Deforestation and financial inclusion in sub-Saharan Africa: evidence from the conditional and unconditional quantile regression approach Open Access

Deforestation remains a pressing global issue, particularly in tropical and subtropical forests, where forest ecosystems are vital for biodiversity, climate regulation, and the livelihoods of millions. In developing regions, especially sub-Saharan Africa (SSA), deforestation is driven by a myriad of factors, including economic, social, and political. The SSA region has experienced significant tree cover loss, with estimates suggesting that Africa lost approximately 10% of its forest cover between 1990 and 2020, translating to an annual average loss of about 4 million hectares, primarily due to agricultural expansion, logging, and infrastructural development (FAO, 2020). The loss not only threatens biodiversity, exacerbates climate change, disrupts local communities, and contributes to the deprivation of ecosystem services. According to the Intergovernmental Panel on Climate Change (IPCC, 2022), deforestation records almost 10% of greenhouse gas (GHG) emissions worldwide, underscoring the urgent need for targeted interventions. Governments in SSA have recognized the urgency of addressing deforestation and have initiated various programs aimed at preserving the forest ecosystem, for instance, the integration of community-based forest management into national strategies and fostering local stewardship of forest resources. Programs designed to enhance access to credit, conservation projects, and training in sustainable agricultural techniques have been implemented across the region. Additionally, governments are increasingly recognizing the importance of integrating financial inclusion strategies into their environmental policies to create a more sustainable balance between economic development and forest conservation.

The intersection of financial inclusion and deforestation has garnered increasing attention, as access to financial resources can influence land-use decisions and conversion efforts. Financial inclusion characterizes the accessibility and handiness of economic systems across all individuals in society, can empower communities to engage in sustainable practices, reduce environmental externalities, and reliance on natural resources, including forest resources (Bu et al., 2023; Gao, 2023; IMF & World Bank, 2020; Shahbaz et al., 2016). Understanding how financial inclusion impacts deforestation is crucial for developing strategies to combat forest degradation and promote sustainable development, particularly in the milieu of attaining the United Nations Sustainable Development Goals (SDG 15), which targets to restore, protect, and promote the sustainable exploitation of mundane ecosystems. Financial inclusion significantly contributes to the deforestation dilemma and many communities in SSA lack access to formal financial services, limiting their ability to invest in sustainable land-usage practices, agricultural activities, or alternative livelihoods that could mitigate forest loss. Without financial resources, the rural population often resorts to unsustainable agricultural methods and illegal logging as an immediate means of survival (Ali et al., 2022; Starfinger, 2021). Without access to credit or savings, farmers and landowners cannot afford the upfront costs associated with sustainable environment practices, such as agroforestry or reforestation, which hinders the transition to less destruction and contributes to deforestation.

The hypothesis of the Environmental Kuznets Curve on deforestation (EKCd) has gained momentum in the inclusive growth debate and theoretical frameworks (Agras and Chapman, 1999; Caravaggio, 2020b, 2020a; Grossman and Krueger, 1991, Grossman and Krueger, 1995). The hypothesis argued that economic growth initially led to environmental degradation, and the trend reverses after an irrefutable level of income is attained. Financial inclusion may accelerate this transition by enabling investment in cleaner technologies and sustainable practices, potentially mitigating deforestation. Mather (1992) proposed the forest transition hypothesis, similar to EKCd, which suggests that a country’s forest cover follows a predictable pattern as it undergoes economic development. The proposition posits two key points: as economies grow, especially in agrarian or developing contexts, forest cover typically declines, whereas the initial phase is characterized by unsustainable use of forest resources to meet economic needs. Eventually, as the country reaches higher levels of economic development, there tends to be a shift towards reforestation and forest conservation. The transition is often facilitated by improved policies, economic diversification, and increased awareness of environmental issues. Financial inclusion facilitates economic diversification, which can reduce pressure on forests, and communities may continue to depend on forestry-related income in its absence, leading to overexploitation and degradation of forest resources (Starfinger et al., 2023).

Despite extensive research on the association between economic prosperity and deforestation, there remains a lack of understanding about the conditional and unconditional distribution of financial inclusion across different quantiles of forest conversions. As a consequence, there is a necessity for a new affirmation and analysis of the link between financial activity and increasing rates of deforestation in the SSA region, among the largest blocks with (sub) tropical forestry. There is a gap in the existing literature; to our understanding, this study is the first to investigate and add to the continuing debate by considering and assessing the conditional and unconditional distribution of financial accessibility on deforestation. The EKCd and forest transition proposition studies are constrained to economic activities, although contrasting tracts of the economy have different suggestions on inclusive growth, climate externalities, and deforestation. The paper is structured into the following sections: Section 2 is a literature review; Section 3 presents data and methodology; Section 4 presents the empirical results, discussion, and policy implications; and finally, Section 5 concludes.

