The environmental impact of remittance inflows in developing countries: evidence from method of moments quantile regression

Despite the importance of remittance inflows as potential source of incomes for recipient households and one of main contributors to the development process in various developing countries, their environmental effects have been largely neglected in empirical literature. To fill this gap, the current study proposes an extension of the conventional environmental Kuznets curve by performing a modified version based on remittance inflows in both low- and middle-income countries from 1980 to 2014. Using the novel method of moments quantile regression with fixed effect, the outcomes provide evidence of an inverted N-shaped Kuznets curve for remittances at higher carbon dioxide emitters in lower middle-income countries, proving that remittances allow recipient households to shift toward clean energy pattern (production/consumption). We find U-shaped curve for remittances from the 40th to 80th quantiles in upper middle-income countries along with monotonic negative effect on carbon dioxide emissions at highest quantiles (90th and 95th). No significant effect on environment has been outlined at lower carbon dioxide emitters for all panels. Regarding economic growth, an inverted N-shaped curve has been observed across all quantiles in upper middle-income countries and from lower to middle quantiles in low-income countries. Finally, financial development, as control variable, exerts significant mixed effect on carbon dioxide emissions, swung between positive at all quantiles in low- and upper middle-income countries and negative at lower quantiles in low-income countries. Some recommendations were further built in the present study.


Introduction
Remittance inflows to the developing countries have been considered as one of the potential sources of funding, susceptive to achieving the Sustainable Development Goals (SDG) defined in the post-2015 Agenda of the United Nations Development Programme (UNDP).
The importance attributed to remittances comes from the surge of their value during recent decades, reaching a large share of GDP in many developing economies and that become Responsible Editor: Nicholas Apergis an important source of income for the recipient households (Rahman et al. 2019;Goschin 2014;Meyer and Shera 2016). Remittance inflows to LMICs have grown rapidly (8.8% in 2017 and 9.6% in 2018), to reach US$ 529,640 million in 2018 up from US$ 31,058 million in 1990 and expected to record $570 billion in 2020 to become their largest source of foreign flows, often overtaking the value of foreign direct investments and official development aid (ODA) (World Bank (2018)).
Given their sheer magnitude and significance for the development process, remittance inflows have aroused a heightened interest of economists and policymakers. In this regard, the emphasis of studies has been focused mainly on the remittances-economic growth relationships (Benhamou and Cassin 2021;Sobiech 2019;Jouini 2015;Senbeta 2013) as well as on their impacts on poverty and income inequality (Vacaflores 2018;Apergis and Cooray 2018;Wagle and Devkota 2018;Bang et al. 2016). Furthermore, a vast literature has treated the effect of remittances on different components of human capital (health, education, labor productivity...) (Azizi 2018;Pilarova and Kandakov 2017;Bouoiyour and Miftah 2016), while another strand of studies has highlighted the relationship between remittances and financial development (Opperman and Adjasi 2019;Coulibaly 2015;Brown et al. 2013).
However, despite the above aspects of remittances, the empirical evidences of their environmental consequences are, surprisingly, quite neglected and not still clear in the literature. Only three studies, to our best knowledge, have tried to investigate remittances-environment relationship (Khan et al. 2020;Rahman et al. 2019;Ahmad et al. 2019). Yet, there are several reasons to argue that remittances are likely to induce CO 2 emissions through various ways: A direct effect could occur when remittances improve the standard of living of households, thereby generating higher levels of consumption (Thapa and Acharya 2017;Amuedo and Pozo 2004;Chami et al. 2005). More specifically, remittances could improve purchasing power, making it easier to acquire household items (automotive vehicles, air condition, computers, electric machines, etc.) and thus putting further pressure on energy consumption and CO 2 emissions. Other direct effects may follow when the implementation of remittances-induced investment results in higher demand for energy. McKenzie and Sasin (2007) have underscored that remittances, as a transitory income, will be invested rather than spent. Remittances could swell the level of domestic investment by providing capital, supporting small firms and promoting real estate investment. This argument is in line with Wagle (2012) who documented that remittances generate a huge boost for growth in real estate, construction and investment in Nepal.
