Energy intensity: a decomposition exercise for Pakistan.
Ullah, Akbar ; Khan, Karim ; Akhtar, Munazza 等
ABSTRACT
In this study an attempt has been made to decompose the change in
energy intensity into changes in efficiency and activities. The study,
covering the period from 1972 to 2011, uses the decomposition method of
Fisher Ideal Index. Our analysis shows that energy intensity has
increased by 45 percent on average between 1972 and 2011. The major
impetus for the increase is inefficiency in its use, i.e. 52 percent of
the increase is caused by the inefficiency. Alternatively, for the same
unit of output we are using more energy now as compared to 1972. Most of
the inefficiencies persist in the consumption of electricity, followed
by gas. The oil sector is relatively efficient as the oil price hikes
have been improving efficiency consistently. However, structural changes
have intensifying effects on the oil intensity. The main deriver of the
increase in aggregate energy intensity is electricity with its average
intensity index of 1.75. The aggregate intensity of oil and gas is
falling, following the recent price and supply crisis. Whereas in
countries like U.S, China, France and India efficiency improvement has
played major role in reducing aggregate energy intensity; in Pakistan,
the worsening of efficiency is a dominant factor in increasing aggregate
energy intensity.
1. INTRODUCTION
Since the recent energy crises, the research in this strand has
increased considerably. A variety of its dimensions have been examined
in the literature. For instance, higher energy prices; instability in
the supplies of its various components; its rapid depletion and global
warming are some of its dimensions, which have been the focus of
discourse among both researchers and policy-makers. Equally, energy
intensity measuring the energy consumption to GDP ratio has been an
important component of energy policies [Ang (2004); Liu and Ang (2007);
Jimenez and Mercado (2013)]. In particular, there is a special focus on
sorting out the contribution of energy efficiency--ratio of sectoral
specific energy consumption to sectoral GDP--to alienate the impact of
efficiency on energy intensity from other relevant factors. This is
because energy efficiency is recognised as one of the most
cost-effective strategies to address crosscutting issues of energy
security, climate change and competitiveness [IDB (2012)]. Consequently,
the information regarding energy intensity, its efficiency or activity
aspects are useful tools for policy decisions and evaluation and are
regularly in practice in most of the advanced countries.
Pakistan is currently faced with severe energy crisis. In
particular, it is facing formidable challenges in meeting its energy
requirements and providing adequate energy to users at affordable costs.
Electricity shortage is continuously widening since 2006-07. For
instance, the gap between demand and supply of electricity increased to
the level of 5000 MW in 2011. (1) Such shortages have adverse
consequences for the economy of Pakistan. According to Abbasi (2011),
energy shortages have cost the county up to 2 percent of GDP per annum.
Similarly, Siddiqui, et al. (2011) proclaim that the loss in industrial
output due to power shortages is estimated to be from 12 percent to 37
percent. They have also forced the closure of hundreds of factories,
paralysing production and exacerbating unemployment. Additionally, they
imperil much-needed investments in development and infrastructure.
Despite these facts, Pakistan's energy intensity per unit of GDP is
higher relative to Countries like India, USA, Germany, Japan and China
[Allcott and Greenstone (2012); IEA (2012c)]. Also, in case of Pakistan,
it has taken a rising trend over time. For instance, the consumption of
oil in 1972 was 12 percent of its consumption level in 2011. Similarly,
it was 9 percent in case of gas and 7 percent in case of electricity. At
the same time, gross value added in 1972 was 14 percent of its 2011
level. These trends show that we are now using more energy per each unit
of economic activity.
However, to overcome energy crisis and achieve energy security, we
must have to bring efficiency in the usage of energy. But before any
successful policy formulation, our academia and policy makers must be
aware of the past trends and current status of the energy intensity. So
far, commendable research has been done on energy issues in Pakistan but
most of the studies have been conducted in the context of changes in
energy prices and their relation to economic growth, inflation and other
macroeconomic indicators [Malik (2007, 2008, 2012); Kiani (2009); Jamil
and Ahmed (2010); Syed (2010); Khan and Ahmad (2011); Siddiqui, et al.
(2011)]. To our knowledge, the only study conducted on energy intensity
in Pakistan is done by Alam and Butt (2001) which uses Sun (1998)
'complete decomposition method'. The Sun (1998) method of
decomposition, based on jointly created and equally distributed
principle, is weaker as compared to recently developed decomposition
techniques. Second, the current energy crises have intensified since
2005, so the study by Alam and Butt (2001) has been a little bit older
now. In this paper, we make an endeavor to address these two issues. The
study provides an empirical decomposition of energy intensity into its
constituent factors, efficiency and economic activity for Pakistan. We
apply Fisher Ideal Index Decomposition Approach (IDA) and cover a period
from 1972 to 2011. In our analysis, we show the effects of change in
efficiency and changes in activities on the change in energy intensity.
This study contributes to the literature in three main aspects. First,
the time of the study is of particular importance. It covers the period
that includes all the three major oil price shocks as well as the recent
energy crisis in Pakistan. Second, instead of considering the overall
energy consumption, we first construct the indices at component level
and then aggregate the individual indices to understand the overall
trends. This has allowed us to see the intensities pattern of oil, gas
and electricity separately. Finally, we have used the most suitable
decomposition tools recommended in most recent literature.
Rest of the paper is organised in four sections. Section 2 reviews
the decomposition methodologies and their empirical implications.
Detailed methodology and the construction of variables are discussed in
Section 3. In Section 4, we provide the details of our empirical
analysis. Section 5 concludes the paper.
2. REVIEW OF DECOMPOSITION LITERATURE
In order to decompose aggregates into their component parts,
different decomposition methodologies have been developed and applied in
empirics. These different methodologies can be broadly divided into four
groups: Shift Share Analysis (SSA), Growth Accounting Analysis (GAA),
Index Decomposition Analysis (IDA) and Structural Decomposition Analysis
(SDA) [Fengling (2004)]. SSA is mainly observed in regional studies and
GAA is used in decomposing identity [Fengling (2004); Szep (2013)]. In
the same manner, the Index Decomposition Analysis (IDA) and Structural
Decomposition Analysis (SDA) are widely used in energy and emission
studies [Ang and Zhang (2000)]. However, the choice between these two
methodologies largely depends on their ease of application and data
requirements. SDA uses information from input-output Tables while IDA
uses aggregate data at the sector-level. The advantage of IDA is its
lowest data requirement along with its strong theoretical foundation
[Hoekstra and Bergh (2003); Fengling (2004); Liu and Ang (2007)]. In
contrast, SDA can distinguish between a range of technological effects
and final demand effects that are not possible in the IDA. However,
Hoekstra and Bergh (2003) have formally shown that IDA techniques can be
transferred to SDA. Similarly Boer (2009) proved that the generalised
Fisher approach, introduced in IDA is equivalent to SDA. Because of this
equivalence and its lower data requirements, IDA remains a popular tool
of decomposition.
