The Effect of Digital Financial Inclusion on Inclusive Growth and Poverty in Emerging and Developing Economies: A System-Generalized Method of Moments Model

Abstract

This study investigates the effect of digital financial inclusion on both inclusive growth and poverty in Emerging and Developing Economies (EMDEs). While previous research has examined the relationship between digital financial inclusion and inclusive growth or poverty, there is a notable gap in the literature regarding the indirect effect of digital financial inclusion on poverty through inclusive growth. Additionally, many existing studies have focused on specific countries, leaving a need for a cross-sectional analysis across various EMDEs, particularly in under-researched regions such as Central Africa, Southern Africa, West Africa, Oceania, and South-Eastern Europe. To address these gaps, this research employs panel data and the System-Generalized Method of Moments (GMM) as the main estimation technique, which provides robust and efficient estimates while addressing potential endogeneity. The study constructs a new digital financial inclusion index using the Principal Component Analysis (PCA) approach to enable consistent cross-country comparisons. The findings reveal that digital financial inclusion has a positive and significant effect on inclusive growth, indicating that as digital financial inclusion increases, inclusive growth increases as well. The results also demonstrate that inclusive growth has a negative and significant effect on poverty, suggesting that equitable economic expansion is a key driver of poverty reduction. These findings provide policymakers and governments in EMDEs with valuable insights, helping them prioritize investments and strategies that leverage digital financial inclusion to foster inclusive growth and alleviate poverty.

1. Introduction

Poverty continues to be one of the biggest challenges faced by developing countries, as most of the people living in these countries do not have access to essential financial services. For the past 20 years, credit unions, savings cooperatives, and microfinance organizations have made significant strides. Yet most of the world’s impoverished communities still lack access to formal financial institutions (Xu et al., 2024). Digital financial services can help alleviate poverty levels by providing impoverished populations with access to basic financial services and products. By adopting and using technology, digital financial inclusion can end the poverty cycle by liberating previously disadvantaged groups and fostering inclusive growth, particularly in EMDEs.

1.1. Digital Finance: Driving Inclusion and Growth

There is currently no consensus on how to define digital finance. However, there is a general agreement that it encompasses all technology and/or infrastructure, products, and services that assist businesses and people in accessing credit facilities, payments, and savings through online channels (Ozili, 2018). Digital finance has attracted significant interest from scholars and policymakers as a means of alleviating poverty and increasing financial inclusion. All participants involved in the financial sector, including users of financial services, companies that supply digital financial services (such as application programming interfaces, web, machine learning, and mobile devices), regulators, decision-makers, and the economy, benefit greatly from digital finance.

Global dialogues emphasize that EMDEs should strive for inclusive growth that goes beyond economic growth (Kouton, 2019), as economic growth alone is not sufficient to achieve equitable growth. Digital financial inclusion can play a crucial role in fostering inclusive growth in EMDEs by providing access to financial services and products to financially excluded individuals and businesses.

1.2. Digital Inclusion, Growth, and Poverty: A Research Gap

Most prior research has focused on either the impact of digital financial inclusion on inclusive growth (Khera et al., 2021; Ozturk & Ullah, 2022; Thaddeus et al., 2020; Zhang et al., 2020) or the impact of financial inclusion on poverty (Lyons et al., 2020; Koomson et al., 2020; Peng & Mao, 2023; Polloni-Silva et al., 2021). These studies have demonstrated that digital finance reduces poverty and is closely linked to the advancement and expansion of inclusive economic growth. Furthermore, research has shown that financial inclusion is a potent tool for combating poverty. However, the existing literature lacks explicit studies examining the effect of digital financial inclusion on both inclusive growth and poverty simultaneously. Put differently, the effect of digital financial inclusion on poverty through the intervention of inclusive growth is not available in the existing literature. Therefore, it is essential to examine how digital financial inclusion contributes to poverty reduction through inclusive growth, as these two topics are closely interconnected and crucial to achieving equitable growth.

Moreover, there is insufficient research on the effects of digital financial inclusion on inclusive growth and poverty from EMDEs’ perspectives, as most studies have focused on specific countries rather than a cross-section of countries. As a result, there has been little attention to simultaneously investigating the effects of digital financial inclusion on inclusive growth and poverty across various EMDEs. Due to the higher prevalence of financial exclusion in EMDEs, they require a more thorough investigation, as they have not been sufficiently covered in the existing literature, thereby creating a gap to explore. In the African region, most studies found that increased digital financial inclusion is associated with a decrease in poverty (Kelikume, 2021; Koomson et al., 2020; Lyons et al., 2020; Wale-Awe & Evans, 2023). Similarly, in the Asian region, results from most studies suggest that digital financial inclusion can significantly lower poverty levels (Peng & Mao, 2023; Tay et al., 2022; Xie, 2023). Additionally, studies in the African and Asian regions also found that digital financial inclusion significantly and positively affects inclusive growth (Afolabi, 2020; Doku et al., 2023; Sarpong & Nketiah-Amponsah, 2022; Wang & Yu, 2024; Zhang et al., 2020; Yang et al., 2024), which ultimately leads to poverty reduction. Other regions of interest in the current study, such as Oceania and South-Eastern Europe, have received limited or no scholarly attention in the existing literature. Therefore, this research will contribute to the scarce literature in these regions. Examining several countries simultaneously can help us identify trends that might not be apparent or obvious in a single country. This may enable broader, more applicable conclusions regarding our specific topic in this study.

1.3. Study Contributions: Digital Inclusion, Growth, and Poverty Analysis

The analytical techniques employed in this studyseek to empirically establish the effect of digital financial inclusion on inclusive growth and poverty in EMDEs. Considering the shortcomings, this study adds to the body of knowledge in four ways.

(a)

Methodology Contribution: System-GMM and PCA Index

First, this study empirically examined the effect of digital financial inclusion on inclusive growth and poverty in EMDEs by employing the System-GMM as the main estimation technique. System-GMM offers reliable and effective estimates, addresses endogeneity, and better fits panel data. The study constructs a new digital financial inclusion index using principal component analysis (PCA). The index enables a cross-country comparison of the level of digital financial inclusion and thus provides a consistent method for monitoring progress across various countries.

(b)

Contribution: Digital Finance and Poverty Through Inclusive Growth

Second, the effect of digital financial inclusion on poverty through the intervention of inclusive growth is not available in the existing literature. Therefore, there is a need to examine the effect of digital financial inclusion on poverty through the lens of inclusive growth. Unlike economic growth, inclusive growth ensures that the benefits of economic growth are allocated equitably, thereby reducing poverty.

(c)

Regional Contribution: Digital Inclusion, Growth, Poverty Analysis

Thirdly, this study investigates the effect of digital financial inclusion on inclusive growth and poverty in Central Africa, Southern Africa, West Africa, Oceania, West Asia, and South Asia, as there are scarce studies in these regions, and hence they require more consideration.

(d)

Methodological Contribution: Justifying Panel Data Use

Finally, this study employed panel data because it provides more insightful data, improves efficiency, captures more variability, and reduces collinearity (Greene, 2005; D. Gujarati, 2003; Wooldridge, 2002). Furthermore, panel data allows for controlling for unobserved country- and time-specific heterogeneity, which may be correlated with the explanatory variables included in the study (Akbar et al., 2011).

1.4. Study Findings: Digital Inclusion, Growth, and Poverty Links

The findings reveal that digital financial inclusion has a positive and significant effect on inclusive growth in EMDEs at a 10% significance level. The sign and magnitude indicate that for each percentage-point increase in digital financial inclusion, inclusive growth increases by 0.7428 units. Digital financial inclusion is crucial to accomplishing inclusive growth as the process and effort of developing a digital society continue to grow (Chen et al., 2022).

Moreover, the findings also demonstrate that inclusive growth has a negative and significant effect on poverty in EMDEs at a 5% significance level. The sign and magnitude indicate that for each percentage-point increase in inclusive growth, poverty decreases by 0.0415 units. Concerns over rising inequality despite economic growth have culminated in an increasing awareness of the significance of inclusive growth and development in international development (Kuss et al., 2022).

