Effects of Energy Consumption, GDP and Microfinance on Sustainable Poverty Reduction: Evidence from a Developing Economy

Abstract

This study examines the combined and individual effects of gross domestic product (GDP), microfinance loan disbursement, per capita power consumption, and total energy consumption on poverty rate in Bangladesh by using annual time series data over the period of 1995–2022. This study determines the direction of causality by applying the Toda–Yamamoto (T–Y) procedure of the Granger causality test in the vector auto regression (VAR) model given the presence of a mixed order of integration of I(1) and I(2). The results of the stationary tests show that all variables except poverty rate are stationary at the I(1) order of integration, according to both the augmented Dicky–Fuller (ADF) and Phillips–Perron (PP) methods, while the poverty rate shows stationarity at the I(2) order in both methods. The T–Y empirical test result shows that the four independent variables combined affect the poverty rate significantly. Power consumption affects both GDP and microfinance and they have bi-directional causality relationship with each other. Our study shows that GDP and total energy consumption individually contribute to poverty reduction. Based on the findings, it is imperative that national policy makers place a greater emphasis on increased domestic production and the proper utilization of power and energy to reduce poverty rates. Policy implications may include strategies to promote sustainable energy development, improve energy efficiency, and provide equitable access to energy services.

1. Introduction

The relationship among GDP growth, energy consumption, and microfinance is crucial in determining how emerging nations structure their economies, especially when it comes to reducing poverty. Energy consumption rises with economic growth, promoting industrial activity and economic development [1]. The difficulty, though, is making sure that this increase results in real poverty reduction. Growth in the GDP, a major measure of economic success, frequently denotes a rise in a country’s wealth. However, there is disagreement regarding how much of this expansion affects a population’s poorer portions [2]. In addition to increased energy consumption, increasing energy efficiency and moderation of energy use also contribute to GDP growth [3]. Unfortunately, the Energy Division of Bangladesh has yet to take any effective measures for energy efficiency and the moderation of energy consumption.

The principal energy sources in Bangladesh are natural gas, petroleum, coal and biomass, all of which produce a higher amount of greenhouse gases (GHGs) than most other sources. Only 3% of total energy demand is met by renewable energies (mainly from solar and hydro). With increased energy consumption, GHG emissions are increasing at a rate of 3.3% per year in the country [4]. Although Bangladesh has one of the lowest per capita GHG emissions, it is listed among the most climate vulnerable countries in the world. Despite Bangladesh’s tiny contribution to global CO₂ emissions, the government has pledged to lessen its influence on the troposphere by employing alternative techniques, which will lower greenhouse gas emissions. In order to reduce GHG emissions, the country has aimed at reducing heavy reliance on fossil fuels, incorporating renewable energy sources into the mix of energy sources used to generate power, as well as implementing cutting edge technology like carbon capture and storage (CCS) with enhanced energy efficiency systems.

In this scenario, microfinance—which offers financial services to those with low incomes—emerges as a crucial tool. Microfinance seeks to empower the economically disadvantaged by providing access to loans, savings, and insurance. This allows people to participate in income-generating activities (IGA) and raise their standard of life [5,6]. Understanding the interaction between macroeconomic variables like GDP growth, energy consumption and microfinance is crucial to comprehending their combined effects on reducing poverty. A positive correlation has been found between greater GDP growth rates and subsequent measures to reduce poverty and increase energy use [7].

This study attempts to investigate how energy consumption, GDP expansion, and microfinance contribute to poverty reduction in a developing nation. More specifically, this paper intends to uncover how these variables interact with one another and contribute to alleviating poverty by analyzing the empirical data. This study offers a thorough analysis that incorporates the effects of GDP growth, energy consumption, and microfinance on poverty alleviation and fills a significant gap in the burgeoning literature by showing the interconnectedness and collective impact of these variables on poverty alleviation. Prior studies have frequently examined these factors in isolation [8,9]. The findings of this study offer valuable insight to suggest policy decisions tailored to the unique challenges and opportunities in developing countries. Moreover, this study employs advanced econometric techniques, such as the Toda–Yamamoto Granger causality test, within a VAR framework, to establish causal relationships among the variables. This robust methodological approach strengthens the evidence supporting policy recommendations by improving the findings’ validity and reliability.

The rest of this study is structured as follows: review of the literature is in Section 2. Data and methods are presented in Section 3. Section 4 presents the results followed by a discussion of those results. The last section concludes the paper and shows the direction for further study.