The earliest Environmental Kuznets Curve for deforestation (EKCd) proposition emerged in the early 1990s and was explored extensively (Bhattarai and Hammig, 2001; Dinda, 2004; Grossman and Krueger, 1991; Panayotou, 1993). The curve suggests that in the initial phases of economic prosperity, deforestation rates rise as a consequence of increased economic productivity activities. As income increases, however, societies tend to invest more in sustainable practices, reforestation, and preservation, paramount to a decrease in rates of deforestation (Panayotou, 1993). The forest transition hypothesis complements the EKCd by asserting that regions transition through distinct phases of forest cover change. Initially, as societies develop, forest cover declines due to agricultural expansion and infrastructure development. However, beyond a critical threshold of economic growth, social and political factors encourage reforestation and conservation efforts (Mather, 1992; Rudel et al., 2005). Extension of EKCd analysis reveals that the increased rates of deforestation, during the initial phases of economic development, can rise with economic progress. Moreover, the forested regions may expand with the increasing level of economic growth, cautions provided by the EKCd proposition about environmental externalities of current and historical economic development paths adopted by most developing economies, such as sub-Saharan Africa (SSA). Caravaggio (2020b, 2020a) provide a detailed EKCd inference across countries, the EKCd and forest transition propositions show the unanswered question of whether the hypotheses prevail or not in the continuing forestry debate, whereas some studies, on the other hand, proved the inexistence of EKCd (Barbier, 2004). This resulted in a fresh investigation on the determinants that influence the rising deforestation in developing and emerging economies.

Financial inclusion, as defined earlier, is argued as the accessibility and usage of financial systems by individuals and entities, particularly in developing economies. It makes a critical contribution to shaping economic behavior and may have a significant implication for environmental outcomes, including deforestation. For smallholder farmers or agribusinesses, credit can fund the purchase of inputs, finance land clearing, and enable shifts to commercial agriculture (Beck et al., 2007; Burgess and Pande, 2005). This often translates into agricultural expansion at the forest frontier in forest-adjacent areas, directly contributing to deforestation. The proposition aligns with the capital-access hypothesis introduced by Angelsen (1999), which posits that improved access to financial capital lowers the cost of converting forest into productive land. Access to finance enhances access to credit, allowing rural communities to invest in sustainable agricultural practices and technologies that reduce the need for deforestation. Conversely, it may reduce reliance on forests as a safety net. In contexts of economic vulnerability, forests often serve as a source of fallback income through fuelwood, charcoal, or non-timber forest products (Byron and Arnold, 1999). Access to formal financial services can help households manage risk without resorting to environmentally degrading practices (Dupas and Robinson, 2013; Karlan et al., 2014). This mechanism aligns with the safety-net substitution theory, suggesting that financial tools can offset the need for the extractive use of forest resources (Devereux, 2002).

Studies in the existing body of literature indicate that microfinance can promote eco-friendly practices by providing resources for agroforestry and alternative livelihoods (Starfinger et al., 2023; Yunus, 2007). Increased financial inclusion facilitates income diversification, and households with diverse income sources are less likely to engage in sustainable logging and land conversion practices, reducing dependence on forest resources (Cavendish, 2000; Duflo et al., 2013; Pattanayak et al., 2010; Starfinger, 2021). Shahbaz et al. (2016) re-examines the asymmetric effects of financial development, based on bank and stock market indicators, on environment quality in Pakistan between 1985 and 2014. The study found that inefficient energy use adversely affects the quality of the environment, and bank-based indicators impede the environment. Financial inclusion policy is essential in promoting the sustainable practices and recovery of the green economy. Gao (2023) assesses the rarity of financial inclusion in developing regions between 1990 and 2020, and found a positive effect of finance and natural hoard on green economic recuperation.

Table 1^[1] presents summaries of prior empirical studies that employed econometric approaches to investigate the impact of the availability and use of finance and macroeconomic causes on the environment, ecology, or biodiversity. Bu et al. (2023) used a panel of E7 economies and a cross-sectional autoregressive distributed lag (CS-ARDL) estimator to investigate the contribution of financial inclusion, natural resources, and urbanization on reducing environmental hazards for a sustainable environment. The study found that improved business through financial inclusion positively contributes to the destruction of the environment through carbon emissions. Ali et al. (2022) employ common correlated mean group (CCEMG), augmented mean group (AMG), and pooled mean group (PMG) on Economic Community of West African States (ECOWAS) economies from 1990 to 2016. They found that financial inclusion and technology demonstrates substantial positive and negative influences on ecological footprints and plays a crucial role on environmental health and carbon emission. In the context of environmental externalities, empirical studies consider utilizing carbon, greenhouse gas emissions, and sustainable practices as variables of environmental deterioration and ignore other ecological factors of degradation, such as agriculture, mining, and deforestation

