1 ELECTRICITY CONSUMPTION AND ECONOMIC GROWTH: A Causality Evidence from Six Economic Corridors of Indonesia, 1984-2010 1 Timotius D. Harsono Lecturer in International Relation, National Resilience Institute RI. Ph.D Candidate, Policy Study, Gadjah Mada University, Yogyakarta, Indonesia Email: [email protected][email protected] (Phone: +62816856773) Mudrajad Kuncoro Professor in Economics, Faculty of Economics & Business, Gadjah Mada University, Yogyakarta, Indonesia Email: [email protected]URL: http://www.mudrajad.com (Phone: +62811254255) ABSTRACT The relationship between electricity consumption and GDP in Indonesia has been examined in some previous studies. However, none of them has tested the causal relationship between electricity consumption and Gross Regional Development Product (GRDP) of regions within a country. Yudhoyono’s government launched Master Plan for Acceleration and Expansion of Indonesia Economic Development (MP3EI) by setting six major economic corridors on 27 of May 2011. However, the regional implementation of electricity development policy has not been elaborated in MP3EI document. Our study will explore further the relationship between electricity consumption and economic growth in six economic corridors as set in the MP3EI. This relationship is important to be identified to understand the policy implications on electricity in the context of regional autonomy and development. Our analysis is a pioneering study of Indonesia’s electricity consumption and economic growth that incorporates “geography” into economic analysis of causality during 1984-2010. Using Granger causality tests and Error Correction Model (ECM), our empirical findings shows that: (1) there is uni-directional short run relationship moving from electricity consumption per capita to GDP per capita for Indonesia as a single region; (2) the causality tests for Sumatra economic corridors shows there is an uni-directional short run relationship moving from electricity consumption to economic growth, meanwhile the tests for Kalimantan shows an opposite short run uni-directional relationship; (3) the causality tests do not show any causal relationship for other corridors (Jawa, Sulawesi, Bali-Nusa Tenggara, and Papua-Kepulauan Maluku). Finally, the findings suggest that fundamental changes in electricity and regional development policies are required. Indonesia needs electricity and regional policy that incorporates regional variations in term of electricity consumption, economic growth, and GRDP per capita. Keywords: Electricity Consumption, GDP, GRDP, Corridors, Granger Causality, ECM 1 This paper is presented at the Kuala Lumpur International Business, Economics, and Law Conference KLIBEL 2013, Kuala Lumpur Malaysia, 8-9 April 2013. I am grateful for the comments from the anonymous reviewers of International Journal of Business, Economics, and Law, volume 2, 2013. Our appreciation also goes to our assistants, Fatkhu Ridho FEP and M. Irka Irfa’Darojat.
Transcript
1
ELECTRICITY CONSUMPTION
AND ECONOMIC GROWTH:
A Causality Evidence from
Six Economic Corridors of Indonesia, 1984-20101
Timotius D. Harsono
Lecturer in International Relation, National Resilience Institute RI.
Yoo, 2005). The empirical studies found some controversial results with respect to the
causal relationship between electricity consumption and economic growth that can be
summarized into four types, as follows (Chen, et al. 2007): First, the uni-directional
causality moving from electricity consumption to economic growth implies that
restrictions on the use of electricity may affect economic growth adversely while
increases in electricity supply may contribute to economic growth. Second, the uni-
directional causality moving from economic growth to electricity consumption would
suggest that a permanent increase in economic growth may result in a permanent
increase in electricity consumption. Third, a bi-directional causal relationship implies
that electricity consumption and economic growth are jointly determined and affected
at the same time. Fourth, the absence of causal relationship implies that: (1) the
electricity consumption is not correlated with economic growth; (2) neither
conservative nor expansive policies related to electricity consumption have any effect
on economic growth.
Indonesia is an excellent laboratory for testing which directions of causality
between electricity consumption and economic growth occur. One of most striking
issues is that generating electric capacity growth in Indonesia has lagged behind the
pace of electricity demand growth, leading to power shortages and a low
electrification ratio. Although Indonesia's generating capacity has increased by more
than a quarter in the last decade, the country has a low electrification ratio compared
to similar income countries. In 2013, around 76% of Indonesia's population had
access to electricity. Indonesian Eastern Regions (Kawasan Timur Indonesia, KTI)
lags behind the Indonesian Western Region (Kawasan Barat Indonesia, KBI), with
some provinces only providing electricity to a third of the population (Figure 1).
Figure 1. Electricity Ratio By Province, Indonesia 2005-2014
Source: Ministry of Energy and Mineral Resource (MEMR) (2011)
US-EIA (2013) indicates two important features of electricity supply. First,
state-owned electric utility company, PT PLN (Perusahaan Listrik Negara), is the
most significant company in the electric power sector. PLN owns and operates about
90% of the country's generating capacity through its subsidiaries, and maintains an
3
effective monopoly over distribution activities. Although the most recent 2009
Electricity Law ends PLN's distribution monopoly, there is a lack of sufficient
regulations to enforce this law. Second, Indonesia had an estimated 44 gigawatts
(GW) of installed capacity in 2011 and generated 192 billion kilowatt-hours (kWh).
In addition, 86% of the power came from conventional thermal sources, with the rest
coming from hydroelectric (9 percent), geothermal (5 percent), and other renewable
sources. Coal accounted for just over half of conventional thermal power.
The outline of this paper is as follows. First, it will highlight some surveys of
literature on causality tests between electricity consumption and GDP. The
relationship between electricity consumption and GDP in Indonesia has been
examined by some previous studies (Murry & Nan, 1996; Yoo, 2006; Chen, et al.,
2006). Second, it will discuss briefly the data and methology used in our study. Third,
our analysis of the relationship between electricity consumption and economic growth
in Indonesia will cover: (a) the causal relationship between national electricity
consumption and Gross Development Product (GDP) and (b) the causal relationship
between Regional Electricity Consumption (REC) and Gross Regional Development
Product (GRDP) of each economic corridor. Our focus of the analysis will be
emphasised on the 6 economic corridors, i.e. Sumatra, Jawa, Kalimantan, Sulawesi,
Bali–Nusa Tenggara, and Papua–Maluku, as set by the recent Master Plan of
Acceleration and Expansion of Indonesia Development (Master Plan Percepatan
Pembangunan Indonesia, MP3EI). Concluding remarks and policy implications are
given in the final section.
