Disaggregate energy consumption, agricultural output and economic growth in Pakistan.
Faridi, Muhammad Zahir ; Murtaza, Ghulam
The basic goal of the study is to analyse the impact of energy
consumption (i.e., electricity, oil and gas) on economic growth and
agriculture sector output in Pakistan. It is desirable to find out
relationship between disaggregate energy consumption, economic growth
and agricultural sector output of Pakistan because energy crisis has
become a central issue now-a-days. Production sector of Pakistan relies
on electricity and gas consumption to large extent and these sources of
energy are falling short because of many reasons which is disrupting
output and consequently exports and real output of the country. To
analyse the relationship, we employed time series data from 1972 to
2011. In order to find out long run and short run effects of energy
consumption on agricultural output and economic growth, ARDL modelling
approach to cointegration is applied after scrutinising the stationarity
of data through ADF Test. Where, bound testing procedure is utilised for
cointegration to judge the existence of long run relationship among
variables and ECM models are formulated for short run analysis. Our
econometric models include agricultural output and economic growth as
dependant variables and electricity, coal and gas consumption as
independent and core variables. The findings of the study indicate that
gas and oil consumption are important determinants of economic growth
and agricultural output.
Keywords: Disaggregate Energy Consumption, Agricultural Output,
Economic Growth, ARDL, Co-integration, Pakistan
1. INTRODUCTION
The performance of an economy is generally measured by sustained
rise in GDP growth over the period of time. The economic growth is the
major goal of macroeconomics. According to neo-classical growth theory,
the core factors of growth are labour and capital. In addition to these
factors; technological progress, human capital development etc. are the
most efficient factors of production. Development of technology and use
of mechanisation in production process require energy at massive scale.
So, energy has become a crucial factor of economic growth indirectly.
Energy is widely regarded as a propelling force behind any economic
activity and indeed plays a vital role in enhancing production.
Therefore, highly important resources of energy will enhance the
technology impact manifold. Quality energy resources can act as
facilitator of technology while less worthy resources can dampen the
power of new technology. Ojinnaka (1998) argued that the consumption of
energy tracks with the national product. Hence, the scale of energy
consumption per capita is an important indicator of economic
modernisation. In general countries that have higher per capita energy
consumption are more developed than those with low level of consumption.
The importance of energy lies in other aspect of
development--increase in foreign earnings when energy products are
exported, transfer of technology in the process of exploration,
production and marketing; increase in employment in energy industries;
improvement of workers welfare through increase in worker's salary
and wages, improvement in infrastructure and socio-economic activities
in the process of energy resource exploitation. Thus in the quest for
optimal development and efficient management of available energy
resources, equitable allocation and efficient utilisation can put the
economy on the part of sustainable growth and development. Arising from
this argument, adequate supply of energy thus becomes central to the
radical transformation of the nation's economy.
The main objective of the study is to investigate the effect of
disaggregate energy consumption on agricultural output and generally
overall growth in Pakistan. Because agriculture is the mainstay of
Pakistan economy and is basic production sector. The manufacturing
sector, services sector and even communication sector have secondary
position albeit their growth rates are higher in absolute terms. The
growth rate of agricultural sector is very low. When structural changes
have occurred, the process of mechanisation has taken place in the
agricultural sector. The use of energy has increased for running the
machinery like tubewells, tractors, threshers etc. Due to shortfall of
energy, the output of agricultural sector has dropped.
One of the interesting features of the study is that it
differentiates short run and the long run effect because it has been
observed that impact of energy consumption varies from short to long run
for the same country. For this purpose, we have employed ARDL modelling
to co-integration to find out long run and short run effect. Unit root
problem of the data is handled by ADF test. The rest of the article is
structured as follows. Trends and structure of energy variables are
given in Section 2. Section 3 provides literature review in detail while
data and methodology are given in Section 4. Empirical results and their
discussion are presented in Section 5. At the end, some policy
implications for energy consumption are suggested on the basis of
empirical results.
2. TRENDS AND SIZE OF PAKISTAN ANNUAL ENERGY CONSUMPTION
Total energy consumption measured in oil consumption is 38.8
million tonnes in the year of 2010-11. Currently gas consumption is the
leading one in total energy consumptions that is 43.2 percent of total
energy consumption. Since 2005-06, Gas, electricity and coal consumption
are equally utilised. Oil consumption stood at second position regarding
usage as its usage is 29 percent of total energy consumption.
We present the trends of energy consumption at disaggregate level
in Pakistan over the last decade. The Figure 1 explains the trends of
annual gas consumption. While, Figure 2 and Figure 3 provide the trends
of annual electricity consumption and annual oil consumption
respectively.
[FIGURE 1 OMITTED]
Gas consumption share is equal to four percent of total energy
consumption during 2005-06 to 20010-11. This is because of the
substitution of gas for expensive energy sources. The consumption of oil
in Pakistan decreased by three percent during the period 2001-2011
because of high prices of oil in the international market. Since the
year 2001 -02, a decreasing trend is observed in the consumption of
petroleum products.
[FIGURE 2 OMITTED]
Yet it is observed that there has been an increase in oil
consumption from 200410, the overall average increase for last ten years
stood at 11 percent per annum. Trends indicate that due to high
volatility in the oil prices consumption intensity is shifting from oil
consumption to some others sources of energy consumption. Figure 2
indicates that the trend of annual electricity consumption (in Giga Watt
Hour) over the last ten years i.e., 2001-2011. Trends show that
electricity consumption increased continuously till 2007 and then fell.
But after the year 2010, there is sharp decline in electricity
consumption. Thus, Gas, electricity and oil consumption trends indicate
an annual increase at an average rate of 5.1 percent, 4.8 percent and
7.7 percent respectively.
[FIGURE 3 OMITTED]
3. LITERATURE REVIEW
Theoretically, neo-classical and endogenous theories both suggest
that energy use and efficiency are drivers of economic growth. Though
there are many studies that find a direct relationship between
productivity and energy consumption in the industrialised world [see
Worrell, el al. (2001)], evidence from the developing world remains
inconclusive. Few disaggregated studies have been conducted on this
issue and the studies using data aggregated at the national or economic
level indicate mixed findings. Further complicating the relationship is
the extent to which economic growth and energy consumption can
theoretically be decoupled, a question raised by ecological economists
who argue thermodynamic laws limit such division. Below is a brief
review of the various theories on the relationship between energy
consumption, energy efficiency and economic growth, followed by a
summary of a select list of empirical studies.
