Capital Formation and Oil Consumption Drive CO₂ Emissions in Ecuador: Evidence from an ARDL Model in Log-First Differences

Abstract

This study investigates the impact of key economic variables on carbon dioxide (CO₂) emissions in Ecuador within the broader context of sustainable development. Annual data from 1990 to 2022 are analyzed using an Autoregressive Distributed Lag (ARDL) model in first logarithmic differences, estimated via Ordinary Least Squares (OLS). The model examines both short- and long-term relationships between CO₂ emissions and three core macroeconomic indicators: gross fixed capital formation (GFCF), GDP per capita, and oil consumption. Descriptive analysis reveals substantial variation in investment and fossil fuel use across the study period. Empirical findings indicate that oil consumption has a positive and statistically significant effect on emissions, while GFCF exhibits a significant negative association in the current period, suggesting the role of cleaner or more efficient investment. Lagged GDP per capita shows a negative effect on emissions, partially supporting the Environmental Kuznets Curve hypothesis. Although renewable energy is discussed in the conceptual framework, it is not included in the current empirical specification—a limitation that will be addressed in future model extensions. The results provide empirical support for directing investments toward low-carbon sectors and accelerating the energy transition, particularly in transport and industry.

1. Introduction

Environmental degradation remains among the most urgent global challenges of our era, with multifaceted economic, social, and health-related impacts that have intensified over recent decades. This phenomenon has been examined from different disciplinary perspectives, particularly in the field of economics, since a country’s policy decisions—whether environmental or trade-related—can significantly influence carbon dioxide (CO₂) emissions and are thus essential for ensuring sustainable economic development [1]. It is important to recall that the Earth’s atmosphere contains a wide array of gases and suspended particles. Although nitrogen and oxygen are the most abundant, they are not inherently harmful. There are others present in smaller concentrations—known as greenhouse gases, such as CO₂—that play a critical role in regulating the planet’s thermal balance, the excessive accumulation of which poses a severe threat to the environment [2]. In fact, the increasing concentration of CO₂ resulting from hydrocarbon combustion has triggering adverse effects at multiple scales: ocean acidification, ozone layer depletion, biodiversity loss, and land use changes, among others. While variations in these factors may occur naturally, anthropogenic greenhouse gas emissions have accelerated such processes to alarming levels. A particularly concerning aspect of rising atmospheric CO₂ is its impact on public health, as the associated global warming contributes to the spread of diseases and the intensification of extreme weather events that endanger human well-being [3].

The primary source of CO₂ emissions is closely linked to fossil energy consumption. Specifically, the combustion of fuels such as natural gas, coal, and oil accounts for the vast majority of energy-related CO₂ emissions. Approximately 93% of energy-related CO₂ emissions come directly from burning hydrocarbons [4]. Each of these fossil fuels contributes to different sectors: natural gas and coal are primarily used for electricity generation, whereas oil remains an essential driver of the global transportation sector. In contrast, promoting the consumption of renewable energy emerges as an effective strategy to halt environmental degradation and mitigate CO₂ emissions. A variety of clean energy sources—solar, wind, oceanic, and geothermal, among others—exist and, with appropriate technology and policy frameworks, can deliver energy services as efficiently as non-renewable sources. Even biofuels may contribute to a cleaner energy mix, provided that their production and use are environmentally sustainable [5]. The transition from a fossil fuel-based energy matrix to one grounded in renewable sources is widely regarded as a pivotal element in addressing the climate crisis.

Industrialization and urbanization also play a decisive role in the generation of CO₂. Recent studies indicate that higher levels of industrial and urban activity are associated with increased emissions of this gas. Ray et al. [6] analyzed selected Asian economies and found a positive and significant relationship between urbanization, industrial advancement, and urban population growth with rising CO₂ emissions. Similarly, in BRICS (Brazil, Russia, India, China, and South Africa), it has been observed that the intensive use of polluting energy sources (measured through CO₂ emissions) contributes to national income growth, thereby hindering reductions in such emissions—particularly when energy consumption constitutes a substantial share of the GDP [7]. A comparable phenomenon occurs even in advanced economies: Sajjad et al. [8], using a quantile regression model of moments to study the G7 countries, conclude that economic growth is accompanied by a significant increase in CO₂ emissions. These findings show the complexity of balancing economic prosperity and environmental sustainability across diverse development contexts.

Regarding the recent evolution of global CO₂ emissions, available data reveal troubling trends. Globally, CO₂ emissions exhibited a sustained upward trajectory until 2018, reaching approximately 35,560,556 kilotons (kt) in that year. Subsequently, in 2020 (the latest year for which data are available), total emissions recorded a decline of approximately 5.94%, largely attributable to the economic slowdown caused by the COVID-19 pandemic [9]. However, this temporary reduction does not address the underlying problem, and CO₂ concentrations remain at historically high levels. In the case of Ecuador, CO₂ emissions have shown a considerable increase over recent decades. According to World Bank data, national CO₂ emissions have followed a sustained upward trend, reaching concerning levels in recent years [10]. This trend shows that protecting the environment is not a priority at the national level yet, highlighting the urgent need to assess the factors driving such emissions. Identifying the economic determinants of CO₂ emissions is essential for designing effective policy solutions that enable the transition toward more sustainable development.

A wide range of empirical studies, both at the global and regional levels, have examined the determinants of CO₂ emissions [11,12]. The literature highlights four key variables that directly influence a country’s carbon footprint: (1) gross fixed capital formation (investment) [13], (2) per capita income (as a proxy for the level of economic development) [14], (3) fossil fuel-based energy consumption (particularly oil, as a proxy for non-renewable energy) [11], and (4) renewable energy consumption [12]. In the following, we draw upon the latest empirical findings:

• Investment, proxied by gross fixed capital formation (GFCF), is typically linked to increased economic and industrial activity, which tends to increase emissions. Rahman and Ahmad [13], using a Nonlinear Autoregressive Distributed Lag (NARDL) model applied to Pakistan, found that a 1% increase in GFCF leads to an approximate 0.58% rise in CO₂ emissions, indicating a positive and significant impact of investment on pollution levels. However, other studies suggest that this effect may vary depending on the context: Adebayo and Kalmaz [15], analyzing the determinants of emissions in Egypt, found that GFCF did not exert a statistically significant influence on CO₂ emissions, whereas energy consumption displayed a positive and significant relationship with environmental degradation in that country. These differences suggest that the link between investment and pollution may depend on how investment is allocated (e.g., towards cleaner or more carbon-intensive sectors) and on the specific conditions of each economy. In fact, Oncu et al. [16], using a global panel dataset, report that while increases in GFCF can raise CO₂ emissions in the short run, the long-term effect of capital accumulation on emissions is often statistically insignificant. This suggests that without environmental regulations guiding investment towards low-carbon technologies, the pollution impact of new capital formation may be muted or vary widely across countries.
• In this context, economic growth, typically measured through GDP per capita, plays a dual role in the environmental discourse. On the one hand, higher income levels generally imply greater consumption of energy and goods, driving emissions during the early stages of development. A study on Southern European countries found a long-term positive relationship: a 1% increase in GDP per capita leads to an estimated 1.77% rise in CO₂ emissions [17]. On the other hand, at more advanced stages of development, economies may have access to cleaner technologies and the capacity to implement stricter environmental regulations, potentially decoupling economic growth from emissions (Environmental Kuznets Curve hypothesis). Wang et al. [18] provide evidence in this regard by reexamining the impact of Foreign Direct Investment (FDI) on carbon emissions under different income levels: their results indicate that when a country’s GDP per capita is relatively low (below approximately USD 541.87), FDI tends to significantly increase CO₂ emissions; however, once the GDP per capita exceeds around USD 46,515, FDI’s influence becomes negative, contributing to emission reductions. Provided that income reaches a sufficient threshold to support the adoption of clean technologies and green policies, economic development may supply the tools necessary to mitigate pollution. Furthermore, recent global research underscores that the growth–emissions relationship might not simply stabilize at high income levels but could eventually intensify again. For instance, Wang et al. [19], analyzing a panel of 214 countries, identified an N-shaped pattern in the income–emissions trajectory: while emissions initially decline after a turning point (consistent with the EKC hypothesis), they begin to rise again once GDP per capita crosses a second high threshold (around USD 73,000). This finding suggests that even in very-high-income economies, continuous growth without additional mitigation measures can lead to reaccelerating CO₂ emissions, highlighting the need for ongoing innovation and policy vigilance. In the case of Ecuador, Borja et al. [20] explored the relationship between economic development and sectoral CO₂ emissions. Using data from 1990 to 2018 and Granger causality tests, the authors found no significant causal relationship between growth (measured by sectoral GDP per capita) and CO₂ emissions in the agriculture and transport sectors. However, in the industrial and services sectors, a long-term causal link was identified. This implies that increased economic activity in Ecuador’s industry and services sectors has been accompanied by higher emissions, in contrast to other sectors where traditional practices or efficiencies may be moderating this relationship.
• The use of non-renewable energy—particularly that derived from oil—has been confirmed as a direct driver of CO₂ emissions. Different studies support the hypothesis that higher fossil fuel consumption leads to environmental degradation. Apergis et al. [21] conducted a study focused on Uzbekistan to estimate the impact of renewable and non-renewable energy consumption on CO₂ emissions, using an Autoregressive Distributed Lag (ARDL) model. Their results indicate that hydroelectric energy consumption (as a renewable source) has a significant negative effect on CO₂ emissions—that is, it contributes to their reduction—whereas consumption of non-renewable sources such as natural gas and oil exerts a significant positive effect, increasing emissions in both the short and long run. However, the magnitude of each fossil fuel’s contribution may vary depending on the specific context. In the case of India, for instance, fossil fuels collectively have a detrimental effect on environmental sustainability, though with notable distinctions: coal is found to be the most harmful fuel in terms of CO₂ emissions, followed by oil, while natural gas consumption has a relatively smaller impact [22]. This finding suggests that although oil consumption clearly contributes to emissions, its effect may be less severe than that of coal, partly due to differences in carbon intensity and fuel efficiency. Similarly, recent evidence from Pakistan indicates that an energy mix heavily dominated by fossil fuels has led to substantial CO₂ growth, whereas the country’s nascent renewable sector has so far had only a negligible effect on overall emissions. In any case, reducing dependence on all fossil fuels remains essential for mitigating climate change.
• Finally, the expansion of renewable energy plays a fundamental role as a mechanism for emissions mitigation. Empirical evidence indicates that higher consumption of clean energy is generally associated with lower CO₂ levels, although complementary conditions are often necessary to maximize this benefit. Traoré and Asongu [23] analyzed Sub-Saharan African countries using a panel VAR model and concluded that there is a negative and significant relationship between renewable energy use and CO₂ emissions: as the share of clean energy in the energy mix increases, emissions tend to decline. However, the impact of energy-related technologies and behaviors can be complex. Nghiem et al. [24] studied OECD countries and found contrasting results when considering certain technological indicators: higher mobile phone penetration appears to help mitigate CO₂ emissions (possibly due to efficiency improvements and digitalization), whereas greater internet usage is associated with increased emissions, highlighting that not all technological innovation automatically ensures a reduced carbon footprint. Additionally, research in emerging economies emphasizes the crucial role of institutional frameworks and public policy in enhancing the beneficial effect of renewables. Specifically, Abbas et al. [25] evaluated the BRICS countries and found that the development of renewable energy, by itself, is insufficient to curb emissions unless accompanied by appropriate market regulations and environmental innovation. In line with this perspective, Fu et al. [26] provide new evidence that renewable energy expansion yields significant CO₂ reductions only once certain development and efficiency thresholds are met. Analyzing 36 major emitting countries, they observe that a renewable energy strategy becomes effective only when a nation’s per capita GDP exceeds roughly USD 16,000 and when energy intensity and carbon intensity fall below specific limits; below those levels, the impact of renewables tends to be limited or even counterproductive due to technological constraints, fossil-dependent industrial structures, or weak environmental governance. Such findings underscore that the adoption of market-based policies, alongside the promotion of green innovation and broader economic structural changes, reinforces the effectiveness of renewables in reducing CO₂ emissions and helps steer countries toward greater environmental sustainability.

Given this context, the present study focuses on analyzing the impact of the aforementioned economic variables—investment (GFCF), economic growth (GDP per capita), fossil energy consumption, and renewable energy consumption—on CO₂ emissions within the Ecuadorian context. Using an econometric approach with recent data, the research aims to quantify the relationship between these variables and emission levels in order to identify the main drivers of rising CO₂ emissions in Ecuador and assess the extent to which clean energy and sustainable growth can help to mitigate the problem. The results of this study provide valuable input for the design of public policies that balance economic development goals with the urgent need to protect the environment.