2. LITERATURE REVIEW
The electricity Law No. 30 year 2009, in particular article 2.2, indicates that the
essentially purpose of this law is to secure the availability of sufficient and good
quality electric power with effordable price to increase the welfare of all Indonesian
people (MEMR, 2011b). The PLN is a state owned company and mandated by the law
to provide electricity service as a single operator. Esentially, the main objectives of
the electricity law is to support the improvement of people’s welfare and promote
economic growth. In order to achieve these objectives, the goverment has prioritized
to fulfill electricity demand of household and industrial sector (MEMR, 2010). In the
last few years, the electrification ratio has increased significantly from 62% (2005) to
67.2% (2010). In 2014, it is projected to reach 80%. The supply of electricity for
industry and business sectors is closely related to economic growth. On the other
hand, the infrastucture development required to increase the electricity supply needs
huge amount of investment. So far, the electricity policies and regulations has been
focused more on the supply side, rather than serving the demand side.
The policy implementation are incorporated in some official documents, such as
Energy Outlook 2008-2010, Annual State Budget and Expenditure Report (Nota
Keuangan & RAPBN), and MP3EI. The general policy in energy sector (including
electricity) is mostly supply and price oriented. The electricity price is driven by
subsidy level, which is mostly aimed to control the inflation level.
The electricity consumption per annum is increasing rapidly in line with the
increasing national economic growth and the changes of paradigm in the society to
meet their day to day life style needs, especially in household and industrial sector
(Table 1). The electricity demand during 2006-2010 had been increasing with average
growth of 6.3% annually. While the electricity infrastructure development can only
provide additional supply with an average growth of 4.4% in the same period.
4
Table 1. Electricity Consumption and Economic Growth, Indonesia 2000-2010
Source: Compiled from BPS (2013); MEMR (2010b)
The disequilibrium between actual demand and the supply of electric power has
created shortage of electricity supply in certain regions, especially those regions
outside PLN interconnective region of Jawa-Madura-Bali. The 1998 economic crisis
has slowed down the development of electricity infrastructure in Indonesia. During
the period of economic recovery, the growth of the installed electric power capacity
was only increased by the rate of 1.13% (MEMR, 2010b).
Electricity consumption of Indonesia increased from 112,609 giga watt hour
(Gwh) in 2006 to 147,297 GWh in 2010, with average increase of 6.6%. Such
increase of electricity comsumption mainly is driven by the increase of consumption
households and industrial sectors. Table 1 shows the trend of electricity consumption
increase was not always followed by economic growth within the period of 2000–
2010. This condition seems to be in line with the finding of Yoo (2006) which argued
that the consumption of electricity in Indonesia did not significantly impact the
economic growth. This condition may be caused by the fact that the consumption of
electricity in Indonesia are used more to meet the need of household sector, while less
portion of electric power are allocated to fullfill the demand of industrial sector and
other economic activities that can stimulate the economic growth strongly (Table 2).
Table 2. Electricity Demand By Group of Users (Gwh), 2006-2010
Group of Users Electric Power Consumption (Gwh)
2006 2007 2008 2009 2010
Household 43.754 47.324 50.184 54.945 59.825
Commercial 18.415 20.608 22.926 24.638 27.157
Industry 43.615 45.802 47.969 46.017 50.985
Others 6.825 7.510 7.940 8.607 9.330
Total 112.609 121.246 129.019 134.207 147.297
Source: MEMR (2010b)
Causality tests between electricity consumption and GDP in Indonesia have been
examined earlier by Murry & Nan (1996) using 1970-1990 period and Yoo (2006)
using 1971-2002 period (Table 3). Yoo (2006) used the standard Granger causality
test and found uni-directional relationship moving from GDP to electricity
consumption. This finding shows a conformity to the finding of Murry & Nan (1996).
In contrast to the finding of Yoo (2006) and Murry & Nan (1996), the causality
test performed by Chen, et al. (2006) using Granger causality test and Error
Correction Model (ECM) found a long run relationship moving from electricity
consumption to economic growth (Table 3). These studies used single country data set
for each country and also applied data panel for 10 selected countries in Asia.
QuantityAnnual
Growth (%)in Million Rp
Annual
Growth (%)Gwh
Annual
Growth (%)
2000 205,132.00 1.12 6.70 6.25 4.92 389 3.20
2001 208,900.60 1.84 7.20 7.46 3.64 405 4.11
2002 212,003.50 1.49 7.60 5.56 4.50 411 1.48
2003 215,152.38 1.49 8.30 9.21 4.78 420 2.19
2004 217,854.10 1.26 10.64 28.19 5.03 459 9.29
2005 220,553.07 1.24 12.45 17.01 5.69 485 5.66
2006 223,013.78 1.12 15.03 20.72 5.50 505 4.12
2007 225,642.50 1.18 17.60 17.10 6.35 579 14.65
2008 228,523.30 1.28 21.70 23.30 6.01 565 -2.42
2009 231,369.50 1.25 24.30 11.98 4.55 560 -0.88
2010 237,641.00 2.71 27.22 12.02 6.50 619 10.54
14.44 5.22 4.72Average
Year
Population GDP per CapitaEconomic
Growth (%)
Electricity
Consumption per
Capita
KWh
5
Table 3. Previous Causality Tests Related to
Electricity Consumption and GDP in Indonesia Authors (Year) Methodology Time Period Causal Relationship
Murry& Nan (1996) Standard Granger causality test 1970–1990 Income Electricity
Yoo (2006) Standard Granger causality test 1971–2002 Income Electricity
Chen (2006) Granger Causality and Error Correction Model
1971-2001 Electricity Income
Our study uses modified Granger causality test with ECM applied to time
series data within the period of 1984-2010. These model are employed to find the
nature of the causal relationship between electricity consumption and economic
growth, both for Indonesia as a single region and each region of the 6 economic
corridors as set by MP3EI document. Our study sought to inject empirical content to
the emerging interest in the causality literature by examining electricity consumption
and regional economic growth in the context of Indonesia’s recent regional autonomy.