By incorporating energy end-use efficiency gains into a
Cobb-Douglas production function, Wei (2007) theorizes about short-term
and long-term effects of increased energy efficiency beginning with the
production function specification as output is a function of labour,
capital and some measures of energy consumption. In the short term,
energy use efficiency is found to lower the cost of non-energy goods and
increase the output of non-energy goods. A 100 percent rebound effect is
evident such that in the short term, energy efficiency gains have no
effect on absolute energy use. In the long term, the impact on
non-energy output of energy end use efficiency is positive. The long
term impact of energy use efficiency on total energy use is lower than
the short-term impact. Wei also finds that energy use efficiency will
increase real energy price in the long term. Van Zon and Yetkiner (2003)
modify the Romer model to include energy consumption of intermediates
and to make them heterogeneous due to endogenous energy-saving technical
change. They found out that energy-saving technical transformation can
enhance economic growth. On the other hand, it may dampen economic
growth with the increase in energy prices that imply that rising real
energy prices consistently will cause to harm economic growth.
Embodied technical change includes improvements in energy
efficiency, thus positively linking improvements in energy efficiency to
economic growth. They conclude that in an environment of rising energy
prices, recycling energy tax proceeds in the form of R&D is
necessary for both energy efficiency growth and output growth. Sorrell
(2009) pointed out that conventional and ecological economists have
conflict on the issue of energy effects on economic growth. The growth
models presented by Neo-classical and new Endogenous growth theories
give little importance to energy consumption as a major factor of
production by giving argument that it has a small share in total cost of
production. Ecological economist contests their point of view by
replying that over the last two centuries, energy inputs are
accelerating economic growth at valuable rate.
For a steady economic growth the role of technological change is of
great importance as earlier growth models have integrated technological
change as an important factor for growth [Solow (1956)]. Energy and raw
material besides labour and capital cause to decrease the statistical
residual. Onakoya, el al. (2013) studied the relationship between energy
consumption and Nigerian economic growth during the period of 1975 to
2010 to find out energy consumption as an important variable for
production. Co-integration results provided evidence of a long run
relationship between energy consumption and economic growth which was
positive. Same results were also found by Paul and Bhattacharya (2004)
who employed Engle-Granger technique to investigate the direction of
relationship between economic growth and energy consumption for India
for the period of 1950-1996. Results revealed that energy consumption
has causality for energy consumption. Hondroyiannis, et al. (2002)
followed the same results in case of Greece by using vector
error-correction estimation on the data from 1960-1996. The findings of
the study indicate the existence of long run relationship.
Oh and Lee (2004) did not find the significant and positive effect
of energy consumption on growth in case of Korea. For Bangladesh,
Mozumder and Marathe (2007) examined a positive relationship between per
capita income and per capita energy consumption. The relationship
between gas consumption and growth was analysed by Apergis and Payne
(2010) to reveal the co-integration among labour, capital, gas
consumption and economic growth. ECM model was employed to find the
bidirectional causality between gas consumption and economic growth but
Yang (2000) opposed this relationship as his study show the absence of
long run relationship between natural gas consumption and real GDP. Same
results of no relationship are also found out by Aqeel and Butt (2001).
Shahbaz and Feridun (2011) investigated the impact of electricity
consumption on economic growth in Pakistan between 1971 and 2008 by
using ARDL technique to identify the long run relationship between
electricity consumption and economic growth. Study gives the evidence of
long run relationship between electricity consumption and economic
growth but inverse is not true. Alam and Butt (2001) investigation
provided the evidence that structural changes also cause to change the
share of various energy consumption variables. And increase in energy is
because of increase in economic activity as well as structural changes.
Javid, et al. (2013) argued that shocks to electricity supply will
have a negative impact on economic growth. Nwosa and Akinbobola (2012)
and Dantama, et al. (2011) come to a conclusion that govt, should adopt
sector specific energy policies rather the one fit-for-all policy by
observing positive aggregate energy consumption and sectoral output.
For Pakistan, Kakar and Khilji (2011) explored the nature of
relationship between economic growth and total energy consumption for
the period 1980-2009 by using Johansen Co- integration and confirmed
that energy consumption is essential for economic growth and any energy
shock may affect the long-run economic development of Pakistan. Ahmad,
et al. (2013) analysed the impact of energy consumption and economic
growth in case of Pakistan employing data from 1975 to 2009. The results
of ordinary least squares test show positive relation between GDP and
energy consumption in Pakistan.
A number of reviews of prior work compel us to make a healthy
endeavour on the concerned issues because a little attention has been
given to agricultural sector regarding energy consumption relationship.
We have observed in the literature review most of the studies are
emphasising on the relationship between overall growth and energy,
manufacturing sector growth and energy. A few studies discuss the
agricultural sector growth and energy. But the present study removes a
number of imperfections of previous studies such as use of energy
consumption and its relationship with overall economic growth instead of
growth in agricultural sector at the disaggregate level. We have used
fresh data on certain variables. An appropriate technique for
co-integration, model specification and proper estimation technique is
employed.
4. DATA AND METHODOLOGY
The present segment consists of data and methodology used to
estimate effects of disaggregates energy consumption on economic growth
and Agricultural output in Pakistan. To order to analyse relationships,
secondary data from year 1972-2011 are employed and Auto Regressive
Distributed Lags (ARDL) technique has been used.
(a) Data Source
The data generated from Pakistan Economic Survey (various issues),
Handbook of Statistics of Pakistan Economy. While, data on variables of
energy consumption, have been obtained from HDIP, Ministry of Petroleum
and Natural Resources. The variables about which data are collected, are
RGDP (Gross Domestic Product) that is used as dependent variable while
RGFCF (Real Gross Fixed Capital Formation), TELF (Total Employed Labour
Force), IR (Inflation Rate), TOC (Total Oil Consumption), TGC (Total Gas
Consumption), TEC (Total Electricity Consumption), AGRI (Agricultural
Output), TELF (Total Employed Labour Force), RAGFCF (Real Agricultural
Gross Fixed Capital Formation), TOC (Total Oil Consumption), TGC (Total
Gas Consumption), TEC (Total Electricity Consumption), ACRDT
(Agricultural Credit).
(b) Methodological Issues
The study is based on time series data. In order to examine the
properties of the time series data, we first examine the stationarity of
data and then decide about the appropriate technique.
(i) Stationarity of Data
In practice, ADF test is used to check the stationary of variables
to see if all the variables are integrated of degree one. In this case,
the variables can be estimated by employing error correction model
because of co-integrated series. However, if all the variables are not
integrated of same degree i.e. some variables are integrated at I (1) or
some are at I(0) or both I(1) and I(0) then ARDL modeling approach will
be employed to identify the existence of long run and short run
relationships among the variables.
(ii) Auto Regressive Distributed Lag Approach to Co-integration
ARDL approach will be applied only on single equation. It will
estimate the long run and short run parameters of model simultaneously.