At the same time, remittances could directly reduce CO 2 emissions when a large portion of them is spent to cleaner energy use (e.g., solar energy) especially in case of high tariff and/or lack of electricity (Sharma et al. 2019). That is, the overall effect of remittances on environment depends on how it is implemented throughout the economy.
On another side, remittances may indirectly affect the environment through several channels, namely financial development and economic growth which are deemed to have a direct impact on CO 2 emissions. Certainly, there is evidence to assume that remittances are highly correlated to these components. In this context, the literature tends to claim that remittances should stimulate macroeconomic variables by increasing aggregate demand, savings, and investment and thus generating multiplier effects on economic growth which in turn affects the environment (Kadozi 2019;Clemens and McKenzie 2018;Ziesemer 2012). In addition, remittances can be brought to bear for health and education spending along with enhancing labor productivity leading to support and improve economic growth (Mamun et al. 2015;Koska et al. 2013;Salas 2014). Similarly, remittances contribute to financial development by providing complementary funding to the business community and relaxing the credit constraints as well as by increasing the demand for financial services (savings, transfer of money, etc.) (Fromentin 2017;Farhani and Ozturk 2015). As a result, remittances as effective contributor to economic growth and financial development could affect indirectly CO 2 emissions. In this context, various studies have investigated financial development-CO 2 emissions links showing mixed effects (Boufateh and Saadaoui 2020;Nasir et al. 2019;Charfeddine and Kahia 2019;Xing et al. 2017;Ayeche et al. 2016;Li et al. 2015). Nevertheless, it is argued that remittances may act as an alternative to financial development (Giuliano and Ruiz-Arranz 2009;Nyamongo et al. 2012). It is assumed that if the banking sector is well developed, the investment credit constraints will therefore be limited, and thereafter, remittances will not be used necessarily for production and hence leading to less release of CO 2 . This substitutability could be conceived as the opposite effects of financial development facing those of remittances on CO 2 emissions.
As for the effect of economic growth on CO 2 emissions, the income per capita is widely considered the most important variable related to environmental degradation. Thence, the most empirical literature has mainly been performed within the EKC framework initiated by Grossman andKrueger (1995) (e.g., Boufateh 2019;Sapkota and Bastola 2017;Sarkodie and Strezov 2019a). Accordingly, the EKC hypothesis exhibits an inverted U-shaped suggesting that income per capita initially leads to deterioration of environment, but after reaching a certain level, the quality of environment improves as income rises. Further, several studies suggest that environmental degradation will start to rise again beyond turning income level, and thus, the relationship turns to be N-shaped (Bhattarai et al. 2009;Alvarez and Balsalobre 2016).
Proceeding from this concept, although remittances have then become an important additional source of income (outside GDP) received directly by households, understanding their underlying effects on CO 2 emissions remains an interesting question that is worth exploring. Thus, we propose to evaluate the EKC hypothesis for remittances (Rem) along with the conventional EKC in LMICs from 1980 to 2014. This study is focused on the underlying idea that remittances, as a specific income, might worsen environmental quality until reaching a turning point and improves it thereafter to rise again beyond a certain level. Hence, the main purpose of this study is to explore the possible N-shaped EKC and investigate whether remittances could affect CO 2 emissions in developing countries. Moreover, we introduce financial development (FD) as a control variable to assess the possible substitutability effects on CO 2 emissions with those of remittances. Further, considering remittances along with a control variable such as financial development is likely to avoid significant bias on income regressors.
