Dynamic links between economic complexity, technological innovation, structural transformation and energy sustainability in newly industrializing countries

Abstract

The active industrial activities, urbanization, globalization and rapid economic development have stimulated the energy demand of newly industrialized countries, prompting renewable energy to be replaced by cheaper and more readily available fossil fuel energy, thus leading to environmental pollution. This study therefore explores the impact of economic complexity, technological innovation and structural transformation on energy efficiency, fossil fuel energy demand and renewable energy generation in newly industrialized countries (NICs). This study differs from earlier studies in that it focuses on improving energy efficiency, renewable energy generation, and reducing fossil fuel energy as indicators of energy conservation and sustainability. This study employs the cross-sectional autoregressive distributed lag (CS-ARDL) technique to investigate both short- and long-term associations and examine the robustness of the results over the period 1985–2023. The results show that in the long run, each unit increase in economic complexity can improve energy efficiency by 0.265%, promote renewable energy generation by 0.327%, and reduce fossil fuel energy demand by 0.228%. Technological innovation has significant incremental impacts on energy efficiency and renewable energy, while having an adverse impact on fossil fuel energy in both the short- and long-term. However, GDP can greatly improve energy efficiency and renewable energy generation while boosting fossil fuel energy demand, but only in the long term. Industrial value added significantly reduces energy efficiency and renewable energy generation, while contributing to fossil fuel energy demand in both short- and long-term. The causality test support feedback hypothesis between GDP and fossil fuel energy. Comparative analysis shows that economic complexity and technological innovation contribute more to supporting energy efficiency and energy production than to mitigating fossil fuels, indicating the necessity of structural adjustment of economic activities and technological innovation. Policymakers should consider the empirical findings of this study to pave the way for a more sustainable and resilient economic and environmental future for newly industrialized countries through energy transition measures that focus on economic maturity and technological innovation to produce renewable energy.

Introduction

Energy needs have become increasingly urgent in the context of climate change, energy conservation imperatives, and global resource constraints^1,2. Greenhouse gas (GHG) emissions, energy resources and economic productivity are important components of sustainable development and largely depend on energy demand^3,4. Energy demand plays a vigorous role in attaining the Sustainable Development Goals such as affordable and clean energy (SDG-7), Industry, innovation and infrastructure (SDG-9), and climate change alleviation (SDG-13)^5,6. Sustainable Development Goal (SDG) 7 states that all aspects of development (climate change mitigation, industry, innovation and infrastructure, water supply and healthcare) can be achieved through access to sustainable, affordable and reliable energy. The world is slowly moving towards sustainable energy goals, and by 2030, nearly 2 billion people will rely on fossil fuel energy and polluting technologies for cooking, while nearly 660 million people do not yet have access to electricity. About 60% of total global greenhouse gas emissions are caused by energy consumption, which is the main contributor to climate change. Access to electricity by the proportion of the World population has boosted from 87% in 2015 to 91% in 2021⁷. The IEA⁸ report predicts that global demand for fossil fuel energy (oil, coal and natural gas) will peak in 2030, but this demand may push the achievement of the Paris Agreement 1.5 °C target too high. In addition, the report shows that demand for fossil fuel energy (coal, oil and natural gas) will remain stable until 2050 as fossil fuel energy consumption increases in newly industrialized economies and decreases in developed countries. In newly industrialized countries, the main source of industrial energy is fossil fuels, directly emitting 9.0Gt of carbon dioxide.

Currently, countries are moving from resource-intensive to economic complexity, focusing more on R&D and innovation to improve energy efficiency and reduce the harmful effects of fossil fuel energy on environmental protection^9,10. The intricacy and diversity of economic activities within a region is called economic complexity¹¹. The complexity of the economy prompts further research and development to introduce environmentally relevant technologies to improve energy efficiency and reduce the demand for fossil fuel energy and environmental harm¹². Furthermore, in countries with higher economic complexity, the proportion of knowledge-based products is greater than the proportion of fossil fuel-consuming energy production. Compared with developed economies, newly industrialized countries (NIC) have lower levels of economic complexity and technological innovation, higher dependence on fossil fuels and greater environmental risks^9,13. In contrast, regions with higher economic complexity are diversifying their energy mix by evaluating more advanced technologies to replace fossil fuel energy needs with renewables^14,15. Therefore, economic complexity can not only improve energy efficiency and reduce the demand for fossil fuel energy, but also promote the development of renewable energy.

Researchers believe that economic complexity mainly affects energy (fossil fuels and renewable energy) from two aspects: technology and scale, thus having a complex impact on energy (fossil fuels and renewable energy)^15,16. On the one hand, higher economic complexity can stimulate the formation and development of renewable energy and reduce fossil fuels through improved technology and production input structures. On the other hand, the economic complexities of countries with higher energy demands and energy-intensive economic frameworks may discourage renewable energy and shift to fossil fuels^17–19. Due to population growth, urbanization, increasing industrial activities and rapid economic development, the demand for energy in newly industrialized countries (China, India, Malaysia, Thailand, the Philippines, South Africa, Turkey, Brazil, and Mexico) has increased dramatically. Newly industrialized countries need more energy to meet the demands of industry and expanding cities. Energy demand in newly industrialized countries is expected to grow at an annual rate of 3–4% over the next decade, compared with modest increase in Organization for Economic Co-operation and Development (OECD) countries²⁰. Thus, increased energy demand in newly industrialized countries with economic complexity may stimulate further consumption of fossil fuels as they become cheaper and more readily available, leading to a decline in the share of renewable energy¹⁷.

Contradictory opinions on economic complexity obscure consideration of how it influences fossil fuel energy demand and renewable energy, dictates further investigation, particularly in newly industrialized countries with higher fossil fuel energy capacity. Although many countries view promoting economic complexity as a strategic tool to promote economic development, such adoption may have ambiguous and different impacts on fossil fuel energy demand and renewable energy¹⁵. Many newly industrialized countries have accepted the essential to diversify their economies and mitigate their reliance on fossil fuels, a shift that requires structural transformation through the adoption of energy-saving technologies or technological innovations²¹. Newly industrialized economies should focus on promoting economic complexity to diversify and complicate their economic activities and developing technological innovation, thereby achieving a sustainable energy growth structure.