The estimated model obtained from the ARDL technique will be unbiased
and efficient. ARDL approach to cointegration is useful for small sample
Narayan (2004). Engel-Granger and Johensan technique are not reliable
for small samples. ARDL gives better results in sample rather than
Johesan co-integration approach. ARDL approach has a drawback because it
is not necessary that all variables are of same order. The variables can
be at I(0) or I(1) or combination of both, the ARDL approach can be
applied. If the variables are stationary at higher order of I(1) then
ARDL is not applicable. ARDL approach consists of two stages. First, the
long run relationship between variables is tested using F-statistics to
determine the significance of the lagged levels variables. Second, the
coefficient of the long run and short run relationship will be examined.
(iii) Bound Testing Procedure
The bound test is based on three basic assumptions that are; first,
use ARDL model after identifying the order of integration of series
Pesaran, et al. (2001). Second, series are not bound to possess the same
order of integration i.e., the regressors can be at I(0) or I(1). Third,
this technique estimates better results in case of small sample size.
The vector auto regression (VAR) of order p, for the economic growth
function can be narrated as Pesaran, et al. (2001);
[Z.sub.t] = [mu] + [p.summation over (i=1)][[beta].sub.i]
[z.sub.t-i] + [[epsilon].sub.t] ... ... ... ... ... ... (1)
Where [x.sub.t] and [y.sub.t] are included in vector [z.sub.t] .
Economic growth (RGDP) and agricultural output (AGRI) are indicated by
[y.sub.t] and [x.sub.t] is the vector matrix which represents a set of
explanatory variables such as [.Xt = RGFCF, TELF, TOC, TEC, TGC, IR] and
[Xt = TELF, RGFCF, TOC, TGC, TEC, ACRDT] for Model-1 and Model-2 and t
denotes time indicator. Vector error correction model (VECM) is given as
below:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
where [DELTA] is the first-difference operator. The long-run
multiplier matrix [lambda] as:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
The diagonal elements of the matrix are unrestricted, so the
selected series can be either 1(0) or 1(1). If [[lambda].sub.YY] = 0,
then Y is 1(1). In contrast, if [[lambda].sub.YY] < 0, then Y is
1(0).
The VECM procedures described above are imperative in the testing
of at most one co-integrating vector between dependent variable
[y.sub.t] and a set of regressors [x.sup.t]. To build up the model,
study uses Pesaran, et al. (2001) postulation of Case V, that is,
unrestricted intercepts and trends.
(c) Description of the Variables
In the present analysis, we have used the variables like employed
labour force and real gross fixed capital formation as theoretical
variables for growth and there are three core variables relating to
energy. Two variables are used as control factors. The explanation and
hypothetical relation of these variables are given below.
Real Gross Domestic Product (RGDP)
Real gross domestic product at factor cost is used as proxy for
economic growth. It is assumed as GDP expands over the period of time,
the economy will grow. RGDP is measured in millions rupees.
Agricultural Output (AGRI)
In order to measure the performance of agricultural sector, we have
used agricultural output measured at current market prices in million
rupees.
Real Gross Fixed Capital Formation (RGFCF)
We have considered real gross fixed capital formation as a proxy
for capital in the present study. It is measured at market prices in
million rupees.
Total Employed Labour Force (TELF)
Labour is used as a core variable in economic growth model. It is
expected that labour contributes positively to economic growth. The
present study uses total employed labour force as a proxy for labour.
Total employed labour force is measured in millions peoples.
Total Oil Consumption (TOC)
Total oil consumption is measured in thousands of tonnes per year.
It is expected that oil consumption has positive relationship with
growth.
Total Gas Consumption (TGC)
It is expected that the utilisation of gas consumption cause to
increase the GDP growth positively. We have used total gas consumption
in million cubic feet (mmcft).
Total Electricity Consumption (TEC)
Use of electricity in production process is an important factor.
Due to shortage of electricity it is expected that total electricity
consumption is contributing negatively to GDP growth as well as to
agriculture output. The total electricity consumption per Annam is
measured in Giga Watt hour (GWh) or ([10.sup.6] Kilo Watt hour).
Agricultural Credit (ACRDT)
Agricultural credit is used as a central variable in the present
analysis. The expected impact of agricultural credit on output is
positive. Agricultural credit is measured in million rupees.
Inflation Rate (IR)
In order to examine the effect of general price level on economic
growth, we have used consumer price index as a proxy for inflation rate.
The inflation rate has negative impact on economic growth because cost
of the production increases, output falls and growth is retarded.
(d) Model Specification
The current study is based on general Neo-classical Production
Function;
Y = A f(L, K) ... ... ... ... ... ... ... (3)
Where, Y = Total output, L = Total employed labour force, K= Total
stock of capital, A= Total productivity factor.
We have employed extended neo-classical growth model by
incorporating energy as a productivity factor as an endogenous variable.
A = f(TOC, TGC, TEC) ... ... ... ... ... ... (4)
Substituting A in Equation (i), we obtained extended growth model.
Y = f(L,K, TOC, TGC, TEC) ... ... ... ... ... (5)
Based on the suggested economic techniques, we have two specified
model. These specified models are given below.
Model-1. Impact of Disaggregate Energy Consumption on Economic
Growth
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)
Where, [DELTA] is the first-difference operator while Ut is a
white-noise disturbance term. This model would estimate the impact of
disaggregate energy consumption on economic growth in which real GDP is
used as dependant variable while real gross fixed capital formation
(proxy for capital), total employed labour force, total oil consumption,
total gas consumption an total electricity consumption are used as
independent variables.
Equation (6) also can be viewed as an ARDL of order (a, b, c, d, e,
f g). Equation (6) indicates that economic growth tends to be influenced
and explained by its past values. The structural lags are established by
using minimum Schwarz Information Criteria (SIC). In our model, we will
use the lagged value of first difference dependent variable and
independent variables for short run and first lagged values of dependent
and independent variables for long run. So, this model is consisted of
both long run and short run coefficients of variables as well. Where
[[beta].sub.1], [[beta].sub.2], [[beta].sub.3], [[beta].sub.4],
[[beta].sub.5], and [[beta].sub.6], [[beta].sub.7] are the short run
coefficients of variables and [[beta].sub.8], [[beta].sub.9],
[[beta].sub.10], [[beta].sub.11], [[beta].sub.12] and [[beta].sub.13],
[[beta].sub.14] are the long run coefficients of variables and
[[beta].sub.0] is the intercept term.
Model-2. Impact of Disaggregate Energy Consumption on Agricultural
Output
The second model would capture the effect of energy consumption on
agricultural output in Pakistan with the help of some explanatory
variables like TELF (Total Employed Labour Force), RGFCF (Real Gross
Fixed Capital Formation), TOC (Total Oil Consumption), TGC (Total Gas
Consumption), TEC (Total Electricity Consumption), ACRDT (Agricultural
Credit); the unrestricted ECM model for Agricultural output is as under;
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)
Where [DELTA] shows the first difference operator and [U.sub.t], is
the residual of the model.