By analyzing 6 regional economic corridors, we attempt to incorporate “geography”
into economic analysis of causality with newer time horizon (1984-2010).
3. METHODOLOGY
3.1. Data
This study will use annual time series data for Indonesia as one single country data
set, as well as for 6 economic corridors within Indonesia, namely Sumatra, Jawa,
Kalimantan, Sulawesi, Bali-Nusa Tenggara, and Papua-Maluku. The annual data for
real GDP is obtained from Badan Pusat Statistik (BPS) and the annual data for electric
power consumption is obtained from PLN. The data period (1984-2010) is determined
based on the availability of regional electricity consumption and regional income data.
The real GDP and GRDP are expressed in rupiah at constant price of year 2000 while
the electric power consumption is expressed in unit of Kilowatt hours (KWh). The
real GDP and real GRDP as well as National Electricity Consumption (NEC) and
Regional Electricity Consumption (REC) are transformed into natural logarithms to
reduce heteroscedasticity. The logarithm variables have its economic meaning since
they are approximated to be viewed as the growth of the respective differenced
variables. The unit roots, cointegration tests, Granger causality tests, and ECM model
estimation are performed by Eviews 6.0.
Instead of using only aggregate data of Indonesia as one unit analysis, this
study will also use a quantitative data of 33 provinces and 14 PLN distributive regions
which are regrouped into 6 economic corridors in line with the MP3EI (Table 4).
Sumatra economic corridor connects its main economic centers from Banda Aceh (in
Aceh province), Medan, Pakanbaru, Jambi, Palembang, and Bandar Lampung (in
Lampung province). Java’s economic corridors connect its main economic centers
from Banten, Jakarta, Bandung, Semarang, Yogyakarta, to Surabaya. Major economic
centers of Kalimantan are Pontianak, Palangkaraya, Banjarmasin, and Samarinda will
be connected with the Corridor Connectivity Lane. Sulawesi corridor is expected to be
Production and Processing Center of Agricultural, Agriculture, Fisheries, Oil and Gas
and Mining National, and become the forefront in serving markets of East Asia,
Australia, Oceania, and America via its major centers (Makasar, Mamuju, Kendali,
Palu, Gorontalo, Manado). Major economic centers of Bali-Nusa Tenggara corridor
are Denpasar, Lombok, Kupang. Papua-Maluku, have been designed to be the Centers for Development of food, fisheries, energy, and national mining that connects its main
economic centers from Ambon,Sofifie,Sorong,Manokwari,Timika,Jayapura, Merauke.
6
Table 4. Six Economic Corridors based on Electricity Consumption Data of 14 PLN Distributive
Regions, GRDP Data of 33 Provinces, and MP3EI’s Theme
14 Distributive Regions of PLN 6 Economic
Corridors
MP3EI’s Theme
1. NAD: Nangroe Aceh Darusalam (NAD) Province
SUMATRA
Center for Production
and Processing of
Natural Resources As
The Nation’s Energy
Reserves
2. North Sumatra: North Sumatra Province
3. West Sumatra Riau: West Sumatra, Riau and
Riau Islands Provinces
4. S2JB, Babel & Lampung: Jambi, South
Sumatra, Bangka Belitung Islands, Bengkulu and
Lampung Provinces
5. West Java and DKI: Special District of Jakarta,
West Java and Banten Provinces
JAVA
Driver for National
Industry and Services
Provision
6. Central Java and DIY: Central Java and
Yogyakarta Special Region Provinces
7. East Java: East Java Province
8. Bali, West Nusa Tenggara and East Nusa
Tenggara: Bali, West Nusa Tenggara and East
Nusa Tenggara Province
BALI – NUSA
TENGGARA
Gateway of Tourism
and Food National
Support
9. West Kalimantan: West Kalimantan Province
KALIMANTAN
Center for Production
and Processing of
National Mining and
Energy Reserves
10. Central-South Kalimantan & East
Kalimantan:
South Kalimantan, Central Kalimantan, and East
Kalimantan Provinces
11. Suluttenggo: North Sulawesi, Gorontalo and
Central Sulawesi Provinces
SULAWESI
Production and
Processing Center of
Agricultural,
Agriculture, Fisheries,
Oil and Gas and
Mining National
12. Sulselra: South Sulawesi, West Sulawesi and
South-East Sulawesi Province
13. Maluku: Maluku and North Maluku Province
PAPUA –
MALUKU
Center for
Development of Food,
Fisheries, Energy, and
National Mining 14. Papua: Papua and West Papua Provinces
This study will use Granger causality test and ECM including stationary and co-
integration test. Prior to applying Granger causality test, it is imparative to ensure that
time series data related to electricity consumption and economic growth are
stationary. The stationarity test is done by applying unit roots tests to find out whether
the two variables (NEC/REC and GDP/GRDP) are stationary at 0 level, first
difference, or second difference. If both variables are not stationary at all level, then
the result of the time series data regression will be spurious, which is indicated by
time series data regression with a high R2 value but the regression coeficient is not
significant (Gujarati, 2003).
In order to make sure the data of two variables are stationary, Augmented Dickey
Fuller (1981) model is used to perform unit root tests. This model is as follows:
∑ (1)
∑ (2)
whereas, y is the variable to be tested and t is the time trend. The first equation uses
no time effect while the second one uses a fixed time effect. The lag length, i, are
7
chosen using Akaike’s Information Criterion (AIC). When the null hypothesis H0 : r =
0 is accepted, this implies that the variable has unit root and is not stationary. The first
difference is thus taken to test for the stationarity of the data set before the Granger-
causality test is implemented (Chen, et al., 2006).