In recent years, the important issue of structural transformation has attracted the attention of numerous researchers^22,23. Economic complexity supports business development by increasing the productivity of national skills through structural changes, attributes, knowledge and skills in the host economy¹⁵. In the context of economic complexity, structural changes in more aspects of services and products can be brought together²⁴. Thus, economic complexity can measure the production structure by expressing the degree of intricacy and show variations in the industrial structure^10,25. The application of skills and knowledge promotes economic sophistication and has the potential to reduce the energy intensity of quality manufacturing¹⁵. The ubiquity of manufactured goods and the degree of product differentiation are also known as economic complexity.

Technological innovation has recently been recognized as a key driver of the energy transition, with energy efficiency continuing to improve through the falling costs and capabilities of underlying technologies such as electric vehicles and solar panels^26,27. Promoting technological innovation in newly industrialized countries can help industries shift from current energy-intensive manufacturing methods to less energy-intensive technologies, thereby reducing reliance on fossil fuels and increasing reliance on renewable energy^28,29. Achieving the 2050 Net Zero Emissions (NZE) scenario requires accelerating clean energy technology innovation. Reaching net-zero CO₂ emissions from the energy sector by 2050 will not require a breakthrough on the level of the original discoveries of batteries, solar, and wind, nor will it require the introduction of entirely new scientific concepts. However, around 35% of CO₂ reductions will need to be achieved through innovation, which plays a significant role in the recently updated 2050 net-zero emissions scenario³⁰. This means that renewable energy technologies have not yet entered the commercial-scale market and are still in the development process for the 2050 net-zero emissions target. As energy supplied by fossil fuels decreases and manufacturing processes improve, continued innovation is also needed to reduce technology costs and improve performance. Existing evidence suggests that the speed and scope of the clean energy transition may be enhanced as new ideas emerge and lead to new technological concepts or materials. But new combinations of existing technologies or enhanced designs can support minimizing the use of critical resources, improving performance, reducing costs, solving new use cases and mitigating other environmental impacts.

The main purpose of this study is to reveal the impact of economic complexity and technological innovation on energy sustainability, especially in the context of increasing energy efficiency, reducing fossil fuel energy demand and adopting renewable energy sources. Energy efficiency, reducing fossil fuel energy demand and switching from fossil fuels to renewable energy sources play a vital role in energy sustainability and climate change mitigation. Thus, this study examines the influence of economic complexity and technological innovation on energy efficiency, fossil fuel energy demand and renewable energy generation, while integrating industrial value added as an indicator of structural change into the model analysis. The current study extends the existing literature by focusing specifically on improving energy efficiency, mitigating fossil fuel energy demand, and renewable energy generation as key indicators of energy sustainability, and explores the role of technological innovation and economic complexity in influencing these factors. This study also aims to provide valuable insights to policymakers and government officials by incorporating industrial value added as a dimension of structural change in empirical investigations aimed at enhancing the sustainability of energy, environment and economic growth.

This study contributes to the literature by exploring the impact of economic complexity and technological innovation on energy efficiency, fossil fuel energy demand and renewable energy generation, which are key components of energy and environmental sustainability goals. Many studies have investigated economic complexity and technological innovation and their impact on environmental emissions, but few have examined their impact on energy efficiency, fossil fuel energy demand and renewable energy generation. Thus, this study determined to fill this gap by revealing the impact of economic complexity and technological innovation on energy efficiency, fossil fuel energy demand and renewable energy generation in the context of newly industrialized countries. Besides, the study takes into account industrial value added as a structural change to reveal its impact on fossil fuel energy demand and renewable energy generation. The newly industrialized countries (China, India, Malaysia, Thailand, Philippines, South Africa, Turkey, Brazil, and Mexico) were selected for this study because the demand for energy specifically fossil fuels energy in these countries has increased dramatically due to globalization, urbanization, increased industrial activity, and rapid economic development. while also considering industry share as an indicator of structural transformation. This research can help stakeholders, officials and policymakers develop effective strategies and provide valuable insights into the economic complexities and the role of technological innovation in supporting the transition from fossil fuel demand to renewable energy generation. Furthermore, this study follows the continuously updated fully modified (CS-ARDL) method introduced by Chudik and Pesaran³¹ and panel causality test developed by Dumitrescu and Hurlin’s³², a second-generation perception that can produce robust and reliable results under cross-sectional dependence and slope heterogeneity.

Literature review

Numerous studies in the literature reveal perceptions of economic complexity and technological innovation and their economic, social, and environmental impacts. A well-established body of research shows that, in addition to recognizing the importance of institutional arrangements, human capital, and physical assets, investing in more complex and uncommon goods and innovations can boost a country’s growth prospects^33–35. Few recent studies have enriched the literature on the impact of economic complexity and technological innovation on various aspects such as environmental degradation^34,36–38, energy consumption^39–43, economic growth^44–47, income inequality^48–50, gender equality^51,52, natural resource rents^53,54, and human development^55,56.

