The role of income of workers on the environment: Does gender matter? Open Access

In recent decades, many scholars have investigated the determinants of environmental degradation, such as globalization (Ang, 2009; Al-Mulali and Ozturk, 2015; Adams et al., 2016; Munir and Ameer, 2018; Van Tran, 2020; Burki and Tahir, 2022; Khan et al., 2022), trade openness (Al-Mulali and Ozturk, 2015; Le et al., 2016; Ling et al., 2020; Udeagha and Ngepah, 2022; Barkat et al., 2025; Pham and Nguyen, 2024), urbanization (Azam and Khan, 2016; Adams and Klobodu, 2017; Yasin et al., 2021; Kahouli et al., 2022), energy consumption (Wang, 2010; Qu et al., 2017), renewable energy (Baek, 2016; Bilgili et al., 2016; Tutak and Brodny, 2022; Zhang et al., 2024; Aydın, 2025), fossil fuel consumption (Kartal et al., 2022; Yousaf et al., 2022; Addai et al., 2024; Eweade et al., 2024; Umair et al., 2025), and inflation (Djedaiet, 2023; Grolleau and Weber, 2024; Jin et al., 2024; Zheng et al., 2024; Hondroyiannis et al., 2025). These studies underline that environmental degradation—through issues such as global warming—poses serious threats to human and ecological systems.

Among the frameworks used to analyze the economy–environment nexus, the Environmental Kuznets Curve (EKC) hypothesis has received the greatest attention. Grossman and Krueger (1991) suggested that economic growth and environmental degradation are linked in an inverted U-shape: pollution rises during early stages of development but declines once higher income levels allow societies to prioritize environmental protection. Numerous studies have confirmed this inverted U-shaped EKC (Ahmed and Long, 2012; Shahbaz et al., 2013; Ahmed and Qazi, 2014; Apergis and Ozturk, 2015; Gyamfi et al., 2021; Sultana et al., 2022; Mitić et al., 2024; Yurtkuran et al., 2025), although others have rejected it (Aslanidis and Iranzo, 2009; Omisakin, 2009; Altıntaş and Kassouri, 2020; Dogan et al., 2020; Massagony and Budiono, 2023; Rahman et al., 2024), or suggested alternative shapes such as an N-shaped curve (Allard et al., 2018; Shahbaz et al., 2019; Awan and Azam, 2022; Shaheen et al., 2022; Shehzad et al., 2022; Achuo and Ojong, 2025).

However, the majority of EKC studies have relied on GDP per capita as a proxy for income. This measure may not fully capture how income is distributed, since higher GDP per capita does not necessarily translate into higher earnings for all individuals. In fact, only a small portion of society may benefit, which means that GDP per capita could be a misleading indicator of citizens’ real economic well-being. This creates an important research gap: little is known about how workers’ actual annual income—not just aggregate GDP—affects environmental outcomes.

To address this gap, the present study investigates whether workers’ annual income follows the EKC hypothesis in 20 OECD countries from 2005 to 2021. I argue that changes in employees’ incomes directly influence their consumption choices, which in turn affect environmental quality. At lower income levels, workers may prioritize utility maximization through higher consumption, which raises pollution. Once their income surpasses a threshold, however, they may adopt more sustainable behaviors and demand stricter environmental standards.

A second contribution of this study is to examine the gender dimension of the EKC. Prior research has shown that men and women have distinct spending patterns (Davies, 2011; Khan and Khalid, 2012; Räty and Carlsson-Kanyama, 2010), which implies that rising incomes could affect the environment differently across genders. For instance, men spend relatively more on transport, dining, and alcohol, while women allocate more to education, housing, and household goods. Yet, to date, no study has tested whether the EKC relationship varies between male and female workers.

Therefore, the study seeks to answer two main research questions:

To answer these questions, I employ feasible generalized least squares (FGLS) and panel-corrected standard error (PCSE) estimators to address heteroscedasticity and autocorrelation in the data. Two proxies of environmental degradation are used—ecological footprint and greenhouse gas emissions—to ensure robust findings.

The remainder of the paper is structured as follows. Section 2 presents the econometric model and methodology. Section 3 reports the findings, and Section 4 concludes with policy implications.