The main intention of our research is to empirically investigate the heterogenous and distributional effects of financial inclusion on forest conversion in the SSA region. We divaricate from existing research that used forest area or net deforestation as a proxy estimate to examine the deforestation debate, as we use a net forest conversion in SSA tropical forests. This research confers to the body of literature in three strands. First, it captures the impacts of demand- and supply-side indices of financial inclusion on different quantiles of forest conversion. We examine the availability and affordability of financial systems in narrowing the conditional and unconditional deforestation gap and gain insights into the challenges and opportunities for leveraging finance to promote sustainable forest growth, reduce forestry loss, and foster more equitable tropical forest distribution. To our knowledge, most studies analyze the causal effect focusing on the average or mean effect, which may seem unlikely that most countries obtain average or even close to average effects, hence causing the heterogeneity effect of finance not to be considered in reducing deforestation rates. Second, we investigate the entire distribution of causal effect by employing an instrumental variable generalized quantile regression approach where multiple endogenous and instrumental variables are applied in the estimator (Powell, 2020). Third, analysis of the dynamics of the SSA region deforestation provides a picture of disparate effects underlying the finance on different tails of forest conversion. The exact nature of the relationship between the accessibility of finance and the initiative to reduce deforestation empirically remains inconclusive.

This study used balanced panel data, and due to data unavailability and consistent with the literature, countries without the key variable for the study were excluded, leading to the selection of 22 ^[2]SSA countries for the period from 2004 to 2020. Data was abstracted from the United Nations Food and Agriculture Organization (FAO), the World Development Indicators (WDI), and the Financial Access Survey (FAS) of the International Monetary Fund (IMF) databases. Table 2 presents the data description, descriptive statistics, and pre-tests, and no serious problems were observed. We incorporate variables in this study as follows. We diverge from the existing studies in the deforestation debates by using the forest conversion, ^[3] whether human-induced or not, to different land uses such as agriculture, animal grazing, mining, and urbanization as a proxy for deforestation, and as dependent variables. The proxy includes the perdurable loss of wood canopy below the threshold of 10% as described by FAO (2020). We included the demand-side indicators as the independent variables, including financial outreach and usage dimensions (the number of bank branches and ATMs per 100,000 adults and 1,000 KM², the number of depositors and deposit accounts with commercial banks per 1,000 adults, and domestic credit to the private sector by banks per GDP) (Ahamed et al., 2021; Kebede et al., 2023).

It is important to incorporate control variables that may impact the relationship across the variables, in line with theoretical and empirical deforestation studies such as Davis (1998), Leach (1992) and Ponce et al. (2021). These studies presented variables such as GDP per capita as indicators of the economic performance of a country, and proposed that the relationship with deforestation follows an inverted U-shape as initially deforestation increases and then decreases at higher income levels, trade openness as a share of GDP for total imports and exports, percentage of the agricultural area as a share of total land, and growing population size with rural-urban disparities in growth. Agricultural land area is a share of land abstracted from FAOSTAT. Agriculture is often a major productivity activity in developing countries, with the majority of the population residing in rural areas. Most economies use agricultural activities as a mechanism to drag individuals and countries out of financial crises due to trends in unemployment and more export opportunities (Dauvergne, 1999). Increasing agricultural practices impact forests, in principle, the expansion of such activities is a driver of deforestation, accounting for more than 80% of forest conversion, whereas large-scale commercial agriculture accounts for more than half in developing countries (FAO, 2020). It is hypothesized that a higher share of land under agriculture will be associated with increased deforestation.

Trade openness is included as another control variable; developing countries export more agricultural products, with top commodities in soybean, oil palm, coffee, and wood in SSA, compared to cocoa among others in Asia. The expansion of commodities that are driven by agricultural activities is the largest driver of deforestation in the tropical and subtropical forests, whereas the shifting in size of agricultural area is among the main causes of forest loss in SSA. Greater trade openness is hypothesized to be associated with higher deforestation. Population growth and disparity between rural and urban areas, urban expansion, mining, industrialization, and infrastructure developments influence increasing deforestation rates in tropical and subtropical economies (Barbier, 2004; Barbier et al., 2010; Ngoma and Yang, 2024). A growing population in an economy often leads to increasing consumption of forest products such as fuelwood and other forest resources, and agricultural farms that boost deforestation (Ngoma and Yang, 2024). Several studies in the existing literature explore the significant effects of growing rural and urban populations. Barbier (2004) argued that a growing population was a significant driver of increasing deforestation rates across tropical and developing economies through increased demand for agricultural land and biomass. Other studies argued that the influence of changes in population size on determining the dynamics in patterns and land use, integrated with increasing agricultural employment, plays a major role in boosting rates of forest cover changes in the region.