Economic complexity and environmental degradation nexus

The literature has reached diverse conclusions with a range of studies showing that economic complexity reduces environmental damage, while some studies show that it increases environmental emissions. Aluko et al.⁵⁷ enriched existing knowledge by revealing the impact of economic complexity on environmental emissions in 35 OECD countries from 1998 to 2017 through a fixed effects model approach. The results show that economic complexity can adversely affect environmental emissions. Dada et al.⁵⁸ in their recent study also support the adverse impact of economic complexity on the ecological footprint of African countries. Similarly, Saqib and Dinca⁵⁹ reveal that, using a panel NARDL approach, positive shocks to economic complexity reduce carbon emissions, while negative shocks promote long-term environmental emissions in major clean energy investing countries from 1995 to 2020. In contrary, Majeed et al.⁶⁰ employ the FMOLS estimator to determine that economic complexity has a strong progressive effect on CO₂ emissions in the long run and validate the EKC hypothesis for OECD countries over the period 1971–2018. More similarly, another study by Ahmed et al.⁶¹ used the CS-ARDL approach from 1984 to 2017 to conclude that economic complexity exacerbates the ecological footprint of emerging countries. Another study by Kelly and Ndeffo⁶² used the Discroll and Kraay fixed effects method and the system GMM estimator covering the period 1998–2017 and concluded that economic complexity exacerbates 22 sub-Saharan African (SSA) country’s environmental emissions. Similar results were obtained by Yasin et al.⁶³, using panel quantile regression from 2000 to 2022, conclude that economic complexity significantly exacerbates environmental degradation in 134 countries. A latest study using the FMOLS method was conducted by Balsalobre-Lorente et al.⁶⁴ demonstrated the inverted U-shaped EKC hypothesis between economic complexity and environmental damage in G20 countries from 1997 to 2018. Environmental pollution caused by increased economic complexity is particularly evident in countries, especially in the early stages of exporting (because exporting goods requires a lot of resources). But at a certain stage, pollution can be reduced by using less polluting resources and technologies^65,66.

Economic complexity and energy nexus

The link between economic complexity and energy demand has not been extensively studied in the literature. Existing research shows that the impact of economic complexity on energy demand has complex consequences. Such as, Can et al.⁶⁷ reveal that between 1971 and 2014, energy consumption augmented in 44 developing countries, while it decreased in 21 developed countries with higher economic complexity. However, Adekoya et al.⁶⁸ determined that energy intensity increases in emerging and developed economies and decreases in other economic level countries with higher economic complexity. Fang et al.⁴¹ empirically illustrate that economic complexity and energy prices reduce energy demand in 25 OECD countries using the AMG approach during 1978–2016. Another study by Dogan et al.⁴⁰ follow a panel ARDL model for the period 1971–2018 and show that economic complexity promotes short-term energy demand while mitigating long-term energy demand for a panel of 63 economies. Rafique et al.⁶⁹ used GMM, FMOLS and DOLS methods to conclude that economic complexity significantly contributes to the demand for renewable energy in G7 and E7 economies. Similarly, Can and Ahmed⁷⁰ contribute to the literature by revealing that demand for renewable energy increases and demand for non-renewable energy decreases as economic complexity increases across 14 EU economies. Another study by Chu⁷¹ shows that the expansion of economic complexity between 1980 and 2017 hinders renewable energy promotion in G7 countries using quantile regression.

Technological innovation and environmental degradation

Technological innovation can help improve environmental quality by diminishing environmental emissions through the use of clean energy such as wind power or solar^72,73. A large number of studies have shown that technological innovation can effectively reduce environmental pollution. Such as, recent research by⁷⁴ pointed out that technological innovation is an important driver of sustainable development, indicating that innovation can promote economic progress without causing harmful effects on the environment. Similarly, Suki et al.⁷⁵ asserted that with the development of technological innovation in Malaysia using the BARDL model, carbon emissions and ecological footprint as indicators of environmental contamination are decreasing. The findings suggest that encouraging domestic and foreign investors to increase investment in renewable energy generation and technological innovation in the Malaysian economy could protect the country’s environment. A panel ARDL model was applied in another study by⁷⁶ concluded that technological innovation significantly reduces environmental pollution levels in South and Southeast Asian economies, but only in the long term. Chu⁷¹ used quantile regression for 1995–2015 to demonstrate that ecosystems in 30 OECD countries can be protected as technology develops. Similarly, another study by Mughal et al.⁴³ followed the FMOLS method from 1990 to 2019 and determined that South Asian countries reduced environmental pollution and improved economic progress by promoting technological innovation. Danish and Ulucak⁷⁷ asserted that with the development of technological innovation, environmental adulteration has significantly reduced in the United States, while the reduction has been slow in China.

Technological innovation and energy nexus

Technological innovation is widely proven to be an important indicator in driving the energy transition towards more sustainable and cleaner energy sources⁷⁸, and energy efficiency is set to increase due to falling costs and technological innovations such as solar panels and electric vehicles^26,27. Chen et al.³⁹ followed the second-generation CS-ARDL process and established that energy efficiency in MENA countries can be improved through higher technological innovation during 1990–2016. Similarly, Wang and Wang⁷⁹ also used the system GMM method to demonstrate the progressive effect of technological innovation on energy efficiency in the east, west, and north of China from 2001 to 2013. The same is true for Zhu et al.⁸⁰ also demonstrated the incremental effect of technological innovation on energy efficiency in China. However, Acheampong et al.⁸¹ document that innovations using a systematic GMM approach directly drove economic growth and energy consumption in Eurozone countries during the period 1995–2019.

The links between economic complexity, technological innovation, and environmental pollution have been extensively studied in the literature, but there are still gaps to be filled in terms of the impacts of economic complexity and technological innovation on energy efficiency, fossil fuel energy demand, and renewable energy generation, especially in the context of newly industrialized countries. In addition, existing studies have focused on the impact of economic complexity and technological innovation on total energy demand and total energy consumption, but there is a lack of research focusing on the impact of economic complexity and technological innovation on fossil fuel energy demand and renewable energy generation.

Theoretical framework, specifications of the model and methodology

Currently, several countries are adopting clean and green energy sources while discouraging the use of fossil fuel energy to achieve environmental and growth sustainability^82–84. To achieve these goals, technological innovations were introduced and structural changes were made^74,85. Concerns about protecting energy supplies in the world energy market, protecting environment from pollution, and reducing dependence on foreign fossil fuels can motivate to promote the transition from fossil fuels to clean energy^82,86. Despite the challenges posed by the transition to renewable energy, many countries are inclined to take important steps to ensure long-term energy security while mitigating environmental pollution^87,88. Although numerous studies have shown that the conversion rate of fossil fuels to renewable energy is low^89,90. As a result, there is a strong desire to accelerate the transition from fossil fuels to clean energy.