Equation (7) also can be viewed as an ARDL of order (p, q, r, s, t,
u, v). Where [phi].sub.li], [phi].sub.2i], [[phi].sub.3i] and
[[phi].sub.4i], [[phi].sub.5i], [[phi].sub.6i], [[phi].sub.7i], are the
short run coefficients of variables and [[gamma].sub.1],
[[gamma].sub.2], [[gamma].sub.3], [[gamma].sub.4, [[gamma].sub.5],
[[gamma].sub.6] and [[gamma].sub.7] are the long run coefficients of
variables and [[phi].sub.0] is the intercept term.
The Wald Test (F-statistics)
After regression of Equation (6) and Equation (7), the Wald test
(F-statistic) is computed to differentiate the long-run relationship
between the concerned variables. The Wald test can be carried out by
imposing restrictions on the estimated long-run coefficients of real
GDP, total employed labour force, real gross fixed capital formation,
total oil consumption, total gas consumption, total electricity
consumption and inflation rate for the Model-1 as under:
The null hypothesis is as follows;
[H.sub.0]: [[beta].sub.8] = [[beta].sub.9] = [[beta].sub.10] =
[[beta].sub.11], = [[beta].sub.12]= [[beta].sub.13] [[beta].sub.14] = 0
(No long-run relationship exists)
Against the alternative hypothesis,
[H.sub.1]: [[beta].sub.8] [not equal to][[beta].sub.9] [not equal
to][[beta].sub.10] [not equal to][[beta].sub.11] [not equal
to][[beta].sub.12] [not equal to][[beta].sub.13] [not equal
to][[beta].sub.14] [not equal to] 0 (A long-run relationship exists)
If the calculated F-statistics does not exceed lower bound value,
we do not reject Null Hypothesis and it is concluded that there is no
existence of long run relationship between RGDP and independent
variables. On the other hand, if the calculated F-statistics exceeds the
value of upper bound, the co-integration exists between RGDP and
independent variables. We will apply the Wald coefficient test on all
lagged explanatory and dependant variables in the model Equations (7).
Our null hypothesis will be that lagged coefficient of explanatory
variables are equal to zero or absent from the model. If we do not
reject the null hypothesis it means long run relationships among
variables do not exist.
Null and alternative hypothesis for Model-2 to apply Wald test is
as follows.
[H.sub.0]: [[gamma].sub.1] = [[gamma].sub.2] = [[gamma].sub.3] =
[[gamma].sub.4] = [[gamma].sub.5] = [[gamma].sub.6 = [[gamma].sub.7] = 0
(No Cointegration Exists)
[H.sub.1]: [[gamma].sub.1][not equal to] [[beta].sub.2] [not equal
to] [[beta].sub.3] [not equal to] [[beta].sub.4] [not equal to]
[[beta].sub.5] [not equal to] [[beta].sub.6] [not equal to]
[[beta].sub.7] [not equal to] 0 (Cointegration Exists)
(d) The Time Horizons
To see the effects of explanatory variables on economic growth in
case of Pakistan both in the short run and long run, we have to estimate
the model which are given Equations (6) and (7) with OLS (Bound test
approach to co-integration) technique and then normalise the resulting
values. The ARDL model for the long run coefficient of Model-1 Equation
(6) is to determine the long run effect of energy consumption on
economic grown in Pakistan,
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] ... ... ... ...
... (8)
The ARDL model for the long run coefficients of Model-2 Equation
(7) is to capture the long run energy consumption effects on
agricultural output in Pakistan.
AGRI = [theta] + [summation] [theta] (TELF) + [summation] [theta]
(RGFCF) + [summation] [theta] (TOC) + [summation] [theta] (TEC) +
[summation] [theta] (TGC) + [summation] [theta] (ACRDT) + [mu] ... (9)
Now we will find the short coefficient of the model with error
correction term. We will use the short run error correction estimates of
ARDL model. The difference between actual and estimated values is
considered as error correction term. Error correction term is defined as
adjustment term showing the time required in the short run to move
toward equilibrium value in the long run. The coefficient of error term
has to be negative and significant. The short run error correction (ECM)
model of Model-1 Equation (6) to find out impact of energy consumption
on economic growth in time adjusting frame work to attain long run
equilibrium is as follows;
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (10)
[ECM.sub.t-1] is lagged error correction term of the model and
[lambda] is the coefficient value of ECM which is the speed of
adjustment.
The short run (ECM) model of Model-2 Equation (7) to find out
impact of energy consumption on Agricultural output in Pakistan in time
adjusting frame work to attain long run equilibrium is as follows.
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (11)
[ECM.sub.t-1] is lagged error correction term of the model and
[omega] is the coefficient value of ECM which is the speed of
adjustment.
The Error Correction Term ([EC.sub.t-1)
The error correction term ([EC.sub.t-1]), which instrument the
adjustment speed in the dynamic model for restoring equilibrium.
Banerjee, et al. (1998) grasped that a highly significant errors
correction term is further proof of the existence of stable long run
relationship. The negative sign of error correction term also give
uni-directional of variables.
4. RESULTS AND DISCUSSIONS
After discussing the data sources, we analyse the impact of
disaggregate energy consumption on economic growth and Agricultural
output on empirical grounds. To analyse these issues, we will provide an
insight to draw some conclusions on the basis of empirical results of
this research. The results are discussed as follows.
(a) Descriptive Analysis
The descriptive statistics of the study are presented in the Table
1. Descriptive statistics consists of procedures used to summarise and
describe the characteristics of a set of data. The table shows the
averages values, standard deviation, skewness, kurtosis and J. Bera
values of the selected variables.
Our study is based on the 41 years of annual observation for the
period 1972-2011. Descriptive statistics on some important variables are
reported in Table 1. We have found that the average agriculture
productivity is 587531.9 million rupees with 737717.6 units'
standard deviation. The mean value of the total oil consumption, total
gas consumption is 10465494 units, 32961.79 units and 550732.2 units
respectively with low variability as compared with mean values. The
value of Jarque- Bera JB test states that residual of the core variables
like RGFCF, TEC, TELF, TGC and TOC are normally distributed. The values
of the co-efficient of skewness show that almost all the variables are
positively skewed expect total oil consumption.
(b) ADF Test for Stationarity
Table 2 explains the summary statistics of ADF test. The results of
the test indicate that some variables are stationary at level and others
are stationary at first difference. The findings of the study provide
the justification of ARDL Approach.
