Economic development and the energy consumption nexus in developing countries: evidence from five South Asian countries

: This paper investigates the relationship between energy use and economic development in five South-Asian


Introduction
Energy is a crucial component of the infrastructure needed for economic growth.The demand for electricity is widespread throughout all economies, homes, and businesses.In addition, electricity is essential for many processes, including the development of agriculture.Since the early nineteen-seventies, electricity usage and income have had a significant positive correlation (Alberini et al. 2011).By analyzing datasets between 1947 and 1974 in the United States, it is found that there is a uni-directional causal connection between energy and GNP (Kalyoncu et al. 2013).Energy consumption that uses electricity is referred to as electricity consumption (Bozkaya 2022).Additionally, it facilitates consistent societal advancement and simplifies long-term economic growth (Yıldırım Durmuş et al. 2019).Once more, the core of any economy is the attraction and maintenance of large inflows of foreign investment through FDI and other international trade procedures.International trade positively affects economic growth and may increase the power demand (Siddika and Ahmad 2022).
Among the South Asian countries, Bangladesh's per capita energy usage is deemed modest.From 1990 to 2014, per capita electricity usage ranged from between 0.05 and 0.31 MWh in Bangladesh.In July 2018, a survey by the BPDB (Bangladesh Power Development Board) revealed that 90% of people have access to power (BPDB 2020).Despite being a country fighting poverty and other development challenges (Ara et al. 2015), Bangladesh's power sector has flourished since its independence.
In India, the electricity sector has a total installed capacity of 228.7 GW.This is insufficient to satisfy the internal demand (Garg et al. 2015).Despite having excess power-generating capability from resources such as biofuels, waste, and nuclear, it lacks enough infrastructure to distribute electricity to someone in need.India is both the world's third-largest manufacturer and user of electricity.By contrast, Pakistan's electricity generation is mostly based on petro-leum, coal, gasoline, biofuels, and nuclear power.Per capita, electricity usage is quite modest, at 0.48 megawatts in 2014.Pakistan's power sector is still in its early stages.For years, harmonizing the country's supply and demand for power has remained a mostly unsolvable issue.As a result, the country faced enormous challenges in modernizing its electrical supply infrastructure.In the case of Sri Lanka, the major electricity generation is thermal and hydropower energy, with some solar and wind power being used in the early stages of development.Even though potential locations are being discovered, the state grid's power generation system does not use additional power sources like nuclear, geothermal, solar thermal, peat, or wave power.In 2014, the overall electricity usage in Sri Lanka exceeded 11.04 terawatt hours in 2014, with per capita consumption at 0.53 megawatts.
Nepal's power usage has been significantly growing for many years.Consumer growth is accelerating as a result of the development of many firms and the use of electrical equipment.Nepal's average capita consumption has remained practically constant for the last twenty years, although the consumption of neighboring countries such as India, Sri Lanka, and Pakistan are rising.
Given the energy generation and consumption scenario, the present study is aimed at the examination of the relationships between economic growth (GDP), foreign direct investment (FDI), energy consumption (EC), and global trade (TR) in a panel of five South Asian nations from 1990 to 2014, namely Pakistan, India, Bangladesh, Sri Lanka, and Nepal.Although there are lot of studies focused on this issue, very few studies cover South Asian countries.This study contributes to the literature by generating new evidence in the context of South Asian countries.Moreover, the existing literature as discussed in the literature review section shows mixed findings relating energy consumption to economic development.
The following sections, section two and three of the article, discuss the literature review and the methodology, respectively.After that, the results of the study are presented and interpreted, followed by a brief discussion.The article ends with some recommendations and concluding remarks.
1. Literature review Mozumder and Marathe (2007) conducted a Granger causality study to explore the relationship between GDP and electricity use.They found that the GDP seemed to have an impact on power use but that there was no relationship between the two.Cheng-Lang et al. (2010), however, found a bidirectional causal connection between electricity use in industry, real GDP, and Total Electricity Usage in Taiwan from 1971Taiwan from -2006. .To determine the causation between Bangladesh's GDP per capita and power usage, Mozumder and Marathe (2007) used the vector and cointegration error correction model.Their finding shows a correlation between GDP per capita and electricity use, but only in one direction.Numerous researchers from all corners of the globe have looked at the correlation between electricity use and economic growth.In a recent study, Jha (2021) utilized the same technique as Jumbe (2004) and found that the increase in GDP has an effect on the amount of electricity used as well as employment in the short term.However, Bozkaya et al. (2022) found a mixed relationship between energy consumption and economic growth.Mozumder and Marathe (2007) utilized an investigation that employed a technique known as Granger causality to determine which way the correlation runs between total annual power usage and GDP.Electricity use was affected by the GDP, and there was no correlation between GDP and electricity consumption discovered by him.The total power used, industrial electricity used, and real GDP in Taiwan were shown to be interconnected in two ways by Cheng-Lang et al. (2010), who analyzed data ranging over 35 years between 1971 and 2006.Employing cointegration and vector error correction, Mozumder and Marathe (2007) investigated the link between Bangladesh's per capita power usage and per unit GDP.The result shows that GDP per capita is causally related to per unit electricity use but only in one direction.
In order to analyze highly predictive economic growth conclusions, Shahbaz et al. ( 2017) additionally considered the general connection between electricity use and the price of oil.The information is broken down into income and OECD, and geographical categories using long-run estimates of parameters, panel cointegration, and Pool Mean Group analyses of the short-term and long-term connections and cointegration of the factors.The empirical findings point to the cointegration of the variables.