Countries are focusing on economic complexity and technological innovation with energy-saving characteristics as measures to accelerate the transition from fossil fuels to renewable energy. Economic complexity indicates the scope and diversity of economic activities within a zone or country¹¹, which affects energy through the reduction of fossil fuel energy use and the adoption of clean energy¹⁵. The adoption of knowledge-intensive technologies associated with greater economic complexity and innovation can improve environmental quality by promoting energy efficiency, renewable energy generation and reducing fossil fuel energy^91,92.

This study hypothesizes that economic complexity and technological innovation have a substantial influence on energy efficiency, fossil fuel energy demand and renewable energy generation, suggesting that economic complexity and technological innovation will improve energy efficiency, reduce fossil fuel energy demand and promote renewable energy generation. More complex economies tend to have a more diverse industrial base, often including sectors that are more energy efficient and use advanced technologies, thereby improving energy efficiency, reducing dependence on fossil fuels, and prioritizing the use of clean energy throughout the economy⁹³. Technological innovations such as energy-saving machinery, new material development, and industrial process optimization directly contribute to improving energy efficiency and replacing fossil fuel energy with renewable energy⁹⁴. The current study uses a panel data set of newly industrialized countries covering the period 1985 to 2023. This study follows the empirical study of Bashir et al.⁹, Yu et al.¹³, and Peng et al.⁹⁵, showing the relationship between economic complexity, technological innovation and CO₂ emissions, and Taghvaee et al.⁴⁴ indicate the link between economic complexity, economic growth, industry and energy consumption, thus, the following equations can be derived.

1

2

3

where EI stands for energy intensity, an alternative to energy efficiency, and can be measured in megajoules per $2017 PPP GDP at purchasing power parity, representing the actions or products that can be produced using a given amount of energy. A reduction in the amount of energy required to produce a product can be viewed as an increase in energy efficiency or a decrease in energy intensity. Thus, a reduction in energy intensity represents an increase in energy efficiency. FFE signifies fossil fuel energy demand, measured as a percentage of total energy demand, EC denotes economic complexity as measured by an index calculated using Standard International Trade Classification (SITC) products. TI represents technological innovation measured as total number of patents, GDP is gross domestic product measured in constant 2015 US dollars, IND is industrial value added measured as a percentage of GDP, and REG reflects renewable energy generation measured in terawatt-hours. Data on fossil fuel energy demand, economic complexity, GDP and industrial value added are from the World Bank. In addition, data on technological innovation can be obtained from the OECD, and data on renewable energy generation can be retrieved from the Statistical Review of World Energy (2024). ln represents the natural logarithm, α reflects the dynamic parameters to be estimated, and μ is the error term. The definitions, measurement units and data sources of each variable are displayed in Table 1.

Table 1.

Variable definitions, measurement units and their data sources.

Variable symbol	Variabledefinition`	Measurement units	Sources
FFE	Fossil fuels energy demand	Percentage of total energy demand	World Bank
EC	Economic complexity	Index	World Bank
TI	Technological Innovation	Total patent	OECD database
GDP	Gross Domestic Product	Constant 2015 US$	World Bank
IND	Industrial Value Addition	Percentage of GDP	World Bank
REG	Renewable Energy Generation	terawatt-hours	World Bank
EI	Energy intensity	Megajoule/$2017 PPP GDP	World Bank

This study examines the role of economic complexity and technological innovation on energy efficiency, fossil fuel energy demand, and renewable energy generation in newly industrialized countries. To achieve these goals, the current study employs several econometric techniques, described below.

Cross-sectional dependence method

The problem of cross-sectional dependence arises more frequently in panel data due to the occurrence of common unobserved shocks and the interdependence of residuals. Ignoring cross-sectional dependence issues can lead to stationarity bias and erroneous results in cointegration and long-run estimation. Thus, it becomes crucial to follow cross-sectional correlation tests in econometric analysis to prevent biased stationarity and errors in long-run cointegration and parameter estimation results. This study follows Pesaran et al.⁹⁶ scaled LM test, Pesaran⁹⁷ cross-sectional dependence (CSD) method, Breusch and Pagan⁹⁸ LM procedure for detecting cross-sectional dependence problems in panel data. The following equations related to these methods can be used to measure cross-sectional dependence.

4

5

6

7

In the above equations, represent the panel time dimension and sample estimate of the pairwise correlation of the residuals, respectively.

Slope heterogeneity approach

Ensuring the homogeneity of parameters across different groups is an important step in the process of panel data analysis. This study follows the method of Pesaran and Yamagata⁹⁹ to address the issue of slope heterogeneity in panel data regression analysis. This method can detect slope heterogeneity and pinpoint the dispersion from arithmetic means in panel data by following a weighted fixed effects pool estimator. The following equations related to the Pesaran and Yamagata slope heterogeneity methods can be used to measure slope heterogeneity problems.

8

9

where are the parameters to be estimated using the ordinary least squares (OLS) technique, and the estimate of the coefficients can be obtained using the weighted fixed effects pooling method. represents the identity matrix and demonstrate a procedure delicate to deviancy from the arithmetic mean (including explanatory factors).

Methods to be followed to detect unit roots

After supporting the existence of cross-sectional correlation and slope heterogeneity in panel data, the next step is to detect the stationarity level of each panel variable through the second-generation panel unit root tests. Because of its limitations, first-generation unit root tests cannot address concerns about cross-sectional correlation and slope heterogeneity. Thus, Cross-section Augmented Dickey-Fuller (CADF) and Cross-section Im-Pesaran-Shin (CIPS) are two second-generation innovative stationarity methods proposed by Pesaran (2007), which can be adopted in the current study to test stationarity of parameters. These tests append cross-sectional components by averaging lags and first differences of the underlying parameters for each cross-section to incorporate sample data cross-sectional dependence and slope heterogeneity. CADF’s stationarity test can check the stationarity level of each individual or the entire panel dynamics, while CIPS checks the unit root of the panel dynamics in the individual series through first differences and uses CADF to increase the lag time. The following equations can be used to measure the panel unit root associated with the CIPS and CADF unit root tests. CADF and CIPS unit root tests are more suitable for testing panel data with a large cross-section (N) and time dimension (T), especially when there are common influencing factors in the data. These tests are designed to handle the cross-sectional dependence common in panel data and may provide more consistent results than traditional unit root tests in this context.

