Exploring the Gender Dimension in Financial Inclusion in India: Insights from the Global Findex Database

Abstract

Despite great progress in recent years in improving access to formal financial services in developing countries, there is still significant access and usage disparity between men and women. We use individual-level data from the World Bank’s Global Findex 2017 database to investigate the differences in the use of accounts by men and women who are account holders and who save and borrow money formally to provide a comprehensive picture of individuals’ financial behaviour in India. Employing treatment effects estimations through the use of propensity scores, we find that although there is improvement in saving behaviour, borrowing behaviour as well as financial resilience among men and women who save and borrow formally as compared to those who do not, women lag significantly behind men in terms of the same. Thus, expanding women’s access to economic opportunity is critical to achieving gender equality in financial inclusion and unlocking the potential for economic empowerment and development. Our findings can aid in the formulation of policies that consider the unique requirements of women in order to increase financial inclusion and close the gender gap in India.

Keywords

Financial inclusion gender gap saving behaviour borrowing behaviour financial resilience Global Findex Data India

Background

The relevance of financial inclusion has been garnering the attention of policymakers and academicians alike, as it is based on the principles of equity and inclusive growth (Chakrabarty, 2012). Financial inclusion, which is broadly defined as the usage of formal financial services, is critical for economic growth. Individuals who are not financially disadvantaged can invest in education and entrepreneurship, thus reducing poverty and boosting economic growth (Beck et al., 2007; Bruhn & Love, 2014). Furthermore, countries with higher levels of financial inclusion have higher GDP growth rates and lower income inequality (Demirgüç-Kunt & Levine, 2009; Demirgüç-Kunt & Singer, 2017; King & Levine, 1993).

The Government of India introduced the Pradhan Mantri Jan-Dhan Yojana (PMJDY) in 2014 as part of its National Mission for Financial Inclusion to help every adult in India open a bank account, inclusive of mobile banking, and approximately 330 million bank accounts were opened as a result (Government of India, 2014). While data show that 77% of females are financially included, there is still a 6% gender disparity in bank account ownership (Demirgüç-Kunt et al., 2020). Some studies suggest that women’s status in India is rooted in society’s patriarchal prejudices, which confine them to a marginalised position at the bottom of the economic, social and political hierarchy. As a result, financial organisations refuse to lend to women since they are deemed high-risk customers with insufficient resources. Women’s limited financial literacy, paired with their time constraints, prevents them from grasping the complexities of saving account ownership. As a result, they are caught in a vicious cycle of low income, low savings, limited credit and low returns on investment (Das Barwa, 2015). Banks are supposed to spend time educating female customers, but there is little evidence that this information is effectively delivered (Sabherwal et al., 2019). These obstacles demonstrate that more work remains to be done in India to provide women with acceptable and useable financial services.

Research evidence shows that people who participate in the financial system are better equipped to manage risk, start or invest in a business and support big expenditures such as school or home improvements (Ashraf et al., 2010; Cull et al., 2014; Dupas & Robinson, 2013). However, we find that there is a lack of empirical evaluation at the individual level to understand the differences in the use of accounts between men and women in terms of their saving and borrowing behaviour, that is, undertaking formal saving and formal borrowing for business purposes, medical purposes and old age as well as financial resilience capacity, that is, the possibility of coming with emergency funds at short notice through formal saving. Given how often women are marginalised in society, a gender viewpoint is especially important.

Most studies for India have been conducted with bank-level data that provide interesting perspectives on the unbanked (Basu & Srivastava, 2005; Burgess & Pande, 2005; Günther, 2017; Kumar, 2013) as well as gender disparities in formal finance (Duflo, 2012; Ghosh & Vinod, 2017; Kaur & Kapuria, 2020; Swamy, 2014). However, these studies mostly use household-level data. Our paper uses individual-level data from the 2017 Global Findex survey for India to investigate disparities between male and female users of financial services in terms of their saving and borrowing behaviour to fulfil certain welfare objectives, viz., starting/operating/expanding a business, preparing for old age and medical purposes as well as their capacity to come up with funds at short notice for emergency purposes. Global Findex statistics have become a tool for tracking progress towards the World Bank’s goal of universal financial access by 2020 and the United Nations Sustainable Development Goals in recent years (Demirgüç-Kunt et al., 2018). To the authors’ knowledge, no previous research has concentrated on these aspects of financial inclusion in India, where the process of financial reforms and liberalisation is still ongoing, giving a fertile testing ground for a thorough investigation into individual financial behaviour. This is our paper’s first contribution.

A second contribution is that in line with N’dri and Kakinaka (2020), we consider the use of formal accounts for formal saving and formal borrowing to be intervention/treatment variables that people choose to improve their well-being. Individuals who formally save and borrow are classified as part of the treatment group, while those who do not are classified as part of the counterfactual group or control group. The unobserved counterfactual dependent variable is imputed by comparing treated and untreated people who have as many pre-treatment features in common as possible. However, because the treatment assignment is not random, there is a risk of selection bias in the measurement of treatment effects (De Janvry et al., 2010; Heckman & Vytlacil, 2007). This is because the reasons for formal saving and formal borrowing might be dependent on men’s and women’s backgrounds, making the decision endogenous. Our study, thus, employs treatment effects estimation using matching and weighting approaches, such as propensity score matching (PSM) and Kernel Weighting, to avoid such an endogeneity problem. The idea behind these strategies is to mimic randomisation in terms of treatment assignment, as seen in randomised controlled trials (N’dri & Kakinaka, 2020). While PSM is widely used in other fields such as public health, where random assignment is not always possible, there are few examples of its application in the financial inclusion field. Using a dataset that was collected as part of a randomised control trial, Cintina and Love (2019) use PSM to measure the impact of microcredit in India. Swain and Floro (2012) also use PSM to measure the impact of participation in bank-connected self-help groups in India. Wellalage and Locke (2020) investigate the impact of remittances on the financial inclusion of refugee migrants using World Bank survey data of 1,041 Syrian refugees and applying inverse probability of treatment weighting propensity score analysis. Thus, to the best of our knowledge, our study is the first to use such approaches to mitigate selection bias in analysing the gender gap in bank account usage in India by women using World Bank survey data.