There are undeniable feedback effects between the consumption of energy and the development of wealth, including oil price and economic expansion.These numbers show that, despite oil prices, rising nations heavily rely upon electricity usage to increase productivity.Saidi et al. (2017) evaluated the correlation between economic growth and energy use across a sample of fifty-three nations using data ranging from 1990 to 2014.The findings demonstrate a lengthy symbiotic link between power use and economic expansion.There exists a short-term in addition to a long-term Granger causality between foreign direct investment and economic growth and also between energy use and economic development, as shown by the results of the causality research conducted by such a worldwide panel.Identical outcomes are displayed separately for the American nations.There exists a correlation between economic development and energy consumption in Africa and the Middle East in both the short-term and the long-term.Moreover, In both the short term and long term, there is linear causation between energy use and economic development in European nations.Both the short-term and the long-term causality relationship between FDI and economic development is demonstrated for nations in Europe, Africa, and the Middle East.
Another paper from Jha (2021) looks at the link between GDP, export and power usage for a selection of Middle Eastern nations.When looking at the whole panel, we discover statistically significant feedback effects between some of these components.Sacko (2004) found that rising energy use significantly affects economic expansion.Moreover, using yearly data from 1950-1951to 1996-1997for India, Ghosh (2002) attempts to evaluate the Granger causation between per capita power usage and GDP per capita for India.This analysis reveals a lack of a long-term stability relationship between the variables, despite the existence of uni-directional Granger causation from economic growth to energy usage with no feedback effect.In another study, Attinay and Karagol (2005) analyzed the causality relationship between electricity use and gross domestic product in Turkey from 1950 to 2000.Both the Dolado-Lutkepohl test, which uses VARs in levels, and the standard Granger causality test, which employs discriminant function data, were subjected to the Granger non-causality test for comparison in the study.A strong correlation between electricity use and income was discovered in both tests.Wolde-Rufael (2006) uses a newly developed cointegration test provided by Pesaran et al. (2001) and uses data from seventeen African countries covering the years 1971-2001.They examine the long-term and causal relationship between real GDP per capita and electricity consumption per capita.Only twelve nations showed Granger causality, and only nine showed a long-run link between real GDP per capita and electricity consumption per capita.There was positive uni-directional causation between real GDP per capita and electricity consumption per capita in six countries, negative bidirectional causality in three countries, and no correlation at all in the other three countries.
Using Granger causality and ECM tests, Bäker and Goodall (2020) looked into the connection between income, energy usage, FDI, and population.They used information from 1970 to 2005 and established a bidirectional causal link between short-term energy usage, income and FDI.By contrast, Chandran et al. (2010) used ARDL analysis to examine the causal link between the variables and reached the same conclusion.Ibrahiem (2015) investigated the link between renewable energy usage, FDI and Egyptian economic growth.The research used an Auto Regressive Distributed Lag (ARDL) bound testing strategy on time series data spanning 1980-2011.The empirical results show that the research variables are co-integrated, indicating a long-lasting link between them.Furthermore, the Granger causality test demonstrates a two-way link between economic growth and the use of renewable energy sources, as well as a one-way causal association between foreign direct investment and economic development.
Bäker and Goodall (2020) employed a multivariate approach to establish the causal connection between energy usage, economic growth, relative pricing, financial development (FD) and foreign investment in Malaysia from 1972 to 2009.The variables perfectly correlate with the limits test and the Johansen-Juselius cointegration test.The data demonstrated that in both the short and long term, Granger's power consumption and economic expansion are mutually causal.By contrast, Bekhet and Othman (2011) used a consumer price index (CPI), electricity consumption (EC), foreign direct investment (FDI), and GDP (gross domestic product) from 1971 to 2009 using the VECM model.Cointegration analysis showed that all variables are related over the long term and are all co-integrated.Furthermore, strong long-run causation from power use to FDI, GDP growth, and inflation was discovered.According to the results, energy consumption is both a key factor in determining Malaysia's economic development and a powerful tool for enforcing the government's energy-saving goals.Long-term economic development is dependent on a reliable energy supply, which policymakers must acknowledge.
Using data from 1960 until 2011, Bento and Moutinho (2016) used an autoregressive distributed lag (ARDL) bounds testing approach for Italy.The findings of the study show that there is a long-run uni-directional Granger causation link between GDP per capita and the production of renewable energy per capita and between non-renewable power generation per capita and the production of renewable energy per capita.Moreover, Rafindadi and Ozturk (2016) looked at how past energy crises in Japan have affected the country's short-and long-run capital, imports, exports, economic expansion, and financial development decisions.The research found that for every percentage point increase in GDP, financial development, imports, and exports in Japan, electricity consumption drops by 0.2429, 0.504, 0.092 and 0.219%.However, it was found that the capital used less energy in every tangible aspect.The research found that an increase of 1% in Japan's growth in the economy, financial development, imports, and exports would increase the country's electrical difficulties by 0.2031, 0.584, 0.0521 and 0.22109%, respectively.Another study by Katırcıoğlu et al. (2016) showed that Canadian energy conservation laws are probably going to hurt output and international trade.In Malaysia, Bhatti et al. (2019) discovered evidence of a uni-directional causal link between power use and exports.In a related analysis, Bhatti et al. (2019) found no causative link between exports and energy generation but found a uni-directional causality linkage between power generation and GDP growth.Using information spanning 1970-2008, claims that export growth in Malaysia was spurred by economic expansion have been disproved.
To summarise, there are mixed findings in existing research.Studies conducted in both developed and developing countries reveal that energy consumption and growth have a strong relationship.However, the direction of the dependency varies.Moreover, in the case of South Asian countries, such studies are very limited in number.Eventually, a new study using panel data may contribute to the existing body of knowledge and guide policymakers.

