Spillover Effects of Financial Development and Globalisation on Environmental Quality in EAEU Countries

1. Introduction

According to the economic literature, integration processes have both positive (trade creation effect (Dalimov, 2009), market integration (Zhuo, 2024) and negative (trade diversion effects (Yang & Martinez-Zarzoso, 2014), resource flow (Caliendo et al., 2021)) effects, exerting a multidirectional impact on welfare in developing nations. Besides direct effects of integration determining directly the dynamics of socio-economic development, there are also indirect, so-called spillover effects of integration. They indirectly affect the welfare of the population through knowledge spillover (Kim & Verdolini, 2023; Otsuka et al., 2023; H. Wang et al., 2023; Xiuwu et al., 2022), technology diffusion (Cyprian En, 2010; Gries et al., 2018), and labour market transformation (Levytska, 2023; Suliman, 1999). A further indirect consequence of integration processes may be the deterioration of the environment and population quality of life

The establishment of a common economic space in integration of countries involves the reduction of trade barriers, unification of norms, standards, rules and procedures, efficient allocation of economic resources, etc. Therefore, the absolute mobility of three economic resources: capital, technology and labour force multiply the benefits of interregional specialisation and cooperation development within the integration association space. Studies by Kurečić & Luša (2020) (Kurečić & Luša, 2014), Cheng, Zhang & Xu (2023) (Cheng et al., 2023), Iwegbu, Justine & Borges Cardoso (2022) (Iwegbu et al., 2022), Barnes, Black, Markowitz & Monaco (2021) (Barnes et al., 2021) confirm the thesis of integration accelerating the economic growth of countries through the expansion of markets, financial development, intensification of production, and regional value chains.

The challenge is an increasing environmental cost as a trade off for higher economic growth. In 1955, Simon Kuznets formulated the hypothesis concerning with the economic growth provides the environmental degradation – the so-called environmental Kuznets curve (EKC) (Kuznets, 1955). Considering the acceleration of integration processes and the dominance of the ‘more money - more production’ paradigm in the global economy, more and more researchers pay attention to the relationship between these processes and the environmental condition (see Table 1).

Therefore, integration processes in the EAEU countries provides the research challenge on increasing of economic growth rates caused by the elimination of trade barriers between countries and the benefits of financial development affect the environment and the improvement of the population quality of life.

Meanwhile, Footprintnetwork.org calculates the index of the ecological footprint, including for individual countries. It is possible to see the risks of a growing ecological deficit for four of the five economies of the Eurasian Economic Union (EAEU) (see Figure 1).

Therefore, the dynamics of trade turnover and GDP of the EAEU countries for 2023 (increasing up to 8% and 3%, respectively), the lack of the environmental empirical studies in the EAEU (see Table 1), and the impact of integration processes on environmental quality within the integration association are extremely relevant.

The aim of the study is to identify the impact of integration spillover effects on the ecological footprint in 5 EAEU countries between 1992 and 2023.

2. Methodology

The general hypothesis of the research concerns with negative impact of integration spillover effects on environmental quality in EAEU countries.

We based on the work of Bui Hoang Ngoca & Nguyen Huynh Mai Tramb (Ngoc & Tram, 2024) on assessing the impact of financial development and globalisation on the environmental quality of ASEAN countries to form the research model and economic strategy.

The following indicators were used to verify the hypothesis:

• GDP per capita (unit: U.S. dollars, at 2015 fixed prices) as a proxy of economic growth provided by the World Bank;
• Gha per capita (unit: points) as a proxy of ecological footprint provided by the Footprintnetwork.org.
• Financial development index (unit: percentage) as a proxy for financial development collected from the International Monetary Fund (IMF);
• KOF globalization index (unit: points) as a proxy of integration process provided by the Swiss Economics Institute.

The data are given in the range: 1990-2023.

The boundaries of the study: 5 EAEU countries.

In order to achieve the research objective, the analysis sequence was carried out through the following steps.

Step 1: 

 Analyze the stationarity of the variables.

Step 2: 

 Check for cross-sectional dependence.

Step 3: 

 Evaluates the consistency of an estimator.

Step 4: 

 Calculate the Moran's I index.

Step 5: 

 Estimate research results using the Spatial Error Model (SEM), Spatial Autoregressive Model (SAR), and Spatial Dubin Model (SDM) or to eliminate spatial models.

