Digital Financial Inclusion and Gender Inequality: Structural Drivers and Financial Participation Gaps in Nigeria

Abstract

Digital financial inclusion (DFI) plays an increasingly important role in expanding financial participation, improving access to payment systems, and supporting inclusive economic development. Yet substantial gender gaps in DFI remain a persistent challenge in many developing economies, including Nigeria. This study examines whether these gaps mainly reflect structural inequalities, particularly unequal access to digital connectivity and economic opportunity, rather than clear evidence of gender-specific behavioral responses between men and women. Using nationally representative microdata from the 2025 Global Findex survey, we analyze three sequential dimensions of DFI: digital account ownership, the use of any digital payment, and merchant digital payments. The empirical analysis combines logit and probit models with nonlinear Fairlie decompositions and interaction analysis focusing on education, income, employment, and internet access. Descriptive results show clear raw gender gaps at the earlier stages of digital financial participation. However, once socioeconomic characteristics and digital access are taken into account, the gender coefficient becomes statistically insignificant across the main specifications. The decomposition results indicate that between 79% and 94% of the observed gaps are statistically attributable to differences in endowments, with internet connectivity and labor market participation emerging as the most important contributors. The interaction results further show that education and connectivity are positively associated with digital financial participation for both men and women, with no clear evidence that these associations differ systematically by gender. Overall, the findings suggest that gender gaps in DFI in Nigeria are mainly associated with structural inequalities rather than clear evidence of gender-specific behavioral responses. The results carry important implications for financial inclusion policy, suggesting that efforts to narrow these gaps should consider not only wider digital access but also stronger support for women’s economic participation within a broader and more inclusive digital financial transition.

1. Introduction

Over the past decade, digital financial services have expanded rapidly across sub-Saharan Africa. Mobile money platforms, fintech applications, and digital payment systems have reduced transaction costs and expanded access to formal finance, particularly in environments where physical banking infrastructure is limited (Demirguc-Kunt et al., 2022; Suri & Jack, 2016). Nigeria, Africa’s most populous country and one of its largest economies, has witnessed substantial growth in digital finance, driven by rising mobile penetration, regulatory reforms, and an active fintech sector.

The scale of this transformation is striking. Electronic payment transactions in Nigeria reportedly reached approximately 1.07 quadrillion Naira in 2024, the highest level ever recorded in the country’s financial system (NIBSS, 2025). Overall financial inclusion rose to about 74% of adults in 2023, while roughly 45% reported using digital financial services, reflecting rapid expansion in mobile and platform-based transactions. Internet penetration and mobile adoption have also increased steadily, reinforcing the infrastructural foundations of digital finance (EFInA, 2024a, 2024b). Together, these figures position Nigeria as one of the most dynamic digital financial markets in sub-Saharan Africa.

Yet persistent gender gaps remain a central concern in African financial development. Across developing economies, women are consistently less likely than men to own financial accounts, use digital payments, or access formal financial services (Demirguc-Kunt et al., 2018). Existing explanations emphasize differences in income, education, employment, and access to digital infrastructure. However, the literature has not yet reached a clear consensus on whether observed gender gaps in digital financial inclusion (DFI) primarily reflect unequal access to resources (structural factors) or persistent differences in behavioral responses after these factors are taken into account. This unresolved distinction represents a key gap in the empirical literature, particularly in the African context.

In Nigeria, digital payments have expanded very quickly over the past few years. However, this expansion has not removed deeper socioeconomic inequalities. Even with the growth of formal financial services, women still face disadvantages in employment, income, and internet access (EFInA, 2024b; World Bank, 2023). The labor market remains largely informal, and digital connectivity is not distributed evenly between men and women, or between urban and rural areas. So, while transaction volumes are reaching record levels, underlying inequalities continue to shape who actually participates. This raises the question of whether digital expansion is accompanied by broader inclusion, or whether it mainly reflects pre-existing structural differences. Similar evidence shows that DFI gaps often remain tied to socioeconomic and digital-access disparities (Sharaf et al., 2026b).

This paper addresses this gap by explicitly distinguishing between structural and residual components of gender disparities in DFI. Rather than focusing only on whether gender gaps exist, the analysis examines whether these gaps persist after conditioning on observable socioeconomic and digital characteristics. In doing so, the paper shifts the focus from descriptive differences to the underlying mechanisms associated with unequal participation.

This distinction matters for policy. If gender disparities are largely structural, interventions should focus on expanding women’s access to education, labor markets, and digital connectivity. If gaps remain after accounting for these characteristics, deeper institutional or behavioral constraints may be at play.

Nigeria provides a useful setting to examine this issue. Women face documented disadvantages in employment, earnings, and internet access (World Bank, 2023) yet digital financial services are widely available. From a social indicators perspective, high transaction volumes and broad formal inclusion do not necessarily imply equivalent depth of participation. Ownership of a digital account and actual engagement in merchant payments capture different levels of economic agency (Demirguc-Kunt et al., 2018). The key empirical question therefore shifts from the expansion of access to the structure of usage: who participates fully in digital finance, and under what conditions?

Using nationally representative microdata from the 2025 Global Findex survey, this paper conceptualizes DFI as a multistage process—digital account ownership, use of any digital payment, and merchant digital payments as an advanced form of engagement. The empirical strategy combines nonlinear probability models with a Fairlie decomposition to distinguish explained from unexplained components of the gender gap. Interaction models further assess whether human capital and digital infrastructure operate differently across genders.

The findings show that while raw gender gaps are sizeable at early stages of digital participation, these differences become statistically insignificant once socioeconomic characteristics and connectivity are taken into account. The decomposition results further indicate that the majority of the observed gap—between 79% and 94%—is associated with differences in observable characteristics, particularly internet access and labor market participation. These results should be interpreted as conditional associations rather than causal effects, given the cross-sectional nature of the data. In particular, variables such as employment, income, and digital usage may be jointly determined, which limits causal interpretation.

This paper contributes to the literature in several ways. First, it provides a clear empirical distinction between structural and residual components of gender disparities in DFI. Second, it conceptualizes DFI as a sequential process rather than a single outcome, allowing for a more nuanced analysis of participation stages. Third, it provides new micro-level evidence from Nigeria, a key African economy that remains underrepresented in detailed gender-based analyses of digital finance. Finally, the findings are interpreted in light of emerging cross-country evidence, which similarly emphasizes the role of resource-based constraints in shaping gender gaps in financial inclusion.

The remainder of the paper is organized as follows. Section 2 reviews the literature. Section 3 describes the data and empirical framework. Section 4 presents the results, which are discussed in Section 5, and Section 6 concludes.

2. Literature Review

Recent studies show that expansion in digital infrastructure and higher levels of financial account ownership do not necessarily translate into effective financial inclusion. A central distinction in this literature is between access to financial services and their actual use. While owning an account or a digital tool is often treated as a key indicator of inclusion, many studies show that it is only a necessary condition, not a sufficient one. Digital financial exclusion appears more clearly at the stage of actual use, that is, digital financial non-use (Gammage et al., 2017; Kowsick & Ramasamy, 2024; Roy & Patro, 2022).

Studies based on Global Findex data show that a noticeable share of individuals own formal financial accounts or digital wallets but do not use them regularly or effectively. This pattern reflects what is widely described as the “use gap,” which cannot be captured by traditional access indicators. Digital payments promote inclusion only when supported by infrastructure, trust, and regulation (Shahen & Sharaf, 2026). As a result, policy efforts that focus primarily on account ownership or expansion of digital channels may improve formal inclusion metrics without necessarily leading to meaningful integration into everyday financial behaviour (Galindo-Manrique & Rojas-Vargas, 2025; Grohmann et al., 2018).