We used two-phase principal component analysis (2s-PCA) ^[7] to create the financial inclusion index, motivated by the multidimensional nature of financial inclusion. The approach is particularly suitable when indicators exhibit varying degrees of correlation within and across dimensions, and the composite index addresses the issues of multicollinearity. The first phase includes domestic credit to the private sector by banks (%, share of GDP), the number of deposits and depositors with commercial banks as a share of 1000 adults to create a financial usage indicator. Then, the financial outreach indicator includes demographic and geographic penetration of financial institutions with the number of bank branches and ATMs (per 100,000 adults and 1000 KM²). Finally, the second phase used to create a financial inclusion index, with the first phase dimensions of financial outreach and usage, defines initiatives that make financial systems accessible, usable, and available to all individuals in a community. The first principal component from the second-stage analysis was extracted as the final financial inclusion index, capturing the shared variance across all dimensions. Primarily, if PCA is employed on a set of variables, either all or some principal components will be used, whereas orthogonal variables are created when all variables are used out of the variables that are intercorrelated, and the number of degrees of freedom in the model will be reduced when some of the variables are used (Jackson, 2003). Kaiser criterion was used to decide on the number of components to be selected. The study proposes the hypothesis that financial inclusion has a heterogeneous effect on deforestation, potentially increasing forest loss in high-conversion regions and reducing it where alternative livelihoods are viable.

Before the main model inference, we used the fixed effects quantile regression (QR) technique that allows distributional conditioning of forest conversion on independent and control variables (Koenker, 2004; Koenker and Bassett, 1978; Machado and Santos Silva, 2019). Traditionally, the relationship between variables in the conditional QR estimator is assumed to be based on unobserved factors. However, the analysis of the estimated slope parameters significantly changes when unobserved variables are observed with the inclusion of other covariates into the QR model, and endogeneity issues of panel data are to be controlled by the QR model. In this study, the QR with the instrumental variable framework was employed as a baseline model, which has several advantages over the conventional QR estimator that was used in several prior studies (Powell, 2020). The QR estimator allows investigation of financial inclusion across the distribution of deforestation rates, identifying whether these effects are stronger or weaker at different points in the distribution. The QR model where $τth$ conditional quantile of the forest conversion indicator expressed as follows

Where $Qlogfit(τ/Xit′)$ denote the forest conversion across different $τth$ conditional quantile as a linear function on $i=1,…,N$ countries and $t=1,…,T$ periods, $Xit′$⁠, $εit$ and $βi$ are respectively vectors of explanatory variables, residual, and slope coefficients of explanatory variables. The ordinary least squares (OLS) model excludes countries with either lower or higher deforestation than medium countries, displaying the mean link between variables is among the weaknesses of linear function analysis; the effect of financial inclusion on high-deforestation countries may differ from that on low-deforestation countries. The conditional QR estimator captures the heterogeneity in the relationship between covariates and forest conversion by modeling the conditional distribution, rather than focusing only on the conditional mean as in OLS (Sarkodie and Strezov, 2019). While conditional QR focuses on the conditional distribution of the forest conversion, unconditional QR is proposed by Firpo et al. (2009) and focuses on how covariates affect the unconditional distribution of the forest conversion, that is, the entire distribution of deforestation outcomes. The estimator is useful and novel in investigating the impact of financial inclusion across the overall distribution, not just conditional on specific characteristics. The recentered influence function (IRF) developed by Firpo et al. (2009) on implementing the unconditional QR approach, the method transforms the dependent variable into the influence function for a specific quantile and then regresses on the covariates. The model using RIF regression is expressed as

Where $Qτ$ is the $τth$ quantile of the unconditional distribution of forest conversion logf1, $I{logf1it≤Qτ}$ denoted an indicator function that equals 1 if logf1 is less than or equal the $τth$ quantile and 0 otherwise, $flogf1(Qτ)$ is the density of forest conversion at the $τth$ quantile. The RIF transformation followed by OLS regression of the transformed forest conversion on the covariates as $RIF(logf1it;Qτ)=X′βτ+εit$ where $βτ$ and $ε$ denoted the vector of coefficients for the $τth$ quantile on the covariates X (financial inclusion measures and control variables) and error term. Unlike conditional QR, unconditional QR does not condition on values of the covariates and hence offers a broader view of how financial inclusion impacts deforestation at different quantiles of distribution, and the estimator is particularly useful in policy analysis as it provides a more general understanding of how covariates shift the entire distribution of outcomes, which is often more relevant for policy design (Byaro et al., 2023). The generalized QR estimator is an approach that enhances the counterfactual distributions for different quantiles of the deforestation rate. The technique solves the endogeneity and nonlinearity problems that exist between variables with the assumption that the method does not rely on average effect estimates. The QR estimator assumes the possibility of the deforestation rate being lower than the quantile function and being constant across control and instrument variables, however, generalized QR allows variation of this probability according to variables whereas some variables will show a high chance of the forest conversion being lower than the quantile function and a low probability predicted by other control variables. The baseline generalized QR with fixed effect and instrumental variables framework employed with the adaptive Markov Chain Monte Carlo (MCMC) sampling used the robustness of the model to cross-sectional dependency (CSD) and stationarity (Opoku and Aluko, 2021; Powell, 2020).