Many economic analysts believe that energy consumption is a vital factor for economic progress. Having access to enough and dependable energy sources may boost industrial and agricultural output, which would boost the economy. Increased wages and the creation of jobs are often correlated with higher GDP growth rates, and both factors can help to reduce poverty. Financial services, such as modest loans and savings accounts given to low-income people or groups who have little or no access to basic banking services are referred to as microfinance. Bangladesh is basically an agriculture-based economy. In recent years, its industrial activity has increased significantly. For the development of both sectors, it is facing a polycrisis on development. A polycrisis is a collection of interconnected socioeconomic, ecological, cultural, institutional crises that occur on a global scale and whose causes cannot be easily removed [10]. Examples of polycrisis in Bangladesh include the following: climate change, loss of biodiversity, pure water scarcity, natural resource crisis, natural calamities like floods and cyclones, global warming, the proliferation of illegal economic activities, institutional corruption, a debt crisis, financial market instability, economic inequality, Rohingya migration, a lack of public infrastructure and services, a rising cost of living, political instability, the erosion of democratic institutions, widespread health issues linked to pollution or lifestyle diseases, mental health problems, cybersecurity law, cybercrime, geopolitical confrontation, international terrorism, the Russia–Ukraine war etc. All these crises have a direct and indirect impact on poverty alleviation. Another paper has examined the effectiveness of sustainable energy development in an era of geopolitical multicrisis [11]. The term “multicrisis” describes the occurrence of multiple crises at the same time, usually as a result of geopolitical risks. The study reveals that these risks can have a significant impact on government effectiveness, the rule of law, and political stability. Examples of these risks include the water crisis, stagflation, COVID-19, population growth, and wars.

The goal of researching microfinance is to better understand how it may empower people economically by allowing them to develop small businesses, make money, and upgrade their quality of life in general [12]. Knowing the relationship between GDP growth, energy use, and microfinance can help foster sustainable economic growth, and leverage microfinance to benefit the poorest segments of society.

Countries in the developing world encounter several barriers, such as poor infrastructure, economic instability, and massive poverty. Exploring these dynamics within the developing world sheds light on the obstacles and chances for reducing poverty in these areas. The results of these investigations may help to guide decision-makers and global development groups in distributing resources wisely. For example, pinpointing areas where more investment in energy could offer the greatest economic benefits or focusing microfinance programs to enhance socio-economic outcomes which can lead to more focused and enduring efforts to reduce poverty. Microfinance is a loan of a small amount extended to the landless and rural people living in utter poverty. The borrowers do not need to place any asset to the lender as collateral or security. The aim of providing the loan is to make the poor class self-sufficient by starting their own small business or agricultural farms. Later they pay off the loan mainly on a weekly basis through small instalments over a period of six months to one year. But the system is not free from limitations; one limitation being that the interest rate is very high, usually more than 20%. Moreover, in case of any loss in the business or farm for which the loan was disbursed, there is no exemption from the repayment. In some instances, some of the micro-borrowers had to sell their land or other assets to repay the loan, and they became poorer after taking the loan.

According to a recent study, microfinance has increased the nation’s GDP in recent years by 8.9% to 11.9% [13], which may be considered as the paramount success of microfinance. A total of 40.86 million members, including 31.53 million borrowers (or roughly 18.55% of the country’s population), are benefited by 731 Microfinance Institutions (MFI) that hold Microcredit Regulatory Authority (MRA) licenses. The failure of microfinance is usually measured in terms of the loan default rate, which is 2.38% [14], a better situation than the commercial banks’ loan default rate which was about 5.13% in the year 2022. No statistics of landless people was found as a result of microloan repayments. However, the MRA has established a housing fund called “Grihayan Tahabil” to assist those who have become landless. By working with non-governmental organizations across the country, Grihayan Tahabil has built 100,000 homes for them. This achievement was made possible by using 174 MRA-licensed MFIS to channel payments totalling BDT 1090 million [14].

However, exploring the effect of energy use, GDP growth, and microfinance on poverty reduction in developing nations provides concrete evidence, informs policy choices, and deepens our understanding of the complex relationships that influence economic progress and social advancement [14] in these settings. According to Lee and Kim, rising energy consumption is linked to higher GDP growth rates and may have an impact on efforts to reduce poverty [15]. According to Wang et al., by providing financial access to marginalized communities, microfinance initiatives in Asia have reduced poverty significantly [16]. The World Bank estimates that Bangladesh’s GDP growth and increased energy consumption have significantly reduced poverty over the past decade. Many efforts to reduce poverty in Bangladesh have been significantly aided by increasing energy consumption. Those efforts to reduce poverty are aided by rising energy consumption in developed nations, which is positively correlated with higher GDP growth rates [17]. Microfinance programs in developed nations have aided in the empowerment of low-income individuals and the reduction of poverty [18].

To comprehend how energy consumption and GDP growth affect poverty levels in different ways, a sector-specific (manufacturing, agriculture, and service) analysis study may be conducted to enable us to trace which industries have the greatest potential for reducing poverty by investing energy in specific areas. It is presumed that higher levels of energy consumption are causally linked to reductions in poverty rates. Policies aimed at expanding energy access and improving energy efficiency could potentially contribute to poverty alleviation by stimulating economic growth and enhancing living standards. Factors such as geographic location, economic development levels, and energy policy frameworks could influence the effectiveness of energy-related interventions in reducing poverty [19].

Bangladesh has made a significant progress in reducing poverty over the last two decades. The Household Income and Expenditure Survey (HIES) conducted in 2016 revealed that the poverty rate had decreased to 24.3 percent, a significant decline from the 40.0 percent recorded in 2005. At present, the poverty rate is 18.7%, with extreme poverty affecting 5.6% of the population [20]. Figure 1 shows the poverty rates of the country from the financial year 1990–1991 to the year 2021–2022. The figure reveals that the poverty rate is declining sharply since the year 2001–2002.