Step 6: 

 Analysis and diagnosis of the model.

Step 7: 

 Eliminate multicollinearity.

3. Results

The spatial regression method proposed by Anselin (Anselin, 1988) extends the classical linear regression model by considering the spatial dependence in data. Spatial dependence can occur in the dependent variable (spatial lag model, SLM) or in model errors (spatial error model, SEM).

3.1. The General Spatial Lag Model (SLM)

In econometric analysis, the stationarity of variables is a key step in the construction of correct and reliable models. Stationarity ensures the stability of time series or panel data statistical properties. It allows ones to avoid false regressions and provide reliable conclusions. This paper analyses the stationarity of variables by several tests for panel data, including the Levin, Lin test & Chu (LLC) (Levin et al., 2002), the IPS test proposed by Im, Pesaran and Shin (IPS) (Im et al., 2003). The use of multiple tests increases reliability and ensures consistency of the results. The test results are presented below (see Table 2).

$$y=\rho Wy+X\beta +\epsilon$$

(1)

where:

• y is a vector of a dependent variable of dimension n×1n×1;
• X is a matrix of independent variables of dimension n×kn×k;
• β is a vector of regression coefficients of dimension k×1k×1;
• W is a spatial weight matrix of dimension n×nn×n, reflecting the spatial relationships between observations;
• p is the coefficient of spatial autoregression;
• ε is the error vector, it is assumed ε∼N(0,σ2I).

Table 2. Result of panel unit root tests.

	IPS Statistic (Levels)	IPS Statistic (First Difference)	LLC Statistic (Levels)	LLC Statistic (First Difference)
Gha_per_person	-2.502845405**	-4.964768713**	-2.502845405**	-4.964768713**
FDI	-1.503180219*	-7.618670237**	-1.503180219*	-7.618670237**
GDP_per_capita	0.077397281*	-3.735265451**	0.077397281*	-3.735265451**
KOF	-3.919361581**	-4.019112779**	-3.919361581**	-4.019112779**

* Data not Stationary at 5%, ** Data Stationary at 5%, Source: authors.

Table 2. Result of panel unit root tests.

	IPS Statistic (Levels)	IPS Statistic (First Difference)	LLC Statistic (Levels)	LLC Statistic (First Difference)
Gha_per_person	-2.502845405**	-4.964768713**	-2.502845405**	-4.964768713**
FDI	-1.503180219*	-7.618670237**	-1.503180219*	-7.618670237**
GDP_per_capita	0.077397281*	-3.735265451**	0.077397281*	-3.735265451**
KOF	-3.919361581**	-4.019112779**	-3.919361581**	-4.019112779**

* Data not Stationary at 5%, ** Data Stationary at 5%, Source: authors.

Based on the results of stationarity tests performed by Levin, Lin methods & Chu (LLC) and Im, Pesaran and Shin (IPS), we found the following:

-

the variables Gha_per_person and KOF are stationary at significance levels of 5% and 10%. It is confirmed by negative and statistically significant values of LLC and IPS. These variables are integrated of the order of zero (I(0)) and have stable statistical properties over time;

-

the variables FDI and GDP_per_capitaare non-stationary in significance levels (the values of LLC and IPS statistics do not allow rejecting the null hypothesis of the presence of a single root at significance levels of 5% and 10%). However, after taking the first differences, these variables become stationary. It indicates their integration of order one (I(1)). This result emphasises the necessity of using first differences of these variables in the construction of econometric models or considering possible cointegration relationships to ensure the correctness and reliability of the obtained conclusions.

3.1.1. The Cross-Dependence Test

After conducting stationarity tests and determining the order of variables integration, the next step is to verify the cross-dependence in the panel data. The cross-dependence test (CD test) allows ones to determine existence of correlations between individual units in the sample not accounted for in the model. The presence of cross-dependence can cause a bias, inefficiency of estimates, and incorrect statistical conclusions.