10

where in the aforesaid equation reflects the average of each cross-section.

The CADF unit root method relies strongly on the CIPS method, as shown in the following CIPS equation.

11

Westerlund method for long-run cointegration

Determining the long-term cointegration among underlying proposed dynamics is an important step before estimating long term model parameters. Thus, this study follows Westerlund (2007) cointegration test because of its robustness and consistent estimates. Furthermore, this test is far superior to traditional cointegration methods and addresses important issues of cross-sectional dependence and slope heterogeneity by assuming panel dynamics in series with first differences and integrating individual-specific effects and lagged differences. The following equations related to the Westerlund method can be used to measure long-run cointegration.

12

is the regression intercept,  = ( ) reflects the deterministic trend; i and t are the total cross-section and time period respectively.

The following equations reflect the test statistics associated with the Group Westerlund test of cointegration

13

14

where the test statistic associated with the Panel Westerlund test can be proved by the following equation

15

16

where measures the adjustment speed to restore long-term equilibrium after a short-term imbalance.

Cross-sectional augmented autoregressive distributed lag (CS-ARDL) method

The study further adopts the CS-ARDL method proposed by Chudik and Pesaran³¹ to reveal the connection between exogenous variables and fossil fuel energy demand (FFE) and renewable energy generation (REG) in the dynamic model. The CS-ARDL method is more efficient and robust than other estimation methods because it addresses issues such as cross-sectional dependence, endogeneity, heterogeneous slope coefficients, and unobserved common factors^100,101. If unobserved common components are overlooked, the results of the analysis may be erroneous. The general equation of the CS-ARDL method can be highlighted as:

17

where represents the lagged dependent variable, is an independent factor that is purely stationary at I(0) or I(1), or mixed stationary for group i, the scalar signifies the coefficient of the lagged dependent variable, denotes the slope coefficient of the explanatory variable, and ωi expresses the group-specific fixed effect error term. Additionally, p and q, and μ represent the cross sectional average lag length, and error term respectively. The long-term CS-ARDL can be depicted by Eq. (18), while the average group coefficient can be described by Eq. (19).

18

19

Robustness testing

Commonly correlated effect mean groups (CCEMG) and augmented mean groups (AMG) are two other second-generation most effective strategies for addressing non-stationary dynamics glitches, predicted by Pesaran¹⁰² and Teal and Eberhardt¹⁰³, respectively. Different from traditional methods, the Commonly Correlated Effects Mean Group (CCEMG) estimator has significant advantages in panel data analysis by effectively addressing slope heterogeneity and cross-sectional dependence. This method calculates the average of the determinants across all cross-sections to eliminate spillover effects due to cross-sectional dependence while omitting trends. In addition to tackle the issues of cross-sectional dependence, slope heterogeneity, and structural fractures, the CCEMG and AMG procedures can also incorporate year-specific adjustments. Consequently, these methods are more robust and adaptable when dealing with unobserved common components^104,105.

Methods for detecting causal relationships between panel variables

Dumitrescu and Hurlin’s³² causal test, an updated version of Granger’s¹⁰⁶ non-causal test, can also be used in the current study to reveal causal relationships between panel variables. The approach is suitable for studies analyzing relationships between variables, especially in the presence of heterogeneity, because it can account for two key dimensions: heterogeneity in the causal relationships themselves (different individuals or groups may have different causal patterns) and heterogeneity in the data generating process (the way data are generated may differ across individuals or groups). This procedure effectively addresses cross-sectional dependence and slope heterogeneity issues and is known for using the Wald statistic of Granger causality to examine cross-sectional dependence and heterogeneous panels. The linear model for the test of causality is highlighted in the following equation.

20

where y and x are stationary series of N individuals within T periods, as shown in Eq. (20). In addition, , , and K in the above linear model denote the intercept fixed effect, slope coefficient and lag length, respectively. The null hypothesis of this approach proposes that there is no causal connection between the panel variables and can be tested against an alternative hypothesis that there is a causal association in at least one cross-sectional unit. The statistics and statistics as described below can be used to test the null hypothesis.

21

22

Empirical findings and interpretation

Cross-sectional dependence and heterogeneous slope results

In today’s context of technological development, globalization, and trade, cross-sectional dependence (CD) does exist among newly industrialized countries. This shows that the economic indicators and dynamics of these countries are interconnected and influence each other, rather than being unrelated. The problem of cross-section dependence has a great impact on the results of panel unit root tests and long-term parameter estimation, so this issue must be considered in the regression analysis. The results of analysis of panel data may be biased or spurious by neglecting the detection of cross-sectional dependence of unobserved common variables^107,108. The results of the Pesaran scaled LM test, the Pesaran cross-sectional dependence (CSD) method, and the Breusch and Pagan LM technique support the existence of cross-sectional correlations based on the significance of the entire dynamics in the model, as shown in Table 2. It turns out that there are deep-rooted interactions among the countries selected for study due to economic and financial shocks. Furthermore, the presence of slope coefficient heterogeneity is confirmed at the 1% significance level, as the slope homogeneity associated with the null hypothesis is rejected and the alternative hypothesis of slope heterogeneity is accepted, as shown in Table 3.

Table 2.

Findings of detecting cross-sectional dependence.

Variables	Breusch-Pagan LM		Pesaran Scaled LM		Pesaran CSD
	Coefficients	P-value	Coefficients	P-value	Coefficients	P-value
(FFE)ln	894.01***	0.000	52.02***	0.005	93.04***	0.000
(EI)ln	734.77***	0.002	56.88***	0.003	50.80***	0.001
(REG)ln	980.16**	0.051	70.29***	0.002	84.95***	0.002
(EC)ln	721.24***	0.004	83.06***	0.001	75.44***	0.003
(TI)ln	653.94***	0.002	71.25**	0.045	66.34***	0.004
(IND)ln	753.19***	0.004	77.25***	0.002	59.27***	0.005

Statistics highlighted with *** signify a significance level of 1%, while statistics highlighted with ** denote a significance level of 5%.