Third, approaches such as PSM and kernel weighting, have the drawback of failing to account for the presence of unobserved factors that can be linked to both the treatment and outcome variables. The presence of such unseen factors can skew our average treatment effects estimates. We, therefore, employ the Rosenbaum approach to account for the selection of unobservables (Rosenbaum, 2002).

Evolution of Financial Inclusion in India

In India, the financial inclusion process can be generally divided into three phases. During the first phase (1960–1990), the emphasis was on channelling of credit to neglected sectors of the economy. Special emphasis was also laid on the marginalised sections of society. The second phase (1990–2005) focused mainly on strengthening financial institutions as part of financial sector reforms. Financial inclusion in this phase was encouraged mainly by the introduction of the SHG–bank linkage programme in the early 1990s and Kisan Credit Cards (KCCs) for providing credit to farmers. The SHG–bank linkage programme was launched by the National Bank for Agriculture and Rural Development in 1992, with policy support from the Reserve Bank of India (RBI), to facilitate collective decision-making by the poor and provide ‘doorstep’ banking. During the third phase (2005 onwards), financial inclusion was explicitly defined as a policy objective, and the thrust was on providing a safe facility of savings deposits through no-frills accounts (Chakrabarty, 2009).

In recent years, there has been a plethora of emerging players, ranging from traditional banks, niche financial entities such as payments banks, small finance banks, micro finance institutions and fintech companies. Banks have been advised by the RBI to put in place financial inclusion plans, consisting of achievements against several parameters such as the number of outlets, basic savings bank deposit accounts, overdraft facilities availed in these accounts, transactions in KCCs and general credit cards as well as transactions through business correspondents–information and communication technology channel. The setting up of the National Centre for Financial Education by regulators and the implementation of the Centre for Financial Literacy project of RBI are two recent initiatives towards improving financial literacy. Further, to measure the extent of financial inclusion in the country, it has been decided to construct and periodically publish a ‘Financial Inclusion Index’. The index will have parameters across the three dimensions of financial inclusion, viz., access, usage and quality (Das, 2021).

Moreover, to mitigate the financial impact of COVID-19–related disruptions, the RBI has taken several policy measures to ease the flow of credit at a lower cost to needy segments such as lowering of policy rate, launching of on-tap liquidity schemes and channelising of liquidity through all-India financial institutions and facilitating financial institutions to resolve stressed loans to individuals, small business and MSMEs. During the pandemic, cash transactions at BC outlets through micro-ATMs have witnessed a significant surge, with more than 94 crore transactions accounting for ₹2.25 lakh crore during 2020–2021. The impact of the digital payment in direct benefit transfers can be discerned from the fact that ₹5.53 lakh crore was transferred digitally across 319 government schemes spread over 54 ministries during 2020–2021 (Das, 2021).

Brief Literature Review on Gender and Financial Inclusion

Global Studies

Demirgüc-Kunt et al. (2013) find that a significant gender gap exists in account ownership, formal saving and formal credit, and Fungácová and Weill (2015) study financial inclusion in China and find that women are less likely to be financially included because of a lack of documentation or because another member of the family has an account. Zins and Weill (2016) find that being a woman increases informal saving and reduces borrowing for business purposes or land purchasing. Delechat et al. (2018) find that a robust negative relationship exists between being female and financial inclusion and legal discrimination, and lack of protection from harassment, including at the workplace and more diffuse gender norms, act as possible explanatory factors. Cabeza-García et al. (2019) provide evidence that greater financial inclusion of women, measured as access to a bank account and access to credit cards, has a positive effect on economic development. Ndoya and Tsala (2021) find that there is a gap in all indicators of access to and use of financial products and services in favour of men in Cameroon. The results also show that the largest contributor to the gender gap in access to financial products and services is income, with a contribution of more than 50%. The largest contributor to the gender gap in the use of financial products and services is education, with an average contribution of more than 35%. Mndolwa and Alhassan (2020) provide evidence to support gender disparities in financial inclusion in Tanzania, mostly pronounced in formal savings (21.3%), formal accounts (10.6%) and mobile money accounts (9.4%). This is reflected in 17.1% less likelihood of female ownership of a formal savings account but a 2% higher likelihood to access formal credit. Gender disparities in financial inclusion are explained by lower levels of education, income and over‐dependence of women on men. Sakyi-Nyarko et al. (2022) find that financial inclusion significantly improves household financial resilience, but this effect does not significantly vary by gender or locality in Ghana. Remittances via mobile money provide significant financial resilience effects, with generally stronger effects in rural than in urban areas, especially for females.

Indian Studies

Ghosh and Vinod (2017) find that on average, female-headed households are 8% less likely to access formal finance and 6% more likely to access informal finance as compared to households that are headed by males. Similarly, households headed by females use 20% less cash loans as compared to male-headed households. For female-headed households, education and wages are more relevant in explaining access to finance, whereas political and social factors are much more germane in explaining the use of finance. Kaur and Kapuria (2020) find that female-headed households, especially those belonging to socially disadvantaged castes, have a lower probability of accessing institutional finance vis-à-vis male-headed households. Kulkarni and Ghosh (2021) find that the economically developed states of Maharashtra, Gujarat, Karnataka, Uttar Pradesh and Tamil Nadu witness a lower percentage of female respondents who own and use a mobile phone for undertaking digital banking transactions. Household income, age and ownership of a smartphone and autonomy in decision-making regarding household finance can influence a woman’s decision about the use of digital means of doing transactions. However, a threat to security and privacy, lack of digital literacy and societal norms are the most prominent deterrents to the acceptance and use of digital finance. Therefore, there is a need to adopt a women-centric and gender-sensitive approach to financial inclusion which shifts the focus of financial inclusion toward mitigating gender inequalities, unjust sociocultural norms and the existing regulatory framework.