To meet the outlined objectives, we conduct the study on a large array of countries divided into three heterogeneous panels based on the World Bank classification of income level (World Bank. 2020): lower-income countries (LIC), lower middle-income countries (LMIC), and upper middle-income countries (UMIC). The focus on these groups is justified by the noticeable ratios of remittance to their GDP and that around 70% of the international migrants come from these countries and hence enjoying from about three-quarters of international remittance transfers (Sobiech 2019;Benhamou and Cassin 2021). Further, breaking down middle-income countries in two sub-samples is to consider their diverse nature and their differentiated level of economic development. Another reason for selecting these countries is their heterogeneity not merely in terms of economic growth but also in terms of remittances and CO 2 emissions. Therefore, we resort to the novel method of moments quantile regression with fixed effect (MM-QR) initiated by Machado and Silva (2019) to reveal whether the effects exerted by remittances alter across different quantiles of CO 2 emissions inside each group. Using this approach allows us to find out a complete picture of the conditional distribution in comparison to OLS technique, especially in countries with low and high emissions.
This study distinguishes from the extant literature in three ways. First, such as the GDP per capita, this paper is the seminal empirical study investigating the effect of remittance on CO 2 emissions through the EKC context. Second, even so a large body of literature has focused on the squared relationship to estimate the EKC; our study considers the cubic relationship to examine the possible N-shaped EKC. Third, to the best of our knowledge, our study is the first to employ the MM-QR to explore the heterogeneous effects as well of remittances as of the rest of the considered variables on CO 2 emissions.
The remainder of the study is structured as follows. "Model specification" presents the model specification. "Data and methodology" describes the corresponding data and the panel quantile methodology. "MM-QR results and discussion" proposes and discusses the empirical results. The conclusion and policy implications are presented in "Conclusion and policy implications."
The coefficient ω is expected to have different sign depending on whether financial development affects the environment.

Method of moments quantile regression
In this study, we employed panel quantile regression introduced by Koenker and Bassett (1978) to explore the impact of income per capita, remittances, and financial development on CO 2 emissions and to assess the N-shaped Kuznets curve hypothesis. This method provides more understanding pictures of the influence of the independent variables by allowing regression's line slopes to differ across the quantiles of the dependent variable which is more powerful compared to the traditional regression techniques focus on the mean effects such as OLS. Moreover, this approach is more accurate in the presence of outliers and if the random error term is not normally distributed (Zhu et al. 2018). Accordantly, using panel quantile regression makes it possible to investigate the CO 2 emissions determinants via the conditional distribution across different panel.
However, quantile regression with individual effects suffers from some problems such as it does not take into account possible unobserved heterogeneity across individuals. Hence, we adopted the method of moments quantile regression with fixed effect recently introduced by Machado and Silva (2019). Based on conditional means, this method makes it possible to estimate the conditional quantiles through combined estimates of the location and scale functions. Indeed, The MM-QR permits the individual effects to influence both the location and scale of the dependent variable Y (CO 2 ) and to impact the entire distribution instead of just shifting location such as in Koenker (2004) and Canay (2011). That is, this method provides information on how the conditional heterogeneous covariance effects of the determinants of CO 2 emissions are identified. In addition, MM-QR is very relevant when one's estimate quantile regression including individual effects and when the explanatory variables possess endogenous properties.
The MM-QR estimates the conditional quantiles of a dependent variable Y whose distribution conditional on a k-vector of covariates X belongs to models of location-scale variant. Y is defined by the following form: where the probability, P{δ i + Z 0 it γ > 0} = 1. (α, β', δ, γ')' are unknown parameters to be estimated.