Data sources
This research spans the years 1990-2014 and focuses on five South Asian countries that are part of the South Asian Association for Regional Cooperation (SAARC): Bangladesh, India, Nepal, Pakistan, and Sri Lanka.Furthermore, we consider Electricity consumption as the dependent variable, whereas foreign direct investment, the GDP per capita, and international trade are the control variables in this estimation.
We extracted all data from The World Bank's World Development Indicators and International Trade (Average of Exports of Goods and Services as a Percentage of GDP) publications provide data on five economic indicators: electricity consumption (kWh per capita), net inflows (percent of GDP), foreign direct investment, GDP per capita (constant 2010 US$), and imports of goods and services.

WDI TR
The term "international trade" (TR) refers to the combined value of a nation's exports of goods and services and its imports of goods and services (percent of GDP).The consolidated worth of kinds of services and goods that a country receives from the entire world is represented by its total imports, whereas the total worth of taxable goods and services that a country exports to the entire world is represented by its export earnings.WDI

Model specification
During the research period, each of the n elements or participants in the panel data set has T observations and a rating of 1.As a result, in the data set, there are n ⋅ T total observations.This research examines the factors of electricity consumption in India, Bangladesh, Pakistan, Nepal, and Sri Lanka from 1990 to 2014 across five SAARC nations: India, Bangladesh, Pakistan, Nepal, and Sri Lanka.The data set is referred to as being balanced if all the data across countries and times are available for the study.However, in this research, some cross-sectional unit observations are overlooked, which is why this research deals with unbalanced data.Total observations are n ⋅ T. whereas n = 5 countries and T = 25 time periods.Thus, the total number of observations should be 5 ⋅ 25 = 125.However, unbalanced data suggests a total of 121 observations in the study.In order to assess the influence of variables on energy consumption, the study considers the growth in the economy, foreign direct investment, and international trade statistics for a panel of five South Asian nations.
The variable data is logarithmically converted.Following modification, the model is subjected to three estimation methods (the pooled regression method, the fixed effects method, and the random effects method).

