Does financial inclusion empower women in emerging countries? A panel data analysis

Women The study using both fixed-effects and panel generalized methods of moments (GMM) estimation techniques covering a panel dataset of 50 emerging economies from 2005 to 2017. We use accessibility to automated teller machine services, availability of bank branches and accessibility to bank assets as three different measures of financial inclusion and gender inequality index, gender development index and political participation of females as the measure of women empowerment. We find evidence that each measure of financial inclusion is associated with lower gender inequality, higher gender development and more opportunities for political involvement of women respectively. Using different econometric estimation method and a board range of inclusion measures confirms the more robust association between financial inclusion and women empowerment. The study, thus, concludes that financial inclusion plays an important role in the welfare of female and girls which would further assist emerging economies to achieve Sustainable Development Goal 5.


Background
Women's empowerment through financial inclusion becomes a cross-cutting issue of discussion among the international development organizations, policymakers and researchers as studies found the persistent presence of gender gap in financial inclusion. Representing "gender equality and empower all females" in 2030 SDG agenda makes this topic more significant. In developing nations, gender inequality in access to banking services is still high. More than half of the unbanked populations in the world are women. 67% of men have a formal account with the banks and financial institutions whereas 59% of the female have access to an account. Since 2011, this gap almost remains constant which is consistently more than 8% (Demirgüç-Kunt et al. 2018). Moreover, girls and females are still facing various economic, social and political limitations and obstacles. They have lower labor force participation rate, lower access to financial services, lower enrollment and compilation rate of education, and lesser rate of political participation. Women spend a larger proportion of their time in the household works and child care than men. In some societies, females are the targets of many violent activities. Employed women are also facing persistent gender-based inequity at their workplace including a greater gender wage gap. These types of inequalities have both social and economic costs including losses of working productivity, higher health care, and legal expenses. Minimizing gender disparities and improving women's welfare is extremely crucial not only for social and moral aspects but also from an economic development standpoint. Full access to financial products and services is a policy tool for women's full participation in the economy and more inclusion of females in the economy would lead to benefits in the growth of GDP (IFC, 2016).
Enhanced and easier access to financial credit facilities help women entrepreneurs to increase the productivity of the enterprise, managerial capacity, and more profits. Women's additional income allows them to save, further productive investment and more consumption (World Bank/OECD, 2013) which facilitates efficient allocation of resources within households that leads to an increase in family expenditure to health care, child nutrition and education (Castilla and Walker, 2013).
Financial inclusion refers to a circumstance in which individuals and businesses, including those currently excluded from the formal banking system, have access to effective financial products and services that meet their needs such as access to credits, daily transactions, savings, and deposit transactions and insurance (AFI, 2016). This is one of the important and proven regulatory instruments to increase macroeconomic welfare, enhance stability, and reduce poverty and inequality (Beck et al. 2007;Cull et al. 2012;Kim 2016). Participation on financial services, particularly access to accounts, helps individual to start a new business and expand an existing one, encourages more investment in schooling and health, and increases savings and consumption which have a greater impact on reducing income inequality, rapid economic growth and women empowerment (Isaac 2014) Thus, this study investigates the main research question-does financial inclusion (FI) empowers women? Besides, this study also analyzes-what is the major contribution of different indicators of FI, such as access to banks, access to ATMs and access to credit facilities, on enhancing women empowerment and how these indicators could help to achieve SDG 5? This study uses the gender inequality index (GII) as the outcome variable because it measures three major dimensions at the same time such as political engagement, participation rate to the labor force and mortality rate.
Lower the value of GII implies lower the gap between males and female which in turn means high women empowerment. Covering panel data of 50 developing countries from 2005-2017, this study finds that more access to banks, ATMs and credit facilities are negatively associated with gender inequality. The results are highly statistically significant and robust to both fixed effects and generalized method of moment estimation technique. This study also considers the gender development index (GDI) and women's political empowerment as the dependent variable instead of GII. Employing fixed effects estimation confirms that FI helps to improve GDI and also increases the opportunity of the political empowerment of females.
This study contributes to the literature in the following three ways. First, the study uses three important proxy indicators of financial inclusion, such as access to banks, access to ATMs and access to banks services and analyzes their impact mainly on reducing gender disparities. This would be the pioneering study to investigate such results in the developing economy perspective. Second, most of the existing papers related to financial inclusion and women's development are based on theoretical perspective and case study. Numerous numbers of literatures are related to financial inclusion and economic growth. An empirical study on women empowerment through financial inclusion is still untouched and has a potential scope of the profound investigation. Third, the study considers three dependent variables that represent women empowerment in a single econometric specification using the latest available annual panel data till 2017, applies both static and dynamic panel data estimation techniques and controls country-specific heterogeneity, endogeneity, and autocorrelation. These techniques provide more robust findings and offer a strong basis for evidence-based policymaking on financial inclusion which helps to promote women empowerment in developing countries.
The remaining part of the study is arranged as follows. Section 2 analyzes existing literature on financial inclusion and women empowerment. Section 3 methodology and data. Section 4 discusses the sources of data and several stylized facts related to our study. Section 5 and 6 present the results and discussion and robustness tests, respectively, while 7 concludes the paper with some policy implications.