Table 2 columns (4) to (9) presents the summary statistics analysis of the utilized variable of the datasets, including mean, median, standard deviation, kurtosis, skewness, and Jarque-Bera (JB) normality test. Results herein suggest that the variables are not normal and skewed to either the left or the right. Before parameter estimation, in the Pesaran (2015, 2004) cross-sectional dependence test, in environmental and economic studies, data from different countries may be correlated due to shared factors such as economic policies, political affiliations, or environmental regulation. Ignoring testing CSD may lead to biased estimates, incorrect inferences, and misspecification in panel QR frameworks. The results in Table 4 column (2) indicate the rejection of the null hypothesis of CSD for each variable included. The test results imply the presence of CSD, which suggests the need to address this feature in the panel QR estimation, also suggestive of slope coefficients heterogeneity. Then, we employ first-generation panel unit root tests of LLC, IPS, and Breitung to ensure estimation validity and selection of the regression technique. According to Table 4, column (3) results related to LLC indicate all variables are stationary at levels, whereas column (4) to (7) results related to IPS and Breitung indicate that not all variables are stationary at the level, but rather at the first difference. To corroborate the stationarity tests, we employed second-generational unit root tests in heterogenous panels with CSD based on the average individual or augmented Dickey-Fuller (PESCADF), proposed by Pesaran (2007) and CIPS, which is parallel to Im et al. (2003). According to results in Table 3 columns (8) and (9), all statistics are significant at 1% LOS for the level, confirming the stationarity of all variables and the possibility concludes that variables are integrated at an order below 1.

To yield consistent estimations, we employed the Hashem Pesaran and Yamagata (2008) test for homogeneity of slope, whereas when the model consists of heterogeneous slopes when staggering homogeneous slopes, it will yield biased and inconsistent outcomes. The test allows for non-normally distributed errors and is particularly relevant where the assumption of common slope coefficients across countries may be inappropriate. According to results in Table 3, delta and adjusted delta statistics for all models reject the null hypothesis that coefficients of slope are homogeneous across cross-sectional units. The results imply that slope coefficients vary across the units in the panel and this means that countries exhibit distinct responses of finance on deforestation, hence the presence of heterogeneity prospect is confirmed. Then we assess the long-run relationship between sets of variables for different models using several tests of cointegration developed by Kao (1999), Pedroni (2004) and Westerlund (2007). Table 5 provides results related to cointegration tests and found that all of the tests were significant at 1% LOS, and the null hypothesis was significantly rejected. Moreover, the cointegration results of these different tests corroborate each other and indicate the cointegration among the selected variables, suggesting a stable long-term relationship. The existence of cointegration suggests that the deviations from the long-run equilibrium are temporary and will eventually correct themselves, which helps in constructing models that will predict long-term trends of financial inclusion in reducing deforestation rates in the region.

After acknowledging the presence of CSD, stationarity, heterogeneity, and long-term equilibrium relationship among variables, we utilized the conditional quantile regression (QR) in panel data with fixed effects, following the frameworks of Machado and Santos Silva (2019). The estimator is useful for analyzing the heterogeneous effects of financial inclusion indicators across the distribution of deforestation rates while accounting for individual effects and allowing for robustness to outliers, with the assumption that financial inclusion is not uniform across the entire distribution of the forest conversion. The QR with fixed effects control for unobserved heterogeneity across countries isolates the impact of covariates on deforestation by controlling for time-invariant characteristics of countries. Results in Table 6 show the method’s appropriateness in explaining usage, outreach, and financial inclusion in the region, rather than using the two-way fixed effect regression approach. Results indicate that financial outreach and inclusion combined variables are significant and negative across all lower 25$th$⁠, median 50$th$⁠, and upper 75$th$ quantiles, whereas financial usage is negative and significant at the 75$th$ quantile. The negative relationship across all quantiles suggests that as financial outreach and inclusion increase, the deforestation rates decrease, irrespective of where in the distribution of countries are located, and consistent throughout the entire distribution. The outcomes indicate that both low- and high-deforestation countries experience a reduction in the forest conversion when there is more financial outreach and inclusion. The fact that financial usage is significant only at the 75$th$ quantile suggests that its impact is most pronounced at the upper end of the deforestation allotment. Countries at the higher end may experience a reduction in the forest conversion as financial usage increases, whereas those in lower or middle quantiles are less affected. The insignificance at the lower quantiles implies that the effect of financial usage is not uniform across the entire distribution.