In the energy decomposition, IDA is extensively used since 1980s.
The earlier literature in which the energy intensity is decomposed into
contributions from structural and efficiency effects left an unexplained
residual term [Bossanyi (1983); Boyd, et al. (1988); Li, et al. (1990);
Howarth, et al. (1991); and Park (1992)]. The long-mean Divisia index,
proposed by Ang and Choi (1997), was an improvement to these earlier
techniques because it leaves no residual term. Since then several other
perfect methods have been developed by different authors. (2) For
instance, Sun (1998) introduced Refined Laspeyres Index (RLI), which is
based on the principle of jointly created and equally distributed
principle. Similarly, Mean Rate of Change Index (MRCI) by Chung and Rhee
(2001) leaves no residue. Further, the decomposition technique of
Albrecht, et al. (2002), which is based on Shapley Value and the Log
Mean Divisia Method and Modified Fisher Ideal Index method by Fengling
(2004) are yet other approaches that are perfect in decomposition.
All of these methods have been extensively used in empirical
studies. (3) In case of energy intensity, different approaches have come
up with different results on the relative roles of the efficiency and
structural effects. For instance, Kander and Lindmark (2004) found that
efficiency was a major factor in the improvement of energy intensity in
case of Sweden. Similarly, a number of other case studies have been
conducted and have come up with similar results [Liao, Fan, and Wei
(2007) for China between 1997-2002; Metcalf (2008) for U.S. between
1970-2001; Sahu and Narayanan (2010) for Indian manufacturing between
1990-2000; Shahiduzzaman and Alam (2012) for Australia between
1978-2009; Song and Zheng (2012) for China between 1995-2009; Szep
(2013) for Czech Republic, Slovakia, Slovenia, Poland and Hungary
between (1990-2009)]. In most of the studies, at disaggregate level
analysis, the share of structural changes increases even when the same
methods of decomposition were used [Karen, et al. (2004) for China; and
Huntington (2010) for U.S. The possible justification is that
aggregations cause the overstatement of the contribution of sub-sector
energy productivity improvements while assigning insufficient weight to
the role of sectoral shift. Despite this disadvantage, the aggregate
level studies are preferred due to their relatively comprehensive
coverage. Likewise, the literature places emphasis on the efficiency
effects in reducing energy intensity, especially in the advanced
countries [Ang (2004); Fengling (2004); Liu and Ang (2007); Ang, et al.
(2009)].
However, the selection of suitable index decomposition method is
very important for getting accurate results. Generally, a method, which
leaves no residual is regarded as the most desirable. Such methods are
referred to as perfect decomposition methods. Other desirable properties
that IDA must satisfy to become a good decomposition method are
adaptability, ease of use and interpretation, consistency in aggregation
and robustness to zero and negative values [Liu and Ang (2007)]. The two
methods that satisfy most of these properties are Log Mean Divisia
techniques and Fisher Ideal Index [Ang (2004); Fengling (2004); Liu and
Ang (2007); Ang, et al. (2009)]. Given its suitability, we use Fisher
Ideal Index in our study like Metcalf (2008) and Song and Zheng (2012).
Besides the perfect decomposition and its robustness to zero-negative
values, the Fisher Ideal Index satisfies time-reversal test, factor
reversal test and proportionality test as well.
3. METHODOLOGY FOR DECOMPOSITION OF ENERGY INTENSITY AND DATA
In this section, we provide the detailed methodology of the study.
Also, we give a description of the construction of our variables.
3.1. The Decomposition Methodology
As is stated earlier, decomposition analysis is used to break down
the aggregate series into understandable and meaningful components. In
this study, our purpose is to use these techniques to decompose the
change in aggregate energy intensity into changes in economic activity
and changes in efficiency. Also, we decompose change in total
consumption. For our analysis, the aggregate energy intensity is defined
as the ratio of total energy consumption to aggregate output of the
economy:
[e.sub.t] = [E.sub.t]/[Y.sub.t] ... (3.1)
[e.sub.t] = [E.sub.t]/[Y.sub.t] = [[summation].sub.i]
[[E.sub.it]/[Y.sub.it]] [[Y.sub.it]/[Y.sub.t]] = [summation] [e.sub.it]
[s.sub.it] ... (3.2)
Where, [E.sub.t] is aggregate energy consumption, [Y.sub.t] is
gross domestic product, [Y.sub.it] is sectoral output and [E.sub.it]
sector specific intensity. We want to choose suitable analytical tools
to decompose the aggregate changes in energy, [DELTA][e.sub.t], into
changes in economic activity and changes in efficiency, that is, into
[DELTA][s.sub.it] and [DELTA][e.sub.it] respectively. For this purpose,
we use Fisher Ideal Index. Fisher Ideal Index has many advantages over
most of the other methods as are mentioned in the previous section.
Using [e.sub.0] to denote aggregate energy intensity in the base
year, we construct energy intensity index and then derive its
decomposition. (4)
[e.sub.t]/[e.sub.0] [congruent to] [I.sub.t] = [I.sup.act.sub.t]
[I.sup.eff.sub.t] ... (3.3)
Where, [I.sup.act.sub.t] is the corresponding activity index and
[I.sup.eff.sub.t] is the efficiency index. As the equation indicates,
the aggregate energy index is decomposed into activity and efficiency
indexes with no residual term and this is guaranteed only by Fisher
ideal index. (5) Once we have these indices, we can easily determine the
amount of change in energy consumption, which is caused by changes in
efficiency and the part that is due to change in activity. Using
[E.sub.0] to denote energy consumption that would have prevailed had
energy intensity not changed since the base year. Following Metcalf
(2008), this is done below:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3.4)
The term [DELTA][E.sub.t] indicates change in energy consumption,
which is the difference between actual consumption in a given year and
the consumption, which would have occurred had energy intensity remained
at 1972 level, that is, [E.sub.t] - [E.sub.0]. As is shown in the
equation, this has enabled us to decompose a given change in energy
consumption, relative to a base year, into changes in efficiency and
changes in activities.