Investigating the effects of digital financial inclusion on inclusive growth and poverty in EMDEs will enable policymakers and governments to identify relevant paths in which digital financial inclusion fosters inclusive growth and reduces poverty. Additionally, this study can help policymakers and governments prioritize capital investments in industries with the greatest prospects for increasing inclusive growth and alleviating poverty.

2. Literature Review

2.1. Theoretical Literature

The theory of Financial Intermediation underpins this study (Goldsmith, 1969; McKinnon, 1973; Shaw, 1973). Financial intermediation is the process by which financial institutions (such as mutual funds, banks, pension funds, and insurance companies) channel funds between those who save (depositors) and those who invest (investors) (Gorton & Winton, 2002). The agency theory and the theory of information asymmetry form the foundation of the financial intermediation theory (Andrieş, 2009). High transaction costs, regulatory constraints, and incomplete information are among the factors that explain why financial intermediaries exist. Nonetheless, information asymmetry is the leading and most widely used approach in the existing literature on financial intermediation theory. The asymmetry concept holds that parties in a transaction have unequal amounts of information, which may result in problems such as adverse selection and moral hazard.

This study investigated the effect of digital financial inclusion on inclusive growth and poverty in EMDEs. The theory of financial intermediation is thus linked to this study as high transaction costs discourage individuals from participating in the formal financial sector, ultimately resulting in high financial exclusion, high poverty levels, and low inclusive growth. Moreover, when market information is not shared equally, this tends to discourage individuals from participating in the financial market, increasing financial exclusion. On the other hand, lower transaction costs tend to increase digital financial inclusion, reduce poverty, and boost inclusive growth. Furthermore, when financial institutions are well-regulated and less risky, it encourages households to participate in the financial sector. Therefore, the theory of financial intermediation is suitable for this study.

2.2. Review of Empirical Literature

Several studies have examined the relationship between digital financial inclusion and either inclusive growth or poverty, using different methodologies, data, and regions in the existing literature. Therefore, this section divides the empirical literature into studies that have examined (1) digital financial inclusion and poverty and (2) digital financial inclusion and inclusive growth.

2.2.1. Digital Financial Inclusion and Poverty

Kelikume (2021) investigated the nexus between poverty reduction, financial inclusion, mobile phones, the informal economy, and the internet. The system-GMM technique and a panel dataset spanning 1995 to 2017 from 42 countries were employed in the study. The findings revealed a positive and significant relationship among internet use, mobile penetration, and poverty alleviation. Moreover, the study found that increased financial inclusion reduces poverty.

Using a two-step System-GMM, Toda causality test, and simultaneous-equations models, Wale-Awe and Evans (2023) examined the causal relationship between the growth-poverty-inequality triangle and digital financial inclusion from 1995 to 2018 in 42 African countries. The cited study found that increased digital financial inclusion leads to poverty alleviation and reduced inequality. The findings also revealed a bi-causality between digital financial inclusion and poverty. Therefore, Wale-Awe and Evans (2023) suggest that enhancing access to digital financial services throughout the African continent is crucial to improving income levels, reducing poverty, and fostering an equitable income distribution.

Focusing on South Asia and Sub-Saharan Africa, Lyons et al. (2020) examined the association between financial and digital inclusion and poverty across seven developing countries. The cited study found that substantial reductions in poverty are associated with increases in several measures of financial and digital inclusion. The study underscores the significance of digital financial services in assisting and maintaining efforts to reduce poverty in developing countries.

In China, Peng and Mao (2023) investigated the impact of digital financial inclusion on urban households’ relative poverty levels. The probit model, the mediating effect model, and the instrumental variable approach were used to perform empirical research. According to the findings, digital financial inclusion lowers the likelihood that urban households will experience poverty. Moreover, the study discovered that digital financial inclusion increases urban families’ engagement in the financial sector and encourages entrepreneurship.

A study by Tay et al. (2022) yielded mixed results. Using a systematic literature review, the cited authors examined digital financial inclusion across countries. The study revealed that, to combat poverty, developing countries, mostly in Asia, embrace and expand digital financial inclusion. However, the findings also showed that, in terms of access to and use of digital financial services, disparities persist by gender, between the wealthy and the poor, and between urban and rural areas in developing countries.

2.2.2. Digital Financial Inclusion and Inclusive Growth

In Sub-Saharan Africa, Sarpong and Nketiah-Amponsah (2022) employed the system-GMM estimation technique to empirically investigate the relationship between financial inclusion and inclusive growth. The study used a panel dataset of 46 countries spanning 2004 to 2018. The results indicate that, compared with other factors, financial services have a measurable and significant impact on inclusive growth. Inclusive growth improves by 0.03 units for every increase in the use of financial services.

To establish whether digital financial inclusion promotes inclusive growth in China, Zhang et al. (2020) compared data from the China Family Panel Studies (CFPS), a representative household survey, with the digital financial inclusion index, which assesses the progress of digital finance in China. The authors found that digital finance increases household income in China, particularly in rural areas. As a result, digital finance helps China’s inclusive growth by closing regional and urban-rural gaps. By efficiently allocating capital, reducing information asymmetry, and mitigating risk, financial intermediation can increase access to financial services and foster economic growth for underprivileged groups.

Yang et al. (2024) investigated the influence of digital financial inclusion on inclusive growth. The study used a panel dataset of 30 provinces in China covering the period from 2011 to 2021. The entropy-weighted method of the fixed-base range was used as the estimation technique for the study. The results indicate that inclusive growth is significantly influenced by digital financial inclusion. Further, the study found that inclusive growth can be achieved through digital financial inclusion by enhancing innovation and human capital.

2.3. Summary of the Empirical Review Findings

Financial inclusion is viewed as one of the key components of inclusive growth and poverty reduction in developing countries (Khan et al., 2020). Furthermore, financial inclusion is considered a strong foundation for a country’s financial infrastructure (Liaqat et al., 2022). Most studies have demonstrated that digital finance reduces poverty and is closely linked to the advancement and expansion of inclusive economic growth. Furthermore, research has shown that financial inclusion is a potent tool for combating poverty, particularly in EMDEs.

However, the existing literature does not explicitly contain studies examining the effect of digital financial inclusion on both inclusive growth and poverty. It is essential to examine how digital financial inclusion affects poverty and inclusive growth, as these two topics are closely connected and important for equitable growth. Given how closely related inclusive growth and poverty are and how they influence each other, it is essential to investigate how digital financial inclusion affects both simultaneously, especially in EMDEs.

Notably, the simultaneous assessment of the impact of digital financial inclusion on both inclusive growth and poverty remains under-researched, hence the rationale for this study. Put differently, the effect of digital financial inclusion on poverty through the intervention of inclusive growth is not available in the existing literature. Therefore, there is a need to investigate the effect of digital financial inclusion on poverty through inclusive growth interventions. Unlike economic growth, inclusive growth ensures that the benefits of economic growth are allocated equitably, thereby reducing poverty.

3. Methodology: Modeling Digital Inclusion’s Impact on Inclusive Growth and Poverty

3.1. Data Selection and Sources

This study was quantitative and used data generated from secondary sources. The study’s time period, spanning 2013 to 2022, was selected primarily because consistent data were available for the key variables across the chosen EMDEs. The year 2013 is particularly important, as it marks a period when several EMDEs began actively collecting and reporting data on digital financial inclusion indicators, which is crucial for constructing the Digital Financial Inclusion Index (DFII). The period captures nearly a decade of rapid acceleration in the adoption of digital finance technologies, making it highly relevant for examining the intended effect of DFI on inclusive growth and poverty alleviation.

The selection of 21 EMDEs addresses the notable gap in the existing literature, which has largely focused on single-country studies. By employing a cross-sectional analysis with a diverse panel dataset, the study can identify broader trends and draw conclusions that are more broadly applicable across developing economies, a finding that might be obscured in single-country investigations. This approach is crucial given the higher prevalence of financial exclusion in EMDEs.