In this context, the literature clearly distinguishes between formal access and functional use. DFI should therefore be understood not only in terms of ownership of accounts or cards, but also in terms of the extent to which these tools are actively used in financial activities such as payments, savings, and regular transactions. Evidence shows that improvements in access indicators often precede—and sometimes are not followed by—similar improvements in usage indicators, revealing a structural disconnection between availability and use (Andrianaivo & Kpodar, 2012; Grohmann et al., 2018).

This distinction becomes even more important when DFI is examined from a gender perspective. Several studies show that the gender gap in digital financial usage is wider and more persistent than the gender gap in access, even in environments that have experienced significant expansion in digital financial services. Women, even when they own accounts or digital financial tools, tend to use them less frequently and less intensively, reflecting a pattern of gendered non-use (Gammage et al., 2017; Özşuca, 2025; Roy & Patro, 2022).

Econometric analyses that control for individual characteristics—such as education, income, age, and digital connectivity—often find that the usage gap remains even after these factors are taken into account. This has led to the argument that digital financial exclusion cannot be explained solely by differences in observable characteristics, but also reflects deeper constraints related to economic empowerment, control over resources, and institutional context (Grohmann et al., 2018; Kulkarni & Ghosh, 2021).

Building on this, recent contributions introduce the concept of “second-stage digital financial exclusion,” which refers to disparities that persist even after access barriers are removed. In this framework, the expansion of digital channels does not automatically lead to equal participation, as structural and institutional constraints may continue to shape outcomes (Gallego-Losada et al., 2023; Gammage et al., 2017). Empirical evidence from several contexts supports this view, showing that gender remains statistically significant even after controlling for socioeconomic variables, suggesting differences in returns to resources or underlying institutional constraints (Özşuca, 2025; Tripathi & Rajeev, 2023).

Evidence from African countries points in a similar direction. Although digital financial services, including mobile payments, have expanded rapidly, usage has not become equal, and socio-economic as well as institutional factors continue to play an important role (Faton et al., 2025; Khchoum & Bardi, 2025). Decomposition studies further indicate that part of the observed gap cannot be fully attributed to measurable characteristics, suggesting the presence of an unexplained component linked to differences in returns or unobserved factors (Morrar et al., 2024; Özşuca, 2019).

At the same time, another strand of the literature emphasizes the role of structural factors in shaping digital financial participation. Differences in employment, income, and access to digital infrastructure—particularly internet connectivity—are consistently identified as key determinants of both access and usage (Demirguc-Kunt et al., 2022; Grohmann et al., 2018; Khera et al., 2022). Employment provides regular income flows and increases the need for financial transactions, internet access lowers transaction costs and enables participation in digital platforms, and income shapes the capacity to engage in financial activities. From this perspective, observed gender disparities may largely reflect unequal distribution of these resources rather than gender-specific preferences or behavioral responses.

This creates a clear tension in the literature. On one hand, several studies highlight persistent gender gaps even after controlling for observable characteristics, pointing to behavioral or institutional constraints. On the other hand, a growing body of evidence suggests that these gaps may diminish substantially once differences in economic and digital endowments are taken into account. The relative importance of structural versus residual factors therefore remains an open empirical question.

Recent cross-country evidence provides useful comparative context for this debate. Yoon et al. (2025) construct an Abnormal Financial Inclusion Index (AbFII) for 100 countries using Global Findex data from 2014 to 2021. Their index captures the level of financial inclusion that exceeds what would be expected from a country’s GDP per capita. They find that countries with higher AbFII are associated with narrower gender and income-group gaps in financial inclusion, particularly in account ownership and digital payments. This evidence is consistent with broader multi-country studies showing that income, labor market participation, and access to digital infrastructure are important in shaping disparities in account ownership and digital payment use (Demirguc-Kunt et al., 2022; Khera et al., 2022). The Nigerian evidence in the present study complements this cross-country literature by showing, at the micro level, that observable structural factors, especially internet access and labor market participation, account for most of the gender gap in digital financial participation.

The literature on digital finance and credit adds another important dimension by focusing on the role of market design and regulation. Payment infrastructure can itself become a source of exclusion when global networks fragment (Sharaf et al., 2026d). Studies show that digitalization may reproduce exclusion through algorithmic decision-making, standardized lending criteria, and institutional biases that disadvantage individuals with limited formal financial histories (Johnen & Mußhoff, 2023; Oluoch & Alhassan, 2026). Similarly, factors such as trust, consumer protection, and product design influence whether individuals translate access into sustained usage, highlighting the importance of institutional context beyond technological availability (Setiawan et al., 2024; Tan et al., 2024).

Experimental and causal studies provide further insight into the relationship between digital usage and empowerment. Evidence shows that digital financial services do not automatically lead to improved outcomes unless they are accompanied by effective control over financial decisions. In some cases, digital tools enhance financial autonomy, while in others they result in only symbolic inclusion without meaningful economic empowerment (Heath & Riley, 2024; Mabrouk et al., 2023).

Overall, the literature suggests that DFI is a multi-stage process that moves from access to active and meaningful use, and that gender disparities in this process may reflect a combination of structural inequalities and residual institutional or behavioral factors. However, there is no clear consensus on the relative importance of these components, and empirical findings vary across contexts.

One possible explanation for this variation is that the relative importance of structural versus residual factors depends on the underlying distribution of resources within each economy. In contexts where inequalities in employment, income, and digital access are large, observed gender gaps may be primarily driven by differences in endowments. In contrast, in settings where access is more equalized, remaining disparities are more likely to reflect institutional or behavioral constraints. This distinction helps explain why the Nigerian case may differ from studies that find more persistent unexplained gender gaps: where basic structural inequalities remain large, the observed gap may be explained more by unequal endowments than by differential returns to the same resources.

Against this background, the present study examines the Nigerian case, where digital financial services have expanded rapidly but underlying inequalities in employment, income, and connectivity remain significant. By focusing on actual usage and applying decomposition techniques, the analysis aims to clarify whether gender gaps in this context are primarily structural or whether a residual gap persists after conditioning on observable characteristics.

3. Data and Methods

3.1. Data Source and Sample Construction

This study uses microdata from the 2025 wave of the Global Findex survey for Nigeria. The Global Findex is a nationally representative household survey administered by the World Bank that collects detailed information on financial account ownership, digital payments, savings, borrowing, and financial access. The survey employs a stratified random sampling design and provides sampling weights to ensure national representativeness. The Nigerian sample includes 1000 adult respondents aged 15 and above. All estimations apply the survey sampling weights. Observations with missing values on key covariates are excluded, resulting in a final estimation sample of 998 individuals for the pooled specifications.

Although the Global Findex survey is nationally representative within its design, the final estimation sample of 998 respondents limits the precision of subgroup and interaction estimates. This is especially relevant for the gender-specific models and interaction specifications, where the effective number of observations is smaller than in the pooled models. The Nigeria-only focus also means that the findings should not be generalized mechanically to other African economies. Instead, the results should be interpreted as evidence from one important national context, with implications that may vary depending on country-specific labor market conditions, digital infrastructure, and institutional settings.