This research’s novelty lies in applying an ARDL model to Ecuador’s context, allowing us to capture both short- and long-term drivers of CO₂ emissions—an approach not previously utilized for this country. Similar ARDL-based analyses in other emerging economies have underscored the dominant roles of energy use and economic activity in shaping emissions. For example, Apergis et al. [21] show that in Uzbekistan the increased consumption of fossil fuels (natural gas, oil) leads to significantly higher CO₂ emissions, whereas greater hydropower use helps reduce emissions in both the short and long run. Likewise, Durmaz et al. [27], examining Latin America’s largest developing economies, find that energy consumption consistently pushes emissions upward with no evidence of an EKC-type turning point. In Bangladesh’s case, Ar Salan et al. [28] report that industrial output and non-renewable energy consumption have a significant positive impact on CO₂ levels, while improvements in agricultural practices can actually help lower emissions. These comparative findings illustrate broad patterns across developing countries—notably, that fossil fuel-driven growth tends to exacerbate carbon emissions, whereas cleaner energy and efficiency gains can mitigate them—even though specific effects vary by nation. By introducing an ARDL analysis for Ecuador, this study fills a gap in the literature and enables a meaningful comparison with similar economies. The evidence obtained will provide fresh insights into how investment, economic growth, and the energy mix influence Ecuador’s carbon footprint, thereby offering valuable guidance for tailoring sustainable development policies in this country.

This paper is structured as follows: Section 2 describes the data used and the econometric methodology employed; Section 3 presents and analyzes the empirical results obtained; Section 4 discusses the findings, policy implications, and future perspectives; finally, Section 5 offers the study’s conclusions and provides recommendations for strengthening economic and environmental sustainability in the country.

2. Methodology

This research adopts a quantitative approach structured into three fundamental stages. First, an exploratory analysis is conducted based on a systematic review of previous studies that have examined the relationship between economic variables and CO₂ emissions, with the aim of establishing a robust conceptual and methodological foundation. Subsequently, in the second stage, a detailed descriptive analysis is performed using available macroeconomic and environmental data for Ecuador, highlighting patterns, trends, and relevant historical behaviors over the study period. Finally, in the third stage, an inferential analysis is carried out through an econometric model that quantifies the relationship between selected economic variables—GFCF, gross domestic product per capita (GDP per capita), and oil consumption—and CO₂ emissions in Ecuador. By applying this procedure, evidence-based findings are produced, and clear recommendations are prepared for public policy and corporate social responsibility actions that can improve the country’s environmental sustainability.

2.1. Variables and Data Collection

This analysis is based on macroeconomic and environmental data for Ecuador, covering the period from 1990 to 2022. The dependent variable in the study is total CO₂ emissions, measured in metric tons. The independent variables selected to evaluate their impact on emissions are GFCF, measured in USD at constant 2015 prices; average daily oil consumption, expressed in barrels; and gross domestic product (GDP) per capita at constant prices.

The data employed in this research were obtained from official and reliable sources, including the World Bank, Our World in Data, and the U.S. Energy Information Administration (EIA). Due to the large numerical magnitudes and inherent variability of these time series, all variables were transformed using the natural logarithm. This procedure, commonly applied in econometric studies, helps stabilize variance, linearize relationships among variables, and satisfy the fundamental assumptions of the linear regression model used in this research.

2.2. Descriptive Analysis of the Variables

To understand the behavior of the selected economic variables and their relationship with CO₂ emissions in Ecuador, a detailed descriptive analysis was conducted, as presented in

Table 1. Descriptive statistics of the variables used in the study (1990–2022).

Statistic	CO2 Emissions	GFCF	Oil Consumption	GDP per Capita
	(Metric Tons)	(USD 2015)	(kWh)	(USD Constant)
Mean	30,246,019.27	17,653,357,074.24	7,370,195	3568.17
Standard Deviation	8,577,191.68	7,031,140,207.63	9,197,633.00	499.13
Skewness	−0.155	0.153	0.248747	0.251
Kurtosis	1.766	1.475	1,821,244.00	1.530
Jarque–Bera	2.225	3.328	2,250,828.00	3.317
Probability	0.329	0.189	0.324518	0.190

.

Table 1 presents the main descriptive statistics of the study variables: CO₂ emissions, GFCF, oil consumption, and GDP per capita.

CO₂ emissions exhibit an annual mean of approximately 30.2 million metric tons, with a substantial standard deviation of about 8.6 million tons, indicating a high degree of dispersion over the period analyzed. The slightly negative skewness (−0.155) suggests a distribution marginally biased toward higher values, while the kurtosis (1.766) indicates a relatively flat (platykurtic) distribution compared to the normal distribution. The Jarque–Bera normality test (statistic: 2.225; p-value: 0.329) does not suggest significant deviations from normality.

Regarding GFCF, the data reveal an annual average of approximately USD 17.7 billion (at constant 2015 prices), with a high degree of variability, as indicated by the standard deviation of about USD 7 billion. The distribution shows a slight positive skewness (0.153) and a kurtosis value of 1.475, indicating a platykurtic shape. The Jarque–Bera test statistic is 3.328 (p = 0.189), suggesting statistical normality.

Per capita energy consumption from petroleum in Ecuador exhibits an annual mean of approximately 7.37 million kWh and a median of 7.28 million kWh, indicating a relatively symmetrical distribution, with values ranging from 6.09 to 9.22 million kWh and a standard deviation of 9.20 million kWh, reflecting appreciable interannual variability. The skewness is slightly positive (0.25), suggesting that a few years recorded atypically high consumption, while the kurtosis (1.82) denotes a distribution flatter than the normal. Finally, the Jarque–Bera test (p = 0.3245) fails to reject the null hypothesis of normality, indicating that the variable displays statistical behavior consistent with a stable consumption pattern, without persistent extreme deviations.

Finally, the GDP per capita presents an annual average of USD 3568.17 (constant prices), with a relatively low standard deviation (USD 499.13), indicating lower dispersion compared to the other variables. The distribution shows a positive skewness (0.251) and a kurtosis of 1.530, again indicating a platykurtic form. The Jarque–Bera test statistic (3.317; p = 0.190) confirms that the distribution does not significantly deviate from normality.

