- Email: [email protected]
Usage of formal financial services in India: Demand barriers or supply constraints?
Accepted Manuscript Usage of formal financial services in India: Demand barriers or supply constraints? Abhishek Kumar, Rama Pal, Rupayan Pal PII:
S0264-9993(17)31753-4
DOI:
https://doi.org/10.1016/j.econmod.2018.11.010
Reference:
ECMODE 4766
To appear in:
Economic Modelling
Received Date: 27 November 2017 Revised Date:
11 November 2018
Accepted Date: 11 November 2018
Please cite this article as: Kumar, A., Pal, R., Pal, R., Usage of formal financial services in India: Demand barriers or supply constraints?, Economic Modelling (2018), doi: https://doi.org/10.1016/ j.econmod.2018.11.010. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Usage of Formal Financial Services in India: Demand Barriers or Supply Constraints?
Abhishek Kumar is PhD student at Indira Gandhi Institute of Development Research (IGIDR), Mumbai. His research interests are in macroeconomics and applied microeconomics and Bayesian methods.
SC
RI PT
Dr. Rama Pal is Assistant Professor of Economics at the Department of Humanities and Social Sciences, IIT Bombay. Her primary research area is Health Economics. Currently, she is working on inclusive growth and development, financial inclusion, migration, health inequality and child labour. International academic journals such as Journal of International Development, Oxford Development Studies, European Journal of Development Research, International Journal of Health Care Finance and Economics, and others have published her research papers.
AC C
EP
TE D
M AN U
Dr. Rupayan Pal, Professor of Economics at IGIDR, India, is an applied micro- economist specializing in the area of strategic interactions among economic agents. His theoretical research has dealt with issues including pricing of network goods, R&D, privatization, environmental regulation, union-oligopoly bargaining and tax competition. He has also researched on some of the policy relevant empirical issues, such as financial inclusion, unionization and tax efficiency. International academic journals such as Mathematical Social Sciences, Journal of Public Economic Theory, Regional Science & Urban Economics, Journal of Policy Modeling, Resource and Energy Economics, Economic Modelling, and others have published his research papers.
ACCEPTED MANUSCRIPT
Usage of Formal Financial Services in India:
November 10, 2018 Abstract:
RI PT
Demand Barriers or Supply Constraints?
SC
Estimating the role of both demand-side and supply-side factors in financial inclusion and its distribution is important for policy making. However, existing literature has primarily focused on
M AN U
supply-side factors. In this context, this paper estimates relative importance of removing demand-side barriers and eliminating supply constraints to enhance financial inclusion in India. It also measures the extent of concentration of usage of formal financial services among richer households. Results suggest that, while availability of banking services has a significant positive effect on usage of formal financial services, its contribution in inducing households to use formal financial services is very small compared to the contribution of factors, such as education,
TE D
income, employment status, gender and social norms, that influence the demand for formal financial services. It highlights the importance of placing greater emphasis on addressing demand-side barriers, rather than on improving physical availability of banking services, to
EP
promote financial inclusion in India.
Key words: Financial inclusion; income-related inequality; demand-side barriers; supply
AC C
constraints; India
JEL classification: O16; O17; D39; G21; O53
ACCEPTED MANUSCRIPT
1. Introduction After the financial crisis, there is a renewed interest in understanding financial development and its effects on economic growth and inequality. Literature provides evidence of positive effect of
RI PT
financial development1 and not necessarily of financial liberalization on growth. Studies find that financial liberalization may not have a direct impact on growth (Demetriades and Luintel, 1996 and 1997) and inequality (Rajan and Zingales, 2003 and 2004; and Ang 2010). However, financial liberalization encourages financial development (Demetriades and Luintel, 1996 and 1997) which in turn facilitates growth and reduces inequality. In contrast, Agnello et al. (2012)
SC
report evidence in support of financial reforms using a panel of 62 countries and claims that financial reforms reduce inequality.2 Cecchetti and Kharroubi (2012) and Chong et al (2016)
M AN U
report a non-linear relation between financial development and growth and suggest that beyond a certain level financial development hurts growth. Further, it is argued that effective financial intermediation plays crucial role in moving households out of poverty – indirectly by stimulating growth (trickle down) and directly by providing savings and credit services to the poor (Burgess and Pande 2005; Kochar 2011; Jeanneney and Kpodar, 2011).
TE D
While the empirical literature on finance and development largely focuses on depth of financial system, the distribution of access to formal financial services has drawn attention of the researchers only recently (Beck et al 2007). Studies in this strand of literature show that easy access to credit reduces poverty (Basu and Mallick, 2008). Similarly, Banerji et al. (2014) report
EP
positive effect of access to Kisan Credit Card on agricultural and infrastructure output and thus, on per capita income of districts in Bihar, India. Using firm level data from 80 developed and
AC C
developing countries around the world, Ayyagari et al. (2008) demonstrate that lack of access to finance restricts the growth of firms significantly. Further, lack of access to adequate finance disproportionately hurts smaller firms (Beck et al. 2005). While studies have mainly focused on impact of greater availability of financial services on growth and productivity, Ahamed and 1
Usually measured with financial deepening indicators such as bank deposit liabilities to GDP ratio (Levine, 2005).
2
Agnello et al. (2012) use the measure of financial liberalization given by Abiad and Mody (2005) and Abiad et al. (2010). This measure is sum of six indicators of financial liberalization, namely, credit controls, interest rate controls, entry barriers, regulations, privatization and international transactions.
1
ACCEPTED MANUSCRIPT
Mallick (2017) extend the discussion and shows that higher availability is also beneficial for the bank stability. Thus, overall there is consensus regarding positive effects of availability of financial services on productivity, growth and equality.
RI PT
Given this positive relation between access to financial services and economic wellbeing, developing countries like India have started giving importance to financial inclusion. The Reserve Bank of India (RBI) defines the financial inclusion as ‘a process of ensuring access to appropriate financial products and services needed by all sections of the society in general and
SC
vulnerable groups such as weaker sections and low income groups in particular at an affordable cost in a fair and transparent manner by mainstream institutional players’ (RBI, 2014). In order
M AN U
to achieve the objectives of financial inclusion, it is important to first measure the extent of financial inclusion. Most of the existing studies consider availability/access to finance as a measure of financial inclusion primarily due to dearth of adequate data on actual use of financial services by households and firms (Honohan, 2008; Bect et al. 2009; and Jeanneney and Kpodar, 2011). However, these outreach indicators of financial inclusion ignore other demand-side factors that may affect actual use of formal financial services such as socio-economic
TE D
background of households, and cost of using formal financial services. One may not use formal financial services, even if there is access, because of low reservation cost of informal financial services and/or high relative prices of financial services compared to the prices of other goods (Kochar 1997, Claessens 2006). Prevailing social/group norms might also play some role to this
EP
effect. Thus, availability of formal financial services is a necessary, but not sufficient condition for use of formal financial services. Only those who use formal financial services can be referred
AC C
as financially included, rest are financially excluded3. Thus, it is more appropriate to consider the data on use of formal financial services, not the mere access to finance.
3
We note here that, in theory, sometimes distinction is being made between two types of financial exclusion. The set of economic agents/households who have access but do not use formal financial services are referred as ‘voluntarily excluded’, and those who do not have access to formal financial services are refereed as ‘involuntarily excluded’. However, given the unavailability of individual/household level data on both use and access, it is not possible to make such distinctions operational.
2
ACCEPTED MANUSCRIPT
In this context, the present paper aims at measuring financial inclusion for India based on usage of formal financial services by Indian households. For this purpose, we use a large-scale household level data for India. We define household to be financially included if it uses either formal credit services or formal savings services or both. Not only formal credit but formal
RI PT
saving services have significant impact on long term asset growth in an economy (Kaboski and Townsend 2005; Ashraf et al 2006). Moreover, access to formal savings services enables the poor to make productive investment, to be less vulnerable to shocks and to smooth consumption expenditure (Dupas and Robinson 2009). To understand the extent of financial inclusion in India,
SC
we examine proportion of households using various financial services and find out how many are left out of the purview of formal financial services. Moreover, it is also required to understand
M AN U
the distribution of such usage across income groups to understand the association of financial inclusion with economic status of household. The present paper attempts to understand both the extent of usage to formal financial services by Indian households and inequality in its distribution across income group. This inequality in financial inclusion is measured using the concentration index and is termed as income-related inequality of financial inclusion. Keeping in mind diversity of the country, we explore the variation across states and sectors, in the extent of
TE D
usage and in income related inequality of financial inclusion. We further examine the factors affecting the household’s usage to formal financial services. Here, we consider both the household level factors as well as the state level supply side factors, i.e., availability of banks, that may affect financial inclusion. We use the Shapley decomposition method (Shorrocks, 1999
EP
and 2013), to understand the relative importance of these demand and supply-side factors in
AC C
determining financial service usage.
The choice of India for the purpose of this analysis rests on several considerations. First, India is an emerging economy that has experienced a rapid economic growth rate since the initiation of major economic reform in 1991. However, income inequality has widened over time and still a large segment of the population lives in poverty (Datt and Ravallion 2002; Deaton and Drèze 2002; Dev and Ravi 2007; Suryanarayana 2008; Cain et al 2010). Therefore, a careful examination of the extent of financial inclusion across income groups in countries like India is of paramount importance. Second, India’s federal structure allows states with partial policy 3
ACCEPTED MANUSCRIPT
autonomy and there is diverse pattern of growth and development across states, which makes it more interesting to examine the issue of financial inclusion in India at the sub-national level. Third, India is one of the few emerging economies for which large scale representative survey
RI PT
data on use of financial services at household level is available.
The empirical findings show low level of financial inclusion in India, particularly among the poor households. The extent of financial inclusion as well as income-related inequality in financial inclusion varies across the states and sectors. Interestingly, while the percentage of
SC
financially included households is lower in the rural sector, the distribution of financially included households is more evenly distributed as compared to the urban sector. It might be
M AN U
because social norms affect rural households’ decision making differently from their urban counterparts. These findings suggest that the different policy measures are warranted in the rural and urban sector. While in the rural sector, higher importance should be on increasing overall access to the financial services, in the urban sector, policy measures should concentrate on improving the access to the poor. Further, while financial inclusion is found to be higher in states with higher average per capita consumption expenditure, neither average per capita consumption
TE D
expenditure nor inequality in consumption expenditure is correlated with income-related inequality in financial inclusion at the state level.
Household-level regression analysis shows that availability of banking services is positively
EP
associated with financial inclusion, along with higher impact for the poor households. Moreover, the association of availability of banking services with the propensity to use financial services by
AC C
a household is stronger than that of other state level factors. Apart from the availability of banking services, most of the household characteristics have significant impact on the probability of financial inclusion. Household characteristics will help us in identifying vulnerable households who are likely to be excluded and we find that the poor, labourers and households belonging to the socially deprived classes are more likely to be financially excluded compared to their respective counterparts in both the rural and urban sectors. Further, self-employed households from the urban areas are less likely to be financially included as compared to the salaried households. This result indicates that small business families in the urban areas are more 4
ACCEPTED MANUSCRIPT
likely to rely on informal sources to meet their need of financial services. It calls for policies to reduce such dependency of small business on informal financial services in order to create a level playing field, which is necessary for employment generation and growth. Moreover, we find discrimination based on gender and social class in the rural areas where households with
RI PT
female head and belonging to socially deprived classes are less likely to be financially included. This discrimination might be result of the overall discrimination that is observed in Indian society (particularly in the rural areas) against these vulnerable groups. Special policies are required to ensure equal opportunity of accessing formal financial services for these vulnerable
SC
groups.
M AN U
Interestingly, while both demand-side barriers and supply constraints hinder the process of financial inclusion, results of Shapley decomposition analysis reveal that demand-side factors play a much greater role compared to supply side factors in inducing households to use formal financial services in India. It points out the importance of placing greater emphasis on removing demand-side barriers than on increasing physical availability of banks in order to foster effective
TE D
financial inclusion in India.
The rest of the paper is organized as follows. The next section describes methodologies to estimate income-related inequality of financial inclusion and presents the econometric model. Section 3 describes the data, variables and steps of estimation. Results are presented in Section 4.
EP
Section 4.1 plots the concentration curves of financial inclusion, estimates concentration index of financial inclusion for sub-national regions in India and analyzes its correlation with the level of
AC C
financial inclusion and other developmental indicators. Section 4.2 discusses the findings from the econometric analysis of the household’s propensity to be financially included and provides estimates of relative importance of removing demand-side barriers and eliminating supply constraints in order to induce households to use formal financial services. Section 5 concludes.
5
ACCEPTED MANUSCRIPT
2. Methodology 2.1 Measurement of Degree of Financial Inclusion Income related inequality in financial inclusion can be appropriately measured using the concentration curves and concentration indices. The concentration curve is the generalized
RI PT
Lorenz curve (Kakwani, 1977).4 In the present case, the concentration curve plots cumulative percentage of households that use formal financial services against cumulative percentage of households arranged according to economic status of household. We consider monthly per capita expenditure (MPCE) as the proxy for economic status. The distance between line of
SC
equality, shown by the 45º line, and the concentration curve shows the inequality. On the other hand, the concentration index (CI) measures the degree of this inequality. The CI is a bivariate measure that quantifies inequality in use of formal financial services across the distribution of
M AN U
households based on their economic status. Let x be the variable representing economic status, i.e. MPCE, of household and y = g(x) be the utilization of formal financial services by the household with MPCE x, where the function y = g(x) is defined as follows.
TE D
1, if the household uses at least one of the formal financial services y = g (x) = 0, otherwise The concentration index of financial inclusion (C) is given by:
2 = − 1 … … … . (1)
EP
where is binary 0 and 1 as explained above, µ = E( y) is the average financial inclusion, is
AC C
the number of household and is fractional rank. Wagstaff (2005) shows that when y i is binary,
1 the minimum value and the maximum value of the concentration index, C, are ( µ − 1 + ) and n 1 (1 − µ + ) , respectively, (see appendix at the end for detailed derivation). For large samples, as n in the present case, the (1/n) term tends to zero, thus, the lower and upper bounds of C can be
4
Initial application of the concentration curves may be found in Mahalanobis (1960).
6
ACCEPTED MANUSCRIPT
considered as ( µ − 1) and (1 − µ ) , respectively. It is clear that the bounds of concentration index shrinks as the average value increases. It implies that the concentration index for financial inclusion may turn out to be low, if the average number of financially included households is high, even though inequality in financial inclusion is high. So, to meaningfully compare the
RI PT
estimates of the concentration index for different regions, it is suggested to divide the original concentration index by (1 − µ ) (Wagstaff, 2005, 2009). The modified concentration index ( CW ) can be written as follows.
SC
2 − 1 … … … . (2) = (1 − )
M AN U
It is evident that, for any value of µ , we have − 1 ≤ CW ≤ 1 . Therefore, for the purpose of the present analysis, we consider the formulation of the concentration index as given by (2).5
Note that lower absolute value of the modified concentration index, CW , of financial inclusion indicates lesser extent of income related inequality in financial inclusion. And, negative
TE D
(positive) value of the concentration index of financial inclusion implies that the use of formal financial services is concentrated among the poorer (richer) households.
2.2 Econometric Model
EP
Next, we want to understand the factors that determine use of formal financial services and examine whether demand- or supply-side factors have more impact on utilization. For this
AC C
purpose, we estimate the logit model for utilization of formal financial services. The reduced form model depicting the use of formal financial services by households can be written as follows.
Yi * = α
+
Z i/ β
+
u i , i = 1, 2…n;
(3)
eu i yi = I (Yi > 0), F (ui ) = . 1 + eui *
5
We note here that there is a growing literature on appropriate formulation of concentration index in the case of binary variable (Erreygers, 2009; 2009b; Wagstaff, 2009; Brekke and Kverndokk, 2012).
7
ACCEPTED MANUSCRIPT
The net benefit of household i from using formal financial services is captured by the latent variable Yi * , and whether household i uses formal financial service or not is indicated by the dummy variable i. I (Yi * > 0) is an indicator function taking the value one if Yi * > 0 and the
RI PT
value zero otherwise. F (u i ) is the cumulative distribution function (CDF) of the error term u i , which has logistic distribution. Z i is the vector of exogenous explanatory variables (state-level factors and household characteristics), α is the constant term, and β is the vector of unknown
SC
parameters to be estimated. Clearly, the probability that household i uses formal financial
eα +Zi β /
pi = E( yi = 1| Zi ) =
1 + eα +Zi β /
.
M AN U
services is
We estimate model (3), considering alternative sets of explanatory variables, by maximum likelihood method with robust standard errors, and compute marginal effects of explanatory variables,
∂ pi ∂ xj
eα + Zi β /
=
(1 + e α + Zi β ) 2 /
β j (j = 1, 2...K), for continuous variables, and discrete changes
TE D
for dummy variables from zero to one.6
EP
3. Data, Variables and Estimation 3.1. Data
We use the household level data from All India Debt and Investment Survey (AIDIS) for the
AC C
year 2002-03. This dataset was collected as a part of the 59th round of the National Sample Survey (NSS), conducted by the Central Statistical Organization, Ministry of Statistics and Program Implementation, Government of India. This representative survey data for India provides micro-level information on households’ savings and borrowings along with socioeconomic characteristics of households.7 Based on availability of data, we estimate income-
6
As indicated by the methodology of the survey, we use sampling weights to compute descriptive statistics and to estimate the model. 7 For more information on the dataset see GoI (2005).
8
ACCEPTED MANUSCRIPT
related inequality in financial inclusion for all the states and the union territories (UTs) in India, but for the econometric analysis we leave out the UTs. We note here that the 29 states together cover more than 99 (98) percent of land-area (population) of India.8
RI PT
3.2. Dependent Variable
A household is termed financially included if it uses any of the formal financial product else financially excluded.9 This categorical variable is dependent variable in regression analysis. It may be noted that the households that do not use any of the formal financial services, can be
SC
referred as completely financially excluded. However, households in the first group are not necessarily fully financially included, since the need for formal financial services of some
M AN U
households in the first group may be met only partially by the formal financial services10. For simplicity of exposition, in this paper, we refer the first group of households as financially included households. Thus, the variable of interest in the present analysis is an indicator variable taking value one if a household use at least one of the formal financial services and the value zero otherwise. We report the percentage of households using at least one of the financial services across states &UTs and sectors, rural and urban, in Table 1 & 2. It is evident that the
8
TE D
extent of financial inclusion is alarmingly low in both rural (34.9%) and urban (49.2%) India.