Crucially, the sample was deliberately chosen to provide a regional contribution by including countries from several under-researched regions. The study’s focus on Central Africa, Southern Africa, West Africa, Oceania, West Asia, and South Asia aims to contribute to the scarce literature in these areas, which have received limited to no scholarly attention from previous studies. The countries used for the study are as follows: Albania, Bangladesh, Benin, Botswana, Burkina Faso, Cameroon, Côte d’Ivoire, Ghana, Guinea, Guinea-Bissau, Lesotho, Mali, Mauritius, Niger, Rwanda, Samoa, Senegal, Solomon Islands, Togo, Zimbabwe, Pakistan.

The following data sources were used: the Financial Access Survey of the IMF (FAS-IMF) and the World Development Indicators (WDI). The data spans 2013 to 2022. The FAS is a supply-side dataset on access to and use of financial products and services, introduced in 2009 to help policymakers evaluate and monitor financial inclusion and compare progress with peers. The WDI is the World Bank’s main collection of development indicators, compiled from various reputable international sources. It offers the most recent and reliable global data.

3.2. Data Selection, Description, and Measurement of Variables

Data were selected based on the various dimensions and indicators used by the study.

3.2.1. Dependent Variable (s)

(a)

Inclusive Growth

This study used GDP per person as a proxy for inclusive growth. GDP per employed person (constant 2017 Purchasing Power Parity $) has been used as a proxy for inclusive growth in several studies (Adeniyi et al., 2021; Kouton, 2019; Olanrewaju et al., 2020; Raheem et al., 2018). The focus of this measure is the people who work in the production process. Inclusive growth ensures that economic growth is distributed equitably.

(b)

Poverty

The presented study employed household consumption expenditure as a proxy for poverty levels of the selected countries due to the lack of poverty data in these countries. Household consumption expenditure has been validated by Akanksha and Mohanty (2010) and, more recently, by Musakwa and Odhiambo (2021). The cited studies employed household consumption expenditure to reflect the income aspect of poverty. Quartey (2008) attests that household consumption expenditure data among the impoverished are typically reported more frequently and accurately than income data. Moreover, Nyasha et al. (2017) also justify using household consumption expenditure as a proxy for poverty as it is found to be a reliable variable. Existing literature has used household consumption expenditure as a proxy for poverty (Dhrifi, 2015; Kaidi et al., 2019; Musakwa & Odhiambo, 2019; Nsiah & Tweneboah, 2024; Quartey, 2008). Data on household consumption expenditure were collected through a single cross-sectional survey for an adequately representative sample of the selected countries, employing households as the unit of observation.

3.2.2. Explanatory Variables: Digital Financial Inclusion

The following dimensions and indicators were used to measure digital financial inclusion.

(a)

Access (Penetration of Basic Financial Services)

Mobile money agent outlets active (per 100,000 adults) and mobile money agent outlets active (per 1000 km²) serve as indicators of access within the geographic or demographic penetration (Banna, 2020). These indicators are important and necessary for the internet, mobile money, and mobile banking to be used as new avenues for accessing financial services. Moreover, in several markets, the availability of mobile money agents in remote areas has significantly increased access to financial services (Khera et al., 2021). Data for these indicators were sourced from the FAS-IMF.

(b)

Usage

For the usage dimension, the number of active mobile money accounts (per 1000 adults), the number of mobile and internet banking transactions (during the reference year) per 1000 adults, and the value of mobile and internet banking transactions (during the reference year) (% of GDP) as indicators (Banik & Roy, 2023; Banna, 2020). Data for these indicators were sourced from the FAS-IMF. Digital financial inclusion can reach more households, providing financial access to individuals who were previously excluded, thereby promoting inclusiveness.

This study further employed the following control variables: Age Dependency Ratio, Gross National Income (GNI) per capita, Education, Inflation, General Government Final Consumption Expenditure (% of GDP) (GOVEXP), Broad money to GDP (Money Supply, Population Growth, Institutional Quality, and GDP Growth.

3.3. Estimation Techniques

The first step in the investigation was to reduce the dimensionality of the data using principal component analysis (PCA). This study contributes to the literature by using dynamic panel methods in the analysis. Dynamic panel methods can produce effective estimators even when heteroscedasticity is present, and serial correlation and endogeneity are corrected. Considering that the current study used dynamic panel data to estimate the effect of digital financial inclusion on inclusive growth and poverty in EMDEs, the System-Generalized Method of Moments (GMM) is employed as the main technique. Moreover, for robustness checks and comparison, this study employs the difference-GMM as an alternative estimation technique.

3.3.1. Principal Component Analysis

Using the approach of Cámara and Tuesta (2014) adopted by Khera et al. (2021) and Ismael and Ali (2021), this study employs a two-stage Principal Component Analysis (PCA) to construct a new financial inclusion index. Constructing a financial inclusion index for EMDEs allows for the assessment of these countries’ progress over time. This allows for comparing financial inclusion levels across countries. PCA is a useful technique to minimize information loss while reducing the number of variables that reflect the same concept (Datta & Singh, 2019).

The PCA technique can decrease datasets to smaller dimensions while retaining as much of the original data as possible. The PCA technique scientifically calculates the weights for each dimension rather than manually assigning them, and it also eliminates correlated variables that do not contribute to decision-making. Moreover, using a two-stage PCA reduces the likelihood that the PCA is skewed towards the weight of indicators for a given dimension.

The index achieves multidimensionality by aggregating indicators across two core dimensions, access (e.g., mobile money agent outlets per 100,000 adults) and usage (e.g., active mobile money accounts and transaction value/volume). These specific indicators are highly relevant in the EMDE context, capturing how digital platforms, especially mobile money, increase service availability in remote areas and facilitate the actual utilization of financial products by the financially excluded.

While alternative weighting schemes, such as simple arithmetic means or raw indicator scores, could be employed, this study utilizes a two-stage PCA to minimize information loss and eliminate the risk of manual weighting bias. A simple mean of standard scores assumes that each dimension of digital financial inclusion contributes identically to the final index. However, the PCA technique scientifically calculates weights by capturing the greatest degree of variation among indicators. This ensures that the constructed DFII is more representative of the actual financial landscape in diverse EMDEs, where the expansion of digital access points (like mobile agents) may have a statistically different impact than the frequency of digital usage.

The PCA method is demonstrated in what follows.

(a)

The first stage of PCA: Estimate the dimensions (access and usage)

This study uses indicators for access and usage dimensions (as per the approach of Khera et al., 2021). This essentially means that the two endogenous variables Y^a and Y^u are unobserved for each index and must be estimated using their corresponding parameters (τ and δ) in the equation below:

$${\mathrm{DFI}}_{\mathrm{i}}={w_{1}Y}_i^a+{w_{2}Y}_i^u+{\mathrm{e}}_{\mathrm{i}}$$

(1)

• where DFI is the digital financial inclusion index of a country i
• e_i is the variation due to error
• w₁, w₂, represent the weights of each dimension; and
• $Y_i^a$, $Y_i^u$ represent dimensions for access and usage. These dimensions are further computed as follows:

$$\begin{array}{c}{Y}_i^a={\mathsf{\tau }}_1 \mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{e} \mathrm{M}\mathrm{o}\mathrm{n}\mathrm{e}\mathrm{y} \mathrm{A}\mathrm{g}\mathrm{e}\mathrm{n}\mathrm{t} \mathrm{p}\mathrm{e}\mathrm{r} \mathrm{100,000} \mathrm{A}\mathrm{d}\mathrm{u}\mathrm{l}\mathrm{t}\mathrm{s}+{\mathsf{\tau }}_2 \mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{e} \mathrm{M}\mathrm{o}\mathrm{n}\mathrm{e}\mathrm{y} \mathrm{A}\mathrm{g}\mathrm{e}\mathrm{n}\mathrm{t}\\ \mathrm{p}\mathrm{e}\mathrm{r} 1000 {\mathrm{km}}^2+{\mathrm{e}}_{\mathrm{i}}\end{array}$$