Overall, these descriptive statistics provide a solid foundation for the subsequent interpretation of the econometric model, facilitating a better understanding of the underlying dynamics and relationships between economic variables and CO₂ emissions in Ecuador.

Figure 1a reveals a generally increasing trend in carbon dioxide (CO₂) emissions in Ecuador from 1990 to approximately 2019, followed by a significant temporary decline in 2020 and subsequent recovery during the 2021–2022 period. This pattern reflects a combination of factors, including industrial expansion and increased reliance on fossil fuels, as well as the economic and social restrictions associated with the COVID-19 pandemic, which temporarily curtailed the country’s economic and energy activity [29].

In addition, historical data provided by the World Bank and reproduced by the YCharts portal indicate a notable drop in CO₂ emissions between 1993 and 1994 [9]. This decline coincides with the economic recession experienced in Ecuador during that period, marked by a sharp devaluation of the sucre, internal fiscal conflicts, and significant restrictions on credit access. These conditions led to a reduction in industrial activity and, consequently, in fossil fuel consumption [30].

Figure 1b illustrates the historical evolution of GFCF in Ecuador over the period 1990–2022, expressed in constant 2015 US dollars. GFCF is widely regarded as a key indicator of long-term economic growth, as it directly reflects the level of capital accumulation within an economy. The overall trend is upward, though marked by significant disruptions during key periods.

During the 1990s, capital investment levels remained relatively stable, fluctuating around USD 9 to 10 billion annually. However, a sharp and abrupt decline occurred in 1999, coinciding with the severe domestic financial crisis that culminated in the official dollarization of the Ecuadorian economy in 2000 [31].

From 2000 onward, GFCF levels exhibited a sustained recovery, with a particularly strong acceleration between 2005 and 2014. This period coincided with the boom in international oil prices, which resulted in a significant increase in Ecuador’s fiscal revenues. Consequently, a large share of these additional resources was allocated to strategic infrastructure projects, public investment, and state-led industrialization initiatives [32].

Between 2017 and 2019, investment showed a modest recovery, although it did not reach the historical peaks recorded in previous years. A subsequent sharp decline in GFCF occurred in 2020, primarily due to the global economic downturn caused by the COVID-19 pandemic, which directly impacted public investment, restricted credit access, and disrupted international supply chains [31].

In the following years (2021–2022), GFCF stabilized at approximately USD 22–23 billion, still below the peak observed in 2014. This stabilization reflects a partial recovery, constrained by ongoing domestic fiscal limitations and an international environment marked by economic uncertainty [30].

Figure 2a shows the historical trend of per capita oil consumption in Ecuador during the period 1990–2022, clearly reflecting the country’s economic dependence on this energy resource.

During the 1990s, per capita oil consumption remained relatively stable, ranging between 6000 and 7000 kWh annually. This pattern can be attributed to the limited energy diversification at the time, marked by heavy reliance on private land transportation and thermal electricity generation, particularly in regions not connected to the national grid [33,34].

Between 1999 and 2001, a sharp decline in per capita consumption is observed, coinciding with the severe economic and financial crisis that culminated in Ecuador’s official dollarization in 2000. This decrease was primarily driven by a contraction in household purchasing power, a drop in domestic consumption, and the widespread paralysis of productive activities [35].

From 2002 to 2013, per capita oil consumption experienced a steady recovery fueled by factors such as the expansion of the vehicle fleet, the growth of land transport, and the implementation of fuel subsidy policies. These measures—mainly financed through oil export revenues—greatly facilitated access to petroleum-based fuels such as diesel and gasoline [31].

In 2020, another notable reduction in per capita oil consumption occurred, largely due to mobility restrictions implemented in response to the COVID-19 pandemic. Lockdown measures significantly curtailed both public and private transportation, leading to a sharp decline in fuel demand. Finally, in 2021 and 2022, consumption gradually recovered, although it remained below the pre-pandemic peak levels [36].

Figure 2b illustrates the evolution of Ecuador’s gross domestic product (GDP) per capita over the period 1990–2022, reflecting the major economic fluctuations experienced by the country over the past three decades.

The initial period (1990–1998) was characterized by structural stagnation, marked by low productivity levels and virtually zero per capita growth, within a context of fiscal instability, inflation, and limited productive investment [37].

Beginning in 2001, following the official adoption of the U.S. dollar as the national currency, the country entered a phase of economic recovery and stabilization. This process was significantly reinforced between 2006 and 2014, coinciding with the boom in international oil and commodity prices. During this period, GDP per capita grew steadily, reaching an average annual growth rate of 2.8% [38,39].

However, starting in 2015, the Ecuadorian economy entered a new phase of deceleration, largely due to the sharp drop in oil prices, compounded by fiscal imbalances and external constraints. This downward trend was significantly exacerbated in 2020 by the COVID-19 pandemic, which triggered one of the most severe economic contractions in the country’s recent history. Real GDP per capita declined sharply due to widespread shutdowns of productive activities, a contraction in trade, and a deterioration in public finances [40,41].

The economic recovery observed in 2021–2022 was moderate, driven by the gradual reopening of the economy and the easing of health-related restrictions. Nonetheless, GDP per capita levels have yet to return to their pre-2015 values, highlighting a slower and more fragile post-pandemic growth trajectory.

2.3. Model Specification and Justification

This study aims to empirically determine the impact of specific economic variables on carbon dioxide (CO₂) emissions in Ecuador, taking as a methodological reference the work of Rahman and Ahmad [13], who analyzed the influence of GFCF, GDP per capita, squared GDP per capita, total coal consumption, and oil consumption on CO₂ emissions. Additionally, the present research builds upon the study conducted by Karedla et al. [42], who also examined the determinants of CO₂ emissions, incorporating further variables such as trade openness, measured through the total volume of imports and exports.