5
The impact of indicators of financial inclusion on women empowerment is still unclear due to the limitation of gender centric disaggregated data as well as fewer empirical studies. Numerous reports and case studies state the descriptive background through which financial inclusion could reduce the gender gap and increase women's participation in the economy. But very few empirical studies are available on nexus between these two macroeconomic factors. At the very beginning, several studies have been conducted on nexus between women's access to credit facilities, particularly access to micro-finance, and gender inequality. But findings from those studies are still controversial. Studies Where, GII it is the Gender Inequality Index, the proxy of women empowerment for country i in year t. lnFI p,it refers to the natural log of financial inclusion measures, p. LF = percentage of female labor force of total labor force, NRR = natural resources rent percentage of GDP, CON = household consumption, TF = trade freedom score, GDP = Annual growth of Gross domestic product. η i represents country fixed effect, λ t indicates the time effects, and ε it is the residual term that also captures the effect of all unobserved omitted variables. The result of the Hausman specification test signifies that the fixed effects model is more appropriate than the random-effects model. As a result, panel fixed effects estimation results are reported directly in this paper.
Moreover, there is a possibility to arise the debate that all independent variables exploit in our fixed effects model are not rigorously exogenous. A model that allows controlling for the endogeneity of independent variables, serial autocorrelation, and heterogeneity through the use of lagged dependent variable and internal instruments is the Arellano and Bond (1991) difference GMM estimator. The study also employs the difference GMM estimation of our baseline model. Two appropriate post estimation tests, test for second-order correlation in the first differenced residuals (the Arellano-Bond test) and test for over-identifying restrictions (the Hansen and Sargan Test) are carried out on the difference GMM estimation. Therefore, the specification for difference GMM estimator is-This study also uses different measures of women empowerment, such as gender development index  Table 1, the descriptive statistics of the data are reported. The total number of observations is 650.
The correlation matrix, presented in Table 2, check out the pattern of relationships between the variables that the study considers for analysis. The gender inequality index is found to be negatively     On the contrary, evidence of positive co-move between financial indicators and GDI is noticeable.
Such opposite co-movement may be most significant between financial inclusion and GII. Intuitively, a high level of financial inclusion initiatives would associate with the lower values of gender disparities and higher values of gender development in developing countries. The study endeavors to confirm this understanding with the econometric specification with additional control variables in the empirical section. The subsequent key hypothesis to be verified in this study is the following-H0: There is no effect of financial inclusion on reducing gender disparities between male and female through access to ATMs, access to bank branches and access to bank loans program.
H1: There is a significant effect of financial inclusion on reducing gender disparities between male and female through access to ATMs, access to bank branches and access to bank loans program.