(α i , δ i ), i =1,….., n, represent the individual i fixed effects and Z includes k-vector of specified components of X. These components are differentiable transformations with element l given by: X it and U it are i.i.d. for any fixed i and through time (t). According to Machado and Silva (2019), U it are orthogonal to X it and standardized to satisfy the moment conditions which do not involve stringent exogeneity. Referring to Eq. (2), the conditional quantile Q y (τ|x) of the dependent variable Y is expressed as follows: where X 0 it includes the independent variables (GDP, GDP 2 , GDP 3 , Rem, Rem 2 , Rem 3 , and FD). Q y (τ|X it ) denotes the quantile distribution of the dependent variable Y it (CO 2 ) which is conditional on the location of explanatory variables X it . The fixed effect of quantile τ for individual i is defined by the In contrast to the standard least-square fixed effect, the individual effects do not evince intercept shift. As timeinvariant parameters, their heterogeneous impacts are permissible to alter across the quantiles of the dependent variable Y. q(τ) is estimated from the following optimization problem: where

Data
We use a sample of 51 countries including three balanced panels presented in Table 1: LIC (10), LMIC (20), and UMIC (21). As mentioned above, this classification stems from the World Bank (2020). We choose not to include high-income countries in the study since their remittance's contribution to the global share of GDP is often minimal. The selection of countries in each panel is restricted by the availability of data especially related to remittances. For a deep understanding of the N-shaped relationship, four variables are adopted in this study: CO 2 emissions (CO 2 ) are used as indicator for environmental degradation; real GDP per capita (GDP) as proxy of economic growth and remittances per capita (Rem) are the annual amounts of remittances divided by the total population and financial development (FD) proxied by the domestic credit to private sector. Annual data cover a time horizon from 1980 to 2014 and were collected from the World Development Indicators (WDI). All variables are expressed in natural logarithms in order to overcome the potential heteroscedasticity problems and to make analysis more meaningful. Table 2 describes the different variables, while Table 3 reports the common sample descriptive statistics.
As reported in Table 3, skewness values are negative and far from zero proving excessive skewness to the left for all variables. The Jarque-Bera statistical test strongly rejects the null hypothesis of normality confirming, once again, that applying OLS estimation will be inconsistent while employing quantile regression remains suitable and more robust for this study. Table 4 illustrates the correlation coefficients between the different variables as well as the variance inflation factor results. All correlation coefficients in absolute values between independent variables are less than 0.7. The results of the VIF test confirm the absence of multicollinearity between independent variables since the average value of 1.28 is less than 5. The results of VIF support our view to consider GDP, Rem, and FD in the same model.

MM-QR results and discussion
The estimate results obtained from MM-QR with fixed effect for different groups of countries are presented in Tables 5, 6, and 7, respectively. For comparison, the graphics of the corresponding MM-QR and OLS estimators for each panel are provided by Figs. 1-3.
The panel quantile results are reported for 5th, 10th, …, 90th, and 95th percentiles of the conditional CO 2 emissions in which the 5th to 30th percentile are used to examine the effect of GDP, Rem, and FD in low CO 2 emitters; from 40th to 60th percentile are adopted to specify the effects in medium CO 2 emitters, while from 70th to 95th percentiles are employed to explore the effects in high CO 2 emitters.
The results from the method of moments quantile regressions showed heterogeneous effects across different quantiles and for different panel countries. The relationship between CO 2 emissions and the explanatory variables is confirmed at some of quantiles and remains inconclusive at other ones. Hence, this finding deserves to be highlighted and explained with carefulness.

Low-income countries
Regarding the estimation conducted for the first group of countries (LIC), presented in Table 5, the coefficients related to GDP, GDP 2 , and GDP 3 are significant and reveal homogeneously negative, positive, and negative respectively along the 5th-60th quantiles and remain insignificant elsewhere. This result supports the validity of an inverted N-EKC for income in both low and medium CO 2 emitters. This interesting finding could be explained by the economic nature of these countries as under-development economies where the high green technology is far from available to enhance the environment in first and/or in advanced development stage. Presumably, the pollution generated by the scale effects was compensated by other natural factors such as the expansion of forests in these regions or the adoption of cleaner practices related to the traditional agriculture susceptive to reduce CO 2 emissions.
