Gregory, P.R., 1976. An intercountry translog model of energy substitution
responses. The American Economic Review 66, 845857.
Hudson, E.A., Jorgenson, W.D., 1974. US energy policy and economic growth 1975
2000. Journal of Economics 5, 461514.
Im, K., Pesaran, H., Shin, Y., 2003. Testing for unit roots in heterogeneous panels. Journal
of Econometrics 115, 5374.
Karagiannis, G., Mergos, G.J., 2002. Estimating theoretically consistent demand systems
using cointegration techniques with application to Greek food data. Economics
Letter 74, 137143.
Khodadad-Khashi, F., 2007. Competition and concentration in manufacturing sector
during rst socioeconomic and cultural development plan of Iran. Iranian Economic
Review 12, 127141.
Kumbhakar, S., Tsionas, E., 2005. Measuring technical and allocative inefciency in the
translog cost system: a Bayesian approach. Journal of Econometrics 126, 355384.
Lau, L.J., Yotopoulos, P.A., 1971. A test for relative efciency and application to Indian
agriculture. The American Economic Review 61, 94109.
Lewbel, A., Ng, S., 2005. Demand systems with nonstationary prices. The Review of
Economics and Statistics 87, 479494 (There is no refrence in the text to this paper,
shall I remove this).
Lim, H., Shumway, C.R., 1997. Technical change and model specication: U.S.
agricultural production. American Journal of Agricultural Economics 79, 543554.
Maddala, G.S., Wu, S., 1999. A comparative study of unit root tests with panel data and a
new simple test. Oxford Bulletin of Economics and Statistics 61, 631652.
Pindyck, S.R., 1979. Inter fuel substitution and the industrial demand for energy: an
international comparison. The Review of Economics and Statistics 61, 169179.
Toda, Y., 1976. Estimation of a cost function when the cost is not minimum: the case of
soviet manufacturing industries. The Review of Economics and Statistics 58,
259268.

Vous aimerez peut-être aussi