Heterogeneous effects are shown from the panel quantile regression results on different levels of income inequality in the region. While conditional QR focuses on the impact of covariates conditional on the value of other variables, we employ the unconditional QR framework that focuses on the entire distribution of the deforestation rates, providing a clear picture of how changes in financial indicators affect the distribution of forest conversion as a whole not just conditional on other variables. Table 7 presents results related to the unconditional QR estimator with time fixed effects, financial inclusion, and outreach, as shown to be negative and significant across all quantiles, whereas financial usage is insignificant negative to lower and upper quantiles, and positive at the median 50$th$ quantile. The findings corroborate that of conditional QR and suggest that greater financial outreach and inclusion reduce deforestation levels across the entire distribution of deforestation rates. The consistency across quantiles implies that access to finance services and broader coverage of banking lead to a broad, economy-wide reduction in deforestation. Findings show distinct responses of financial inclusion on forest conversion across different quantiles, reflecting variability that has been in line with Shahbaz et al. (2016), who observed that the benefits of financial access depend heavily on existing ecological conditions. This supports the notion that policies aimed at enhancing financial inclusion must be tailored to regional specifics to effectively combat deforestation.

In countries with low or very high levels of deforestation, increased financial usage does not emerge to have a substantial effect, suggesting that financial services are not the key factor in shaping environmental outcomes in regions with low deforestation rates. Results align with the hypothesis that improved financial access may lead to better sustainable practices. This is supported by Bu et al. (2023), who highlight that financial inclusion empowers communities to adopt sustainable agricultural practices, potentially mitigating detrimental economic activities such as illegal logging or land conversion for agriculture. Similarly, in countries with very high deforestation, financial usage may not play a critical role, likely due to the dominant drivers of deforestation related to large-scale industrial activities or entrenched economic practices that are less influenced by financial usage. The positive coefficient at the median quantile suggests that in countries with moderate levels of deforestation, increased financial usage may contribute to higher deforestation. This could be due to the fact that increased access to credit and financial services might influence investment in activities that are more capital-intensive and environmentally damaging such as agriculture expansion, logging, and land conversion for commercial use.

On control covariates, the economic performance proxied by GDP per capita, forest rents are significant and positive, whereas trade openness is negative and significant across all quantiles. Population growth is significant and negative at the 50$th$ quantile with financial usage, outreach, and inclusion, whereas significantly positive at the 35$th$ quantile with financial usage and inclusion. The agriculture area is significantly positive at the 25$th$ quantile with financial outreach and inclusion, while negative at the 50$th$ and 75$th$ quantiles with financial outreach, usage, and inclusion. The positive and significant relationship between GDP per capita, forest rents, and forest conversion across all quantiles suggest that as income levels rise in the SSA region, deforestation increases, indicating that higher income levels influence economic activities that put pressure on forests, and conversion increases as economic returns from forest increase, particularly across all quantiles of deforestation rates.

The positive correlation between GDP per capita and forest conversion revealed in our study can be aligned with findings by Agras and Chapman (1999) and Grossman and Krueger (1995), who discuss the Environmental Kuznets Curve (EKC) theory. They posit that economic growth initially leads to increased pollution and resource exploitation before societies begin to prioritize environmental sustainability as wealth increases. While our findings resonate with this theory, they also underline the need for a transition to sustainable practices as economies mature. The negative relationship for trade openness suggests that lower levels of deforestation are influenced by higher trade. This could be due to countries shifting toward more diversified and less land-intensive industries as they open up to international markets, reducing reliance on deforestation for economic growth.

The negative relationship of population growth at the 50$th$ quantile suggests that countries with moderate levels of deforestation are linked with higher population growth, accompanied by better land management practices that reduce pressure on forests. However, the positive relationship at the lower quantile implies that for regions with low deforestation, population growth is linked to increasing deforestation due to population pressure, leading to agricultural expansion or increased demand for land and resources, particularly in rural areas. An increase in agricultural area is associated with higher deforestation, and expansion of agricultural activities is indicated to be among the drivers of deforestation at lower quantiles, where the forest is still relatively abundant and more vulnerable to conversion for farming. Furthermore, at higher quantiles, an increase in agricultural area is associated with a decrease in deforestation. In regions with high deforestation, land is already extensively converted for agriculture, and further expansion has a diminishing effect on deforestation. The approach evaluates the asymmetric effects of economic prosperity policies indicated to have a little effect on the median, but a large effect on countries’ forest conversion at the bottom and top of the deforestation distribution. Our results indicate that agricultural expansion often coincides with increased forest conversion, aligning with studies by Mather (1992) and Rudel et al. (2005), which suggests that as economies develop, there is often an initial phase of forest loss due to agricultural practices. However, as societies reach higher economic thresholds, there tends to be a shift towards reforestation and sustainable land management practices.