9
EP
Latest All India Debt and Investment Survey (AIDIS) from 70th round doesn’t contain data on consumption expenditure which makes it unsuitable for the exercise being done in this paper. Therefore, we have tried to substantiate our two main results i.e. poor people are less likely to be financially included and banking presence increases financial inclusion especially among poor people using the Global Financial Inclusion (Global Findex) Database 2014. It is clear from Table 11 that poor people are less likely to have account in any financial institution and more of poor people complain about less supply of financial services.
AC C
The data set provides detailed information on sources of borrowings and savings instruments used by the households. Out of fifteen different sources of borrowings provided in the data set, we consider the institutional sources of borrowings, namely, government, cooperative society, commercial bank, insurance, provident fund, financial corporation, financial company and other institutional agencies, to estimate utilization of formal credit services. Similarly, households’ savings with the institutional agencies, namely, government or reserve bank of India certificates and bonds, deposits in post office, cooperative society/bank, commercial bank and non-banking company, and insurance premium, annuity certificates and provident fund, are considered as formal savings. If a household reports either borrowing from at least one institutional source or savings with at least one institutional agency or both, we categorize that household in the first group – ‘households utilizing formal financial services’.
10
We also note here that, though there are differences between access and use of formal financial services, most of the existing studies have considered these two synonymously, primarily due to dearth of adequate data on actual use of financial services by households and firms (Honohan 2008, Beck et al 2009, Jeanneney and Kpodar 2011).
9
ACCEPTED MANUSCRIPT
Moreover, there is substantial variation (range is 55%) in terms of usage of formal financial services across states & UTs in India.
3.3. Explanatory Variables
RI PT
The determinants of the household use of formal financial services can be classified broadly into two categories, namely, household characteristics and state level factors, as indicated in Table 3. Definitions of explanatory variables are detailed in the Appendix. Table 4 presents the summary statistics of the variables used in the present analysis. In order to alienate the effects of household
SC
income clearly, we consider MPCE quintile groups, and include dummy variables for the last four groups, Consumption2 – Consumption5, in the set of explanatory variables, while
M AN U
Consumption1 is the bottom quintile.11 Apart from the income, employment type (i.e. source of earnings) of a household might also affect its probability to use formal financial services. We can categorize the households according to the type of primary employment, which is the main source of income, into three groups, namely, salaried, self-employment and casual labor. We consider the dummy variables Self Employed and Laborer as explanatory variables (Salaried serves as the base category) in the regression. Summary statistics shows that 48.8 percent
TE D
households earn their livelihood through self-employment, which is the largest in size out of three broad categories (Table 4). In contrast, only 21 percent households’ principal earnings come from salaried employment. The probability to use formal financial services is expected to be lowest (highest), if a household is primarily dependent on casual labor supply (salaried
EP
employment).
AC C
In order to control for possible effect of household’s education on use of formal financial services, we construct the variables Illiterate, Literate, Primary, Middle and Secondary and above, based on highest level of education of household head. Summary statistics reveals that in Indian states only 20 percent households’ heads were having secondary or higher level of education in 2002-03; whereas 40.8 percent households have illiterate head. It is also expected that female headed households are more vulnerable and are less likely to be financially included, 11
Our results are not sensitive to the consideration of alternative categorizations (e.g. deciles) of household according to MPCE or to the use of MPCE directly.
10
ACCEPTED MANUSCRIPT
compared to the male headed households. Therefore, we expect to have negative effect of the variable Female on probability to use formal financial services. In order to control for the size of the households, we consider the variable Household Size as an explanatory variable.
RI PT
Households belonging to socially disadvantageous communities are expected to have lower propensity to be financially included, because of possible discrimination against them in the formal sector and/or due to societal culture and the extent of discrimination could be different in rural and urban areas. In India socially disadvantageous communities are identified on the basis
SC
of castes, namely, scheduled castes (SC); scheduled tribes (ST); and other backward classes (OBC). Remaining households are considered as general caste (General). So we use dummy
M AN U
variables SC, ST and OBC as explanatory variables in the regression to assess the impact of social groups. To control for the possible effect of the place of residence, we consider the dummy variable Rural while estimating the model using full sample, and estimate the model separately for rural and urban sectors.
Moreover, we link the household-level data to state-level measures of (a) availability of banking
TE D
services, (b) level of infrastructural facilities and (c) urbanization. Data on banking services is collated from Basic Statistical Returns of Scheduled Commercial Banks, published by the Reserve Bank of India. Data on per capita net state domestic product (NSDP) come from CSO. The source of data on state-wise tele-density is the annual report of the Department of
EP
Telecommunications, Ministry of Communications & Information Technology, Government of India. state-wise population, land-area and urbanization data come from Census of India for the
AC C
year 2001.
We consider the variable Bank Penetration, which is defined as the number of bank offices per lakh population, as a measure of availability of banking services. In order to avoid possible endogeneity problem, we use one year lagged values of Bank Penetration in the regression. Higher Bank Penetration implies greater access to formal financial services. Therefore, it is expected that Bank Penetration would be positively associated with the household use of formal financial services. For instance, Beck et al. (2007) demonstrate that indicators of outreach of 11
ACCEPTED MANUSCRIPT
banking services are positively associated with hard-to-collect micro-level statistics on household use of formal financial service. However, to the best of our knowledge, the issue of whether such positive association is uniform across different income groups of households or not has not received much attention in the literature so far. It is often argued that there is need to
RI PT
enhance the supply of banking services in order to promote financial inclusion particularly among the poor. However, for such policies to be effective, availability of banking services should have differential impact on use of formal financial services by the households across income groups, with positive bias to the relatively poor households. In order to examine whether
SC
that is indeed true or not, we introduce interaction variables (Bank Penetration)* Consumptionj, j = 2, 3, 4, and 5, in the regression. Table 4 shows that in the sample states the average number
M AN U
of bank offices per lakh population was 7.79, which seems to be quite low. Also, there was considerable variation in terms of Bank Penetration across states in India. Bank Penetration was lowest (3.57) in Nagaland and highest (24.64) in Goa during the period of study.
It may be argued that the variable Bank Penetration controls for only one aspect of availability of banking services, namely, demographic penetration of bank offices. And, thus, geographic bank
TE D
penetration, which is defined as number of bank offices per thousand square kilometer area, percapita availability of deposit services and per-capita availability of credit services should also be considered as indicators of outreach of banking services, which measures different aspects of availability of banking services. To check the robustness of our results we consider two
EP
alternative variables, (a) Outreach Index and (b) Outreach Index PCA, in place of Bank Penetration in separate regressions. Outreach Index and Outreach Index PCA are two indices of
AC C
banking sector’s outreach. Each of these two indices combine the above mentioned four indicators, including demographic penetration, into a single measure of banking sector’s outreach by employing (i) the method of Chakravarty and Pal (2013) and (ii) principal component analysis, respectively, as described in the Appendix.12 Both the indices are bounded between zero and one, where value closer to one indicates higher availability of banking
12
The correlation coefficient of ranks of states based on Bank Penetration and Outreach Index (Bank Penetration and Outreach Index PCA) is 0.8473 (0.8690), which is significant at one percent level. Therefore, it seems that Bank Penetration is a good proxy of availability of banking services.
12
ACCEPTED MANUSCRIPT
services. The average values (standard deviations) of Outreach Index and Outreach Index PCA were, respectively, 0.226 and 0.144 (0.161 and 0.138) in the year 2001 (see Table 4). Clearly, the overall measures of banking sector’s outreach also indicate that, during the period of study, supply of banking services was quite low on an average and varied widely across states in India.
RI PT
We also control for possible implications of the level of infrastructural facilities and extent of urbanization in a state, by considering the variables Tele-density and Urbanization as explanatory variables.
SC
3.4. Estimation of the Econometric Model
First, we estimate the model given in equation (3) by considering the explanatory variables
M AN U
pertaining to household characteristics using full sample, results of which are reported in Table 5. Second, we estimate the model separately for rural and urban sectors, to check whether association of household characteristics with the dependent variable varies across sectors (Table 6). In both the first and second regressions, we control for state specific unobserved factors using dummy variables for states appropriately. Third, we drop the state dummies and introduce the state level factors, Bank Penetration, Tele-density and Urbanization, in the regression (see
TE D
Model I in Table 7). It helps us to examine the association of specific state level variables of interest with the dependent variable. 13 Fourth, we consider interaction of the variable Bank Penetration with the dummy variables Consumption2, Consumption3, Consumption4 and Consumption5, in order to examine whether there is any differential impact of availability of
EP
banking services on use of formal financial services by households from different income class (see Model II in Table 7). Since required data to measure availability of banking services in rural
AC C
and urban sectors separately is not available, we consider the full sample to estimate the model when state-level variables are also considered as explanatory variables. Fifth, to check the sensitivity of our results to the measure of availability of banking services considered, we reestimate the last two specifications of the model by replacing the variable Bank Penetration with alternative measures of availability of banking services, Outreach Index and Outreach Index
13
Needles to mention here, we cannot consider state level variables along with the state dummies, since that would lead to the problem of multicollinearity.
13
ACCEPTED MANUSCRIPT
PCA, separately (see Table 8 and Table 9). In order to avoid possible endogeneity problem, we consider lagged values of the state level variables in the regression.14
3.5. Estimation of Relative Importance of Demand and Supply
RI PT
Finally, in order to understand the relative importance of demand- and supply-side factors which influence formal financial service utilization, we carry out the Shapley decomposition analysis. The explanatory variables are grouped into three categories, namely, (1) household characteristics, representing the demand side, (2) availability of banking services, representing
SC
the supply side; and (3) other region-specific variables. The Shapley method decomposes the pseudo R-square15 and estimates the contribution of each of the categories to total variation in
M AN U
the financial inclusion (Shorrocks, 1999 and 2013).
4. Empirical Findings
In this section, we first analyze the distribution of utilization of formal financial services across economic status of household. Next, we examine various factors affecting utilization of formal
TE D
financial services.
4.1. Concentration Curves and Estimates of Concentration Index of Financial Inclusion across States and Sectors in India
First, we plot the concentration curve of financial inclusion. In order to understand the relative
EP
difference between inequalities in financial inclusion and in economic wellbeing, we also plot
AC C
the Lorenz curve for MPCE along with the concentration curve of financial inclusion.
Note that the concentration curve of financial inclusion for all India lies everywhere below the 45o line, but it lies everywhere above the Lorenz curve for MPCE (see Figure 1). It implies that
14
We note here that possible endogeneity problem may not be fully avoided by considering the lagged values of explanatory variables always. However, note that the focus of this analysis is to examine the association of statelevel explanatory variables with the dependent variable, not the causal effects. Therefore, it seems that the issue of endogeneity does not deserve much attention in the present context. 15 We use the ‘shapley2’ software package in STATA for decomposition analysis (Juarez, 2012).
14
ACCEPTED MANUSCRIPT
there is income related inequality in financial inclusion in India, and the income-related inequality in financial inclusion is lower than inequality in consumption expenditure.
Further, it seems that both income related inequality in financial inclusion and inequality in
RI PT
MPCE differ across sectors, rural and urban, in India. In particular, it may be observed from Figure 2 that (a) inequality in MPCE is higher in urban India than that in rural India and (b) the concentration curve of financial inclusion is relatively closer to the Lorenz curve of MPCE for rural India as compared to that for urban India. It seems to indicate that income related inequality
SC
in financial inclusion closely follow inequality in MPCE in rural sector, but not so in urban sector. However, from Figure 2 it appears to be difficult to assess the relative position of rural
M AN U
India in terms of income related inequality in financial inclusion vis-à-vis that of urban India from the concentration curves. In order to address this issue we estimate the modified concentration index of financial inclusion ( CW ), using the formula as given by equation (2), and compare the values of CW for rural and urban India.
Table 1 reports the estimates of concentration index of financial inclusion, along with the
TE D
percentage of households using formal financial services, average MPCE and Gini coefficient of MPCE, for India as well as for each of the 35 states and UTs in India. Estimated concentration index of financial inclusion for all India is found to be 0.345, which implies that financial inclusion is concentrated among richer households. In other words, financial exclusion is
EP
disproportionately higher among the relatively poor households compared to their richer counterparts. It is also evident that there is considerable variation in terms of income related
AC C
inequality in financial inclusion across Indian states and UTs. In 14 out of 35 states and UTs the estimated concentration index is greater than that for all India. The concentration index is highest, 0.624, in Meghalaya; which is followed by Chandigarh, Sikkim, Mizoram, Manipur and Gujarat. Whereas, it is lowest in Dadra Nagar Haveli; which is followed by Andaman and Nicobar Island, Arunachal Pradesh, Uttaranchal, Kerala and Haryana. Interestingly, in Dadra Nagar Haveli financial inclusion is more concentrated among poorer households, which is clearly an exception. However, the extent of income related inequality in financial inclusion is quite low, -0.066, in Dadra Nagar Haveli. 15
ACCEPTED MANUSCRIPT
Also, note that although only 22.4 percent of households in Arunachal Pradesh use formal financial services, income related inequality in financial inclusion is quite low in this state. At the same time, in Chandigarh, 66.9 percent households use formal financial services, but its
RI PT
estimated concentration index is as large as 0.612. Therefore, it seems that increase of the proportion of households using formal financial services in a state/UT need not necessarily reduce inequality in financial inclusion across income groups or foster financial inclusion among the poor households in that state/UT. There is no significant relation between the percentage of
SC
households using formal financial services and concentration index of financial inclusion (the correlation coefficient between these two variables is negative, -0.28, but not significant at 10
M AN U
percent level).
We also find that the correlation coefficients between concentration index of financial inclusion and average MPCE is negative, but very low (-0.017) and not significant at 10 percent level. That is, income related inequality in financial inclusion does not appear to be significantly associated with overall economic wellbeing. However, there is significant (at one percent level)
TE D
and positive correlation between the percentage of households using formal financial services and average MPCE across states and UTs, which seems to be in line with the results from crosscountry studies (Honohan, 2004; Beck et al., 2009). Therefore, it seems that higher level of economic wellbeing is positively associated with higher level of financial inclusion on an
AC C
households.
EP
average, but not necessarily with higher level of financial inclusion among the relatively poor
Further, we find that there is very weak correlation between concentration index of financial inclusion and Gini coefficient of MPCE (the correlation coefficient is -0.037, which is not significant at 10 percent level). Nonetheless, high (low) income inequality may coexist with low (high) income related inequality in financial inclusion in some cases, as observed in case of Uttaranchal and Arunachal Pradesh (Manipur, Meghalaya and Punjab). It seems to indicate that the policy measures that are effective to reduce income inequality may not prove to be equally effective in reducing income related inequality in financial inclusion. 16
ACCEPTED MANUSCRIPT
These results show that there is correlation between financial inclusion and economic wellbeing. However, improving economic wellbeing does not immediately imply higher usage of formal financial services by both rich and poor. So, in order to achieve higher financial inclusion by
RI PT
needy, we have to examine factors other than income that facilitate financial inclusion. We analyze such factors at the household level and report the results in the next sub-section.
Considering rural sector and urban sector separately, we find that both the percentage of
SC
households using formal financial services and the concentration index of financial inclusion are lower in rural India compared to that in urban India (see Table 2). This is similar to the rural-
M AN U
urban comparisons of average MPCE and Gini coefficient of MPCE in India. The utilization rate, i.e., percentage of households using formal financial services, and concentration index are, respectively, about 41 percent and 29 percent higher in urban India compared to that in rural India. Clearly, there is rural-urban divide both in terms of use of formal financial services and income related inequality in financial inclusion. Nonetheless, more than 50 percent households
TE D
are completely excluded from the formal financial system even in urban India.
As before, there is considerable variation both in terms of percentage of households using formal financial service and concentration index of financial inclusion in both rural and urban sectors across states and UTs. It seems to be important to note that the range of concentration index of
EP
financial inclusion across states and UTs is much higher in rural sector (0.89) compared to that in urban sector (0.78). Also note that, the percentage of households using formal financial services is higher in urban sector than in rural sector of each of the states and UTs, except in Delhi, as in
AC C
case of all India.16 However, concentration index of financial inclusion is higher in rural sector, compared to that in urban, of ten states and UTs (Dadra Nagar Haveli, Delhi, Jammu & Kashmir, Jharkhand, Lakshadweep, Maharashtra, Meghalaya, Mizoram, Nagaland and Sikkim), and in Manipur there is no difference between the sectors. Also, only in three UTs (Andaman and Nicobar Island, Chandigarh and Dadra Nagar Haveli) the concentration index of financial
16
In Delhi the percentage figure is marginally in favor of the rural sector.
17
ACCEPTED MANUSCRIPT
inclusion is negative. That is, only in these three UTs utilization of formal financial services is concentrated among relatively poor households.
We also find that the correlation coefficient between the concentration index of financial
RI PT
inclusion and the percentage of households using formal financial services is negative, -0.48, and significant (at one percent level) in the case of rural sector. However, in urban sector there is no significant correlation between these two variables, as observed in the case of both combined. It implies that increase in overall financial inclusion can have differential consequence on income
SC
related inequality in financial inclusion across these two sector.
M AN U
There is positive and significant, at one percent level, correlation between average MPCE and percentage of households using formal financial services in both rural sectors (0.55) and in urban sector (0.52), as in the case of both the sectors combined (0.56). However, there is no significant (at ten percent level) correlation (a) between concentration index of financial inclusion and average MPCE and (b) between concentration index of financial inclusion and Gini coefficient of MPCE, even when we consider rural and urban sectors separately. These results strengthen the
TE D
arguments that, though higher level of per capita income may be positively associated with higher level of financial inclusion, (a) higher level of per capita income in a region need not necessarily be associated with lower level of income related inequality in financial inclusion and (b) income related inequality in financial inclusion should be treated differently from income
EP
inequality.
AC C
4.2. Demand-side and Supply-side Factors associated with Household’s Propensity to Use Formal Financial Services: What Matters More? Given the presence of high income related inequality in use of financial services, it becomes imperative to analyze the factors that determine use of financial services. The focus of this subsection is twofold, first, to examine whether greater availability of banking services is linked to higher utilization of financial services, particularly among the poor households and second, to understand the relative importance of supply-side availability of banking services in determining actual utilization.
The empirical findings show that availability of banking services does 18
ACCEPTED MANUSCRIPT
increase the utilization and the effect is higher for the poorer households (Table 7 – Table 9). However, the relative contribution of banking services in the Pseudo-R2 (explained variation) is very small around four percent, whereas the household characteristics together account for the
(Table 10). We explain the results below in detail.