The effect of Rem on CO 2 emissions is negative but insignificant across all quantiles meaning that EKC-Rem is not supported for these countries. The insignificant effects of remittances on environmental pollution could be explained by the fact that, given the weakness of GDP per capita in such countries, remittances, as additional incomes, are entirely devoted to the consumption of basic necessities, health coverage and education, generating in turn fairly low levels of CO 2 emissions. Rem 2 reveals homogeneously positive, significant and gradually increased from 60th to 95th quantiles and The effect of financial development on CO 2 emissions is negative and significant and decreases gradually (in absolute value) in low quantiles (5th-30th). However, its effect becomes irrelevant when moving from the 40th to 95th quantiles reflecting an under development financial markets in such countries. The negative effect of (FD) on environment, found in this study, is confirmed by previous studies which point out that financial development reduces energy consumption and CO 2 emissions by increasing energy efficient technologies (Pata 2018;Shahzad et al. 2017). However, these studies were conducted on developed countries, suggesting that resorting to bank credits in low-income countries is essentially to meet certain consumption needs not emitter of CO 2 . Nevertheless, despite these mitigated effects, we can also observe that the impact of FD on CO 2 emissions arises when the effect of remittances dissolves, reflecting a possible substitutability between the two components.
The estimates also validate a U-shaped EKC from locational dimension. Indeed, the results showed that GDP per capita and its squared values (GDP 2 ) have effects with opposite signs on the location and scale assuming that increasing GDP (GDP 2 ) reduces (increases) the average CO 2 emissions but also increases (reduces) the variance of CO 2 across quantiles. However, GDP 3 has no effect on location even if increases the dispersion of observed CO 2 through all quantiles. Furthermore, despite their insignificant effects on the average CO 2 emissions, Rem 2 and Rem 3 increase the scale of the observed carbon dioxide emissions since the coefficients of Rem 2 and Rem 3 are significantly dispersed across quantiles as depicted by Fig. 1. Table 6 illustrates the outcomes of MM-QREG for (LMIC). The empirical results deduced from MM-QR provide insignificant coefficients of GDP as well of their second and third polynomial excluding the existence of a relationship between environmental degradation and income. This result is very consistent with the study of (Allard et al. 2018) underscoring that, in lower middle-income countries, the N-shaped Kuznets curve is depicted only in few quantiles. In line with these authors view, the disconnection between economic growth and CO 2 release assumes that these countries have not yet achieved a desired level of income at the development stage allowing to intensify the production tools and to adopt technologies susceptive to influence the environment, whether in terms of pollution or in terms of mitigation. This disconnection could be also consistent with the idea that these economies are mainly focused on non-polluting sectors, namely traditional agricultural activities and light industries, which are far from using energy levels likely to release a large polluting amount of CO 2 .

Lower middle-income countries
The MM-QR estimates show that the effects of remittances on CO 2 emissions are insignificant at low and middle quantiles (5th-60th), while their effects become negative and statistically significant from higher quantiles (70th, 80th, 90th, 95th). When moving from the 70th to 95th quantile, we can notice a gradual dwindling of CO 2 emissions and that the effect of remittances becomes more significant inferring that remittances are inclined to the adoption of cleaner technologies. The negative effect of remittances on carbon dioxide has been found by Khan et al. (2020) in India assuming that received remittances lead to enhance the Similarly, at higher quantiles, the relationship between Rem 2 and CO 2 emissions is homogeneously positive and significant, whereas CO 2 emissions decrease with increasing Rem 3 . That is, the MM-QR estimates corroborate the validity of an inverted N-shaped Kuznets curve for remittances in higher quantiles. Even it is difficult to explain this interesting finding, remittances are probably spent, at initial stages, for the use of cleaner energy in order to overcome the lack of electricity for lighting, hydraulic pumping, etc., or for clean agricultural activities such as trees-cropping likely to reduce CO 2 release. Thereafter, recording high level, along with the weakness of incomes, remittances would be oriented toward polluting production inducing hence environmental degradation. Further, reaching large higher values, remittances would possibly invest in eco-friendly technology which in turn leads to less energy consumption and CO 2 emissions. Accordantly, when reaching higher level, remittances allow recipient households to shift toward clean energy pattern in terms of production or consumption.