In this context, the present study extends prior contributions by applying an ARDL framework to the Ecuadorian case and focusing on three core macroeconomic determinants—gross fixed capital formation (GFCF), GDP per capita, and oil consumption—so as to capture both short- and long-term effects on CO₂ emissions. Therefore, the general relationship proposed between the explanatory variables and the dependent variable (CO₂ emissions) is formally specified by Equation (1).

$$ln\left({\mathrm{CO}}_{2t}\right)={\beta }_1+{\beta }_{2}ln\left({\mathrm{GFCF}}_t\right)+{\beta }_{3}ln\left({\mathrm{GDPpc}}_t\right)+{\beta }_{4}ln\left({\mathrm{Oil}}_t\right)+{\beta }_{5}t+U_t$$

(1)

In this model, ${\mathrm{CO}}_{2t}$ represents the carbon dioxide emissions generated during the study period t; ${\mathrm{GFCF}}_t$ denotes the gross fixed capital formation in year t; ${\mathrm{GDPpc}}_t$ refers to the GDP per capita in year t; ${\mathrm{Oil}}_t$ corresponds to oil consumption in year t; t is the linear time trend variable; and $U_t$ is the stochastic error term in period t.

The inspection of the time series graphs reveals a clear linear trend in the growth of these variables, which justifies the inclusion of the linear time trend (t) in the econometric model. The time trend allows us to capture structural or temporal effects influencing CO₂ emissions, which is essential for a comprehensive and accurate analysis.

The econometric model employed in this research is an Autoregressive Distributed Lag (ARDL) model, which utilizes logarithmic differences in the variables to satisfy the statistical assumptions required for econometric analysis. This methodological approach is strongly supported by the recent empirical literature. For instance, Rahman and Ahmad [13] applied an ARDL model to evaluate the economic impact on CO₂ emissions in Pakistan. Similarly, Mehmood [43] used this methodology to examine the influence of various economic variables on CO₂ emissions in South Asian economies such as Pakistan, India, Bangladesh, and Sri Lanka. The ARDL model is particularly suitable for time series data, as it allows for the simultaneous inclusion of both contemporaneous and lagged variables, thereby capturing dynamic short-term and long-term relationships among the variables under study.

For the estimation of the proposed model, the econometric software EViews (version 12) was used. This robust tool supports advanced time series analysis, facilitates the efficient inclusion of lagged variables, and enables the application of transformations such as logarithmic differencing, which are essential for stabilizing variance and linearizing nonlinear relationships.

This study employs the following ARDL Error Correction Model (ECM) specification to examine the dynamic and long-term relationships between carbon dioxide emissions and key macroeconomic variables. The econometric model specified in this study is formally defined in Equation (2).

$$\begin{array}{cc}\hfill \Delta ln{\left({\mathrm{CO}}_2\right)}_t& =\rho \left[ln{\left({\mathrm{CO}}_2\right)}_{t-1}-{\beta }_{1}ln{\left(\mathrm{GFCF}\right)}_{t-1}-{\beta }_{2}ln{\left(\mathrm{GDPpc}\right)}_{t-1}-{\beta }_{3}ln{\left(\mathrm{Oil}\right)}_{t-1}-c\right]\hfill \\ \hfill & +\sum _{i=1}^{p-1}{\psi }_i\Delta ln{\left({\mathrm{CO}}_2\right)}_{t-i}+\sum _{j=0}^{q_1-1}{\gamma }_{1j}\Delta ln{\left(\mathrm{GFCF}\right)}_{t-j}+\sum _{k=0}^{q_2-1}{\gamma }_{2k}\Delta ln{\left(\mathrm{GDPpc}\right)}_{t-k}\hfill \\ \hfill & +\sum _{m=0}^{q_3-1}{\gamma }_{3m}\Delta ln{\left(\mathrm{Oil}\right)}_{t-m}+u_t,\hfill \end{array}$$

(2)

where

• $\Delta$: The first-difference operator, used to transform the series and ensure stationarity.
• $ln{\left({\mathrm{CO}}_2\right)}_t$: Natural logarithm of (per capita) carbon dioxide emissions at time t.
• $ln{\left(\mathrm{GFCF}\right)}_t$: Natural logarithm of gross fixed capital formation (investment), a proxy for capital accumulation.
• $ln{\left(\mathrm{GDPpc}\right)}_t$: Natural logarithm of GDP per capita, capturing the effect of economic growth.
• $ln{\left(\mathrm{Oil}\right)}_t$: Natural logarithm of oil consumption, representing fossil fuel use in the energy mix.
• $\rho$: Speed-of-adjustment parameter, indicating how quickly deviations from the long-run equilibrium are corrected.
• ${\beta }_1,{\beta }_2,{\beta }_3$: Long-run elasticity coefficients; percent change in CO₂ emissions from a 1% change in the respective regressor.
• c: Intercept term in the long-run (level) relationship.
• ${\psi }_i,{\gamma }_{1j},{\gamma }_{2k},{\gamma }_{3m}$: Short-run dynamic coefficients measuring the immediate effects of changes in $\Delta ln{\left({\mathrm{CO}}_2\right)}_{t-i}$, $\Delta ln{\left(\mathrm{GFCF}\right)}_{t-j}$, $\Delta ln{\left(\mathrm{GDPpc}\right)}_{t-k}$, and $\Delta ln{\left(\mathrm{Oil}\right)}_{t-m}$, respectively (with summation ranges as given in Equation (2)).
• $u_t$: Stochastic disturbance term, assumed to be i.i.d.

As illustrated in Figure 3, the ARDL framework distinguishes between contemporaneous effects of $\Delta ln{\left(\mathrm{GFCF}\right)}_t$, $\Delta ln{\left(\mathrm{GDPpc}\right)}_t$, and $\Delta ln{\left(\mathrm{Oil}\right)}_t$ on $\Delta ln{\left({\mathrm{CO}}_2\right)}_t$, their lagged effects, the autoregressive component $\Delta ln{\left({\mathrm{CO}}_2\right)}_{t-1}$, and the linear time trend t.

2.4. Pre-Estimation Tests: Stationarity, Lag Selection, and Cointegration

To determine the order of integration of the variables in the econometric specification, Augmented Dickey–Fuller (ADF) unit root tests were performed under specifications with a constant trend and with a constant plus linear trend. As reported in

Table 2. ADF unit root tests (first differences) with intercept and linear trend.