Null hypothesis (H0): there is no cross-dependence in the residuals:

$$H0:cov\left({\epsilon }_{it},{\epsilon }_{jt}\right)=0 \forall i\ne j; \forall t$$

(2)

Alternative hypothesis (H1): there is a cross-dependence in the residuals:

$$H1:cov\left({\epsilon }_{it},{\epsilon }_{jt}\right)\ne 0 \mathsf{д}\mathsf{л}\mathsf{я} \mathsf{н}\mathsf{е}\mathsf{к}\mathsf{o}\mathsf{т}\mathsf{o}\mathsf{р}\mathsf{ы}\mathsf{х} i\ne j; \forall t$$

(3)

The CD test statistics are calculated as follows:

$$CD=\sqrt{{\displaystyle \frac{2T}{N\left(N-1\right)}}}{\sum }_{i=1}^{N-1}{\sum }_{j=i+1}^N{\widehat{\rho }}_{ij}$$

(4)

where:

• N is the number of individual units (cross sections);
• T is the number of time periods;
• $\rho$ – assessment of the paired correlation between the residuals:

  $${\widehat{\rho }}_{ij}={\displaystyle \frac{\sum _{t=1}^T{\widehat{\epsilon }}_{it}{\widehat{\epsilon }}_{jt}}{\sqrt{\sum _{t=1}^T{\widehat{\epsilon }}_{it}^2}\sqrt{\sum _{t=1}^T{\widehat{\epsilon }}_{jt}^2}}}$$

  (5)

Table 3. CD-test result.

Variable cross_sectional_dependence_results	Statistic	P_Value
Gha_per_person	7.104943697	1.20372E-12
FDI	11.48649614	1.54242E-30
GDP_per_capita	17.22569383	1.70372E-66
KOF	16.92342036	3.0235E-64

Source: authors.

Table 3. CD-test result.

Variable cross_sectional_dependence_results	Statistic	P_Value
Gha_per_person	7.104943697	1.20372E-12
FDI	11.48649614	1.54242E-30
GDP_per_capita	17.22569383	1.70372E-66
KOF	16.92342036	3.0235E-64

Source: authors.

According to CD test results, all the studied variables have statistically significant cross-dependence between individual units in the data panel. Low P-values (significantly less than 0.05) allow ones to confidently reject the null hypothesis of the absence of such a dependence. The inclusion of appropriate adjustments in the econometric model will help ones to avoid bias in estimates and improve the quality of research results.

Once a significant cross-sectional relationship between individual units has been identified using the CD test, there is a need to correctly select a model for further analysis of the panel data. The presence of cross-sectional dependence indicates traditional panel data models may not be sufficient to adequately describe the structure of data and a more rigorous approach to model specification is required.

3.1.2. Evaluates the Consistency of an Estimator

Particularly, it is necessary to determine whether to use a fixed effects (FE) model or a random effects (RE) model to account for individual differences in observations and to ensure unbiased and efficient parameter estimations. To address this issue, the Hausman test (Hausman, 1978) is used to determine whether individual effects are correlated with explanatory variables. The results of the Hausman test will help to select the most appropriate model and ensure the correctness of the subsequent econometric analysis.

The Hausman test is conducted to test the hypothesis on the difference between the coefficient estimates of the FE and RE models being statistically insignificant. It indicates the random effects model is preferred. Further, we introduce the formal setting of the test, describe its hypotheses, and interpret the results obtained.

Null hypothesis (H0): Differences in coefficient estimates between FE and RE models are statistically insignificant. It implies the individual effects are not correlated with the explanatory variables and the RE model produces consistent and efficient estimates.

$$Н0:cov\left({\alpha }_i,x_{it}\right)=0$$

(6)

Alternative hypothesis (H1): Differences in coefficient estimates between FE and RE models are statistically significant. It implies the individual effects are correlated with explanatory variables and the FE model is preferred.

$$Н1:cov\left({\alpha }_i,x_{it}\right)\ne 0$$

(7)

Hausman test statistics is calculated as follows:

$$Н={\left({\widehat{\beta }}_{RE}-{\widehat{\beta }}_{FE}\right)}^{\prime }{\left[Var\left({\widehat{\beta }}_{FE}\right)-Var\left({\widehat{\beta }}_{RE}\right)\right]}^{-1}({\widehat{\beta }}_{RE}-{\widehat{\beta }}_{FE})$$

(8)

where:

• ${\widehat{\beta }}_{RE}$ – vector of coefficient estimates from a model with random effects;
• ${\widehat{\beta }}_{FE}$ – vector of coefficient estimates from a fixed-effects model;
• $Var\left({\widehat{\beta }}_{FE}\right),Var\left({\widehat{\beta }}_{RE}\right)$ – covariance matrices of coefficient estimates for the RE and FE models, respectively.