ADF model is applied to test the stationary using time fixed effects and no time
effects methods. The time fixed effects methods with trend and intercept is used to
find out at what level the data is stationary. Trend and intercept are to show the result
of unit root test using Random Walk Model (RWM) which indicates the deterministic
trend and the time shift. No time effects method is to show the result of unit root test
without using trend and intercept, hence it does not indicate the time shift. The result
of ADF test will be compared to McKinnon critical value at certain level (α) to find
out if the data is stationary.
It is important to select the optimal lag to be applied in this model because if the
lag used in this stationarity test is too small, then the residual of the regresion will not
show white noise process. At this point, the ADF model cannot estimate the error
accurately. In the contrary, if the selection of lag is too big, it may reduce the
capability to reject the null hypotesis (H0), because if the parameter value is too big
then it will reduce the degree of freedom (Harris, 1995: 65). The selection of optimal
lag can lead to the most appropriate finding with uncorrelated residual.
ADF model applied in this study is to emphazise more on the consistency rather
than efficiency. Therefore, the automatic selection method is performed using EViews
6.0 to determine the optimal lag by using the value of AIC as referrence. The AIC
value is used because it is more consistent compared to the value obtained using
Schwarz Information Criteria (SIC) and Hannan-Quinn Information Criteria (HQIC)
(Shatland, 2008).
Johansen cointegration test is applied to find out whether NEC/REC and
GDP/GRDP has a long run relationship. The optimal lag for each corridor is selected
based on each of its AIC value. If the NEC/REC and GDP/GRDP variables are
stationary at the first difference, then both variables are most probably cointegrated,
hence, Error Correction Terms is applied to perform modified ECM causality test to
find out whether there is any long run causal relationship.
The relationship between electricity consumption and GDP, as well as GDRP,
will be examined by applying Granger causality test and ECM as has been performed
by Chen (2006). These methods are used to show the relationship between electricity
consumption (NEC/REC) and economic growth (GDP/GRDP) in Indonesia, as well
as in each of those 6 economic corridors. The characteristic of the relationship
includes: (1) Is there any uni-directional relationship moving from NEC/REC to
GDP/GRDP?; (2) Is there any uni-directional relationship moving from GDP/GRDP
to NEC/REC?; (3) Is there any bi-directional relationship between GDP/GRDP and
NEC/REC?; (4) Is there no relationship at all between GDP/GRDP and NEC/REC?
(5) Is the nature of the relationship (if any) of both variables show a long run or short
run relationship. It is imparative to understand the characteristic of the relationship
between electricity consumption and the economic growth since the nature of such
relationship has different policy implication toward both sectoral related to electricity
and regional policy. The result of this causality tests implies that the past events have
impacted on the present events, not the future ones.
In the case NEC and GDP variables are not stationary, but become stationary
after unit roots test at the first diferrence, and after further Johansen cointegration test
both variables are not cointegrated, then standard Granger causality test is applied, as
shown in the following equations:
8
∑ ∑
(3)
∑ ∑
(4)
If the Johansen cointegration shows that both variables are cointegrated, then
modified ECM Granger causality test is applied by using Error Correction Term, as
follows:
∑ ∑
(5)
∑ ∑
(6)
whereas, NEC is national electricity consumption and GDP is Gross Domestic
Product in the form of natural logarithm, Δ is diferrence operator, q is the lag, α is the
parameter to be estimated, µ is serially uncorrected error term, and is error correction term which is obtained from cointegration test. In case the unit roots and
the cointegration test show hat the data is stationary and cointegrated, then modified
ECM causality test is used.
Equations 3 and 4 shows that ∆NEC has short run effects on ∆GDP if the
estimated coefisien (α12s) is stastically significant as shown by the value of statistik F.
Equations 5 and 6 show the application of ECM by including the dependent variables
with each of its lag and the disequilibrium of the previous error correction term, . The long run causal relationship can be determined by the level of significance of the
coefisien ( ) for error correction term based on t statistic test. By going through the same procedure as defined in equation 5, we can test the long run causal relationship
from ∆GDP to ∆NEC (equation 6).
The causality tests between REC and GRDP variables for 6 economic corridors
also use the same model and procedures as the causality test for Indonesia as a single
region. The Method is based on the Granger causality to find the direction of causal
relationship, whereas the ECM is applied by using Error Correction Term to find out
whether there is any long run causal relationship. The equations can be stated as
follows:
∑ ∑
(7)
∑ ∑
(8)
Modified Granger causality and ECM are applied to examine whether there is
any long run causal relationship by obtaining the residual value and use it as Error
Correction Term, as stated in the following equations:
∑ ∑
(9)
∑ ∑
(10)
whereas, REC is regional electricity consumption and GRDP is gross regional
development product in the form of natural logarithm.
9
4. FINDINGS AND ANALYSIS
4.1. Unit Roots and Cointegration Tests
4.1.1. Test Result for Indonesia (as a Country Data Set)
The result of ADF test using No Time Effects model shows that the NEC variable is
not stationary at level as well as at first difference with significance of 5%. However,
NEC variable becomes stationary at the first difference with significance of 5% if
time fixed effects is applied. On the other hand, the ADF test using both No Time
Effects and Time Effects model shows that the GDP variable is not stationary at level
with significance of 5%, however, GDP variable becomes stationary at the first
difference with significance of 5% (Table 5).