In other ways, a strong significant effect of FD on CO 2 emissions is observed when using MM-QR. A positive association exists between financial development and CO 2 emissions across all quantiles at 1% significance level suggesting that the banking sector, in these countries, remains the most important source of investment funding and that the banking credits contribute a priori to the financing of economic projects emitters of CO 2 . It is noteworthy to point out that starting from lower toward higher quantiles, the effect of FD on CO 2 is increasingly upsurging. Indeed, the increase in FD by 1% rises CO 2 emissions by 0.344-0.638% implying that more financial development leads to more environmental degradation.    For location and scale parameters estimates, while the Nshaped EKC for remittances is not validated from locational dimension, the effects of Rem, Rem 2 , and Rem 3 have an increase variance across quantiles as showed in Fig. 2. Financial development not only significantly affects the average CO 2 emissions from the locational dimension but also increases the dispersion of observed CO 2 across quantiles.

Upper middle-income countries
The MM-QR results showed evidence from an inverted Nshaped EKC for GDP in (UMIC) at all quantiles. Indeed, as seen in Table 7, the coefficients related to GDP and their quadratic and cubic values are significant and reveal homogeneously negative, positive, and negative respectively along all quantiles. The decline of CO 2 could be due to technological transfer and may be a compensation effect generated by highenergy efficiency against the scale effect-induced pollution, which further leading to a negative effect of GDP on CO 2 emanation (Sarkodie and Strezov 2019b;Allard et al. 2018).
From the middle (40th) to higher quantiles (95th), remittances generate a strong gradual negative effect on CO 2 emissions with 1% level of significance. Similar to the low-income countries, the effect of remittances remains insignificant at lower quantiles. Remittances squared cause a negative effect from the 20th to the 80th quantiles, showing inconclusive aggregated weight on CO 2 emissions. In addition, the effect of Rem 3 on CO 2 emissions is positively significant across all quantiles providing evidence that reaching higher values, remittances would cause a rise in CO 2 emanation. This statement has been validated by Khan et al. (2020) suggesting that, in the long-term, remittances escalate the CO 2 emissions in some BRICS countries which are part of upper middle-income countries. Nevertheless, the signs of Rem, Rem 2 , and Rem 3 support the validity of a U-shaped EKC at the 40th to 80th quantiles. Moreover, at the highest quantiles (90th and 95th), the insignificant effect of Rem 2 indicates that the relationship between remittances and CO 2 emissions is negatively monotonic excluding the validity of the EKC hypothesis.
Such as in LMIC, financial development increases CO 2 emissions across all quantiles at 1% significance level. FD gradually increases CO 2 emissions from the lower to the higher quantiles meaning that FD-environment degradation relationship is highly quantile dependent.
The positive connection between FD and environment has been outlined by previous studies (Kayani et al. 2020;Wang et al. 2019;Bekhet and Othman 2017;Ito 2017;Jebli et al. 2016;Gunasekaran et al. 2014) which have pointed out that, in case of developed financial markets, the investors can get credits at low rates allowing purchasing machinery increasing thus the release of carbon dioxide. Further, higher is the financial development, more are foreign investors attracted and then contributing to the environmental degradation.
However, it should be noted that in low quantiles, the effect of FD surges when the effect of remittances is inconclusive, but starting from the middle to higher quantiles, this effect is in part compensated by the negative effect of remittances when credits banking seems to become more expensive.
According to the location and scale estimates, the results showed that all explanatory variables strongly affect the average CO 2 emissions and that only remittance coefficient has higher negative dispersion of CO 2 across quantiles as presented in Fig. 3.

Conclusion and policy implications
The main purpose of this study is to investigate whether remittance inflow could affect the environment in developing countries. Examining the relationship between remittances and environment is for great importance given the magnitude  In line with this feature, this study seeks to evaluate the validity of the environmental N-shaped Kuznets curve hypothesis for remittances along with the conventional EKC in LIC, LMIC, and UMIC countries from 1980 to 2014. In this study, financial development is considered as control variable.
We have adopted the recent method of moments quantile regression with fixed effect introduced by Machado and Silva (2019) to explore the effects of the considered variables on different quantiles of CO 2 emissions.
The empirical results showed heterogeneous effects of remittances, income, and financial development on CO 2 emissions across different quantiles in each panel country.