Data and Methodology

Data Source

With financing from the Bill and Melinda Gates Foundation, the World Bank established the Global Findex database in 2011, a complete data collection on how adults invest, borrow, make payments and manage risk. It has now become a keystone of global efforts to foster financial inclusion. The first round of surveys, conducted in cooperation with Gallup, Inc. in 2011, was followed by a second in 2014 and a third in 2017. The measurements in the 2017 Global Findex database are based on survey data from nearly 150,000 people in 144 economies, representing over 97% of the world’s population. The target population is the entire civilian, non-institutionalised population aged 15 and up. Data weighting is used to ensure that each country’s sample is nationally representative. The base sampling weight, which compensates for uneven probabilities of selection based on household size, and the post-stratification weight, which corrects for sampling and non-response errors, make up the final weights. Post-stratification weights are calculated using economy-level population statistics on gender and age, as well as education or socio-economic status when accurate data are available (Demirgüç-Kunt et al., 2018).

The survey was undertaken in India from 21 April to 2 June 2017. Three thousand people were interviewed face to face in Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil and Telugu. North-eastern states and outlying islands, which account for less than 10% of the population, are not included in the sample (Demirgüç-Kunt et al., 2018).

Main Indicators and Sample Set

People put money aside for future needs, such as a significant purchase, an investment in school or a business, or their own needs in old age or in the event of an emergency. They may also choose to borrow if they are faced with more immediate expenses. Findex data from around the world reveal how and why people save and borrow, as well as their financial resilience in the face of unexpected expenses. For account owners, formal saving is identified through the following survey question: In the past 12 months, have you, personally, saved or set aside any money by using an account at a bank or another type of formal financial institution? The survey, then, asks about two specific reasons for saving in the past year—for old age and to start, operate or expand a business (Demirgüç-Kunt et al., 2018). We refer to this variable as saving behaviour. Second, to identify formal borrowing, the survey asks both men and women the following question: In the past 12 months, have you, by yourself or together with someone else, borrowed any money from a bank or another type of formal financial institution? The survey further asks whether people have borrowed money in the past 12 months for medical purposes or to start, operate or expand a business (Demirgüç-Kunt et al., 2018). We use the term ‘borrowing behaviour’ to indicate this variable.

Finally, financial inclusion is a means to an end since people are better able to manage financial risk when they have a safe place to save money and obtain credit when they need it. To learn more about how financially resilient people all over the world are when faced with unexpected expenses, the 2017 Global Findex survey asks respondents who have savings, whether or not it would be possible to come up with an amount equal to 1/20 of gross national income per capita in local currency within the next month (Demirgüç-Kunt et al., 2018). We denote this variable as financial resilience. After adjusting for missing observations and ambiguous responses, our sample set for treatment effects estimations consists of 1,402 individuals, out of which 739 are female and 663 are male.

Methodology

PSM

PSM estimators have been frequently utilised in assessment research to estimate average treatment effects in observational settings where treatment selection bias is a concern (Abadie & Imbens, 2009; Rosenbaum & Rubin, 1983). A set of participants gets exposed to a treatment while others are not in observational studies like ours. However, there is no randomisation, and the data are collected after the subjects have been exposed to the treatment. Due to this, PSM is especially beneficial because it can reduce bias and confounding. It is possible to ex post pair individuals in the data from both the treatment and control groups based on their propensity score or a participant’s probability of belonging to the treatment group given his or her observational features when matching on propensity score values.

PSM estimators present two advantages over other possible methods. First, instead of matching on several covariates with close or exact values, units with dissimilar covariate values can be matched on equal or similar propensity score values (Abadie & Imbens, 2009; Imbens & Rubin, 2015). A second advantage of the PSM method is that it balances the distribution of the observed baseline covariates between treated and control units with the same propensity score (Abadie & Imbens, 2009; Austin & Stuart, 2015). The average treatment effect on the treated is defined as the average difference between these matched subjects (ATET) (Rosenbaum & Rubin, 1983), which is our measure of interest in this study.

A common first step in generating a propensity score is to perform a logit regression with treatment as the outcome variable and potential confounders as explanatory variables. Trade-offs between variables’ impacts on bias (distance of estimated treatment effect from true effect) and efficiency govern covariate selection (precision of estimated treatment effect) (Garrido et al., 2014). Following the estimation of propensity scores, the ATET can be estimated as follows (N’dri & Kakinaka, 2020):

A T E T = E [Y_{1} | D = 1, (x)] - E [Y_{0} | D = 0, (x)],

where p(x) is the estimated propensity score, E [Y₁|D = 1, p(x)] is the expected outcome for the units that receive treatment (D = 1), E [Y₀|D = 0, p(x)] is the expected outcome for the treated units’ best matches, and x is the set of relevant pre-treatment characteristics of the respondents, viz., age, educational level, within-economy household quintile and employment status. The outcome variables are saving behaviour, financial resilience and borrowing behaviour. Details of both outcome variables and pre-treatment characteristics of the respondents can be found in Table 1.

Table 1.

Variables and Their Categories Used in the Study.

Variables	Categories
Saved using account at FI in past 12 months (formal saving)	Yes, no
Saved for business/old age/both using account at FI in past 12 months (savings behaviour)	Yes, no
Borrowed using account at FI in past 12 months (formal borrowing)	Yes, no
Borrowed for business/medical purposes/both using account at FI in past 12 months (borrowing behaviour)	Yes, no
Possibility of coming up with emergency funds at short notice for individuals if saved formally at FI in past 12 months (financial resilience)	Yes, no
Age (in years)	15–90
Educational level	Completed primary or less, secondary, completed tertiary or more
Within-economy household income quintile	Poorest 20%, Second 20%, Middle 20%, Fourth 20%, Richest 20%
Employment status	In workforce, out of workforce

Source: Global Findex Database, 2017.

Note: FI = financial institution.

PSM is based on three assumptions. The conditional independence assumption, often known as confoundedness, is the first assumption. It asserts that following conditioning on covariates, no unobservable variable should affect both the likelihood of treatment and the outcome of interest. The second assumption is the independent and identically distributed observations assumption, which states that each unit of analysis’ potential outcome and treatment status are independent of all other units in the sample’s possible outcomes and treatment status. The common support or overlap condition is the third assumption, which states that every observation has a positive probability of being in both the treatment and control groups (N’dri & Kakinaka, 2020).