Results And Discussion
The preliminary assessment is employed using OLS and panel fixed effects model. All the variables are in percentage form except GII and trade freedom index. Table 3 illustrates the results using these two methods. The results in this table validate the main hypotheses that higher the access to ATMs, bank branches and loan facilities in a country the lower is the disparities between male and female, ceteris paribus. All three indicators of financial inclusion have a negative coefficient and in every case are strongly statistically significant at a 1% level. Panel data estimation using OLS could raise bias results because of omitted variables and unobserved country characteristics. The fixed-effects model can control this unobserved heterogeneity, thus our discussion is mainly based on fixed effects estimation. Moving from column 4 to column 6, the fixed effects estimation results shows that (1) a one percent increase in banks ATMs decreases GII by 2.61 points, (2) a one percent raise of bank branches cut GII by 3.37 points and (3) a unit percent increase of people's access to bank loans reduce GII by 0.19 points, ceteris paribus. This means that more access to saving and checking accounts for any formal and informal financial transaction through automated teller machines, deposit machines, bank branches or mobile banking encourages more savings; investment and household consumption empower the women. Access to bank loans and advances affects consumption patterns positively, generates more income and employment opportunities for women that would help to reduce the gender gap in the developing countries. These enhanced financial inclusion initiatives including financial literacy programs encourage women's labor force participation; develop entrepreneurship, helping to set up new SMEs and boost inclusive growth (IMF 2018).   For the control variable female labor force percentage of the total labor force, the study gets expected negative sign signifying that the higher the inclusion of female labor force in a country, the lower is the gender gap and the coefficient is statistically significant at 1% level. The coefficient of natural resources rents over GDP is also highly significant and has theoretically expected positive sign, meaning that the higher the natural resources rents, the higher the gender disparities. The positive sign of the coefficient of the control variable domestic consumption implies that the higher the household consumption in an economy, the higher is the GII and all coefficients are economically significant in all cases but statistically significant only in the FE model with main explanatory variable access to loan facilities (column 6). The coefficient of domestic credit to private sectors appears to be statistically highly significant negative effects on GII except it affects through access to loan facilities which is statistically insignificant and has no evidence of explanation at least 10% level. The coefficients of freedom to trade are negative and statistically strongly significant only when financial inclusion affects GII through access of people to bank branches and access to more loan facilities. In every instance and model, the coefficients of GDP growth are positive but statistically insignificant.
Controlling year effect in the fixed effects models (Column 7-9) points out, there is no significant change in baseline results. Moreover, the effect of GDP growth is found highly significant indicating that GDP growth reduces gender inequality.

Robustness Test
This part of the study is related to the test of the robustness of the findings obtained from the fixed effects estimations. The robustness test is split into following two sections:

Robustness test using the alternative method of estimation and controlling endogeneity
Modified Wald test and Pesaran CD test respectively find out the presence of group-wise heteroskedasticity and cross-sectional dependency in the fixed effects regression model results (Table 4). To overcome the potential problem of endogeneity, country-level heterogeneity, heteroskedasticity and serial autocorrelation in the fixed effects models, the study employs difference GMM estimation of dynamic panel data proposed by Arellano and Bond (1991). The Arellano and Bond estimator is fit for datasets with larger cross-sections and few periods. This study follows Roodman (2009). The first difference of all the control variables are treated as standard instruments, and the first lag of GII and GDP growth serve as GMM-type instrument in our dynamic estimation. Table 4 illustrates the one step Arellano-Bond Dynamic Panel-Data Estimation results. The dynamic panel data estimation results suggest that all the coefficient of financial inclusion indicators is negative and significant indicating that more access to ATMs, bank branches and loan facilities lower the gender gap, hence, empower women. All findings are coinciding with the results of OLS and fixed-effects models (Table 3) only the variation is in the difference in the level of significance. A one percent increase of access to bank's ATMs, branches and loans cut down the gender inequality index by 2.334, 3.319 and 0.363 points respectively, on average in the short run, Ceteris Paribas. The estimates are statistically significant at 1, 5, and 10% level respectively. (ii) the instrument set is (2 3) and (iii) *** p < 0.01, ** p < 0.05, * p < 0.1 Table 4 Arellano-Bond dynamic panel data estimation results The coefficient of female labor force and natural resources rent is also the same as the baseline results in fixed effects estimation with expected sign. The effects of these coefficients are strongly statistically significant at a 1% level which is similar to the level of significance in fixed-effects estimation. The study does not find any statistical significance of at least a 10% level to describe the coefficient of household consumption and domestic credit impacts on empowering women. But the coefficient of trade freedom is found a positive sign and statistically significant at the 10% level when the channel of the impact of financial inclusion is access to loan facilities by the people. In the fixedeffects model, the study found a negative and highly significant impact of trade freedom on GII.
Economic growth measures by GDP growth rate are strongly significant at a 1% level and a negative sign is expected in all cases. The lag of dependent variable GII is also highly and positively significant indicating that the gender gap is persistent over the period and developing countries may face gender difference trap. A study on aid impact on gender inequality Pickbourn and Ndikumana (2016) also find a similar result. Unequal socio-cultural norms and structures regarding initial access to resources and economic opportunities are one of the main causes of persistent of the inequality trap.
GMM estimator has to satisfy several diagnostic checks including Sargan and Hansen test for instrument validity through over-identifying restrictions, and AR (1) and AR (2) tests for serial correlation. These test results are reported at the last three rows of Table 4 signifies that the study does not reject the null hypothesis of the validity of over-identifying restrictions, thus there is no indication of model misspecification in all specifications, and there is no evidence of first-order autocorrelation AR (1), but only in the second-order autocorrelation AR (2). These diagnostic tests confirming that in all models underlying assumptions for the selection of valid instruments are satisfied.
6.2 Robustness test using the substitute dependent variables Table 5 Robustness test using the substitute dependent variables Table 5 Robustness test using the substitute dependent variables (1) (2) (3) (4) (5)   Table 5 illustrates the results of the baseline model using fixed effects estimation techniques where the control variables are identical. Instead of our main dependent variable GII, the study uses GDI (Column 1-3) and political empowerment (column 4-6) as an independent variable. Moving from columns 1 to 6, it is observed that the indicators of financial inclusion have strong positive effects on GDI and the political empowerment of women. All the results are as our expectation and statistically significant at a 1% level indicating that more geographical outreach of automated teller machines, commercial banks outlets and access to loan facilities not only help to improve the gender development but also enhance the political empowerment of females in the developing nations. These findings are in support of our baseline fixed effects model results ( Table 5). The study does not find any statistically significant relationship to describe the impact of GDP on GDI and the political empowerment of females. But the addition of females to the total labor market has a strong positive relationship with both dependent variables. We find the freedom to execute international trade does not have significant positive effects on women's political empowerment but it significantly facilitates women empowerment by increasing of gender development index. As natural resources rent increases gender disparities, so the negative impact of this control is expected and significant at 1% level to GDI and political empowerment indicators. Therefore, the study can conclude that financial inclusion helps to diminish gender inequality, assists to improve gender development and empower women's political engagement.

Conclusions
This study brings to light the relationship between financial inclusion and women empowerment and contributes to the literature that more access to the financial premises and services reduces gender disparities, promotes women empowerment that would help developing countries to achieve SDGs, particularly SDG 5. This paper uses three indicators of financial inclusion such as access to ATMs, access to bank branches and access to loans from the Financial Access Survey, IMF database and tests the link between each of them and gender inequality index in fixed effects and difference The important policy implication of this study is that developing countries are likely to obtain extensive benefits from the inclusion of more female people to banking channels towards achieving the SDGs. This also helps regulatory authorities and the government to set up an appropriate legal and regulatory framework in support of women-centric financial inclusion policies. Moreover, women targeted financial education programs, gender-responsive financial strategies, and customized product sales and promotion policies are effective strategies for women empowerment. This study would benefit from considering more gender-specific disaggregated data on financial inclusion which are currently not available for long-run impact analysis on women empowerment (Global Findex database of World Bank only consist of data for three years). The availability of disaggregated data would offer the possible scope for future empirical analysis in this field on both single country and cross-country perspective. Appendix B See Table 6   Table 6 Sources of data Appendix C See Table 7   Table 7 Test of multicolinearity 20 Declarations

Ethics approval and consent to participate
Ethical approval and consent to participate is not applicable for this study.

Consent for publication
Not Applicable

Availability of data and material
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.