The cross-sectional nature of the data also limits causal interpretation. The empirical analysis identifies conditional associations between gender, socioeconomic characteristics, digital connectivity, and digital financial participation. It does not establish one-way causal effects. In particular, employment, income, internet access, and digital finance usage may be jointly determined. For example, employment may increase the need for digital transactions, but digital financial access may also support employment or income-generating activities. Similarly, internet access may enable digital payment use, but individuals who are more financially active may also be more likely to invest in connectivity. For this reason, the estimates are interpreted as associations rather than causal effects.

Despite these limitations, the Global Findex is well suited to the purpose of this study because it provides nationally comparable microdata on financial access, digital payments, and socioeconomic characteristics. The empirical strategy is designed not to claim causality, but to examine whether observed gender gaps persist after conditioning on observable characteristics and to decompose the extent to which these gaps are statistically attributable to differences in endowments.

3.2. Outcome Variables

DFI is conceptualized as a multi-stage process. Three binary outcomes are constructed. The first captures digital account ownership, where $Y_{1i}=1$ if individual $i$ reports owning a digital financial account and zero otherwise. The second measures use of any digital payment, defined as $Y_{2i}=1$ if the respondent reports using any form of digital payment in the past year. The third captures merchant digital payments, where $Y_{3i}=1$ if the individual reports making digital payments to merchants. These three measures represent progressively more intensive stages of digital financial engagement.

This multi-stage structure is important because account ownership, general digital payment use, and merchant payments reflect different levels of participation. Account ownership captures access, any digital payment captures basic use, and merchant payments capture a more active form of participation in the digital economy. This allows the analysis to move beyond a single measure of financial inclusion and examine whether gender gaps differ across stages of digital engagement.

3.3. Baseline Empirical Specification

To assess whether a gender gap persists after conditioning on observable characteristics, the following logit specification is estimated for each outcome $\mathrm{k}\in \{1,2,3\}$

$$\mathrm{Pr}\left(Y_{ki}=1\right|X_i)=\mathsf{\Lambda }({\mathsf{\alpha }}_k+{\mathsf{\beta }}_k{Female}_i+{\mathrm{X}}_i^{\prime }{\mathsf{\gamma }}_k)$$

(1)

where $\mathsf{\Lambda }(\cdot )$ denotes the logistic cumulative distribution function, ${Female}_i$ is a binary indicator equal to one for women, and $X_i$ is a vector of socioeconomic and connectivity characteristics. The parameter ${\mathsf{\beta }}_k$ represents the conditional gender difference, while ${\mathsf{\gamma }}_k$ represents the vector of slope coefficients associated with the controls. The control vector includes age, education categories (with primary education or less as the reference group), income quintiles (with the lowest quintile as the base category), employment status, urban residence, mobile phone ownership where applicable, and internet access. Robust standard errors are used in all estimations. As a robustness exercise, all specifications are re-estimated using probit models.

The baseline specification is used to examine whether the raw gender gap remains after accounting for observable socioeconomic and digital access characteristics. The coefficient on the female indicator should therefore be interpreted as a conditional association, not as evidence of causal discrimination or a direct behavioral effect. A statistically insignificant female coefficient does not mean that gender inequality is absent; rather, it indicates that the observed difference is largely accounted for by differences in measured characteristics included in the model.

The use of logit models is appropriate because all three outcomes are binary. Probit models are used as a robustness check to assess whether the findings are sensitive to the choice of nonlinear probability model. The analysis reports average marginal effects where appropriate, since these are easier to interpret than raw logit coefficients and express changes in predicted probabilities.

3.4. Gender-Specific Models

To explore whether the determinants of digital financial participation differ across genders, separate logit models are estimated for men and women. For each gender $g\in \{Male,Female\},$ the following model is estimated:

$$\mathrm{Pr}\left(Y_{ki}=1\right|X_i,G_i=g)=\mathsf{\Lambda }({\mathsf{\alpha }}_{kg}+{\mathrm{X}}_i^{\prime }{\mathsf{\gamma }}_{kg})$$

(2)

Allowing coefficients to vary fully by gender makes it possible to assess whether education, income, employment, and digital connectivity operate differently for men and women.

These gender-specific models are used as exploratory heterogeneity checks. Because the sample is split by gender, the estimates are less precise than the pooled models and should be interpreted cautiously. Their purpose is not to provide definitive evidence of separate causal pathways, but to examine whether the observed associations appear similar or different across men and women.

3.5. Human Capital Mechanism: Gender × Education

To examine whether the association between education and digital financial participation differs by gender, interaction terms between gender and education categories are introduced. The specification takes the following form:

$$\mathrm{Pr}\left(Y_{ki}=1\right|X_i)=\mathsf{\Lambda }({\mathsf{\alpha }}_k+{\beta }_k{Female}_i+{\delta }_k{Educ}_i+{\theta }_k{(Female}_i\times {Educ}_i)+{\mathrm{Z}}_i^{\prime }{\mathsf{\gamma }}_k)$$

(3)

where ${Educ}_i$ denotes education category indicators and ${(Female}_i\times {Educ}_i$) captures differential associations between education and digital financial participation by gender. The vector $Z_i$ contains the remaining control variables. The interaction coefficients ${\theta }_k$ test whether the association between human capital and digital financial participation differs systematically between men and women.

These interaction models are included to examine mechanisms rather than to claim causality. Education may be associated with digital financial participation by improving literacy, confidence, and ability to navigate digital platforms. However, education may also be correlated with income, employment, location, and household background. Therefore, the interaction terms are interpreted as evidence on whether the education-participation association differs by gender, not as causal estimates of the effect of education.

3.6. Digital Infrastructure Mechanism: Gender × Internet

To assess whether digital infrastructure moderates gender differences, a second interaction specification is estimated:

$$\mathrm{Pr}\left(Y_{ki}=1\right|X_i)=\mathsf{\Lambda }({\mathsf{\alpha }}_k+{\beta }_k{Female}_i+{\varphi }_k{Internet}_i+{\psi }_k{(Female}_i\times {Internet}_i)+{\mathrm{W}}_i^{\prime }{\mathsf{\gamma }}_k)$$

(4)

where ${Internet}_i$ indicates access to the internet and ${(Female}_i\times {Internet}_i)$ captures gender-specific associations between connectivity and digital financial participation. The vector $W_i$ includes age, education, income quintiles, employment status, urban residence, and phone ownership where applicable. The parameter ${\psi }_k$ evaluates whether internet access alters the conditional gender gap in digital participation.

Internet access is treated as a key digital-infrastructure mechanism because digital financial services often require connectivity, platform access, and the ability to complete transactions remotely. At the same time, internet access may be endogenous to digital finance use, since financially active individuals may be more likely to obtain or maintain connectivity. For this reason, the estimates are interpreted as conditional associations between connectivity and digital participation, not as causal effects of internet access.