Even if the overlap assumption is met, there may be a significant gap between the propensity scores of the two closest persons eligible for matching, resulting in unsatisfactory matches. The caliper constraint is used in this study to avoid this issue (Caliendo & Kopeinig, 2008), which sets a maximum distance between matched units as a limit. If the distance exceeds this threshold, the treated observation is removed from the analysis to prevent receiving skewed results. The caliper in our study is calculated as 0.2 of the standard deviation of the logit of the propensity score and is considered optimal (Austin, 2011; Garrido et al., 2014).

Kernel Weighting

PSM demands that treatment and control groups overlap in an area of common support to connect participants with the same propensity scores in order to derive causal conclusions from data (Imbens & Rubin, 2015). Individuals who fall outside of this range must therefore be excluded from the analysis. A kernel weight can be used to estimate the counterfactual instead of eliminating unpaired individuals from the comparison group and reducing the sample size. Kernel matching (also known as kernel weighting) is a technique that can be useful in studies that use survey data with sampling weights, such as ours (DuGoff et al., 2014; Imbens, 2000).

Each treated individual is assigned a weight of one in kernel weighting. To construct a match for each treated individual, a weighted composite of comparison observations is employed, with comparison individuals being weighted by their propensity score distance from treated individuals within a range or bandwidth of the propensity score. Kernel weighting improves precision by preserving sample size while reducing bias by giving superior matches more weight. Untreated individuals with similar propensity ratings to treated ones are given higher weights via a kernel function (DiNardo & Tobias, 2001). Kernel weights are well suited to ATET estimation. We can acquire more reliable treatment effects estimates by employing kernel weighting as a robustness check (Garrido et al., 2014).

Covariate Balance Checks

It is also crucial to see if, regardless of treatment, people with the same propensity score have a similar distribution of observable variables or features. After PSM and kernel weighting have been applied, this check can be performed by assessing the balance in the observable features of individuals between treatment and comparison groups (Austin & Stuart, 2015). When a covariate’s distribution does not differ across treatment thresholds, it is considered to be balanced. We calculate standardised differences that take both means and variances into consideration (Austin, 2009; Rosenbaum & Rubin, 1985). The standardised mean difference of a perfectly balanced covariate is zero, and the variance ratio is one (StataCorp, 2013).

Sensitivity Analyses

On the basis of matching/weighting processes, a significant influence of formal saving and formal borrowing on financial behaviour of men and women is estimated in our study. However, if the unobservable qualities affect both the treatment and outcome variables at the same time, a hidden bias may emerge, compromising the results’ robustness. We use sensitivity analysis, based on Rosenbaum’s boundness technique, to determine how strongly hidden biases may influence our results (Rosenbaum, 2002). As our outcome variable is binary, we compute the Mantel–Haenszel (MH) test statistic following Becker and Caliendo (2007) and look for evidence of overestimation of our treatment effects figures because of the presence of unobservables.

If Γ is the ratio of the odds of receiving treatment for two matched individuals i and j with different unobserved characteristics, then in line with Rosenbaum (2002), we can write:

1 / Γ \leq P_{i} / (1 - P_{i}) / P_{j} / (1 - P_{j}) \leq Γ,

where P_i and P_j are the true treatment probabilities that depend on both the observables and the unobservables. Following this, we can vary the values of Γ, commencing from 1, and test if there is overestimation of the true effect, that is, whether the estimated effect remains significant across values of Γ (Caliendo & Tübbicke, 2020).

Results

Table 2 reports our sample statistics. We find that 24% of men use their bank account to save formally while only 17% of women do so. Nineteen per cent of men save for business/old age/both while only 14% of women undertake the same. Women also lag behind men when it comes to financial resilience. While 25% of men are able to come up with emergency funds, only 16% of women are able to do so. Finally, while 10% of men borrow formally using their account, only 6% of women do so in comparison. Twenty-two per cent of men borrow for business/medical reasons/both while only 18% of women do so.

Table 2.

Descriptive Statistics (in %).

Variables	Female (n = 739)	Male (n = 663)
Formal saving	17	24
Savings behaviour	14	19
Formal borrowing	6	10
Borrowing behaviour	18	22
Financial resilience	16	25

Source: Authors’ own calculations.

We further substantiate these figures using empirical methods to highlight financial behaviour of men and women. Table 3 reports the estimated ATETs of formal saving and formal borrowing on our three outcome variables, viz., saving behaviour, financial resilience and borrowing behaviour under PSM and kernel weighting frameworks. The results present clear evidence supporting the fact that use of financial services leads to better welfare at the individual level. Once an individual undertakes use of accounts through formal saving and formal borrowing, the three outcome measures are improved by 0.363–0.392, 0.212–0.235 and 0.307–0.375 points, respectively for women while the corresponding improvements for men are 0.405–0.407, 0.302–0.310 and 0.378–0.406 points, respectively. However, these results indicate that women lag significantly behind men in terms of their financial behaviour, regardless of being account holders as well as users.

Table 3.

Estimates of ATETs of Financial Behaviour Using PSM and Kernel Weighting.

Sample Groups	Female (n = 739)		Male (n = 663)
Technique/Outcomes	PSM	Kernel weighting	PSM	Kernel weighting
Savings behaviour	0.392 (0.056) ***	0.363 (0.048) **	0.405 (0.055) ***	0.407 (0.047) ***
Borrowing behaviour	0.307 (0.095) ***	0.375 (0.084) ***	0.378 (0.087) ***	0.406 (0.065) ***
Financial resilience	0.235 (0.053) ***	0.212 (0.048) ***	0.310 (0.064) ***	0.302 (0.053) ***

Source: Authors’ own calculations.

Note: Figures in parentheses are standard errors; ***, ** and * denote the significance at the 1%, 5% and 10% levels, respectively.

Results of balance checks post-treatment effects estimation with respect to PSM and kernel weighting are shown in Tables 4 -5 and 6- 7, respectively. They illustrate that although we find substantial differences on many unweighted covariates between treatment and control groups in the raw data, once we use matching and weighting techniques to balance the treatment and comparison groups, we obtain good balance on all covariates—all standardised differences are close to 0, and nearly all variance ratios are close to 1. Finally, Tables 8 and 9 highlight the results of our sensitivity analyses for women and men, respectively. We find that the assumption of overestimation gets rejected even up to a Γ of 5 which means that, in order to invalidate our results, the unmeasured factor would have to increase the odds of receiving treatment by 5 times compared to an individual without these characteristics. Therefore, we conclude that our PSM and kernel weighting results are quite robust to unobserved confounders.