2. Literature Reviews

Growth in GDP is frequently regarded as a crucial indicator of economic progress. According to Dollar and Kraay, rising average incomes benefit the poor, and economic expansion generally reduces poverty [22]. However, depending on the inclusiveness of growth and the initial levels of inequality, the magnitude of growth can vary [23]. GDP growth does not always result in significant reductions in poverty in developing nations, where economic disparities are frequently pronounced [24].

Microfinance is considered to be a strong tool for reducing poverty because it extends financial services to the underserved. Armendáriz and Morduch emphasize the potential of microfinance to empower individuals with low incomes by facilitating investment in small businesses, enhancing credit access, and encouraging financial inclusion [5]. Clients’ income, consumption, and overall economic well-being have all improved because of microfinance’s positive effects on poverty reduction [25,26].

A complicated dynamic is created when energy consumption, GDP expansion, and microfinance interact. Improved energy access can help microfinance initiatives by enabling micro-entrepreneurs to operate more effectively and expand their businesses [27]. Furthermore, the effects on poverty reduction may be amplified when microfinance is incorporated into broader economic policies that promote energy access and sustainable growth [28].

Empirical research in developing nations suggests a variety of outcomes. For instance, a Nigerian study showed that poverty reduction and economic growth were positively impacted by an increase in energy consumption [29]. Similarly, both financial reform and energy use significantly contributed to the reduction of poverty in Kenya [30]. Both Nigeria and Kenya have a similar economic condition to Bangladesh, in terms of the per capita income that ranges from USD 2100 to USD 2600. Another study examined the energy and poverty nexus in BRICS nations. Although these giant economies are spread over four continents with diverse economic standards, they have shown that energy consumption and GDP growth have significant impact upon poverty reduction [31]. Another study examined the association between microfinance and poverty rate of 1132 microfinance institutions in 57 developing economies. The econometric findings consistently show a substantial and negative correlation between the poverty rate and the number of microloans per capita [32]. However, the success of these strategies is contingent on the socioeconomic conditions of the nation and the implementation of complementary policies that guarantee equitable benefit distribution and inclusive growth. In a recent study, a significant relationship has been identified between energy use and income per capita in the context of Bangladesh [33].

Access to clean and sufficient energy enhances the quality of life for individuals, empowering them to pursue self-sufficiency and entrepreneurial endeavors. It also enables them to participate in activities that yield higher productivity and improved health standards, ultimately leading to increased productivity and higher earning potential [31]. Several studies concur with the research regarding the connection between energy use and poverty alleviation [34]. Significant effects of energy use on Bangladesh’s efforts to reduce poverty have been highlighted in recent research. A study highlights that by facilitating economic activity and raising living standards, expanding access to dependable electricity may significantly reduce poverty [35]. Better health, education, and economic prospects for low-income households have been associated with increased energy access, which promotes inclusive growth [36]. Growth in GDP is a key indicator of economic progress, and its contribution to reducing poverty is especially important in developing nations like Bangladesh [37]. Recent studies and reports focus on employment creation, income distribution, and structural economic reforms as some of the ways GDP growth has contributed to the country’s alleviation of poverty. According to the most recent update from the World Bank, Bangladesh has seen significant economic growth over the past few decades, which has helped reduce poverty. According to the global poverty line, an income below USD 2.15 per day, poverty in Bangladesh decreased from 11.8% in 2010 to 5.0% in 2022. Strong exports in the ready-made garment (RMG) sector, robust remittance inflows, and stable macroeconomic conditions are for the cause of this decrease [38].

To maintain growth and reduce poverty, structural reforms are necessary. The World Bank emphasizes that Bangladesh must address issues like inflation, vulnerabilities in the financial sector, and external pressures to maintain growth and further reduce poverty. Stabilizing the economy and supporting continued growth necessitates monetary, exchange rate, and financial sector policy adjustments. Diversification of the economy beyond the RMG sector is crucial for creating more employment opportunities and reducing poverty [39]. By enhancing human capital and improving infrastructure, Bangladesh can attract more private investment, which is vital for job creation and sustained poverty reduction [36].

Additionally, research has demonstrated that poverty can be alleviated through financial inclusion [40], bolstering efforts to reduce poverty. Financial services make it possible for low-income households to pay for cutting-edge energy solutions [41]. These findings emphasize the significance of energy access to strategies for reducing poverty, and the need for policies in Bangladesh that integrate energy expansion with efforts to promote financial inclusion.

Despite the progress, Bangladesh faces challenges such as inflation, energy shortages, and financial sector vulnerabilities, which can hinder economic growth and poverty reduction efforts. The World Bank projects a subdued GDP growth rate of 5.6% for the fiscal year 2024, compared to the pre-pandemic average growth rate of 6.6%. Addressing these challenges through comprehensive policy reforms is essential for achieving long-term poverty reduction goals [42]. In Bangladesh, economic policies, structural reforms, and external economic conditions all have an impact on the complex relationship between GDP growth and poverty reduction. Even though GDP growth has made a big difference in reducing poverty, more work is needed to solve problems and keep this progress going.

