Financial Inclusion and Fertility: An Empirical Study

This article is an edited version of the research paper titled Financial Inclusion and Fertility: An Empirical Study submitted by the author at the Global Governance and Public Health Vulnerabilities in the Global South Conference, jointly organised by the Rethinking Economics India Network (REIN) and the Young Scholars’ Initiative (South Asia Working Group) on October 2, 2021. The conference was a part of the REIN 2021 Annual Event.

The author, Govindapuram Suresh, is a PhD scholar at Indian Institute of Technology Tirupati, Andhra Pradesh. He is a recipient of CBSE Merit Scholarship (for FIVE Years), has cleared UGC NET, got NFSC award (declined), and is currently receiving MoE PhD Fellowship. His area of interest includes development economic, micro, macro, Indian economy, agricultural, rural, urban and socio- economic issues, especially poverty, inequality and discrimination aspects.

The traditional emphasis on the beneficial effects of financial development has increasingly been subjected to scrutiny in the light of the devastating financial crises that the developing world has faced in the recent past in addition to the widening socio-economic inequities in terms of access to financial services that have accompanied financial liberalization. It is being increasingly recognized that the fruits of financial liberalization have not trickled down to the vulnerable sections of society, and instead have served to exacerbate inequalities and deepen economic insecurities for the poor. It is in this context that policymakers and academics across the developing world have increasingly started to emphasize the importance of improving access to financial markets as opposed to merely expanding and developing financial markets. Defined as “a process that ensures the ease of access, availability, and usage of the formal financial system for all members of an economy”, financial inclusion has come to be seen as a crucial tool of social development in many developing countries. Financial inclusion has been shown to have a significant impact on economic growth, reducing poverty and inequality, and on a range of educational and health-related outcomes.

The relationship between financial inclusion and fertility, although an important one, has been overlooked by the existing literature, which is quite surprising for the following reasons. First, fertility rates have always been viewed as important indicators of development ever since Malthus’s theoretical endeavors in the 18th century. It has been widely suggested that changes in fertility rates can have significant effects on economic growth, savings rates, technological progress, and women’s empowerment. Secondly, the lacuna is surprising because there is vast literature on the relationship between financial development and fertility rates which has emphasized the important role that well-developed financial markets can have on fertility rates. Given that financial development and financial inclusion are distinct concepts, quite contradictory to each other, the relationship between financial inclusion and fertility is crucial and one that needs to be examined carefully.

The term financial development refers to the process by which financial markets reduce transaction costs, diversify risks, improve economic coordination, enable better allocation of resources, and thus enable economies to save and innovate at faster rates. Seen in this manner there can be little doubt that financial development is crucial for making financial services and institutions more accessible to individuals. Yet in the context of developing countries where massive social, cultural, and legal barriers prevent individuals from partaking in economic life, very often individuals from marginalized backgrounds are excluded from financial markets. Thus, it has been widely noted that gender, caste, race, ethnicity, and other social markers are often correlated with economic and social outcomes. Lack of access to credit and the general non-availability of financial services to underprivileged sections compounds existing social, cultural, political, legal exclusions. Financially excluded people face different forms of social and economic discrimination at different levels in society and such exclusion can potentially co-exist with high levels of financial development. Thus, we must understand the differences between financial development and financial inclusion. The former focuses on the availability of quality financial services, appropriate institutions, and well-developed markets, while the latter is focused entirely on the issue of accessibility. It is within this context that there have been several attempts to bring to the forefront of policymaking the issue of financial inclusion, as best exemplified by the Grameen Bank experiments in Bangladesh by Nobel laureate Mohammad Yunus (2009).

This study aims to address the missing link between financial inclusion and fertility. Theoretically, the relationship between financial inclusion and fertility (as addressed by the miniscule existing literature on it) is ambiguous. Traditionally the standard “old age security” hypothesis which suggests that households’ decision regarding family size is motivated by the future income that children earn and share with their parents during their old age, predicts a negative relationship between financial development and fertility.

This study employs a panel of 120 countries between 2004 and 2018[1] to explore the relationship between financial inclusion and fertility rates. The period is chosen based on the data availability of our financial inclusion variables that we explain in more detail below. Moreover, to incorporate the multi-faceted and complex linkages between the two, the empirical specification departs from the usual linear specifications that have been commonplace within the literature and explore potential non-monotonicities in the finance-fertility relationship to better capture the dynamics between the two variables. To do so the paper uses both fixed-effect with a quadratic term as given below in equation (1) and semi-parametric regression as given in equation (2):

Yi,t = β0 + β1IFI i,t + β2IFI2i,t + β3X i,t + ∈ i,t (1)

Yi,t = β0 + m(IFI i,t) + β1X i,t + ∈ i,t (2)

where Y is the Fertility rate. X is the vector of explanatory variables affecting the Fertility rate. IFI is the financial inclusion index variable, IFI2 is the squared term to show the possibility of non-linearity. Further, ‘i’ indicates country, ‘t’ is for time periods, and ‘∈’ is an error term.