Bounds Test for Co-integration
In the first step the existence of the long run relationship among
the variables is needed. We have used Bound Testing Approach in order to
examine the long run relationship. Table 3 interprets the findings of
Wald-Test (F-Statistics) for long-run relationship.
The value of F-statistics based on Wald test is given in second
column. The upper bound values are reported in third column of Table 2.
The results of the test indicate that there exits long-run relationship
among the variables in both models.
Estimates of Energy Consumption and Economic Growth
The long-run estimates of the model-1 are reported in Table 4. The
dependant variable is economic growth which is proxied as real GDP
whereas RGFCF, TELF, TOC, TEC and TGC, IR are independent variables.
We have observed that the value of regression coefficient of Real
Gross Fixed Capital Formation (RGFCF) that is 604.54 which means that
the one unit increase in Real Gross Fixed Capital Formation increases
the economic growth (RGDP) by 604.54 units and this effect is strong and
statistically significant. The expansion of infrastructure directly
stimulates productive activities. The other channel may be that
investment spending in various projects raises overall productivity and
economic growth. Our results stay in line with Khan and Reinhart (1990);
Blomstrom, et al. (1994) who find positive relationship between
investment and growth.
The coefficient of the employed labour force is although positive
but insignificant. Our findings are matched with conventional
neo-classical theories of growth [see Barrow and Sala-i-Martin (1995)].
The core variables of the study are energy variables i.e., total
energy consumption, total gas consumption and total electricity
consumption. We have noted in the present study that total oil
consumption directly influence the economic growth. The value of the
coefficient of oil consumption is 0.90 which means that an increase of
one unit in total oil consumption raises real GDP about 0.90 units. The
same results are found in the short run. The findings support the
theoretical results. The reason may be that the wheel of the economic
life cannot be run without oil now-a-days because of mechanisation and
technological progress.
We have observed that the coefficient of total gas consumption is
positive and highly significant. The real GDP increases almost 15.6
units due to one unit increase in total gas consumption. It is noted
that the third variable of the energy turns out to be negative. The
coefficient of the total electricity consumption is (-346.85) and
statistically significant. The short run findings also indicate negative
impact on growth. The analysis concludes that electricity is considered
as limiting factor to economic growth in Pakistan. The reason may be
that the continuous short fall of the electricity and electricity supply
shock are the main causes of growth deterioration. Our results support
the [Javaid, et al. (2013); Kakar and Khilji (2011); Shahbaz, et al.
(2013); Onakoya, et al. (2013) and Yuan, et al. (2007)] findings.
The inflation rate is used as control variable in the growth model.
The analysis concludes that the effect of inflation rate on economic
growth is negative and statistically insignificant. Theoretically, it is
sound because rising prices cause an increase in the cost of production.
As a result production decreases and ultimately economic growth
declines.
Interpretation of Error Correction Term ([EC.sub.t-1])
The coefficient of ecmt-1 for Model-1 is equal to (-0.38) for the
short-run model and implies that deviation from the long-term economic
growth is corrected by 38 percent over each year at 1 percent level of
significance.
Estimates of Disaggregate Energy Consumption and Agricultural
Output
The value of regression coefficient of real Gross Fixed Capital
Formation (RGFCF) is 8.92 which means that the one unit increase in real
Gross Fixed Capital formation raises the Agricultural output by 8.92
units. The reason may be that investments in agriculture input industry
like tractors, thrashers, tube wells and pesticides increase along with
an increase in the income of the farmer. Therefore, per capita saving
rate increases and ultimately growth per capita increases [Barro
(1991)].
We have observed that the value of regression coefficient of
Employed Labour Force (TELF) is 3033. This means that the one unit
increase in Employed Labour Force increases agricultural output by 3033
units and the result is statistically insignificant.
We have found that the coefficient of total gas consumption is 1.81
and statistically highly significant. The agricultural product increases
by about 1.8 units due to one unit increase in total gas consumption.
The results may be justified on the sound reasoning that fertiliser and
pesticides producing industries have shifted their production process
from electricity usage to gas usage considering it cheaper source of
energy.
The estimated coefficient of Total Oil Consumption is .054. This
means that the one unit increase in Total Oil Consumption increases
agricultural output by .054 units. The estimated coefficient of Total
Electricity Consumption (TEC) is-9.41 which implies that agricultural
output is affected negatively by electricity consumption and is
statistically significant. The agricultural credit is contributing
positively in boosting up economic growth as coefficient of agricultural
credit is 8.83 and is significant. Results are consistent with [Ayaz, et
al. (2011)]. Formal credit directly influences agricultural productivity
through investment and financing of fertilisers and seeds [Qureshi and
Shah (1992); Jehanzeb, et al. (2008)].
Interpretation of Error Correction Term (E[C.sub.t-1])
The value of Ecmt-1 for Model-2 is (-.46) which implies that the
short run variables approach to long run variables by 46 percent each
year. Negative and significant value of error correction term also
provides further proof of existence of long run and unidirectional
relationship [Bannerjee, et al. (1998)].
Diagnostic Tests
J-B normality test for residual is conducted to see residuals are
normally distributed or not because one of the assumptions of CLRM is
residuals are normally distributed with zero mean and constant variance.
Breusch-Godfrey LM test is conducted to check the serial autocorrelation
in our model. Autoregressive conditional heteroskedasticity (ARCH) is
conducted to check the autocorrelation in the variance of error term.
The outcomes of all these tests are reported in the Tables 8 and 10.
Stability Test
In order to check the stability of the Models, we plot the
cumulative sum of recursive residuals CUSCUM and cumulative sum of
recursive residuals of square CUSUMS. The results show that coefficients
in our estimated models are stable as the graph of CUSUM and CUSUMS
statistics lies in the critical bounds. The absence of divergence in
CUSUM and CUSUMS graphs confirm that in our ARDL Models, short run and
long run estimates are stable.
[GRAPHIC OMITTED]
[GRAPHIC OMITTED]
[GRAPHIC OMITTED]
[GRAPHIC OMITTED]
5. CONCLUSIONS
In this study, we have analysed the impact of disaggregate energy
consumption on economic growth and Agricultural output on empirical
grounds with respect to Pakistan. Study has used ADF test which
indicates mixed results with different order of integration. Existence
of long run relationship among variables is examined for both models.
Long run estimation and error correction representation of both models
have been discussed and their interpretations are made. Findings of the
study conclude that disaggregate energy consumption, economic growth and
agricultural output are interlinked with each other in short as well as
in long run.