The Hausman test statistic value is 0.2648, and the corresponding P-value is 0.9665. Since P-values are significantly higher than standard significance levels (e.g., 5% or 10%), there is no reason to reject the null hypothesis for no significant differences between the FE and RE model estimates.

It implies the individual effects are not correlated with the explanatory variables and the RE model produces consistent and efficient estimates. Therefore, the RE model is preferred for further data analysis.

3.1.3. Moran’s I index

However, despite the correctness of the RE model in the context of individual effects, it remains necessary to consider the possible influence of spatial dependence between observations. The presence of spatial autocorrelation can disrupt the assumptions of classical econometric models on the independence of observations. It can cause biased and inefficient parameter estimates. Therefore, the next stage of the research involves testing the data for the presence of spatial autocorrelation using Moran's I index (Moran, 1950). It identifies and quantifies the degree of spatial dependence in geographic data.

Moran's I index is used to identify and measure the degree of spatial autocorrelation in geographic data. Spatial autocorrelation concerns with the values of a variable in one location depend on the values in neighboring ones. The presence of spatial autocorrelation can disrupt the assumptions of classical econometric models on the independence of observations. It can cause biased and inefficient parameter estimates. Therefore, it is important to identify and consider the autocorrelation in modelling.

Moran's I index is calculated as follows:

$$I={\displaystyle \frac{N}{W}}·{\displaystyle \frac{{\sum }_{i=1}^N{{\sum }_{j=1}^N\omega }_{ij}\left(x_i-\overline{x}\right)\left(x_i-\overline{x}\right)}{{{{\sum }_{j=1}^N(x}_i-\overline{x})}^2}}$$

(9)

where:

• – total number of observations;
• – the value of the variable in location i;
• – the average value of the variable for all observations;
• – elements of the spatial weight matrix W reflecting the degree of proximity between locations i and j;
• $W=\sum _{i=1}^N\sum _{j=1}^N{\omega }_{ij}$ – the sum of all the elements of the weight matrix.

Null hypothesis H0: there is no spatial autocorrelation in the data; the values of the variable are randomly distributed in space.

$$H0:I=E\left[I\right]\approx -{\displaystyle \frac{1}{N-1}}$$

(10)

Alternative hypothesis H1: spatial autocorrelation is present; the values of the variable have a spatial dependence.

$$H1:I\ne E\left[I\right]$$

(11)

The results of spatial autocorrelation testing are presented in Table 4.

Although the results are statistically significant (low P-values), Moran's I index values are very small in absolute value and close to zero. Although the test formally indicates a significant spatial autocorrelation, its magnitude is almost negligible. Since there is no significant spatial autocorrelation, there is no necessity to apply complex spatial models such as the Spatial Durbin Model (SDM).

Considering the identified significant cross-correlation between individual units and the absence of significant spatial autocorrelation, there is a need to adjust the model to account for the impact of common factors not captured in the original specification. Although the random effects (RE) model is recognised as the preferred model, ignoring cross-sectional dependence may cause bias and inefficient parameter estimates.

3.1.4. Common Correlated Effects Mean Group Estimator

In this regard, it is rational to apply the Common Correlated Effects Mean Group estimator (CCEMG) proposed by Pesaran (2006) (Pesaran, 2006). This method can effectively account for the influence of uncontrolled common factors in panel data. Moreover, it may be a source of cross-sectional dependence between individual units. The use of CCEMG provides consistent and asymptotically normal estimates of the model coefficients in the presence of common factors correlated with the explanatory variables.

Consider a panel regression model:

$$y_{it}={\alpha }_i+{\beta }^{\prime }{x}_{it}+u_{it}$$

(12)

where:

• – the dependent variable for the individual unit i at time t.
• – vector of explanatory variables of dimension .
• – individual fixed effect.
• – vector of the coefficients of dimension .
• – - model error accounting for unobserved common factors.

The error is assumed to consist of unobserved common factors and random error:

$$u_{it}={\lambda }_i^{\prime }{f}_t+{\epsilon }_{it}$$

(13)

where:

• – is a vector of unobservable general dimension factors .
• – vector of individual loads (factor loadings) of dimension .
• – a random error, ${\epsilon }_{it}~IID(0,{\sigma }^2)$.

