The Gender Health Gap in Ghana: Exploring the Role of Financial Inclusion

doi:10.21203/rs.3.rs-3918162/v1

Download PDF

Research Article

The Gender Health Gap in Ghana: Exploring the Role of Financial Inclusion

https://doi.org/10.21203/rs.3.rs-3918162/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

This paper investigates the impact of financial inclusion on health and the gender-health differences in Ghana using micro data from the sixth wave of the Ghana Living Standards Survey (GLSS) and relying on the IV-probit and 2SLS-IV techniques. The findings suggested significant gender health differences, with female individuals reporting lower health than their male counterparts. Additionally, financial inclusion matters for health and the gender-health gap, as people who have higher level of financial inclusion reports being healthier than their counterparts who are less included, and there is no significant health difference across gender for people with higher level of financial inclusion. This finding is novel and important for policy implications, as financial inclusion may help reduce the gender-health gap in Ghana.

JEL: J16, G2, I12

Gender

Financial inclusion

Health gap

Developing countries

Ghana

Gender health inequality has long been of interest to policy makers and academics in developing countries. Despite many years of policy interventions, there are still gender differences in many aspects of life, especially health, where many outcomes are more favourable for men. It is established in existing literature that, women report more illnesses, have worse health outcomes, and exhibit higher health care utilization compared to men, despite their higher life expectancy (Takahashi, Jang Kino and Kawachi, 2020; Zhang, d’Uva and Doorslaer, 2015; Chun, Khang, Kim and Cho, 2008). There have been many attempts to explain this phenomenon, reasons such as the “illness behavior” (this comprises biological risk factors, acquired risk factors and psychosocial aspects of symptoms and care, as well as health reporting behavior and prior health care) and household energy poverty have been cited as part of the causal factors of the health gap across gender (see Verbrugge, 1989; Molarius and Janson, 2002; Case and Paxson, 2005; Mestl and Aunan, 2005; Malmusi et al., 2012; Iddrisu, Phimister and Zangelidis, forthcoming).

Financial exclusion is another potential reason for the health gap across genders but to the authors’ knowledge this is yet to receive empirical attention. Higher levels of income play a significant role in maintaining better health outcomes through investment in health. Since levels of income in developing countries are reportedly low, policy alternatives that can improve income levels in a sustainable way should be explored. Among such policies that has been recognized as being capable of poverty reduction and enhancing general welfare is financial inclusion (FI) (see Mohammed et al., 2017; Danquah et al., 2017; Churchill and Marisetty, 2019; Bukari and Koomson, 2020). Financial inclusion is defined as the access to and effective use of a variety of suitable financial services by adults, including mobile money services (Demirgüç-Kunt et al., 2018; Demirgüç- Kunt et al., 2017). Previous research found that FI helps to improve health (eg. Sarma and Pais, 2011; Koomson and Ibrahim, 2018; Njiru and Letema, 2018; Li, 2018; Gyasi et al., 2019; Stein and Yannelis, 2019; Inoue, 2019; Matekenya et al., 2020). Specifically, Koomson and Ibrahim (2018) noted that access to and usage of financial services may encourage households to start household enterprises to earn extra income that can be invested in health for the households. And investment in health is an important pathway through which financial inclusion affects health outcomes.

Financial inclusion is increasing steadily in developing countries. Specifically, the Alliance for Financial Inclusion (2015) indicated that Ghana has made much progress in the attainment of universal financial access after signing the Maya Declaration in 2012. Data from the Bank of Ghana (2016) has also shown that between the 2007–2014 period, there has been an increment of over 113% in the number of bank branches in Ghana, showing a rise of about 514 additional bank branches in bank branch penetration. At the same time, there were positive developments in primary indicators of FI. For example, the number of bank depositors per 1000 adults rose to about 501.2 in 2014 from 183.6 in 2005 (World Bank, 2017). However, this increase seems to be skewed towards men. There is evidence of gender differences in financial inclusion, where females are largely excluded (Fanta and Mutsonziwa, 2016; Mndolwa and Alhassan, 2020; Ghosh, 2022). In the case of Ghana, Demirgüç-Kunt et al., (2018) indicated that about 62% of males have a transaction account compared to about 54% of females. Therefore, reconciling women’s less inclusion in the financial sector and the gender-health gap, which is less favourable to women, suggest that their financial exclusion could potentially be a causal factor of the gender health gap. Yet, empirical attention has not been given to FI’s role in the gender-health gap.

This study investigates the role of FI in the gender-health gap in Ghana for people between the ages of 18 and 75 years expected to be active in the financial sector. Specifically, using self-reported health measures, we first investigate to ascertain the health difference across gender as argued in the literature. It is important to confirm this before exploring the role FI plays. Further, using multidimensional FI measures, the paper investigates the health effects of FI. Finally, the paper investigates the role of FI in the gender-health gap.