The empirical analysis of disaggregate consumption on economic
growth and on agricultural output leads to a number of conclusions for
policy formulation. Electricity consumption and economic growth puts
some essential policy implications on the economy of Pakistan. The
unidirectional relationship of electricity consumption to economic
growth and agricultural output leads us to draw a conclusion that
shortage of electricity supply at the prevailing level can harm
Pakistan's economic growth and agricultural output. As, consumption
of electricity can influence national and agricultural output as it is
the main source of energy, that is why it is significant to maintain the
supply of electricity according to its demand. And since in cyclical
sense economic fluctuations are also caused due to changes in
electricity consumption which implies that electricity may be a leading
indicator for business cycle. Another important implication is that as
oil consumption and gas consumption are contributing positively to
economic growth and agricultural growth, therefore, Pakistan energy
sources (i.e., oil, coal and gas) other than electricity should be
enhanced for sustainable economic growth because Pakistan production
sectors like agricultural sector also rely on electricity consumption
mainly and increasing demand of electricity as compared to its supply
and insufficient installed capacity reduce agricultural as well as
national output.
Comments
The paper titled "Disaggregate Energy Consumption, Agriculture
Output and Economic Growth: An ARDL Modelling Approach to
Co-integration" touches upon an important subject of policy making
in the context of economic growth.
Having said that let me point out some of the weakness which to my
understanding if improved can make this paper very useful both for
academia and policymakers.
(i) The use of the econometrics technique such as ARDL is a norm
now a day, hence putting the same in the title does not signifies it any
more. So I would recommend that the authors should stick to the economic
title only.
(ii) Putting such words on the title page, e.g. in the abstract
data ranges from 1972-2011 from a reliable source ... leaves the reader
thinking that is it an insider job. Such mentioning is taken care of in
the data and methodology section. Subsequently these sources are also
not mentioned in data section.
(iii) There is a need to carefully review and state the objectives
of the study.
(iv) There is no discussion of why the authors have just picker the
agricultural out put to be a representative of the sub components of GDP
and leave out other potential sectors which are also contributing to the
economic growth such as manufacturing, communications etc.
(v) The descriptive analysis is out of context and does not help
the reader in establishing the linkage between the variables of
interest. The use of data is also not appropriate such as Figure 2 is
totally out of context and not discussed at all.
(vi) There is a confusion across the paper as to authors are
focusing on the consumption, the efficient consumption of the energy
sources at disaggregated level or the energy-mix in use.
(vii) The qualification such as.... The study also fulfils a number
of imperfections of previous studies such as not using appropriate
technique for co-integration, model specification and methodological
issues requires the literature review to be set accordingly and the next
coming sections such as the methodology etc. to further qualify that.
(viii) Variables abbreviations such as RGFCF, TELF ... does not
convey its description.
(ix) There is no model as such, mentioning the variables in a
simple production function with variables of interest a arguments of a
function is not a model.
(x) Tables when placed needs an explanation, e.g., Table 1: on the
descriptive statistics.
(xi) For estimation of the regression ARDL approach is used:
(a) The results for unit root test are not provided for inclusion
of intercept and trend or there is no plot of the data. Further for
robust results often PP test is also applied but not in this case.
(b) Now once it was observed that all the variables at 1(1) except
1R and RGFCF (which 1 don't know what these are) which are 1(0).
Then a simple cointegration method like Johanson and Jusilus or Engle
and Granger was more appropriate leaving these two, as the ARDL is
adopted if the variables under consideration have different order of
integrations (i.e. a mix of 1(0) and 1(1)).
(c) While comparing the wald-F test for existence of cointegration
Pesaran, et al. (2001) tables are used, which were for large samples
(500-1000), for our case where the total observations are around 40 we
have to use the tables provided by Naryan (2005) otherwise it may get
non-parsimonious results as the F-test used here has a non-standard
distribution and depends on the (1) Variables being 1(0) or 1(1), (2) No
of repressors, (3) Intercept and/or trends and the (4) sample size. So
we can not use the old tables for exploring the critical bound.
(d) The Cusum and Cusum Square tests are showing a unique picture,
where the bounded line appears only the latter years and not for the
whole sample period, please explain.
(xii) In the results description what bothers me is the results and
there explanation, e.g. 1 unit Oil consumption leading to an increase of
about just 0.90 units in the GDP. First I am unaware as to the use of
Oil consumption units used here are in energy units or expenditures on
oil consumption. Second these results are somewhat unexpected also not
validated with the help of other studies.
(xiii) There is a strong possibility of multicollinearity in the
estimation as both the Oil consumption and electricity usage is taken as
explanatory variables.
(xiv) Further the results are totally in abeyance of any possible
explanation, e.g. electricity consumption presents a negative
relationship with economic growth.
(xv) Further taking inflation to be linked with economic growth
means we may need to explore the sacrifice ratio, but that has to be
through the demand side, whereas in Pakistan inflation may be arising
from the supply side.
(xvi) Investment in human capital is not synonymous to R&D.
(xvii) The paper needs a through reading and then editing.
(xviii) Conclusions are based on empirical work which is largely
not reflecting the true logics. Further basing policy recommendations
which are not arrived at from the authors estimation should not be put
forth.
Over all the study needs a thorough revision both in the context of
theoretical understanding and the econometric methodology on how to
estimate it.
Mahmood Khalid
Pakistan Institute of Development Economics, Islamabad.
REFERENCES
Ahmed, W., K. Zaman, S. Taj, R. Rustam, M. Waseem, and M. Shabir
(2013) Economic Growth and Energy Consumption Nexus in Pakistan. South
Asian Journal of Global Business Research 2:2, 251-275.
Alam, S. and M. S. 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.
Apergis, N. and E. Payne (2010) Natural Gas Consumption and
Economic Growth: A Panel Investigation of 67 Countries. Applied Energy
87, 2759-2763.
Aqeel, A. and M. S. Butt (2001) The Relationship between Energy
Consumption and Economic Growth in Pakistan. Asia- Pac. Dev. J. 8,
101-110.
Ayaz, S., S. Anwar, M. H. Sial and Z. Flussain (2011) Role of
Agricultural Credit on Production Efficiency of Farming Sector in
Pakistan - A Data Envelopment Analysis. Pak. J. LifeSoc. Sci. 9:1,
38-44.
Banerjee, A., J. J. Dolado, and R. Mestre (1998) Error-Correction
Mechanism Tests for Cointegration in a Single-Equation Framework.
Journal of Time Series Analysis 19, 267-283.
Barro, R. (1991) Economic Growth in a Cross-section of Countries.
Quarterly Journal of Economics 106, 407-443.
Barro, R. J. and X. Sala-i-Martin (1995) Economic Growth. New York:
McGraw-Hill.
Blomstrom, M., R. Lipsey, and M. Zejan (1994) What Explains Growth
in Developing Countries? In W. Baumol, R. Nelson and E. Wolff (eds.)