In a study on center and periphery in Bosnia and Herzegovina, several social and spatial indicators for regional disparities have been identified. The indicators have shown statistical relationship with regional disparity [43]. Bangladesh has a small land area of 147,000 square kilometers with a huge population of 170 million. Geographically the country is almost homologous, and spatial difference is low. Even though the country has some regional disparities in terms of development and poverty, the national level poverty rate is 18.7%. But different levels of poverty are found among the eight Divisional areas. Khulna (14.8%), Chattogram (15.8%), and Rajshahi (16.7%) have the lowest poverty rates while Rangpur (24.8%), Mymensingh (24.2%), and Barishal (26.9%) have the highest poverty rates. The central region with the capital city Dhaka (17.9%) and the Sylhet Division (17.4%) have medium poverty rates. Among the regions, Chattogram, Rajshahi, Dhaka, and Sylhet have a natural gas supply facility. Industrialization in Bangladesh has occurred more in the area where there is a natural gas supply. Electricity supply is almost identical all over the country. Chattogram and Khulna have seaports and Dhaka, Chattogram, and Sylhet have international airports. It is worth mentioning that the Dhaka, Chattogram, and Rajshahi Divisions, where poverty rates are relatively low, have a higher number of microfinance institutions (MFIs). On the contrary, Mymensingh, Barisal and Sylhet, where poverty rates are higher, have a lower number of MFIs. So, energy, microfinance activities, and infrastructure facilities may be some of the reasons for regional disparity in the country. Division-wise, the number of MFIs and poverty rates are shown in Figure 2.

In general, the research demonstrates how energy use, GDP growth, and microfinance influence outcomes for reducing poverty. Despite the importance of energy and economic expansion, microfinance’s facilitation of inclusive processes ultimately determines their impact on poverty alleviation. By providing empirical evidence from a developing nation, this study aims to build on previous research by shedding light on how these variables interact in a specific context to influence poverty reduction.

Based on the literature reviews, it is observed that a few studies investigate the combined effects of energy access, GDP growth, and microfinance on poverty reduction. To the best of the authors’ knowledge, how integrated policy approaches the combination of energy, economic growth, and microfinance initiatives can be optimized for poverty reduction is absent in the existing literature; this necessitates the need for more context-specific studies that consider local economic, social, and macroeconomic factors in evaluating the combined impact.

3. Data and Methodology

3.1. Data and Variables

Annual poverty rates have been used as dependent variables; whereas GDP, per capita power consumption, total energy consumption, and microfinance loan disbursement have been used as independent variables. Data for 27 years, starting from 1995 to 2022 have been collected from various sources. Collected data for this research is time series data by nature. An analysis has been conducted using EViews v.10 software. Different variables and their definitions with sources are shown in

Table 1. List of different variables, definitions, and data sources.

Variable	Definition	Unit	Source of Data
POVR	Poverty Rate	%	World Bank
GDP	Gross Domestic Products	Billion USD	World Bank
PCON	Per Capita Power Consumption	KWh	Bangladesh Power Development Board
TEC	Total Energy Consumption	Kiloton Oil Equivalent (ktoe)	World Energy Information
MFL	Microfinance loan disbursement	Million USD	Microcredit Regulatory Authority, Bangladesh

.

Figure 3 shows the line charts of the four independent variables. All the variables exhibit upward sloping trends, which means GDP, microfinance, power, and energy consumptions are increasing over time. Each of the lines has an intercept and a trend.

3.2. Research Models and Their Significance

It is important to select an appropriate methodology for analyzing the time series data. The descriptive statistics present the basic characteristics of the series such as central tendencies and dispersion of data [48]. The aim of this analysis is to forecast the value of the dependent data using given values of the independent data. The choice of methodology for a time series analysis primarily depends on the results of unit root tests that detect whether the variable is stationary or not.

When all the series data are stationary, the methodology becomes very specific. In such cases, unbiased estimates can be obtained using ordinary least square (OLS) or vector autoregressive (VAR) models [49]. Three scenarios will determine the approaches to be used if none of the variables are stationary. Johansen’s cointegration test can be used to identify co-integrating relationships between the dependent and independent variables when all variables are categorized as an I(1) order [50]. A vector error correction model (VECM) is used if cointegration is present. The creation of models that concentrate on the dynamics of the variables over the short and long terms can be aided by this model. The Autoregressive Distributed Lag (ARDL) model is used when the I(0) and I(1) orders are mixed. Short-run dynamics and long-run equilibrium are merged through the development of a dynamic error correction model (ECM) from the ARDL. To estimate the VAR model in the data’s levels and evaluate the causal relationship between variables, the Toda–Yamamoto (T–Y) model is utilized when all the data are non-stationary in different orders with I(0), I(1), and I(2) [51].

Finally, the models undergo diagnostic tests to evaluate normal distribution, serial correlation, and heteroscedasticity. Valid interpretation of the results is possible when the functional form and residuals exhibit no serial correlation, adhere to a normal distribution, and exhibit homoscedasticity [52]. For stability test of parameters, two tests are applied: the cumulative sum (CUSUM) of recursive residuals and the cumulative sum of squares (CUSUMSQ) tests [53]. The following equations have been employed for data analysis for each model. Selection of methods for time series forecasting is shown in Figure 4.