We employ cross-sectional data from the sixth round of the Ghana Living Standard Survey (GLSS 6) and employ the OLS, binomial probit, and instrumental variable methods for this exercise. The GLSS 6 is used because of its widest countrywide coverage and allows for subsample modelling to further explore the gender differences in the FI-health gap relationship. The results confirm the health gap across gender in Ghana, with male individuals being more likely to be healthy than their female counterparts. Second, the results also show that FI has positive health effects, which potentially may play a role in closing the gender-health gap. Particularly, while people with higher levels of financial inclusion are healthier than their counterparts with lower levels, the results indicate no significant health difference across gender for people with higher levels of financial inclusion. The rest of the paper is organised as follows: Section 2 presents literature on the link between financial inclusion and the gender-health gap. Section 3 discusses the data used, the model and the estimation strategy employed to achieve our stated objectives and description statistics. Section 4 contains the results and analysis. Finally, in section 5, we conclude.

Financial inclusion (FI) is one of the biggest challenges many developing countries face. Despite its welfare-improving effects (Demirgüç-Kunt et al., 2017; Mohammed et al., 2017; Koomson et al., 2020a), many developing countries still face difficulties getting their people included in the financial sector, due to either financial illiteracy, inadequate financial infrastructure, or weak financial institutions. Previous literature measures FI using various indicators, including a multidimensional FI index. In cross-country studies, measures such as the number of branches per adult (financial access), percentage of adults that own an account in regulated institutions (financial usage), and stock market access, among others, are often used (see, for example, Iddrisu and Turkson, 2020; Amidu et al., 2022). At the household level, the use of multidimensional FI measure is common (see Zhang and Posso, 2017; Koomson et al., 2020a; Churchill et al., 2020b; Koomson and Danquah, 2021) and may comprise ownership of bank or mobile money account, ownership of insurance, access to credit/loan, and receipt of financial remittance from the bank or through mobile money among others. This potentially implies the presence of endogeneity bias since FI may depend on the availability of and distance to financial institutions.

Financial inclusion can potentially facilitate the household’s investments, including investment in health (Kuri and Laha, 2011; Abdul-Mumuni and Koomson, 2019). Previous research found that FI is associated with improvements in health among others (eg. Sarma and Pais, 2011; Koomson and Ibrahim, 2018; Njiru and Letema, 2018; Li, 2018; Gyasi et al., 2019; Stein and Yannelis, 2019; Inoue, 2019; Matekenya et al., 2020). For example, Koomson and Ibrahim (2018) noted that access to and usage of financial services may encourage households to start household enterprises to earn an extra income that can be invested in health for the households. Also, FI cushions against and hinders health-related challenges, especially in later life (Gyasi et al., 2019).

Given the arguments on the importance of FI in the improvement of health, women, compared to men, are reportedly less included in the financial sector, particularly in developing countries (see Fanta and Mutsonziwa, 2016; Mndolwa and Alhassan, 2020; Ndoya and Tsala, 2021; Ghosh, 2022). Specifically, in Cameroon, Ndoya and Tsala (2021) find a gap in all indicators of access to and use of financial products and services in favour of men. Demirgüç-Kunt et al. (2018) in Ghana find about 62% males to have a transaction account compared to about 54% females. At the same time, there are reports of gender health gaps which tend to be less favourable to women. Studies have shown that, compared to men, women report more illnesses and have worse health outcomes (Chun, Khang, Kim and Cho, 2008; Zhang, d’Uva and Doorslaer, 2015; Takahashi, Jang Kino and Kawachi, 2020). Reconciling their financial exclusion with their health deprivation presents an open empirical question regarding the role of FI on the gender health gap. Incidentally, this has not been given empirical attention in existing literature as (to the authors’ knowledge) there is no empirical evidence on the FI-gender health gap nexus.

3.1 Main model

The empirical framework follows the work of Gangadharan and Valenzuela (2001), where health outcomes are assumed to be determined by the environment and income. Also, the health outcomes of a population improve as the standard of living improves. We incorporate FI into this relationship giving its importance in income generation, investment in health and improving the standard of living. The relationship between these variables and health outcomes is expressed as:

$$H=f\left(FI,I, E, W\right) \left(1\right)$$

where $H$ represents the health outcome of the population, $FI$ represents their financial inclusion status, $I$ is the income level, $E$ represents the environment, and $W$ refers to other factors that could influence the population’s health outcomes.

Transforming the functional relationship into an econometric model and representing it in a cross-sectional framework, we gradually build the model as follows:

We first investigate the gender-health gap and the health effects of FI as in Eq. (2) below,

$${healthy}_{i}={\beta }_{1}+{\beta }_{2}{female}_{i}+{\beta }_{3}{fi}_{i}+{\sum }_{j=4}^{k}{\beta }_{j}{X}_{ij}+{\mu }_{t}+{\epsilon }_{i} \left(2\right)$$

Consequently, the full model is specified in Eq. (3) as follows,

$${healthy}_{i}={\beta }_{1}+{\beta }_{2}{female}_{i}+{\beta }_{3}{fi}_{i}+{\beta }_{4}{(female}_{i}*{fi}_{i})+{\sum }_{j=5}^{k}{\beta }_{j}{X}_{ij}+{\mu }_{t}+{\epsilon }_{i} (3)$$

Where ${healthy}_{i}$ is the health status of the individual $i$, ${female}_{i}$ is the gender of the individual$i$, ${fi}_{i}$ is the financial inclusion status of the individual $i$, ${(female}_{i}*{fi}_{i})$ is the corresponding interaction. ${X}_{ij}$ is a vector of $k$ variables controlling for individual and household characteristics that affect health outcomes, such as age, education, marital status, household income, type of household cooking and lighting fuels, household size and location of the household, $\beta {\prime }s$ are the parameter vectors, ${\mu }_{t}$ represents district fixed effects which control for unobserved district characteristics, and ${\epsilon }_{i}$ is the random error term of the equation.