RI PT
largest portion in the total explained variation in the utilization of formal financial services
The econometric analysis gives several interesting results. First and foremost, availability of banking services is positively associated with a household’s propensity to use formal financial
SC
services. The estimated coefficient of the Bank Penetration variable is positive and significant at one percent level (Model I in Table 7) and thus bank increases the probability of financial
M AN U
inclusion. Moreover, we find that the marginal effects of (Consumption3*Bank Penetration), (Consumption4*Bank Penetration) and (Consumption5*Bank Penetration) are negative and significant, but the marginal effect of (Consumption2*Bank Penetration) is insignificant; while the marginal effects of Consumption2, Consumption3, Consumption4, Consumption5 and Bank Penetration continue to be positive and significant (see Model II in Table 7). Therefore, we can say that presence of banks increases the probability of financial inclusion especially among poor
TE D
people. Recent policies of government of India of taking banks to the people seems logical in light of the above finding. These results hold true, if we consider Outreach Index or Outreach Index PCA in place of Bank Penetration (see Model I and Model II in Table 8 & 9). In other words, our results are robust to consideration of alternative measures of availability of banking
EP
services.
AC C
Household’s propensity to use formal financial services increases as its economic status improves. The coefficients of the dummy variables Consumption2– Consumption5 are positive and significant, and the difference between the marginal effects of two consecutive MPCE quintile groups is greater for higher quintile groups (Table 5). These results remain valid, if we estimate the model separately for rural and urban sectors or consider alternative specifications of the model (Table 6 – Table 9). That is, this result indicates that larger proportion of poor households do not use formal financial services, compared to that of relatively rich households, even after controlling for the other factors. It strengthens the findings of Section 3 that there is 19
ACCEPTED MANUSCRIPT
income related inequality in financial inclusion and larger proportion of relatively poor households are financially excluded compared to their richer counterparts in India.
The econometric analysis also shows that there is positive (negative) and significant association
RI PT
of the level of infrastructural facilities (urbanization) in a state with the probability to use formal financial services by a household in that state. Interestingly, we find that the marginal effect of availability of banking services is greater than that of level of infrastructural facilities. This result indicates stronger association of availability of banking services with household’s propensity to
SC
use formal financial services as compared to other infrastructural facilities.
M AN U
We also find that the coefficient of the dummy variable Rural is positive and significant (see Table 5). However, descriptive statistics in Table 2 shows that lower proportion of rural households are financially included compared to that of urban households. Therefore, it seems that the degrees of association of sector specific unobserved factors are positively biased towards rural sector. Breaking the dependent variable into two, we observe that formal savings (credit) services are utilized by 40.05 percent and 21.38 percent (13.54 percent and 19.45 percent) of
TE D
urban and rural households, respectively. That is, while higher percentage of urban households use formal savings services than that of rural households, utilization of formal credit services is greater in the rural sector compared to urban sector. Therefore, it seems that rural focus of government’s credit policy (such as, priority sector lending norms, Kisan credit card scheme,
EP
etc.) has facilitated financial inclusion in rural sector vis-à-vis urban sector.
AC C
Let us now turn to the associations of other characteristics of households with their propensity to use formal financial services. We find that the coefficients of Self Employed and Laborer are negative and significant at one percent level, and the absolute value of the marginal effect of Laborer is greater than that of Self Employed (see Table 5 – Table 9). It confirms commonly held perceptions that (a) principal economic activity of a household plays crucial role in determining whether that household would use formal financial services or not and (b) the propensity to use formal financial services is highest for households with salaried employment as the main source
20
ACCEPTED MANUSCRIPT
of income and lowest for households who are primarily dependent on casual labor supply, ceteris paribus.
The econometric analysis also reveals that the probability of a household to be financially
RI PT
included increases at an increasing rate with the increase in level of education of the household head. In other words, as expected, greater proportion of households with relatively lower level of education are financially excluded. Interestingly, we find that the marginal effect of Household Size is always positive and significant. It implies that, ceteris paribus, the probability to use
SC
formal financial services is higher for larger households as the probability of using formal financial services by at least one member of a household increases with the increase in number of
M AN U
members.
Negative and significant marginal effect of the dummy variable Female, when we estimate the model using full sample, seems to imply that the propensity of female headed households to use formal financial services is less than the male headed households in India (see Table 5). However, separate estimates for the two sectors, rural and urban, reveals that the gender of
TE D
household head plays significant role only in rural India, not in urban India (see Table 6).17 In other words, unlike as in the case of rural India, female headed households are not significantly different from male headed households as far as utilization of formal financial services is concerned. We find similar results for households belonging to socially disadvantageous
EP
communities. The marginal effects of SC, ST and OBC are negative and significant, if we consider both rural and urban sector together or consider only rural sector. On the other hand, in
AC C
the case of urban sector, marginal effects of SC and ST turn out to be insignificant, while the marginal effect of OBC is negative and significant at 5% level. It implies that, in urban India, the propensity to use formal financial services is lower for OBCs as compared to the general category households. However, in rural India the propensity to use formal financial services is the lowest for ST households, followed by SC households and OBC households, ceteris paribus (see Table 6).
17
The marginal effect of Female in case of urban sector is negative but not significant at 5% level.
21
ACCEPTED MANUSCRIPT
There are differences between the two sectors, rural and urban, in terms of magnitudes of marginal effects of other household characteristics as well. It is easy to observe that marginal effects of Consumption2 – Consumption5 are greater in urban sector than that it rural sector. It justifies the higher concentration index of financial inclusion in urban India compared to that in
RI PT
rural India, as observed in Section 3. Degrees of association of employment status and household size with the propensity to use formal financial services are also stronger in urban sector compared to that in rural sector. Rural-urban comparison of the marginal effects of dummy variables for educational levels leads to mixed results. We find that the marginal effects of
SC
Primary, Middle, and Secondary and above are higher in urban sector than in rural sector, but the
M AN U
marginal effect of Literate is higher in rural sector than in urban sector.
The above results show that various household-specific factors along with supply-side variables affect household’s utilization of formal financial services. The Shapley decomposition analysis gives the relative contributions of these factors in predicting the utilization. We decompose the pseudo- (explained variation 18 ) and calculate the contribution of
three broad groups of
variables, household level variables, availability of banking services and other region-specific
TE D
factors in pseudo- . We carry out this analysis for all the three specifications with different indicators of availability banking services of Model 1 in as reported in table: 7, 8, 919. We find that the relative contribution of availability of banking services is lowest at 4.4 percent when we consider either bank penetration or outreach index (Table 10). On the other hand, the household
EP
characteristics, representing demand side barriers, contribute between 83.54 and 84.86 percent depending on the model20. The results are consistent across models with very less variation in the
18
AC C
contributions of different groups. Thus, the decomposition analysis points out the importance of There is unexplained variation of financial inclusion too but we can’t decompose that as it is not being explained by the model. We can only decompose the proportion of explained variation to get the source of it. 19
For decomposition analysis models having interaction terms between demand-side (consumption categories) and supply-side (banking penetration/ outreach) factors are not considered, because it is not feasible to attribute contribution of those interaction terms to either of the two sides – demand or supply. 20
It may be noted that the supply-side factors are considered at the state-level due to the non-availability of data at more granule level. It is likely that the contribution of the supply-side factors will change if more localized information on physical access is available.
22
ACCEPTED MANUSCRIPT
demand-side factors rather than physical availability of banking services as the major barriers to actual utilization of formal financial services.
5. Concluding Remarks
RI PT
In this paper we have analyzed the distribution and determinants of usage of formal financial service across sub-national regions in India using a representative household level survey data, linked to state-level factors. The main contributions of this paper are threefold. First, to the best of our knowledge, this is the first attempt to study distribution of usage of formal financial
SC
services across income using concentration indices. Second, it provides empirical evidence of the household’s propensity to use formal financial services in a developing country using both
M AN U
demand and supply-side factors. We also estimate their relative contribution in determining the utilization of formal financial services. Third, it provides an assessment of the role of banking services to promote financial inclusion, particularly among the poor households.
We find that the extent of financial inclusion is low, particularly among the poor, in India. We also find that there are considerable variations in terms of concentration index of financial
TE D
inclusion and the level of financial inclusion across Indian states and UTs as well as across sectors – rural and urban. Both the level of financial inclusion and estimated concentration index of financial inclusion are lower in rural sectors compared to that in urban sectors. Nonetheless, about a half of the urban households are financially excluded. Our findings seem to indicate that
EP
increase in level of financial inclusion can have differential consequence on income related
AC C
inequality in financial inclusion across sectors.
Though there is positive correlation between overall economic wellbeing (average MPCE) and the percentage of financially included households at state level, there is no significant correlation of concentration index of financial inclusion with overall economic wellbeing or with the percentage of financially included households.
We also demonstrate that income related
inequality in financial inclusion deserves special attention of the policy makers.
23
ACCEPTED MANUSCRIPT
Results of our econometric analysis demonstrate that availability of banking services is positively associated with a household’s propensity to use formal financial services and, in particular, the degree of association is higher for the poor and marginally non-poor households. It seems to indicate that greater availability of banking services can foster financial inclusion,
RI PT
particularly among the poor. Moreover, compared to the availability of infrastructural facilities and extent of urbanization in a state, availability of banking services in a state is more strongly associated with the propensity of a household to be financially included. These results are robust
SC
to alternative measures of availability of banking services.
We also find that the level of education, employment status and household size are significantly
M AN U
associated with the use of formal financial services by a household in both rural and urban sectors. Rural-urban comparison indicates that income and employment status have stronger association with urban household’s propensity to be financially included, compared to that of a rural household. An important finding is that self-employed households from the urban areas are less likely to be financially included as compared to the salaried households, which indicates that small business families in the urban areas are more likely to rely on informal sources to meet
TE D
their need of financial services.
Interestingly, we find that the gender of household head and social group are not significantly associated with use of formal financial services by urban households. However, in rural India,
EP
the probability to use formal financial services by socially deprived classes and female headed households is significantly lower than their counterparts. In urban India, only the OBC
AC C
households have higher probability to be financially excluded compared to the households belonging to other social groups including general category households.
We have evidence suggesting that both demand- and supply-side factors affect the probability of utilizing formal financial services. However, household characteristics representing demand side, contribute the largest part in determining the usage, while physical availability of banking services plays a very minor role. This result emphasizes policy measures to eliminate the demand-side barriers along with increasing availability of banking services. 24
ACCEPTED MANUSCRIPT
It may be noted that this paper considers household as financially excluded if that household does not use any of the financial services provided by institutional agencies; otherwise that
RI PT
household is considered as financially included. That is, we do not distinguish between partial and complete financial inclusion of a household. In order to assess the extent of financial inclusion of a household, it is important to appropriately estimate the need of different financial services by that household, which is beyond the scope of the present paper. It might also be
SC
interesting to assess the implications of government policies on dynamics of socio-economic inequality in financial inclusion. We leave these for future research.
M AN U
Acknowledgements:
References
TE D
The authors would like to thank the editor and referees for insightful comments. Authors would also like to thank Satya R. Chakravarty and the participants in the ‘Workshop on Measuring Financial Inclusion from the Demand Side’, conducted by the Center for Advanced Financial Research and Learning (CAFRAL) at NIBM, Pune, for comments and helpful discussions. An earlier draft of the paper was circulated as “Rama Pal & Rupayan Pal, 2012. Income related inequality in financial inclusion and role of banks: Evidence on financial exclusion in India, IGIDR-WP-2012-013, Indira Gandhi Institute of Development Research (IGIDR), India.” Usual disclaimer applies.
AC C
EP
Abiad, A., Detragiache, E. and Tressel, T., 2010. A new database of financial reforms. IMF Staff Papers, 57(2), pp.281-302. Abiad, A. and Mody, A., 2005. Financial reform: What shakes it? What shapes it? American Economic Review, 95(1), pp.66-88. Agnello, L., Mallick, S.K. and Sousa, R.M., 2012. Financial reforms and income inequality. Economics Letters, 116(3), pp.583-587. Ahamed, M.M. and Mallick, S.K. 2017. Is Financial Inclusion Good for Bank Stability? International Evidence. Journal of Economics Behavior and Organization, forthcoming. Ang, J.B., 2010. Finance and Inequality: the Case of India. Southern Economic Journal, 76(3), pp.738-761. Ashraf, N., Karlan, D. and Yin, W., 2006. Tying Odysseus to the mast: Evidence from a commitment savings product in the Philippines. The Quarterly Journal of Economics, 121(2), pp.635-672. 25
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Ayyagari, M., Demirgüç-Kunt, A. and Maksimovic, V., 2008. How Important are Financing Constraints? The Role of Finance in the Business Environment. The World Bank Economic Review, 22(3), pp.483-516. Banerji, S., Gangopadhyay, K., Gupta, P. and Mallick S., 2014. Financial intermediation and economic development: Evidence from Bihar, in Ashima Goyal (ed.), Handbook of the Indian Economy in the 21st Century: Understanding the Inherent Dynamism, Oxford University Press, pp.898-929. Basu, S. and Mallick, S., 2008. When Does Growth Trickle Down to the Poor? The Indian Case. Cambridge Journal of Economics, 32(3), pp.461-477. Beck, T., Demirgüç-Kunt, A. and Honohan, P., 2009. Access to financial services: Measurement, impact, and policies. The World Bank Research Observer, 24(1), pp.119-145. Beck, T., Demirgüç-Kunt, A. and Levine, R., 2007. Finance, inequality and the poor. Journal of economic growth, 12(1), pp.27-49. Beck, T., Demirgüç‐Kunt, A.S.L.I. and Maksimovic, V., 2005. Financial and legal constraints to growth: Does firm size matter?. The Journal of Finance, 60(1), pp.137-177. Brekke, K.A. and Kverndokk, S., 2012. Inadequate Bivariate Measures of Health Inequality: The Impact of Income Distribution. The Scandinavian Journal of Economics, 114(2), pp.323333. Burgess, R. and Pande, R., 2005. Do Rural Banks Matter? Evidence from the Indian Social Banking Experiment. American Economic Review, 95(3), pp.780-795. Cain, J.S., Hasan, R., Magsombol, R. and Tandon, A., 2010. Accounting for Inequality in India: Evidence from Household Expenditures. World Development, 38(3), pp.282-297. Chakravarty, S. R., and Pal, R., 2013. Measuring Financial Inclusion: An Axiomatic Approach. Journal of Policy Modeling, 35(5), pp. 813-83. Cecchetti, S., and Kharroubi, E., 2012. Reassessing the Impact of Finance on Growth. BIS Working Paper No 381, Bank for International Settlements. Chong, E., and Varela Sandoval, F., 2016. Does Financial Development Help Growth or Hurt It, or Neither? unpublished manuscript. Dabla-Norris E., M.E., Kochhar, M.K., Suphaphiphat, M.N., Ricka, M.F. and Tsounta, E., 2015a. Causes and consequences of income inequality: a global perspective. International Monetary Fund. Dabla-Norris, E., Ji, Y., Townsend, R.M. and Unsal, D.F., 2015 b. Distinguishing Constraints on Financial Inclusion and Their Impact on GDP, TFP, and Inequality (No. w20821). National Bureau of Economic Research. Datt, G. and Ravallion, M., 2002. Is India's economic growth leaving the poor behind?. The journal of economic perspectives, 16(3), pp.89-108. Deaton, A. and Dreze, J., 2002. Poverty and inequality in India: a re-examination. Economic and political weekly, pp.3729-3748.
26
ACCEPTED MANUSCRIPT
SC
RI PT
Demetriades, P.O. and Luintel, K.B., 1996. Financial development, economic growth and banking sector controls: evidence from India. The Economic Journal, pp.359-374. Demetriades, P.O. and Luintel, K.B., 1997. The direct costs of financial repression: evidence from India. The Review of Economics and Statistics, 79(2), pp.311-320. Dev, S.M. and Ravi, C., 2007. Poverty and inequality: All-India and states, 19832005. Economic and Political weekly, pp.509-521. Dupas, P. and Robinson, J., 2013. Savings constraints and microenterprise development: Evidence from a field experiment in Kenya. American Economic Journal: Applied Economics, 5(1), pp.163-192. Erreygers, G., 2009. Correcting the concentration index. Journal of health economics, 28(2), pp.504-515.
AC C
EP
TE D
M AN U
Erreygers, G., 2009 b. Correcting the concentration index: a reply to Wagstaff. Journal of Health Economics, 28(2), pp.521-524. Government of India., 2005. Household assets and liabilities in India: NSS 59th round. National Sample Survey Organization, Ministry of Statistics and Programme Implementation. New Delhi. Honohan, P., 2004. Financial Development, Growth and Poverty: How Close are. Financial development and economic growth: Explaining the links, p.1. Honohan, P., 2008. Cross-country variation in household access to financial services. Journal of Banking & Finance, 32(11), pp.2493-2500. Jeanneney, S.G. and Kpodar, K., 2011. Financial development and poverty reduction: Can there be a benefit without a cost?. The Journal of development studies, 47(1), pp.143-163. Juarez, F.C. 2012. SHAPLEY2: Stata Module to Compute Additive Decomposition of Estimation Statistics by Regressors or Groups of Regressors. Statistical Software Components S457543, Boston College Department of Economics, revised 17 June 2015. Kaboski, J.P. and Townsend, R.M., 2005. Policies and Impact: An Analysis of Village‐Level Microfinance Institutions. Journal of the European Economic Association, 3(1), pp.1-50. Kakwani, N.C., 1977. Applications of Lorenz curves in economic analysis. Econometrica: Journal of the Econometric Society, pp.719-727. Kochar, A., 2011. The distributive consequences of social banking: a microempirical analysis of the Indian experience. Economic Development and Cultural Change, 59(2), pp.251-280. Levine, R., 2005. Finance and Growth: Theory and Evidence. Handbook of Economic Growth, 1, pp.865-934. Mahalanobis, P.C., 1960. A method of fractile graphical analysis. Econometrica: Journal of the Econometric Society, pp.325-351. Rajan, R.G. and Zingales, L., 2003. The great reversals: the politics of financial development in the twentieth century. Journal of financial economics, 69(1), pp.5-50.
27
ACCEPTED MANUSCRIPT
RI PT
Rajan, R.G. and Zingales, L., 2004. Saving capitalism from the capitalists: Unleashing the power of financial markets to create wealth and spread opportunity. Princeton University Press. RBI 2014 Financial Intermediation for All – Economic Growth with Equity, URL: https://rbi.org.in/scripts/FS_Speeches.aspx?Id=912&fn=2754. Last accessed: May 03, 2018. Shorrocks, A. F. 1999. Decomposition Procedures for Distributional Analysis: a Unified Framework Based on the Shapley Value. Unpublished Manuscript. Department of Economics, University of Essex.
Shorrocks, A. F. 2013. Decomposition Procedures for Distributional Analysis: a Unified
SC
Framework Based on the Shapley Value. Journal of Economic Inequality. 11(1). pp. 99126.