3.3. Stationarity Test

A variable is stationary when its probability distribution always remains the same. In a stationary time series, the mean, standard deviation, and autocovariance remain fixed at all periods; while a non-stationary time series can display changes in variance, mean, or both as time progresses. In this case, the null hypothesis is “there is a unit root in the time series data”. When the p-value is less than 5%, the null hypothesis is rejected, and the alternative hypothesis of “there is no unit root” is accepted. The absence of unit root means the data are stationary. Moreover, it may be noted that a crucial characteristic of time series data is stationarity, which denotes that the statistical characteristics of the data remain constant across time. The ADF test determines the stationarity of data by looking for unit roots. The PP test is a stationarity test that is like the ADF test but uses different assumptions and techniques, which makes it robust in some situations. The ADF and PP tests are enhanced by the KPSS test.

3.3.1. ADF Unit Root Test

Equation (1) is employed to conduct the ADF unit root test:

$$∆Y_t=\mu +\beta t+\alpha Y_{t-1}+\sum _{i=1}^k{c}_i∆Y_{t-i}+{\epsilon }_t$$

(1)

where, μ = a constant, ΔY_t = first difference of Y_t or Y_t − Y_t−1, α = coefficient of Y_t−1, t = time and ε_t = error term.

3.3.2. PP Unit Root Test

The subsequent formula is employed for the PP test:

$$∆Y_t=\alpha +\theta Y_{t-1}+{\epsilon }_t$$

(2)

where, ε_t (white noise error term) is I(0) and heteroskedastic, Y_t−1 is a deterministic trend component and α = constant.

3.3.3. KPSS Unit Root Test

The KPSS test utilizes linear regression as its foundation. It decomposes a series into three distinct components: a deterministic trend (β_t), a random walk (r_t), and a stationary error (ε_t). These components are incorporated into Equation (3).

X_t = r_t + β_t + ε_t

(3)

If the variable exhibits stationarity, it will possess a constant term for an intercept, or the variables will demonstrate stationarity with respect to a fixed level [54].

3.4. Toda–Yamamoto Granger Causality Test in VAR Approach

As noted by Yamamoto and Toda, the Toda–Yamamoto procedure is effective in mitigating the bias introduced by non-stationarity in Granger causality tests within a VAR framework. In VAR models, it is crucial to account for non-stationarity in the time series data. The Toda–Yamamoto test allows for the inclusion of lagged terms in the model to test for causality even when the variables may be non-stationary. Standard Granger causality tests can produce biased results when variables are non-stationary. The Toda–Yamamoto procedure adjusts for this by employing additional lagged terms and applying appropriate transformations to the variables. Unlike traditional Granger tests, which assume a linear causal relationship, the Toda–Yamamoto test allows for the possibility of bidirectional causality or feedback effects between variables in a VAR framework [55].

The Toda–Yamamoto Granger causality test was utilized to establish a causal relationship between energy and economic growth in Bangladesh [56]. Ahmed et al. examined the causal relationship between financial growth and the decrease of poverty in Bangladesh using the Toda–Yamamoto technique [57]. A study that investigated the causal association between energy consumption and the decrease of poverty in developing nations used the T–Y test [58]. Through application of the Toda–Yamamoto Granger causality test, the authors discovered a substantial causal association between energy consumption and the decrease of poverty in developing nations. Rahman et al. used the Toda–Yamamoto procedure to investigate how changes in energy consumption affect poverty rates across different regions. The authors identified bidirectional causality between energy consumption and poverty rates across different regions [59]. They suggested that policies promoting energy access and efficiency could potentially reduce poverty by stimulating economic growth and improving living standards.

The Granger Causality test procedure by Toda and Yamamoto commences with the selection of the optimal lag length k through the implementation of the standard lag selection process. Subsequently, one needs to determine the maximum order of integration, or d_max [60]. The maximum difference value, d_max, will be 2 if the stationarity tests verify that the series are stationary at I(0), I(1), and I(2). The lag order k (k ≥ d_max) must be greater than or equal to d_max for the model to be considered valid. Estimating a VAR model of order (k + d_max) comes next and performing a block exogeneity Wald test is crucial to assess causality.

To conduct the Granger causality test using the Toda–Yamamoto procedure, it is essential to modify the basic econometric model, denoted by Equation (6), into the VAR system that follows:

POVR_t = α₀ + M + ϒ_1t

(4)

GDPt = β₀ + M + ϒ_2t

(5)

PCON_t = λ₀ + M + ϒ_3t

(6)

TEC_t = θ₀ + M + ϒ_4t

(7)

MFL_t = δ₀ + M + ϒ_5t

(8)

where, α₀, β₀, λ₀, θ₀, and δ₀ are constant terms.