Table 5. Result for Unit Roots Test between NEC and GDP Variable in
Indonesia, 1984-2010
Region Variable
ADF
No Time Effects Time Fixed Effects
Lag Level Lag ∆ Lag Level Lag ∆
Indonesia NEC 1
1.54
( 0.9664) 0
-1.59
( 0.1025) 0
-0.98
( 0.9291) 0
-4.34*
( 0.0106)
GDP 0 1.93
( 0.9846) 0
-4.17*
( 0.0002) 1
-3.35
( 0.0800) 3
-3.64*
( 0.0491)
Notes: ∆ denotes first differences. All variables are in natural logarithms; NEC and GDP in per capita;
ADF = Augmented Dickey Fuller; * Means that the null of the unit root in the ADF tests is
rejected; The lag lengths are selected using AIC; Number in parentheses are P-value
The result of stationary test using ADF model with time fixed effect as well as no
time effects show that the data of NEC and GDP is stationary at the first difference,
and therefore we can proceed with Johansen cointegration test. Table 6 shows the
result of cointegration tests between NEC and GDP variables. This test shows that the
trace statistic value and maximum Eigenvalue is smaller than the significance of 5%.
This result shows no cointegration between NEC and GDP. In this case, we cannot
proceed by applying ECM to find whether there is any long run relationship between
the two variables. At this point, standard Granger causality test is applied to examine
whether there is any short run relationship.
Table 6. Results for Cointegration Test between NEC and GDP Variable
in Indonesia, 1984-2011
Region Null Hypothesis Lag
Johansen Test
Statistic Cointegraton
Trace Max-eigen
Indonesia R=0
1 15.99982 15.1362
No R≤1 0.863584 0.863584
Notes: NEC and GDP in per capita; r = the number of the vector cointegration; * means that the null
hyptothesis of no cointegration relationship is rejected at the 5% level; The optimal lag
lengths are selected using Akaike’s information criterion
10
4.1.2. Test Result for Six Economic Corridors
The result of stationary test on both variables (REC and GRDP) using ADF method is
shown in Table 7. ADF tests indicate that REC for Sumatra, Jawa, and Kalimantan
corridors are not stationary at 0 level as well as at the first difference with the
significance of 5%, but become stationary at the first difference with the significance
of 10%. The GRDP variable for the three corridors are stationary at the first difference
with the significance of 5%. The REC and GDRP in Sulawesi, Bali-Nusa Tenggara,
and Papua-Kepulauan Maluku corridors are stationary at the first difference with the
significance of 5%.
Table 7. Results for Unit Roots Test between REC and GRDP Variable in 6
Indonesia’s Economic Corridor, 1984-2010
Corridors Variable
ADF
No Time Effects Time Fixed Effects
Lag Level Lag ∆ Lag Level Lag ∆
Sumatra REC 2
1.88
(0.9827) 1 -1.76**
(0.0741) 2 -1.79
(0.6737) 3 -2.71
(0.2409)
GRDP 0 1.10
(0.9252) 0 -4.49*
(0.0001) 1 -3.07
( 0.1327) 1 -4.05*
(0.0204)
Jawa REC 4
0.23 ( 0.7442) 4
-1.8** ( 0.0644) 2
-1.03
(0.9199) 1 -4.34*
( 0.0015)
GRDP 4 3.25
(0.9992) 0 -4.13
(0.0002) 0 -2.78
(0.2156) 3 -4.04*
(0.0222)
Kalimantan REC 1
1.34
(0.9510) 1 -1.22
( 0.1948) 0 -1.53
( 0.7889) 0 -3.44**
(0.0676)
GRDP 2 0.20
(0.7380) 1 -5.11*
(0.0000) 1 -3.34
(0.0820) 1 -4.98*
( 0.0028)
Sulawesi REC 0
9.25
(1.0000) 1 -1.21
(0.1995) 0 -2.39
(0.3718) 0 -4.32*
(0.0111)
GRDP 0 2.40
(0.9946) 0 -4.36*
(0.0001) 0 -1.74
(0.7000) 2 -3.23
(0.1027)
Bali – Nusa
Tenggara
REC 1 2.36
(0.9940) 5 -1.23
(0.1921) 0 -0.74
(0.9580) 0 -4.98*
(0.0026)
GRDP 0 1.64
(0.9725) 0 -5.01*
(0.0000) 0 -2.68
(0.2513) 3 -3.12
( 0.1242)
Papua –
Kepulauan
Maluku
REC 0 4.39
(1.0000) 0 -2.53*
(0.0137) 0 -2.13
(0.5041) 0 -4.82*
(0.0039)
GRDP 0 0.13
(0.7165) 0 -5.28*
(0.0000) 0 -3.00
( 0.1498) 0 -5.10*
(0.0020)
Notes: ∆ denotes first differences. All variables are in natural logarithms; REC and GRDP in per
capita; ADF = Augmented Dickey Fuller; * Means that the null of the unit root in the ADF tests
is reejected; ** Means that the null of the unit root in the ADF tests is rejected; The lag
lengths are selected using AIC; The number in parentheses are P-value.
Table 8 shows the Johansen cointegration test for variable REC and GRDP for
each corridor during the period of 1984-2010. The optimal lag selected for each
corridor is determined by using the AIC value as reference. The cointegration test
result shows that the trace statistic value and maximum eigenvalue is greater than the
critical value at the significance of 5% for Sumatra, and Kalimantan, Bali–Nusa
Tenggara, and Papua-Kepulauan Maluku. Therefore, the hypotesis which states there
is cointegration between the two variables in Sumatra corridor are proven. As for
Jawa and Sulawesi corridors, the trace statistic value and maximum eigenvalue are
smaller than the critical value at the significance of 5%. The result of these tests
indicates that for Sumatra, Kalimantan, and Bali-Nusa Tenggara, hence, Error
Correction Model may be applied to examine whether there is any long run causality
relationship in those three corridors.