Table 4.

Covariate Balance Checks—PSM: Savings Behaviour and Financial Resilience.

Covariates	Female (n = 739)		Male (n = 663)
	Std. Dif.	Var. Ratio	Std. Dif.	Var. Ratio
	Raw (Weighted)	Raw (Weighted)	Raw (Weighted)	Raw (Weighted)
Age (in years)	0.213 (-0.012)	0.777 (0.797)	0.325 (0.058)	1.065 (1.171)
Educational level
Secondary	0.063 (0.028)	1.112 (1.239)	0.089 (0.019)	1.060 (1.178)
Completed tertiary or more	0.239 (-0.026)	2.290 (0.937)	0.259 (-0.078)	1.904 (0.758)
Within-economy household quintile
Second 20%	-0.340 (0.025)	0.572 (1.062)	-0.227 (0.058)	0.680 (1.141)
Middle 20%	0.062 (-0.024)	1.100 (0.804)	-0.091 (0.000)	0.869 (1.000)
Fourth 20%	0.225 (0.026)	1.438 (1.378)	0.160 (0.065)	1.302 (1.309)
Richest 20%	0.333 (-0.018)	1.609 (0.984)	0.414 (-0.092)	1.528 (0.961)
Employment status	0.248 (0.065)	1.071 (1.004)	0.284 (-0.034)	0.696 (1.064)

Source: Authors’ own calculations.

Note: Std. Dif. = standardised difference; Var. Ratio = variance ratio.

Table 5.

Covariate Balance Checks—PSM: Borrowing Behaviour.

Covariates	Female (n = 739)		Male (n = 663)
	Std. Dif.	Var. Ratio	Std. Dif	Var. Ratio
	Raw (Matched)	Raw (Matched)	Raw (Matched)	Raw (Matched)
Age (in years)	−0.055 (−0.036)	0.926 (0.914)	0.374 (−0.015)	1.151 (1.306)
Educational level
Secondary	0.199 (0.011)	1.344 (1.148)	0.007 (0.029)	1.018 (1.108)
Completed tertiary or more	0.166 (−0.076)	1.799 (0.822)	−0.146 (0.067)	0.638 (1.312)
Within-economy household quintile
Second 20%	−0.317 (−0.035)	0.589 (0.755)	−0.142 (−0.093)	0.797 (0.732)
Middle 20%	−0.026 (0.096)	0.983 (1.470)	−0.189 (0.092)	0.725 (1.242)
Fourth 20%	0.313 (−0.051)	1.582 (0.955)	0.033 (0.024)	1.073 (1.263)
Richest 20%	−0.018 (−0.061)	0.990 (0.901)	0.178 (0.066)	1.230 (1.065)
Employment status	0.218 (−0.094)	1.073 (1.009)	0.218 (−0.040)	0.760 (1.071)

Source: Authors’ own calculations.

Note: Std. Dif.= standardised difference; Var. Ratio = variance ratio.

Table 6.

Covariate Balance Checks—Kernel Weighting: Savings Behaviour and Financial Resilience.

Covariates	Female (n = 739)		Male (n = 663)
	Std. Dif.	Var. Ratio	Std. Dif.	Var. Ratio
	Raw (Weighted)	Raw (Weighted)	Raw (Weighted)	Raw (Weighted)
Age (in years)	0.213 (0.002)	0.777 (0.739)	0.325 (0.033)	1.065 (0.920
Educational level
Secondary	0.063 (0.032)	1.112 (1.054)	0.089 (0.065)	1.060 (1.042)
Completed tertiary or more	0.239 (−0.053)	2.290 (0.865)	0.259 (−0.098)	1.904 (0.857)
Within-economy household quintile
Second 20%	−0.340 (−0.018)	0.572 (0.963)	−0.227 (0.013)	0.680 (1.037)
Middle 20%	0.062 (0.009)	1.100 (1.018)	−0.091 (0.048)	0.869 (1.098)
Fourth 20%	0.225 (−0.011)	1.438 (0.994)	0.160 (0.026)	1.302 (1.039)
Richest 20%	0.333 (0.043)	1.609 (1.052)	0.414 (−0.044)	1.528 (0.982)
Employment status	0.248 (−0.014)	1.071 (1.006)	0.284 (0.019)	0.696 (0.970)

Source: Authors’ own calculations.

Note: Std. Dif. = standardised difference; Var. Ratio = variance ratio.

Table 7.

Covariate Balance Checks—Kernel Weighting: Borrowing Behaviour.

Covariates	Female (n = 739)		Male (n = 663)
	Std. Dif.	Var. Ratio	Std. Dif.	Var. Ratio
	Raw (Weighted)	Raw (Weighted)	Raw (Weighted)	Raw (Weighted)
Age (in years)	−0.055 (−0.044)	0.926 (0.973)	0.374 (0.016)	1.151 (1.155)
Educational level
Secondary	0.199 (0.016)	1.344 (1.041)	0.007 (0.032)	1.018 (1.040)
Completed tertiary or more	0.166 (−0.085)	1.799 (0.809)	−0.146 (−0.024)	0.638 (0.923)
Within-economy household quintile
Second 20%	−0.317 (−0.017)	0.589 (0.980)	−0.142 (−0.072)	0.797 (0.890)
Middle 20%	−0.026 (0.108)	0.983 (1.264)	−0.189 (0.051)	0.725 (1.147)
Fourth 20%	0.313 (0.048)	1.582 (1.083)	0.033 (0.069)	1.073 (1.143)
Richest 20%	−0.018 (−0.102)	0.990 (0.869)	0.178 (0.062)	1.230 (1.075)
Employment status	0.218 (0.002)	1.073 (1.023)	0.218 (−0.059)	0.760 (1.133)

Source: Authors’ own calculations.

Note: Std. Dif. = standardised difference; Var. Ratio = variance ratio.

Table 8.

MH Bounds Sensitivity Analysis (Female, n = 739).