Table 3.

Slope heterogeneity results from Pesaran and Yamagata methods.

Models	Test-statistics	Value	p-value
Model-EI		74.85***	0.008
		59.04***	0.005
Model-FFE		60.75***	0.003
		70.14***	0.001
Model-REG		76.90***	0.008
		79.48***	0.001

Statistics highlighted with *** signify a significance level of 1%, while statistics highlighted with ** denote a significance level of 5%.

Panel unit root and long term cointegration results

The results of the CADF and CIPS unit root tests are shown in Table 4, reflecting that for newly industrialized countries (NICs), the entire dynamics of the series such as energy intensity, fossil fuel energy, renewable energy generation and economic complexity, technological innovation and industrial added value are non-stationary at the level, but can be transformed into stationary by taking the first-order differential. Thus, it is established that the panel variables in the model are stationary at the first order differential.

Table 4.

Results from CADF and CIPS panel unit root techniques.

Dynamics	CADF		First difference	p-value	CIPS		First difference	p-value
	Level	p-value			Level	p-value
(FFE)ln	-1.548	-0.142	-2.297***	-0.002	-3.836	-0.176	-4.283***	-0.003
(EI)ln	-1.475	-0.106	-3.576***	-0.004	-3.648	-0.150	-4.427***	-0.007
(REG)ln	-1.196	-0.154	-2.809***	0.001	-3.148	-0.187	-3.879***	-0.002
(EC)ln	-1.649	-0.191	-3.837***	0.000	-2.197	-0.168	-3.276***	-0.004
(TI)ln	-1.342	-0.153	-3.069***	0.001	-2.782	-0.135	-4.340***	0.001
(IND)ln	-1.349	-0.149	-2.824***	0.000	-2.620	-0.152	-3.039***	0.001

Statistics highlighted with *** signify a significance level of 1%, while statistics highlighted with ** denote a significance level of 5%.

Next, the panel cointegration approach of Westerlund (2007) is applied to examine the long-run cointegration associations among panel dynamics in the model. The results in Table 5 indicate that there is a long-term relationship based on the significance of the group statistics and one panel statistic in the proposed three different models, reflecting the rejection of the null hypothesis that there is no cointegration between the panel dynamics. After confirming the long-run cointegration among the panel dynamics, the long-run parameters can be estimated using the CS-ARDL approach along with AMG and CCEMG methods to ensure the robustness of the results.

Table 5.

Long-run cointegration results of Westerlund (2007) method.

	Model-EI	Model-FFE	Model-REG
Gt	-5.310***	-4.337***	-5.073***
Ga	-4.872***	-5.995***	-4.328***
Pt	-6.436	-5.025	-6.602
Pa	-5.759***	-4.837**	-5.756**

Statistics highlighted with *** signify a significance level of 1%, while statistics highlighted with ** denote a significance level of 5%.

Long-term and short-term coefficient estimation results

After validating the existence of long-run cointegration linkages between energy intensity, fossil fuel energy, renewable energy generation, and regressors, the next step is to reveal the long-run parameter estimates. This study adopts the CS-ARDL approach to reveal the short- and long-term dynamic linkages between energy intensity, fossil fuel energy, renewable energy generation, and exogenous variables. The short- and long-term parameter estimation results reported in Table 6 indicate that a 1% expansion in economic complexity can lead to significant reductions in energy intensity and fossil fuel energy by 0.276% and 0.239% respectively, while boosting renewable energy generation by 0.338% in the long term. However, in the short term, economic complexities have an insignificant impact on energy intensity, fossil fuel energy and renewable energy generation in newly industrialized countries. This result is highly innovative and represents a significant contribution to literature. Moreover, economic complexity can provide leverage for reducing demand for fossil fuel energy and provide policymakers with a route to follow when designing regional and national economic expansion and energy and environmental sustainability strategies. Economic complexity can help compress environmental contamination through diversified product mix and technological advances in well-managed production systems. The consequences of economic complexity squeeze demand for fossil fuel energy, promote energy efficiency and the use of renewable energy consistent with Adekoya et al.⁶⁸, Can et al.⁶⁷, Fang et al.⁴¹, Dogan et al.⁴⁰, Rafique et al.⁶⁹, Can and Ahmed⁷⁰.

Table 6.

Results of the CS-ARDL method for short-term and long-term coefficient estimation.

Variables	Long-term			Short-term
	Coefficients	p-values	Std. Error	Coefficients	p values	Std. Error
EI = f(EC, TI, GDP, IND)
lnEC	-0.276***	-0.000	0.232	-0.297	-0.118	0.135
lnTI	-0.291***	-0.001	0.102	-0.203*	-0.079	0.187
lnGDP	0.289***	0.000	0.054	-0.129**	-0.035	0.108
lnIND	0.208**	0.019	0.114	0.326**	0.038	0.094
C	0.398***	0.006	0.109	0.180***	0.009	0.087
ECM(-1)				-0.198***	-0.000	0.115
FFE = f(EC, TI, GDP. IND)
lnEC	-0.239***	-0.003	0.286	-0.116	-0.134	0.164
lnTI	-0.293***	-0.002	0.183	-0.165**	-0.048	0.027
lnGDP	0.438***	0.009	0.225	0.238**	0.067	0.195
lnIND	0.383***	0.006	0.176	0.225*	0.074	0.043
C	0.445		0.169	0.331	0.239	0.158
ECM(-1)				-0.763***	-0.006	0.193
REG = f(EC, TI, GDP, IND)
lnEC	0.338***	0.004	0.276	-0.229	-0.149	0.198
lnTI	0.343***	0.003	0.198	0.278*	0.095	0.165
lnGDP	-0.304***	-0.003	0.032	-0.129**	-0.036	0.105
lnIND	-0.287***	-0.009	0.198	0.298	0.118	0.073
C	0.353***	0.003	0.185	0.117***	0.006	0.043
ECM(-1)				-0.140***	-0.003	0.190

Statistics highlighted with *** signify a significance level of 1%, while statistics highlighted with ** denote a significance level of 5%.