11
Table 8. Results for Cointegration Test between REC and GRDP Variable in
6 Indonesia’s Economic Corridors, 1984-2010
Corridors Null
Hypothesis
Optimal
Lag
Johansen Test Statistic Cointegration
Max-eigen Trace
Sumatra R=0
1 21.06203* 23.64630*
Yes R≤1 2.584269 2.584269
Jawa R=0
1 14.50134 14.50134
No R≤1 0.020588 0.020588
Kalimantan R=0
5 21.71951* 30.03826*
Yes R≤1 8.318755 8.318755
Sulawesi R=0
1 10.43199 10.78349
No R≤1 0.351501 0.351501
Bali-
Nusa Tenggara
R=0 1
20.77115* 21.14443* Yes
R≤1 0.373289 0.373289
Papua-
Kepulauan Maluku
R=0 1
9.299099 12.79955 No
R≤1 3.500455 3.500455
Notes: REC and GRDP in per capita; r = the number of the vector cointegration; * means that the
null hyptothesis of no cointegration relationship is rejected at the 5% level; The optimal
lag lengths are selected using Akaike’s information criterion.
4.2. Causality Tests
4.2.1. The result of Causality Tests for Indonesia
The results of Granger causality test for Indonesia is obtained by applying data within
the period of 1984-2010 (Table 9). The result of F statistic shows the evidence of
short run uni-directional relationship moving from economic growth per capita
(∆GDP per capita) to electric consumption growth per capita (∆NEC per capita). The
short run causal relationship between ∆NEC per capita and ∆GDP per capita can be
seen at Table 9 column (3). The statistic F shows that the probability of ∆NEC per
capita does not have any influence to ∆GDP per capita is only 0.0884 or smaller than
the significance of 10%, therefore the null hypotesis which states that ∆NEC do not
have any influence on ∆GDP is rejected. In the contrary, the probability of ∆GDP per
capita does not influence ∆NEC per capita is 0.2319 or it is greater than significance
of 10%. This finding shows that the growth of electricity consumption in Indonesia
has influenced the growth of GDP. In other words, the findings shows that the
electricity consumption in Indonesia tend to be demand driven.
Table 9 column (4) shows that the result of t statistic test does not prove any
long run relationship both from electricity consumption growth per capita (NEC per
capita) to economic growth per capita (GDP per capita) or from economic growth to electricity consumption growth. This is reflected by the insignificant t-value between
those two variables.
The result of F statistic test does not show any joint short run/long run
relationship moving from electricity consumption growth per capita to economic
growth per capita. This F statistic test also does not show joint short/long run
relationship from the opposite direction. The absence of joint short run/long run
relationship, is shown by Table 9 column (5), whereas the result of F statistic test
indicates that the probability of NEC per capita does not influence the GDP per
capita is 0.9509 or it is greater that the significant level of 5%. This test also shows
that the probability of GDP per capita does not influence the NEC per capita is 0.7540 or it is greater that the significant level of 5%.
12
Table 9. Results for Causality Test between NEC and GDP Variable
in Indonesia, 1984-2010
Region
(1)
Null Hypothesis
(2)
Sources of Causation
Short run
(3)
Long run
(4)
Joint (short
run/long run)
(5)
F-statistics t-statistics F-statistics
∆NEC ∆GDP εt-1 ∆NEC,
εt-1
∆GDP,
εt-1
Indonesia ∆NEC does not cause ∆GDP
3.1785** (0.0884)
1.369681 (0.2988)
0.00388 (0.9509)
∆GDP does not cause ∆NEC 1.51130
(0.2319)
1.065393
(0.1853)
0.10093
(0.7540)
Notes: NEC and GDP in per capita; The lag length are selected using Akaike’s information criterion;
* Means that the null hypothesis of no causation is is rejected at the 5% level; ** Means that
the null hypothesis of no causation is is rejected at the 10% level; The number in parentheses
are P-value.
The finding of causality test in this study does support the finding of Chen, et al.
(2006), whereas the causal relationship is uni-directional moving from electricity
consumption to GDP. However, the finding of this study shows short run relationship,
whilst the relationship found by Chen (2006) is long run. On the other hand, the
finding in this study challenges the finding of Yoo (2006) and Murry & Nan (1996) in
which both findings show uni-directional relationship moving from economic growth
to electricity consumption growth.
The finding in this study shows the presence of short term relationship moving
from electricity consumption per capita to economic growth per capita. This finding
suggests that in short run, the implementation of conservation policy by limiting
electricity consumption may negatively impact the economic growth. Whilst the
increase of electricity demand can lead to an increase the consumption level which
will give positive contribution to the economic growth.
4.2.2. The Result of Causality Tests in Six Corridors
To understand the nature of causal relationship of each economic corridor within
Indonesia, it is necessary to perform causality test by applying the electricity
consumption data of 14 PLN distributive regions and the regional economic data of
33 provinces. Data at provincial level are regrouped into six economic corridors in
line with the directive of MP3EI document. Considering the size of Indonesia territory
and the geographical location of each corridor, it is useful to run causality test for the
six corridors to find out the relationship of regional electric consumption and the
GRDP for each corridor. The objective of this test is to provide evidence that both the
causality between electricity consumption and the economic growth in each region
and their regional variations will offer new insights for regional and national
electricity policy.
The general direction of the policy formulated in MP3EI document essentially is
to accelerate and expand the economic development in each corridor with the main
objective is to create high and sustainable economic growth (above 7%), to reach all
regions and more balanced growth (CMEA,2011). To create high quality economic
growth, the presence of infrastructure development in all economic corridors is
absolutely required to support its regional development.
Table 10 shows the economic indicators and the characteristic of electricity
consumption in each economic corridor indicate regional variations matter. Jawa is
one of the smallest corridor in terms of size of island among others but by far it has
13
the highest concentration of 55.78% of total population, followed by Sumatra with
22.43% of total population. Jawa and Sumatra regions represent the KBI with 30.63%
of the total area and 78.21% of the total population of Indonesia. In terms of GRDP,
KBI contributes 82.12% of the national GDP. In contrast, the other four corridors
within KTI, i.e. Kalimantan, Sulawesi, Bali-Nusa Tenggara, and Papua-Maluku, only
account for 6.11%, 7.17%, 5.78%, and 2.73% of the total Indonesia’s population
respectively.