The dependent variable of interest is the fertility rate (FR) which is defined as the number of total births per woman. The control variables include infant mortality rate (IMR) which refers to the number of infant deaths per 1000 live births. Improvements in infant mortality can reduce fertility rates but in general, the relationship is ambiguous because, levels of fertility may be depending on levels of mortality (Basso et al. 2014, Zakaria et al 2017; Lai and Yip 2019). We also take into account per capita income (GDPPC) to capture the effect of income and development levels on fertility (Becker 1960, Filoso and Papagni, 2015; Zakaria et al., 2016; Basso et al. 2014). Previous literature has noted the importance of education in the context of fertility. Higher levels of education generally result in lower fertility rates, since parents are interested in the quality of their children rather than the quantity of education (Becker 1960). The education component is measured by the secondary school enrolment rate (EDN) (Filoso and Papagni, 2015; Zakaria et al., 2016; Basso et al. 2014). We further take into account the extent of urbanization as measured by the urban population as a percentage of the total population (Urban) because previous studies have shown that urbanization increases the cost of raising children (Galloway et al. 1998). Another feature that affects the opportunity costs of childbearing is the age-dependency (AgeDe) which we include in our control variable list (Zakaria et al., 2016). We finally also add inflation (Infl) and trade openness (TRADE), as measures of macroeconomic stability (Filoso and Papagni, 2015). To smooth out fluctuations and to account for missing data, we consider three-year averages for all the variables. All variables are transformed in logarithms. These variables are sourced from World Bank’s World Development Indicators (WDI).

Following the literature on financial inclusion2, we use data on the number of bank accounts per 1000 population (Depositors), Number of bank Branches per 100000 population (BRANCHES), and Number of ATMs per 100000 per population from World Banks’ World Development Indicators (WDI). We also take usage indicators of volume of credit and deposits relative to GDP from the IMF Financial Access Survey (IMF FAS). This study uses these indicators of financial inclusion to compute a financial inclusion index using Principal component analysis (PCA) as has been done in the literature previously as well (Sarma 2008; Sarma and Pais 2011; Honohan 2008; Cámara and Tuesta, 2014; Park and Mercado 2015; Chakravarty and Pal 2013; Demirguc-Kunt and Klapper 2012). PCA optimizes the combinations of different sub-indices and helps identify those combinations that give the most variance. It transforms a large set of variables into a smaller one that still contains most of the information in the large set.

For getting efficient estimators (for equation 1), we used Fixed Effect (FE) model. It accounts by creating dummies for the countries and allows intercept to shift in each country. That is, differences between individuals can be accommodated from different intercepts. Further, the FE estimator takes into account heterogeneity between countries and time-invariant unobserved factors influencing fertility, also possibly correlated to our selected regressors. The time-invariant characters are unique to the individual countries and should not be correlated with other individual country characteristics. FE model controls for time-invariant characteristics. To confirm the use of fixed effects, we run a Hausman test where the null hypothesis is that the preferred model is random effects vs. the alternative the fixed effects. The fixed effect model is accordingly considered.

We controlled the problems of heteroscedasticity and auto-correlation in the FE model. The quadratic terms in equation 1 capture potential non-linearities in the relationship being analyzed. To statistically test it, the SLM test for the existence of a U-shape relationship is employed as a confirmatory test. The SLM test explores the monotonic or non-monotonic relationship between financial inclusion and fertility. These tests enable us to understand the marginal impact of financial inclusion at a certain point and after a certain point where financial inclusion no longer contributes to fertility or may have a positive outcome after reaching the threshold. Most of the previous techniques focus on the significance of the quadratic term and the range of extremum points. Lind and Mehlum (2010) used Sasabuchi’s (1980) likelihood ratio techniques to prove a stronger test to identify the presence of a U-shaped relationship between independent and dependent variables.

Equation 1 imposes a functional form on the relationship between financial inclusion and fertility but the regression specification in equation 2 relaxes this assumption by utilizing semi-parametric specification, by following Baltagi and Li (2002), which allows the effect of the financial inclusion index to be modeled by an unrestricted functional form. Where Y is a dependent variable fertility rate, X is a vector of control variables and Financial Inclusion Index enters non-parametrically to see the non-linear relationship. The semi-parametric model does not impose any functional restriction on the relationship between financial inclusion and fertility rates and helps to capture the existence of non-linearities in the data (Verardi 2013; Baltagi and Li 2002; Zouaoui et al. 2018).