ϒ_1t to ϒ_5t = Residuals of the model.

$$\begin{array}{ll}\mathrm{M}& \equiv {\sum }_{i=1}^k{\alpha }_{1i}{POVR}_{t-i}+{\sum }_{i=k+1}^{k+dmax}{\alpha }_{2i}{POVR}_{t-1}+{\sum }_{i=1}^k{\beta }_{1i}{GDP}_{t-i}+{\sum }_{i=k+1}^{k+dmax}{\beta }_{2i}{GDP}_{t-i}\\ & +{\sum }_{i=1}^k{\lambda }_{1i}{PCON}_{t-i}+{\sum }_{i=k+1}^{k+dmax}{\lambda }_{2i}{PCON}_{t-i}+{\sum }_{i=1}^k{\theta }_{1i}{TEC}_{t-i}+{\sum }_{i=k+1}^{k+dmax}{\theta }_{2i}{TEC}_{t-i}\\ & +{\sum }_{i=1}^k{\delta }_{1i}{MFL}_{t-i}+{\sum }_{i=k+1}^{k+dmax}{\delta }_{2i}{MFL}_{t-i}\end{array}$$

(9)

Based on Equation (9), POVR is induced by GDP when β_1i ≠ 0, for all i = 1,2, …, k. The T–Y method examines the linear or nonlinear constraints on the initial k coefficient matrices utilizing conventional asymptotic theory [51]. It may be noted that three steps are required to implement the T–Y. At the very onset, we need to test each variable to find out the d_max by using tests like the ADF unit root test, the PP unit root test, and the KPSS test to identify the maximum order of integration (d_max) of the variable. Afterwards, we need to select the optimal lag length (k) based on several lag length selection criteria. Finally, in the third step, we need to test the vector auto regression (VAR) model for causality.

4. Results and Discussion

4.1. Descriptive Statistics

Table 2. Descriptive Statistics of the variables.

	POVR	GDP	PCON	TEC	MFL
Mean	35.19259	158.3507	179.3067	28898.96	6066.259
Median	34.90000	102.4800	141.9700	27003.00	3780.000
Maximum	50.10000	460.2000	422.1300	49,306.00	19,384.00
Minimum	18.90000	46.44000	60.21000	15,797.00	670.0000
Std. Dev.	11.21602	128.6975	108.5556	10354.26	6077.084
Skewness	−0.011892	1.049987	0.830236	0.429058	0.995965
Kurtosis	1.519760	2.727282	2.479596	1.918785	2.545120
Jarque-Bera	2.465634	5.044802	3.406484	2.143564	4.696536
Probability	0.291470	0.080267	0.182092	0.342398	0.095534
Observations	27	27	27	27	27

displays the descriptive statistics of the variables, and these statistics primarily compute the measures of dispersions and central tendency. Over the years, the average poverty rate has been 35.19%, with a top rate of 50.1% and a minimum rate of 18.9%. At −0.012 and 1.520, the skewness is negative, and the kurtosis is positive. Over the years, the average GDP was 158.35 billion USD, with a maximum of 460.20, and a minimum of 46.44. Both the skewness and the kurtosis, at 1.05 and 2.73, are favorable. The average electricity consumption is 179.31 kWh, with minimum and maximum values of 60.21 and 422.13 kWh, respectively. At 0.83 and 2.48, the skewness and kurtosis scores are positive. The total energy consumption average is 28,898.96 ktoe, with minimum and maximum values of 15,797 and 49,306, respectively. The values of 1.92 and 0.42 indicate that kurtosis and skewness are both positive. Microfinance loans range from a minimum of 670 USD to a high of 19,384 USD, with an average of 6066.26 million USD. At 1.00 and 2.55, the skewness and kurtosis are both positive. The skewness of all variables is within −1 to +1 value, mean excellent condition in terms of normality [61,62]. All the variables have similarities according to the nature of their descriptive statistics.

4.2. Unit Root Test

The statement “the variable has a unit root (or non-stationary)” is the null hypothesis for the unit root test. The null hypothesis is rejected, and the data is considered stationary if the likelihood is less than 5%. Our computation for the stationary test reveals that the unit root tests have yielded a jumbled order of integration. At the first order of integration, a majority of the variables are stationary according to both ADF and PP methods. Only the dependent variable poverty rate shows stationarity at I(2) data in both methods. When there is mixed integration of I(1) and I(2), we can use the Toda–Yamamoto method to develop a cointegration relationship among the variables.

Table 3. Summary results of ADF and PP unit root tests.

ADF Test
Variables	Level		1st Difference		2nd Difference		Order of Integration
	t-Statistic	Probability	t-Statistic	Probability	t-Statistic	Probability
POVR	−1.622424	0.7559	−1.534807	0.7906	−4.262635	0.0132 *	I(2)
GDP	0.797355	0.9995	−4.362089	0.0103 *	-	-	I(1)
PCON	0.398057	0.9979	−7.526857	0.0000 *	-	-	I(1)
TEC	−1.056841	0.9167	−7.269787	0.0000 *	-	-	I(1)
MFL	3.333218	1.0000	−5.338551	0.0012 *	-	-	I(1)
PP Test
	t-Statistic	Probability	t-Statistic	Probability	t-Statistic	Probability
POVR	−2.061981	0.5427	−1.534807	0.7906	−7.492427	0.0000 *	I(2)
GDP	0.797355	0.9995	−4.362089	0.0103 *	-	-	I(1)
PCON	1.529097	1.0000	−7.340959	0.0000 *	-	-	I(1)
TEC	−0.916105	0.9388	−10.08038	0.0000 *	-	-	I(1)
MFL	−0.522999	0.9755	−5.324481	0.0012 *	-	-	I(1)

Note: * Indicates the rejection of null hypothesis of “the variable has a unit root”.

displays the summary results of ADF and PP unit root tests.