AC C
EP
TE D
M AN U
Suryanarayana, M.H., 2008. What Is Exclusive about ‘Inclusive Growth’?. Economic and Political Weekly, pp.93-101. Wagstaff, A., 2005. The bounds of the concentration index when the variable of interest is binary, with an application to immunization inequality. Health economics, 14(4), pp.429432. Wagstaff, A., 2009. Correcting the concentration index: a comment. Journal of Health Economics, 28(2), pp.516-520.
28
ACCEPTED MANUSCRIPT
APPENDIX Derivation of Concentration Index
RI PT
Let x be the variable representing economic status, i.e. MPCE, of household and y = g(x) be the utilization of formal financial services by the household with MPCE x, where the function y = g(x) is defined as follows.
SC
1, if the household uses at least one of the formal financial services y = g (x) = 0, otherwise
Let F(x) be the distribution function of MPCE and f(x) be the probability density function of MPCE. Now, the
M AN U
proportion of households having MPCE less than or equal to x and using formal financial services, out of total number of households using formal financial services, can be represented by F1[ g ( x)] as follows.
() =
(
1 () () !, ()
where () = ) () ()! represents expected number of of households using formal financial services.
TE D
The relation between () and F(x) can be called as the concentration curve of y, i.e., concentration curve of financial inclusion (see Kakwani, 1977).
AC C
EP
From the properties of distribution functions, we have:
lim () = 0
→
lim () = 1
→(
And one can write
! () () () = ! ()
Following Kakwani (1977), we define the concentration index for financial inclusion as one minus twice the area under the concentration curve for . Therefore, our concentration index for financial inclusion is given by (
= 1 − 2 () ()! .
29
ACCEPTED MANUSCRIPT
Using integration by parts ( ( ! () = 1 − 2 * () ()! − + , - ()!. ! / !
⟹ = 1 − 2 *1 − ⟹ =1−2+
(
Now, () = ) () ()! = ,
( 2 () ()()! ()
( 2 () ()()! ()
( 2 () * () ()()! − / () 2
(
( 1 () ()()! / ()
M AN U
⟹=
() () ()!/ ()
SC
⟹ = −1 +
(
RI PT
⟹ = 1 − 2 * ()()|( −
and 345(), () = ) :() − 6()7; :() − ; ()!
( 1 ( ⟹ 345(), () = ()() ()! − () ()! 2
TE D
(
( ( 1 − 6()7 () ()! + (()) ()! 2
Since ) () ()! = we can write above expression as
( 1 345(), () = ()() ()! − (()) 2
EP
AC C
Using this we can write
=
2 8345(), ()9 ()
One can write the discrete case of the above formulae as
where
µ = E ( y) ,
=
2 345,
r denotes the individual’s fractional rank in the MPCE distribution and n is number of
households in the sample.
30
ACCEPTED MANUSCRIPT
Using the fact that ̅ = 1/2
∑ ( − )( − 1/2)
345, =
∑ − ∑ − ∑ +
⟹ 345, =
1 1 1 1 − − + 2 2 2
SC
⟹ 345, =
∑ ( − =)( − ̅ )
RI PT
345, =
1 1 ⟹ 345, = − 2
Therefore, = : ∑ − ; E
M AN U
=
2 − 1
For a given > 0, concentration index is maximum when poorest @ individuals are not included financially i.e. for =
TE D
them = 0 and for rich − @ individuals = 1. Thus we have, =
EP
AC C
an can be written as
2 1 A 0 + 0 + 0 + 0 … + @ + 1 + @ + 2 + @ + 3 + ⋯ + D − 1 −@
Using the sum of natural numbers from @ + 1 to :
GH
2 @+1 @+2 @+3 A 0 + 0 + 0 + 0 … + + + + ⋯ + 1 D − 1 −@
⟹=
Using =
GH
=
@+1+ −1
, we can write:
= 1−+
1
Minimum value of can be derived in the same way and is given by = − 1 +
31
ACCEPTED MANUSCRIPT
APPENDIX (Continued) EXPLANATORY VARIABLES: DEFINITIONS AND DATA SOURCES Variable
Definition
Data Source
Household Size
ST
OBC
Place of Residence: Rural (Ref: Urban)
-do-
-do-
-do-
-do-
The variable ‘Household Size’ is defined as the number of members in the household -do-
It takes the value 1 if household belongs to General category; otherwise 0. It takes the value 1 if household belongs to Scheduled Caste; otherwise 0. It takes the value 1 if household belongs to Scheduled Tribes; otherwise 0. It takes the value 1 if household belongs to Other Backward Castes; otherwise 0.
AC C
Social Group: General (Base category) SC
AIDIS, 2002-2003
It takes the value 1 if the household head is female; otherwise 0.
EP
Gender: Female (Ref: Male)
TE D
M AN U
SC
RI PT
Household Characteristics Income: Consumption1- Consumption i takes the value 1 if the household belongs to the ith consumption quintile group based on MPCE; otherwise 0. Consumption5 (Base category: Consumption1) Employment Status: Salaried (Base It takes the value 1 if the household’s principal earning is from category) salaried employment; otherwise 0. Self Employed It takes the value 1 if the household’s principal earning is from self employment; otherwise 0. Laborer It takes the value 1 if the household’s principal earning is from casual/agricultural/non-agricultural labor; otherwise 0. Education: Illiterate(Base It takes the value 1 if head of household is illiterate; otherwise 0. category) otherwise Literate It takes the value 1 if head of household is literate; otherwise 0.otherwise Primary It takes the value 1 if highest education completed by head is primary; otherwise 0. Middle It takes the value 1 if the highest education completed by head is middle level; otherwise 0. Secondary and It takes the value 1 if the highest education completed by head is above secondary level or above; otherwise 0.
-do-
It takes the value 1 if household resides in rural area; otherwise 0.
32
ACCEPTED MANUSCRIPT
APPENDIX (Continued) Variable
Definition
Data Source
The variable ‘Outreach Index’ is the index of outreach of banking services, which is constructed by considering (a) number of bank offices per lakh population, (b) number of bank offices per thousand square kilometer area, (c) the number of deposit accounts per thousand population and (d) number of credit accounts per thousand people as indicators of outreach of banking services employing the formula, as in Chakravarty and Pal (2013), r
SC
Outreach Index
Basic Statistical Returns of Scheduled Commercial Banks, Reserve Bank of India. (Data on population and land area comes from Census of India).
RI PT
State Level Factors Availability of Banking Services: Bank The variable ‘Bank Penetration’ is measured as number of bank Penetration offices per lakh population. (1 lakh =100000)
M AN U
1 k x − mi (Outreach Index 1) = ∑ i , where r=0.75, k=4, k i =1 M i − mi xi denotes the value of the i-th indicator, mi (Mi) denotes the minimum (maximum) value of the i-th indicator.
The variable ‘Teledensity’, which is a proxy for infrastructural facility, is defined as the number of telephones per hundred population.
AC C
EP
Infrastructure: Teledensity
The variable ‘Outreach Index PCA’ is the index of outreach of banking services constructed by employing well-known principal component analysis (PCA), by considering the following four indicators of outreach of banking services: (a) number of bank offices per lakh population, (b) number of bank offices per thousand square kilometer area, (c) the number of deposit accounts per thousand population and (d) number of credit accounts per thousand people.
TE D
Outreach Index PCA
Urbanization
The variable ‘Urbanization’ is measured as the percentage of urban population to total population.
33
Department of Telecommunication, Ministry of Communications & Information Technology, Government of India. Census of India
ACCEPTED MANUSCRIPT
Table 1: State-wise Estimates of Concentration Index of Financial Inclusion Average MPCE
Gini Coefficient
0.092 0.267 0.112 0.355 0.298 0.612 0.374 0.243 0.277 -0.066 0.262 0.465 0.195 0.291 0.444 0.393 0.296 0.184 0.228 0.307 0.490 0.624 0.503 0.428 0.351 0.318 0.449 0.438 0.234 0.532 0.293 0.339 0.259 0.179 0.328 0.345
1047.52 635.78 632.22 563.94 426.96 1497.54 443.60 958.21 1175.82 955.89 992.37 771.03 824.98 773.95 501.23 771.78 670.52 792.60 996.32 752.03 644.76 724.32 983.13 535.37 964.52 402.24 885.21 894.05 607.19 710.64 719.84 561.01 516.17 1034.07 598.99 634.90
0.226 0.322 0.364 0.264 0.241 0.372 0.313 0.350 0.321 0.195 0.346 0.321 0.287 0.300 0.320 0.229 0.310 0.300 0.197 0.343 0.191 0.230 0.229 0.305 0.204 0.320 0.329 0.281 0.278 0.308 0.321 0.287 0.284 0.516 0.323 0.333
SC
RI PT
Concentration Index
M AN U
EP
Andaman &Nicobar (UT) Andhra Pradesh Arunachal Pradesh Assam Bihar Chandigarh(UT) Chhattisgarh Daman and Diu (UT) Delhi Dadra Nagar Haveli (UT) Goa Gujarat Haryana Himachal Pradesh Jharkhand Jammu &Kashmir Karnataka Kerala Lakshadweep(UT) Maharashtra Manipur Meghalaya Mizoram Madhya Pradesh Nagaland Orissa Pondicherry(UT) Punjab Rajasthan Sikkim Tamil Nadu Tripura Uttar Pradesh Uttaranchal West Bengal All-India
Percentage of Households Using Formal Financial Services 62.3 37.8 22.4 31.8 18.4 66.9 29.7 46.8 28.9 57.0 47.6 44.7 45.6 62.5 27.5 40.8 36.1 73.5 51.0 50.2 25.5 22.7 22.4 37.8 37.5 31.7 52.9 46.2 30.1 36.9 41.4 34.3 35.3 47.7 41.2 38.8
TE D
State/UT
AC C
‘(UT)’ indicates union territory. Source: Estimation based on All-India Debt and Investment Survey,2002-03
34
ACCEPTED MANUSCRIPT
Table 2: State-wise and Sector-wise Estimates of Concentration Index of Financial Inclusion
M AN U
TE D
AC C
Gini Coefficient
78.4 43.2 41.1 63.1 37.0 75.7 52.9 83.9 29.2 77.7 64.3 58.6 53.8 74.7 57.8 61.9 51.4 73.2 71.4 63.1 47.3 44.3 49.6 57.1 60.7 52.7 53.2 52.2 45.3 71.4 49.2 61.3 42.1 64.4 57.2 49.2
Average MPCE
0.196 0.252 0.316 0.234 0.199 0.213 0.258 0.330 0.202 0.159 0.332 0.266 0.232 0.279 0.226 0.205 0.220 0.257 0.225 0.263 0.174 0.193 0.206 0.225 0.157 0.240 0.245 0.259 0.220 0.277 0.255 0.234 0.231 0.551 0.242 0.268
0.654 0.421 0.297 0.481 0.336 0.823 0.710 0.541 0.259 0.043 0.756 0.600 0.441 0.481 0.235 0.137 0.426 0.201 0.208 0.320 0.450 0.416 0.403 0.544 0.224 0.490 0.494 0.517 0.361 0.358 0.379 0.615 0.426 0.535 0.363 0.374
1238.23 958.91 1112.09 916.79 717.31 1544.69 764.75 1720.71 1243.22 1070.18 1191.72 1055.23 1127.92 1254.07 923.00 1021.12 989.50 1046.22 989.91 1053.02 764.43 1110.46 1159.90 856.64 1170.14 830.68 1007.84 1064.30 896.94 1144.14 1007.87 1005.93 755.44 920.27 955.49 970.85
0.246 0.356 0.487 0.269 0.311 0.374 0.348 0.315 0.323 0.252 0.330 0.314 0.332 0.276 0.302 0.219 0.306 0.348 0.171 0.305 0.195 0.219 0.230 0.325 0.226 0.358 0.345 0.292 0.303 0.321 0.336 0.291 0.329 0.325 0.337 0.331
RI PT
Utilization*
946.20 520.83 557.57 520.04 391.15 1028.39 384.72 876.03 782.67 899.20 822.16 610.37 708.13 712.33 393.33 694.80 511.87 703.97 1004.22 524.59 598.89 653.17 874.86 432.46 848.08 334.71 672.43 800.88 516.72 640.41 577.12 489.17 448.47 1063.42 477.45 508.80
Concentration Index
Gini Coefficient
-0.130 0.202 0.010 0.278 0.257 -0.240 0.325 0.163 0.296 -0.082 0.080 0.427 0.099 0.294 0.287 0.442 0.227 0.134 0.390 0.367 0.450 0.592 0.620 0.327 0.353 0.266 0.358 0.430 0.094 0.651 0.216 0.338 0.209 0.112 0.270 0.290
SC
Average MPCE
74.0 32.6 20.3 31.3 24.3 60.7 36.6 49.1 29.6 64.3 61.9 40.3 47.0 62.1 27.4 42.4 36.5 66.5 55.4 53.9 19.8 17.2 22.6 36.5 33.6 34.6 48.2 44.4 31.3 46.6 36.4 43.7 36.1 52.2 40.7 34.9
EP
Andaman & Nicobar (UT) Andhra Pradesh Arunachal Pradesh Assam Bihar Chandigarh (UT) Chhattisgarh Daman and Diu (UT) Delhi Dadra Nagar Haveli (UT) Goa Gujarat Haryana Himachal Pradesh Jharkhand Jammu & Kashmir Karnataka Kerala Lakshadweep (UT) Maharashtra Manipur Meghalaya Mizoram Madhya Pradesh Nagaland Orissa Pondicherry (UT) Punjab Rajasthan Sikkim Tamil Nadu Tripura Uttar Pradesh Uttaranchal West Bengal All-India
Urban
Concentration Index
Rural Utilization*
State/UT
‘(UT)’ indicates union territory. * Utilization refers to percentage of households using formal financial services. Source: Estimation based on All-India Debt and Investment Survey,2002-03
35
ACCEPTED MANUSCRIPT
Table 3: Variables Associated with Household Use of Formal Financial Services Household Characteristics
Availability of Banking Services Infrastructure Urbanization
Income Employment Status Education Household Size
0.212 0.205 0.199 0.202 0.182 0.210 0.488 0.302 0.408 0.119 0.137 0.136 0.200 0.110 4.885 0.305 0.208 0.081 0.407 0.747 7.789 0.226 0.161 4.767 25.795
AC C
EP
Consumption1 Consumption2 Consumption3 Consumption4 Consumption5 Salaried Self Employed Laborer Illiterate Literate Primary Middle Secondary and above Female Household Size General SC ST OBC Rural Bank Penetration Outreach Index Outreach Index PCA Teledensity Urbanization Number of Observations
Standard Deviation 0.409 0.404 0.399 0.401 0.386 0.408 0.500 0.459 0.491 0.324 0.344 0.342 0.400 0.313 2.524 0.460 0.406 0.273 0.491 0.435 3.981 0.144 0.138 3.040 11.069
Minimum
Maximum
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 3.57 0.02 0.01 1.10 9.79 129026
1 1 1 1 1 1 1 1 1 1 1 1 1 1 35 1 1 1 1 1 24.64 0.64 0.61 15.25 49.77
M AN U
Mean
TE D
Variable
SC
Table 4: Summary Statistics of Explanatory Variables
36
Social Group Gender Place of Residence
RI PT
State-level Variables
ACCEPTED MANUSCRIPT
Table 5: Determinants of Use of Formal Financial Services by Households Marginal Effect 0.068 0.125 0.196 0.320 -0.036 -0.114 0.072 0.092 0.116 0.194 -0.018 0.029 -0.034 -0.075 -0.027 0.065
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.023 0.000 0.000 0.000 0.000 0.000 0.000
RI PT
Coefficient 0.343 0.628 0.983 1.606 -0.179 -0.571 0.363 0.460 0.585 0.977 -0.092 0.145 -0.169 -0.376 -0.136 0.328 -2.222
SC
Explanatory Variables Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Rural Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
M AN U
129026 3913.01 0.000 -106100000 0.118
Note: State dummies are included in the regression analysis.
Table 6: Sector Wise Determinants of Use of Formal Financial Services by Households Coefficient
Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
0.332 0.602 0.953 1.542 -0.024 -0.433 0.381 0.474 0.560 0.847 -0.122 0.127 -0.226 -0.445 -0.158 -1.864
Rural Marginal Effect 0.065 0.119 0.188 0.304 -0.005 -0.085 0.075 0.093 0.110 0.167 -0.024 0.025 -0.045 -0.088 -0.031
AC C
EP
TE D
Explanatory Variables
84411 2323.04 0.000 -78640822 0.096
p-value
Coefficient
0.000 0.000 0.000 0.000 0.613 0.000 0.000 0.000 0.000 0.000 0.014 0.000 0.000 0.000 0.000 0.000
0.451 0.847 1.225 1.861 -0.363 -0.769 0.365 0.500 0.750 1.271 0.054 0.206 0.027 -0.146 -0.075 -2.865
Urban Marginal Effect 0.091 0.171 0.247 0.375 -0.073 -0.155 0.074 0.101 0.151 0.256 0.011 0.042 0.005 -0.029 -0.015 44615 1902.36 0.000 -27123534 0.149
Note: State dummies are included in the regression analysis.
37
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.442 0.000 0.664 0.298 0.124 0.000
ACCEPTED MANUSCRIPT
Table 7: Availability of Banking Services and Use of Formal Financial Services
0.318 0.564 0.876 1.469 -0.170 -0.626 0.353 0.431 0.565 0.987 -0.163 0.144 -0.141 -0.332 -0.107 0.295
Bank Penetration
0.088
Consumption2* Bank Penetration Consumption3* Bank Penetration Consumption4* Bank Penetration Consumption5* Bank Penetration 0.052 -0.016 -2.490
0.017
0.010 -0.003
129026 4256.32 0.000 -105200000 0.125
TE D
Teledensity Urbanization Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
p-value
Coefficient
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.313 0.760 1.529 2.655 -0.160 -0.640 0.350 0.426 0.567 0.981 -0.166 0.145 -0.144 -0.345 -0.108 0.308
AC C
EP
Note: Regression analysis includes all Indian states.