3.2. The Construction of Variables
We carry out the decomposition analysis for various components of
energy, i.e. oil, gas and electricity. Together, these three comprise
about 90 percent of the total energy consumption in Pakistan. Here, we
present the details of the data of all the three components. The energy
year book reports the oil consumption data under six headings: household
consumption; industrial consumption; agricultural consumption; transport
consumption; power consumption; and other government consumption. In
order to construct the indices, we need the contribution of each of
these sectors to the national gross valued added. For this purpose, we
make certain matching operations. For instance, for the share of
household sector in gross value added, we use household final
consumption expenditure. (6) Similarly, for the share of industrial
sector, we take into account the industrial value added net of the
electricity and gas distribution. This is because electricity and gas
distribution is considered to compute the share of power sector in the
total gross value added. The gross value added accruing from transport,
storage, and communication is taken as the share of transport sector.
For agricultural sector, its share in gross valued added is considered
for the analysis.
The data on the consumption of gas is reported for seven sectors,
i.e., household, commercial, industrial, cement, fertiliser, power, and
transport sectors. To make it congruent to the national accounts data,
we merge the data of cement and fertilisers with the industrial
consumption of gas. In the same manner, the data of transport
consumption is merged into the data of commercial sector. In the overall
contribution of commercial sector to gross value added, we include the
value additions of transport, storage and communications, wholesale and
retail trade, and finance and insurance. The shares of the remaining
sectors are constructed in a similar way to those, which are constructed
in the case of oil sector.
In the case of electricity, the consumption of traction, street
light and other government sector are eliminated from the total
electricity consumption. This is not going to make any difference
because collectively the share of these sectors in the total consumption
of electricity is less than 7 percent. Consequently, for our analysis of
the electricity sector, we consider four sectors, i.e. household,
commercial, industrial, and agriculture.
Finally, the gross value addition of each of the mentioned sectors
is in constant prices of 2000. The data is taken from four sources:
Energy Year Books; Statistical Supplements; Hand Book of Statistics; and
the World Development Indicators of the World Bank. The descriptive
statistics of the various sectors is summarised in the appendix.
4. RESULTS OF THE DECOMPOSITION ANALYSIS
In order to carry out the detailed analysis of the changes in
intensity, we have done separate analysis for each of the three major
energy components. In this section, we provide the details of our
empirical results one by one.
4.1. Oil Energy
The decomposition of oil intensity between 1972 and 2011 is
presented in Figure 4.1. Overall, the oil intensity is 22 percent higher
in 2011 as compared to that of 1972. The highest intensity is in the
year 2000, which is 88 percent higher as compared to that of 1972. On
average per annum, intensity is 37 percent higher than that of 1972.
However, the indices of activity and efficiency are giving divergent
patterns. For instance, the activity index is 24 percent higher in 2011
as compared to 1972 while the efficiency index is 1 percent lower in
2011 as compared to its base value. The highest value of activity index
is 1.48, which is in the year 2000 while the highest value of efficiency
index is 1.38 in 1996. Besides, the activity index remains above its
1972 level for whole time period covered in this study. Since 1980, the
indices show rising trends for the onward two decades with the activity
index dominating the efficiency index. This means that during this
period the share of oil using sectors increased in relative terms.
However, after 2000 we have experienced sharp reduction in oil intensity
with efficiency as a dominant factor in this change.
[FIGURE 4.1 OMITTED]
The data of oil consumption indicate that total oil consumption in
2011 would have been 3387509 tonnes lower had the energy intensity
remained at its 1772 level. (7) Equation 3.4. can be used to decompose
this change into the changes in activities and a change in efficiency.
According to our analysis, change in economic activities cause oil
consumption to increase by 3596498 tonnes in 2011 as compared to 1972.
In contrast, the change in efficiency causes oil consumption to drop by
217989 tonnes in 2011. The year wise details of this analysis are given
in the appendix.
As we compare the trends in our indices with the changes in oil
prices, some interesting results emerge. The global economy has
experienced three big shocks in oil prices. The first was in 1973 when
the Organisation of Petroleum Exporting Countries (OPEC countries)
imposed embargo on oil exports in response to Arab-Israel war. The
second shock occurred in 1979, which was mainly caused by the Islamic
Revolution in Iran. From 1983 to 1998, oil prices remained stable both
in domestic as well as international markets. However, since 1999 the
world is experiencing a third big oil price shock in the global history.
In their meeting in March 1999, the OPEC countries agreed to cut the oil
production with a view to increasing the prices of crude oil to around
$20 per barrel. As a consequence, the oil prices very quickly surpassed
the $20 per barrel with a dramatic increase in the new century. For
instance, in 2003-04 oil prices were 11 percent higher than those of
2002-03 and around 41 percent higher in the following year compared to
those of 2003-04. In the same manner, in 2007-08 oil prices were 53
percent higher as compared to those in the preceding year. Continuing
with the rising pattern, the prices reached to a record level of about
$150 per barrel in 2008-09.
The comparison of oil price history with Figure 4.1, in particular
with the efficiency index, shows that the indices remained almost stable
during the whole 1970s. In 1980, efficiency started improving, which
continued until 1984. During this period, the efficiency was better than
that of 1972. Onwards, the indices have steadily increased and this
increasing trend continues up to 1998. After 2000, the aggregate
intensity strongly falls and the dominant factor for this fall can be
seen as the efficiency index. For instance, the efficiency index falls
from 1.16 in 2003 to 0.82 in 2004 and during the same period, oil prices
increased by 41 percent as compared to preceding year. This trend holds
not only for the international prices but also for the domestic
variation in the prices of furnace oil, HOBC, HSD etc. If the relation
holds true, it implies that whenever oil prices increase, we increase
efficiency in its use. This is an important implication and requires
in-depth analysis. After the year 2000, the activity index also shows
declining trend; however, it is not as pronounced as the efficiency
index.
4.2. Gas Energy
In this sub-section, we extend our analysis to the case of gas. The
decomposition analysis of gas intensity, shown in Figure 4.2, indicates
that gas intensity in 2011 is 58 percent higher as compared to that of
1972. The intensity is the highest in 2005, which is 87 percent higher
than that of 1972. On average, the gas intensity is 43 percent higher
than 1972 for the onward period. The index of activity is dominated by
the index of efficiency. For instance, the activity index in 2011 is 06
percent higher as we compare it with that of the 1972. In contrast, the
efficiency index is 50 percent higher in 2011. The highest value of
activity index is 1.27 in 1999 while the highest value of efficiency
index is 1.81 in 2008. In the same way, the lowest value of the
efficiency index is estimated at 0.97 in 1974. In general, the
efficiency index remains above its 1972 level for most of the period
covered in this study. As is evident from the figure below, the
aggregate intensity index is strongly driven by the efficiency index in
case of gas consumption.