Table 10. Economic and Electricity Consumption Indicator in 6 Indonesia’s
Economic Corridors (2010)
Corridors Population GRDP per capita
Electricity
Consumption
per capita
% Millons % Rp Millions % KWh
Sumatra 22.43 50.63 93.98 9.24 60.30 388.64
Jawa 55.78 125.97 109.44 10.76 135.75 874.70
Kalimantan 6.11 13.78 140.33 13.80 57.74 372.13
Sulawesi 7.17 16.21 67.02 6.59 48.58 313.12
Bali-Nusa Tenggara 5.78 13.07 47.80 4.70 50.57 325.90
Papua-Kepulauan Maluku 2.73 6.16 63.63 6.25 30.40 195.93
Total (Indonesia) 100.0 225.82 100.0 9.84 100.0 644.46
Sources: Calculated from BPS (2011); PLN (2011)
KBI region with only representing 30.63% of the national territory has a portion
of 78.21% of national population and 82.12% of GDP. In contrast, KTI region,
representing 69.37% of the national territory, has only a portion of 21.79% population
and contributing only 17.88% of national GDP. The imbalanced population
concentration and economic activities between KBI and KTI regions have created
serious disparity across regions, between and within islands in the Indonesia’s
economic development (Kuncoro, 2013, 2012).
In terms of income per capita, Kalimantan with Rp13.8 millions per capita has
the highest income per capita among all other regions, followed by Jawa (Rp10.76
million per capita), Sumatra (Rp9.24 million per capita), Sulawesi (Rp6.59 million
capita), Papua-Maluku (Rp6.25 million per capita), and Bali-Nusa Tenggara (Rp4.7
million per capita). In terms of electricity consumption, the highest electricity
consumption per capita is in Jawa (874.70 Kwh per capita), followed by Sumatra
(388.64 Kwh per capita), Kalimantan (372.13 Kwh per capita), Bali-Nusa Tenggara
(325.90 Kwh per capita), Sulawesi (313.12 Kwh per capita), and Papua-Kepulauan
Maluku (195.93 Kwh per capita).
Figure 2 shows the regional typology by REC per capita and GRDP per capita
of the six economic corridors. By calculating the average REC per capita as the
vertical axis and the average GRDP per capita as the horizontal axis, the 33 provinces
in Indonesia can be divided into four groups, namely: high electricity consumption
and high income, high income but low electricity consumption, high electricity
consumption but low income, and relatively backward provinces with low electricity
consumption and low income. Jawa corridor has GRDP per capita and REC per capita
far aboved the national average. Whilst, Sumatra and Kalimantan have GRDP per
capita higher than national average, but their electricity consumption per capita are
lower than national average. The three other regions, those are Sulawesi, Bali-Nusa
Tenggara and Papua-Kepualuan Maluku have both REC per capita and GRDP per
capita below the national average.
14
Figure 2. Scatter Plot of GRDP Per Capita and
REC Per Capita in 6 Indonesia’s Economic Corridors, 1984-2010
Sources: Calculated from PLN (2011) and BPS (2011)
In the case of Sumatra, the test result shows a short run uni-directional
relationship moving from electricity consumption (REC per capita) to economic
growth (GRDP per capita). This finding suggests that in the short period of time, any
attempt to subdue the electricity consumption will have effect on economic growth in
Sumatra region. Whilst, the increase of electricity consumption will have a positive
implication on the economic growth of Sumatra region. Figures 3 shows a strong
concurrent upward trend for both electricity consumption and economic growth per
capita. Since 1998 this growth has been consistently staying above the national
average level. This trend indicates that this region is economically more resilient than
Jawa, Sulawesi, Bali-Nusa Tenggara, and Papua-Maluku. Unlike the other corridors,
Sumatra shows a persistent trend of increasing growth of electricity consumption and
GRDP per capita even during Asian crisis (1997-1998) and recent global crisis (2008-
2009).
Figure 3. REC per capita and GRDP per capita, Sumatra 1984-2010
Sources: Calculated from BPS and PLN (2011)
In the case of Kalimantan, the causality tests show a short run uni-directional
relationship moving from economic growth (GRDP per capita) to electricity
consumption (REC per capita) (Table 8). This finding suggests that an increase of
15
economic growth has resulted to an increase of electricity consumption in this region.
Figures 4 shows a strong concurrent upward trend for both electricity consumption
and economic growth per capita. Since 2001, this growth has been consistently high
and staying above the average national level. Recently, the income per capita of
Kalimantan corridor has been the highest among all regions in Indonesia. Similar to
Sumatra, this trend indicates that this region is economically resilient during the
recent global crisis. However, during the Asian Crisis 1997-1998, GRDP per capita
along Kalimantan corridor dropped slightly below that of the national level but REC
per capita persisted to increase above that of the national level.
Figure 4. REC per capita and GRDP per capita, Kalimantan 1984-2010
Sources: Calculated from BPS and PLN (2011)
While both Sumatra and Kalimantan regions show that they have short run uni-
directional relationship (but in opposite direction), the other 4 regions (Jawa,
Sulawesi, Bali - Nusa Tenggara, and Papua–Kepulauan Maluku) do not show any
causal relationship between REC and GRDP. Unlike the other 4 other regions,
Sumatra and Kalimantan regions also consistently show a strong concurrent upward
trend for both electricity consumption and economic growth per capita.
The absence of causal relationship is shown in Jawa, Sulawesi, Bali-Nusa
Tenggara, and Papua-Kepulauan Maluku. This finding suggests that electricity
consumption in these four regions were not correlated with economic growth. It also
implies that neither expansive nor conservative policies related to electricity
consumption did have any effects on economic growth. The finding shows that no
causal relationship between electricity consumption and its regional economic growth
is quite surprising, considering that Jawa is the only region that has more electric
consumption in industrial sector rather than household sector. Furthermore, this
region representing the highest portion of population and GRDP (55.78% and 58.66%
respectively).