Convergence of Productivity: Cross-National Studies and Historical
Evidence. Oxford and New York: Oxford University Press. 243-59.
Dantama, Y. U., Y. Z. Abdullahi, and N. Inuwa (2011) Energy
Consumption-Economic Growth Nexus in Nigeria: An Empirical Assessment
Based on ARDL Bound Test Approach. European Scientific Journal 8:12.
Hondroyiannis, G., S. Lolos, and E. Papapetrou (2002) Energy
Consumption and Economic Growth: Assessing the Evidence from Greece.
Energy Economics 24, 319336.
Javid, A. Y., M. Javid, and Z. A. Awan (2013) Electricity
Consumption and Economic Growth: Evidence from Pakistan. Economics and
Business Letters 2, 21-32.
Kakar, Z. K. and B. A. Khilji (2011) Energy Consumption and
Economic Growth in Pakistan. Journal of International Academic Research
11:1.
Khan, M. S. and C. M. Reinhart (1990) Private Investment and
Economic Growth in the Developing Countries. World Development
(January).
Mozumder, P. and A. Marathe (2007) Causality Relationship between
Electricity Consumption and GDP in Bangladesh. Energy Policy 35,
395-402.
Nwosa, P. I. and T. O. Akinbobola (2012) Aggregate Energy
Consumption and Sectoral Output in Nigeria. African Research Review 6:4,
206-215.
Oh, Wankeun and Kihoon Lee (2004) Energy Consumption and Economic
Growth in Korea: Testing the Causality Relation. Journal of Policy
Modeling 26: 973-981.
Ojinnaka, 1. P. (1998) Energy Crisis in Nigeria: The Role of
Natural Gas. CBN Bulletin 22:4, 8-12.
Onakoya, A. O., O. A. Jimi - Salami, and B. O. Odedairo (2013)
Energy Consumption and Nigerian Economic Growth: An Empirical Analysis.
European Scientific Journal 9:4, 25-40.
Paul, S. and R. N. Bhattacharya (2004) Causality between Energy
Consumption and Economic Growth in India: A Note on Conflicting Results.
Energy Economics 26, 977-983.
Pesaran, M. H. and Y. Shin (1999) An Autoregressive Distributed Lag
Modelling Approach to Cointegration Analysis. In S. Strom, A. Holly and
P. Diamond (eds.) Econometrics and Economic Theory in the 20th Century:
The Ragner Frisch Centennial Symposium. Cambridge: Cambridge University
Press.
Pesaran, M. H., Y. Shin, and R. J. Smith (2001) Bounds Testing
Approaches to the Analysis of Level Relationships. Journal of Applied
Econometrics 16:3, 289-326.
Pew Centre on Global Climate Change and the National Commission on
Energy Policy.
Qureshi, S. K. and A. H. Shah (1992) A Critical Review of Rural
Credit Policy in Pakistan. The Pakistan Development Review 31:4 871801.
Shah, M., K. Khan, H. Jehanzeb and Z. Khan (2008) Impact Of
Agricultural Credit On Farm Productivity And Income Of Farmers In
Mountainous Agriculture In Northern Pakistan: A Case Study of Selected
Villages in District Chitral. Sarhad. J. Agri 24:4.
Shahbaz, ML and M. Feridun (2011) Electricity Consumption and
Economic Growth Empirical Evidence from Pakistan. Qual Quant.
Shahbaz, M, H. H. Lean, and A. Farooq (2013) Natural Gas
Consumption and Economic Growth in Pakistan. Renewable and Sustainable
Energy Reviews 18, 8794.
Solow, R. M. (1956) A Contribution to the Theory of Economic
Growth. The Quarterly Journal of Economics, 65-94.
Sorrell, S. (2009) Jevons' Paradox revisited: The Evidence for
Backfire from Improved Energy Efficiency. Energy Policy 37, 2310-2317.
Van Zon and Yetkiner (2003) An Endogenous Growth Model with
Embodied Energy-saving Technical Change. Resource and Energy Economics
25, 81-103.
Wei, T. (2007) Impact of Energy Efficiency Gains on Output and
Energy Use with Cobb-Douglas Production Function. Energy Policy
35,2023-2030.
Worrell, E., L. Price, and LBNL (2001) Pew Centre/NCEP 10-50.
Improving Industrial Energy Efficiency in the U.S.: Technologies and
Policies for 2010 to 2050. Workshop proceedings. "The 10-50
Solution: Technologies and Policies for a Low-Carbon Future."
Yang, H. Y. (2000) A Note on the Causal Relationship Between Energy
and GDP in Taiwan. Energy Economics 22:3, 309-317.
Yuan, J., C. Zhao, S. Yu, and Z. Hu (2007) Electricity Consumption
and Economic Growth in China: Cointegration and Co-feature Analysis.
Energy Economics 29:6, 1179-1191.
Muhammad Zahir Faridi, PhD <zahirfaridi@bzu.edu.pk> is
Assistant Professor at Bahauddin Zakariya University. Multan, Pakistan.
Ghulam Murtaza <GM.QAUI@gmail.com, ghulammurtaza_14@pide.org.pk>
is Visiting Lecturer at Bahauddin Zakariya University, Multan, Pakistan.
Authors' Note: The authors gratefully acknowledge the comments
and suggestions received from Mahmood Khalid, PIDE at the 29th AGM and
Conference, PSDE.
Table 1
Descriptive Statistics of Variables
Variables Mean Std. Dev. Skewness
AGRI 587531.9 737717.6 1.70
IR 9.633333 5.732839 1.87
RGDP 1507061 1991864 1.07
RGFCF 8910.988 5118.798 0.76
TEC 32961.79 22153.06 0.40
TELF 31.31373 8.480152 0.26
TGC 550732.2 371132 0.78
TOC 10465494 5656927 -0.00
ACRDT 43420 66478 2.00
Variables Kurtosis J.Bera Prob.
AGRI 5.41 28.33 0.00
IR 7.08 50.07 0.00
RGDP 2.38 8.16 0.01
RGFCF 2.87 3.81 0.14
TEC 1.96 2.77 0.24
TELF 1.98 2.13 0.34
TGC 2.41 4.53 0.10
TOC 1.44 3.93 0.13
ACRDT 5.85 39.39 0.00
Source: Authors' calculations.
Table 2
Results of ADF Test
ADF Statistic ADF (With
(At Level) First Difference)
Trend Trend
and and Order of
Variables Intercept Intercept Intercept Intercept Integration
IR -3.252 -3.394 -- -- I(0)
ACRDT 3.503 2.740 -1.848 -2.754 I(1)
RGDP 0.648 -1.289 -5.966 -6.346 I(1)
TELF 1.728 -0.477 -7.827 -8.092 I(1)
RGFCF -0.602 -3.344 -- -- I(0)
TOC -0.993 -3.568 -- -- I(0)
TGC 1.414 -2.831 -3.783 -2.948 I(1)
TEC -2.076 -3.229 -- -- I(0)
AGRI 3.503 2.740 -1.848 -3.754 I(1)
Note: Results are based on authors' calculations.