One key methodological issue of concern is the potential endogeneity that may bias the estimated effect of financial inclusion on health in our model. For this reason, our identification strategy relies on the use of instrumental variables. Estimation routines such as IV-Probit are appropriate when accounting for endogeneity. Thus, our preferred method is the IV-Probit estimation. With non-linear models, it is difficult to find the marginal effect of an interaction term, therefore, as a robustness check and to help explain the interaction terms we use the Ordinary Least Squares (OLS) and the 2SLS-IV estimations. Least squares, as a natural approach to estimation, makes explicit use of the structure of the model as laid out in the equations above. In addition, least squares, even for non-linear probability models, enjoy robustness compared to other estimators, in the sense that even if the actual model is not a linear regression, the regression line fit by least squares is an optimal linear predictor for the dependent variable. Finally, under the very specific assumptions of the classical model, by one reasonable criterion, least squares will be the most efficient use of the data (Greene, 2003).

Following arguments in previous literature that financial inclusion may depend on financial literacy, adequate financial infrastructure, or strong financial institutions, the estimation of our empirical equations may suffer from endogeneity bias caused by measurement error or omitted variable bias. This is because inadequate financial infrastructure (such as bank branches) leading to long distances to financial institutions and lack of financial literacy may be partly accountable for financial exclusion. Existing studies identified that distance to the nearest bank branch affects FI, in the sense that longer distance to the nearest bank branch is associated with higher financial, in-kind and psychological costs, which worsens FI (Demirgüç-Kunt and Klapper, 2012; Koomson et al., 2020a; Churchill et al., 2020b). Also, studies that used distance as an instrument (eg. Churchill and Marisetty, 2019; Koomson et al., 2020a; Churchill et al., 2020b; Koomson and Danquah, 2021) supported its validity using microfinance and other forms of financial institutions’ operational modalities in rural areas which have increased access to banks across locations. Reiter and Peprah (2015) noted that the provision of rural- and area-specific products by Ghanaian banks had been identified as playing a major role in ensuring success in the country’s microfinance industry. The presence of endogeneity is suspected to lead to biased and inconsistent estimates of the relationship between FI and health outcomes (Koomson, Villano, Hadley, 2020a; Churchill and Marisetty, 2019). Therefore, to account for any potential endogeneity and improve the empirical analysis, we estimate instrumental variable (IV) regressions (two-stage least squares-2SLS and IV-probit). Our identification strategy relies on using instrumental variables as follows: (i) average distance to the nearest bank branch, (ii) number of bank branches in a district, (iii) a binary variable indicating the availability of at least one bank branch in a community (bank in community = 1), and (iv) interaction terms between the ‘female’ dummy and each of the above instruments.

3.2 Data

This study employs micro-level data from the sixth wave of the Ghana Living Standards Survey (GLSS), administered in 2012/2013. The GLSS is a nationally representative household survey, the sampling frame for the survey is the population living in private households in Ghana. The above sample frame is divided into primary and secondary sampling units. The primary sampling unit is the census enumerated areas (EAs) formed within the then ten administrative regions of Ghana based on proportional allocation using the population in each of the regions. On the other hand, the second sampling unit is the households living in each of the enumeration areas. We considered survey 6 due to the wider coverage of households and availability of observations. The sixth round had a total of 18,000 households. Out of this, 16,772 were successfully interviewed, comprising a response rate of 93.2 percent. From the above survey, the main sample for this study is restricted to people aged 18 to 75 years. While this ensures that we center the study on financially active people, it is also consistent with Ghanaian laws, which set the legal age of bank account ownership to 18 years and above.

3.3 Variable measurement

Based on the information from the GLSS sample, we constructed the variables of interest for the empirical analysis. The health outcome variable is a binary variable (healthy) that takes the value 1 for individuals who reported not suffering from any illness during the last two weeks before the interview. Injury incidence is negligible, so excluded from the analysis. Unfortunately, this is the only health measure in the dataset we could rely on in the data. This is a potential limitation to the study, however, self-reported health measures have been widely used in previous literature (eg. Boadi and Kuitunen, 2006; Cundale et al., 2017). Thus, the use of self-reported measures is widely accepted and can provide accurate and efficient assessments of objective states (Cleary, 1997). We control for the gender of the individual, represented by the female binary variable (female = 1).

Following previous literature (Zhang and Posso, 2017; Churchill and Marisetty, 2019; Koomson et al., 2020a; Koomson et al., 2020b; Churchill et al., 2020b; Koomson and Danquah, 2021), we use a multidimensional FI measure. Consistent with Koomson and Danquah (2021), the study considers four dimensions of FI (ownership of bank account; access to credit/loan; ownership of insurance; and receipt of financial remittance from the bank or through mobile money) as detailed in Table 1. Each dimension is equally weighted 0.25 and used to generate an individual financial inclusion score in line with Eq. (6) below.

$${fi}_{i}={w}_{1}{I}_{1}+{w}_{2}{I}_{2}+\dots +{w}_{n}{I}_{n} \left(6\right)$$

where ${fi}_{i}$ is the individual financial inclusion score, ${I}_{i}$ is a binary variable that is 1 if an individual satisfies dimension $i$ and 0 otherwise. ${w}_{i}$ is the weight attached to dimension $i$ with ${\sum }_{i=1}^{n}{w}_{i}=1$.