Outcome/Treatment Variables	Γ	MH Test Statistic		Significance Level
Outcome/Treatment Variables		O/estn of TE	U/estn of TE	O/estn of TE	U/estn of TE
Saving behaviour	1	11.006	11.006	0.000	0.000
	1.5	8.680	13.711	0.000	0.000
	2	7.186	15.916	0.000	0.000
	2.5	6.097	17.829	0.000	0.000
	3	5.241	19.546	0.000	0.000
	3.5	4.538	21.116	0.000	0.000
	4	3.940	22.574	0.000	0.000
	4.5	3.420	23.940	0.000	0.000
	5	2.961	25.231	0.002	0.000
Borrowing behaviour	1	6.217	6.217	0.000	0.000
	1.5	4.629	8.046	0.000	0.000
	2	3.603	9.525	0.000	0.000
	2.5	2.849	10.801	0.002	0.000
	3	2.255	11.941	0.012	0.000
	3.5	1.764	12.979	0.039	0.000
	4	1.346	13.940	0.089	0.000
	4.5	0.980	14.839	0.064	0.000
	5	0.655	15.685	0.076	0.000
Financial resilience	1	7.374	7.374	0.000	0.000
	1.5	5.303	9.692	0.000	0.000
	2	3.928	11.534	0.000	0.000
	2.5	2.897	13.108	0.002	0.000
	3	2.071	14.504	0.019	0.000
	3.5	1.381	15.773	0.084	0.000
	4	0.786	16.944	0.021	0.000
	4.5	0.263	18.037	0.039	0.000
	5	−0.020	19.065	0.050	0.000

Source: Authors’ own calculations.

Note: Γ is the odds of differential assignment due to unobserved factors. O/estn= overestimation; U/estn = underestimation; TE = treatment effects

Table 9.

MH Bounds Sensitivity Analysis (Male, n = 663).

Outcome/Treatment Variables	Γ	MH Test Statistic		Significance Level
Outcome/Treatment Variables		O/estn of TE	U/estn of TE	O/estn of TE	U/estn of TE
Saving behaviour	1	12.514	12.514	0.000	0.000
	1.5	10.101	15.288	0.000	0.000
	2	8.534	17.528	0.000	0.000
	2.5	7.383	19.464	0.000	0.000
	3	6.476	21.198	0.000	0.000
	3.5	5.728	22.783	0.000	0.000
	4	5.092	24.254	0.000	0.000
	4.5	4.539	25.633	0.000	0.000
	5	4.050	26.936	0.000	0.000
Borrowing behaviour	1	7.332	7.332	0.000	0.000
	1.5	5.484	9.439	0.000	0.000
	2	4.276	11.131	0.000	0.000
	2.5	3.383	12.588	0.000	0.000
	3	2.675	13.885	0.004	0.000
	3.5	2.087	15.068	0.018	0.000
	4	1.584	16.161	0.057	0.000
	4.5	1.145	17.184	0.061	0.000
	5	0.754	18.147	0.072	0.000
Financial resilience	1	10.231	10.231	0.000	0.000
	1.5	7.879	12.849	0.000	0.000
	2	6.310	14.912	0.000	0.000
	2.5	5.135	16.667	0.000	0.000
	3	4.196	18.222	0.000	0.000
	3.5	3.414	19.633	0.000	0.000
	4	2.742	20.935	0.003	0.000
	4.5	2.153	22.150	0.016	0.000
	5	1.629	23.295	0.052	0.000

Source: Authors’ own calculations.

Note: Γ is the odds of differential assignment due to unobserved factors. O/estn = overestimation; U/estn = underestimation; TE = treatment effects.

Discussion

Financial inclusion is synonymous with using formal financial services, and there is an assumption that using formal financial products is associated with higher welfare. This study examines the differences between men and women in usage of accounts following financial inclusion, that is, ownership of a bank account. Using individual-level data from the 2017 Global Findex database on 739 women and 663 men, we find that although saving behaviour, borrowing behaviour as well as financial resilience improve within men who save and borrow formally as well as within women who do the same, overall, women lag significantly behind men in terms of usage of accounts for all three purposes. Thus, there is a need to encourage women to make greater use of their bank accounts because it enhances women’s lives by giving them a voice and empowering them to make better decisions. Our findings, thus, add to the literature (Ghosh & Vinod, 2017; Kaur & Kapuria, 2020; Mndolwa & Alhassan, 2020; Ndoya & Tsala, 2021) in general as well as specifically for India on gender gap and use of formal finance. It is also important to acknowledge that broader social constraints related to intra-household bargaining power and the social status of women in societies like India, which limit the broader impact of financial inclusion on women’s economic empowerment. It is therefore crucial to recognise these constraints to ensure financial inclusion can have a transformational impact (Ng’Weno et al., 2018).

Although scaling up big government programmes, such as PMJDY, can provide a quick path to banking the unbanked and women, it also necessitates localised solutions and multi-layered programming to address chronic challenges and exclusions (Arnold & Gammage, 2019). The RBI has directed banks to undertake a pilot financial inclusion campaign and take a structured approach to financial inclusion, particularly for women. This includes the creation of sex-disaggregated financial inclusion plans and their approval by bank boards for urgent execution. It also entails the establishment of financial literacy centres that hold monthly financial awareness camps in order to provide them with the required abilities to comprehend fundamental banking operations (Das Barwa, 2015). Additionally, financial service providers need to develop products that are more affordable and accessible to women, with favourable terms such as tiered know-your-customer requirements and low interest rates to encourage borrowing. Such products should match customers’ levels of financial literacy and be easy to understand. Moreover, fintech has the potential to overcome financial obstacles that women face, and smartphones are becoming increasingly common, even in the most remote and economically underdeveloped areas. Lowering the cost of opening a bank account allows women to open and retain accounts as well as make transactions on a daily basis. These developments combine a variety of services, such as deposits and withdrawals, as well as payments for goods and services, making it easier for women who, between working and caring for a family, may not have enough time to visit a bank. The ease with which transactions can be completed can free up money that would otherwise be used for transportation to financial institutions. As demonstrated by Suri and Jack (2016), the ease with which technologies can be used can help women better control their time, freeing up time for more productive activities.