38
0.000
0.000 0.000 0.000
Model II Marginal Effect 0.061 0.149 0.300 0.521 -0.031 -0.126 0.069 0.084 0.111 0.193 -0.033 0.029 -0.028 -0.068 -0.021 0.060
RI PT
Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Rural
Model I Marginal Effect 0.063 0.111 0.173 0.290 -0.034 -0.123 0.070 0.085 0.111 0.194 -0.032 0.028 -0.028 -0.065 -0.021 0.058
SC
Coefficient
M AN U
Explanatory Variables
p-value 0.019 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.168
0.033
0.000
-0.004 -0.039 -0.106 -0.174
-0.001 -0.008 -0.021 -0.034
0.857 0.057 0.000 0.000
0.057 -0.016 -2.995
0.011 -0.003
0.000 0.000 0.000
129026 4421.12 0.000 -104900000 0.127
ACCEPTED MANUSCRIPT
Table 8: Availability of Banking Services and Use of Formal Financial Services: Sensitivity Analysis I
0.331 0.589 0.914 1.523 -0.172 -0.628 0.357 0.432 0.562 0.983 -0.176 0.148 -0.145 -0.254 -0.105 0.303
Outreach Index (Consumption2* Outreach Index) (Consumption3* Outreach Index) (Consumption4* Outreach Index) (Consumption5* Outreach Index) Teledensity Urbanization Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
3.698
0.728
0.000
0.001 -0.005
0.544 0.000 0.000
TE D
0.007 -0.023 -2.377
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.000
Coefficient
129026 4151.55 0.000 -105100000 0.126
AC C
EP
Note: Regression analysis includes all Indian states.
39
Model II Marginal Effect 0.050 0.155 0.303 0.522 -0.031 -0.127 0.069 0.083 0.111 0.192 -0.036 0.030 -0.030 -0.050 -0.021 0.062
RI PT
Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Rural
p-value
0.254 0.792 1.552 2.674 -0.158 -0.652 0.354 0.423 0.570 0.985 -0.185 0.151 -0.152 -0.256 -0.105 0.317
SC
Model I Marginal Effect 0.065 0.116 0.180 0.300 -0.034 -0.124 0.070 0.085 0.111 0.193 -0.035 0.029 -0.028 -0.050 -0.021 0.060
Coefficient
M AN U
Explanatory Variables
6.226 0.136 -1.393 -3.352 -5.276 0.011 -0.020 -2.950
1.216 0.027 -0.272 -0.655 -1.030 0.002 -0.004 129026 4553.99 0.000 -104500000 0.131
p-value 0.009 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.788 0.004 0.000 0.000 0.314 0.000 0.000
ACCEPTED MANUSCRIPT
Table 9: Availability of Banking Services and Use of Formal Financial Services: Sensitivity Analysis II
0.329 0.586 0.909 1.512 -0.172 -0.619 0.356 0.434 0.562 0.980 -0.163 0.145 -0.149 -0.289 -0.111 0.303
Outreach Index PCA (Consumption2* Outreach Index PCA) (Consumption3* Outreach Index PCA) (Consumption4* Outreach Index PCA) (Consumption5* Outreach Index PCA) Teledensity Urbanization Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
2.580
Coefficient
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.276 0.735 1.396 2.389 -0.158 -0.646 0.353 0.424 0.568 0.982 -0.177 0.149 -0.155 -0.281 -0.111 0.317
0.509
0.000
0.009 -0.004
0.000 0.000 0.000
129026 4147.88 0.000 -105300000 0.124
TE D
0.046 -0.021 -2.140
p-value
AC C
EP
Note: Regression analysis includes all Indian states.
40
Model II Marginal Effect 0.054 0.144 0.273 0.467 -0.031 -0.126 0.069 0.083 0.111 0.192 -0.035 0.029 -0.030 -0.055 -0.022 0.062
RI PT
Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Rural
Model I Marginal Effect 0.065 0.116 0.179 0.298 -0.034 -0.122 0.070 0.086 0.111 0.193 -0.032 0.029 -0.029 -0.057 -0.022 0.060
SC
Coefficient
M AN U
Explanatory Variables
6.091 0.015 -1.756 -4.019 -6.051 0.037 -0.019 -2.609
1.191 0.003 -0.343 -0.786 -1.184 0.007 -0.004 129026 4538.18 0.000 -104600000 0.130
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.979 0.002 0.000 0.000 0.001 0.000 0.000
ACCEPTED MANUSCRIPT
Table 10: Contribution of Various Factor Groups to Pseudo-R2 Based on Shapley Decomposition Outreach Index
Household Factors
84.86
83.54
Availability of Banking Sector Regional/Supply-side Factors
4.41
4.40
10.73
12.06
Outreach Index (PCA)
RI PT
Bank Penetration
84.41 3.70
11.89
SC
Factors
Note: The regression models considered for the decomposition are Model I reported in Table 7, Table 8 and Table 9.
M AN U
Table 11: Income and Financial Inclusion No. of Respondents
No Account
Not Saved
Not Enough Money
Too far Away
Expensive
0- 20 %
398
52.26
78.14
58.80
23.15
27.22
20-40 %
594
51.35
70.20
54.26
20.82
47.55
40-60 %
669
50.52
58.59
48.73
24.08
34.96
60-80 %
609
80-100 %
730
TE D
Income Group
42.36
50.90
41.97
17.88
31.61
29.32
45.07
29.65
14.16
20.22
AC C
EP
Source: Global Financial Inclusion (Global Findex) Database 2014 Notes: ‘No Account’ and ‘Not Saved’ gives us the percentage of people in that income group having no account and no saving, respectively, in last 12 months. The remaining three column gives the percentage of people that gave mentioned reason for not having account.
41
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 1: Concentration Curve of Financial Inclusion and Lorenz Curve of Consumption Expenditure (All-India)
Source: Estimation based on All-India Debt and Investment Survey, 2002-03
AC C
EP
TE D
Figure 2: Concentration Curves of Financial Inclusion and Lorenz Curves of Consumption Expenditure for Rural and Urban Sectors in India
Source: Estimation based on All-India Debt and Investment Survey,2002-03
42
ACCEPTED MANUSCRIPT
November 10, 2018 Title:
Usage of Formal Financial Services in India: Demand Barriers or Supply
RI PT
Constraints? Abstract:
Estimating the role of both demand-side and supply-side factors in financial inclusion and its distribution is important for policy making. However, existing literature has primarily focused on
SC
supply-side factors. In this context, this paper estimates relative importance of removing demand-side barriers and eliminating supply constraints to enhance financial inclusion in India. It also measures the extent of concentration of usage of formal financial services among richer
M AN U
households. Results suggest that, while availability of banking services has a significant positive effect on usage of formal financial services, its contribution in inducing households to use formal financial services is very small compared to the contribution of factors, such as education, income, employment status, gender and social norms, that influence the demand for formal financial services. It highlights the importance of placing greater emphasis on addressing demand-side barriers, rather than on improving physical availability of banking services, to
TE D
promote financial inclusion in India.
Key words: Financial inclusion; income-related inequality; demand-side barriers; supply
EP
constraints; India
JEL classification: O16; O17; D39; G21; O53
AC C
Highlights:
1. Demand-side barriers, along with availability factors, affect financial inclusion 2. Removal of demand barriers is much more important than relaxing supply constraints 3. Bank outreach contributes below 5% in explained variation in financial inclusion 4. Gender and caste based discrimination in financial inclusion is less in urban area 5. Self-employed households in urban area rely more on informal financial services Authors Ordering: Abhishek Kumar, Rama Pal, Rupayan Pal.
S0264-9993(17)31753-4
DOI:
https://doi.org/10.1016/j.econmod.2018.11.010
Reference:
ECMODE 4766
To appear in:
Economic Modelling
Received Date: 27 November 2017 Revised Date:
11 November 2018
Accepted Date: 11 November 2018
Please cite this article as: Kumar, A., Pal, R., Pal, R., Usage of formal financial services in India: Demand barriers or supply constraints?, Economic Modelling (2018), doi: https://doi.org/10.1016/ j.econmod.2018.11.010. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Usage of Formal Financial Services in India: Demand Barriers or Supply Constraints?
Abhishek Kumar is PhD student at Indira Gandhi Institute of Development Research (IGIDR), Mumbai. His research interests are in macroeconomics and applied microeconomics and Bayesian methods.
SC
RI PT
Dr. Rama Pal is Assistant Professor of Economics at the Department of Humanities and Social Sciences, IIT Bombay. Her primary research area is Health Economics. Currently, she is working on inclusive growth and development, financial inclusion, migration, health inequality and child labour. International academic journals such as Journal of International Development, Oxford Development Studies, European Journal of Development Research, International Journal of Health Care Finance and Economics, and others have published her research papers.
AC C
EP
TE D
M AN U
Dr. Rupayan Pal, Professor of Economics at IGIDR, India, is an applied micro- economist specializing in the area of strategic interactions among economic agents. His theoretical research has dealt with issues including pricing of network goods, R&D, privatization, environmental regulation, union-oligopoly bargaining and tax competition. He has also researched on some of the policy relevant empirical issues, such as financial inclusion, unionization and tax efficiency. International academic journals such as Mathematical Social Sciences, Journal of Public Economic Theory, Regional Science & Urban Economics, Journal of Policy Modeling, Resource and Energy Economics, Economic Modelling, and others have published his research papers.
ACCEPTED MANUSCRIPT
Usage of Formal Financial Services in India:
November 10, 2018 Abstract:
RI PT
Demand Barriers or Supply Constraints?
SC
Estimating the role of both demand-side and supply-side factors in financial inclusion and its distribution is important for policy making. However, existing literature has primarily focused on
M AN U
supply-side factors. In this context, this paper estimates relative importance of removing demand-side barriers and eliminating supply constraints to enhance financial inclusion in India. It also measures the extent of concentration of usage of formal financial services among richer households. Results suggest that, while availability of banking services has a significant positive effect on usage of formal financial services, its contribution in inducing households to use formal financial services is very small compared to the contribution of factors, such as education,
TE D
income, employment status, gender and social norms, that influence the demand for formal financial services. It highlights the importance of placing greater emphasis on addressing demand-side barriers, rather than on improving physical availability of banking services, to
EP
promote financial inclusion in India.
Key words: Financial inclusion; income-related inequality; demand-side barriers; supply
AC C
constraints; India
JEL classification: O16; O17; D39; G21; O53
ACCEPTED MANUSCRIPT
1. Introduction After the financial crisis, there is a renewed interest in understanding financial development and its effects on economic growth and inequality. Literature provides evidence of positive effect of
RI PT
financial development1 and not necessarily of financial liberalization on growth. Studies find that financial liberalization may not have a direct impact on growth (Demetriades and Luintel, 1996 and 1997) and inequality (Rajan and Zingales, 2003 and 2004; and Ang 2010). However, financial liberalization encourages financial development (Demetriades and Luintel, 1996 and 1997) which in turn facilitates growth and reduces inequality. In contrast, Agnello et al. (2012)
SC
report evidence in support of financial reforms using a panel of 62 countries and claims that financial reforms reduce inequality.2 Cecchetti and Kharroubi (2012) and Chong et al (2016)
M AN U
report a non-linear relation between financial development and growth and suggest that beyond a certain level financial development hurts growth. Further, it is argued that effective financial intermediation plays crucial role in moving households out of poverty – indirectly by stimulating growth (trickle down) and directly by providing savings and credit services to the poor (Burgess and Pande 2005; Kochar 2011; Jeanneney and Kpodar, 2011).
TE D
While the empirical literature on finance and development largely focuses on depth of financial system, the distribution of access to formal financial services has drawn attention of the researchers only recently (Beck et al 2007). Studies in this strand of literature show that easy access to credit reduces poverty (Basu and Mallick, 2008). Similarly, Banerji et al. (2014) report
EP
positive effect of access to Kisan Credit Card on agricultural and infrastructure output and thus, on per capita income of districts in Bihar, India. Using firm level data from 80 developed and
AC C
developing countries around the world, Ayyagari et al. (2008) demonstrate that lack of access to finance restricts the growth of firms significantly. Further, lack of access to adequate finance disproportionately hurts smaller firms (Beck et al. 2005). While studies have mainly focused on impact of greater availability of financial services on growth and productivity, Ahamed and 1
Usually measured with financial deepening indicators such as bank deposit liabilities to GDP ratio (Levine, 2005).
2
Agnello et al. (2012) use the measure of financial liberalization given by Abiad and Mody (2005) and Abiad et al. (2010). This measure is sum of six indicators of financial liberalization, namely, credit controls, interest rate controls, entry barriers, regulations, privatization and international transactions.
1
ACCEPTED MANUSCRIPT
Mallick (2017) extend the discussion and shows that higher availability is also beneficial for the bank stability. Thus, overall there is consensus regarding positive effects of availability of financial services on productivity, growth and equality.
RI PT
Given this positive relation between access to financial services and economic wellbeing, developing countries like India have started giving importance to financial inclusion. The Reserve Bank of India (RBI) defines the financial inclusion as ‘a process of ensuring access to appropriate financial products and services needed by all sections of the society in general and
SC
vulnerable groups such as weaker sections and low income groups in particular at an affordable cost in a fair and transparent manner by mainstream institutional players’ (RBI, 2014). In order
M AN U
to achieve the objectives of financial inclusion, it is important to first measure the extent of financial inclusion. Most of the existing studies consider availability/access to finance as a measure of financial inclusion primarily due to dearth of adequate data on actual use of financial services by households and firms (Honohan, 2008; Bect et al. 2009; and Jeanneney and Kpodar, 2011). However, these outreach indicators of financial inclusion ignore other demand-side factors that may affect actual use of formal financial services such as socio-economic
TE D
background of households, and cost of using formal financial services. One may not use formal financial services, even if there is access, because of low reservation cost of informal financial services and/or high relative prices of financial services compared to the prices of other goods (Kochar 1997, Claessens 2006). Prevailing social/group norms might also play some role to this
EP
effect. Thus, availability of formal financial services is a necessary, but not sufficient condition for use of formal financial services. Only those who use formal financial services can be referred
AC C
as financially included, rest are financially excluded3. Thus, it is more appropriate to consider the data on use of formal financial services, not the mere access to finance.
3
We note here that, in theory, sometimes distinction is being made between two types of financial exclusion. The set of economic agents/households who have access but do not use formal financial services are referred as ‘voluntarily excluded’, and those who do not have access to formal financial services are refereed as ‘involuntarily excluded’. However, given the unavailability of individual/household level data on both use and access, it is not possible to make such distinctions operational.
2
ACCEPTED MANUSCRIPT
In this context, the present paper aims at measuring financial inclusion for India based on usage of formal financial services by Indian households. For this purpose, we use a large-scale household level data for India. We define household to be financially included if it uses either formal credit services or formal savings services or both. Not only formal credit but formal
RI PT
saving services have significant impact on long term asset growth in an economy (Kaboski and Townsend 2005; Ashraf et al 2006). Moreover, access to formal savings services enables the poor to make productive investment, to be less vulnerable to shocks and to smooth consumption expenditure (Dupas and Robinson 2009). To understand the extent of financial inclusion in India,
SC
we examine proportion of households using various financial services and find out how many are left out of the purview of formal financial services. Moreover, it is also required to understand
M AN U
the distribution of such usage across income groups to understand the association of financial inclusion with economic status of household. The present paper attempts to understand both the extent of usage to formal financial services by Indian households and inequality in its distribution across income group. This inequality in financial inclusion is measured using the concentration index and is termed as income-related inequality of financial inclusion. Keeping in mind diversity of the country, we explore the variation across states and sectors, in the extent of
TE D
usage and in income related inequality of financial inclusion. We further examine the factors affecting the household’s usage to formal financial services. Here, we consider both the household level factors as well as the state level supply side factors, i.e., availability of banks, that may affect financial inclusion. We use the Shapley decomposition method (Shorrocks, 1999
EP
and 2013), to understand the relative importance of these demand and supply-side factors in
AC C
determining financial service usage.
The choice of India for the purpose of this analysis rests on several considerations. First, India is an emerging economy that has experienced a rapid economic growth rate since the initiation of major economic reform in 1991. However, income inequality has widened over time and still a large segment of the population lives in poverty (Datt and Ravallion 2002; Deaton and Drèze 2002; Dev and Ravi 2007; Suryanarayana 2008; Cain et al 2010). Therefore, a careful examination of the extent of financial inclusion across income groups in countries like India is of paramount importance. Second, India’s federal structure allows states with partial policy 3
ACCEPTED MANUSCRIPT
autonomy and there is diverse pattern of growth and development across states, which makes it more interesting to examine the issue of financial inclusion in India at the sub-national level. Third, India is one of the few emerging economies for which large scale representative survey
RI PT
data on use of financial services at household level is available.
The empirical findings show low level of financial inclusion in India, particularly among the poor households. The extent of financial inclusion as well as income-related inequality in financial inclusion varies across the states and sectors. Interestingly, while the percentage of
SC
financially included households is lower in the rural sector, the distribution of financially included households is more evenly distributed as compared to the urban sector. It might be
M AN U
because social norms affect rural households’ decision making differently from their urban counterparts. These findings suggest that the different policy measures are warranted in the rural and urban sector. While in the rural sector, higher importance should be on increasing overall access to the financial services, in the urban sector, policy measures should concentrate on improving the access to the poor. Further, while financial inclusion is found to be higher in states with higher average per capita consumption expenditure, neither average per capita consumption
TE D
expenditure nor inequality in consumption expenditure is correlated with income-related inequality in financial inclusion at the state level.
Household-level regression analysis shows that availability of banking services is positively
EP
associated with financial inclusion, along with higher impact for the poor households. Moreover, the association of availability of banking services with the propensity to use financial services by
AC C
a household is stronger than that of other state level factors. Apart from the availability of banking services, most of the household characteristics have significant impact on the probability of financial inclusion. Household characteristics will help us in identifying vulnerable households who are likely to be excluded and we find that the poor, labourers and households belonging to the socially deprived classes are more likely to be financially excluded compared to their respective counterparts in both the rural and urban sectors. Further, self-employed households from the urban areas are less likely to be financially included as compared to the salaried households. This result indicates that small business families in the urban areas are more 4
ACCEPTED MANUSCRIPT
likely to rely on informal sources to meet their need of financial services. It calls for policies to reduce such dependency of small business on informal financial services in order to create a level playing field, which is necessary for employment generation and growth. Moreover, we find discrimination based on gender and social class in the rural areas where households with
RI PT
female head and belonging to socially deprived classes are less likely to be financially included. This discrimination might be result of the overall discrimination that is observed in Indian society (particularly in the rural areas) against these vulnerable groups. Special policies are required to ensure equal opportunity of accessing formal financial services for these vulnerable
SC
groups.
M AN U
Interestingly, while both demand-side barriers and supply constraints hinder the process of financial inclusion, results of Shapley decomposition analysis reveal that demand-side factors play a much greater role compared to supply side factors in inducing households to use formal financial services in India. It points out the importance of placing greater emphasis on removing demand-side barriers than on increasing physical availability of banks in order to foster effective
TE D
financial inclusion in India.
The rest of the paper is organized as follows. The next section describes methodologies to estimate income-related inequality of financial inclusion and presents the econometric model. Section 3 describes the data, variables and steps of estimation. Results are presented in Section 4.