The dataset includes 20 OECD countries, namely Austria, Belgium, Canada, Czechia, Denmark, Estonia, France, Ireland, Luxembourg, Mexico, the Netherlands, Norway, Poland, Portugal, Slovakia, Slovenia, Sweden, Switzerland, the United Kingdom, and the United States, that were compiled from three different sources between 2005 and 2021. There are two main reasons behind choosing OECD countries in this study. First, the reliable and consistent data from OECD countries is essential for robust empirical analysis, as it covers a broad range of economic and environmental indicators. These countries are also known to be highly industrialized and have a significant impact on environmental degradation, making them particularly relevant for our study of the relationship between employee annual income and environmental degradation. Statistics show that fossil fuels account for 78% of the energy supply in OECD countries. The significant role OECD countries play in polluting the environment was exemplified by their accounting for 31% of global production-based emissions and 36% of global demand-based emissions in 2020 (OECD, 2025). The final decision was to include 20 OECD countries because data were missing across certain variables in some countries.

The Ecological Footprint Index, greenhouse gas emissions, and trade openness were gathered from the Global Footprint Network and World Development Indicators (WDI), but the International Labour Organization (ILO) provided the remaining variables. Table 1 provides some details about the variables.

For specifying the econometric model, the Kuznets theory introduced by Grossman and Krueger (1991) was used. I also employed three variables such as trade openness (Al-Mulali and Ozturk, 2015; Le et al., 2016; Ling et al., 2020; Udeagha and Ngepah, 2022; Barkat et al., 2025), fossil fuel consumption (Kartal et al., 2022; Yousaf et al., 2022; Addai et al., 2024; Eweade et al., 2024; Umair et al., 2025), and inflation (Djedaiet, 2023; Grolleau and Weber, 2024; Jin et al., 2024; Zheng et al., 2024; Hondroyiannis et al., 2025). The final econometric models are defined as follows:

I estimate the parameters by using the feasible generalized least squares (FGLS) method. The FGLS is an excellent approach to dealing with heterogeneity, serial correlation. While this technique has its advantages, it also has some drawbacks. FGLS has a significant problem with the underestimation of SEs in finite samples. The analytical performance of FGLS is poor if the true error variance-covariance matrix is unknown. PCSE, introduced by Beck and Katz in 1995, is a method for addressing the shortcomings. It is deemed a superior estimator to the FGLS one in many aspects (Reed and Ye, 2011; Appiah et al., 2019). According to Beck and Katz (1995), the PCSE estimator generates more precise SE estimates without any decrease in efficiency (Reed and Webb, 2010). This technique is also useful for dealing with heterogeneity, serial correlation, as well as cross-sectional dependence. This claim about the efficiency of the PCSE estimator was rejected by Chen et al. (2010). The PCSE estimator was found to have a lower efficiency than FGLS, except when the number of periods is near the number of cross sections. Nevertheless, I use both techniques to estimate the six models.

In equations (1) to (3), I utilize the ecological footprint (measured as consumption per capita) as an indicator of environmental degradation, similar to previous studies such as Ibrahiem and Hanafy (2020), Qaiser Gillani et al. (2021), Chu (2022), and Shojaeenia (2024). This indicator is represented by ED in this study. In addition, I employed another indicator of environmental degradation, which is shown by GHG in equations (4) to (6), which refers to total greenhouse gas emissions. Total greenhouse gas emissions in kt of CO2 equivalent are composed of CO2 totals excluding short-cycle biomass burning (such as agricultural waste burning and savanna burning) but including other biomass burning (such as forest fires, post-burn decay, peat fires, and decay of drained peat lands), all anthropogenic CH4 sources, N2O sources, and F-gases (HFCs, PFCs and SF6) ^[1].

Tinc demonstrates the real earnings of employees, while Tinc² is the square of the real earnings of employees ^[2]. I used these two variables to test the possible inverted U-shaped relationship between the real earnings of employees and environmental degradation. For example, in equation (1), when $α1$ is positive and $α2$ is negative, the inverted U-shaped curve is confirmed.