Moreover, the intensity index goes through two notable upward
spikes, one in around 1981; and the second is the most prolonged one
beginning in 2000 and lasts up to 2008. 2008 onwards, we have
experienced a declining trend in gas intensity with efficiency as the
dominant factor in this change. One factor for the higher intensity in
the beginning of 21st century can be the policies of the Musharraf
administration, which converted most of electricity or oil run
industries to gas. For instance, one critical sector in this regard is
transport sector. In 1998, transport sector was using 490 (mm cft) of
gas, which increased to 1 13055 (mm cft) in 2011. Again, a striking
feature is that the increase is mainly dominated by the efficiency
changes rather than the activity changes.
[FIGURE 4.2 OMITTED]
The gas consumption data indicate that total gas consumption of
1240672(mm cft) in 2011 would have been 784286.2 (mm cft) had energy
intensity remained at its 1772 level. The increase is jointly shared by
the activity and efficiency factors. For instance, the increase in gas
consumption caused by the change in economic activity is estimated to
55458.55 (mm cft) in 2011 as we compared it to the level of 1972.
Similarly, the change in efficiency causes gas consumption to increase
by 400927.3 (mm cft), in 2011 compared to the level of 1972. The year
wise details of the changes in indices are given in the appendix.
4.3. Electricity Energy
This section provides the blueprints of the intensity of
electricity. The details of the decomposition analysis are shown in
figure 4.3. Overall, the intensity of the electricity is 110 percent
higher in 2011 as compared to that of 1972. This difference is highest
in the year 2007 where the index is 122 percent higher than that of the
base year. For all the periods onwards the base year, the index is on
average 75 percent higher than that of the base year. So far the
decomposition is concerned; the activity index is 08 percent lower
whereas the efficiency index is 129 percent higher compared to their
corresponding values in 1972. The efficiency effect is dominating the
activity effect for almost the whole period covered in this study. The
highest value that the activity index takes is 1.03 in 1980 while the
highest value that efficiency index takes is 2.33 in 2003. The
efficiency index remains above its base year value for almost the whole
period covered. As a consequence, the aggregate intensity index is
perfectly guided by the efficiency index in case of electricity
consumption. As is shown in the figure, both the aggregate index and the
efficiency index are increasing over time while the activity index
remains static and sometimes is slightly below its 1972 level. The
analysis shows that each unit of output produced in Pakistan uses more
and more electricity with each passing year.
[FIGURE 4.3 OMITTED]
Given our analysis, the total electricity consumption of 71845
(GWH) in 2011 would have been 34215.9 (GWH) had the energy intensity
remained at its base year level. This translates that our electricity
consumption is 110 percent higher than would have been had the energy
intensity remained at the level of the base year. We have shown above
that most of the increase in the consumption of electricity is mainly
driven by the inefficiency in the use of electricity. For instance, the
change in economic activity causes the consumption of electricity to
decrease by 4386.27 (GWH) in 2011 as compared to 1972. In contrast, the
change in efficiency causes electricity consumption to increase by
42015.37(GWH) in 2011 as compared to 1972. The year-wise details for the
whole period of the analysis are given in the appendix.
4.4. Discussion
We have shown above that the changes in efficiency guide the
changes in intensity in case of gas and electricity while change in
activities is a dominant factor in case of oil. There are definitely
some cases where an increase in efficiency index in one sector was
accompanied by a corresponding decrease in other sector. (8) In order to
truly understand the dynamics of efficiency indices, we aggregate the
indices and check whether the change in efficiency is mere a transfer of
activity from one component to another or is just a real wastage of
energy. In this regard, we have taken the weighted average of the
respective individual indices of each component. Weights are given
according to the contribution of each energy component in the total
energy consumption. This analysis is shown in Figure 4.4. The resulting
aggregate activity index smoothly increases and reaches its maximum
value 1.4 in 1999. This trend is mostly explained by the oil energy.
However, after 2000, the index is falling, which may be due to the
severe shocks to the gas supply and oil prices, which are evident
throughout the first decade of this century. This demonstrates that the
share of sectors using energy in the total gross value added is falling
after 2000. The main justification is that in the initial decades of
independence, our economy was moving away from less energy intensive
agricultural sector to more intensive industrial sector. But in recent
decades, the trend is completely different: both the agriculture and
industry are losing their shares to another less energy intensive
services sector. Given the sectoral transformation of the economy, this
result is not surprising.
[FIGURE 4.4 OMITTED]
The aggregate efficiency index shows that increasing energy use per
each unit of output is the dominant factor of the increase in intensity
during the study period. The index is smoothly increasing for most of
the period with the highest value of 1.6 for the year 2008. Ultimately,
this has guided the change in energy intensity in Pakistan since 1972.
It is clearly evident that after 1998, the fluctuations in this index
are caused by the indices of gas and oil.
In conclusion, it is stated that energy intensity in Pakistan
increases by 45 percent on average between 1972 and 2011. A critical
feature of the increase is that 52 percent of the increase is caused by
the inefficiency associated with the use of energy. Alternatively, for
the same unit of output, we are using more energy now as compared to the
per unit use of 1972. Most of the inefficiencies arise in the
consumption of electricity, followed by gas sector. This translates into
that oil sector is relatively efficient as we compare it with gas and
electricity. In particular, the oil price hikes have beneficial effects
on improving the efficiency in oil sector.
5. CONCLUSION
The study is motivated by the recent energy crises in Pakistan.
Most of the existing literature in case of Pakistan deals with the
energy prices and their impact on other macroeconomic variables like
economic growth, inflation, employment etc. To our knowledge, there is
no commendable work on the energy intensity side. We are filling this
gap by providing a complete analysis of the energy intensity. Our
analysis of energy is components-wise to see the trends in different
components like oil, gas and electricity. The study, covering the period
from 1972 to 2011, uses the decomposition method of Fisher Ideal Index.
We disaggregate the change in aggregate energy intensity into changes in
the efficiency and changes in the activities. Our analysis shows that
aggregate energy intensity has steadily increased over time until the
recent years. The consumption of electricity in 1972 was 7 percent of
its consumption in 2011. At the same time, the gross value added in 1972
was 14 percent of its level in 2011. This implies that growth in the
consumption of electricity is greater than the growth rate of gross
value added.