(2)

$$\begin{array}{c}{Y}_i^u={\mathsf{\delta }}_1 \mathrm{A}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e} \mathrm{M}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{e} \mathrm{M}\mathrm{o}\mathrm{n}\mathrm{e}\mathrm{y}+{\mathsf{\delta }}_2 \mathrm{N}\mathrm{o} \mathrm{o}\mathrm{f} \mathrm{I}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{n}\mathrm{e}\mathrm{t} \mathrm{B}\mathrm{a}\mathrm{n}\mathrm{k}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{T}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}+{\mathsf{\delta }}_3\\ \mathrm{V}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e} \mathrm{o}\mathrm{f} \mathrm{I}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{n}\mathrm{e}\mathrm{t} \mathrm{B}\mathrm{a}\mathrm{n}\mathrm{k}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{T}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{s}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}+{\mathrm{e}}_{\mathrm{i}}\end{array}$$

(3)

Significant differences exist between country-specific values of the various financial inclusion indicators. Therefore, to improve data comparability, the Min–Max approach is used to normalize the dataset. By subtracting the minimum value and dividing it by the range of the indicators’ values, this procedure puts all the distinct indicators within an equal range between 0 and 1 (Le et al., 2019). The normalization formula is presented in Equation (4) below:

$${\mathrm{X}}_{\mathrm{i},\mathrm{d}}={\displaystyle \frac{{(x}_i-m_i)}{(M_i-m_i)}}$$

(4)

where x_i indicates the value of indicator i, $m_i$ is the minimum value of indicator i and $M_i$ is the maximum value of dimension i. X_i,d represents the standardized value of indicator i of dimension d.

The estimator of each dimension is represented by the following weighted averages.

$$Y_i^a={\displaystyle \frac{{\sum }_{j,k=1}^p{\lambda }_j^a{PC}_{ki}^a}{{\sum }_{j=1}^p{\lambda }_j^a}}$$

(5)

$$Y_i^u={\displaystyle \frac{{\sum }_{j,k=1}^p{\lambda }_j^u{PC}_{ki}^u}{{\sum }_{j=1}^p{\lambda }_j^u}}$$

(6)

The correlation matrix is defined as Rp (pxp) of the p standardized indicators for each dimension. ${\lambda }_j$ (j = 1, …, p) is described as the jth eigenvalue of each of the dimensions. j represents the number of principal components (PC). The first PC accounts for the greatest degree of variation in the explanatory variables. γj(px1) is the eigenvector of the correlation matrix. The assumption made is that λ1 > λ2 > ⋯ > λp and represents where P_k (k = 1, …, p) as the k-th PC (weights).

The second stage of PCA: After obtaining the dimension indices, another PCA is conducted to determine the weights for financial inclusion.

The overall financial inclusion index is computed by combining all the separate indices into Equation (1) to estimate the parameter (λ):

$${TFI}_i={\displaystyle \frac{{\sum }_{j,k=1}^p{\lambda }_j{PC}_{ki}}{{\sum }_{j=1}^p{\lambda }_j}}$$

(7)

PC_ki (which represents each component) is expressed as a linear combination of the three dimensions and their respective eigenvectors of the correlation matrices as follows:

$${\mathrm{PC}}_{1\mathrm{i}}={\mathsf{\pi }}_{11}{Y}_i^a+{\mathsf{\pi }}_{12}{Y}_i^u$$

(8)

$${\mathrm{PC}}_{2\mathrm{i}}={\mathsf{\pi }}_{21}{Y}_i^a+{\mathsf{\pi }}_{22}{Y}_i^u$$

(9)

By replacing the PC_ki as written in Equations (8) and (9), the financial inclusion index can therefore be expressed as follows:

$${TFI}_i={\displaystyle \frac{{\sum }_{j=1}^3{\lambda }_j({\mathsf{\pi }}_{j1}{Y}_i^a+{\mathsf{\pi }}_{j2}{Y}_i^u)}{{\sum }_{j=1}^3{\lambda }_j}}$$

(10)

Equation (10) can thus be written as a linear equation for the weighted average of the financial inclusion index:

$${\mathrm{DFI}}_{\mathrm{i}}=w_1{Y}_i^a+w_2{Y}_i^u+{\mathrm{e}}_{\mathrm{i}}$$

(11)

The digital financial inclusion index is then used to classify the various countries into three groups.

0.5 $<DFI\le 1$ illustrates high digital financial inclusion.

0.3 $\le DFI<0.5$ demonstrates medium digital financial inclusion.

0 $\le DFI<0.3$ shows low digital financial inclusion.

To ensure that the results are not a byproduct of the weights assigned by the PCA, a sensitivity analysis was conducted by re-estimating the main model using an alternative Digital Financial Inclusion Index constructed through a simple arithmetic mean (equal weighting). All five underlying indicators were first normalized using min-max scaling to a 0–1 range to ensure comparability. The correlation between the PCA-derived index and this equal-weighted index was found to be exceptionally high (r = 0.9937), suggesting that the structural variation in digital financial inclusion across the sampled EMDEs is consistent regardless of the weighting scheme. As shown in

Table 1. Robustness and Index Weighting Scheme.

Variable	(1) Baseline (PCA Weights)	(2) Robustness (Equal Weights)
Digital Finance Index	0.506 *	0.498 *
	(0.182)	(0.189)
Statistical Significance	p < 0.01	p < 0.01
AR(2) p-value	0.45	0.455
Hansen p-value	0.312	0.318

Note: Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

, the coefficient for the equal-weighted index remains positive and statistically significant (p < 0.01), with the magnitude of the effect staying nearly identical to the baseline PCA model. This confirms that the observed relationship is robust and not sensitive to the specific statistical weights generated by the PCA.

3.3.2. Panel Estimation Techniques: System-Generalized Method of Moments (GMM)

This study employed system-GMM to estimate the effect of digital financial inclusion on inclusive growth and poverty in EMDEs. Hansen (1982) was the first to introduce GMM approaches. Holtz-Eakin, Newey, and Rosen further refined them (Holtz-Eakin et al., 1988), while Arellano and Bond (1991) advocated for the difference-GMM. The difference-GMM estimator is designed for data with numerous individuals but for few periods, that is, small T and large N panels.

The system-GMM was then outlined by Arellano and Bover (1995) and extensively developed by Blundell and Bond (1998). The system-GMM formulates supplementary orthogonality conditions that make more valid instruments accessible and achieve efficiency gains. System-GMM addresses heteroscedasticity problems, bias of omitted variables, endogeneity between variables, and autocorrelation (Anjum et al., 2022). Employing the panel data technique enables the endogeneity of the model or regressors to be addressed while addressing several periods and individual effects (Labra Lillo & Torrecillas, 2018). Furthermore, dynamic panel methods can produce estimators even when heteroscedasticity is present, and serial correlation and endogeneity are corrected.

This study used the system-GMM instead of the difference-GMM estimator as the System-GMM offers reliable and effective estimates, solves the endogeneity issue, and better fits panel studies. Like difference-GMM, the system-GMM estimator is developed for data with many individuals but for few periods, that is, small T and large N panels. The System-GMM produces more accurate estimates with a minimal bias relative to the difference- GMM (Blundell & Bond, 1998).

The System-GMM is the most appropriate estimation technique because it directly addresses the critical endogeneity concerns inherent in analyzing the relationships between Digital Financial Inclusion (DFI), Inclusive Growth (IG), and Poverty in EMDEs. The primary source of endogeneity is reverse causality or simultaneity bias, particularly between DFI and IG. While the study hypothesizes that DFI positively affects IG, it is highly probable that higher levels of inclusive growth spur greater investment in and adoption of digital financial services, simultaneously driving DFI. Additionally, since the models are dynamic, including the lagged dependent variable (${IG}_{it-1}\mathrm{a}\mathrm{n}\mathrm{d}{\mathrm{P}\mathrm{O}\mathrm{V}}_{it-1}$) as a regressor, the standard Fixed-Effects estimator would be inconsistent. System-GMM is further justified because it explicitly controls for unobserved country-specific heterogeneity ($u_{it}$), which is likely correlated with the explanatory variables, such as institutional quality or historical development paths.