The ADF test is employed to test for non-stationarity, while the Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test is used to test for stationarity to ascertain the highest level of integration. The ADF and KPSS tests will determine the D_max. The following table shows the summary results of the KPSS test where we find the maximum order of integration is one. In ADF test maximum order is 2. Accordingly, D_max = 2 for this time series data. The KPSS test results are shown in

Table 4. KPSS test results.

Variables	Critical Value at 5% Level	Test Statistic		Order of Integration
		Level	1st Difference
POVR	0.146000	0.107213 *	-	I(0)
GDP	0.146000	0.204117	0.108774 *	I(1)
PCON	0.146000	0.202690	0.075719 *	I(1)
TEC	0.146000	0.209401	0.128500 *	I(1)
MFL	0.146000	0.195540	0.061069 *	I(1)

Note: * Indicates the rejection of null hypothesis of “the variable has a unit root”.

.

4.3. Lag Length Selection

Prior to conducting the Toda–Yamamoto causality test, it is essential to determine the most suitable lag length. Through the evaluation of various criteria such as the Akaike information criterion (AIC), sequential modified (LR), Schwarz information criterion (SC), Hannan–Quinn information criterion (HQ), and the final prediction error (FPE), it was found that out of five criteria, four criteria exhibit a lag length of two. As a result, a lag length of two (k = 2) has been identified as optimal for the Toda–Yamamoto causality test.

Table 5. VAR lag length selection.

Lag	LogL	LR	FPE	AIC	SC	HQ
0	−709.1115	NA	4.45 × 1018	57.12892	57.37269	57.19653
1	−552.2572	238.4185	1.22 × 1014	46.58057	48.04322 *	46.98625
2	−516.1662	40.42183 *	6.64 × 1013 *	45.69330 *	48.37483	46.43704 *

* lag order selected by the criterion.

lists the values of the VAR lag length selection.

4.4. Toda–Yamamoto Test Results

The maximum order of integration (d_max), as determined by the ADF and KPSS tests, was found to be two. It is essential that k should be ≥d_max to create a viable model. Lastly, to determine the direction of the causal relationship, it is critical to estimate a (k + d_max) the order of VAR and perform the T–Y test. In the present scenario, k = 2, d_max = 2, and the order of VAR is 4. Toda–Yamamoto Granger Causality tests results are given in

Table 6. Toda–Yamamoto Granger Causality Tests results.

VAR Granger Causality/Block Exogeneity Wald test
Dependent variable: POVR
Variable	Chi-Square	df	Probability
GDP	11.78580	2	0.0028 *
PCON	0.362448	2	0.8342
TEC	8.258243	2	0.0161 *
MFL	3.238735	2	0.1980
All	19.85529	8	0.0109 *
Dependent variable: GDP
Variable	Chi-Square	df	Probability
POVR	2.833778	2	0.2425
PCON	13.31018	2	0.0013 *
TEC	2.871965	2	0.2379
MFL	9.583745	2	0.0083 *
All	25.57793	8	0.0012
Dependent variable: PCON
Variable	Chi-Square	df	Probability
POVR	0.746399	2	0.6885
GDP	5.440485	2	0.0659
TEC	3.969587	2	0.1374
MFL	2.861684	2	0.2391
All	23.87339	8	0.0024
Dependent variable: TEC
Variable	Chi-Square	df	Probability
POVR	4.607531	2	0.0999
GDP	7.641475	2	0.0219 *
PCON	2.616844	2	0.2702
MFL	10.63353	2	0.0049 *
All	34.17210	8	0.0000
Dependent variable: MFL
Variable	Chi-Square	df	Probability
POVR	1.322733	2	0.5161
GDP	13.75693	2	0.0010 *
PCON	12.53200	2	0.0019 *
TEC	1.485388	2	0.4758
All	63.33311	8	0.0000

Note: * Indicates that the variable affects the dependent variable.

.

The test results show that the null hypothesis for the T–Y test, “GDP does not granger cause poverty rate”, is rejected at the 5% significance level, indicating that GDP has an impact on the rate of poverty. Additionally, the results reject the null hypothesis, “all variables combinedly do not granger cause poverty rate”, at the same 5% significance level. This means that four independent variables combined to affect the poverty rate, since the probability of all variables is 0.0109, a value less than 5%. In a similar manner, power consumption affects both GDP with a probability of 0.0013 and microfinance with a probability of 0.0083. Thus, GDP and microfinance loans have bi-directional causality relationship with each other. Again, microfinance and GDP affect the total energy consumption with probabilities of 0.0219 and 0.0083, respectively. Finally, GDP and total energy consumption with probabilities of 0.0010 and 0.0019, individually contribute to poverty reduction. So, there is a chain relationship among the variables that starts with power consumption and ends with poverty rate. Microfinance, GDP, and total energy consumption stand in between them. It is evident from the T–Y analysis that GDP, microfinance, power, and energy consumption make a significant contribution to poverty reduction in Bangladesh directly or indirectly. Figure 5 summarizes the relationship among the variables.