EP
Section 4.1 plots the concentration curves of financial inclusion, estimates concentration index of financial inclusion for sub-national regions in India and analyzes its correlation with the level of
AC C
financial inclusion and other developmental indicators. Section 4.2 discusses the findings from the econometric analysis of the household’s propensity to be financially included and provides estimates of relative importance of removing demand-side barriers and eliminating supply constraints in order to induce households to use formal financial services. Section 5 concludes.
5
ACCEPTED MANUSCRIPT
2. Methodology 2.1 Measurement of Degree of Financial Inclusion Income related inequality in financial inclusion can be appropriately measured using the concentration curves and concentration indices. The concentration curve is the generalized
RI PT
Lorenz curve (Kakwani, 1977).4 In the present case, the concentration curve plots cumulative percentage of households that use formal financial services against cumulative percentage of households arranged according to economic status of household. We consider monthly per capita expenditure (MPCE) as the proxy for economic status. The distance between line of
SC
equality, shown by the 45º line, and the concentration curve shows the inequality. On the other hand, the concentration index (CI) measures the degree of this inequality. The CI is a bivariate measure that quantifies inequality in use of formal financial services across the distribution of
M AN U
households based on their economic status. Let x be the variable representing economic status, i.e. MPCE, of household and y = g(x) be the utilization of formal financial services by the household with MPCE x, where the function y = g(x) is defined as follows.
TE D
1, if the household uses at least one of the formal financial services y = g (x) = 0, otherwise The concentration index of financial inclusion (C) is given by:
2 = − 1 … … … . (1)
EP
where is binary 0 and 1 as explained above, µ = E( y) is the average financial inclusion, is
AC C
the number of household and is fractional rank. Wagstaff (2005) shows that when y i is binary,
1 the minimum value and the maximum value of the concentration index, C, are ( µ − 1 + ) and n 1 (1 − µ + ) , respectively, (see appendix at the end for detailed derivation). For large samples, as n in the present case, the (1/n) term tends to zero, thus, the lower and upper bounds of C can be
4
Initial application of the concentration curves may be found in Mahalanobis (1960).
6
ACCEPTED MANUSCRIPT
considered as ( µ − 1) and (1 − µ ) , respectively. It is clear that the bounds of concentration index shrinks as the average value increases. It implies that the concentration index for financial inclusion may turn out to be low, if the average number of financially included households is high, even though inequality in financial inclusion is high. So, to meaningfully compare the
RI PT
estimates of the concentration index for different regions, it is suggested to divide the original concentration index by (1 − µ ) (Wagstaff, 2005, 2009). The modified concentration index ( CW ) can be written as follows.
SC
2 − 1 … … … . (2) = (1 − )
M AN U
It is evident that, for any value of µ , we have − 1 ≤ CW ≤ 1 . Therefore, for the purpose of the present analysis, we consider the formulation of the concentration index as given by (2).5
Note that lower absolute value of the modified concentration index, CW , of financial inclusion indicates lesser extent of income related inequality in financial inclusion. And, negative
TE D
(positive) value of the concentration index of financial inclusion implies that the use of formal financial services is concentrated among the poorer (richer) households.
2.2 Econometric Model
EP
Next, we want to understand the factors that determine use of formal financial services and examine whether demand- or supply-side factors have more impact on utilization. For this
AC C
purpose, we estimate the logit model for utilization of formal financial services. The reduced form model depicting the use of formal financial services by households can be written as follows.
Yi * = α
+
Z i/ β
+
u i , i = 1, 2…n;
(3)
eu i yi = I (Yi > 0), F (ui ) = . 1 + eui *
5
We note here that there is a growing literature on appropriate formulation of concentration index in the case of binary variable (Erreygers, 2009; 2009b; Wagstaff, 2009; Brekke and Kverndokk, 2012).
7
ACCEPTED MANUSCRIPT
The net benefit of household i from using formal financial services is captured by the latent variable Yi * , and whether household i uses formal financial service or not is indicated by the dummy variable i. I (Yi * > 0) is an indicator function taking the value one if Yi * > 0 and the
RI PT
value zero otherwise. F (u i ) is the cumulative distribution function (CDF) of the error term u i , which has logistic distribution. Z i is the vector of exogenous explanatory variables (state-level factors and household characteristics), α is the constant term, and β is the vector of unknown
SC
parameters to be estimated. Clearly, the probability that household i uses formal financial
eα +Zi β /
pi = E( yi = 1| Zi ) =
1 + eα +Zi β /
.
M AN U
services is
We estimate model (3), considering alternative sets of explanatory variables, by maximum likelihood method with robust standard errors, and compute marginal effects of explanatory variables,
∂ pi ∂ xj
eα + Zi β /
=
(1 + e α + Zi β ) 2 /
β j (j = 1, 2...K), for continuous variables, and discrete changes
TE D
for dummy variables from zero to one.6
EP
3. Data, Variables and Estimation 3.1. Data
We use the household level data from All India Debt and Investment Survey (AIDIS) for the
AC C
year 2002-03. This dataset was collected as a part of the 59th round of the National Sample Survey (NSS), conducted by the Central Statistical Organization, Ministry of Statistics and Program Implementation, Government of India. This representative survey data for India provides micro-level information on households’ savings and borrowings along with socioeconomic characteristics of households.7 Based on availability of data, we estimate income-
6
As indicated by the methodology of the survey, we use sampling weights to compute descriptive statistics and to estimate the model. 7 For more information on the dataset see GoI (2005).
8
ACCEPTED MANUSCRIPT
related inequality in financial inclusion for all the states and the union territories (UTs) in India, but for the econometric analysis we leave out the UTs. We note here that the 29 states together cover more than 99 (98) percent of land-area (population) of India.8
RI PT
3.2. Dependent Variable
A household is termed financially included if it uses any of the formal financial product else financially excluded.9 This categorical variable is dependent variable in regression analysis. It may be noted that the households that do not use any of the formal financial services, can be
SC
referred as completely financially excluded. However, households in the first group are not necessarily fully financially included, since the need for formal financial services of some
M AN U
households in the first group may be met only partially by the formal financial services10. For simplicity of exposition, in this paper, we refer the first group of households as financially included households. Thus, the variable of interest in the present analysis is an indicator variable taking value one if a household use at least one of the formal financial services and the value zero otherwise. We report the percentage of households using at least one of the financial services across states &UTs and sectors, rural and urban, in Table 1 & 2. It is evident that the
8
TE D
extent of financial inclusion is alarmingly low in both rural (34.9%) and urban (49.2%) India.
9
EP
Latest All India Debt and Investment Survey (AIDIS) from 70th round doesn’t contain data on consumption expenditure which makes it unsuitable for the exercise being done in this paper. Therefore, we have tried to substantiate our two main results i.e. poor people are less likely to be financially included and banking presence increases financial inclusion especially among poor people using the Global Financial Inclusion (Global Findex) Database 2014. It is clear from Table 11 that poor people are less likely to have account in any financial institution and more of poor people complain about less supply of financial services.
AC C
The data set provides detailed information on sources of borrowings and savings instruments used by the households. Out of fifteen different sources of borrowings provided in the data set, we consider the institutional sources of borrowings, namely, government, cooperative society, commercial bank, insurance, provident fund, financial corporation, financial company and other institutional agencies, to estimate utilization of formal credit services. Similarly, households’ savings with the institutional agencies, namely, government or reserve bank of India certificates and bonds, deposits in post office, cooperative society/bank, commercial bank and non-banking company, and insurance premium, annuity certificates and provident fund, are considered as formal savings. If a household reports either borrowing from at least one institutional source or savings with at least one institutional agency or both, we categorize that household in the first group – ‘households utilizing formal financial services’.
10
We also note here that, though there are differences between access and use of formal financial services, most of the existing studies have considered these two synonymously, primarily due to dearth of adequate data on actual use of financial services by households and firms (Honohan 2008, Beck et al 2009, Jeanneney and Kpodar 2011).
9
ACCEPTED MANUSCRIPT
Moreover, there is substantial variation (range is 55%) in terms of usage of formal financial services across states & UTs in India.
3.3. Explanatory Variables
RI PT
The determinants of the household use of formal financial services can be classified broadly into two categories, namely, household characteristics and state level factors, as indicated in Table 3. Definitions of explanatory variables are detailed in the Appendix. Table 4 presents the summary statistics of the variables used in the present analysis. In order to alienate the effects of household
SC
income clearly, we consider MPCE quintile groups, and include dummy variables for the last four groups, Consumption2 – Consumption5, in the set of explanatory variables, while
M AN U
Consumption1 is the bottom quintile.11 Apart from the income, employment type (i.e. source of earnings) of a household might also affect its probability to use formal financial services. We can categorize the households according to the type of primary employment, which is the main source of income, into three groups, namely, salaried, self-employment and casual labor. We consider the dummy variables Self Employed and Laborer as explanatory variables (Salaried serves as the base category) in the regression. Summary statistics shows that 48.8 percent
TE D
households earn their livelihood through self-employment, which is the largest in size out of three broad categories (Table 4). In contrast, only 21 percent households’ principal earnings come from salaried employment. The probability to use formal financial services is expected to be lowest (highest), if a household is primarily dependent on casual labor supply (salaried
EP
employment).
AC C
In order to control for possible effect of household’s education on use of formal financial services, we construct the variables Illiterate, Literate, Primary, Middle and Secondary and above, based on highest level of education of household head. Summary statistics reveals that in Indian states only 20 percent households’ heads were having secondary or higher level of education in 2002-03; whereas 40.8 percent households have illiterate head. It is also expected that female headed households are more vulnerable and are less likely to be financially included, 11
Our results are not sensitive to the consideration of alternative categorizations (e.g. deciles) of household according to MPCE or to the use of MPCE directly.
10
ACCEPTED MANUSCRIPT
compared to the male headed households. Therefore, we expect to have negative effect of the variable Female on probability to use formal financial services. In order to control for the size of the households, we consider the variable Household Size as an explanatory variable.
RI PT
Households belonging to socially disadvantageous communities are expected to have lower propensity to be financially included, because of possible discrimination against them in the formal sector and/or due to societal culture and the extent of discrimination could be different in rural and urban areas. In India socially disadvantageous communities are identified on the basis
SC
of castes, namely, scheduled castes (SC); scheduled tribes (ST); and other backward classes (OBC). Remaining households are considered as general caste (General). So we use dummy
M AN U
variables SC, ST and OBC as explanatory variables in the regression to assess the impact of social groups. To control for the possible effect of the place of residence, we consider the dummy variable Rural while estimating the model using full sample, and estimate the model separately for rural and urban sectors.
Moreover, we link the household-level data to state-level measures of (a) availability of banking
TE D
services, (b) level of infrastructural facilities and (c) urbanization. Data on banking services is collated from Basic Statistical Returns of Scheduled Commercial Banks, published by the Reserve Bank of India. Data on per capita net state domestic product (NSDP) come from CSO. The source of data on state-wise tele-density is the annual report of the Department of
EP
Telecommunications, Ministry of Communications & Information Technology, Government of India. state-wise population, land-area and urbanization data come from Census of India for the
AC C
year 2001.
We consider the variable Bank Penetration, which is defined as the number of bank offices per lakh population, as a measure of availability of banking services. In order to avoid possible endogeneity problem, we use one year lagged values of Bank Penetration in the regression. Higher Bank Penetration implies greater access to formal financial services. Therefore, it is expected that Bank Penetration would be positively associated with the household use of formal financial services. For instance, Beck et al. (2007) demonstrate that indicators of outreach of 11
ACCEPTED MANUSCRIPT
banking services are positively associated with hard-to-collect micro-level statistics on household use of formal financial service. However, to the best of our knowledge, the issue of whether such positive association is uniform across different income groups of households or not has not received much attention in the literature so far. It is often argued that there is need to
RI PT
enhance the supply of banking services in order to promote financial inclusion particularly among the poor. However, for such policies to be effective, availability of banking services should have differential impact on use of formal financial services by the households across income groups, with positive bias to the relatively poor households. In order to examine whether
SC
that is indeed true or not, we introduce interaction variables (Bank Penetration)* Consumptionj, j = 2, 3, 4, and 5, in the regression. Table 4 shows that in the sample states the average number
M AN U
of bank offices per lakh population was 7.79, which seems to be quite low. Also, there was considerable variation in terms of Bank Penetration across states in India. Bank Penetration was lowest (3.57) in Nagaland and highest (24.64) in Goa during the period of study.
It may be argued that the variable Bank Penetration controls for only one aspect of availability of banking services, namely, demographic penetration of bank offices. And, thus, geographic bank
TE D
penetration, which is defined as number of bank offices per thousand square kilometer area, percapita availability of deposit services and per-capita availability of credit services should also be considered as indicators of outreach of banking services, which measures different aspects of availability of banking services. To check the robustness of our results we consider two
EP
alternative variables, (a) Outreach Index and (b) Outreach Index PCA, in place of Bank Penetration in separate regressions. Outreach Index and Outreach Index PCA are two indices of
AC C
banking sector’s outreach. Each of these two indices combine the above mentioned four indicators, including demographic penetration, into a single measure of banking sector’s outreach by employing (i) the method of Chakravarty and Pal (2013) and (ii) principal component analysis, respectively, as described in the Appendix.12 Both the indices are bounded between zero and one, where value closer to one indicates higher availability of banking
12
The correlation coefficient of ranks of states based on Bank Penetration and Outreach Index (Bank Penetration and Outreach Index PCA) is 0.8473 (0.8690), which is significant at one percent level. Therefore, it seems that Bank Penetration is a good proxy of availability of banking services.
12
ACCEPTED MANUSCRIPT
services. The average values (standard deviations) of Outreach Index and Outreach Index PCA were, respectively, 0.226 and 0.144 (0.161 and 0.138) in the year 2001 (see Table 4). Clearly, the overall measures of banking sector’s outreach also indicate that, during the period of study, supply of banking services was quite low on an average and varied widely across states in India.
RI PT
We also control for possible implications of the level of infrastructural facilities and extent of urbanization in a state, by considering the variables Tele-density and Urbanization as explanatory variables.
SC
3.4. Estimation of the Econometric Model
First, we estimate the model given in equation (3) by considering the explanatory variables
M AN U
pertaining to household characteristics using full sample, results of which are reported in Table 5. Second, we estimate the model separately for rural and urban sectors, to check whether association of household characteristics with the dependent variable varies across sectors (Table 6). In both the first and second regressions, we control for state specific unobserved factors using dummy variables for states appropriately. Third, we drop the state dummies and introduce the state level factors, Bank Penetration, Tele-density and Urbanization, in the regression (see
TE D
Model I in Table 7). It helps us to examine the association of specific state level variables of interest with the dependent variable. 13 Fourth, we consider interaction of the variable Bank Penetration with the dummy variables Consumption2, Consumption3, Consumption4 and Consumption5, in order to examine whether there is any differential impact of availability of
EP
banking services on use of formal financial services by households from different income class (see Model II in Table 7). Since required data to measure availability of banking services in rural
AC C
and urban sectors separately is not available, we consider the full sample to estimate the model when state-level variables are also considered as explanatory variables. Fifth, to check the sensitivity of our results to the measure of availability of banking services considered, we reestimate the last two specifications of the model by replacing the variable Bank Penetration with alternative measures of availability of banking services, Outreach Index and Outreach Index
13
Needles to mention here, we cannot consider state level variables along with the state dummies, since that would lead to the problem of multicollinearity.
13
ACCEPTED MANUSCRIPT
PCA, separately (see Table 8 and Table 9). In order to avoid possible endogeneity problem, we consider lagged values of the state level variables in the regression.14
3.5. Estimation of Relative Importance of Demand and Supply
RI PT
Finally, in order to understand the relative importance of demand- and supply-side factors which influence formal financial service utilization, we carry out the Shapley decomposition analysis. The explanatory variables are grouped into three categories, namely, (1) household characteristics, representing the demand side, (2) availability of banking services, representing
SC
the supply side; and (3) other region-specific variables. The Shapley method decomposes the pseudo R-square15 and estimates the contribution of each of the categories to total variation in
M AN U
the financial inclusion (Shorrocks, 1999 and 2013).
4. Empirical Findings
In this section, we first analyze the distribution of utilization of formal financial services across economic status of household. Next, we examine various factors affecting utilization of formal
TE D
financial services.
4.1. Concentration Curves and Estimates of Concentration Index of Financial Inclusion across States and Sectors in India
First, we plot the concentration curve of financial inclusion. In order to understand the relative
EP
difference between inequalities in financial inclusion and in economic wellbeing, we also plot
AC C
the Lorenz curve for MPCE along with the concentration curve of financial inclusion.
Note that the concentration curve of financial inclusion for all India lies everywhere below the 45o line, but it lies everywhere above the Lorenz curve for MPCE (see Figure 1). It implies that
14
We note here that possible endogeneity problem may not be fully avoided by considering the lagged values of explanatory variables always. However, note that the focus of this analysis is to examine the association of statelevel explanatory variables with the dependent variable, not the causal effects. Therefore, it seems that the issue of endogeneity does not deserve much attention in the present context. 15 We use the ‘shapley2’ software package in STATA for decomposition analysis (Juarez, 2012).
14
ACCEPTED MANUSCRIPT
there is income related inequality in financial inclusion in India, and the income-related inequality in financial inclusion is lower than inequality in consumption expenditure.
Further, it seems that both income related inequality in financial inclusion and inequality in
RI PT
MPCE differ across sectors, rural and urban, in India. In particular, it may be observed from Figure 2 that (a) inequality in MPCE is higher in urban India than that in rural India and (b) the concentration curve of financial inclusion is relatively closer to the Lorenz curve of MPCE for rural India as compared to that for urban India. It seems to indicate that income related inequality
SC
in financial inclusion closely follow inequality in MPCE in rural sector, but not so in urban sector. However, from Figure 2 it appears to be difficult to assess the relative position of rural
M AN U
India in terms of income related inequality in financial inclusion vis-à-vis that of urban India from the concentration curves. In order to address this issue we estimate the modified concentration index of financial inclusion ( CW ), using the formula as given by equation (2), and compare the values of CW for rural and urban India.