Panel estimation techniques
The Pooled Regression Method [PRM]: When independent time series are combined with data from many persons, a pooled model is produced.When OLS [ordinary least square] is applied to a pooled model, pooled least square estimation is performed.Whenever data are prior homogeneous, it is employed.To estimate the constant slope and intercept, the pooled regression method analyses the model's data, regardless of the time or cross-sectional unit.In other circumstances, it was anticipated that all countries and years would have the same slope and intercept.The random and fixed effects are ignored in this combined estimate.
In the study, the pooled model is Equation 1.Note that in Equation 1, β 0 is the intercept, β 1 is the slope (coefficient or parameter estimate) of economic growth, β 2 is the slope of foreign direct investment, β 3 is the slope of international trade, and ε i is the error term.
The explanation of the variables is the same as before.In the model, the i th denotes the i th countries for the period.The independent and dependent variables are varied over time and countries, but the intercept and slope coefficient is the same for all the countries and time and is assumed to ignore individual heterogeneity.If individual effect α i (cross-sectional or time -specific effect) does not exist (α i = 0), ordinary least squares (OLS) produce efficiency, and the parameters are unbiased and consistent.
OLS consists of five key assumptions (Ababneh 2020; Kennedy 2008).a) The dependent variable is expressed as just a linear function of the response variable and the error (disturbance) term, as required by linearity.
b) When a disturbance is said to be homogeneous, it means that its expected value is zero or that it is not connected to any regressors.c) Disturbances are unrelated to one another and have the same variance (3.a homoskedasticity) (3.b non-autocorrelation).
d) The independent variable's observations are stable in repeated samples without measurement mistakes rather than stochastic.
e) The full rank assumption asserts that independent variables do not have a perfect linear connection (no multicollinearity).
For pooled least square estimation, the error term assumptions are: a) Zero mean of the error term E(ε it ) = 0.
b) The variance of the error terms is the same i.e. homoscedasticity var.(ε it ) = δ 2 .c) Uncorrelated error terms cov(ε it ε js ) = 0, where, i ≠ j, and t ≠ s. d) Uncorrelated errors term and explanatory variables cov(ε it x it ) = 0.It is assumed that there exists unobserved heterogeneity among the individuals detected by α i .The main query is whether the regressors and the individual-specific effects α i are correlates.
We have a fixed effects model if they are correlated.We have a random effects model if they are not correlated.
The Fixed Effects Method [FEM]: Individual variations in intercepts are studied in a fixed effect model, considering uniform slopes and variance.Individual particular effects are regarded as an element of the intercept because they are time-invariant, α i maybe related to other regressors.This FE model is calculated by least squares dummy variable (LSDV) regression (OLS with a set of dummies).
The FE model permits correlations between the individual-specific effects α i and the regressors x.This research involves intercepts as α i .Every individual has a unique intercept term and identical slope parameters. (2) This study can retrieve the particular individual impacts after assessment as: To put it another way, the leftover variation of the dependent variable that the regressors are unable to explain is called individual-specific effects.In the regressors x, time dummies can be included.
The fixed effects model for the study is: if there is a constant slope coefficient across all individuals and time.However, the intercept term β 0i varies depending on the country, not depending on the individuals' different time periods.
The individual intercept term, often known as the FE, captures the individual heterogeneity (the unique characteristics of each country regardless of time).