3.7. Fairlie Decomposition

To identify the sources of observed gender gaps, the nonlinear Fairlie (2005) decomposition method is applied. let ${\overline{Y}}_m$ and ${\overline{Y}}_f$ denote the average predicted probabilities for men and women. The total gender gap is defined as:

$$\Delta ={\overline{Y}}_m-{\overline{Y}}_f$$

(5)

Following Fairlie (2005), the gap can be decomposed as:

$$\Delta =\left[{\displaystyle {\displaystyle {\sum }_{i=1}^{N_M}}}{\displaystyle \frac{F\left(X_{Mi}\widehat{\beta }\right)}{N_M}}-{\displaystyle {\displaystyle {\sum }_{i=1}^{N_F}}}{\displaystyle \frac{F\left(X_{Fi}\widehat{\beta }\right)}{N_F}}\right]$$

(6)

which can be separated into an explained component reflecting differences in observable characteristics and an unexplained component reflecting differences in coefficients or unobserved factors:

$$\Delta =\underset{explained}{\underbrace{\left[\sum F\left(X_{Mi}\widehat{\beta }\right)-\sum F\left(X_{Fi}^{*}\widehat{\beta }\right)\right]}}+\underset{unexplained}{\underbrace{\left[\sum F\left(X_{Fi}^{*}\widehat{\beta }\right)-\sum F\left(X_{Fi}\widehat{\beta }\right)\right]}}$$

(7)

This approach quantifies the contribution of education, income, employment, and digital connectivity to the overall gender gap in DFI.

The Fairlie decomposition is appropriate because the outcomes are binary and the study seeks to distinguish between explained and unexplained components of the gender gap. The explained component captures the part of the gap statistically attributable to differences in observed characteristics, such as education, employment, income, mobile phone ownership, and internet access. The unexplained component captures differences that remain after these characteristics are taken into account, which may reflect unobserved factors, differences in coefficients, institutional constraints, measurement limits, or other residual influences. It should not be interpreted as direct evidence of discrimination or behavioral difference.

This decomposition is especially useful for the research question because the paper is not only interested in whether a gender gap exists, but also in whether the gap is mainly associated with unequal endowments or with residual differences after observable characteristics are considered.

4. Results

4.1. Descriptive Evidence

Table 1. Descriptive statistics by gender.

Variable	Men	Women
Age	33.67	31.95
Primary education (educ1)	0.366	0.511
Secondary education (educ2)	0.616	0.476
Tertiary education (educ3)	0.018	0.013
Income quintile 1 (lowest)	0.134	0.264
Income quintile 2	0.182	0.220
Income quintile 3	0.209	0.189
Income quintile 4	0.228	0.174
Income quintile 5 (highest)	0.247	0.153
Employed	0.798	0.650
Urban residence	0.482	0.463
Owns mobile phone	0.858	0.817
Internet access	0.447	0.307
Digital account ownership	0.619	0.444
Any digital payment use	0.631	0.459
Merchant digital payments	0.338	0.230

Note: All figures are weighted means using Global Findex sampling weights.

presents weighted descriptive statistics for digital financial outcomes by gender in Nigeria.

The descriptive evidence reveals substantial raw gender gaps at the earlier stages of digital participation. Digital account ownership is reported by 61.9 percent of men compared to 44.4 percent of women, a gap of approximately 17 percentage points. A similar disparity is observed for the use of any digital payment (63.1 percent for men versus 45.9 percent for women). Participation declines for both groups at the merchant payment stage, though a gap remains (33.8 percent for men and 23 percent for women).

Gender differences are also evident in key socioeconomic characteristics. Women are more concentrated in the lowest income quintiles and exhibit lower employment rates and lower internet access. These differences are economically meaningful and suggest that the observed digital participation gap may reflect unequal access to resources rather than gender-specific preferences or behaviors. The multivariate analysis below evaluates this distinction formally.

4.2. Baseline Logit Estimates

Table 2. Determinants of digital financial inclusion in Nigeria. Average marginal effects from logit models.

Variables	Digital Account	Any Digital Payment	Merchant Digital Payment
Female	−0.016	−0.022	−0.003
	(0.023)	(0.024)	(0.029)
Age	0.002 **	0.003 ***	0.003 ***
	(0.001)	(0.001)	(0.001)
Secondary education	0.126 ***	0.070 **	0.094 **
	(0.036)	(0.035)	(0.046)
Tertiary education	0.193 ***	0.202 ***	0.193 **
	(0.072)	(0.060)	(0.091)
Income quintile 2	0.035	0.016	0.023
	(0.045)	(0.046)	(0.052)
Income quintile 3	0.062	0.082 *	−0.009
	(0.045)	(0.045)	(0.054)
Income quintile 4	0.181 ***	0.178 ***	0.034
	(0.044)	(0.043)	(0.052)
Income quintile 5	0.136 ***	0.134 ***	0.085 *
	(0.042)	(0.042)	(0.050)
Employed	0.137 ***	0.154 ***	0.144 ***
	(0.026)	(0.027)	(0.037)
Urban	0.032	0.018	0.041
	(0.024)	(0.024)	(0.029)
Owns mobile phone	0.218 ***	0.231 ***	0.283 ***
	(0.049)	(0.050)	(0.101)
Internet access	0.252 ***	0.256 ***	0.296 ***
	(0.020)	(0.021)	(0.026)
Observations	998	998	998

Note: *, **, *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. AMEs represent percentage-point changes in predicted probabilities. Robust standard errors in parentheses.

reports the AMEs from logit models for the three DFI outcomes.

Across all three models, the coefficient on the female indicator is negative but statistically insignificant. This finding does not imply equality in outcomes; rather, it indicates that once differences in education, income, employment, and connectivity are accounted for, there is no statistically significant residual gender difference after conditioning on these observed characteristics. The raw gender gap observed in Table 1 therefore appears to be largely associated with observable structural differences.

In contrast, several covariates display strong and consistent associations. Education exhibits a positive gradient. Individuals with secondary education are significantly more likely to own digital accounts and use digital payments relative to those with primary education or less. The magnitude of this association increases for higher education levels, particularly for merchant digital payments.

Employment status is also strongly related to all three outcomes. It is associated with a 14–15 percentage-point higher probability of digital engagement across stages, with especially large effects for advanced usage. This highlights the importance of labor market attachment in facilitating financial integration.

Connectivity variables are among the strongest correlates in the models. Mobile phone ownership and internet access have large and highly significant associations at all stages, with internet access associated with an increase in the predicted probability of merchant digital payments of nearly 30 percentage points. In magnitude, connectivity effects exceed those of education and income, underscoring the structural importance of digital infrastructure.

Income effects are nonlinear, with higher-income individuals more likely to participate at early stages, while income gradients weaken for advanced usage. Model diagnostics are reported at the end of the Results section in Section 4.9.

4.3. Fairlie Decomposition of Gender Gaps

To identify the sources of gender disparities, we apply the nonlinear Fairlie decomposition, which separates the observed gender gap into an explained component (arising from differences in characteristics) and an unexplained component (arising from differences in coefficients or unobserved factors).

Table 3. Fairlie Decomposition of Gender Gaps in Digital Financial Inclusion.

Panel A. Aggregate Decomposition
Outcome	Digital Account Ownership	Any Digital Payment Use	Merchant Digital Payments
Predicted probability (Men)	0.759	0.754	0.455
Predicted probability (Women)	0.641	0.628	0.365
Total gender gap	0.118	0.126	0.089
Explained component	0.095	0.099	0.084
Unexplained component	0.023	0.027	0.005
Share explained (%)	80.9%	78.8%	93.9%
Panel B. Contributions to the Explained Component
Variable	Digital Account Ownership	Any Digital Payment Use	Merchant Digital Payments
Age	0.003	0.008	0.008
Secondary education	0.011	0.005	0.005
Tertiary education	0.001	0.002	0.002
Income quintile 2	−0.002	−0.001	−0.001
Income quintile 3	0.005	0.007	0.000
Income quintile 4	0.017	0.014	0.000
Income quintile 5	0.001	0.001	0.001
Employed	0.014	0.021	0.022
Urban residence	0.000	−0.001	−0.001
Mobile phone ownership	0.015	0.015	—
Internet access	0.029	0.028	0.048

Note: Values are rounded to three decimal places. Phone ownership is omitted in the merchant payment decomposition due to perfect prediction. Internet access is included because it is one of the main contributors to the explained gender gap.

presents the Fairlie decomposition results for the three dimensions of DFI.