4.5. Residual Diagnostic Tests

4.5.1. Normality Jarque–Bera Test

The Jarque–Bera test is used to assess the normality of the residuals or errors in a regression model. It tests whether the distribution of the residuals follows a normal (Gaussian) distribution [62], which is one of the basic assumptions of linear regression. In this case, the Jarque–Bera value probability is 0.518264, which is higher than 5%. As a result, the null hypothesis that the data are normally distributed is not successfully refuted. When data is regularly dispersed, the preceding test outcomes gain greater dependability. The Jarque–Bera Normality test results are plotted in Figure 6.

4.5.2. Serial Correlation and Heteroskedasticity Test

Test results of serial correlation and Heteroskedasticity are displayed in

Table 7. Serial Correlation and Heteroskedasticity Test results.

Serial Correlation LM Test
F-statistic	0.782533	Probability: F (2,20)	0.4708
Obs *R-squared	1.959502	Probability: Chi-Square (2)	0.3754
Heteroskedasticity Test
F-statistic	1.580400	Probability: F (4,22)	0.2147
Obs *R-squared	6.026608	Probability: Chi-Square (4)	0.1972
Scaled explained SS	3.287824	Probability: Chi-Square (4)	0.5109

. The likelihood of a chi-square, as determined by the Breusch–Godfrey serial correlation LM test, is 0.375, a number more than 5%. Therefore, the null hypothesis that “there is no serial correlation within the variables” cannot be rejected. The likelihood of a chi-square, as determined by the Breusch–Pagan–Godfrey heteroscedasticity test is 0.197, a number more than 5%. Because of this, the null hypothesis that “there is no heteroscedasticity in the data set” cannot be rejected. Since the variables do not exhibit serial correlation or heteroscedasticity, the estimates derived from the test findings are reliable, efficient, and objective.

4.5.3. Model Stability Tests

Figure 7a depicts that the model’s long-run coefficients are stable in accordance with the CUSUM test as the blue lines are in between the upper and the lower critical line at 5% level of significance. It means the time series has no systematic change in the regression coefficient or there is no possibility of structural breaks in the residuals. Figure 7b shows the CUSUMSQ test where the blue line goes beyond the red line. It means that there is the possibility of a sudden break in the long-run stability of the model.

5. Conclusions

This study analyzed the causal relationship among poverty rate, per capita power consumption, total energy consumption, gross domestic product (GDP), and microfinance loans in the context of a developing economy, Bangladesh. Yearly time series data spanning 1995–2022 were collected from various secondary sources and relevant econometric models were applied to analyze the data. The yearly poverty rates show a decreasing trend with the increasing GDP, power consumption, total energy consumption, and microfinance over the years. Our calculation for the stationary test shows a mixed order of integration with I(0), I(1) and I(2) from the unit root tests. For this combination of orders, the Toda–Yamamoto test has been applied to determine the cointegration among the variables.

The Toda–Yamamoto test result shows that the four independent variables combinedly affect the poverty rate with a p-value of 0.0109 which is statistically significant. Power consumption affects both GDP and microfinance. GDP and microfinance loans have bi-directional causality relationship with each other. Moreover, microfinance and GDP affect the total energy consumption. Finally, GDP and total energy consumption individually contribute to poverty reduction. Therefore, it is evident from the T–Y analysis that GDP, microfinance, power and energy usage all significantly contribute to the country’s efforts to reduce poverty. The average GDP growth rate of the economy is approximately 6% over the last decade, which is expected to continue in the coming years. Microfinance institutions, such as Grameen Bank, BRAC, Proshika etc. are performing well to alleviate poverty countrywide. Even though the terms and conditions to get access to microcredit and the repayment procedures should be more liberal. On the other hand, the power and energy sector has been suffering from many problems and inefficiencies, making the country one of the lowest energy-consuming economies in the world. The Energy Division of Bangladesh could not implement any effective activity for energy efficiency and moderation of energy use, which could have played an important role in the higher energy consumption in poverty alleviation. The national policy-makers should emphasize more on the increased production and proper utilization of power and energy to reduce the poverty rate in Bangladesh. Policy implications may include strategies to promote sustainable energy development, improve energy efficiency, moderation of energy use, and ensure easy and fair access to energy services. Integrating energy considerations into poverty reduction strategies can lead to more effective and targeted interventions, particularly in developing regions.

Future study may attempt to cover a wider range of data, including foreign direct investment (FDI), remittance, ready-made garment (RMG) exports, etc. to examine their impacts on poverty alleviation to conduct more exhaustive research. This research lays the groundwork for future academic inquiry by noting areas that require more research, such as the importance of renewable energy sources and the long-term viability of microfinance programs. This contribution is valuable for academics, researchers, and policy-makers seeking to build on the existing body of knowledge and address unresolved questions in the field.