The technological innovation coefficient for energy intensity and fossil fuel energy are significantly negative, while for renewable energy power generation is significantly positive, confirming that economic complexity strongly condenses energy intensity and fossil fuel energy, while contributes to renewable energy power generation in both the short and long term. Over the long term, a 1% upsurge in technological innovation can considerably diminish energy intensity and fossil fuel energy by 0.291% and 0.293% respectively, while stimulate renewable energy generation by 0.343%. Similarly, in the short term, every 1% upsurge in technological innovation can sensitively shrink energy intensity and fossil fuel energy by 0.203% and 0.165% respectively, while expanding renewable energy power generation by 0.278%. The negative correlation between technological innovation, energy intensity and fossil fuel energy and the progressive link between technological innovation and renewable energy generation are understandable given that technological progress is critical for newly industrialized countries to assess and implement sustainable development. In addition, technological innovation can improve energy efficiency by reducing the use of fossil fuel energy and promoting renewable energy, thereby mitigating environmental pollution. Discoveries of economic complexity and technological innovation in energy and environmental sustainability are unique and promising. Technological innovation also helps reduce environmental contamination, as it is a major source of innovative production and economic complexity. Studies include Chu⁷¹, Raihan, and Tuspekova⁷², Zhang et al.⁷³, Suki et al.⁷⁵, Kiani et al.⁷⁶ supporting improvements in environmental quality through technological innovation consistent with current research.

The gross domestic product (GDP) coefficient is statistically significant, positive for energy intensity and fossil fuel energy, while negative for renewable energy generation. In the long run, for every 1% upsurge in GDP, energy intensity and fossil fuel power generation can considerably promote by 0.289% and 0.438% respectively, while renewable energy power generation can robustly decrease by 0.304%. Likewise, in the short term, GDP also strongly promotes energy intensity and fossil fuel energy, while reducing the use of renewable energy. A progressive link between economic expansion, energy intensity and fossil fuel energy seems feasible because newly industrialized countries are developing rapidly and their GDP has grown sharply in recent decades. Energy demand in newly industrialized countries is increasing with the economic expansion, which accelerate the releasing of carbon dioxide. In addition, energy demand is growing faster than the supply of clean energy, so fossil fuels are expected to meet the growth in energy demand in newly industrialized countries. For energy intensity and fossil fuel energy, the coefficient of industrial value added is significant and positive, but it is not conducive to renewable energy generation in the long run. In the long run, every 1% increase in industrial added value can effectively promote 0.208% of energy intensity and 0.383% of fossil energy power generation, while reducing renewable energy power generation by 0.287%. In the short term, every 1% increase in industrial added value can effectively drive the growth of energy intensity and fossil fuel energy by 0.326% and 0.225% respectively, while the impact on renewable energy power generation is insignificant. Higher industrial growth and manufacturing value added in newly industrialized countries may increase energy intensity and demand for fossil fuel energy, while inhibiting renewable energy generation, making them more vulnerable to environmental contamination. This finding demonstrates the unique and special role of economic growth and industrial value added in promoting fossil fuels and renewable energy. The negative significant error term (ECM-1) verifies the existence of long-term correlations among panel variables, thus supporting the long-term convergence of the model.

Robustness check

The long-term coefficient estimation results of the above CS-ARDL method can be verified using the augmented mean group (AMG) and common correlation effect mean group (CCEMG) methods. The results of the AMG and CCEMG methods are in good agreement with the results of CS-ARDL method reported in Table 7, reflecting the significance of economic complexity and technological innovation in mitigating energy intensity and fossil fuel energy demand while strongly promoting renewable energy generation. Economic growth and industrial value added have meaningfully increased energy intensity and demand for fossil fuel energy, while seriously hampering the generation of renewable energy in newly industrialized countries in the long term.

Table 7.

Results of the AMG and CCEMG method for long-term coefficient estimation.

Variables	AMG		CCEMG		p-values	Std. error
	Coefficients	p-values	Std. Error	Coefficients
EI = f(EC, TI, GDP, IND)
lnEC	-0.219***	-0.009	0.218	–0.228**	-0.039	0.107
lnTI	-0.207***	-0.007	0.116	–0.259***	-0.002	0.183
lnGDP	0.180***	0.003	0.095	0.107***	0.001	0.159
lnIND	0.279***	0.002	0.181	0.279**	0.024	0.084
C	0.223***	0.008	0.118	0.280***	0.008	0.039
RMSE	0.086				0.082
FFE = f(EC, TI, GDP. IND)
lnEC	-0.399***	-0.004	0.197	–0.383**	-0.057	0.186
lnTI	-0.302***	-0.003	0.184	-0.323***	-0.009	0.039
lnGDP	0.483***	0.001	0.303	0.403**	0.048	0.189
lnIND	0.484***	0.006	0.154	0.447**	0.042	0.074
C	0.379	0.181	0.253	0.350	0.179	0.112
RMSE	0.079				0.078
REG = f(EC, TI, GDP, IND)
lnEC	0.179***	0.005	0.276	0.318**	0.049	0.198
lnTI	0.184***	0.005	0.198	0.213***	0.009	0.165
lnGDP	-0.364***	-0.002	0.053	–0.386***	-0.009	0.183
lnIND	-0.350***	-0.005	0.117	0.337**	0.028	0.075
C	0.373***	0.004	0.181	0.332***	0.009	0.089
RMSE	0.097				0.092

Statistics highlighted with *** signify a significance level of 1%, while statistics highlighted with ** denote a significance level of 5%.