The panel quantile regression reveals interesting patterns about the effect of remittances on CO 2 emissions. In particular, remittances begin to exert negatively significant effects on CO 2 emissions mainly at higher quantiles in (LMIC) and from medium to higher quantiles in (UMIC). On the contrary, we do not find significant spillover from remittances to CO 2 emissions in low quantiles meaning that in low CO 2 emitters, remittances support households for their subsistence needs away from what could harm the environment.
The analysis of remittance coefficients and their second and third polynomial values support the validity of an inverted N-shaped Kuznets curve for remittances in LMIC at higher quantiles highlighting the importance of remittance as a contributor to the pollution decrease. Afterward, a U-shaped EKC is observed from 40th to the 80th quantiles in (UMIC). However, given the values of Rem and Rem 2 , a monotonic negative effect of remittances on environment is found in UMIC at highest quantiles, excluding the existence of EKC. The negative monotonic effect on CO 2 emissions in UMIC indicates that the quest to improve remittance inflows in these countries should enhance environmental quality. Nevertheless, the positive value of Rem 3 in UMIC argues that the environmental gains resulting from remittances can be negatively affected if some measures of mitigation and awareness are not adopted to reduce pollution.
When considering the relationship between GDP per capita and CO 2 emissions, our results support an inverted N-shaped Kuznets curve for income at all quantiles in UMIC and at both lower and middle quantiles (5th-60th) in LIC. The empirical results found from LIC and UMIC are consistent with the idea that increasing production should not always causes environmental damages but in opposite could improve environmental quality. This consistent effect is occurred especially when improving energy efficiency and technological innovation.
As for financial development, this component exerts mixed effect on CO 2 emissions across different panel with respect to the degree of financial markets development. We find a positive effect at all quantiles in both LMIC and UMIC. In LIC, financial development led to a negative aggregated effect on CO 2 emissions from lower quantiles to the first middle quantile (5th-40th).
Another interesting result is that remittances reveal a substitute funding tool to the financial development especially when dealing with underdeveloped financial markets.
To conclude, our findings have several implications and provide useful insights to policymakers. First, reliance only on mean regression results may induce ineffective policy implications. Second, policies need to be established individually for each country depending on their specificities in terms of remittances, incomes and financial system. Third, it is inferred that, beyond the role played in terms of economic and financial development and in terms of funding support for households, the remittances are a powerful tool allowing affecting the environment depending on how they are allocated. Hence, it is at first imperative to avoid structural inefficiencies and bureaucracy issues in order to facilitate received remittances and to be formally sent. Besides, reducing remittances costs and considering incentives to increase remittance inflows to the country remains a suitable target under sustainable development. Fourth, there is evidence to implement effective financial system to soften investment credits granting intended for clean technology and for projects funding which are complying with the environment. Meanwhile, promoting research and development for cleaner production remains a suitable action to ensure sustainable environment.
For further researches, it should be argued that the relationships between environment and the considered variables are more complex and should be taken with great carefulness. Incorporating other control variables and classifying the countries according to other criteria as well as the availability of large dataset are also required to more accurately capture the different effects and may explicit the inconclusive results in various quantiles in our study.
Abbreviations EKC, environmental Kuznets curve; LMICs, low-and middle-income countries; LIC, low-income countries; LMIC, lower middle-income countries; UMIC, upper middle-income countries; SDG, the Sustainable Development Goals; UNDP, United Nations Development Programme; ODA, official development aid; WDI, World Development Indicators; GDP, real gross domestic product per capita; CO 2 , carbon dioxide emissions; Rem, remittances per capita; FD, financial development; MM-QR, method of moments quantile regression; OLS, ordinary least square; VIF, variance inflation factor Availability of data and materials The datasets analyzed during the current study are available in the World Development Indicators in World Bank Data.
Author contribution Mahmoud el Batanony: Conceptualization Imed Attiaoui: Methodology; Formal analysis and investigation; Writing original draft preparation Ibrahim Ali Ali: Software; Formal analysis and investigation Nahla Nasser: Collection of the data; Writing, review and editing