Table 1 reports the estimates of concentration index of financial inclusion, along with the
TE D
percentage of households using formal financial services, average MPCE and Gini coefficient of MPCE, for India as well as for each of the 35 states and UTs in India. Estimated concentration index of financial inclusion for all India is found to be 0.345, which implies that financial inclusion is concentrated among richer households. In other words, financial exclusion is
EP
disproportionately higher among the relatively poor households compared to their richer counterparts. It is also evident that there is considerable variation in terms of income related
AC C
inequality in financial inclusion across Indian states and UTs. In 14 out of 35 states and UTs the estimated concentration index is greater than that for all India. The concentration index is highest, 0.624, in Meghalaya; which is followed by Chandigarh, Sikkim, Mizoram, Manipur and Gujarat. Whereas, it is lowest in Dadra Nagar Haveli; which is followed by Andaman and Nicobar Island, Arunachal Pradesh, Uttaranchal, Kerala and Haryana. Interestingly, in Dadra Nagar Haveli financial inclusion is more concentrated among poorer households, which is clearly an exception. However, the extent of income related inequality in financial inclusion is quite low, -0.066, in Dadra Nagar Haveli. 15
ACCEPTED MANUSCRIPT
Also, note that although only 22.4 percent of households in Arunachal Pradesh use formal financial services, income related inequality in financial inclusion is quite low in this state. At the same time, in Chandigarh, 66.9 percent households use formal financial services, but its
RI PT
estimated concentration index is as large as 0.612. Therefore, it seems that increase of the proportion of households using formal financial services in a state/UT need not necessarily reduce inequality in financial inclusion across income groups or foster financial inclusion among the poor households in that state/UT. There is no significant relation between the percentage of
SC
households using formal financial services and concentration index of financial inclusion (the correlation coefficient between these two variables is negative, -0.28, but not significant at 10
M AN U
percent level).
We also find that the correlation coefficients between concentration index of financial inclusion and average MPCE is negative, but very low (-0.017) and not significant at 10 percent level. That is, income related inequality in financial inclusion does not appear to be significantly associated with overall economic wellbeing. However, there is significant (at one percent level)
TE D
and positive correlation between the percentage of households using formal financial services and average MPCE across states and UTs, which seems to be in line with the results from crosscountry studies (Honohan, 2004; Beck et al., 2009). Therefore, it seems that higher level of economic wellbeing is positively associated with higher level of financial inclusion on an
AC C
households.
EP
average, but not necessarily with higher level of financial inclusion among the relatively poor
Further, we find that there is very weak correlation between concentration index of financial inclusion and Gini coefficient of MPCE (the correlation coefficient is -0.037, which is not significant at 10 percent level). Nonetheless, high (low) income inequality may coexist with low (high) income related inequality in financial inclusion in some cases, as observed in case of Uttaranchal and Arunachal Pradesh (Manipur, Meghalaya and Punjab). It seems to indicate that the policy measures that are effective to reduce income inequality may not prove to be equally effective in reducing income related inequality in financial inclusion. 16
ACCEPTED MANUSCRIPT
These results show that there is correlation between financial inclusion and economic wellbeing. However, improving economic wellbeing does not immediately imply higher usage of formal financial services by both rich and poor. So, in order to achieve higher financial inclusion by
RI PT
needy, we have to examine factors other than income that facilitate financial inclusion. We analyze such factors at the household level and report the results in the next sub-section.
Considering rural sector and urban sector separately, we find that both the percentage of
SC
households using formal financial services and the concentration index of financial inclusion are lower in rural India compared to that in urban India (see Table 2). This is similar to the rural-
M AN U
urban comparisons of average MPCE and Gini coefficient of MPCE in India. The utilization rate, i.e., percentage of households using formal financial services, and concentration index are, respectively, about 41 percent and 29 percent higher in urban India compared to that in rural India. Clearly, there is rural-urban divide both in terms of use of formal financial services and income related inequality in financial inclusion. Nonetheless, more than 50 percent households
TE D
are completely excluded from the formal financial system even in urban India.
As before, there is considerable variation both in terms of percentage of households using formal financial service and concentration index of financial inclusion in both rural and urban sectors across states and UTs. It seems to be important to note that the range of concentration index of
EP
financial inclusion across states and UTs is much higher in rural sector (0.89) compared to that in urban sector (0.78). Also note that, the percentage of households using formal financial services is higher in urban sector than in rural sector of each of the states and UTs, except in Delhi, as in
AC C
case of all India.16 However, concentration index of financial inclusion is higher in rural sector, compared to that in urban, of ten states and UTs (Dadra Nagar Haveli, Delhi, Jammu & Kashmir, Jharkhand, Lakshadweep, Maharashtra, Meghalaya, Mizoram, Nagaland and Sikkim), and in Manipur there is no difference between the sectors. Also, only in three UTs (Andaman and Nicobar Island, Chandigarh and Dadra Nagar Haveli) the concentration index of financial
16
In Delhi the percentage figure is marginally in favor of the rural sector.
17
ACCEPTED MANUSCRIPT
inclusion is negative. That is, only in these three UTs utilization of formal financial services is concentrated among relatively poor households.
We also find that the correlation coefficient between the concentration index of financial
RI PT
inclusion and the percentage of households using formal financial services is negative, -0.48, and significant (at one percent level) in the case of rural sector. However, in urban sector there is no significant correlation between these two variables, as observed in the case of both combined. It implies that increase in overall financial inclusion can have differential consequence on income
SC
related inequality in financial inclusion across these two sector.
M AN U
There is positive and significant, at one percent level, correlation between average MPCE and percentage of households using formal financial services in both rural sectors (0.55) and in urban sector (0.52), as in the case of both the sectors combined (0.56). However, there is no significant (at ten percent level) correlation (a) between concentration index of financial inclusion and average MPCE and (b) between concentration index of financial inclusion and Gini coefficient of MPCE, even when we consider rural and urban sectors separately. These results strengthen the
TE D
arguments that, though higher level of per capita income may be positively associated with higher level of financial inclusion, (a) higher level of per capita income in a region need not necessarily be associated with lower level of income related inequality in financial inclusion and (b) income related inequality in financial inclusion should be treated differently from income
EP
inequality.
AC C
4.2. Demand-side and Supply-side Factors associated with Household’s Propensity to Use Formal Financial Services: What Matters More? Given the presence of high income related inequality in use of financial services, it becomes imperative to analyze the factors that determine use of financial services. The focus of this subsection is twofold, first, to examine whether greater availability of banking services is linked to higher utilization of financial services, particularly among the poor households and second, to understand the relative importance of supply-side availability of banking services in determining actual utilization.
The empirical findings show that availability of banking services does 18
ACCEPTED MANUSCRIPT
increase the utilization and the effect is higher for the poorer households (Table 7 – Table 9). However, the relative contribution of banking services in the Pseudo-R2 (explained variation) is very small around four percent, whereas the household characteristics together account for the
(Table 10). We explain the results below in detail.
RI PT
largest portion in the total explained variation in the utilization of formal financial services
The econometric analysis gives several interesting results. First and foremost, availability of banking services is positively associated with a household’s propensity to use formal financial
SC
services. The estimated coefficient of the Bank Penetration variable is positive and significant at one percent level (Model I in Table 7) and thus bank increases the probability of financial
M AN U
inclusion. Moreover, we find that the marginal effects of (Consumption3*Bank Penetration), (Consumption4*Bank Penetration) and (Consumption5*Bank Penetration) are negative and significant, but the marginal effect of (Consumption2*Bank Penetration) is insignificant; while the marginal effects of Consumption2, Consumption3, Consumption4, Consumption5 and Bank Penetration continue to be positive and significant (see Model II in Table 7). Therefore, we can say that presence of banks increases the probability of financial inclusion especially among poor
TE D
people. Recent policies of government of India of taking banks to the people seems logical in light of the above finding. These results hold true, if we consider Outreach Index or Outreach Index PCA in place of Bank Penetration (see Model I and Model II in Table 8 & 9). In other words, our results are robust to consideration of alternative measures of availability of banking
EP
services.
AC C
Household’s propensity to use formal financial services increases as its economic status improves. The coefficients of the dummy variables Consumption2– Consumption5 are positive and significant, and the difference between the marginal effects of two consecutive MPCE quintile groups is greater for higher quintile groups (Table 5). These results remain valid, if we estimate the model separately for rural and urban sectors or consider alternative specifications of the model (Table 6 – Table 9). That is, this result indicates that larger proportion of poor households do not use formal financial services, compared to that of relatively rich households, even after controlling for the other factors. It strengthens the findings of Section 3 that there is 19
ACCEPTED MANUSCRIPT
income related inequality in financial inclusion and larger proportion of relatively poor households are financially excluded compared to their richer counterparts in India.
The econometric analysis also shows that there is positive (negative) and significant association
RI PT
of the level of infrastructural facilities (urbanization) in a state with the probability to use formal financial services by a household in that state. Interestingly, we find that the marginal effect of availability of banking services is greater than that of level of infrastructural facilities. This result indicates stronger association of availability of banking services with household’s propensity to
SC
use formal financial services as compared to other infrastructural facilities.
M AN U
We also find that the coefficient of the dummy variable Rural is positive and significant (see Table 5). However, descriptive statistics in Table 2 shows that lower proportion of rural households are financially included compared to that of urban households. Therefore, it seems that the degrees of association of sector specific unobserved factors are positively biased towards rural sector. Breaking the dependent variable into two, we observe that formal savings (credit) services are utilized by 40.05 percent and 21.38 percent (13.54 percent and 19.45 percent) of
TE D
urban and rural households, respectively. That is, while higher percentage of urban households use formal savings services than that of rural households, utilization of formal credit services is greater in the rural sector compared to urban sector. Therefore, it seems that rural focus of government’s credit policy (such as, priority sector lending norms, Kisan credit card scheme,
EP
etc.) has facilitated financial inclusion in rural sector vis-à-vis urban sector.
AC C
Let us now turn to the associations of other characteristics of households with their propensity to use formal financial services. We find that the coefficients of Self Employed and Laborer are negative and significant at one percent level, and the absolute value of the marginal effect of Laborer is greater than that of Self Employed (see Table 5 – Table 9). It confirms commonly held perceptions that (a) principal economic activity of a household plays crucial role in determining whether that household would use formal financial services or not and (b) the propensity to use formal financial services is highest for households with salaried employment as the main source
20
ACCEPTED MANUSCRIPT
of income and lowest for households who are primarily dependent on casual labor supply, ceteris paribus.
The econometric analysis also reveals that the probability of a household to be financially
RI PT
included increases at an increasing rate with the increase in level of education of the household head. In other words, as expected, greater proportion of households with relatively lower level of education are financially excluded. Interestingly, we find that the marginal effect of Household Size is always positive and significant. It implies that, ceteris paribus, the probability to use
SC
formal financial services is higher for larger households as the probability of using formal financial services by at least one member of a household increases with the increase in number of
M AN U
members.
Negative and significant marginal effect of the dummy variable Female, when we estimate the model using full sample, seems to imply that the propensity of female headed households to use formal financial services is less than the male headed households in India (see Table 5). However, separate estimates for the two sectors, rural and urban, reveals that the gender of
TE D
household head plays significant role only in rural India, not in urban India (see Table 6).17 In other words, unlike as in the case of rural India, female headed households are not significantly different from male headed households as far as utilization of formal financial services is concerned. We find similar results for households belonging to socially disadvantageous
EP
communities. The marginal effects of SC, ST and OBC are negative and significant, if we consider both rural and urban sector together or consider only rural sector. On the other hand, in
AC C
the case of urban sector, marginal effects of SC and ST turn out to be insignificant, while the marginal effect of OBC is negative and significant at 5% level. It implies that, in urban India, the propensity to use formal financial services is lower for OBCs as compared to the general category households. However, in rural India the propensity to use formal financial services is the lowest for ST households, followed by SC households and OBC households, ceteris paribus (see Table 6).
17
The marginal effect of Female in case of urban sector is negative but not significant at 5% level.
21
ACCEPTED MANUSCRIPT
There are differences between the two sectors, rural and urban, in terms of magnitudes of marginal effects of other household characteristics as well. It is easy to observe that marginal effects of Consumption2 – Consumption5 are greater in urban sector than that it rural sector. It justifies the higher concentration index of financial inclusion in urban India compared to that in
RI PT
rural India, as observed in Section 3. Degrees of association of employment status and household size with the propensity to use formal financial services are also stronger in urban sector compared to that in rural sector. Rural-urban comparison of the marginal effects of dummy variables for educational levels leads to mixed results. We find that the marginal effects of
SC
Primary, Middle, and Secondary and above are higher in urban sector than in rural sector, but the
M AN U
marginal effect of Literate is higher in rural sector than in urban sector.
The above results show that various household-specific factors along with supply-side variables affect household’s utilization of formal financial services. The Shapley decomposition analysis gives the relative contributions of these factors in predicting the utilization. We decompose the pseudo- (explained variation 18 ) and calculate the contribution of
three broad groups of
variables, household level variables, availability of banking services and other region-specific
TE D
factors in pseudo- . We carry out this analysis for all the three specifications with different indicators of availability banking services of Model 1 in as reported in table: 7, 8, 919. We find that the relative contribution of availability of banking services is lowest at 4.4 percent when we consider either bank penetration or outreach index (Table 10). On the other hand, the household
EP
characteristics, representing demand side barriers, contribute between 83.54 and 84.86 percent depending on the model20. The results are consistent across models with very less variation in the
18
AC C
contributions of different groups. Thus, the decomposition analysis points out the importance of There is unexplained variation of financial inclusion too but we can’t decompose that as it is not being explained by the model. We can only decompose the proportion of explained variation to get the source of it. 19
For decomposition analysis models having interaction terms between demand-side (consumption categories) and supply-side (banking penetration/ outreach) factors are not considered, because it is not feasible to attribute contribution of those interaction terms to either of the two sides – demand or supply. 20
It may be noted that the supply-side factors are considered at the state-level due to the non-availability of data at more granule level. It is likely that the contribution of the supply-side factors will change if more localized information on physical access is available.
22
ACCEPTED MANUSCRIPT
demand-side factors rather than physical availability of banking services as the major barriers to actual utilization of formal financial services.
5. Concluding Remarks
RI PT
In this paper we have analyzed the distribution and determinants of usage of formal financial service across sub-national regions in India using a representative household level survey data, linked to state-level factors. The main contributions of this paper are threefold. First, to the best of our knowledge, this is the first attempt to study distribution of usage of formal financial
SC
services across income using concentration indices. Second, it provides empirical evidence of the household’s propensity to use formal financial services in a developing country using both
M AN U
demand and supply-side factors. We also estimate their relative contribution in determining the utilization of formal financial services. Third, it provides an assessment of the role of banking services to promote financial inclusion, particularly among the poor households.
We find that the extent of financial inclusion is low, particularly among the poor, in India. We also find that there are considerable variations in terms of concentration index of financial
TE D
inclusion and the level of financial inclusion across Indian states and UTs as well as across sectors – rural and urban. Both the level of financial inclusion and estimated concentration index of financial inclusion are lower in rural sectors compared to that in urban sectors. Nonetheless, about a half of the urban households are financially excluded. Our findings seem to indicate that
EP
increase in level of financial inclusion can have differential consequence on income related
AC C
inequality in financial inclusion across sectors.
Though there is positive correlation between overall economic wellbeing (average MPCE) and the percentage of financially included households at state level, there is no significant correlation of concentration index of financial inclusion with overall economic wellbeing or with the percentage of financially included households.
We also demonstrate that income related
inequality in financial inclusion deserves special attention of the policy makers.
23
ACCEPTED MANUSCRIPT
Results of our econometric analysis demonstrate that availability of banking services is positively associated with a household’s propensity to use formal financial services and, in particular, the degree of association is higher for the poor and marginally non-poor households. It seems to indicate that greater availability of banking services can foster financial inclusion,
RI PT
particularly among the poor. Moreover, compared to the availability of infrastructural facilities and extent of urbanization in a state, availability of banking services in a state is more strongly associated with the propensity of a household to be financially included. These results are robust
SC
to alternative measures of availability of banking services.
We also find that the level of education, employment status and household size are significantly
M AN U
associated with the use of formal financial services by a household in both rural and urban sectors. Rural-urban comparison indicates that income and employment status have stronger association with urban household’s propensity to be financially included, compared to that of a rural household. An important finding is that self-employed households from the urban areas are less likely to be financially included as compared to the salaried households, which indicates that small business families in the urban areas are more likely to rely on informal sources to meet
TE D
their need of financial services.
Interestingly, we find that the gender of household head and social group are not significantly associated with use of formal financial services by urban households. However, in rural India,
EP
the probability to use formal financial services by socially deprived classes and female headed households is significantly lower than their counterparts. In urban India, only the OBC
AC C
households have higher probability to be financially excluded compared to the households belonging to other social groups including general category households.
We have evidence suggesting that both demand- and supply-side factors affect the probability of utilizing formal financial services. However, household characteristics representing demand side, contribute the largest part in determining the usage, while physical availability of banking services plays a very minor role. This result emphasizes policy measures to eliminate the demand-side barriers along with increasing availability of banking services. 24
ACCEPTED MANUSCRIPT
It may be noted that this paper considers household as financially excluded if that household does not use any of the financial services provided by institutional agencies; otherwise that
RI PT
household is considered as financially included. That is, we do not distinguish between partial and complete financial inclusion of a household. In order to assess the extent of financial inclusion of a household, it is important to appropriately estimate the need of different financial services by that household, which is beyond the scope of the present paper. It might also be
SC
interesting to assess the implications of government policies on dynamics of socio-economic inequality in financial inclusion. We leave these for future research.
M AN U
Acknowledgements:
References
TE D
The authors would like to thank the editor and referees for insightful comments. Authors would also like to thank Satya R. Chakravarty and the participants in the ‘Workshop on Measuring Financial Inclusion from the Demand Side’, conducted by the Center for Advanced Financial Research and Learning (CAFRAL) at NIBM, Pune, for comments and helpful discussions. An earlier draft of the paper was circulated as “Rama Pal & Rupayan Pal, 2012. Income related inequality in financial inclusion and role of banks: Evidence on financial exclusion in India, IGIDR-WP-2012-013, Indira Gandhi Institute of Development Research (IGIDR), India.” Usual disclaimer applies.