For digital account ownership, men have a predicted probability of 0.759 compared to 0.641 for women, producing a gap of 11.8 percentage points. Of this difference, 9.5 percentage points—about 81 percent—is explained by observable characteristics. The remaining 2.3 percentage points are unexplained by the covariates included in the model. This indicates that most of the measured gap is statistically attributable to differences in endowments, but it does not imply that all institutional or behavioral barriers are absent. Connectivity variables are among the main contributors. Internet access alone explains nearly 3 percentage points of the gap, while mobile phone ownership accounts for about 1.5 percentage points. Employment differences contribute roughly 1.4 percentage points, and higher income quintiles, especially the fourth quintile, also play a meaningful role. Urban residence contributes negligibly. These results suggest that the gender gap in account ownership is strongly associated with unequal access to digital and economic resources.

For any digital payment use, the predicted probabilities are 0.754 for men and 0.628 for women, implying a gap of 12.6 percentage points. Approximately 9.9 percentage points—around 79 percent—are explained by differences in characteristics. Internet access again emerges as the most important factor, explaining close to 2.8 percentage points of the gap. Employment differences account for about 2.1 percentage points, while income disparities, particularly in higher quintiles, also matter. Education contributes positively but more modestly. The unexplained component is 2.7 percentage points and is discussed further below.

At the merchant payment stage, roughly 94 percent of the 8.9 percentage-point gap is explained by observable characteristics, leaving an unexplained component of 0.5 percentage points. This suggests that observable characteristics explain most of the measured merchant-payment gap. However, the small residual should still be interpreted cautiously because it may reflect unobserved constraints, measurement limits, or differences in how similar resources translate into actual use.

The unexplained component is smaller than the explained component, but it should not be dismissed. For digital account ownership and any digital payment use, the unexplained components are 2.3 and 2.7 percentage points, respectively. In a country the size of Nigeria, even a small percentage-point residual may correspond to a large number of adults. This residual may reflect unobserved factors not captured in the Global Findex data, including social norms, intra-household control over financial decisions, trust in digital platforms, product design, measurement limits, or differences in how similar resources translate into actual use. Therefore, the decomposition results should be interpreted as evidence that observable endowments explain most of the measured gap, not as evidence that all institutional or behavioral barriers are absent.

Overall, these findings indicate that observable structural factors explain most of the measured gender gap in Nigeria’s DFI. The results point especially to digital connectivity and labor market participation as important channels, while also leaving room for residual institutional or unobserved constraints.

4.4. Gender-Specific Marginal Effects (Heterogeneity by Gender)

Table 4. Gender-Specific Average Marginal Effects (AMEs) from Logit Models.

	Digital Account Ownership		Any Digital Payment Use		Merchant Digital Payments
Variable	Men	Women	Men	Women	Men	Women
Age	0.001 (0.001)	0.004 ** (0.002)	0.003 * (0.001)	0.004 ** (0.002)	0.003 * (0.002)	0.004 * (0.002)
Secondary education	0.078 (0.049)	0.170 *** (0.058)	0.038 (0.048)	0.106 * (0.056)	0.082 (0.070)	0.085 (0.068)
Tertiary education	0.105 (0.095)	0.306 *** (0.105)	0.149 ** (0.073)	0.276 ** (0.108)	0.180 (0.126)	0.226 * (0.135)
Income Q2	0.077 (0.062)	0.006 (0.065)	0.037 (0.060)	0.003 (0.070)	0.082 (0.078)	−0.037 (0.077)
Income Q3	0.103 * (0.061)	−0.002 (0.066)	0.099 * (0.057)	0.055 (0.070)	0.006 (0.080)	−0.024 (0.077)
Income Q4	0.212 *** (0.055)	0.133 ** (0.068)	0.180 *** (0.053)	0.178 ** (0.070)	0.027 (0.075)	0.057 (0.080)
Income Q5	0.143 ** (0.055)	0.124 * (0.066)	0.118 ** (0.052)	0.158 ** (0.070)	0.079 (0.073)	0.113 (0.076)
Employed	0.070 * (0.040)	0.182 *** (0.038)	0.114 *** (0.039)	0.184 *** (0.040)	0.195 *** (0.065)	0.096 ** (0.049)
Urban	−0.003 (0.031)	0.057 (0.037)	0.015 (0.032)	0.011 (0.040)	0.026 (0.042)	0.063 (0.043)
Owns mobile phone	0.317 *** (0.075)	0.144 ** (0.072)	0.310 *** (0.080)	0.170 ** (0.068)	—	0.133 (0.107)
Internet access	0.252 *** (0.027)	0.261 *** (0.031)	0.251 *** (0.027)	0.263 *** (0.033)	0.316 *** (0.038)	0.294 *** (0.036)
Observations	552	446	552	446	517	446

Notes. Average marginal effects reported. Robust standard errors in parentheses. *, **, *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Phone ownership omitted in the male merchant payment model due to perfect prediction.

presents the gender-specific average marginal effects from separate logit models estimated for men and women across the three outcomes.

The gender-specific estimates reveal some heterogeneity in the associations between socioeconomic characteristics and DFI across stages of adoption in Nigeria. Because the sample is split by gender, these estimates should be interpreted cautiously and treated as exploratory. They are useful for identifying possible differences in pathways into digital finance, but they should not be read as definitive evidence of gender-specific causal effects.

For digital account ownership, age and education show stronger marginal associations for women than for men. For example, tertiary education is associated with an approximately 30-percentage-point higher probability of account ownership among women, compared with a smaller and statistically insignificant association among men. This pattern suggests that education may be especially relevant for women’s entry into formal digital finance, although the estimates should be interpreted with caution because the tertiary-education group is small.

Employment also shows gender-specific patterns. While employment is positively associated with account ownership for both groups, the association is larger for women. This suggests that labor market attachment may be particularly important for women’s digital financial participation, possibly because employment increases regular income flows, transaction needs, and contact with formal financial systems. However, this should be interpreted as an association rather than a causal effect.

Income gradients are present for both genders, particularly at higher quintiles, but they appear somewhat steeper for men at the upper end for basic account ownership. However, for advanced usage, such as merchant digital payments, income becomes less consistently significant for both groups. This may indicate that income is more important for entry into basic digital finance than for all forms of advanced digital use, once connectivity and employment are considered.

Connectivity variables remain among the strongest correlates across all outcomes. Internet access shows large, positive, and highly significant marginal effects for both men and women at every stage. The magnitude of the internet association is similar across genders, indicating that connected women and men show broadly comparable predicted participation patterns. Mobile phone ownership is also strongly associated with digital participation, though its association is more consistent in the earlier stages and is not always statistically significant in the most advanced stage.

For merchant digital payments, employment and internet access remain the strongest correlates for both men and women. The gender difference in marginal effects narrows at this stage, suggesting that once individuals have the necessary connectivity and economic linkages, men and women may show more similar patterns of advanced digital payment use. This interpretation is consistent with the broader structural argument of the paper, but it remains descriptive rather than causal.