A complete model considering common factors:

$$y_{it}={\alpha }_i+{\beta }^{\prime }{x}_{it}+{\lambda }_i^{\prime }{f}_t{+\epsilon }_{it}$$

(14)

However, since they are $f_t$unobservable, it is impossible to directly include them in the model. The CCEMG method suggests using the average values of variables for all individuals as a proxy for unobserved common factors.

Enter the cross-sectional average values:

$${\overline{y}}_t={\displaystyle \frac{1}{N}}{\sum }_{i=1}^N{y}_{it}{\overline{x}}_t={\displaystyle \frac{1}{N}}{\sum }_{i=1}^N{x}_{it}$$

(15)

Using these averages, specification of the model is as follows:

$$y_{it}={\alpha }_i+{\beta }^{\prime }{x}_{it}+c^{\prime }\overline{z_t}{+\epsilon }_{it}$$

(16)

where:

• $\overline{z_t}$ – a vector containing the average values of the dependent and explanatory variables: $\overline{z_t}=({\overline{y}}_t,{\overline{x}\prime }_t)$’
• $c$ – a vector of coefficients reflecting the influence of general factors through average values.

CCEMG final model:

$$y_{it}={\alpha }_i+{\beta }^{\prime }{x}_{it}+{\varphi }^{\prime }{\overline{x}}_t+{\psi }^{\prime }{\overline{y}}_t{+\epsilon }_{it}$$

(17)

where:

• $\varphi$ and $\psi$are the coefficients corresponding to the average values of the explanatory and dependent variables, respectively.

The final model specification for the data we have:

$$Gh{a}_{pe{r\_capita}_{it}}={\alpha }_i+{\beta }_{1}FD{I}_{it}+{\beta }_{2}GD{P}_{pe{r\_capita}_{it}}+{\beta }_{3}KO{F}_{it}+{\theta }_1{\overline{FDI}}_t+{\theta }_2\overline{GDP\_per\_capit{a}_t}+{\theta }_3{\overline{KOF}}_t{+\epsilon }_{it}$$

(18)

where:

• ${\overline{FDI}}_t,\overline{GDP\_per\_capit{a}_t}$,${\overline{KOF}}_t$ are the average values of the corresponding variables at time t;
• ${\alpha }_i$ – individual effect for each country;
• ${\beta }_1,{\beta }_2,{\beta }_3$ – coefficients of influence of individual variables;
• ${\theta }_1,{\theta }_2,{\theta }_3$ – coefficients of influence of the average values of variables (proxies for common factors).

Table 5. Results of model estimation using CCEMG.

Variable	Coefficient estimation	The standard error	t-statistics	p-value	Value
Constant	6.8679	0.78465	8.7529	4.339e-15	***
FDI	-1.7236	1.5836	-1.0884	0.27819
GDP_per_capita	0.00051559	0.000054257	9.5027	< 2.2e-16	***
KOF	0.070859	0.028408	2.4944	0.01372	*
mean_FDI	-0.14440	5.1264	-0.0282	0.97757
mean_GDP_per_capita	0.00024937	0.00016427	1.5180	0.13115
mean_KOF	-0.18384	0.040502	-4.5390	1.162e-05	***

Source: authors.

Table 5. Results of model estimation using CCEMG.

Variable	Coefficient estimation	The standard error	t-statistics	p-value	Value
Constant	6.8679	0.78465	8.7529	4.339e-15	***
FDI	-1.7236	1.5836	-1.0884	0.27819
GDP_per_capita	0.00051559	0.000054257	9.5027	< 2.2e-16	***
KOF	0.070859	0.028408	2.4944	0.01372	*
mean_FDI	-0.14440	5.1264	-0.0282	0.97757
mean_GDP_per_capita	0.00024937	0.00016427	1.5180	0.13115
mean_KOF	-0.18384	0.040502	-4.5390	1.162e-05	***

Source: authors.