Variable	Test Form (Exogenous Terms)	ADF Statistic	1% Crit. Val.	5% Crit. Val.	10% Crit. Val.	p-Value	Order
$ln\left({\mathrm{CO}}_2\right)$	Constant, Linear Trend	$-8.593153$	$-4.284580$	$-3.562882$	$-3.215267$	$0.0000$	$I\left(1\right)$
$ln\left(\mathrm{GFCF}\right)$	Constant, Linear Trend	$-5.423176$	$-4.296729$	$-3.568379$	$-3.218382$	$0.0006$	$I\left(1\right)$
$ln\left(\mathrm{GDPpc}\right)$	Constant, Linear Trend	$-4.372268$	$-4.284580$	$-3.562882$	$-3.215267$	$0.0081$	$I\left(1\right)$
$ln\left({\mathrm{Oil}}_{\mathrm{kWh},\mathrm{pc}}\right)$	Constant, Linear Trend	$-5.313438$	$-4.284580$	$-3.562882$	$-3.215267$	$0.0008$	$I\left(1\right)$

Notes: ADF tests are performed on first-difference series with intercept and linear trend. Critical values correspond to the selected specification. Order indicates the inferred integration order.

, all series—the logarithm of CO₂ emissions ($ln\left({\mathrm{CO}}_2\right)$); the logarithm of gross fixed capital formation ($ln\left(\mathrm{GFCF}\right)$); the logarithm of GDP per capita ($ln\left(\mathrm{GDPpc}\right)$); and the logarithm of per capita oil consumption in kilowatt-hours ($ln\left({\mathrm{Oil}}_{\mathrm{kWh},\mathrm{pc}}\right)$)—are stationary in first differences, i.e., first-order integration, $I\left(1\right)$. In each case, the ADF test statistics in first differences are more negative than the corresponding critical values at the 1%, 5%, and 10% significance levels (with p-values $<0.05$), allowing us to reject the null of a unit root. Since no series is integrated to an order of two, $I\left(2\right)$, the set of variables satisfies the fundamental requirement for the ARDL framework, which permits a mixture of $I\left(0\right)$ and $I\left(1\right)$ regressors but excludes $I\left(2\right)$ processes. This confirms the methodological suitability of ARDL for the subsequent estimation.

2.5. Post-Estimation Diagnostics

To validate the reliability of inference from the ARDL specification, we examine the residuals for heteroskedasticity and serial correlation.

Table 3. Post-estimation heteroskedasticity diagnostics (Breusch–Pagan–Godfrey).

Test	Null Hypothesis	Stat.	df	p-Val.	Conclusion
Breusch–Pagan–Godfrey (F-statistic)	Homoskedasticity	1.0359	(18, 10)	0.4967	Fail to reject $H_0$—no evidence of heteroskedasticity
Breusch–Pagan–Godfrey (Obs*$R^2$)	Homoskedasticity	18.8764	18	0.3995	Fail to reject $H_0$—no evidence of heteroskedasticity
Breusch–Pagan–Godfrey (scaled explained SS)	Homoskedasticity	2.3580	18	1.0000	Fail to reject $H_0$—no evidence of heteroskedasticity

Notes: Columns use fixed widths (3 cm, 2 cm, 1 cm, 1 cm, 1 cm, 3 cm) to prevent overflow. Tests were computed on residuals from the baseline ARDL specification. df = degrees of freedom. The null hypothesis ($H_0$) is homoskedasticity.

reports three variants of the Breusch–Pagan–Godfrey (BPG) test for heteroskedasticity (F-statistic, Obs*$R^2$, and scaled explained SS). In all cases, the associated p-values (0.4967, 0.3995, and 1.0000, respectively) are well above conventional significance thresholds (1%, 5%, and 10%). We therefore fail to reject the null hypothesis of homoskedasticity, indicating no statistical evidence of non-constant error variance in the baseline model.

Regarding serial correlation, the Durbin–Watson statistic is $\mathrm{DW}=2.15$, which lies close to the no autocorrelation benchmark of 2 and does not suggest first-order residual autocorrelation. Taken together, these diagnostics support the adequacy of the ARDL specification and the validity of the reported standard errors under the assumptions of constant variance and no first-order autocorrelation.

3. Results

This study examined the impact of various economic variables on CO₂ emissions in Ecuador. The independent variables considered in the analysis were GFCF, GDP per capita, and oil consumption. The methodological approach consisted of an ARDL econometric model using logarithmic differences, estimated with the EViews statistical software. This technique enables the identification of both immediate and lagged effects of economic variables on CO₂ emissions.

3.1. General Model Analysis

Table 4. Estimated coefficients of the ARDL model in first logarithmic differences.

Variable	Coefficient	Std. Error	t-Statistic	p-Value
C	$9.749580$	$2.255634$	$4.322324$	$0.0015$
@TREND	$0.021123$	$0.005728$	$3.687643$	$0.0042$
$\Delta ln{\left({\mathrm{CO}}_2\right)}_{t-1}$	$-1.107305$	$0.113750$	$-9.734534$	$0.0000$
$\Delta ln{\left({\mathrm{CO}}_2\right)}_{t-2}$	$-1.060674$	$0.114632$	$-9.252831$	$0.0000$
$\Delta ln{\left({\mathrm{CO}}_2\right)}_{t-3}$	$-0.497503$	$0.106321$	$-4.679246$	$0.0009$
$\Delta ln{\left(\mathrm{FBCF}\right)}_t$	$-0.126422$	$0.137672$	$-0.918282$	$0.3801$
$\Delta ln{\left(\mathrm{FBCF}\right)}_{t-1}$	$-0.339369$	$0.145944$	$-2.325341$	$0.0424$
$\Delta ln{\left(\mathrm{GDPpc}\right)}_t$	$1.512569$	$0.676064$	$2.237318$	$0.0492$
$\Delta ln{\left(\mathrm{GDPpc}\right)}_{t-1}$	$2.968316$	$0.771428$	$3.847823$	$0.0032$
$\Delta ln{\left(\mathrm{GDPpc}\right)}_{t-2}$	$1.668007$	$0.499278$	$3.340837$	$0.0075$
$\Delta ln{\left(\mathrm{GDPpc}\right)}_{t-3}$	$2.385040$	$0.511204$	$4.665533$	$0.0009$
$\Delta ln{\left({\mathrm{Oil}}_{\mathrm{kWh}}\right)}_t$	$0.470256$	$0.179830$	$2.614995$	$0.0258$
$\Delta ln{\left({\mathrm{Oil}}_{\mathrm{kWh}}\right)}_{t-1}$	$-0.424420$	$0.189103$	$-2.244391$	$0.0486$
$\Delta ln{\left({\mathrm{Oil}}_{\mathrm{kWh}}\right)}_{t-2}$	$-0.159278$	$0.231016$	$-0.689468$	$0.5062$
$\Delta ln{\left({\mathrm{Oil}}_{\mathrm{kWh}}\right)}_{t-3}$	$-1.158515$	$0.215810$	$-5.368218$	$0.0003$
$EC{M}_{t-1}$ (CointEq(-1)∗)	$-0.350695$	$0.082018$	$-4.275819$	$0.0016$
Model statistics
$R^2$	0.964262	Mean dependent var		0.018851
Adjusted $R^2$	0.923026	S.D. dependent var		0.169264
S.E. of regression	0.046961	Akaike info criterion		$-2.977895$
Sum squared resid	0.028669	Schwarz criterion		$-2.223525$
Log likelihood	5.917948	Hannan–Quinn criter.		$-2.741636$
F-statistic	2.338388	Durbin–Watson stat		2.150187
Prob(F-statistic)	0.000001