The System-GMM technique provides a superior solution compared to alternatives like the Fixed-Effects model or simple Instrumental Variable (IV) regression. Unlike the Fixed-Effects estimator, System-GMM uses internal instruments (lagged values of the variables themselves) to ensure consistent estimates in the presence of dynamic effects and fixed country effects. Moreover, it is preferred over the difference-GMM because it yields more accurate estimates with minimal bias, especially when the variables are persistent, as is common with inclusive growth and poverty series.

This study’s period is from 2013 to 2022, with 21 EMDEs used. The System-GMM is suitable, as alluded to earlier, provided N > T to avoid large instruments. The study uses an annual panel. Another way to evaluate whether the estimation technique should be static or dynamic is to check whether the dependent variable is lagged and whether it is part of the regressors. If it lags, then the suitable approach is dynamic (D. N. Gujarati et al., 2012).

The System-GMM is demonstrated below:

(a)

Model 1

$${IG}_{it}={IG}_{it-1}+{\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{F}\mathrm{I}}_{it}+{\mathsf{\alpha }}_2{\mathrm{I}\mathrm{N}\mathrm{F}}_{it}+{{\mathsf{\alpha }}_3{\mathrm{P}\mathrm{G}}_{it}+{\mathsf{\alpha }}_4{\mathrm{G}\mathrm{O}\mathrm{V}\mathrm{E}\mathrm{X}\mathrm{P}}_{it}+{\mathsf{\alpha }}_5{\mathrm{G}\mathrm{N}\mathrm{I}}_{it}+{\mathsf{\alpha }}_6{\mathrm{D}\mathrm{P}\mathrm{R}}_{it}+{\mathsf{\alpha }}_7{\mathrm{I}\mathrm{Q}\mathrm{I}}_{it}+u_{it}+\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(12)

where ${IG}_{it}$ represents inclusive growth, ${IG}_{it-1}$ is the inclusive growth lag, $\mathsf{\alpha }$ is the coefficient of the estimated equation, ${\mathrm{I}\mathrm{N}\mathrm{F}}_{it}$ denotes the inflation, ${\mathrm{P}\mathrm{G}}_{it}$ is the Population growth (annual %), ${\mathrm{G}\mathrm{O}\mathrm{V}\mathrm{E}\mathrm{X}\mathrm{P}}_{it}$ is the government expenditure, ${\mathrm{G}\mathrm{N}\mathrm{I}}_{it}$ is GNI per capita, ${\mathrm{D}\mathrm{P}\mathrm{R}}_{it}$ is the age dependency ratio, and ${\mathrm{I}\mathrm{Q}\mathrm{I}}_{it}$ is the institutional quality.

u_it is the fixed effect by country, ε_i,t is the error term, i is the country index, and t is the period (years).

(b)

Model 2

$${POV}_{it}={\mathrm{P}\mathrm{O}\mathrm{V}}_{it-1}+{\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{I}\mathrm{G}}_{it}+{\mathsf{\beta }}_2{\mathrm{I}\mathrm{N}\mathrm{F}}_{it}+{{\mathsf{\beta }}_3\mathrm{B}\mathrm{M}+{\mathsf{\beta }}_4\mathrm{I}\mathrm{Q}{\mathrm{I}}_{it}+{\mathsf{\beta }}_5{\mathrm{G}\mathrm{D}\mathrm{P}\mathrm{G}\mathrm{R}\mathrm{O}\mathrm{W}\mathrm{T}\mathrm{H}}_{it}+{\mathsf{\beta }}_6{\mathrm{P}\mathrm{G}}_{it}+{\mathsf{\beta }}_7{\mathrm{D}\mathrm{F}\mathrm{I}}_{it}+u_{it}+\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(13)

where ${POV}_{it}$ represents poverty, ${POV}_{it-1}$ is the poverty lag, $\mathsf{\beta }$ is the coefficient of the estimated equation, ${IG}_{it}$ represents inclusive growth, ${\mathrm{I}\mathrm{N}\mathrm{F}}_{it}$ denotes the inflation, ${\mathrm{B}\mathrm{M}}_{it}$ is the broad money to GDP, $\mathrm{I}\mathrm{Q}{\mathrm{I}}_{it}$ is the institutional quality index, ${\mathrm{G}\mathrm{D}\mathrm{P}\mathrm{G}\mathrm{R}\mathrm{O}\mathrm{W}\mathrm{T}\mathrm{H}}_{it}$ is Gross Domestic Product growth, ${\mathrm{P}\mathrm{G}}_{it}$ is the population growth, and ${\mathrm{D}\mathrm{F}\mathrm{I}}_{it}$ is digital financial inclusion.

u_it is the fixed effect by country, ε_i,t is the error term, i is the country index, and t is the period (years).

3.4. Diagnostic Tests

This section outlines the diagnostic tests conducted to ensure the robustness and validity of the empirical model. Specifically, it presents the results of the Sargan/Hansen test for overidentifying restrictions, the Arellano–Bond test for autocorrelation, and the Breusch–Pagan test for heteroskedasticity (

Table 2. Heteroskedasticity—Model 1.

Test	Chi-Squared (1)	Prob > Chi-Squared
Breusch–Pagan	0.36	0.5506

Source: Calculated by the author using Stata 18.

and

Table 3. Heteroskedasticity—Model 2.

Test	Chi-Squared (1)	Prob > Chi-Squared
Breusch–Pagan	2.20	0.1379

Source: Calculated by the author using Stata 18.

), each designed to assess different aspects of model integrity.

The p-values of the Breusch–Pagan test for both Model 1 and Model 2 are greater than 5%, indicating no problem of heteroskedasticity in the data. In other words, the null hypothesis could not be rejected as there was insufficient evidence against it.

Multicollinearity—Variance Inflation Factor

The VIF was employed to test for multicollinearity. The results show that the independent variables’ VIFs are less than 10, indicating no multicollinearity (see

Table 4. Variance Inflation Factor Test.

Variable	VIF	1/VIF
GNI	6.14	0.162779
BM	5.16	0.193663
PG	4.46	0.224053
TO	3.7	0.270285
IQI	2.65	0.377376
GOVEXP	2.65	0.377597
POV	1.9	0.526329
INF	1.46	0.682774
GDP Growth	1.44	0.692971
EDU	1.42	0.703686
DFII	1.29	0.77326
Mean VIF	2.94

MS: Broad money to GDP (Money Supply); IG: Inclusive Growth; PG: Population growth (annual %); IQI: Institutional Quality Index; POV: Poverty; EDU: Education; GOVEXP: General government final consumption expenditure (% of GDP); GNI: GNI per capita (constant 2015 US$); INF: Inflation. Source. Calculated by the author using Stata 18.

, which follows).

4. Empirical Results and Discussion

This section presents the empirical findings of the study, beginning with descriptive statistics to characterize the data, followed by diagnostic tests including unit root and multicollinearity assessments, and culminating in an examination of pairwise correlations across the models employed.

Table 5. Unit Root Test Results for Digital Financial Inclusion Dataset.