Table 1

Dimensions and weights for multidimensional FI.
Dimensions & weight	Details
Bank account (0.25)	The individual has a bank account (bank account includes savings, current, fixed deposit, or microfinance account).
Loan/Credit (0.25)	The individual has access to a loan/credit from a bank, microfinance institution or other formal institution.
Insurance (0.25)	The individual has access to medical, life, property, unemployment/income or family insurance.
Financial remittance (0.25)	The individual lives in a household that receives financial remittance from a bank, money transfer service provider or through mobile money.

Also, to control for other factors that may influence individuals’ health outcome, we include as covariates: (i) the log of equivalized household income, (ii) household energy poverty represented by main type of cooking fuel -‘dirty cooking fuel’ which is 1 for dung cake, kerosene, firewood, charcoal or any other biomass fuel and 0 for electricity and gas, and lighting fuel-‘dirty lighting fuel’ which is 1 for kerosene, candles, gas or any other biomass fuel and 0 for electricity (iii) the household size, with a minimum of 1-person household and a maximum of 22 members (iv) the age of the individual in years, where a minimum age of 18 and a maximum of 75 years was considered (v) controls for marital status (never married, married, cohabitating, divorce and widowed), (vi) indicators for educational level of the individual (no education, primary, middle, secondary and tertiary), (vii) the individuals’ employment status and finally, (vii) whether the household is in an urban or rural area.

3.4 Descriptive Statistics

Table 2 presents a summary of the mean values of the variables used in the study as discussed in section 3.3 above. Data in the table show differences in household demographics between males and females. The sample size under consideration is 22,606, out of which 11,505 are males comprising about 51% and 11,101 are females. There is a significant health difference across gender in favour of males. The percentage of those reported being healthy in the total sample is about 86%, with about 89% and 83% in the male and female sub-samples, respectively. On the financial inclusion indices, there are slight differences across gender in favour of males. FI for the total sample averaged 0.143, with averages of 0.147 and 0.140 respectively for males and females.

For marital status, male heads record higher percentages of the following categories; never married and married compared to female heads. However, female heads are the highest in the cohabitating, divorce, and widowed category. In terms of level of education, there is a significant difference in higher education attainment in favour of males. Precisely, females with no education and primary level education are more than their male counterparts. However, males record the highest in the middle, secondary and tertiary education level categories compared to males. Males, on average, are older and belong to households with larger sizes than their female counterparts. Again, out of about 47% households in rural areas of Ghana, about 50% of males live in those households compared to about 45% of females.

Regarding employment and income, the percentage of employed males is slightly lower than employed females, although the difference is insignificant. This may partly drive the income differences in favour of females, with an average of GHS 128.381 equalized income.

Table 2

Descriptive Statistics
	Total	Male	Female	T-test Difference
Sample	22,606	11,505	11,101
Healthy	0.859	0.885	0.832	0.053***
Financial Inclusion	0.143	0.147	0.140	0.007**
Marital Status	---	---	---	---
Never married	0.341	0.391	0.289
Married	0.459	0.469	0.449
Cohabitating	0.098	0.091	0.105
Divorce	0.067	0.039	0.096
Widowed	0.035	0.010	0.061
Education Level	1.894	2.043	1.739	0.304***
No education	0.156	0.123	0.190
Primary	0.141	0.119	0.163
Middle	0.436	0.447	0.425
Secondary	0.187	0.214	0.161
Tertiary	0.079	0.097	0.061
Age	34.779	35.484	34.048	1.435***
Household Size	4.980	4.928	5.033	-0.105**
Rural	0.474	0.501	0.447	0.054***
Employed	0.558	0.556	0.559	-0.0037
Equalized Income	2153.996	2090.953	2219.333	-128.381***
Dirty cooking fuel	0.766	0.778	0.754	0.024***
Dirty lighting fuel	0.043	0.041	0.046	-0.005*
‘---’ means not applicable. * p < 0.01, p < 0.05, * p < 0.10 (Here, a simple t-test is performed by household head gender and revealed significant differences in the variables).

Finally, on the type of cooking and lighting fuels used in households which reportedly have effects on health, about 78% of males live in households that use ‘dirty fuel’ (dung cake, firewood, charcoal etc., other than electricity and gas) for cooking compared to about 75% females. However, about 5% of females are in households that use ‘dirty fuel’ (kerosene, candles, gas, etc., other than electricity) for lighting compared to about 4% males².