The Random Effects Method [REM]:
Since the premise of a random effect model is that individual effect (heterogeneity) is unrelated to any regressor, the study assumes error variance on a group-by-group basis (or times).Therefore, is a part of the overall erratic heterogeneity or the combined error term.It is for this reason that a random effect model is also referred to as an error component model.There is no variation in the regressors' intercepts or slopes when compared.Individual specific errors, not intercepts, distinguish individuals (or time intervals) from one another.
The random effects model for the study is: where If the composite error term is defined as v it = α i + ε it , then the above equation may be rewritten as follows: All of the fixed effects assertions as well as the extra condition that is independent of all independent variables across all time periods make up an ideal random effect's assumption.Use first differencing or fixed effects initially if it is believed that the unobserved impact α i is associated with any explanatory factors.
Hausman specification Test: The Durbin-Wu-Hausman test (sometimes referred to as the Hausman specification test) is a statistical hypothesis test in econometrics that is named after James Durbin, De-Min Wu, and Jerry A. Hausman.The test assesses the consistency of an estimate in comparison to an alternative, less efficient estimator that is acknowledged to be consistent.This helps in determining whether or not a given statistical model adequately describes the data.The Hausman test is used to choose the appropriate effect between the RE and FE model.It's being used to evaluate the estimated coefficient of the FE model to that of the RE model.
The hypothesis for the Hausman is: If the probability of the cross-sectional chi-square is more than 5% level, we do not reject the Null hypothesis.This means that the Random effect estimators would be proper to explain the model.In the case of a chi-square value of less than 5%, we reject the Null hypothesis, which stands for using fixed effect estimators.

Results
Table 2.The results of all pooled least square, and fixed and random effect models (dependent Variable: L(EC)) Tabela 2. Wyniki wszystkich połączonych modeli najmniejszych kwadratów oraz modeli z efektami stałymi i losowymi (zmienna zależna: L(EC)) Table 2 presents the results of all pooled least square, fixed and random effect models where dependent variable: LEC.Here, the slope coefficient of International Trade Ln(TR) is negative (-7.638074) which implies that as international trade increased by 1%, electricity consumption decreased by 7.63%, but it is extremely significant at a significance level compared with fewer than 1%.However, from the above result, it's clear that Electricity Consumption Ln(EC) is highly influenced by the Foreign Direct Investment Ln(FDI) and the Economic Growth Ln(GDP) with a positive slope coefficients of 160.2319 and 0.081013, respectively, as well as being very significant at the 1% significance level.
In the fixed effect model, all of the variables' coefficients have positive values that are statistically significant at the 1% level or below.The coefficient value of Ln(GDP), Ln(FDI), Ln(TR) are 0.186621, 23.82336, and 5.662425, respectively.This indicates that electricity consumption (EC) increases by 0.18, 23.82 and 5.66% as GDP, FDI and TR are increased by 1%.Although all the coefficients of the variables shows a positive relationship, Electricity Consumption Ln(EC) is more influenced by Ln(FDI).
The results of the random effect model are shown in Table 2, where all coefficients of the variables are positive attributes, and all of them are statistically significant at the 1% level.The coefficient values of Ln(GDP), Ln(FDI), Ln(TR) are 0.172996, 33.69695 and 4.187515, respectively.However, the coefficient values of all variables mean that a 1% increase in GDP trends increases the EC by 0.17%.Furthermore, if FDI and TR increase by 1%, then EC increases by 33.69% and 4.18%, respectively.Foreign Direct Ln(FDI) is a highly influencing matter of electricity consumption.
From the results in Table 2, it is clear that Electricity Consumption Ln(EC) is highly influenced by the Foreign Direct Investment Ln(FDI) and the Economic Growth Ln(GDP) with a positive slope.

Hausman Specification Test
To investigate the appropriate model between the fixed and random effect models, the Hausment Test is commonly used in empirical studies.Here we consider: Null hypothesis (H 0 ): the random effect model is appropriate, and Alternative hypothesis (H A ): the random effect model is not appropriate.Table 3 reports the results of the Hausman specification test.The test reports that the p-value for the chi-square statistic is 0.00, which is less than 5%; so we can't accept the null hypothesis indicating rejection of the of the Hausman test.This means that we accept the alternative hypothesis.As a result, the fixed effect model is the most appropriate because it consistently and effectively explains the variables.
Thus, we conclude that according to the fixed effect model, FDI is the most influential indicator to increase EC; however, the rest of the model supports the fixed effect model.