From 2020 onwards, as a result of the COVID-19 epidemic, due to the loss of jobs and income, the pandemic has put a great burden on the economic status of individuals and households. The government promptly announced to send 500 rupees each month to women’s PMJDY accounts for the first three months of the lockdown. Unfortunately, due to a lack of gender-disaggregated data in the banking industry as a whole, only PMJDY account holders were eligible for the incentive, leaving many other worthy women out (Nageswaran & Kale, 2021). An examination of the structural barriers faced by women reveals recurring trends that have yet to be addressed by existing policy-level financial inclusion programmes. Some of these issues are social, such as lack of mobility and gender roles that limit women’s economic participation and limit earning and saving opportunities, while others are market related, such as low digital literacy, lack of entrepreneurial skills or market information and a lack of personal collaterals or credit history. Other structural flaws can be seen in the ownership of identification documents, mobile phone usage and internet access (Arora, 2021). As a result, India still has to address the widening digital divide that restricts women’s access at a policy level, especially during a pandemic when physical realities are drastically altered. Therefore, a gender-inclusive approach can help the government and fintech organisations collaborate more effectively. Despite the fact that the pandemic has disproportionately affected women, causing them extreme financial hardship as well as massive job losses due to their overrepresentation in the informal sector, they can begin to re-envision a smoother path from isolation to inclusion with the help of financial technology.

Conclusion

To conclude, more integration of women into the formal financial sector is recommended, since financial empowerment of women has welfare benefits, both from a wealth and well-being perspective. Significant usage of bank accounts can contribute to increasing self-confidence and ability of women to plan better for their futures. It can provide women with the opportunity to mix and interact with women from other communities, thereby increasing their knowledge about issues that are of interest to them (Desai & Tarozzi, 2011). The necessity of the hour is for political advocacy as well as concerted efforts to alleviate social vulnerabilities and increase women’s financial inclusion.

Further, while the strength of our study is that since it is based on national-level data, making the findings representative of the country, the limitation we face is our inability to remark on the association between financial inclusion and gender gap taking into account the impact of COVID-19 because our dataset is limited to 2017. Also, information pertaining to the usage of bank accounts may be susceptible to recall bias, which could affect the precision of the findings. As more updated datasets become available and accessible, there will be fertile ground for future research.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

ORCID iD

Sanjukta Sarkar

References

Abadie

, & Imbens

G. W.

(2009). Matching on the estimated propensity score. No. w15301. National Bureau of Economic Research .

Arnold

, & Gammage

(2019). Gender and financial inclusion: The critical role for holistic programming. Development in Practice , 29(8), 965–973.

Arora

(2021). Fintech industry: The great equaliser for women. https://www.orfonline.org/expert-speak/fintech-industry-the-great-equaliser-for-women/

Ashraf

, Karlan

, & Yin

(2010). Female empowerment: Impact of a commitment savings product in the Philippines. World Development , 38(3), 333–344.

Austin

P. C.

(2009). Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Statistics in Medicine , 28(25), 3083–3107.

Austin

P. C.

(2011). Optimal caliper widths for propensity‐score matching when estimating differences in means and differences in proportions in observational studies. Pharmaceutical Statistics , 10(2), 150–161.

Austin

P. C.

, & Stuart

E. A.

(2015). Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Statistics in Medicine , 34(28), 3661–3679.

Basu

, & Srivastava

(2005). Exploring possibilities: Microfinance and rural credit access for the poor in India. Economic and Political Weekly , 1747–1756.

Beck

, Demirgüç-Kunt

, & Levine

(2007). Finance, inequality and the poor. Journal of Economic Growth , 12(1), 27–49.

10.

Becker

S. O.

, & Caliendo

(2007). Sensitivity analysis for average treatment effects. The Stata Journal , 7(1), 71–83.

11.

Bruhn

, & Love

(2014). The real impact of improved access to finance: Evidence from Mexico. The Journal of Finance , 69(3), 1347–1376.

12.

Burgess

, & Pande

(2005). Do rural banks matter? Evidence from the Indian social banking experiment. American Economic Review , 95(3), 780–795.

13.

Cabeza-García

, Del Brio

E. B.

, & Oscanoa-Victorio

M. L.

(2019). Female financial inclusion and its impacts on inclusive economic development . In Women’s studies international forum (p. 102300). Pergamon.

14.

Caliendo

, & Kopeinig

(2008). Some practical guidance for the implementation of propensity score matching. Journal of Economic Surveys , 22(1), 31–72.

15.

Caliendo

, & Tübbicke

(2020). New evidence on long-term effects of start-up subsidies: Matching estimates and their robustness. Empirical Economics , 59(4), 1605–1631.

16.

Chakrabarty

K. C.

(2009). Banking: Key driver for inclusive growth. RBI Monthly Bulletin , 9, 1479–1486.

17.

Chakrabarty

K. C.

(2012). Financial inclusion: Issues in measurement and analysis. In Irving fisher committee workshop on financial inclusion indicators . Malaysia.

18.

Cintina

, & Love

(2019). Re-evaluating microfinance: Evidence from propensity score matching. The World Bank Economic Review , 33(1), 95–115.

19.

Cull

, Ehrbeck

, & Holle

(2014). Financial inclusion and development: Recent impact evidence (No. 88169, pp. 1–12). The World Bank.

20.

Das Barwa

(2015). Women’s financial inclusion case study. Using sex-disaggregated data to promote women’s financial inclusion in India.

21.

Das

(2021). Financial inclusion—Past, present and future. Inaugural address by Shri Shaktikanta Das, Governor, Reserve Bank of India, delivered at the Economic Times Financial Inclusion Summit on Thursday, July 15, 2021.

22.

Delechat

M. C. C.

, Newiak

M. M.

, Xu

, Yang

M. F.

, & Aslan

(2018). What is driving women’s financial inclusion across countries? International Monetary Fund.

23.

Demirgüç-Kunt

, & Klapper

(2013). Measuring financial inclusion: Explaining variation in use of financial services across and within countries. Brookings Papers on Economic Activity , 2013(1), 279–340.

24.

Demirgüç-Kunt

, Klapper

, Singer

, & Ansar

(2018). The global findex database 2017: Measuring financial inclusion and the fintech revolution . World Bank Publications.