4.1 Empirical Estimations and Discussions

This sub-section analyses the empirical results to examine the gender-health gap, the health effects of financial inclusion and the role that financial inclusion plays in the gender-health gap in Ghana. Table 3 presents the results of the Probit, IV-Probit, OLS, and 2SLS-IV regressions that use micro-level data from the sixth round of the Ghana living standards survey (GLSS), accounting for the respective households’ income level, household size, type of main cooking and lighting fuel and location, as well as the individuals’ age, education level, marital status, and employment status. To gradually build the health model, we begin by assessing the gender-health and the health effects of financial inclusion; columns (1), (3), (5), and (7) are estimated using the financial inclusion variables together with the individual and household characteristics as specified in Eq. (2). Finally, in columns (2), (4), (6), and (8), we incorporate the role financial inclusion may play in the gender-health gap by estimating the full model with the interaction term, Eq. (3). All regressions are corrected for robust clustered standard errors, controlled for district effects.

Table 3

Estimates of Equations (1) and (2) using Probit, OLS, IV-Probit and 2SLS instrumental variables (IV)
Dependent Variable: Health Status (healthy = 1, illness = 0)
	Probit (dy/dx)			IV-Probit (dy/dx)		OLS		2SLS-IV
Independent Variables	(1)	(2)	(3)		(4)	(5)	(6)	(7)	(8)
Female	-0.055***	-0.0467***	-0.064***		-0.1040***	-0.0540***	-0.0461***	-0.0631***	-0.135***
	(0.00482)	(0.00593)	(0.006)		(0.049)	(0.00480)	(0.00593)	(0.00626)	(0.0496)
Fin. Inclusion index	-0.00179	0.0301	1.357***		1.215***	-0.00387	0.0236	1.335***	1.141***
	(0.0123)	(0.0183)	(0.314)		(0.346)	(0.0123)	(0.0154)	(0.307)	(0.316)
Female*F. Inclusion		-0.266**(c)			0.855(c)		-0.0550**		0.500
		(0.113)			(1.017)		(0.0241)		(0.339)
Under identification test								59.023(0.000)	59.125(0.000)
Hansen J (overid)								1.441(0.2300)	1.495(0.4736)
Endogeneity test								28.133(0.000)	30.083(0.000)
F-stat								29.842	14.979
District Effect	Yes	Yes	Yes		Yes	Yes	Yes	Yes	Yes
Other controls	Yes	Yes	Yes		Yes	Yes	Yes	Yes	Yes
Observations	22,606	22,606	22,606		22,606	22,606	22,606	22,606	22,606
R-squared						0.034	0.034	-0.464	-0.528
Wald test of exogeneity			32.18(0.000)		43.33(0.000)
Robust standard errors in parentheses, * p < 0.01, p < 0.05, * p < 0.1. For the under-identification, Hansen J. (overidentification) and endogeneity tests, we report the test values with p-values in parenthesis. Also, For the exogeneity tests, we reported the chi sqr. test values with p-values in parenthesis. Complete estimates are provided in Appendix I (OLS and 2SLS_IV) and II (Probit and IV-Probit). (c) means coefficient is reported rather than marginal effects.

While our Probit and OLS estimates may be economically meaningful, the issue of potential endogeneity bias remains. To improve the estimates and account for any potential endogeneity, the IV-Probit and 2SLS-IV regressions are presented in columns (3), (4), (7), and (8) of Table 3. The choice of instruments is supported by the corresponding tests, particularly for the 2SLS-IV, the F-statistics on the test for weak identification of the endogenous regressors (Fin. Inclusion index and Female*Fin. Inclusion) are reported as 29.842 (column 7) and 14.979 (column 8), for equations (2) and (3), respectively. These values exceed the Stock-Yogo (2005) critical values indicating that the endogenous regressors are strongly identified. Furthermore, the test statistics of under-identification and over-identification (Hansen J.), reported at the bottom of Table 3 columns 7 and 8, suggest that the instruments are relevant and the overidentifying restrictions are exogenous, respectively. Thus, the chosen instruments are well-identified. Finally, the endogeneity test rejects the null hypothesis of exogeneity, thus supporting the use of instrumental variables. As a result, the instrumental variable estimates (IV-Probit and 2SLS-IV) are our preferred estimates, as the results account for potential endogeneity and allow us to identify the causal effect of financial inclusion on health. In particular, we rely on the IV-probit while the 2SLS results serve as robustness check and also help with the interpretation of the interaction terms.

The results from Table 3 confirmed the health difference across gender. In line with the findings and arguments in the literature, including that of Verbrugge (1989), Malmusi et al. (2012) and Zhang, d’Uva and Doorslaer (2015) in America, Spain, and China respectively, the coefficient of the gender dummy (Female) is negative and statistically significant at the conventional levels across all regressions in Tables 3. In each case, it indicates that females, on average, are likely to report being ill compared to their male counterparts. Referring to the Probit estimates in column (1), the results suggest that females are about 0.06 percent less likely to report being healthy (not report illness) than males. The probable health difference became more pronounced after accounting for endogeneity in column (3), indicating corrections made to the bias of the Probit estimator. Based on the IV-probit estimates in column (3), females are found to be about 0.06 percent less likely to report being healthy than their male counterparts, all else equal. This finding aligns with our expectations and is consistent with the existing literature. Next, we analyse the health effect of financial inclusion. In line with the argument of previous literature (Sarma and Pais, 2011; Koomson and Ibrahim, 2018; Njiru and Letema, 2018; Li, 2018; Gyasi et al., 2019; Stein and Yannelis, 2019; Inoue, 2019; Matekenya et al., 2020), the results of the IV estimates (IV-probit and 2SLS-IV) show (at 1% significance level) that people who have higher level of financial inclusion are healthier than their counterparts who are less included. Here, the insignificance of the Probit and OLS results may be indicative of their anticipated bias. Referring to the IV-Probit estimates in column (3), people with higher levels of financial inclusion are about 1.14 percent more likely to be healthy than their counterparts with low financial inclusion, all else equal. This finding is intuitive since higher levels of financial inclusion may be associated with higher investment in health.