Overall, these results indicate that the absence of a statistically significant female coefficient in pooled models does not imply equality in the pathways into digital finance. Rather, the gender-specific estimates suggest that some socioeconomic characteristics, especially education and employment, may be more strongly associated with women’s digital participation, while connectivity infrastructure is strongly associated with participation for both groups. The evidence therefore supports a policy focus not only on expanding digital infrastructure, but also on strengthening women’s human capital and labor market integration as complementary conditions for DFI.

4.5. Human Capital Mechanism: Gender × Education Interaction

Table 5. Gender × Education Interaction (Human Capital Mechanism) (Logit models; robust standard errors).

Variables	Digital Account Ownership	Any Digital Payment Use	Merchant Digital Payments
Female	−0.280 (0.399)	−0.218 (0.393)	−0.102 (0.414)
Secondary education	0.773 ** (0.325)	0.444 (0.328)	0.450 (0.317)
Tertiary education	0.923 (0.837)	1.440 * (0.754)	0.840 (0.607)
Female × Secondary	0.181 (0.451)	0.065 (0.444)	0.087 (0.449)
Female × Tertiary	1.190 (1.242)	0.396 (1.207)	0.422 (0.916)
Controls	Yes	Yes	Yes
Observations	998	998	998
Pseudo R2	0.330	0.306	0.184

Notes: Controls include age, income quintiles, employment, urban residence, phone ownership (where applicable), and internet access. Robust standard errors are reported in parentheses. * and ** denote statistical significance at the 10% and 5% levels, respectively.

examines whether the association between education and digital financial participation differs by gender. The main education terms are positive in most specifications, suggesting that education is associated with higher digital financial participation overall. However, the interaction terms between gender and education are not statistically significant across the three outcomes. This means that the data do not provide clear evidence that the education-participation relationship differs systematically between men and women.

This finding is important for the paper’s structural interpretation. It suggests that education matters as a general resource for digital inclusion, but the measured returns to education are not clearly gender-specific in these specifications. In other words, education may help both men and women engage with digital finance, but it does not appear, on its own, to remove the structural conditions that create gender gaps.

The result should also be interpreted in light of sample-size limitations. The interaction estimates have relatively large standard errors, especially for tertiary education, reflecting the small number of respondents in that category. Therefore, the absence of statistically significant interaction effects should not be read as proof that gender differences in educational returns do not exist. Rather, it indicates that such differences are not clearly detected in this sample.

4.6. Digital Infrastructure Mechanism: Gender × Internet Interaction

Table 6. Gender × Internet Interaction (Predictive Margins) (Logit models; predicted probabilities).

Outcome	No Internet (Male)	No Internet (Female)	Internet (Male)	Internet (Female)
Digital Account	0.557	0.550	0.865	0.835
Any Digital Payment	0.543	0.535	0.859	0.817
Merchant Payment	0.225	0.209	0.557	0.566

Notes: Values are predicted probabilities from logit models estimated with robust standard errors. Controls include age, education, income quintiles, employment status, and urban residence (and phone ownership where applicable).

examines whether internet access alters the gender gap in digital financial participation. The predicted probabilities show that internet access is strongly associated with higher participation for both men and women. For digital account ownership, predicted participation rises from 0.557 to 0.865 for men and from 0.550 to 0.835 for women. A similar pattern appears for any digital payment use. These results suggest that connectivity is a major entry point into digital finance for both groups.

The merchant payment result is especially important. Among respondents without internet access, predicted merchant payment use is 0.225 for men and 0.209 for women. Among respondents with internet access, the predicted probability rises to 0.557 for men and 0.566 for women. This means that connected women slightly exceed connected men in predicted merchant payment use, although the difference is small and should not be over interpreted.

There are two possible interpretations. First, this pattern is consistent with the idea that once a key infrastructure barrier is removed, women can participate in advanced digital payments at rates similar to men. This supports the structural interpretation of the paper. Second, the result may reflect sample composition among internet-connected women, who may be more urban, economically active, or already more integrated into digital markets. Because the data are cross-sectional and the subgroup is smaller, the result should be interpreted as suggestive rather than causal.

Overall, Table 6 shows that internet access is strongly associated with digital participation, but connectivity alone is not a complete solution. Its effect operates alongside employment, income, education, and broader institutional conditions. This reinforces the view that digital infrastructure is necessary for inclusion, but not sufficient by itself to eliminate gender disparities.

4.7. Robustness Checks: Probit Estimates

Table 7. Probit Estimates: Average Marginal Effects of Digital Financial Participation.

Variables	Digital Account Ownership	Any Digital Payment Use	Merchant Digital Payments
Female	−0.012	−0.021	−0.001
	(0.023)	(0.024)	(0.028)
Age	0.002 **	0.003 ***	0.003 ***
	(0.001)	(0.001)	(0.001)
Secondary Educ.	0.129 ***	0.069 *	0.095 **
	(0.037)	(0.036)	(0.044)
Higher Educ.	0.189 **	0.208 ***	0.197 **
	(0.075)	(0.060)	(0.090)
Income Q2	0.036	0.009	0.026
	(0.046)	(0.046)	(0.051)
Income Q3	0.063	0.081 *	−0.004
	(0.046)	(0.046)	(0.053)
Income Q4	0.183 ***	0.172 ***	0.042
	(0.044)	(0.043)	(0.051)
Income Q5	0.141 ***	0.136 ***	0.090 *
	(0.043)	(0.043)	(0.049)
Employed	0.137 ***	0.155 ***	0.148 ***
	(0.027)	(0.027)	(0.037)
Urban	0.032	0.021	0.040
	(0.023)	(0.024)	(0.028)
Phone Ownership	0.213 ***	0.231 ***	0.247 ***
	(0.047)	(0.048)	(0.082)
Internet Access	0.253 ***	0.258 ***	0.300 ***
	(0.021)	(0.021)	(0.026)
Observations	998	998	998

Notes: Robust standard errors in parentheses. Base categories are primary education or less and lowest income quintile. *, **, *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

presents average marginal effects derived from probit specifications.

The probit estimates closely mirror the logit findings. The marginal effect of being female remains negative but statistically insignificant across outcomes. Magnitudes are small relative to the effects of connectivity and employment.

Internet access increases participation probabilities by roughly 25–30 percentage points, while phone ownership and employment remain strongly associated with digital engagement.

The consistency between logit and probit specifications supports the main finding that the measured gender disparities in DFI in Nigeria are largely associated with unequal structural endowments rather than a statistically robust residual gender effect.

4.8. Additional Subgroup Robustness Checks

To address the possibility that the main findings are driven by location-specific patterns, the baseline logit models are re-estimated separately for rural and urban respondents. The results are reported in

Table 8. Additional Subgroup Robustness Checks: Urban and Rural Samples. Average marginal effects from logit models.