According to our results, energy intensity in Pakistan increased by
45 percent on average between 1972 and 2011 and 52 percent of the
increase is due to the worsening efficiency in its use. In other words,
the use of energy per unit of output has increased significantly since
1972. The inefficiencies in the consumption of electricity are
dominating, followed by the inefficiencies in gas sector. For instance,
in case of electricity, the average index of efficiency is 1.82, which
implies that a given amount of output is now produced with 1.82 (GWH) of
electricity on average whereas the same amount of output required only 1
(GWH) in 1972. In comparison, the average value of activity index is
0.97, again in case of electricity, which is showing that the share of
electricity intensive sectors in total gross value added is declining.
In case of the consumption of gas, the average value of efficiency index
is 1.29 and the average value of activity index is 1.11. It implies that
the level of efficiency also has declined in case of the consumption of
gas. Unlike electricity and gas sectors, the oil sector is relatively
efficient. For instance, the average value of efficiency index in case
of oil is 1.08 while the average value of activity index is 1.26.
In summary, the change in aggregate intensity is mainly caused by
the inefficiencies. The main deriver of the change in aggregate energy
intensity is electricity with its average intensity index of 1.75. The
aggregate intensity of oil and gas is falling following the recent
prices and supply crisis. Countries like China, USA, and India have
experienced significant gains in the aggregate energy intensity over the
past years. In these countries, improvements in efficiency are regarded
as the catalyst behind the changes in energy intensity. In contrast, in
Pakistan efficiency has worsened over time instead of improving. In
general, the change in efficiency index in Pakistan seems to be somehow
related to the price changes. For instance, whenever the price hikes
have taken place, oil efficiency index has improved. Additionally, our
analysis shows that energy sources with relatively low prices are more
prone to increased inefficiencies. This is an important area for future
research; in particular with relevance to policy formulation.
Akbar Ullah <akbarullalif3ipidc.org.pk> is Staff Economist,
Pakistan Institute of Development Economics (PIDE), Islamabad. Karim
Khan <karim.khan@pide.org.pk> is Assistant Professor, Pakistan
Institute of Development Economics (PIDE), Islamabad. Munazza Akhtar
<munazza.akhtar@skt.umt.edu.pk> is Lecturer, University of
Management and Technology (UMT), Sialkot Campus.
APPENDIX
The Construction of Fisher Ideal Index is given below
First we construct Laspeyres and Paasche activity and efficiency
indexes. The Laspeyres activity and efficiency indexes are
[L.sup.act.sub.t] = [summation over (i)]
[e.sub.i0][s.sub.it]/[summation over (i)] [e.sub.i0][s.sub.i0] ... (A1)
[L.sup.ef.sub.t] = [summation over (i)]
[e.sub.it][s.sub.i0]/[summation over (i)] [e.sub.i0][s.sub.i0] ... (A2)
The Paasche activity and efficiency indexes are
[P.sup.act.sub.t] = [summation over (i)]
[e.sub.it][s.sub.it]/[summation over (i)] [e.sub.it][s.sub.i0] ... (A3)
[P.sup.ef.sub.t] = [summation over (i)]
[e.sub.it][s.sub.it]/[summation over (i)] [e.sub.i0][s.sub.it] ... (A4)
Now the Fisher ideal indexes for activity and efficiency are given
as
[I.sup.act.sub.t] = [square root of ([L.sup.act.sub.t]
[P.sup.act.sub.t])] ... (A5)
[I.sup.ef.sub.t] = [square root of ([L.sup.ef.sub.t]
[P.sup.ef.sub.t])] ... (A6)
Using this we can construct the aggregate energy index (2.3) as
below
[e.sub.t]/[e.sub.0] [congruent to] [I.sub.t] = [I.sup.act.sub.t]
[I.sup.eff.sub.t] ... (A7)
Table A1
Descriptive Statistics of Sectors Included in the Analysis
Economic Activity
Sector Components Mean SD
Household Oil 2209813 1074383
Gas -- --
Electricity -- --
Commercial Oil 830726 526754
Gas -- --
Electricity
Transport Oil 274094 168383
Gas -- --
Electricity -- --
Agriculture Oil 686997 293882
Gas -- --
Electricity -- --
Industry Oil 698696 410210
Gas -- --
Electricity -- --
Industry Net Oil 617727 373753
of Electricity Gas -- --
Electricity -- --
Electricity Oil 80969 45366
Gas -- --
Electricity -- --
Intensity
Sector Components Mean SD
Household Oil 0.3242 0.2077
Gas 0.0317 0.0171
Electricity 0.005 0.0027
Commercial Oil -- --
Gas 0.0262 0.0171
Electricity 0.0027 0.0004
Transport Oil 20.4162 2.9852
Gas -- --
Electricity -- --
Agriculture Oil 0.4062 0.2536
Gas -- --
Electricity 0.0065 0.0016
Industry Oil -- --
Gas -- --
Electricity 0.0152 0.0017
Industry Net Oil 1.8379 0.8543
of Electricity Gas 0.4281 0.0531
Electricity -- --
Electricity Oil 29.98 20.7199
Gas 2.3744 0.8548
Electricity -- --
Source: SPB, WB, Ministry of Finance, Hydrocarbon Development
Institute of Pakistan.