Variable	Levin–Lin–Chui		Hadri LM
	Level		Level
	Constant	Constant + Trend	Constant	Constant + Trend
DFII	−5.3821 ***	−1.6054 **	15.1448 ***	1.4404 *
	(0.0000)	(0.0542)	(0.0000)	(0.0749)
IG	−2.9160 ***	−3.9809 ***	15.1065 ***	6.0057 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0000)
POV	−1.7285 **	−6.2368 ***	22.1750 ***	1.3221 *
	(0.0420)	(0.0000)	(0.0000)	(0.0931)
DPR	9.2393	−3.1925 ***	22.4054 ***	13.7787 ***
	(1.0000)	(0.0007)	(0.0000)	(0.0000)
IQI	−7.5343 ***	−8.7303 ***	8.6789 ***	3.7162 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0001)
INF	1.8300	−4.0878 ***	5.8073 ***	−0.1282
	(0.9664)	(0.0000)	(0.0000)	(0.5510)
BM	−1.4876 *	−4.9217 ***	16.2548 ***	2.0409 **
	(0.0684)	(0.0000)	(0.0000)	(0.0206)
GDP Growth	−5.8753 *** (0.0000)	−8.6426 *** (0.0000)	−0.0757 (0.5302)	−0.0783 (0.5312)
EDU	−4.2329 ***	−6.8878***	15.9326 ***	5.6781 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0000)
GNI	7.2128	9.8274	3.6377 ***	3.8019 ***
	(1.0000)	(1.0000)	(0.0001)	(0.0001)
PG	1.7091	−3.1519 ***	15.8107 ***	10.9203 ***
	(0.9563)	(0.0008)	(0.0000)	(0.0000)
GOVEXP	−3.8951 ***	−2.0130 **	8.4703 ***	3.1472 ***
	(0.0000)	(0.0221)	(0.0000)	(0.0008)

DFII: Digital Financial Inclusion Index; IG: Inclusive Growth; POV: Poverty; DPR: Age Dependency Ratio; IQI: Institutional Quality Index; INF: Inflation; BM: Broad money to GDP; EDU: Education; GNI: GNI per capita (constant 2015 US$); PG: Population growth (annual %); GOVEXP: General government final consumption expenditure (% of GDP). ***, **, * denote the significance at the 1%, 5%, and 10% levels, respectively. Note: () denotes the p-values. Source: Calculated by the author in Stata 18.

below depicts the outcomes of the Levin–Lin–Chu and Hadri LM unit root tests for investigating the effect of digital financial inclusion on inclusive growth and poverty in EMDEs.

Both tests indicate that digital financial inclusion, inclusive growth, poverty, age dependency ratio, institutional quality, inflation, broad money to GDP, GDP growth, primary enrolment as a percentage of gross enrolment trade (% of GDP), Gross National Income (GNI), population growth, trade (% GDP), and government expenditure are all stationary at level [I (0)], indicating that they are valid, reliable and appropriate for the study, and thus will not generate spurious results.

4.1. System-GMM Results—Model 1

This section demonstrates the regression results on the effect of digital financial inclusion on inclusive growth in EMDEs.

Table 5 illustrates the results of the various panel methods employed by the study. Although System-GMM is the primary estimation technique in this study, a difference-GMM was estimated for comparison and robustness checks. There are seven explanatory variables employed in GMM estimation techniques: digital financial inclusion, inflation, population growth, general government final consumption expenditure (% of GDP), GNI per capita (constant 2015 US$), age dependency ratio, and institutional quality.

Table 6. System-GMM and Difference-GMM Outcomes.

Model
Variables	System-GMM	Difference-GMM
lnIG (−1)	0.7428 *** (0.1398)	0.7754 *** (0.0845)
lnDFII	0.0227 * (0.0128)	−0.00125392
INF	−0.0003 *** (0.0000)	−0.0001 (0.0001)
PG	−0.0972 *** (0.0266)	−0.0698 ** (0.0001)
lnGOVEXP	−00630 (0.0491)	−0.1520 (0.1026)
lnGNI	0.1621 (0.1622)	0.2084 (0.2969)
lnDPR	0.3452 ** (0.1543)	−1.2349 ** (0.5498)
IQI	0.0286 (0.0272)	−0.0285 (0.0322)
Intercept	−0.0000 (0.9428)
	Diagnostics
R-squared
Autocorrelation—Arrelano–Bond test (AR2)	0.96 [0.336]	0.67 [0.502
Sargan’s test of overidentifying restrictions	6.78 [0.452]	0.48 [0.786]
Hansen’s test of overidentifying restrictions	4.83 [0.681]	0.93 [0.628]
Friedman’s test of cross-sectional independence	17.810 [0.5999]	17.810 [0.5999]

IG: Inclusive Growth; DFII: Digital Financial Inclusion Index; INF: Inflation; PG: Population growth (annual %); GOVEXP: General government final consumption expenditure (% of GDP); GNI: GNI per capita (constant 2015 US$); DPR: Age dependency ratio; IQI: Institutional Quality Index. Note: Inclusive Growth is the dependent variable. The number of observations is 189. ***, **, * denote the significance at the 1%, 5%, and 10% levels, respectively. () denote the robust standard errors. [] denote the probability values. Source: Calculated by the author in Stata 18.

further shows that four (digital financial inclusion, inflation, population growth, and age dependency ratio) of the seven explanatory variables were significant. The coefficient values suggest that explanatory variables vary in the degree of impact.

Digital financial inclusion is not significant under the difference-GMM estimation technique. However, other control variables, such as population growth and age dependency ratio, are significant at 10%. Nonetheless, the difference-GMM has weaker instruments relative to System-GMM, making it unsuitable for the present study. This is because it is claimed that the first difference-GMM exhibits finite sample bias when the lags of the dependent variable have a weak correlation with the dependent variable’s first difference in the subsequent period (Blundell & Bond, 1998).

Therefore, the System-GMM estimation technique is the basis for interpreting the findings. This is because the system-GMM estimation technique provides the best outcome relative to the difference-GMM. Additional orthogonality restrictions provided by system-GMM increase the efficiency and reliability of instruments. This study employed the xtabond2 as suggested by Roodman (2009).

The main explanatory variable (digital financial inclusion) is significant under the system-GMM estimation technique. On the main independent variable, findings reveal that digital financial inclusion has a positive and significant effect on inclusive growth in EMDEs at a 10% significance level. The sign and magnitude indicate that when digital financial inclusion increases by a percentage point, inclusive growth will increase by 0.7428 units. Digital financial inclusion is regarded as one of the crucial elements to accomplishing inclusive growth as the process and effort of developing a digital society continues to grow (Chen et al., 2022).

Financially excluded individuals can have the opportunity to participate in the formal financial sector by gaining access to essential financial services and products through available digital financial platforms. The digital finance sector is anticipated to grow, with increased financial access and inclusion, and inclusive growth due to the development and sharing of information and data through the internet (Zhang et al., 2020). A stronger economy includes more people emerging from this increased participation in the financial sector. The findings are consistent with those of Zhang et al. (2020) and Yang et al. (2024), who found that digital financial inclusion has a positive, significant effect on inclusive growth.

Inflation, as a control variable, significantly and negatively affects inclusive growth in EMDEs at the 1% significance level. This indicates an inverse relationship between inflation and inclusive growth. Inclusive growth will decrease by 0.0003 for a percentage inflation increase. Inflation decreases the value of money. The purchasing power of the public is reduced by high, volatile inflation, particularly for individuals in low-paying occupations. Therefore, purchasing necessities may become difficult for individuals, hindering their ability to participate in the economy. Sajid and Ali (2018) argue that rising inflation makes it difficult for developing countries to achieve economic growth. The best monetary policies for inclusive growth are those geared toward maintaining low inflation and steady aggregate demand (Anand et al., 2014). Ahiadorme (2022) found a positive relationship between decreased income disparity, improved poverty alleviation, and increased inclusivity when low inflation and steady economic growth exist in a country. The research findings corroborate those of Ghouse et al. (2022), who found that inflation significantly and negatively affects inclusive growth in low- and middle-income countries.

Population growth was negative and significant at a 1% significance level. The sign and magnitude indicate that a percentage increase in population growth will lead to a 0.0972 decrease in inclusive growth. The global population is expected to grow to 9.7 billion by 2050 and 10.9 billion by 2100 (Gu et al., 2021). Increased demand for limited resources such as water, food, cloth, land, etc., can be under pressure due to a population increase. This may result in fewer individuals having access to necessities. The findings are consistent with Pham et al. (2024), who also showed that population growth has a negative, significant impact on inclusive growth. Tella and Alimi (2016) also found that population growth deteriorates inclusive growth.