	Rural			Urban
Variables	Digital Account Ownership	Any Digital Payment Use	Merchant Digital Payments	Digital Account Ownership	Any Digital Payment Use	Merchant Digital Payments
Female	−0.065 *	−0.032	−0.037	0.001	−0.031	0.020
	(0.037)	(0.038)	(0.042)	(0.029)	(0.031)	(0.040)
Age	0.003 *	0.005 ***	0.005 ***	0.001	0.002	0.002
	(0.002)	(0.002)	(0.002)	(0.001)	(0.001)	(0.002)
Secondary education	0.051	0.040	0.052	0.208 ***	0.119 **	0.140 *
	(0.049)	(0.050)	(0.058)	(0.059)	(0.054)	(0.073)
Tertiary education	0.130	0.113	0.158	0.255 ***	0.263 ***	0.246 **
	(0.116)	(0.124)	(0.206)	(0.090)	(0.070)	(0.110)
Income quintile 2	0.088	0.029	0.087	−0.007	0.015	−0.044
	(0.063)	(0.062)	(0.069)	(0.065)	(0.069)	(0.080)
Income quintile 3	0.098	0.118 *	0.067	0.016	0.042	−0.103
	(0.063)	(0.061)	(0.069)	(0.066)	(0.068)	(0.084)
Income quintile 4	0.261 ***	0.221 ***	0.091	0.125 **	0.150 ***	−0.020
	(0.068)	(0.067)	(0.074)	(0.054)	(0.057)	(0.076)
Income quintile 5	0.184 ***	0.182 ***	0.094	0.099 *	0.104 *	0.059
	(0.063)	(0.063)	(0.070)	(0.055)	(0.058)	(0.073)
Employed	0.091 **	0.108 **	0.111 *	0.154 ***	0.174 ***	0.170 ***
	(0.044)	(0.044)	(0.060)	(0.031)	(0.033)	(0.048)
Owns mobile phone	0.217 ***	0.217 ***	0.298 **	0.245 ***	0.280 ***	0.277 *
	(0.074)	(0.068)	(0.139)	(0.071)	(0.083)	(0.151)
Internet access	0.296 ***	0.284 ***	0.277 ***	0.203 ***	0.221 ***	0.307 ***
	(0.030)	(0.031)	(0.035)	(0.028)	(0.028)	(0.039)
Observations	469	469	469	529	529	529

Notes: Average marginal effects are reported. Robust standard errors are in parentheses. Base categories are primary education or less and the lowest income quintile. The models are estimated separately for rural and urban respondents; therefore, urban residence is not included as a covariate. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

. These subgroup checks are exploratory because splitting the sample reduces statistical precision. However, they provide useful evidence on whether the main findings are sensitive to rural–urban differences.

The results show that the female coefficient remains statistically insignificant in the urban models across all three outcomes. In the rural models, the female coefficient is negative for all three outcomes, but it is statistically significant only at the 10% level for digital account ownership. This suggests that some residual gender difference may remain in rural account ownership, although this result should be interpreted cautiously given the smaller subgroup sample and the exploratory nature of the analysis.

The broader pattern remains consistent with the main results. Internet access is strongly associated with digital financial participation in both rural and urban samples across all three outcomes. Employment is also positively associated with participation in both locations, although the magnitude and precision vary by outcome. Education appears more strongly associated with participation in the urban sample, while higher income quintiles are more consistently related to account ownership and digital payment use in the rural sample.

Overall, the subgroup results support the main conclusion that digital connectivity and labor market attachment remain central correlates of DFI across locations. At the same time, the rural digital-account result suggests that location-specific constraints may still matter and should be considered in policy design. These findings reinforce the need to interpret the gender-specific and interaction results cautiously, especially when the sample is split into smaller groups.

4.9. Model Diagnostics

As a final check,

Table 9. Diagnostic Checks for Logit Models.

Diagnostic	Digital Account Ownership	Any Digital Payment Use	Merchant Digital Payments
Mean VIF	1.52	1.52	1.52
Max VIF	2.49	2.49	2.49
Hosmer–Lemeshow χ2 (p-value)	8.92 (0.349)	11.20 (0.190)	10.09 (0.259)
Link test: _hatsq p-value	0.070	0.147	0.285
Observations	998	998	998

Notes: Multicollinearity assessed using VIFs from linear probability models. Hosmer–Lemeshow tests use 10 quantile groups. Link test reports significance of squared predicted values.

reports diagnostic tests for the baseline logit models.

The variance inflation factors indicate no serious multicollinearity concerns, with a mean VIF of 1.52 and a maximum VIF of 2.49. The Hosmer–Lemeshow tests do not reject model fit across the three outcomes. The link tests also provide no strong evidence of misspecification, as the squared fitted terms are not statistically significant. Overall, these checks support the reliability of the baseline logit specifications.

5. Discussion

The findings of this study indicate that the gender gap in digital financial inclusion (DFI) in Nigeria is mainly associated with the unequal distribution of economic and digital resources, rather than clear evidence of gender-specific behavioral responses. Although the descriptive statistics show a wide gap across the three stages—digital account ownership, use of digital payments, and merchant payments—the gap shrinks substantially once education, income, employment, and internet access are controlled for. The loss of statistical significance of the female coefficient in the pooled models does not mean that gender inequality is absent. Rather, it suggests that the observed gender gap is largely accounted for by measurable socioeconomic and digital-access characteristics. This interpretation is consistent with capability-based approaches, which argue that economic inclusion depends on the availability of institutional and social resources that enable individuals to transform access into effective participation (Hilbert, 2011; Sen, 1999). From a social indicators perspective, DFI reflects not only technological expansion but also the degree of individuals’ economic agency and integration into formal systems, as highlighted in recent research (Shahen & Sharaf, 2025).

The Fairlie decomposition results support this structural interpretation, but they should be read with caution. Between 79% and 94% of the measured gender gap across the different stages is explained by observable characteristics, mainly digital connectivity, labor market integration, and income levels. At the same time, the unexplained component should not be dismissed. For digital account ownership and any digital payment use, the unexplained components remain around 2.3 and 2.7 percentage points, respectively. In a country the size of Nigeria, even a small percentage-point residual may represent a large number of adults. This residual may capture factors not observed in the Global Findex data, such as social norms, intra-household control over financial decisions, trust in digital platforms, product design, financial autonomy, measurement limits, or differences in how similar resources translate into actual use. Regulatory design may also shape trust and effective use (Sharaf et al., 2026c). Accordingly, the decomposition results should be interpreted as evidence that observable endowments explain most of the measured gap, not as evidence that all institutional or behavioral constraints are absent.

From a broader comparative perspective, the Nigerian case fits within economies where digital exclusion is closely connected to labor market and income inequalities. Cross-country analyses suggest that gender gaps in account ownership and digital payments narrow substantially once structural variables are incorporated into empirical models (Khera et al., 2022). World Bank assessments similarly emphasize that digital divides often mirror broader gaps in economic empowerment rather than differences in willingness to engage with financial tools (Demirguc-Kunt et al., 2022). Recent cross-country evidence by Yoon et al. (2025) also provides useful context. Their study shows that countries with higher levels of financial inclusion beyond what would be predicted by GDP per capita are associated with narrower gender gaps in account ownership and digital payments. The present study complements this country-level evidence by showing, at the individual level, that digital connectivity and labor market participation account for much of the observed gender gap in Nigeria.

The results are also consistent with literature distinguishing between the access gap and the use gap, where digital infrastructure is treated as a basic condition for inclusion (Hilbert, 2011). The large association of internet access with digital participation—around 25–30 percentage points across the main models—suggests that connectivity functions as an important entry point into digital finance. This is in line with GSMA reports documenting that usage gaps are closely linked to connectivity and infrastructure gaps (GSMA, 2022, 2023). However, internet access should not be treated as a stand-alone solution. Connectivity may enable participation, but its value depends on whether individuals also have income, employment links, financial knowledge, trust in digital systems, and opportunities to use digital payments in daily life.