Discussion
The average of the pooled OLS results of the study shows that the SAARC region's energy consumption is significantly and favorably impacted by GDP per capita, FDI, and international trade (TR).More specifically, Electricity Consumption Ln(EC) is highly influenced by the Foreign Direct Investment Ln(FDI) and the Economic Growth Ln(GDP).This finding is common in the context of developing countries.For instance, in the case of Bangladesh, Ahamed (2014) argued how nuclear energy production can contribute to growth.Many other studies also support this study.For instance, Rabab Mudakkar et al. (2013) found a significant correlation between EC and FDI for SAARC countries in both the short and long term.Alam et al. (2015) investigate that after accelerating the FDI inflow, GDP per capita increases and thus energy demand increases rapidly over time.Furthermore, Mudakkar et al. (2013), Khan et al. (2014), Mozahid et al. (2022) and many other empirical studies have investigated a positive relation EC, GDP, FDI and TR but FDI inflow increases the energy demand radically.The uniqueness of the findings of the study relies on the context of developing countries.Specifically, in the SAARC countries, such findings as those similar to previous studies making the existing body of knowledge regarding energy consumption and economic development nexus.

Recommendations
1.As FDI is an influencing matter to boost electricity consumption, more emphasis should be given to promoting this sector.For this, the rate of tariff-based FDI should be simple to attract foreigners.2. To foster international trade, trade creation among the South Asia region should be built up. 3.As has already been mentioned, the study's findings emphasize that increased electricity consumption is necessary for South Asia to experience higher economic growth.As a result, the government should give high priority to issues relating to proper electricity distribution systems and management solutions in addition to power generation in its short-and medium-term plans.4. The power structure may change as a result of the use of alternative and renewable energy sources.The SAARC region's electricity crisis has a lot of room for improvement.The energy that the sun provides (such as solar energy) is considerably more than what is currently needed in terms of electricity.There is also a lot of potential in the wind, waves, and tides.It is important to realize that traditional energy sources like gas and fuel are running out, whereas renewable energy sources might be one of the most important sources of electricity in the future.As a result of the above findings, the focus should be placed on producing more power and increasing investment.The answer to the issue of whether power consumption alone might encourage economic expansion; the answer is definitely no since one of the determining elements is the utilization of power.The government should promote a business-friendly climate in addition to increasing electricity production to attract more foreign and domestic firms.Only in such a scenario will having more power result in more economic activity, otherwise it would be expensive.In this context, the government may implement legislative measures to boost power production and entice domestic and international investors to invest in energy and other industries.

Concluding remarks
The goal of this study was to inspect the association between FDI, GDP per capita, and international trade (TR) on energy consumption (EC) of SAARC nations for the data from 1990 to 2014.In order to evaluate the energy-growth nexus, which includes GDP per capita, FDI, and international trade (TR) metrics in the SAARC area, this study has employed a variety of panel data methodologies, including pooled OLS, a fixed effects model, and a random effects model.Additionally, the Hausman test, a statistical test for model specification, is used to compare several options, such as whether the fixed effect model is superior to the random effect model.The average of the pooled OLS results show that the SAARC region's energy consumption is significantly and favorably impacted by GDP per capita, FDI, and international trade (TR).
The majority of the exogenous variables generally have a considerable impact on the region of SAARC's energy consumption, according to the analysis of a fixed effect model.The findings support the region's high prevalence of energy-led growth, energy-led foreign direct investment, and energy-led total returns.This suggests that increased power consumption causes more investment, which results in increased economic growth.The study's random effect model indicates that every variable has a significant impact on the region's power consumption, meaning that SAARC nations will face more challenging difficulties than before and will need robust policy measures to protect themselves.Our findings show that FDI had a significant impact upon power consumption and the area of SAARC's energy demand, resulting in the entry of new technology and an increase in both economic growth and energy consumption.The fixed effect model is regarded as the optimum model for examining the relationship between variables, according to a model specification test.
Słowa kluczowe: zużycie energii elektrycznej, energia, PKB per capita, BIZ, handel międzynarodowy, dane panelowe, region SAARC in the composite error in each time period, the v it serially correlated across time.The errors take the following assumptions: a) Zero mean of the error term [E(α i ) = 0].b) The variance of the error terms are the same i.e. homoscedasticity [var(α i ) = δ 2 ]. c) Uncorrelated error terms [cov(α i , α j ) = 0] where i ≠ j.

Table 1 .
Variable description It includes the reinvestment of profits, equity capital, as well as other short-and long-term capital.The World Bank provides statistics as a percentage of GDP.