I also plan to examine this linkage between the real earnings of female employees and the two environmental indicators, and that is why I apply Feinc and Feinc², which are the real earnings of female employees and the square of real earnings of female employees, respectively. Finally, to do the same examination for males, Minc and Minc² are utilized, which represent the real earnings of male employees and their squared value, respectively. Furthermore, $Open$ shows the trade openness and is measured as exports plus imports divided by GDP. Trade openness can have an impact on the environment through three channels. The scale channel indicates that trade openness can cause an improvement in economic activities, which damages the environment. The technological effect describes how trade enhances production techniques and innovations, leading to energy efficiency and a cleaner environment. Trade can also alter the composition of output around the world through the composition channel, leading to the transfer of high-polluting goods from rich countries to poor countries. As a result, wealthy nations with strict environmental laws focus on producing clean goods, and poor nations tend to produce goods that release a significant amount of pollution into the atmosphere (Le et al., 2016; Yu et al., 2019; Sun et al., 2020; Chhabra et al., 2023).

Inflation is the rate at which the consumer price index increases. High inflation could cause consumers to switch to cheaper, less eco-friendly products, resulting in an increase in environmental pollution. However, the presence of rising prices could result in a decrease in private spending, particularly if households experience real income reductions, which would reduce emissions. Inflation is also viewed as a tax on consumption, which hurts consumption. So, during times of higher inflation, individuals typically boost their savings. Furthermore, customers may opt for products that use less energy when the price level rises. Companies that replace old machines to reduce production costs could expect the same effect. People may also choose to consume from second-hand markets due to price increases. Because these products are already made, buying second-hand ones is a more environmentally friendly option than buying new ones (Grolleau and Weber, 2024).

Fossil demonstrates fossil fuel consumption (% of total final energy consumption). The burning of fossil fuels results in the emission of a significant amount of carbon and greenhouse gases into the atmosphere. The trapping of greenhouse gases in the atmosphere leads to global warming. It increases the sea level, extreme weather events, and drought in tropical regions, increases tornadoes, hurricanes, and food shortages (Rani et al., 2023). So, an increase in fossil fuel consumption harms the environment. Furthermore, $i$ denotes country, $t$ presents time, and $ϵit$ is the error term.

Table 2 displays the correlation matrix. There is a positive relationship between the ecological footprint and all independent variables. The correlation matrix also reveals the positive link between GHG and the dependent variables. Since there is a high correlation between total, female, and male income, I need to examine their effects on environmental degradation in separate models.

Table 3 describes the results of Pesaran's (2021) cross-sectional dependence test (CD). According to the CD test, all variables have a cross-sectional dependence since the null hypothesis of cross-sectional independence is rejected at the 1% level.

Table 4 shows the outcome of the Hardi LM test. The null hypothesis is “All panels are stationary,” which is rejected at the level. However, at the first difference, the null hypothesis is not rejected. So the variables are I(1).

Since all variables are integrated of order one, I(1), their individual time series are non-stationary, meaning their means and variances change over time. Regressing non-stationary variables on each other can produce spurious regression results, where significant relationships appear even if no true long-term relationship exists. Cointegration tests are therefore necessary to determine whether a stable long-run equilibrium relationship exists among these non-stationary variables. If the variables are cointegrated, it implies that while they may drift in the short run, a linear combination of them is stationary, and they move together over time, justifying meaningful long-term inference. Table 5 exhibits the results of Pedroni (1999) and Kao (1999) residual cointegration tests. In models 1 to 6, the null hypothesis is rejected at the 1% level according to the three different approaches. According to the Modified Dickey-Fuller t and Augmented Dickey-Fuller t approaches, the null hypothesis is rejected at the 5% and 10% levels, respectively. The Pedroni cointegration test also provides evidence of a long-run relationship among variables at 1% according to three different approaches.

To select the best technique to estimate the models, first of all, I need to conduct two tests to choose between the fixed-effect model, the random-effect model, and the pooled model. The initial option is the F test, which assists us in choosing between the fixed-effect model and the pooled regression model. It can be inferred that the fixed-effect model is superior to the pooled model when the P-value is below 5%. As the P-value is lower than 5%, I can conclude that the fixed-effect model has an advantage over the pooled one in all six models.

The Hausman test is used to choose between the fixed and random-effect models. The random-effect model has an advantage over the fixed-effect model when the P-value of the Hausman test is greater than 5%. The null hypothesis of non-systematic coefficient differences for all six models is not confirmed, as shown in Table 6. Thus, the fixed-effect method is chosen.