Notes: Numbers are shown in math mode to render the proper minus sign. ∆ denotes first differences. $EC{M}_{t-1}$ is the lagged error correction term (EViews output CointEq(−1)∗). Oil is measured per capita in kWh. The specification includes a linear trend.

reports the estimates of the baseline ARDL(4, 2, 4, 4) specification in first logarithmic differences. The model exhibits high explanatory power, with $R^2=0.964$ and adjusted $R^2=0.923$, indicating that over 92% of the variation in $\Delta ln\left({\mathrm{CO}}_2\right)$ is explained by the included regressors. The Durbin–Watson statistic is $\mathrm{DW}=2.15$, which does not suggest first-order residual autocorrelation.

The error correction coefficient, $EC{M}_{t-1}$ (EViews output CointEq(-1)), is equal to $-0.3507$ and is statistically significant at the 1% level. The negative sign is consistent with theoretical expectations and implies that approximately $35\%$ of any deviation from the long-run equilibrium is corrected in each period (year), confirming a stable long-term relationship.

Regarding long-term co-movement, the bound testing results (see Table 4) show an F-statistic of $3.5159$, which exceeds the $I\left(0\right)$ bound at the 10% level, providing weak evidence of cointegration. This is supported more robustly by the t-bounds statistic of $-4.2758$, which exceeds the $I\left(1\right)$ critical values at the 10% and 5% levels.

Several short–term dynamics are statistically significant. Lagged changes in $\Delta ln\left({\mathrm{CO}}_2\right)$ display strong negative effects, indicating a rapid short-term adjustment. Changes in $\Delta ln\left(\mathrm{GDPpc}\right)$ are positive and significant across multiple lags, suggesting that short–term economic growth is associated with higher CO₂ emissions. Changes in $\Delta ln\left({\mathrm{Oil}}_{\mathrm{kWh},\mathrm{pc}}\right)$ show mixed signs—positive contemporaneous values and negative values at longer lags—pointing to a complex interaction between oil-based energy use and emissions, potentially reflecting efficiency improvements or substitution effects over time.

Overall, the evidence supports a stable long-run equilibrium among CO₂ emissions, economic growth, capital formation, and oil-based energy consumption, alongside meaningful short-run fluctuations driven by macroeconomic and energy variables. The estimated speed of adjustment (about 35% per year) highlights a moderate pace at which environmental–economic dynamics revert to equilibrium following shocks.

3.2. Coefficient Analysis

3.2.1. Gross Fixed Capital Formation $\Delta ln\left(\mathrm{GFCF}\right)$

The estimated coefficient is $-0.8130$ (p = 0.0394), indicating a statistically significant and negative relationship at the 5% level between gross capital investment and CO₂ emissions. Specifically, a 1% increase in GFCF is associated with a 0.813% reduction in emissions, ceteris paribus. This finding contrasts with conventional theory, which posits that increased investment typically leads to greater production and energy use. However, this result may suggest that part of Ecuador’s investment is directed toward more efficient or less carbon-intensive technologies.

3.2.2. Gross Domestic Product per Capita $\Delta ln\left(\mathrm{GDPpc}\right)$

The coefficient of 1.8709 is not statistically significant (p = 0.2174), though its positive sign aligns with the hypothesis that higher per capita income increases the demand for goods, services, and energy, and therefore emissions. The lack of significance may stem from nonlinearities or interactions with other factors such as energy mix.

3.2.3. Oil Consumption $\Delta ln\left(\mathrm{Oil}\right)$

The estimated coefficient is 1.2116 and marginally significant at the 10% level (p = 0.0536). This suggests that a 1% increase in oil consumption is associated with a 1.21% increase in CO₂ emissions. This finding is consistent with Ecuador’s energy matrix, where oil remains the dominant source, especially in transport and thermal electricity generation.

3.3. Lag Structure Analysis

3.3.1. Lagged CO₂ $\Delta ln{\left({\mathrm{CO}}_2\right)}_{t-1}$

The coefficient of $-0.5777$ is highly significant (p = 0.0017), indicating strong persistence in emissions. That is, past levels significantly influence current emissions, a typical pattern in environmental data. This implies that public policies or structural changes may take time to achieve substantive reductions.

3.3.2. Lagged GFCF $\Delta ln{\left(\mathrm{GFCF}\right)}_{t-1}$

The coefficient is 0.3966 and not statistically significant (p = 0.2463), suggesting that the short-term effect of capital investment on emissions does not persist into the following period. Its positive sign contrasts with the contemporary effect, indicating a possible nonlinear or delayed impact.

3.3.3. Lagged GDP per Capita $\Delta ln{\left(\mathrm{GDPpc}\right)}_{t-1}$

The coefficient of $-3.3005$ is significant at the 5% level (p = 0.0456). This result suggests that increases in per capita GDP in the previous period are associated with reductions in current CO₂ emissions, potentially reflecting improved energy efficiency or shifts in consumption patterns.

3.3.4. Lagged Oil Consumption $\Delta ln{\left(\mathrm{Oil}\right)}_{t-1}$

The coefficient of 1.6300 is significant at the 5% level (p = 0.0171), reinforcing the conclusion that oil consumption has a sustained and direct effect on CO₂ emissions. This highlights the urgency for energy transition policies.