Before the main models’ estimation, we employed the generalized QR estimator developed by Powell (2016) as a baseline model for robustness and sensitivity inference. The generalized QR estimator discourse ia an essential issue presented by conventional QR in panel data, and the incorporation of additional covariates changes the inference of the estimated parameters on the treatment covariates, the problem addressed by the generalized estimator, and generates unconditional quantile treatment effects in the presence of additional controls. We employ the estimator via numerical optimization procedures via the Monte Carlo Markov Chain (MCMC) optimization for $τ∈(0,1)$⁠. All exogenous explanatory and additional control variables are included as instruments. Table 8 presents results related to the generalized QR estimator, and a slight difference is observed in the coefficients of financial inclusion, confirming the robustness of our model, negative and significant.

Results remain consistent when applying the baseline model as a reference, which strengthens the case for the original findings. Financial outreach is positive and significant at the 25$th$ quantile, whereas financial usage is significantly positive at the 25$th$ quantile and negative at the 75$th$ quantile. The findings of the link between economic growth and forest conversion in the SSA countries may vary from other research in different regions. Methodology, control variables, time of inference, and GDP per capita are positive and significant, and international trade is significant and negative across all quantiles, agricultural area is positive and significant at the median and 75$th$ quantiles, whereas financial inclusion significantly positive across all quantiles with financial outreach and usage. Forest rents are different in sign of coefficients, significantly negative at the 25$th$ quantile, while positive at the median and 75$th$ quantiles. Moreover, the growing population in the region shows to be significant and positive across all quantiles with financial inclusion and financial usage, while negative at the 25$th$ quantile and positive at the median and 75$th$ quantiles with financial outreach.

In this sub-section, we include alternative measures of financial inclusion; the eight indices provided by the IMF- Global Financial Development Index (GFDI) database are abstracted as the panel of supply-side indicators of financial inclusion, including financial institutions and markets access, efficiency, and depth. According to results in Table 9 related to the fixed conditional QR estimator, the composite measure of financial institutions and accessibility of financial institutions are significant and negative, while that of financial markets is positive across all quantiles. However, the depth or size of financial institutions is significantly positive at the 25$th$ quantile and negative at the 75$th$ quantile, while efficiency in financial institutions is significantly negative at the median and upper 75$th$ quantiles of forest conversion. Furthermore, the depth of the financial market is significantly negative at the median and upper quantiles, while efficiency in the financial market is significantly positive at the 25$th$ quantile and negative at the 50$th$ quantile. The findings suggest that significant negative financial institutions, including the availability of banking, credit access, and other financial services, reduce forest conversion regardless of whether deforestation rates are low, moderate, or high.

This implies that well-developed and accessible financial institutions may facilitate investment in more sustainable activities that reduce the need to exploit forests for economic survival. For instance, access to formal credit and insurance can help growing rural populations and businesses transition to more sustainable agricultural or non-extractive industries. The positive link between the financial market, involving capital markets and more complex products, and deforestation implies that greater financial market activity enables investments in capital-intensive projects, which lead to higher deforestation. We emphasize that governments should implement environmental safeguards and green certification for financial products that fund large-scale projects that contribute to deforestation. Additionally, trade openness is negative, and population growth and GDP per capita are positive and significant across all quantiles. Different results are shown on the agricultural area, positive with financial institutions access and negative with depth and efficiency in financial institutions, whereas forest rents are significantly negative at the 25$th$ quantile and positive at the 50$th$ and 75$th$ quantiles with access, efficiency, and depth of financial institutions across all quantiles.