Tables 7 and 8 show the results of autocorrelation and heteroskedasticity tests, which make estimates inefficient (Nawaz and Rahman, 2023). The Wooldridge test is displayed in Table 7, which is employed as the autocorrelation test for the fixed-effect model. It is shown that the data present a first-order autocorrelation issue under the fixed-effect model for all models since the null hypothesis, which is “no first-order autocorrelation,” is rejected.

The modified Wald test for groupwise heteroskedasticity in the fixed effects model is shown in Table 8. The probability value is below 5%; hence, the null hypothesis of homoscedasticity is rejected.

Table 9 presents the results of the six estimates using the FGLS technique. In the first three estimates, the ecological footprint is the dependent variable, while I selected greenhouse gas emissions (GHG) as the dependent variable in the second three estimates. The result of the first estimate demonstrates that the total income of employees has an inverted U-shaped relationship with the dependent variable since the estimated coefficients of total income and its squares are 0.0898 and −0.000934, respectively. In the fourth estimate, when the dependent variable changed to GHG, the magnitude of the coefficient changed. However, an inverted U-shaped linkage exists between total income and GHG. I also examine the influence of income based on the different genders on the environment. According to the second estimate, for women, income and the square of income possess positive and negative impacts on the ecological footprint, which means female income also follows an inverted U-shaped relationship. There is no difference in terms of how female income affects the environment when I altered the dependent variable to GHG in the fifth estimate. In estimates 3 and 6, male income also has an inverted U-shaped relationship with the ecological footprint and GHG. According to Table 10, when I employed the PCSE technique, I also acquired the same outcomes, illustrating that the results of the first six estimates using the FGLS technique are reliable. This inverted U-shaped linkage between total, female, and male income with these two environmental indicators shows that increasing employees’ income can increase environmental degradation in the first stage, since employees use more goods and services, which emit pollution. However, in the next stage, when their income passes the threshold, they will change their lifestyles and consumption in favor of the environment, and try to consume eco-friendly goods and services. This is consistent with Kuznet’s theory; both men and women seem to follow this process. When I computed and compared the maximum points of the inverted U-shaped graphs related to men and women, I found that men need a higher income on average to modify their routines and care more about the environment than women, according to the outcomes. For example, estimate 2 demonstrates that when the annual income of female employees exceeds 39.061 thousand Euros, they start to change their behaviors. However, for those who are male, this figure is 54.482 thousand Euros. Women reach the turning point at lower income levels, since their spending shifts earlier toward less polluting goods and services. In addition, social norms and roles may reinforce this pattern: women are often more directly responsible for household well-being and may feel a stronger sense of social responsibility toward the environment and future generations. This sense of responsibility can lead to more environmentally conscious consumption choices at earlier stages of income growth.

By contrast, men typically require higher income levels before such a shift occurs. This is because their consumption patterns are more strongly oriented toward status-related or luxury goods such as mobility, leisure, and entertainment, which provide utility primarily at higher levels of income. In other words, men need to reach the point where the consumption of these goods becomes saturated and no longer significantly increases their satisfaction. Only after this stage—when further income growth no longer enhances utility through luxury consumption—do men begin to allocate resources in ways that reflect greater concern for the environment. Thus, both the structure of male consumption bundles and women’s comparatively stronger sense of social and intergenerational responsibility explain why the turning point occurs later for men and earlier for women.

Moreover, in estimates 7 to 12, trade openness harms the environment, similar to estimates 1 to 6. For example, in the third estimation, ecological footprint goes up by 0.0159 units when trade openness increases one percent. This is because higher trade intensifies overall production and consumption (scale effect) and shifts production structures toward competitive but environmentally harmful sectors such as manufacturing and heavy industry (composition effect). Although openness can theoretically facilitate the diffusion of cleaner technologies (technique effect), in this study, the negative effects dominate, leading to environmental degradation. Ang (2009), Al-Mulali and Ozturk (2015), Munir and Ameer (2018), Van Tran (2020), and Akhayere et al. (2023) concluded the same outcome in terms of the influence of trade openness on the environment.