Table A2
Oil Consumption Decomposition
Change
E-E^ Activity Change Due Efficiency Due to
Year (tonnes) Index to Activity Index Efficiency
1972 0 1.00 0 1.00 0
1973 -71123.5 1.03 72154.99 0.94 -143278
1974 -99043.7 1.04 85272.88 0.93 -184317
1975 194409.1 1.02 41235.18 1.06 153173.9
1976 -15271.2 1.01 14929.45 0.99 -30200.7
1977 115246.5 1.02 51754.2 1.02 63492.28
1978 125501.8 1.04 120104.4 1.00 5397.38
1979 260575 1.07 226153.2 1.01 34421.72
1980 124849.2 1.08 254885.2 0.96 -130036
1981 85207.14 1.30 941485.5 0.79 -856278
1982 355269.5 1.29 1007013 0.85 -651744
1983 729035 1.27 1055317 0.93 -326282
1984 1150859 1.30 1251281 0.98 -100422
1985 1375348 1.28 1296462 1.02 78886.04
1986 1519733 1.26 1284095 1.04 235637.8
1987 2265023 1.28 1524955 1.13 740067.9
1988 2699635 1.30 1754870 1.15 944764.6
1989 2966072 1.24 1532993 1.22 1433079
1990 3544186 1.27 1833546 1.25 1710640
1991 3294799 1.29 1995895 1.18 1298904
1992 3842105 1.33 2464075 1.17 1378030
1993 4694140 1.37 2870828 1.22 1823312
1994 5603384 1.37 3106205 1.29 2497180
1995 5972654 1.39 3417268 1.28 2555386
1996 7047755 1.35 3425000 1.38 3622754
1997 6927567 1.37 3549863 1.35 3377704
1998 7679937 1.43 4397130 1.31 3282807
1999 7349618 1.44 4500968 1.26 2848649
2000 8151257 1.48 5075029 1.27 3076228
2001 7822836 1.46 4947586 1.25 2875250
2002 6749384 1.42 4499642 1.19 2249742
2003 5977893 1.36 4021050 1.16 1956843
2004 2140302 1.45 4494238 0.82 -2353936
2005 2400226 1.40 4444422 0.86 -2044195
2006 1617995 1.30 3494033 0.87 -1876038
2007 3009863 1.25 3347383 0.98 -337520
2008 3760082 1.10 1500483 1.15 2259600
2009 3292826 1.30 4108234 0.95 -815408
2010 4120181 1.29 4237332 0.99 -117152
2011 3378509 1.24 3596498 0.99 -217989
See text for Construction.
Table A3
Gas Consumption Decomposition
Change
E-E^ Activity Change Due Efficiency Due to
Year (mm eft) Index to activity Index Efficiency
1972 0 1.00 0 1.00 0
1973 8213.916 1.04 5362.295 1.02 2851.621
1974 -11474.9 1.09 -19557.6 0.97 8082.723
1975 23808.13 1.03 3889.607 1.15 19918.52
1976 19652.4 1.04 5600.364 1.10 14052.04
1977 27860.49 1.08 11433.79 1.11 16426.71
1978 27389.14 1.09 13890.4 1.09 13498.74
1979 36083.72 1.11 19236.49 1.10 16847.23
1980 55737.73 1.16 30087.99 1.14 25649.74
1981 81504.18 1.00 1036.355 1.44 80467.82
1982 88396.04 1.01 2743.793 1.43 85652.24
1983 90009.15 1.00 -208.002 1.43 90217.15
1984 82220.45 1.04 9948.178 1.32 72272.27
1985 81194.71 1.02 4156.875 1.32 77037.83
1986 84537.01 1.03 8233.405 1.30 76303.6
1987 89890.42 1.06 17273.95 1.26 72616.48
1988 102317.7 1.12 36863.71 1.22 65454.03
1989 96523.41 1.13 43193.57 1.17 53329.84
1990 130896.1 1.18 62182.63 1.20 68713.52
1991 136895.7 1.20 71403 1.18 65492.68
1992 133301.8 1.21 79328.93 1.14 53972.92
1993 150814.6 1.24 91252.58 1.15 59562
1994 174298.7 1.23 95780.51 1.19 78518.17
1995 151260.9 1.23 98560.72 1.12 52700.18
1996 161239.7 1.23 105035.9 1.12 56203.77
1997 168991.7 1.21 95931.31 1.15 73060.35
1998 164100.5 1.23 106536.4 1.12 57564.07
1999 173455.2 1.27 131420.5 1.08 42034.72
2000 234326.5 1.16 89841.59 1.27 144485
2001 278199.6 1.14 79586.36 1.38 198613.3
2002 319491.4 1.10 59631.08 1.49 259860.4
2003 343280.9 1.03 19946.36 1.60 323334.6
2004 482849.4 1.14 101100.6 1.63 381748.8
2005 541543.6 1.20 156607.3 1.56 384936.4
2006 567841.6 1.09 76096.43 1.72 491745.1
2007 520711.4 1.11 98938.83 1.57 421772.6
2008 549193 0.97 -31969.9 1.81 581162.8
2009 530910.8 1.08 77380.62 1.59 453530.1
2010 516655.6 1.10 96539.04 1.52 420116.5
2011 456385.8 1.06 55458.55 1.50 400927.3
See text for Construction.
Table A4
Electricity Consumption Decomposition
Activity Due to Efficiency Due to
Year E-E^(Gwh) Index Activity Index Efficiency
1972 0 1.00 0 1.00 0
1973 147.1306 1.00 14.25343 1.03 132.8772
1974 -20.094 1.01 76.58612 0.98 -96.6802
1975 271.3448 0.98 -92.2112 1.09 363.556
1976 210.6708 0.99 -34.184 1.04 244.8548
1977 263.4431 1.00 1.67046 1.04 261.7726
1978 967.3281 1.00 -1.69634 1.15 969.0245
1979 1142.583 1.02 127.4104 1.14 1015.173
1980 1899.576 1.03 272.4253 1.21 1627.15
1981 2314.522 0.97 -309.645 1.33 2624.167
1982 3040.936 0.97 -300.997 1.39 3341.933
1983 3748.127 0.96 -503.867 1.47 4251.994
1984 4830.246 0.96 -482.532 1.57 5312.778
1985 5155.73 0.96 -488.386 1.55 5644.115
1986 6518.233 0.95 -670.08 1.67 7188.312
1987 8319.675 0.96 -697.942 1.79 9017.617
1988 10721.78 0.97 -576.98 1.93 11298.76
1989 11343.9 0.96 -729.227 1.95 12073.13
1990 13043.9 0.96 -703.4 2.03 13747.3
1991 15001.11 0.95 -1043.28 2.15 16044.38
1992 16013.47 0.97 -706.488 2.13 16719.95
1993 18444.27 0.97 -650.369 2.23 19094.64
1994 17850.8 0.97 -840.403 2.16 18691.2
1995 19730.41 0.96 -1070.89 2.23 20801.31
1996 20562.7 0.96 -1156.37 2.21 21719.07
1997 20358.5 0.95 -1475.6 2.20 21834.11
1998 20956.88 0.96 -1298.36 2.18 22255.24
1999 19312.85 0.97 -893.759 2.02 20206.61
2000 20781.92 0.94 -1908.99 2.12 22690.91
2001 23440.61 0.94 -2022.77 2.24 25463.38
2002 24872.55 0.93 -2499.79 2.30 27372.35
2003 25961.16 0.91 -3157.89 2.33 29119.05
2004 28765.17 0.95 -2030.25 2.28 30795.41
2005 30233.22 0.97 -1383.39 2.19 31616.61
2006 34602.73 0.97 -1169.79 2.27 35772.53
2007 37391.74 0.98 -905.152 2.27 38296.89
2008 36803.11 0.94 -2926.09 2.30 39729.19
2009 33439.63 0.93 -3361.8 2.19 36801.43
2010 36181.78 0.92 -3992.01 2.27 40173.79
2011 37629.1 0.92 -4386.27 2.29 42015.37
See text for Construction.