The findings further reveal that the age dependency ratio positively and significantly affects inclusive growth at a 5% significance level. The sign and magnitude indicate that when the age dependency ratio increases by 1 percentage point, inclusive growth increases by 0.3452. This finding is surprising, as an increase in the age dependency ratio often results in economic difficulties, suggesting more people are dependent than those who qualify to work or seek employment. Nonetheless, the findings are consistent with those of Tella and Alimi (2016), who found that age dependency positively and significantly affects inclusive growth.

The rest of the variables—general government final consumption expenditure (% of GDP), GNI per capita (constant 2015 US$), and institutional quality—were, however, found to be insignificant in both the system-GMM and difference-GMM estimation techniques.

The insignificance of these variables can be attributed to practical challenges within Emerging and Developing Economies (EMDEs). For government expenditure on inclusive growth (Model 1), the lack of a significant positive link may stem from inefficiency or corruption in resource allocation, preventing funds from effectively translating into equitable economic expansion. For the institutional quality index, its insignificance might indicate that the level of institutional quality in the selected EMDEs has not yet reached a sufficient threshold to exert a powerful, independent influence on inclusive growth or poverty reduction, especially when facing strong counter-effects from significant variables like inflation and population growth.

4.2. System-GMM Results—Model 2

This section presents the regression results of inclusive growth’s effect on poverty in EMDEs.

Table 7. System-GMM and Difference-GMM Outcomes.

Model
Variables	System-GMM	Difference-GMM
lnPOV (−1)	1.0364 *** (0.0148)	1.0256 *** (0.0636)
lnIG	−0.0415 ** (0.0206)	−0.0378 (0.1294)
INF	0.0001 ** (0.0000)	−2.58 × 10−6 (0.0000)
lnBM	0.0240 (0.0233)	0.0576 * (0.0333)
IQI	0.0209 (0.0190)	−0.0625 * (0.0357)
GDPGrowth	−0.0050 ** (0.0018)	0.0059 ** (0.0020)
lnDFII	−0.0240 * (0.0138)	0.0198 (0.0132)
lnGOVEXP	−0.0548 ** (0.0257)	−0.0751 (0.0631)
PG	0.0001 (0.00923)	0.1172 *** (0.0362)
Intercept	−0.7571 ** (0.3104)
	Diagnostics
Autocorrelation—Arrelano–Bond test (AR2)	0.47 [0.639]	0.67 [0.501]
Sargan test of overidentifying restrictions	8.15 [0.700]	4.69 [0.455]
Hansen test of overidentifying restrictions	7.58 [0.751]	4.73 [0.449]
Frees’ test of cross-sectional independence (at 5% significance level)	3.489 [0.3429]	3.489 [0.3429]

POV: Poverty; IG: Inclusive Growth; INF: Inflation; BM: Broad Money to GDP (Money Supply); IQI: Institutional Quality Index; GDP Growth: Gross Domestic Product growth; DFII: Digital Financial Inclusion Index; General government final consumption expenditure (% of GDP); PG: Population Growth. Note: Poverty is the dependent variable. The number of observations is 189. ***, **, * denote the significance at the 1%, 5%, and 10% levels, respectively. () denote the robust standard errors. [] denote the probability values. Source: Calculated by the author in Stata 18.

illustrates the results of the system-GMM and difference-GMM estimation techniques. Although System-GMM was the main estimation technique, the difference-GMM was employed for comparison and robustness checks. GMM estimation techniques and eight explanatory variables were used. These are: inclusive growth, inflation, broad money to GDP (Money Supply); institutional quality index, GDP Growth, digital financial inclusion index, general government final consumption expenditure (% of GDP), and population Growth. Table 6 further shows that five (5) of the eight (8) explanatory variables are significant. The coefficient values show that the explanatory variables vary in the degree of impact.

The System-GMM estimation technique was the basis for interpreting Model 2 results for reasons already stated in the Model 1 results discussion section. The main explanatory variable (inclusive growth) is significant only under the system-GMM estimation technique. The findings reveal that inclusive growth has a negative, significant effect on poverty in EMDEs at a 5% significance level. The sign and magnitude indicate that when inclusive growth increases by a percentage point, poverty will decrease by 0.0415%.

Concerns over rising inequality despite economic growth have culminated in an increasing awareness of the significance of inclusive growth and development in international development (Kuss et al., 2022). This confirms that inclusive growth is critical in poverty alleviation, particularly in EMDEs where poverty is more prevalent. Effective and fair distribution of resources for the greater good of society is essential to advance an inclusive economy (Singh, 2017) and poverty reduction. The findings are consistent with Amponsah et al. (2023), who found that inclusive growth decreases the adverse effects of income disparity on poverty and can help in reducing it.

As a control variable, inflation significantly and positively affects poverty in EMDEs. For a percentage point increase in inflation, poverty will increase by a small magnitude of 0.0001. Inflation is one of the major concerns for policymakers in both advanced economies and EMDEs (Hassan et al., 2016). Basic goods and products become more expensive because of inflation, making it particularly difficult for households in EMDEs, especially for some of the selected countries in our data, such as Zimbabwe, Pakistan, Senegal, Samoa, Rwanda, Mauritius, Guinea, Ghana, Burkina Faso, and Botswana, which in 2022 each had annual inflation rates above 10%. Our results are consistent with those of Omoniyi (2018) and Sehrawat and Giri (2018). Baloch et al. (2020) also found that the purchasing power of individuals is reduced by inflation, thus forcing them to settle for products of lower quality that exacerbate poverty.

Gross domestic product (GDP) growth was found to be significant at a 5% significance level. The sign and magnitude suggest that a percentage point increase in GDP growth will lead to a marginal decrease of 0.0050 units in poverty. The United Nations aims to end extreme poverty—as measured by the proportion of individuals who live on less than USD1.25 a day. However, recent studies suggest that economic growth by itself is unlikely to reach this goal (Anderson et al., 2018). The findings corroborate those of Anyanwu et al. (2013), who found that an increase in real per capita GDP has a considerable negative impact on poverty levels, thus contributing positively to poverty alleviation. Although economic growth is essential, comprehensive policies that promote economic inclusion should be considered and implemented effectively and efficiently to ultimately experience poverty reduction through economic growth.

Digital financial inclusion negatively and significantly affects poverty in EMDEs at a 10% significance level. For a 1 percentage point increase in digital financial inclusion, poverty will decrease by 0.0240. The World Bank considers individuals financially included if they have a bank account or a digital wallet (Ogbeide & Igbinigie, 2019). Most impoverished individuals are unable to access financial products and services, such as loans, through traditional financial institutions because physical branches are at times too far and it is expensive for individuals to reach them. Nonetheless, digital financial services enable impoverished individuals to access essential products and services without visiting a branch. This provides individuals with the opportunity to manage their money more efficiently and effectively. The findings align with Lee et al. (2023), who demonstrated that the growth of digital financial inclusion significantly lowers and stabilizes the overall poverty level.

Government expenditure negatively and significantly affects poverty in the selected EMDEs of our study at a 5% significance level. The sign and magnitude suggest that a percentage point increase in government expenditure will lead to a decrease of 0.0548 units in poverty. Global economic growth and the fight against poverty are significantly facilitated by government expenditure (Chude et al., 2019). However, for government expenditure to be effective, it needs to be targeted at the right initiatives that improve and empower society. Nevertheless, in some cases, resources might be squandered and thus fail to reach the intended target market due to corruption. These findings align with those reported by Mehmood and Sadiq (2010). However, the remaining variables—broad money to GDP, institutional quality, and population growth—were found to be insignificant under the system-GMM estimation technique.

Table 8. Leave-One-Out (LOO) Sensitivity Test.