Despite estimates 7 to 9, which show the unfavorable impact of inflation is insignificant, other estimations illustrate that this effect is positive and significant. It means inflation harms the environment. For example, in the first estimation, a 1% increase in inflation causes a 0.0543 increase in ecological footprint. This is because higher inflation often leads consumers to switch toward cheaper goods, which are usually less environmentally friendly. These goods often involve lower production standards, higher energy intensity, and lower durability, which contribute to higher emissions and waste. This outcome is consistent with studies such as Hondroyiannis et al. (2025).

Fossil fuel consumption also has an adverse effect on the environment, according to the outcomes of all estimations. Fossil fuel consumption directly increases the release of CO₂ and other greenhouse gases, which can decrease the quality of the environment. For instance, in the first estimation, when fossil fuel consumption increases by one percent, the ecological footprint rises by 0.0193 units. This result is in line with outcomes of studies such as Zhang et al. (2022), Umair et al. (2025), and Rani et al. (2023).

Finding the determinants of environmental degradation has been a significant concern for researchers in recent decades. Scholars found that many factors, such as globalization, trade openness, urbanization, Fossil fuel consumption, and renewable energy consumption, affect the environment. However, in previous studies, scholars have not paid attention to the role of the annual income of workers on the environment. To fill this gap, I assessed the association between employees’ income and environmental degradation in 20 OECD countries, Austria, Belgium, Canada, Czechia, Denmark, Estonia, France, Ireland, Luxembourg, Mexico, the Netherlands, Norway, Poland, Portugal, Slovakia, Slovenia, Sweden, Switzerland, the United Kingdom, and the United States from 2005 to 2021. The ecological footprint and total greenhouse gas emissions (GHG) were selected as two proxies for environmental degradation, and I also employed two econometric techniques, FGLS and PCSE, to reach authoritative and robust outcomes.

The results show an inverted U-shaped relationship between worker income and environmental degradation. It means that, at a lower income level, the quality of the environment can be worsened by an increase in annual income. However, two ecological indices decrease when the annual income of workers reaches the maximum level (turning points). This is because workers usually do not care about the environment and related laws when their incomes are low, and they try to maximize their utility by using different goods and services. However, they try to do more to preserve the environment after passing a specific income level.

I also did this gender-based assessment and found that both male and female annual incomes possess an inverted U-shaped relationship with the ecological indices in this study. However, comparing the maximum points of curves related to male and female workers demonstrates that female workers need a lower annual income than male workers to change their lifestyles and care about the environment. For instance, the turning point for women’s income occurs at a lower level (around 39,000 Euros), while for men it occurs at a higher level (around 54,000 Euros). This suggests that women shift earlier toward less polluting goods and services, partly because of their greater sense of social norms and intergenerational responsibility, while men only adjust after their consumption of luxury or status-related goods becomes saturated.

Furthermore, the environment will be harmed when trade openness increases. This result is supported by the theoretical background as well as previous studies. In addition, the findings indicate that inflation also worsens environmental quality, since higher inflation drives consumers to purchase cheaper and less durable goods, which are typically more polluting. Likewise, fossil fuel consumption has a consistently harmful impact, as greater reliance on fossil energy directly raises emissions and ecological footprint. However, in estimates 7 to 9, this unfavorable effect of inflation is insignificant.

According to the outcomes, policymakers should not be worried about the increase in workers’ annual income in the long term, since they start to curb their ecological footprint after passing a specific level of income. Moreover, suppose governments want to increase the wages of workers as an environmental policy. In that case, it seems that focusing on the increase in the wages of women can bring fewer costs for companies and governments since women can pay attention to the environment at a lower level of income. However, since the sample is relatively homogeneous, the outcomes cannot be generalized to other contexts, which is another limitation of the study.

Future studies are recommended to investigate the inverted U-shaped relationship between workers’ income and environmental degradation using larger and more diverse datasets. Future studies can also examine these effects based on countries’ income levels and stages of development. It can strengthen the literature and help policymakers to answer this question: “Should they do anything when the income of workers goes up?” or “Does the unfavorable environmental effect of workers’ income recover automatically in the long run?”.

The author would like to thank the anonymous reviewers and the editor for their valuable and constructive comments, which significantly improved the quality of this paper. The author is also grateful to Dr Elena Meschi for her insightful suggestions and guidance on an earlier version of the manuscript.