Model statistics:

• Total Sum of Squares: 475.42
• Residual Sum of Squares: 147.53
• Coefficient of determination (R-squared): 0.69857
• Adjusted coefficient of determination (Adj. R-squared): 0.68635
• F-model statistics: 57.1654
• Degrees of freedom (df): 6 (numerator), 148 (denominator)
• p-value of the F-test: < 2.22e-16

The model explains about 69.86% of the variation in the dependent variable. It indicates a good level of its explanatory power The constant is significant at the level of 1% and shows the basic value of the environmental indicator with zero values of independent variables. GDP_per_capita has a positive and statistically significant impact on the environmental indicator (p < 0.001). It indicates the increase in GDP per capita is related to the increase in the environmental indicator. Indeed, KOF also has a positive and statistically significant effect on the environmental indicator (p < 0.05). It suggests a relationship between the level of globalisation and environmental performance. However, mean_KOF has a negative and statistically significant effect (p < 0.001). It may reflect the impact of globalisation trends on the environmental indicator as a whole. The remaining variables (FDI, mean_FDI, mean_GDP_per_capita) are not statistically significant. It indicates the absence of identified influence on the dependent variable within the framework of the model.

The provided correlation matrix shows the degree of relationship between the model variables (see Table 6).

Observations of the correlation matrix are as follows:

• high correlation between FDI and GDP_per_capita (0.8031): financial development and GDP per capita are closely related;
• strong correlation between KOF and mean_KOF (0.8616): expected, since mean_KOF is the average value of KOF;
• a very high correlation between the average values of the variables:

  ◦

  mean_FDI и mean_GDP_per_capita: 0.9290;

  ◦

  mean_FDI и mean_KOF: 0.9420;

  ◦

  mean_GDP_per_capita и mean_KOF: 0.9280.

3.1.5. Variance Inflation Factor Analysis

VIF measures the degree of multicollinearity between explanatory variables. VIF values above 5 or 10 indicate a serious multicollinearity problem (see Table 7).

Variables with high multicollinearity (VIF > 10):

• KOF: 14.137
• mean_FDI: 12.153
• mean_GDP_per_capita: 10.011
• mean_KOF: 21.334

Variables with moderate multicollinearity (VIF close to 10):

• FDI: 8.876

Reasons for multicollinearity:

• the inclusion of the average values of the variables: mean_FDI, mean_GDP_per_capita, and mean_KOF strongly correlate with the corresponding individual variables and with each other;
• economic relationships: variables reflect related economic processes (globalisation, financial development, economic growth).

Exclusion of the average values of variables:

◦

the average values cause high multicollinearity;

◦

consider alternative methods of accounting for common factors, for example, models with fixed time effects.

After detecting high multicollinearity in the inclusion of the mean values of the variables in the model (under the common correlated effects method, CCE), it it was decided to re-estimate the model without their inclusion. It allows us to assess the effect of exclusion the average values on the quality of the model and eliminating the problem of multicollinearity. The results of the new model and their comparison with previous results are presented below (see Table 8).

Model statistics:

• Total Sum of Squares: 475.42
• Residual Sum of Squares: 147.53
• Coefficient of determination (R-squared): 0.68969
• Adjusted coefficient of determination (Adj. R-squared): 0.68635
• F-model statistics: 57.1654
• Degrees of freedom (df): 3 (numerator), 121 (denominator)
• p-value of the F-test: < 2.22e-16

The reassessment of the model without including the average values of the variables was successful. The problem of multicollinearity was eliminated, while the main results and interpretations were maintained. It confirms the expediency of using a fixed-time effects model to account for common factors and ensure reliable and interpretable research results.

4. Discussion

The findings of this study shed light on the spillover effects of financial development, economic growth, and integration on ecological footprints in EAEU from 1990 to 2023. The final model with fixed time effects uses the variables FDI, GDP_per_capita, and KOF. The model obtained showed a high coefficient of determination (R-squared ~ 69%). It indicates a good level of explanatory power while the problem of multicollinearity was successfully eliminated and the main results of the model remained stable after adjustment.

The results of the research are as follows:

Firstly, economic growth has a positive and statistically significant impact on ecological footprints. It corresponds to the hypothesis put forward in the work, as well as the EKC. This result is similar to studies by Alola, Bekun and Sarkodie (2019) (Alola et al., 2019) for 16 EU countries in the period 1997-2014, Danish, Hassan, Baloch, Mahmood & Zhang (2019) (Danish et al., 2019) for Pakistan in the period 1971-2014, Ahmed, Zafar (2020) (Ahmed et al., 2020) for G7 countries from 1971 to 2014, Nathaniel, Anyanwu and Shah (2020) (Nathaniel et al., 2020) for the countries of the MENA region, Ngoc & Huynh (2024) (Ngoc & Tram, 2024) for 10 ASEAN countries between 1992 and 2021.