The negative relationship of conditional QR results suggests that policies promoting financial outreach and inclusion may have a broad and uniform positive impact on reducing deforestation rates, potentially reducing financial disparities across the growing population in the SSA region. The heterogeneity implies the importance of policy interventions aimed at expanding financial access, such as banking services, which could play an important role in reducing deforestation, benefiting both low- and high-performing countries. Policies or interventions focused on increasing financial usage, such as borrowing, saving, or using financial products, might benefit higher-income countries more than the lower-performing groups. The negative coefficient at the 75$th$ quantile could indicate that excess financial usage, particularly at higher levels of forest conversion distribution, might lead to negative outcomes such as over-indebtedness or inefficiency in financial management. This could be an early warning for policymakers to consider the risks of encouraging financial usage without proper financial literacy or safeguards, especially for countries at the higher end of deforestation distribution. Findings imply that policy should focus not only on increasing financial usage but also on enhancing financial literacy and regulating financial services to avoid potential negative consequences. Unconditional QRs are often employed in policy analysis to investigate how a policy impacts the entire distribution of forest conversion.  Our study understands how financial accessibility and usage impact the entire distribution of forest conversion across all quantiles without conditioning selected set of control variables. Policymakers often care about the distributional effects, and using both conditional and unconditional QR allows for more comprehensive policy evaluation by addressing both individual-level and aggregate-level impacts. In this study, we captured the diverse effects of financial indicators as heterogeneity revealed within countries for conditional QR, while insights were provided into the broader, population-wide effects for the unconditional QR approach.

Expanding outreach and inclusion, particularly in rural and underdeveloped areas, could serve as a sustainable development strategy that reduces pressure on the forest. Greater access to financial services likely boosts individuals and businesses to engage in more sustainable livelihoods, such as transitioning from extractive logging to more productive, sustainable ventures such as agroforestry, eco-tourism, or other non-extractive industries. The uniform negative effect suggests that financial inclusion policies may benefit all countries in the region, those with both high and low deforestation rates. Access to finance may help diversify income sources and enable communities to rely less on deforestation for economic survival. The divergence in the effects of financial markets and institutions across quantiles signifies the need for targeted financial policies. In countries with moderate deforestation, where financial usage increases deforestation, there might be a need for better regulation of credit for environmentally harmful activities. Financial institutions could be encouraged to lend for sustainable activities, such as reforestation, renewable energy, or sustainable agriculture, rather than practices that exacerbate deforestation. To mitigate the unintended positive impact of financial usage on deforestation, policy should focus on green financing initiatives that tie credit to environmentally sustainable projects, which would ensure that increased financial accessibility does not lead to environmental degradation.

Reducing deforestation within and across countries is among the aims of the SDGs, and increasing easy accessibility of financial services and products becomes a policy priority to improve inclusive growth and achieve sustainable economic development. Financial inclusion plays a crucial role in reducing the forest conversion gap between lower- and higher-income individuals in a society. Theoretical and empirical outcomes related to the relationship between financial inclusion and deforestation are controversial; little is known concerning the impact on the finance-deforestation nexus in the SSA region. Financial inclusion is more powerful in transforming and enhancing real-time forest and sustainable management activities that enhance income distributions. Extended financial services can provide easy access and flow of income from physical locations of financial institutions and markets to underserved individuals who rely on the forest for economic survival. This study employs conditional and unconditional quantile regression (QR) to examine the heterogeneous impacts of demand- and supply-side indicators of financial inclusion on forest conversion in SSA economies between 2004 and 2020. Results indicate that significant negative financial inclusion and access to financial institutions reduce deforestation, whereas financial usage and markets increase deforestation rates. The positive impact of financial markets on forest conversion suggests the need for regulatory policies to ensure investments influenced by financial markets do not lead to forest and environmental degradation. Governments and policymakers should consider implementing environmental safeguards and green certification for financial products that fund large-scale projects, such as agricultural expansion or infrastructure development, which contribute to deforestation.

To counteract the potentially negative impact of financial institutions and inclusion on forests, policymakers could promote the development of green bond markets and sustainable finance initiatives that channel capital into environmentally friendly projects. Encouraging investments in reforestation, renewable energy, and sustainable infrastructure would reduce market expansion on deforestation, despite stronger accessibility of financial institutions likely providing the necessary support for sustainable and inclusive economic activities, which in turn reduce reliance on forest exploitation. Expanding access to financial services in rural and forested areas is key to reducing deforestation. Governments and development organizations should focus on financial inclusion initiatives that increase access to credit, savings, and insurance for low-income and forest-dependent communities. Mobile banking and digital services can play an important role in reaching underserved areas, and these services should be paired with programs that encourage investments in sustainable practices. This research is not without restrictions, prior research on the deforestation debate has used various macroeconomic variables that may impact forest variables, such as technological expansion, debt, price indices, and quality of government. There are other possible covariates including afforestation policies, taxes and prices on agricultural and forest products that could be applicable for inclusion in this inference, even though data were not accessible for all economies in the region. Moreover, the longer period of forest conversion data available for this research was deemed inadequate to describe several changes in seemingly uninterrupted forest wealth and environmental externalities. No matter what limitations are mentioned, a systematic change of economic pursuit while strengthening international trade and growing populations within SSA economies will curtail the loss of forest area and resources.