The Nigerian findings differ partly from studies that document a persistent unexplained gender gap at more advanced stages of digital use, even after socioeconomic characteristics are taken into account. In some contexts, movement toward business-related or merchant transactions is associated with wider gaps because women may face stronger institutional, normative, or autonomy-related constraints (GSMA, 2022; van Deursen & van Dijk, 2019). In contrast, the Fairlie decomposition in Nigeria shows that the merchant payment gap is mostly explained by observed characteristics, with a smaller unexplained component. One possible explanation is that Nigeria still has large underlying inequalities in employment, income, and internet access. In such a setting, observed gender gaps may be driven more by unequal endowments than by differential returns to the same resources. This helps reconcile the Nigerian results with studies such as (Morrar et al., 2024) and (Özşuca, 2025), where unexplained or second-stage gaps remain more visible. The difference may therefore reflect context, measurement, and the stage of digital participation being examined, rather than a contradiction in the literature.

The results also qualify more deterministic arguments that digitalization automatically reproduces existing inequalities at all stages of use. Critical literature emphasizes that financial technology may amplify structural imbalances if it is not accompanied by institutional reforms (BIS, 2021; Seguino, 2011). The Nigerian evidence suggests a more nuanced pattern. Digital finance does not appear to be associated with a large additional residual gender gap once socioeconomic and digital-access characteristics are considered. However, it also does not overcome the unequal distribution of those characteristics. In this sense, digitalization mirrors existing structural inequalities more than it independently resolves them.

The internal pathways into inclusion also matter. The gender-specific models show that education and employment are strongly associated with women’s digital financial participation. This finding is consistent with literature linking economic empowerment to financial participation (Kabeer, 1999; Seguino, 2011). Employment may matter because it provides regular income flows, increases the need for payments and transfers, and brings individuals into contact with formal financial systems. Education may matter because it improves the ability to understand, trust, and use digital tools. Internet access lowers the practical cost of participation by enabling remote transactions and platform-based payments. These pathways help explain why the gender gap narrows when employment, income, education, and connectivity are included in the models. At the same time, existing research increasingly questions the assumption that education and digital access generate uniform returns across genders, showing that women may derive lower benefits from these resources because of structural and labor market constraints (Sharaf et al., 2026a). Yet, the insignificant interaction terms between gender and education indicate that education is not clearly detected as a distinct gender-specific corrective channel in this sample. Instead, it appears to support participation within the broader distribution of opportunities.

The interaction results further refine the interpretation of the main findings. The gender × education models do not show statistically significant interaction effects, which suggests that the association between education and DFI is not clearly different for men and women in this sample. This should be interpreted cautiously because the standard errors are large, especially for tertiary education. The gender × internet margins show that internet access is strongly associated with higher participation for both men and women. In merchant payments, connected women slightly exceed connected men in predicted participation, but this small difference should not be over interpreted. It may suggest that once connectivity barriers are removed, women can participate in advanced digital payments at rates similar to men. It may also reflect the composition of connected women, who may be more urban, employed, educated, or already integrated into digital markets.

The consistency of results across logit and probit specifications supports the stability of the main findings. Still, these robustness checks are limited because they test sensitivity to model form rather than all possible sources of heterogeneity. The added rural–urban subgroup checks provide useful evidence that the main results are not driven only by one location group. However, further splits by age, education, or income would reduce precision given the sample size. For this reason, the gender-specific and interaction results are best interpreted as exploratory evidence rather than definitive subgroup conclusions.

Overall, these findings position Nigeria in an intermediate place within the international debate on gender and digitalization. They support the view that gender gaps in DFI are strongly associated with the unequal distribution of resources and capabilities. At the same time, they do not fully support the view that advanced usage necessarily produces a large independent return gap. Nor do they support the optimistic claim that connectivity alone is sufficient to close gender gaps, since connectivity operates within the structure of labor markets, income distribution, and institutional conditions.

Therefore, the Nigerian case reframes the theoretical question on gender and digitalization. The issue is not whether women are less inclined to use financial technology, but whether the economic and digital environment gives them the resources needed to convert access into regular and meaningful use. When those resources are more evenly distributed, digital financial outcomes may become more similar. When they are unevenly distributed, digital expansion can reproduce existing inequalities even without a large residual gender coefficient. Digital exclusion, in this sense, is better understood as a problem of capability distribution within the economy, rather than as a purely technological or behavioral gap.

6. Conclusions

This study examined whether gender gaps in DFI in Nigeria are mainly linked to structural inequalities or whether they persist after observable characteristics are taken into account. Using nationally representative microdata from the 2025 Global Findex survey, DFI was conceptualized as a multistage process—digital account ownership, use of any digital payment, and merchant digital payments as a more advanced form of engagement. Nonlinear probability models, decomposition techniques, and interaction specifications were used to distinguish between differences in endowments and residual differences after observable characteristics are considered.

The results are consistent across the main models. Raw gender gaps are visible at each stage of digital participation. However, once differences in education, income, employment, and digital connectivity are taken into account, the estimated female coefficient becomes small and statistically insignificant. Decomposition results indicate that the majority of the observed gap is statistically attributable to differences in observable characteristics, particularly internet access and labor market attachment. In the Nigerian context, gender disparities in digital finance therefore appear to be mainly associated with unequal access to economic and digital resources rather than clear evidence of gender-specific behavioral responses.

The policy implications are clear and actionable. Reducing gender gaps in digital finance requires more than expanding access to digital tools. For regulators, the findings suggest that financial inclusion strategies should move beyond account ownership targets and monitor actual use of digital financial services, especially by gender and location. For banks and fintech firms, the results point to the importance of designing products that reduce onboarding barriers, improve trust, support low-income users, and respond to women’s transaction needs. For digital inclusion programs, expanding affordable internet access and mobile connectivity remains essential, but it should be combined with financial literacy, consumer protection, and support for women’s economic participation. For labor market and inclusion policies, strengthening women’s employment opportunities may be one of the most effective ways to deepen digital financial participation.

The analysis is subject to several limitations. The data are cross-sectional and therefore do not permit causal interpretation. Employment, income, internet access, and digital finance usage may influence one another, so the estimated relationships should be interpreted as conditional associations rather than one-way effects. Behavioral mechanisms, such as social norms, intra-household bargaining dynamics, trust, or control over financial decisions, cannot be directly observed in the Global Findex data. The final estimation sample of 998 respondents also limits the precision of subgroup and interaction estimates, and the Nigeria-only focus means that the findings should not be generalized mechanically to other African economies.

Future research can take this analysis further by using panel data, experimental designs, or qualitative interviews. Panel data or experiments would help examine how changes in employment, income, or connectivity are associated with women’s digital financial engagement over time. Larger samples would allow more detailed subgroup analysis by rural or urban residence, age, education, and income. Qualitative research could further examine how women experience digital financial services, including issues of trust, autonomy, product design, and control over financial decisions.

Overall, the evidence suggests that gender gaps in Nigeria’s digital financial system are mainly associated with structural inequalities. As digital finance continues to expand across Africa, addressing the underlying economic and infrastructural inequalities that shape participation will be central to ensuring that digital transformation translates into inclusive development. The main lesson is that digital access matters, but it is not enough on its own. A more inclusive digital financial system requires connectivity, employment opportunities, consumer protection, and economic agency to advance together.