Finally, the results indicate that financial inclusion may play a role in the gender-health gap, in the sense that a higher level of financial inclusion can potentially reduce the health gap across gender. Specifically, in the IV-probit and 2SLS-IV estimates (columns 4 and 8), although the coefficients of the female dummy indicates that males with a lower level of financial inclusion are about 0.10 percent (column 4) less likely to report being healthy compared to their counterparts with a higher level of financial inclusion, the coefficients of the interaction term ‘Female*F. Inclusion’ are positive but statistically insignificant, indicating no significant health difference across gender for people with higher levels of financial inclusion. Similar estimates are reported for the OLS and 2SLS-IV on all the hypotheses.

As a robustness check of the health effects of financial inclusion across gender, we provided estimates for gender sub-samples in Table 4 (complete estimates of both Probit and Least Squares are found in Appendices III and IV). Relying on the IV-estimates, these sub-sample estimates provided results that are consistent with that of Table 3. The coefficients remain positive and significant, suggesting that, in both sub-samples, people with a higher level of financial inclusion are more likely to be healthy than their counterparts with lower level. For the female sub-sample, in column (1) those with higher level of financial inclusion are about 2.28 percent more likely to be healthy, while in column (3), those with a higher level of financial inclusion are about 0.82 percent more likely to be healthy for the male sub-sample. It should be noted that a comparison of the coefficients across the two sub-samples does not suggest statistically significant gender differences in the magnitude of the estimated effects. The z values³ provided in Table 4 were all below 1.96, thus, failing to reject the null hypothesis that ${\beta }_{Female}={\beta }_{Male}$, and one cannot conclude that financial inclusion affects females differently than males.

Table 4

Estimates of Equations (1) and (2) using 2SLS-IV for gender Sub-Samples
Dependent Variable: Health Status (healthy = 1, illness = 0)
	Female sub-sample		Male sub-sample		${\beta }_{Female}={\beta }_{Male}$
Independent Variables	(1)	(2)	(3)	(4)	Z values
	IV-Probit	2SLS-IV	IV-Probit	2SLS-IV	IV-Probit	2SLS-IV
	(dy/dx)		(dy/dx)
Fin. Inclusion	2.278***	2.263***	0.817***	0.843***	1.946	1.942
	(0.687)	(0.662)	(0.303)	(0.310)
Under identification test		22.677(0.00)		41.244(0.00)
Hansen J (overid)		0.392(0.531)		0.217(0.641)
Endogeneity test		24.779(0.00)		8.698(0.003)
F-stat		11.492		20.836
District Effect	Yes	Yes	Yes	Yes
Other controls	Yes	Yes	Yes	Yes
Observations	11,101	11,101	11,505	11,505
R-squared		-1.248		-0.193
Wald test of exogeneity	9.32(0.0023)		26.49(0.000)
Robust standard errors in parentheses, * p < 0.01, p < 0.05, * p < 0.1. For the under-identification, Hansen J. (overidentification) and endogeneity tests, we report the test values with p-values in parenthesis. Also, For the exogeneity tests, we reported the chi sqr. test values with p-values in parenthesis Complete estimates are provided in Appendix III (OLS and 2SLS-IV) and IV (Probit and IV-Probit).

In sum, our analysis provided three key findings: (i) there is a gender-health gap in Ghana, where females are less likely to report being healthy than their male counterparts, (ii) financial inclusion has adverse effects on individuals’ health, as people with lower levels of financial inclusion report lower health, and finally (iii) our estimates suggest that financial inclusion may contribute to closing the gender-health gap, a finding that potentially may have significant policy implications.

The link between financial inclusion and the gender-health gap has been under research in empirical literature. Previous literature suggests that financial inclusion promotes investments in health. It is also established in the existing literature that, women report more illnesses, have worse health outcomes, and exhibit higher health care utilization compared to men, despite their higher life expectancy. Financial inclusion is widely argued to help improve people’s income level, and higher income level plays a significant role in maintaining better health outcomes. Incidentally, gender difference in financial inclusion exists where females are largely excluded, and this could be partly responsible for the health difference across gender. As a result, the current study contributes to the literature by exploring the role of financial inclusion on the gender-health gap in Ghana. Specifically, we used micro-level data from the sixth round of the Ghana living standards survey (GLSS), restricted the sample to people between the ages 18 to 75 years and employed various identification strategies to investigate; (i) the gender-health gap using self-reported health measure, (ii) the paper investigates the health effects of FI using multidimensional FI measures, and (iii) the role of FI in the gender health-gap.