REFERENCES
Abbasi, Z. (2011) Energy Crisis Costs 2 Percent of GDP Annually.
Business Recorder, July 07.
Alam, S. and S. M. Butt, (2001) Assessing Energy Consumption and
Energy Intensity Changes in Pakistan: An Application of Complete
Decomposition Model. The Pakistan Development Review 40:2, 135-147.
Albrecht, J., D. Francois, and K. M. Schoors (2002) A Shapley
Decomposition of Carbon Emissions without Residuals. Energy Policy 30:9,
727-736.
Allcott, H. and M. Greenstone (2012) Is There an Energy Efficiency
Gap. Journal of Economic Perspectives 26, 03-28.
Ang, B. W. and F. L. Liu, (2003) Eight Methods for Decomposing the
Aggregate Energy Intensity of Industry. Applied Energy 76(1-3), 15-23.
Ang, B. W. (2004) Decomposition Analysis for Policymaking in
Energy: Which Is the Preferred Method? Energy Policy 32:9, 1131-39.
Ang, B. W. and F. Q. Zhang (2000) A Survey of Index Decomposition
Analysis in Energy and Environmental Studies. Energy 25, 1149-1176.
Ang, B. W. and K. H. Choi (1997) Decomposition of Aggregate Energy
and Gas Emission Intensities for Industry: A Refined Divisia Index
Method. The Energy Journal 18:3, 59-74.
Ang, B. W., H. C. Huangand, and A. R. Mu (2009) Properties and
Linkages of Some Index Decomposition Analysis Methods. Energy Policy 37,
4624-32.
Boer, D. P. (2009) Generalised Fisher Index or Siegel-Shapley
Decomposition? Energy Economics 31:5, 810-814.
Bossanyi, E. (1983) UK Primary Energy Consumption and the Changing
Structure of Final Demand. Energy Policy 7:6, 253-258.
Boyd, G. A. and J. M. Roop (2004) A Note on the Fisher Ideal Index
Decomposition for Structural Change in Energy Intensity. The Energy
Journal 25:1, 87-101.
Boyd, G. A., D. A. Hanson, and T. N. S. Sterner (1988)
Decomposition of Changes in Energy Intensity: A Comparison of the
Divisia Index and Other Methods. Energy Economics 10:4, 309-12.
Chung, H. S. and H. C. Rhee (2001a) Residual-Free Decomposition of
the Sources of Carbon Dioxide Emissions: A Case of the Korean
Industries. Energy 26:1, 15-30.
Diewert, W. E. (2001) The Consumer Price Index and Index Number
Theory: A Survey Vancouver. Department of Economics, University of
British Columbia. (Department Paper: 0102).
Fengling, L. (2004) Decomposition Analysis Applied to Energy: Some
Methodological Issues. A Thesis Submitted for the Degree of Doctor of
Philosophy, Department of Industrial and Systems Engineering, National
University of Singapore.
Fisher, I. (1921) The Best Form of Index Number. Quarterly
Publications of the American Statistical Association 17:133, 533-537.
Hatzigeorgiou, E., H. Polatidis, and D. Haralambopoulos (2008) CO2
Emissions in Greece for 1990-2002: A Decomposition Analysis and
Comparison of Results using the Arithmetic Mean Divisia Index and
Logarithmic Mean Divisia Index Techniques. Energy 33:3, 492-499.
Hoeksrta, R. and J. Bergh (2003) Comparing Structural Decomposition
Analysis and Index. Energy Economics 25:1, 39-64.
Howarth, R. B. (1991) Energy use in US Manufacturing: The Impacts
of the Energy Shocks on Sectoral Output, Industry Structure and Energy
Intensity. The Journal of Energy and Development 14:2, 175-191.
Huntington, H. G. (2010) Structural Change and U.S. Energy Use:
Recent Patterns. Energy Journal 31, 25-39.
IEA (International Energy Agency) (2012) Energy Indicators System:
Index Construction Methodology. Paris, France: IEA.
Inter-American Development Bank (IDB) (2012) Justification de la
Intervention del Gobierno en el Mercado de Eficiencia Energetic, Serie
Sobre Eficiencia Energetica.
Jamil, F. and E. Ahmad (2011) Income and Price Elasticities of
Electricity Demand: Aggregate and Sector-Wise Analyses. Energy Policy
39:9, 5519-5527.
Jimenez, R. and J. Mercado (2013) Energy Intensity: A Decomposition
and Counter Factual Exercise for Latin American Countries. (IDB-WP:
144).
Kander, A. and M. Lindmark (2004) Energy Consumption, Pollutant
Emissions and Growth in the Long Run: Sweden through 200 Years. European
Review of Economics History 6, 297-335.
Karen, F-V. et al. (2004) What is Driving China's Decline in
Energy Intensity? Resource and Energy Economics 26:1, 77-97.
Khan, M. A. and A. Ahmed (2011) Macroeconomics Effect of Global
Food and Oil Price Shock to Pakistan Economy: A Structural Vector
Autoregressive (SVAR) Analysis. Pakistan Institute of Development
Economics, Islamabad. (PIDE Working Paper Series).
Kiani, A. (2009) Impact of High Oil Prices on Pakistan's
Economic Growth. International Journal of Business and Social Science
2:17.
Li, J. W., R. M. Shresthaand, and W. K. Foell (1990) Structural
Change and Energy Use: The Case of the Manufacturing Sector in Taiwan.
Energy Economics 12:2, 109-115.
Liao, H., Y. Fang, and Y. Wei (2007) What Induced China's
Energy Intensity to Fluctuate: 1997-20061 Energy Policy 35:9, 4640-4649.
Liu, N. and W. B. Ang (2007) Factors Shaping Aggregate Energy
Intensity Trend for Industry: Energy Intensity versus Product Mix.
Energy Economics 29:4, 609-635.
Mairet, N. and F. Decellas (2009) Determinants of Energy Demand in
the French Service Sector: A Decomposition Analysis. Energy Policy 37:7,
2734-2744.