3.4. Time Trend

The linear time trend (t) has a positive coefficient (0.0036) but is not statistically significant (p = 0.3122). This suggests that, after accounting for the economic variables and their lags, there is no significant linear temporal trend in CO₂ emissions. Instead, the variation in emissions appears to be driven more by economic determinants than by time alone.

3.5. Model Interpretation

The estimated ARDL model indicates that both oil consumption and economic growth have meaningful—albeit sometimes divergent—short- and long-term effects on CO₂ emissions in Ecuador. In particular, oil consumption exerts both immediate and lagged statistically significant effects, reinforcing the necessity for a shift toward cleaner energy sources. Capital investment has a negative and significant contemporary effect but exhibits ambiguity in its lagged impact. Meanwhile, GDP per capita shows a complex dynamic: insignificant in the current period but negative and significant in the lag, potentially signaling improvements in energy efficiency as income rises.

4. Discussion

This research identified that certain economic variables significantly influence environmental degradation, broadly aligning with findings from previous studies. Specifically, in the Ecuadorian case, GFCF was found to have a negative relationship with CO₂ emissions. This may suggest that increased capital investment is being directed toward cleaner or more efficient technologies, thereby reducing the country’s carbon footprint. However, the environmental effects of investment differ across contexts. For instance, a study in Saudi Arabia, employing a Vector Error Correction (VEC) model, concluded that internet usage significantly and positively contributes to increased CO₂ emissions, while capital investment did not exhibit a significant effect in that country [44]. Conversely, in the BRICS economies, investments in technological innovation have been associated with increases in carbon emissions both in the short and long run [45]. This latter result underscores the importance of directing technological development toward sustainable alternatives, as unchecked technological progress may hinder the achievement of sustainable development goals by exacerbating CO₂ emissions.

The positive association between GDP per capita growth and CO₂ emissions is consistent with the Environmental Kuznets Curve hypothesis in its early stage, where rising income levels tend to increase emissions due to industrialization and energy demand [46,47]. The mixed short-term effects of oil-based energy use may reflect immediate emission increases from consumption, followed by adjustments due to energy efficiency measures or substitution effects in subsequent years.

Furthermore, the results confirm that oil consumption has a positive impact on atmospheric CO₂ emissions—an outcome fully consistent with economic theory and previous empirical research. Indeed, a panel data analysis for five Association of Southeast Asian Nations (ASEAN) countries reported that oil consumption has a positive and significant relationship with CO₂ emissions [48]. This study evaluated environmental degradation drivers—including population, oil consumption, tourism, and corruption—and found that among these, fossil fuel usage is one of the main contributors to rising CO₂ levels. However, not all economic sectors contribute equally to emissions stemming from oil consumption. In Mexico, for example, a cointegration analysis revealed that the electricity, transportation, manufacturing, and construction sectors are the primary sources of emissions due to non-renewable energy use. Consequently, it is recommended that public policy focuses on promoting a transition to renewable energy in these sectors [49]. This sector-specific evidence highlights the importance of designing differentiated strategies to mitigate environmental impacts, prioritizing the adoption of clean technologies, particularly in industries with high levels of pollution.

5. Conclusions

The results of this research demonstrate that macroeconomic variables such as GFCF, GDP per capita, and oil consumption have a significant relationship with CO₂ emissions in Ecuador. Using a logarithmic difference ARDL econometric model, it was found that GFCF exhibits a negative relationship with CO₂ emissions. This suggests that investment in less carbon-intensive sectors or cleaner technologies may be mitigating the country’s environmental footprint.

This finding contrasts with previous studies conducted in other regions, where infrastructure investment is generally associated with increases in emissions [50,51,52], highlighting a possible structural specificity of the Ecuadorian context. The results also imply an urgent need to redirect public credit and tax incentives towards low-carbon technologies, retire diesel and petrol subsidies to support electrified transport corridors, and mandate energy management standards for energy-intensive industries while offering concessional finance for retrofits.

Additionally, oil consumption maintains a positive and statistically significant relationship with emissions, reaffirming the role of fossil fuels as one of the main drivers of atmospheric pollution. This link is particularly relevant given that, despite progress in the penetration of renewable energy sources within the national electricity matrix [53,54], the transport sector—and other key production activities—remain heavily dependent on petroleum derivatives [55]. At the same time, GDP per capita displays a dual pattern: its current value is positively related to CO₂ emissions, whereas its one-period lag bears a negative and statistically significant coefficient, offering preliminary evidence for the Environmental Kuznets Curve (EKC) in the Ecuadorian context [46,47]. This result suggests that once a certain income threshold is reached, institutional and technological mechanisms may be activated that decouple economic growth from environmental degradation.

In the estimated ARDL error correction model, the coefficient of the first lagged difference in GFCF presents a negative and statistically significant value (−0.3394; p = 0.0424), indicating that, in the short term, recent increases in fixed capital investment are associated with a reduction in CO₂ emissions.

In the Ecuadorian context, this finding can be interpreted in light of the recent composition of public and private investment, primarily oriented towards clean energy infrastructure projects and technological modernization. The strong drive for hydroelectric generation, rural electrification, and improvements in the efficiency of the national electricity grid have reduced dependence on fossil fuel sources for energy production, which explains the negative short-term relationship between GFCF and CO₂ emissions.

The literature supports this outcome. Studies on Balkan countries have shown that GFCF influences CO₂ emissions and is one of the mechanisms correcting disequilibria both in the short and long run, highlighting its dynamic role in the energy transition [56]. Likewise, analyses in India demonstrate that the positive relationship between GFCF and emissions only emerged after economic liberalization, suggesting that, under a policy framework oriented towards sustainability, investment in fixed capital can have a mitigating effect on emissions [57].

These results are consistent with Ecuador’s renewable energy transition strategy, where capital investments have been primarily directed towards low-carbon sectors, generating positive environmental impacts in the short term.

Although Ecuador faces persistent economic and sociopolitical headwinds, its structural advantages—abundant renewable energy potential, a relatively compact electricity grid, and an emerging green finance framework—position the country to embark decisively on a low-carbon development path. The evidence presented here underlines the urgency of adopting targeted measures in transport, industry, and power generation, backed by clear fiscal incentives and robust governance mechanisms, to replace fossil fuel dependence with clean energy alternatives. Taken together, these findings offer a solid empirical basis for formulating public policies that fuse economic and environmental objectives while fulfilling Ecuador’s obligations under international sustainability agreements.