The findings suggested a confirmation of the gender-health gap in Ghana. Again, people who are financially included (have higher financial inclusion) are healthier than their less included counterparts. Finally, there was an indication that a higher level of financial inclusion has the potential to reduce the health gap across gender as there was no significant health difference across gender for people with a higher level of financial inclusion. The study concludes that there is a gender health gap which tends to be less favourable to women in Ghana. Financial inclusion has positive health effects which may potentially contribute to closing the gender-health gap through health investment. Thus, policy makers of developing countries should be mindful that policies promoting financial inclusion may help address the gender-health gap in Ghana and other developing countries with similar characteristics. This is possible through improved access to financial services and conducting financial literacy programs. The limitation of this study is largely on the data. The unavailability of panel datasets and alternative health measures in the dataset to help explore the relationship over time and for robustness checks, respectively, are potential limitations future studies can investigate.

Declarations: No funds, grants, or other support was received

Conflicts of interest/Competing interests; No conflict of interest exists, and the authors have no relevant financial or non-financial interests to disclose.

Availability of data; The data that support the findings of this study are openly available in the Ghana Statistical Service (GSS) data repository at https://www.statsghana.gov.gh/gssda tadow nload spage.php

Author Contribution

AGI conceive and produced an initial draft. JIM, NB, and ID reviewed and add their feedback

Abdul-Mumuni, A., and Koomson, I., 2019. Household remittance inflows and child education in Ghana: exploring the gender and locational dimensions. J. Econ. Res. (JER) 24 (2), pp. 197–222.
Allen, F., Demirgüç-Kunt, A., Klapper, L. F., & Martinez Peria, M. S. (2016). The foundations of financial inclusion: Understanding ownership and use of formal accounts. Journal of Financial Intermediation, 27, pp. 1–30.
Alliance for Financial Inclusion, 2015. 2015 Maya declaration report: Commitments into action. Alliance for Financial Inclusion http://www.afi-global.org/sites/default/files/ publications/ maya_report_2015-final.pdf.
Amidu, M., Iddrisu, A.G., Andani, A. and Agyeman, B., 2022. Bank Pricing Behaviour, Monetary Policy and Inclusive Finance in Africa. In The Economics of Banking and Finance in Africa: Developments in Africa’s Financial Systems (pp. 335-369). Cham: Springer International Publishing.
Baland, J.M., Bardhan, P., Das, S., Mookherjee, D. and Sarkar, R., 2007. 'Managing the environmental consequences of growth: forest degradation in the Indian Mid-Himalayas', Indian Policy Forum, 2006/7 (3), pp215-266.
Bank of Ghana, 2016. List of Banks in Ghana, Various Years. Bank of Ghana, Accra, Ghana.
Bruhn, M. and Love, I., 2014. The real impact of improved access to finance: Evidence from Mexico. The Journal of Finance, 69(3), pp.1347-1376.
Bukari, C., Koomson, I., 2020. Adoption of mobile money for healthcare utilization and spending in rural Ghana. Moving from the Millennium to the Sustainable Development Goals. Palgrave Macmillan, Singapore, pp. 37–60.
Case, A. and Paxson, C., 2005. Sex differences in morbidity and mortality. Demography, 42(2), pp.189-214.
Churchill, S.A., Marisetty, V.B., 2019. Financial inclusion and poverty: a tale of forty-five thousand households. Appl. Econ., pp. 1–12
Chun, H., Khang, Y.H., Kim, I.H. and Cho, S.I., 2008. Explaining gender differences in ill-health in South Korea: the roles of socio-structural, psychosocial, and behavioral factors. Social science & medicine, 67(6), pp.988-1001.
Cleary, P.D., 1997. Subjective and objective measures of health: which is better when?. Journal of health services research & policy, 2(1), pp.3-4.
Clogg, C.C., Petkova, E. and Haritou, A., 1995. Statistical methods for comparing regression coefficients between models. American journal of sociology, 100(5), pp.1261-1293.
Danquah, M., Quartey, P., Iddrisu, A.M., 2017. Access to financial services via rural and community banks and poverty reduction in rural households in Ghana. J. Afr. Dev. 19 (2), pp. 67–76.
Demirgüç-Kunt, A., Klapper, L., Singer, D., 2017. Financial Inclusion and Inclusive Growth: A Review of Recent Empirical Evidence. The World Bank.
Demirgüç-Kunt, A., Klapper, L., Singer, D., Ansar, S., Hess, J., 2018. The Global Findex Database 2017: Measuring Financial Inclusion and the Fintech Revolution. The World Bank.
Demirgüç-Kunt, A., Córdova, E.L., Pería, M.S.M. and Woodruff, C., 2011. Remittances and banking sector breadth and depth: Evidence from Mexico. Journal of Development Economics, 95(2), pp.229-241.
Fanta, A.B. and Mutsonziwa, K., 2016. Gender and financial inclusion. Policy research paper, (01), p.1.
Gangadharan, L. and Valenzuela, M.R., 2001. Interrelationships between income, health and the environment: extending the Environmental Kuznets Curve hypothesis. Ecological Economics, 36(3), pp.513-531.
Ghosh, S., 2022. Gender and financial inclusion: does technology make a difference?. Gender, Technology and Development, pp.1-19.
Greene, W.H., 2003. Econometric analysis. Pearson Education India.
Gyasi, R.M., Adam, A.M., Phillips, D.R., 2019. Financial inclusion, health-seeking behavior, and health outcomes among older adults in Ghana. Res. Aging 41 (8), 794–820.
Iddrisu, A.G. and Ebo Turkson, F., 2020. Political business cycles, bank pricing behaviour and financial inclusion in Africa. Cogent Economics & Finance, 8(1), p.1762286.
Iddrisu, A.G., Phimister, E. and Zangelidis, A., forthcoming. The role of Energy-Poverty in the Gender Health Gap in Ghana.
Koomson, I. and Danquah, M., 2021. Financial inclusion and energy poverty: Empirical evidence from Ghana. Energy economics, 94, p.105085.
Koomson, I., Villano, R.A., Hadley, D., 2020a. Effect of financial inclusion on poverty and vulnerability to poverty: evidence using a multidimensional measure of financial inclusion. Soc. Indic. Res. 25 (4), 375–387.
Kuri, P.K., Laha, A., 2011. Financial inclusion and human development in India: an interstate analysis. Indian J. Hum. Dev. 5 (1), 61–77.
Malmusi, D., Artazcoz, L., Benach, J. and Borrell, C., 2012. Perception or real illness? How chronic conditions contribute to gender inequalities in self-rated health. The European Journal of Public Health, 22(6), pp.781-786.
Mestl, H.E.S. and Aunan, K., 2005. The impact of household solid fuel use on population exposure and health in Shanxi province, China. WIT Transactions on Ecology and the Environment, 85.
Mohammed, J.I., Mensah, L. and Gyeke-Dako, A., 2017. Financial inclusion and poverty reduction in Sub-Saharan Africa. African Finance Journal, 19(1), pp.1-22.
Molarius, A. and Janson, S., 2002. Self-rated health, chronic diseases, and symptoms among middle-aged and elderly men and women. Journal of clinical epidemiology, 55(4), pp.364-370.
Mndolwa, F.D. and Alhassan, A.L., 2020. Gender disparities in financial inclusion: Insights from Tanzania. African Development Review, 32(4), pp.578-590.
Ndoya, H.H. and Tsala, C.O., 2021. What drive gender gap in financial inclusion? Evidence from Cameroon. African Development Review, 33(4), pp.674-687.
Takahashi, S., Jang, S.N., Kino, S. and Kawachi, I., 2020. Gender inequalities in poor self-rated health: cross-national comparison of South Korea and Japan. Social Science & Medicine, 252, p.112919.
Verbrugge, L.M., 1989. The twain meet: empirical explanations of sex differences in health and mortality. Journal of health and social behavior, 30(3), pp.282-304.
World Bank, 2017. World Development Indicators Data Base. The World Bank.
Zhang, H., d’Uva, T.B. and Van Doorslaer, E., 2015. The gender health gap in China: A decomposition analysis. Economics & Human Biology, 18, pp.13-26.