Malik, A. (2007) Crude Oil Prices, Monetary Policy and Output: Case
of Pakistan. Pakistan Institute of Development of Economics, Islamabad.
Malik, A. (2008) How Pakistan is Coping with the Challenges of High
Oil Pricey Pakistan Institute of Development Economics, Islamabad.
Malik, A. (2012) Power Crisis in Pakistan: A Crisis in Governance?
Pakistan Institute of Development Economics. (PIDE Monograph Series).
Metcalf, G. (2008) An Empirical Analysis of Energy Intensity and
its Determinants at the State Level. The Energy Journal 29:3, 1-26.
Pakistan, Government of (2012) Statistical Supplement. Ministry of
Finance.
Pakistan, Government of (Various Issues,) Pakistan Energy Yearbook.
Hydrocarbon Development Institute of Pakistan, Ministry of Petroleum and
Natural Resources.
Park, S. H. (1992) Decomposition of Industrial Energy Consumption:
An Alternative Method. Energy Economics 14L:4, 265-70.
Reddy, B. S. and B. K. Ray (2011) Understanding Industrial Energy
Use: Physical Energy Intensity Changes in Indian Manufacturing Sector.
Energy Policy 39:11, 7234-43.
Sahu, S. and K. Narayanan (2010) Decomposition of Industrial Energy
Consumption in Indian Manufacturing: The Energy Intensity Approach.
(MPRA Paper No. 21719).
Shahiduzzaman, M. D. and, A. Khorshed (2012) Changes in Energy
Efficiency in Australia: A Decomposition of Aggregate Energy Intensity
Using Logarithmic Mean Divisia Approach. (MPRA Paper: 36250).
Siddique, Rehana, Hafiz Hanzala Jalil, M. Nasir, Wasim Shahid
Malik, and Mahood Khalid (2011) Cities of Pakistan. Pakistan Institute
of Development Economics, Islamabad. (PIDE Working Papers, 2011:75).
Song, F. and X. Zheng (2012) What Drives the Change in China's
Energy Intensity: Combining Decomposition Analysis and Econometric
Analysis at the Provincial Level. Energy Policy 51, 445^453
State Bank of Pakistan (2010) Hand Book of Statistics. Karachi:
State Bank of Pakistan.
Sun, J. W. (1998b) Accounting for Energy Use in China, 1984-94.
Energy 23:10, 835-949.
Sun, J. W. (1998a) Changes in Energy Consumption and Energy
Intensity: A Complete Decomposition Model. Energy Economics 20:1,
85-100.
Syed, N. I. (2010) Measuring the Impact of Changing Oil Prices and
Other Macroeconomic Variables on GDP in Context of Pakistan Economy.
International Research Journal of Finance and Economics. ISSN:
1450-2887.
Szep, S. T. (2013) Eight Methods for Decomposing the Aggregate
Energy Intensity of the Economic Structure. Journal of Theory
Methodology Practice 9:1, 77-84.
Zhao, X., C. Ma, and D. Hong (2010) Why did China's Energy
Intensity Increase During 1998-2006: Decomposition and Policy Analysis.
Energy Policy 38:3, 1379-1388.
Comments
Energy intensity in Pakistan is more than double to that of the
world average and more than five times to that of Japan and the UK. For
resource constrained economies like Pakistan it is more cost effective
to increase its energy security and ease supply constraints through
efficiency in its use and conservation compared to exploiting/ building
new sources of energy.
Energy efficiency is regarded as an important component in national
energy strategy but so far it is the most neglected area in
Pakistan's energy strategies and plans. In this context, this study
is a significant attempt in terms of examining the change in energy
intensity over the years and decomposing this change in terms of
activity and efficiency. I have few suggestions:
Besides correcting some typo's and editorial mistakes which
made it difficult for the reader to understand, some in-depth analysis
on changes in energy intensity for the main economic sectors would make
this study even more interesting and useful. Because when analysing
energy intensity for Pakistan it is important as well as useful from
policy perspective to identify those economic activities that are
crucial to reduce energy consumption. Similarly, there is a need to
highlight potential strategies and measures for improving the efficiency
of final energy use with reference to particular economic activities.
Since you have already collected energy consumption details at the
sector level which might help you in some sort of discussion on
efficiency of energy end uses for sub-activities. Moreover, while
discussing and analysing energy intensities some discussion on
electrification over the years would make the discussion more valuable.
Likewise, when you are discussing positive changes in oil intensity it
may be because of negative changes in other sources of energy.
Similarly, in the analysis of your results you can make some
comparison with the studies for other countries. For instance, China has
decreased its energy intensity significantly through improvement of
energy efficiency; whereas structural-mix changes played a low, but
positive role in decreasing the energy intensity. But in Pakistan as
your results show it's the opposite. Similarly in India, energy
efficiency also played a positive role, but, the industrial structure
has become more energy-intensive because of the increasing share of
energy-intensive sub-sectors, which offsets the impact of energy
efficiency on energy intensity. So you need to discuss on how other
countries have enhanced efficiency in the use of energy.
Even, you can compare your results with the previous study on
Pakistan by Alam and Butt (2001) (you mentioned in your paper) to
highlight the significance of your study and your results.
Afia Malik
Pakistan Institute of Development Economics, Islamabad.
(1) During the month of May in 2011, the shortfall had surpassed
7000 MW. See for the details Malik (2012).
(2) A method is regarded as perfect if it leaves no residual term.
(3) Boyd and Roop (2004) and Metcalf (2008) used Fisher ideal index
for U.S, Hatzigeorgiou, et al. (2008) used arithmetic mean Divisia
method for Greece, Mairet and Decellas (2009) used log mean Divisia
method for France, Sahu and Narayanan (2010) used the Laspeyres index
approach and the Divisia index approach for Indian manufacturing
industries for 1990-2008, Zhao, et al. (2010) used log mean Divisia
method while Song and Zheng (2012) used Fisher ideal index for China,
Szep (2013) examined energy intensity for Czech Republic, Slovakia,
Slovenia, Poland and Hungary between 1990 and 2009 using eight different
decomposition methods (results were almost same for all of the methods).
(4) We are broadly following Boyd and Roop (2004).
(5) For instance, see the appendix for details.
(6) We are following Metcalf (2008) in this calculation.
(7) Note that it does not include the consumption under the other
government heading.
(8) For instance see the oil and gas efficiency indices after 2000.