Secondly, integration processes have a positive and statistically significant impact on ecological footprints. It corresponds to the hypothesis put forward in the work, as well as with international trade theories. According to those theories, if a specific government injects more money into the economy, it will directly affect the competitiveness of that country's goods with those of neighboring countries (through exchange rate effects), thus indirectly affecting the economic growth of neighboring countries. Accelerating economic growth by increasing commodity turnover can negatively impact energy quality through increased energy consumption, increased consumer demand, and industrial development. At the same time, the results obtained contradict the conclusions of the studies of Xiao, Tan, Huang, Li & Luo (2022) (Xiao et al., 2022), Murshed, Ahmed, Kumpamool, Bassim & Elheddad (2021) (Murshed et al., 2021), Lv, Zhu, & Du (2024) (X. Lv et al., 2024), Qi, Liu, & Ding (2023) (Qi et al., 2023). According to these researches, the integration allows ones to redistribute environmental costs between countries, reduce environmental damage through knowledge spillover and diffusion of technologies.

Hence, financial development does not have a statistically significant impact on ecological footprints in the EAEU countries over the long term.

5. Conclusions

The data obtained revealed several blind spots in studies on the socio-economic development of the EAEU countries. It allowed us to formulate a number of research questions to be addressed in our next researches:

-

effect of knowledge spillover and diffusion of technologies in the EAEU space on the quality of production (green transformation, fossil fuel energy vs renewable energy) and, accordingly, environmental quality;

-

relation of the quality of the environment, the level of income and consumption and living standards in the EAEU countries;

-

effect of financial development in Russia on the level of energy consumption, CO2 emission, and economic growth in other EAEU countries.

The practical significance of the research involves a number of policy recommendations for the EAEU countries.

Economy policy framework:

• stimulation of renewable energy sources using (solar, wind, hydro) and increase the energy efficiency of equipment through the implementation of strict emission standards for products manufactured in the EAEU;
• introduction of turnover penalties for champion enterprises in CO2 emissions in the EAEU countries;
• tax incentives and subsidies for companies producing zero-carbon products;
• formation of a "low-carbon trade" model for the integration association: tariff incentives for exports and imports of goods with a low carbon footprint; increasing of tariffs for imports of products manufactured in violation of environmental standards;
• establishing ‘green’ economic zones in the EAEU for companies committing to produce zero-carbon footprint products;
• establishing an interstate eco-lending system for countries to subsidise loan rates for companies implementing the ESG agenda.

Institutional framework:

• design a single integrated system of air, water and soil quality monitoring for timely response to environmental problems in the integration association area;
• establishing joint programmes on R&D and transfer of environmentally friendly technologies between developed and developing countries in the EAEU;
• formation of the interregional fund for sustainable development to prevent and eliminate environmental damage in the EAEU countries;
• development of a regulatory framework and ensuring control over the relocation of harmful production to countries with more environmentally lenient regulations;
• introduction of mandatory disclosure of environmental risks in the construction of industrial facilities and formation of ESG ranking in the EAEU.

Social framework:

• organise educational campaigns to raise the level of environmental awareness of the population starting from school; involve schoolchildren and students in the activities of environmental groups and eco-volunteering;
• expanding companies' commitments under the ESG agenda;
• increasing the eco-awareness of the population through the implementation of social advertising and corporate culture of companies;
• households tax deductions and subsidies for the installation of solar panels, construction of energy-efficient buildings, and the introduction of environmentally friendly technologies;
• educating the public on the possibilities of investing in environmentally responsible projects through bonds, stocks, and funds;
• increased grant support for eco-research aimed at developing new technologies to reduce pollution, recycle waste, and use renewable energy sources;
• implementing policies to integrate environmentally responsible behaviour into daily life: from developing infrastructure for separate waste collection to bonus programmes for those who use eco-bags, refuse plastic bags, or buy local products.

Hopefully, the data obtained and the problems addressed will activate a new wave of applied research on the impact of spillover effects of integration on living standards in the EAEU countries.