The correlation between the independent variables is generally low (< 0.70). The low correlations between the variables suggests less collinearity among them which will not cause estimation issues.
Following the works of (Clogg et al., 1995; Paternoster et al., 1998), the formula is argued to be appropriate for testing for the difference between two regression coefficients.

No competing interests reported.

Appendices.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

The Gender Health Gap in Ghana: Exploring the Role of Financial Inclusion

Status:

Version 1

Abstract

1. Introduction

2. The link between FI and the gender-health gap

3. Model and estimation strategy

3.1 Main model

3.2 Data

3.3 Variable measurement

3.4 Descriptive Statistics

4. Empirical Results

4.1 Empirical Estimations and Discussions

5. Summary, Conclusion and Policy Implications

Declarations

References

Footnotes

Additional Declarations

Supplementary Files

Status:

Version 1

Dependent Variable: Health Status (healthy = 1, illness = 0)
	Female sub-sample		Male sub-sample		\({\beta }_{Female}={\beta }_{Male}\)
Independent Variables	(1)	(2)	(3)	(4)	Z values
	IV-Probit	2SLS-IV	IV-Probit	2SLS-IV	IV-Probit	2SLS-IV
	(dy/dx)		(dy/dx)
Fin. Inclusion	2.278***	2.263***	0.817***	0.843***	1.946	1.942
	(0.687)	(0.662)	(0.303)	(0.310)
Under identification test		22.677(0.00)		41.244(0.00)
Hansen J (overid)		0.392(0.531)		0.217(0.641)
Endogeneity test		24.779(0.00)		8.698(0.003)
F-stat		11.492		20.836
District Effect	Yes	Yes	Yes	Yes
Other controls	Yes	Yes	Yes	Yes
Observations	11,101	11,101	11,505	11,505
R-squared		-1.248		-0.193
Wald test of exogeneity	9.32(0.0023)		26.49(0.000)
Robust standard errors in parentheses, * p < 0.01, p < 0.05, * p < 0.1. For the under-identification, Hansen J. (overidentification) and endogeneity tests, we report the test values with p-values in parenthesis. Also, For the exogeneity tests, we reported the chi sqr. test values with p-values in parenthesis Complete estimates are provided in Appendix III (OLS and 2SLS-IV) and IV (Probit and IV-Probit).