Eventhough Jawa has the bigest portion of GRDP and also the highest
consumption of electricity per capita, but the income per capita is significantly lower
than Kalimantan. In the contrary to Sumatra and Kalimantan, the trend of Jawa shows
that the rate of growth in electricity consumption has started to decline and the rate of
economic growth has started to slow down (Figure 5).
16
Figure 5. REC per capita and GRDP per capita, Jawa 1984-2010
Sources: Calculated from BPS and PLN (2011)
The trends of all 3 other regions, as shown by Figure 6 to 8, indicate that the
rate of growth on the electricity consumption has started to decline, while their
economic rate of growth significantly lower compared to Kalimantan, Jawa, and
Sumatra. These regions can be catagorized as regions with low growth of income and
electricity consumption.
Figure 6. REC per capita and GRDP per capita, Sulawesi 1984-2010
Sources: Calculated from BPS and PLN (2011)
The growth of electric power supply in the regions outside Jawa are mostly
allocated to meet the consumption sector, i.e. household sector, rather than industrial
or commercial sector. Although the causal relationship did not occur significantly in
this corridor, however Jawa is the only corridor which has the portion of electricity
consumption of industrial sector (41,27%) higher than household sector (36,61%).
Whilst, Sumatera and Kalimantan regions, with the causal relationship are found,
show much lower portion electricity consumption of industrial sector (18.94% and
6.90% respectively) than household sector (51.31% and 59.94% respectively) Table
11 (column 2).
17
Figure 7. REC per capita and GRDP per capita, Bali-Nusa Tenggara 1984-2010
Sources: Calculated from BPS and PLN (2011)
Figure 8. REC per capita and GRDP per capita,
Papua-Kepulauan Maluku 1984-2010
Sources: Calculated from BPS and PLN (2011)
The electricity consumption in the other three corridors, those are Sulawesi,
Bali–Nusa Tenggara, and Papua–Maluku, are also dominated by household sector. As
shown in Table 11, the portion of household sector is between 47.83% to 59.94%,
while the portion of industrial sector is only between 1,23% to 14,96%. The last
supports Kuncoro (2012b)’s finding that those regions are nonindustrial areas with
very few large and medium manufacturing industries. The increase of consumption
portion of household sector in Sumatra (2009-2011) had been the most significant,
followed by Sulawesi, Bali-Nusa Tenggara, Kalimantan, and Papua-Maluku regions.
As for Jawa, the consumption portion had decreased during the same time period.
5. CONCLUSIONS
Using Granger causality tests and Error Correction Model (ECM), applied to the
quantitative data within the period of 1984-2010, our empirical findings shows that:
(1) there is uni-directional short run relationship moving from electricity consumption
18
per capita to GDP per capita for Indonesia as a single country data set; (2) the
causality tests for Sumatera economic corridors shows there is a uni-directional short
run relationship moving from electricity consumption to economic growth, meanwhile
the test for Kalimantan shows an opposite short run uni-directional relationship;
(3) the causality tests do not show any causal relationship for other corridors (Jawa,
Sulawesi, Bali-Nusa Tenggara, and Papua-Kepulauan Maluku).
Table 11. Electricity Consumption by Household and Industry Sectors in 6
Indonesia’s Economic Corridors (%), 2009-2011
Corridors 2009
(1)
2010
2)
2011
(3) Sumatra
- Household Sector 51.25 51.31 54.52
- Industry Sector 20.23 18.94 18.58
Jawa
- Household Sector 37.30 36.61 36.76
- Industry Sector 40.47 41.27 41.57
Kalimantan
- Household Sector 60.25 59.94 61.59
- Industry Sector 7.19 6.90 6.72
Sulawesi
- Household Sector 52.46 53.94 54.83
- Industry Sector 16.13 14.96 14.53
Bali-Nusa Tenggara
- Household Sector 48.01 47.83 49.66
- Industry Sector 3.43 3.24 3.15
Papua-Kep Maluku
- Household Sector 57.95 57.66 58.97
- Industry Sector 1.07 1.23 1.03
Source: PLN (2012)
To implement MP3EI succesfully, it requires a thorough understanding of
Indonesia’s growth poles and electricity consumptions (and generating power supply)
for each economic corridor. Electricity generating capacity growth in Indonesia has
lagged behind the pace of electricity demand growth, leading to power shortages and
a low electrification ratio. This problem leads to some major challenges to implement
MP3EI (Kuncoro, 2013a, 2012a). Our analysis is a pioneering study of Indonesia’s
electricity consumption and economic growth that incorporates “geography” into
economic analysis of causality during 1984-2010. Before the 1990s, mainstream
economics neglected economic geography, meaning the study of where economic
activity takes place and why (Kuncoro, 2013b, 2012b). Spatial aspects remained a
blind spot for many economists because of their inability to properly model the
various aspects of industrial locations (Krugman, 1995: 31-7). Our study sought to
inject empirical content to the emerging interest in the causality literature by
examining electricity consumption and regional economic growth in the context of
Indonesia’s recent regional autonomy.
Finally, the finding suggest that fundamental changes in electricity and
regional development policies are required. Indonesia needs electricity and regional
development policy that incorporate regional variations in term of electricity
consumption, economic growth, and GRDP per capita (Figure 10). Using the recent
perspectives in the literature of causality, our study demonstrates that Indonesia
represents an excellent example of both the uneven geographic distribution of
electricity consumption and the relationship between electricity consumption and
regional development.
19
Figure 10. National Share of GDRP and Regional Electricity Consumption (2010)
(% share of total Indonesia)
Souce: Compiled from BPS (2011a, 2011c); MEMR (2010b)
REFERENCES
Altinay, G. & Karagol, E. (2005). Electricity consumption and economic growth: Evidence
from Turkey. Energy Economics, Vol. 27, 849-856.
Badan Pusat Statistik (BPS). (2011a). Produk Domestik Regional Bruto Atas Dasar Harga
Konstan 2000 Menurut Provinsi, 2004-2010 [Regional Gross Domestic Products based
on Constant Price 2000 by Provinces, 2004-2010], Retrieved July 25, 2012, from