25.

Demirgüç-Kunt

, Klapper

, Singer

, Ansar

, & Hess

(2020). The global findex database 2017: Measuring financial inclusion and the fintech revolution . World Bank Publications.

26.

Demirgüç-Kunt

, & Levine

(2009). Finance and inequality: Theory and evidence. Annual Review of Financial Economics , 1(1), 287–318.

27.

Demirgüç-Kunt

, & Singer

(2017). Financial inclusion and inclusive growth: A review of recent empirical evidence. World Bank Policy Research Working Paper , 8040.

28.

Desai

, & Tarozzi

(2011). Microcredit, family planning programs, and contraceptive behavior: Evidence from a field experiment in Ethio-pia. Demography , 48(2), 749–782.

29.

De Janvry

, Dustan

, & Sadoulet

(2010). Recent advances in impact analysis methods for ex post impact assessments of agricultural technology: Options for the CGIAR. Unpublished working paper, University of California-Berkeley.

30.

DiNardo

, & Tobias

J. L.

(2001). Nonparametric density and regression estimation. Journal of Economic Perspectives , 15(4), 11–28.

31.

Duflo

(2012). Women empowerment and economic development. Journal of Economic Literature , 50(4), 1051–1079.

32.

DuGoff

E. H.

, Schuler

, & Stuart

E. A.

(2014). Generalizing observational study results: Applying propensity score methods to complex surveys. Health Services Research , 49(1), 284–303.

33.

Dupas

, & Robinson

(2013). Savings constraints and microenterprise development: Evidence from a field experiment in Kenya. American Economic Journal: Applied Economics , 5(1), 163–192.

34.

Fungáčová

, & Weill

(2015). Understanding financial inclusion in China. China Economic Review , 34, 196–206.

35.

Garrido

M. M.

, Kelley

A. S.

, Paris

, Roza

, Meier

D. E.

, Morrison

R. S.

, & Aldridge

M. D.

(2014). Methods for constructing and assessing propensity scores. Health Services Research , 49(5), 1701–1720.

36.

Ghosh

, & Vinod

(2017). What constrains financial inclusion for women? Evidence from Indian micro data. World Development , 92, 60–81.

37.

Government of India. (2014). National health mission (NHM). Ministry of Health & Family Welfare, Government of India.

38.

Günther

(2017). The progress of financial inclusion in India: Insights from multiple waves of survey data. SSRN 2946954.

39.

Heckman

J. J.

, & Vytlacil

E. J.

(2007). Econometric evaluation of social programs, part I: Causal models, structural models and econometric policy evaluation. Handbook of Econometrics , 6, 4779–4874.

40.

Imbens

G. W.

(2000). The role of the propensity score in estimating dose-response functions. Biometrika , 87(3), 706–710.

41.

Imbens

G. W.

, & Rubin

D. B.

(2015). Causal inference in statistics, social, and biomedical sciences . Cambridge University Press.

42.

Kaur

, & Kapuria

(2020). Determinants of financial inclusion in rural India: Does gender matter? International Journal of Social Economics , 47(6), 747–767.

43.

King

R. G.

, & Levine

(1993). Finance and growth: Schumpeter might be right. The Quarterly Journal of Economics , 108(3), 717–737.

44.

Kulkarni

, & Ghosh

(2021). Gender disparity in the digitalization of financial services: Challenges and promises for women’s financial inclusion in India. Gender, Technology and Development , 25(2), 233–250.

45.

Kumar

S. M.

(2013). Does access to formal agricultural credit depend on caste? World Development , 43, 315–328.

46.

Mndolwa

F. D.

, & Alhassan

A. L.

(2020). Gender disparities in financial inclusion: Insights from Tanzania. African Development Review , 32(4), 578–590.

47.

Nageswaran

V. A.

, & Kale

(2021). How India could speed the financial inclusion of women. https://www.livemint.com/opinion/columns/how-india-could-speed-up-the-financial-inclusion-ofwomen-11615220290174.html

48.

Ndoya

H. H.

, & Tsala

C. O.

(2021). What drive gender gap in financial inclusion? Evidence from Cameroon. African Development Review , 33(4), 674–687.

49.

N’dri

L. M.

, & Kakinaka

(2020). Financial inclusion, mobile money, and individual welfare: The case of Burkina Faso. Telecommunications Policy , 44(3).

50.

Ng’weno

, Oldja

, Hassan

, & Kapoor

(2018). Demand-side review of financial inclusion for women in entrepreneurship and smallholder agriculture.

51.

Rosenbaum

P. R.

(2002). Overt bias in observational studies. In Observational studies (pp. 71–104). Springer.

52.

Rosenbaum

P. R.

, & Rubin

D. B.

(1983). The central role of the propensity score in observational studies for causal effects. Biometrika , 70(1), 41–55.

53.

Rosenbaum

P. R.

, & Rubin

D. B.

(1985). Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. The American Statistician , 39(1), 33–38.

54.

Sabherwal

, Sharma

, & Trivedi

(2019). Using direct benefit transfers to transfer benefits to women: A perspective from In-dia. Development in Practice , 29(8), 1001–1013.

55.

Sakyi-Nyarko

, Ahmad

A. H.

, & Green

C. J.

(2022). The gender-differential effect of financial inclusion on household financial resilience. The Journal of Development Studies , 58(4), 692–712.

56.

StataCorp

L. P.

(2013). Stata time-series reference manual. Stata Press College Station, TX .

57.

Suri

, & Jack

(2016). The long-run poverty and gender impacts of mobile money. Science , 354(6317), 1288–1292.

58.

Swain

R. B.

, & Floro

(2012). Assessing the effect of microfinance on vulnerability and poverty among low income households. Journal of Development Studies , 48(5), 605–618.

59.

Swamy

(2014). Financial inclusion, gender dimension, and economic impact on poor households. World Development , 56, 1–15.

60.

Wellalage

N. H.

, & Locke

(2020). Remittance and financial inclusion in refugee migrants: Inverse probability of treatment weighting using the propensity score. Applied Economics , 52(9), 929–950.

61.

Zins

, & Weill

(2016). The determinants of financial inclusion in Africa. Review of Development Finance , 6(1), 46–57.