Table 1 gives results for the fixed-effect model for a sample of 120 countries. Models 1–6 add an independent variable to the model. All models give evidence of a non-linear relationship between fertility and financial inclusion (Lai and Yip 2019). Initially, the financial inclusion index (IFI) is inverse with coefficients between -0.45 to -0.50 and significantly (at 1 percent) related to fertility, and then the squared term of IFI shows positive with a coefficient between 0.45 to 0.64 and significant at 1 percent. It suggests that the relationship between fertility and financial inclusion is non-linear and it’s a U-shaped relationship. Hence, there is a negative relationship between fertility and financial inclusion (Zakaria et al .2017) and later it shows a positive relationship. It shows the income effect and substitution effect between income and fertility (Becker 1960). The levels of education significantly (at 1 percent) influence fertility levels, as education increases, fertility is decreasing in all the models, with the coefficients between -0.10 to -0.15. This means that educated couples will have a lesser number of children as compared with couples having lower education levels, these results are similar to previous studies (Filoso and Papagni, 2015; Zakaria et al., 2016; Basso et al.2014). The GDP per capita income and infant mortality rates are not statistically significant and are not robust. As female labor increases in the workforce, it reduces fertility rates, since the couples are interested in working more than having kids. The rate of urbanization is inversely related to fertility rates, as urbanization increases, fertility decreases, since, in Urban areas, couples prefer working than having kids, since the cost of living is more in urban areas than in rural areas (Zakaria et al., 2016; Lai and Yip 2019). Infant mortality doesn’t have a significant relationship in our analysis as age-dependency increases, fertility also increases. Other macroeconomic indicators: trade openness and inflation significantly influence levels of fertility and the results were similar to Filoso and Papagni, 2015. The age-dependency, trade openness, and inflation rates were used to check the robustness of the models. In these results, GDP per capita and female labor force participation rates are not robust.

Figure 1 shows a relationship between fertility and the index of financial inclusion for a sample of all 120 countries. This gives evidence of the non-linear relationship between financial inclusion and fertility rates, got by employing the semi-parametric analysis. The shape of the curve is U-shape. That is, there is a U-shaped relationship between fertility and financial inclusion. These results are like table 1.

Similarly explores, table 2 elaborates on results for a sample of 70 Low- and middle-income countries. In the first case index of financial inclusion is inversely and significantly related to fertility and the squared term shows a positive and significant relation with fertility. For IFI, the coefficients range from -0.54 to -0.63, and the squared term is between 0.42 to 0.64. and these are significant at the 1 percent level. This relationship indicates that there is a U-shaped relationship between fertility and financial inclusion. Furthermore, these results were tested with the SLM test, and we found that U- test results were significant in all the models, this indicates that there exists a U-shaped relationship between fertility and financial inclusion. Other macroeconomic variables such as levels of education are inversely related to fertility and its significant at 1 percent in all the models, except in model 6. The GDP per-capita is significant in models 3 and 6 and in Figure other models, it’s not significant. Infant mortality rates (IMR) and female labor force participation are not robust. further, levels of urbanization are negatively (-0.32 and -0.16) affecting fertility at 1% and 5% in models 5 and 6, respectively. Age -dependency, trade openness, and inflation rates are robust and significantly influence levels of fertility.

Figure 2 shows a relationship between fertility and the index of financial inclusion for a sample of 70 low- and middle-income countries. This gives evidence of the non-linear relationship between financial inclusion and fertility rates, got by employing the semi-parametric analysis. The shape of the curve is U-shape. That is, there is a U- shaped relationship between fertility and financial inclusion. These results are similar to table 2. Both figure 1 and 2 shows the U-shape relationship between fertility and financial inclusion. The U-shape is steeper for all the sample countries (Figure 1) and it is a little flatter for low- and middle-income countries (in figure 2).

To the presence of a U-shaped relationship between financial inclusion and fertility (i.e, IFI, IFI2, and fertility). In both Table 1 and Table 2, U-Test results significantly in all the models which show that the relationship between fertility and financial inclusion is in a U-shape. The p-values suggest that we reject the null hypothesis and accept the alternative hypothesis that there is a U-shaped relationship between two variables.

Semiparametric results are presented in Table 3, these results are similar to results in Tables 1 and 2. The levels of education are negatively and significantly affecting fertility. For Low- and middle-income countries it is -0.04 and for all sample countries, it is 0.0.7. Urbanization is inversely related to fertility (-1.8 and -1.9) and it is significant at a 5% level of significance for both the sample countries. Age-dependency, trade openness, and inflation are robust and significantly influence fertility. Further, GDP per-capita and infant mortality are not robust. The semiparametric results are similar to panel fixed-effects models.

Thus we can conclude from this study that fertility plays an important role in the study of demography and economics. We find a U-shaped relationship between financial inclusion and fertility for all 120 sample countries and 70 low- and middle-income countries and our estimated results are consistent with previous studies.

[1] We consider only those country which have data for at least 80 percent of the period, and other countries dropped from analysis.