Bivariate causality results

The current study also uses the DH causality method to examine the causal associations between energy intensity (EI), fossil fuel energy demand (FFE), renewable energy generation (REG) and important regressors. The findings reported in Table 8 support the feedback hypothesis between fossil fuel energy (FFE) and GDP, as both GDP and fossil fuel energy have a bilateral causal relationship in newly industrialized countries. In addition, there is a bidirectional causal relationship between energy intensity and technological innovation; and between economic complexity and renewable energy generation. The analysis results also prove the unidirectional causal relationship from fossil fuel energy to economic complexity; from industrial value added to fossil fuel energy; from renewable energy generation to technological innovation, economic growth and industrial value added.

Table 8.

DH panel causality results between panel variables.

Variables	W-statistics	Zbar-statistics	Probability	Inferences
lnFFE ⇎ lnEC	1.392***	3.643	0.002	FFE → EC
lnEC ⇎ lnFFE	1.403	3.154	0.158
lnEI ⇎ lnTI	2.574***	0.173	0.005	EI ↔ TI
lnTI ⇎ lnEI	2.402***	3.547	0.003
lnFFE ⇎ lnGDP	3.524***	4.364	0.004	GDP ↔ FFE
lnGDP ⇎ lnFFE	1.079***	0.570	0.002
lnFFE ⇎ lnIND	2.520	5.651	0.198	IND → FFE
lnIND ⇎ lnFFE	2.405***	0.529	0.004
lnREG ⇎ lnEC	1.509***	0.476	0.009	REG ↔ EC
lnEC ⇎ lnREG	2.753***	2.409	0.007
lnREG ⇎ lnTI	1.402***	0.908	0.001	REG → TI
lnTI ⇎ lnREG	3.309	4.446	0.175
lnREG ⇎ lnGDP	3.635***	3.598	0.005	REG → GDP
lnGDP ⇎ lnREG	2.408	4.569	0.289
lnREG ⇎ lnIND	3.526***	4.549	0.008	REG → IND
lnIND ⇎ lnREG	1.508	0.437	0.140

Statistics highlighted with *** signify a significance level of 1%, while statistics highlighted with ** denote a significance level of 5%. → , ↔ and ⇎ respectively represent unidirectional causation, bilateral causation and no homogeneous cause.

Conclusion and policy recommendations

Increased industrial activities, urbanization, globalization and rapid economic development have expanded the demand for energy in newly industrialized countries, prompting fossil fuels to replace renewable energy sources in a cheaper and more accessible way, leading to environmental pollution. Therefore, this study adopts the CS-ARDL approach to reveal how economic complexity and technological innovation affect the demand for fossil fuel energy and renewable energy generation in newly industrialized countries from 1990 to 2022, and uses the AMG and CCEMG methods for robustness checks. The results show that economic complexity and technological innovation reduce energy intensity and the demand for fossil fuel energy, which means that economic complexity and technological innovation in newly industrialized countries helps improve environmental quality by reducing the use and dependence on fossil fuel energy. This result suggests that economic complexity and technological innovation may have energy-depressing properties in newly industrialized countries. However, economic complexity and technological innovation have stimulated the development of energy efficiency, thereby improving environmental quality by promoting the production of clean energy. This empirical result clearly shows that the amount of green production technology and knowledge can increase with the development of technological innovation and economic complexity, thereby improving green output and environmental sustainability. GDP has a strong incremental impact on energy intensity and demand for fossil fuel energy while significantly reducing renewable energy generation in the short and long term. Industrial value added also has a strong incremental impact on energy intensity and fossil fuel energy, while a significant adverse impact on renewable energy generation. The impact of industry on energy intensity and fossil fuel energy is both short- and long-term, while the impact of industry on renewable energy generation is only long-term. The results of the causality test support the feedback hypothesis between fossil fuel energy (FFE) and GDP because there is a bilateral causal relationship between GDP and fossil fuel energy in newly industrialized countries. In addition, there is a bidirectional causal relationship between energy intensity and technological innovation; and between economic complexity and renewable energy generation. The analysis results also prove the unidirectional causal relationship from fossil fuel energy to economic complexity; from industrial value added to fossil fuel energy; from renewable energy generation to technological innovation, economic growth and industrial value added.

In order to improve the economic sector structure in an environmentally friendly manner, policymakers should adopt new strategies based on the above empirical analysis. These goals can be achieved by developing strategies to encourage economic activity, such as boosting the share of other inputs in production processes relative to fossil fuel energy, which promotes the complexity of achieving energy savings. Furthermore, promoting economic sophistication and technological innovation in newly industrialized countries can help industry shift from current energy-intensive manufacturing methods to less energy-intensive technologies, thereby reducing energy intensity and reliance on fossil fuels and increasing reliance on renewable energy. Furthermore, strategically speaking, newly industrialized countries face common energy and environmental challenges that should be addressed by encouraging cooperation and knowledge sharing among them. Selected groups of countries can work together through regional initiatives to pool resources and expertise to address common issues related to energy conservation and sustainable development.

Newly industrialized countries can reduce their dependence on fossil fuels and promote growth and environmental sustainability by seriously considering the transition to renewable and clean energy in their structural transformation. To achieve this, these countries should support the development of emerging industries by investing heavily in green technology, infrastructure and research and development. Newly industrialized countries can pave the way for a more sustainable and resilient economic and environmental future through energy transition measures focused on producing renewable energy through economic maturity and technological innovation.

The main limitation of the current study is that it focuses only on newly industrialized countries. In this regard, the view in this study that promoting energy transformation and green growth through developing economic complexity and technological innovation is limited to newly industrialized countries and is worth exploring for both developed and developing countries. Thus, future research on developed and developing economies could consider a more comprehensive insight into the energy-conservation green characteristics of economic complexity and technological innovation.

Author contributions

Yutong Li and Yu Zirui contributed to the conceptualization of the study, data curation, methodology used, analysis and final validation. Jie Mao contributed to study revision, literature search and approved the final revision.

Data availability

The panel variable data used in the analysis can be accessed through the following link: https://databank.worldbank.org/source/world-development-indicators. Although, data on technological innovation measured in total patents can be retrieved from http://oe.cd/ipstats.

Competing interests

The authors declare no competing interests.