Variable	(1) Baseline (N = 21)	(2) Excl. Ghana	(3) Excl. Rwanda	(4) Excl. Zimbabwe
L.lnIG	0.7428 ***	0.7395 ***	0.7451 ***	0.7412 ***
	−0.1398	−0.141	−0.1385	−0.1402
lnDFII	0.0227 *	0.0215 *	0.0234 *	0.0221 *
	−0.0128	−0.0131	−0.0125	−0.0129
Control Variables	Included	Included	Included	Included
Time Dummies	Yes	Yes	Yes	Yes
Hansen p-value	0.681	0.675	0.692	0.679
AR(2) p-value	0.336	0.331	0.34	0.334

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

below presents the stability of the DFII coefficient when excluding high-leverage countries identified by GNI and digital index volume.

The System-GMM models for both Inclusive Growth (Model 1) and Poverty (Model 2) were systematically re-estimated by excluding one country at a time from the 21 EMDEs. As shown in Table 8, the coefficient for DFII remained positive and statistically significant (p < 0.10) in every iteration, confirming that no single country exerts undue influence on the results.

The COVID-19 Interaction and Time-Period Sensitivity test as shown in

Table 9. COVID-19 Interaction and Time-Period Sensitivity.

Variable	Coefficient	Robust Std. Error	p-Value
L.lnIG	0.7381 ***	0.1412	0
lnDFII	0.0219 *	0.0131	0.094
COVID (2020–2022)	0.0045	0.0062	0.468
lnDFII\times COVID	0.0082	0.0105	0.435
Intercept	0	0.9428	0.999
Controls	Included
AR(2) p-value	0.342
Hansen test p-value	0.672

Note: The primary effect of DFII remains statistically significant at the 10% level, while the COVID interaction term is non-significant, indicating that the impact is not pandemic-dependent. ***, **, * denote the significance at the 1%, 5%, and 10% levels respectively.

assesses whether the digital acceleration during the pandemic (2020–2022) disproportionately drives the inclusive growth effect.

A dummy variable was introduced for the post-2019 period (2020–2022) and an interaction term (DFII\times COVID). The baseline effect of DFII remained significant, while the interaction term was not statistically significant. This confirms that the observed relationship is a generalizable structural trend in EMDEs and not merely a “COVID-driven” spike.

5. Concluding Remarks

This study sought to investigate how digital financial inclusion affects poverty through inclusive growth EMDEs. This was achieved through the application of panel estimation techniques.

This chapter explores the theoretical foundations of the effect of digital financial inclusion on inclusive growth and poverty in EMDEs through the perspective of the theory of financial intermediation. Financial intermediation is also based on the transaction costs approach established by Benston and Smith (1976) and Fama (1980). High transaction costs deter people from taking part in the formal financial sector, which eventually leads to greater financial exclusion, high poverty levels, and low inclusive growth. Conversely, lower transaction costs typically lead to greater inclusive growth, poverty reduction, and increased digital financial inclusion.

The literature review demonstrates that in most regions with EMDEs, digital financial inclusion plays a role in increasing inclusive growth. Moreover, existing literature also shows that poverty can be reduced through inclusive growth. Additionally, financial inclusion is a solid foundation for the financial infrastructure of any country (Liaqat et al., 2022). Most studies in our empirical literature demonstrated that digital finance does reduce poverty and is closely related to the advancement and expansion of inclusive economic growth. Furthermore, studies have shown that financial inclusion is a powerful tool for reducing poverty, particularly in EMDEs.

This study constructed a new multidimensional digital financial inclusion index for 21 EMDEs using weights derived from the PCA in aggregating indicators for two dimensions, that is, access and usage, following Cámara and Tuesta (2014) and Khera et al. (2021). The index is a useful tool for evaluating the degree of digital financial inclusion in various EMDEs and tracking their development over time. In our analysis, using data spanning from 2013 to 2022, we found that none of the countries we selected demonstrated a high level of digital financial inclusion in the time frame we examined, applying the ranking systems developed by Cámara and Tuesta (2014). Zimbabwe and Rwanda demonstrated a moderate level of digital financial inclusion, whereas Ghana showed a medium level of digital financial inclusion in 2020 and 2021. Moreover, the EMDEs selected for the study all fall under the low-level category since their average level of digital financial inclusion from 2013 to 2021 is less than 30%. Considering this, policymakers and governments should take it upon themselves to ensure that correct policies and programs are made available to meet the needs of the underserved and previously disadvantaged populations in EMDEs.

The main objective of this study was to employ the system-GMM estimation technique to examine how digital financial inclusion affects poverty through inclusive growth EMDEs. This was achieved by first estimating the effect of digital financial inclusion on inclusive growth (scenario 1) and then estimating the effect of inclusive growth on poverty in EMDEs (scenario 2). Under the first scenario, the study found that digital financial inclusion (the main explanatory variable) was significant under the system-GMM estimation technique. The results show that on the main independent variable (that is, inclusive growth), digital financial inclusion has a positive and significant effect on inclusive growth in EMDEs at a 10% significance level. The sign and magnitude indicate that when digital financial inclusion increases by a percentage point, inclusive growth will increase by a magnitude of 0.7428 units.

Some control variables also showed that they were significant to the dependent variable, digital financial inclusion. These control variables included inflation, population growth, and the age dependency ratio. Nonetheless, general government final consumption expenditure (% of GDP), GNI per capita (constant 2015 US$), and institutional quality were found to be insignificant to digital financial inclusion. Policymakers and governments should consider prioritizing capital or investments in industries with the greatest prospects of increasing inclusive growth through digital financial inclusion. This can be made possible through increasing access to digital financial services and products, particularly in areas where financial exclusion is severe.

Under the second scenario, the study found that inclusive growth (the main explanatory variable) is significant under the system-GMM estimation technique. The findings demonstrate that inclusive growth has a negative, significant effect on poverty in EMDEs at the 5% significance level. The sign and magnitude indicate that for each percentage-point increase in inclusive growth, poverty decreases by 0.0415 units. Some control variables also demonstrated that they were significant to the dependent variable, poverty. These control variables included inflation, Gross domestic product (GDP) growth, digital financial inclusion, and general government final consumption expenditure (% of GDP). The rest of the variables, that is, broad money to GDP, institutional quality, and population growth, were, however, found to be insignificant under the system-GMM estimation technique. By developing channels and policies that reduce income inequality and increase job opportunities for all, policymakers and governments can alleviate poverty through inclusive growth in EMDEs. When job opportunities increase, workforce participation is also likely to rise, particularly among poor individuals who are often unemployed.

5.1. Policy Implications and Interventions

Policy Recommendations and Interventions Based on the empirical evidence that digital financial inclusion (DFI) positively impacts inclusive growth and alleviates poverty, policymakers in EMDEs should prioritize the following targeted interventions. First, regarding digital infrastructure, governments must incentivize the expansion of mobile money agent networks in remote and rural regions. Our findings suggest that increasing the density of these physical touchpoints is critical for transitioning underserved populations into the formal financial sector. Furthermore, investment in high-speed internet and mobile network coverage is essential to lower the high transaction costs that currently discourage participation in formal finance.

Second, a robust regulatory framework is required to manage information asymmetry and ensure financial stability. Policymakers should implement regulatory sandboxes that allow Fintech companies to test innovative digital products in a controlled environment, fostering innovation while protecting vulnerable consumers. Additionally, given that inflation and population growth were found to hinder inclusive growth, monetary authorities must maintain stable price levels to preserve the purchasing power of the poor. Finally, government expenditure should be specifically targeted toward digital literacy programs to ensure that the usage dimension of our DFI index, such as the frequency and value of internet transactions, can be fully leveraged by all citizens to drive equitable economic expansion.

5.2. Limitations and Future Research Directions

The selection of the 21 EMDEs and the 2013–2022 time period was primarily driven by the availability and reliability of secondary data. While digital financial services are expanding globally, formal reporting on digital financial inclusion remains severely limited in many developing countries. This study utilized the Financial Access Survey (FAS-IMF) and World Development Indicators (WDI), which are the most reputable sources for the metrics used in the study; however, many countries were excluded due to missing data for critical indicators like mobile money agent outlets and digital transaction values.

Future research should aim to expand the scope of this study by incorporating a broader longitudinal dataset as more recent post-pandemic data becomes available.