Table 3
Results of Bound Test for Co-integration
F-statistic Upper Bound
Equations Calculated Critical Value Conclusion
Model-1 Equation (6) 7.42 4.90 (99%) Co-integration
RGDP / RGFCF, TELF, [0.0002] exists
TOC, TGC, TEC, IR
Model-2 Equation (7) 13.51 4.90 (99%) Co-integration
ARGI / RGFCF, TELF, [0.000] exists
TOC, TGC, TEC, ACRDT
Note: Computed F-statistic: 7.42 and 13.51 (Significant at 1 percent
marginal values). Critical Values at k = 7-1=6 and k =7-1=6 are
cited from Pesaran, et al. (1999), Table Cl (V), Case V:
Unrestricted intercept and Unrestricted trend. The numbers in
parenthesis shows the probabilities of F-statistic.
Table 4
Long-run Results of Disaggregate Energy Consumption and Economic
Growth
Estimated Long Run Coefficients using the ARDL Approach
ARDL(1,0,2,0,1,2,1) selected based on Schwarz Bayesian Criterion
Dependent Variable is RGDP
Regressor Coefficient Standard Error T-Ratio [Prob]
RGFCF 604.54 332.51 1.81 [.083]
TELF 588561 523156 1.12 [.273]
TOC .90 .29 3.00 [.007]
TGC 15.63 6.30 2.47 [.021]
TEC -346.85 157.78 -2.19 [.039]
IR -69002 60625.9 -1.13 [.267]
C -1.17 9592168 -1.22 [ 235]
T -779741 351826.6 -2.21 [.037]
Note: Results are based on Authors' calculations using Microfit 4.1.
Table 5
Short Run Estimates of Disaggregate Energy Consumption on Economic
Growth
ARDL (1,0,2,0,1,2,1) selected based on Schwarz Bayesian Criterion
Dependent variable is dRGDP
Regressor Coefficient Standard Error T-Ratio[Prob]
dRGFCF 229.9852 87.6733 2.6232[.014]
dTELF 153071.4 101752.5 1.5044[.145]
dTELF1 -205340.1 101328.0 -2.0265[.053]
dTOC .34272 .077428 4.4263[.000]
dTGC 10.8104 2.3522 4.5958[.000]
dTEC -92.0338 52.6229 -1.7489[.092]
dTEC1 -152.8186 54.7032 -2.7936[.010]
dIR 2370.6 16643.9 -14243[.888]
dC -4460483 3357701 -1.3284[.196]
dT -296635.0 113131.6 -2.6220[.014]
ecm(-1) -.38043 .11781 -3.2290[.003]
ectn = RGDP - 604.54*RGFCF - 588561.3*TELF - .90*TOC - 15.63*TGC +
346.8594*TEC + 69002.1*IR + 1.17E*C + 779741.9*T
R-Squared .76189 R-Bar-Squared .61036
DW-statistic 2.3488 F-stat. F(10, 26) 7.0393[.000]
Note: Results are based on Authors' calculations using Microfit 4.1.
Table 6
Long-Run Estimates of Disaggregate Energy Consumption and
Agricultural Output
ARDL(1,2,0,2,1,2,2) selected based on Schwarz Bayesian Criterion
Dependent Variable is AGRI
Regressor Coefficient Standard Error T-Ratio[Prob]
RGFCF 8.92 23.65 .377[.710]
TELF 3033 12712 .238[.814]
TOC .054 .017 3.114[.006]
TGC 1.81 .47 3.817[.001]
TEC -9.41 8.23 -1.142[.267]
ACRDT 8.83 .95 9.260[.000]
C -208966 231783 -.901[.379]
T -33472 11111 -3.012[.007]
Note: Results are based on Authors' calculations using Microfit 4.1.
Table 7
Short Run Effects of Disaggregate Energy Consumption on Agricultural
Output
Error Correction Representation for the Selected ARDL Model
ARDL(1,2,0,2,1,2,2) selected based on Schwarz Bayesian Criterion
Dependent Variable is dAGRI
Regressor Coefficient Standard Error T-Ratio[Prob]
dRGFCF -1.03 7.19 -.14[.887]
dRGFCF1 -17.82 7.89 -2.25[.033]
dTELF 1409.2 5927 .23[.814]
dTOC .007 .006 1.15[.259]
dTOC1 -.016 .007 -2.07[.049]
dTGC .104 .18 55[.582]
dTEC -6.78 5.71 -1.18[.247]
dTECl -23.28 5.10 -4.55[.000]
dACRDT .769 1.19 64[.527]
dACRDT1 -1.92 1.17 -1,63[. 114]
dC -97067 109958 -.88[.386]
dT -15548 5548.90 -2.80[.010]
ecm(-1) -.464 .102 -4.52[.000]
ecm = AGRI -8.9289*RGFCF -3033.7*TELF -.054952*TOC -1.8115*TGC
+9.4111*TEC -8.8307*ACRDT + 208966.6*C + 33472.2T
R-Squared .98 R-Bar-Squared .97
DW-statistic 1.83 F-stat. F(12, 24) 121.65[.000]
Note: Results are based on Authors' calculations using Microfit 4.
Table 8
Diagnostic Test
Diagnostic Tests of Model-1 [RGDP | RGFCF, TELF, TEC, TGC, TOC, IR]
Test Statistics LM Version F Version
A:Serial Correlation*CHSQ (1) =1.6304[.202] F(1, 21)= .96801 [.336]
B:Functional Form *CHSQ(1) =3.6478[.066] F(1, 21)= 2.2968[.145]
C:Normality CHSQ (2) =2.1778[.337] Not applicable
D:Heteroscedasticity*CHSQ (1) =.36585[.545] F(1, 35)= .34953[.558]
Source: Authors' calculation using Microfit 4.1.
Table 9
Diagnostic Test
Diagnostic Tests of Model-2 [AGRI | RGFCF, TELF, TEC, TGC, TOC, ACRDT]
Test Statistics LM Version F Version *
A:Serial Correlation*CHSQ (1) = .53399[.465] F(1, 18)= ,26358[.614]
B:FunctionaI Form*CHSQ (1) = 2.5889[.118] F(1, 18)= 1.3542[.260]
C:Normality*CHSQ (2) = 2.4167[.299] Not applicable
D:Heteroscedasticity*CHSQ (1) = .46974[.493] F(1, 35)= .45007[.507]
Source: Authors' calculation using Microfit 4.1.