AC C
EP
Abiad, A., Detragiache, E. and Tressel, T., 2010. A new database of financial reforms. IMF Staff Papers, 57(2), pp.281-302. Abiad, A. and Mody, A., 2005. Financial reform: What shakes it? What shapes it? American Economic Review, 95(1), pp.66-88. Agnello, L., Mallick, S.K. and Sousa, R.M., 2012. Financial reforms and income inequality. Economics Letters, 116(3), pp.583-587. Ahamed, M.M. and Mallick, S.K. 2017. Is Financial Inclusion Good for Bank Stability? International Evidence. Journal of Economics Behavior and Organization, forthcoming. Ang, J.B., 2010. Finance and Inequality: the Case of India. Southern Economic Journal, 76(3), pp.738-761. Ashraf, N., Karlan, D. and Yin, W., 2006. Tying Odysseus to the mast: Evidence from a commitment savings product in the Philippines. The Quarterly Journal of Economics, 121(2), pp.635-672. 25
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Ayyagari, M., Demirgüç-Kunt, A. and Maksimovic, V., 2008. How Important are Financing Constraints? The Role of Finance in the Business Environment. The World Bank Economic Review, 22(3), pp.483-516. Banerji, S., Gangopadhyay, K., Gupta, P. and Mallick S., 2014. Financial intermediation and economic development: Evidence from Bihar, in Ashima Goyal (ed.), Handbook of the Indian Economy in the 21st Century: Understanding the Inherent Dynamism, Oxford University Press, pp.898-929. Basu, S. and Mallick, S., 2008. When Does Growth Trickle Down to the Poor? The Indian Case. Cambridge Journal of Economics, 32(3), pp.461-477. Beck, T., Demirgüç-Kunt, A. and Honohan, P., 2009. Access to financial services: Measurement, impact, and policies. The World Bank Research Observer, 24(1), pp.119-145. Beck, T., Demirgüç-Kunt, A. and Levine, R., 2007. Finance, inequality and the poor. Journal of economic growth, 12(1), pp.27-49. Beck, T., Demirgüç‐Kunt, A.S.L.I. and Maksimovic, V., 2005. Financial and legal constraints to growth: Does firm size matter?. The Journal of Finance, 60(1), pp.137-177. Brekke, K.A. and Kverndokk, S., 2012. Inadequate Bivariate Measures of Health Inequality: The Impact of Income Distribution. The Scandinavian Journal of Economics, 114(2), pp.323333. Burgess, R. and Pande, R., 2005. Do Rural Banks Matter? Evidence from the Indian Social Banking Experiment. American Economic Review, 95(3), pp.780-795. Cain, J.S., Hasan, R., Magsombol, R. and Tandon, A., 2010. Accounting for Inequality in India: Evidence from Household Expenditures. World Development, 38(3), pp.282-297. Chakravarty, S. R., and Pal, R., 2013. Measuring Financial Inclusion: An Axiomatic Approach. Journal of Policy Modeling, 35(5), pp. 813-83. Cecchetti, S., and Kharroubi, E., 2012. Reassessing the Impact of Finance on Growth. BIS Working Paper No 381, Bank for International Settlements. Chong, E., and Varela Sandoval, F., 2016. Does Financial Development Help Growth or Hurt It, or Neither? unpublished manuscript. Dabla-Norris E., M.E., Kochhar, M.K., Suphaphiphat, M.N., Ricka, M.F. and Tsounta, E., 2015a. Causes and consequences of income inequality: a global perspective. International Monetary Fund. Dabla-Norris, E., Ji, Y., Townsend, R.M. and Unsal, D.F., 2015 b. Distinguishing Constraints on Financial Inclusion and Their Impact on GDP, TFP, and Inequality (No. w20821). National Bureau of Economic Research. Datt, G. and Ravallion, M., 2002. Is India's economic growth leaving the poor behind?. The journal of economic perspectives, 16(3), pp.89-108. Deaton, A. and Dreze, J., 2002. Poverty and inequality in India: a re-examination. Economic and political weekly, pp.3729-3748.
26
ACCEPTED MANUSCRIPT
SC
RI PT
Demetriades, P.O. and Luintel, K.B., 1996. Financial development, economic growth and banking sector controls: evidence from India. The Economic Journal, pp.359-374. Demetriades, P.O. and Luintel, K.B., 1997. The direct costs of financial repression: evidence from India. The Review of Economics and Statistics, 79(2), pp.311-320. Dev, S.M. and Ravi, C., 2007. Poverty and inequality: All-India and states, 19832005. Economic and Political weekly, pp.509-521. Dupas, P. and Robinson, J., 2013. Savings constraints and microenterprise development: Evidence from a field experiment in Kenya. American Economic Journal: Applied Economics, 5(1), pp.163-192. Erreygers, G., 2009. Correcting the concentration index. Journal of health economics, 28(2), pp.504-515.
AC C
EP
TE D
M AN U
Erreygers, G., 2009 b. Correcting the concentration index: a reply to Wagstaff. Journal of Health Economics, 28(2), pp.521-524. Government of India., 2005. Household assets and liabilities in India: NSS 59th round. National Sample Survey Organization, Ministry of Statistics and Programme Implementation. New Delhi. Honohan, P., 2004. Financial Development, Growth and Poverty: How Close are. Financial development and economic growth: Explaining the links, p.1. Honohan, P., 2008. Cross-country variation in household access to financial services. Journal of Banking & Finance, 32(11), pp.2493-2500. Jeanneney, S.G. and Kpodar, K., 2011. Financial development and poverty reduction: Can there be a benefit without a cost?. The Journal of development studies, 47(1), pp.143-163. Juarez, F.C. 2012. SHAPLEY2: Stata Module to Compute Additive Decomposition of Estimation Statistics by Regressors or Groups of Regressors. Statistical Software Components S457543, Boston College Department of Economics, revised 17 June 2015. Kaboski, J.P. and Townsend, R.M., 2005. Policies and Impact: An Analysis of Village‐Level Microfinance Institutions. Journal of the European Economic Association, 3(1), pp.1-50. Kakwani, N.C., 1977. Applications of Lorenz curves in economic analysis. Econometrica: Journal of the Econometric Society, pp.719-727. Kochar, A., 2011. The distributive consequences of social banking: a microempirical analysis of the Indian experience. Economic Development and Cultural Change, 59(2), pp.251-280. Levine, R., 2005. Finance and Growth: Theory and Evidence. Handbook of Economic Growth, 1, pp.865-934. Mahalanobis, P.C., 1960. A method of fractile graphical analysis. Econometrica: Journal of the Econometric Society, pp.325-351. Rajan, R.G. and Zingales, L., 2003. The great reversals: the politics of financial development in the twentieth century. Journal of financial economics, 69(1), pp.5-50.
27
ACCEPTED MANUSCRIPT
RI PT
Rajan, R.G. and Zingales, L., 2004. Saving capitalism from the capitalists: Unleashing the power of financial markets to create wealth and spread opportunity. Princeton University Press. RBI 2014 Financial Intermediation for All – Economic Growth with Equity, URL: https://rbi.org.in/scripts/FS_Speeches.aspx?Id=912&fn=2754. Last accessed: May 03, 2018. Shorrocks, A. F. 1999. Decomposition Procedures for Distributional Analysis: a Unified Framework Based on the Shapley Value. Unpublished Manuscript. Department of Economics, University of Essex.
Shorrocks, A. F. 2013. Decomposition Procedures for Distributional Analysis: a Unified
SC
Framework Based on the Shapley Value. Journal of Economic Inequality. 11(1). pp. 99126.
AC C
EP
TE D
M AN U
Suryanarayana, M.H., 2008. What Is Exclusive about ‘Inclusive Growth’?. Economic and Political Weekly, pp.93-101. Wagstaff, A., 2005. The bounds of the concentration index when the variable of interest is binary, with an application to immunization inequality. Health economics, 14(4), pp.429432. Wagstaff, A., 2009. Correcting the concentration index: a comment. Journal of Health Economics, 28(2), pp.516-520.
28
ACCEPTED MANUSCRIPT
APPENDIX Derivation of Concentration Index
RI PT
Let x be the variable representing economic status, i.e. MPCE, of household and y = g(x) be the utilization of formal financial services by the household with MPCE x, where the function y = g(x) is defined as follows.
SC
1, if the household uses at least one of the formal financial services y = g (x) = 0, otherwise
Let F(x) be the distribution function of MPCE and f(x) be the probability density function of MPCE. Now, the
M AN U
proportion of households having MPCE less than or equal to x and using formal financial services, out of total number of households using formal financial services, can be represented by F1[ g ( x)] as follows.
() =
(
1 () () !, ()
where () = ) () ()! represents expected number of of households using formal financial services.
TE D
The relation between () and F(x) can be called as the concentration curve of y, i.e., concentration curve of financial inclusion (see Kakwani, 1977).
AC C
EP
From the properties of distribution functions, we have:
lim () = 0
→
lim () = 1
→(
And one can write
! () () () = ! ()
Following Kakwani (1977), we define the concentration index for financial inclusion as one minus twice the area under the concentration curve for . Therefore, our concentration index for financial inclusion is given by (
= 1 − 2 () ()! .
29
ACCEPTED MANUSCRIPT
Using integration by parts ( ( ! () = 1 − 2 * () ()! − + , - ()!. ! / !
⟹ = 1 − 2 *1 − ⟹ =1−2+
(
Now, () = ) () ()! = ,
( 2 () ()()! ()
( 2 () ()()! ()
( 2 () * () ()()! − / () 2
(
( 1 () ()()! / ()
M AN U
⟹=
() () ()!/ ()
SC
⟹ = −1 +
(
RI PT
⟹ = 1 − 2 * ()()|( −
and 345(), () = ) :() − 6()7; :() − ; ()!
( 1 ( ⟹ 345(), () = ()() ()! − () ()! 2
TE D
(
( ( 1 − 6()7 () ()! + (()) ()! 2
Since ) () ()! = we can write above expression as
( 1 345(), () = ()() ()! − (()) 2
EP
AC C
Using this we can write
=
2 8345(), ()9 ()
One can write the discrete case of the above formulae as
where
µ = E ( y) ,
=
2 345,
r denotes the individual’s fractional rank in the MPCE distribution and n is number of
households in the sample.
30
ACCEPTED MANUSCRIPT
Using the fact that ̅ = 1/2
∑ ( − )( − 1/2)
345, =
∑ − ∑ − ∑ +
⟹ 345, =
1 1 1 1 − − + 2 2 2
SC
⟹ 345, =
∑ ( − =)( − ̅ )
RI PT
345, =
1 1 ⟹ 345, = − 2
Therefore, = : ∑ − ; E
M AN U
=
2 − 1
For a given > 0, concentration index is maximum when poorest @ individuals are not included financially i.e. for =
TE D
them = 0 and for rich − @ individuals = 1. Thus we have, =
EP
AC C
an can be written as
2 1 A 0 + 0 + 0 + 0 … + @ + 1 + @ + 2 + @ + 3 + ⋯ + D − 1 −@
Using the sum of natural numbers from @ + 1 to :
GH
2 @+1 @+2 @+3 A 0 + 0 + 0 + 0 … + + + + ⋯ + 1 D − 1 −@
⟹=
Using =
GH
=
@+1+ −1
, we can write:
= 1−+
1
Minimum value of can be derived in the same way and is given by = − 1 +
31
ACCEPTED MANUSCRIPT
APPENDIX (Continued) EXPLANATORY VARIABLES: DEFINITIONS AND DATA SOURCES Variable
Definition
Data Source
Household Size
ST
OBC
Place of Residence: Rural (Ref: Urban)
-do-
-do-
-do-
-do-
The variable ‘Household Size’ is defined as the number of members in the household -do-
It takes the value 1 if household belongs to General category; otherwise 0. It takes the value 1 if household belongs to Scheduled Caste; otherwise 0. It takes the value 1 if household belongs to Scheduled Tribes; otherwise 0. It takes the value 1 if household belongs to Other Backward Castes; otherwise 0.
AC C
Social Group: General (Base category) SC
AIDIS, 2002-2003
It takes the value 1 if the household head is female; otherwise 0.
EP
Gender: Female (Ref: Male)
TE D
M AN U
SC
RI PT
Household Characteristics Income: Consumption1- Consumption i takes the value 1 if the household belongs to the ith consumption quintile group based on MPCE; otherwise 0. Consumption5 (Base category: Consumption1) Employment Status: Salaried (Base It takes the value 1 if the household’s principal earning is from category) salaried employment; otherwise 0. Self Employed It takes the value 1 if the household’s principal earning is from self employment; otherwise 0. Laborer It takes the value 1 if the household’s principal earning is from casual/agricultural/non-agricultural labor; otherwise 0. Education: Illiterate(Base It takes the value 1 if head of household is illiterate; otherwise 0. category) otherwise Literate It takes the value 1 if head of household is literate; otherwise 0.otherwise Primary It takes the value 1 if highest education completed by head is primary; otherwise 0. Middle It takes the value 1 if the highest education completed by head is middle level; otherwise 0. Secondary and It takes the value 1 if the highest education completed by head is above secondary level or above; otherwise 0.
-do-
It takes the value 1 if household resides in rural area; otherwise 0.
32
ACCEPTED MANUSCRIPT
APPENDIX (Continued) Variable
Definition
Data Source
The variable ‘Outreach Index’ is the index of outreach of banking services, which is constructed by considering (a) number of bank offices per lakh population, (b) number of bank offices per thousand square kilometer area, (c) the number of deposit accounts per thousand population and (d) number of credit accounts per thousand people as indicators of outreach of banking services employing the formula, as in Chakravarty and Pal (2013), r
SC
Outreach Index
Basic Statistical Returns of Scheduled Commercial Banks, Reserve Bank of India. (Data on population and land area comes from Census of India).
RI PT
State Level Factors Availability of Banking Services: Bank The variable ‘Bank Penetration’ is measured as number of bank Penetration offices per lakh population. (1 lakh =100000)
M AN U
1 k x − mi (Outreach Index 1) = ∑ i , where r=0.75, k=4, k i =1 M i − mi xi denotes the value of the i-th indicator, mi (Mi) denotes the minimum (maximum) value of the i-th indicator.
The variable ‘Teledensity’, which is a proxy for infrastructural facility, is defined as the number of telephones per hundred population.
AC C
EP
Infrastructure: Teledensity
The variable ‘Outreach Index PCA’ is the index of outreach of banking services constructed by employing well-known principal component analysis (PCA), by considering the following four indicators of outreach of banking services: (a) number of bank offices per lakh population, (b) number of bank offices per thousand square kilometer area, (c) the number of deposit accounts per thousand population and (d) number of credit accounts per thousand people.
TE D
Outreach Index PCA
Urbanization
The variable ‘Urbanization’ is measured as the percentage of urban population to total population.
33
Department of Telecommunication, Ministry of Communications & Information Technology, Government of India. Census of India
ACCEPTED MANUSCRIPT
Table 1: State-wise Estimates of Concentration Index of Financial Inclusion Average MPCE
Gini Coefficient
0.092 0.267 0.112 0.355 0.298 0.612 0.374 0.243 0.277 -0.066 0.262 0.465 0.195 0.291 0.444 0.393 0.296 0.184 0.228 0.307 0.490 0.624 0.503 0.428 0.351 0.318 0.449 0.438 0.234 0.532 0.293 0.339 0.259 0.179 0.328 0.345
1047.52 635.78 632.22 563.94 426.96 1497.54 443.60 958.21 1175.82 955.89 992.37 771.03 824.98 773.95 501.23 771.78 670.52 792.60 996.32 752.03 644.76 724.32 983.13 535.37 964.52 402.24 885.21 894.05 607.19 710.64 719.84 561.01 516.17 1034.07 598.99 634.90
0.226 0.322 0.364 0.264 0.241 0.372 0.313 0.350 0.321 0.195 0.346 0.321 0.287 0.300 0.320 0.229 0.310 0.300 0.197 0.343 0.191 0.230 0.229 0.305 0.204 0.320 0.329 0.281 0.278 0.308 0.321 0.287 0.284 0.516 0.323 0.333
SC
RI PT
Concentration Index
M AN U
EP
Andaman &Nicobar (UT) Andhra Pradesh Arunachal Pradesh Assam Bihar Chandigarh(UT) Chhattisgarh Daman and Diu (UT) Delhi Dadra Nagar Haveli (UT) Goa Gujarat Haryana Himachal Pradesh Jharkhand Jammu &Kashmir Karnataka Kerala Lakshadweep(UT) Maharashtra Manipur Meghalaya Mizoram Madhya Pradesh Nagaland Orissa Pondicherry(UT) Punjab Rajasthan Sikkim Tamil Nadu Tripura Uttar Pradesh Uttaranchal West Bengal All-India
Percentage of Households Using Formal Financial Services 62.3 37.8 22.4 31.8 18.4 66.9 29.7 46.8 28.9 57.0 47.6 44.7 45.6 62.5 27.5 40.8 36.1 73.5 51.0 50.2 25.5 22.7 22.4 37.8 37.5 31.7 52.9 46.2 30.1 36.9 41.4 34.3 35.3 47.7 41.2 38.8
TE D
State/UT
AC C
‘(UT)’ indicates union territory. Source: Estimation based on All-India Debt and Investment Survey,2002-03
34
ACCEPTED MANUSCRIPT
Table 2: State-wise and Sector-wise Estimates of Concentration Index of Financial Inclusion
M AN U
TE D
AC C
Gini Coefficient
78.4 43.2 41.1 63.1 37.0 75.7 52.9 83.9 29.2 77.7 64.3 58.6 53.8 74.7 57.8 61.9 51.4 73.2 71.4 63.1 47.3 44.3 49.6 57.1 60.7 52.7 53.2 52.2 45.3 71.4 49.2 61.3 42.1 64.4 57.2 49.2
Average MPCE
0.196 0.252 0.316 0.234 0.199 0.213 0.258 0.330 0.202 0.159 0.332 0.266 0.232 0.279 0.226 0.205 0.220 0.257 0.225 0.263 0.174 0.193 0.206 0.225 0.157 0.240 0.245 0.259 0.220 0.277 0.255 0.234 0.231 0.551 0.242 0.268
0.654 0.421 0.297 0.481 0.336 0.823 0.710 0.541 0.259 0.043 0.756 0.600 0.441 0.481 0.235 0.137 0.426 0.201 0.208 0.320 0.450 0.416 0.403 0.544 0.224 0.490 0.494 0.517 0.361 0.358 0.379 0.615 0.426 0.535 0.363 0.374
1238.23 958.91 1112.09 916.79 717.31 1544.69 764.75 1720.71 1243.22 1070.18 1191.72 1055.23 1127.92 1254.07 923.00 1021.12 989.50 1046.22 989.91 1053.02 764.43 1110.46 1159.90 856.64 1170.14 830.68 1007.84 1064.30 896.94 1144.14 1007.87 1005.93 755.44 920.27 955.49 970.85
0.246 0.356 0.487 0.269 0.311 0.374 0.348 0.315 0.323 0.252 0.330 0.314 0.332 0.276 0.302 0.219 0.306 0.348 0.171 0.305 0.195 0.219 0.230 0.325 0.226 0.358 0.345 0.292 0.303 0.321 0.336 0.291 0.329 0.325 0.337 0.331
RI PT
Utilization*
946.20 520.83 557.57 520.04 391.15 1028.39 384.72 876.03 782.67 899.20 822.16 610.37 708.13 712.33 393.33 694.80 511.87 703.97 1004.22 524.59 598.89 653.17 874.86 432.46 848.08 334.71 672.43 800.88 516.72 640.41 577.12 489.17 448.47 1063.42 477.45 508.80
Concentration Index
Gini Coefficient
-0.130 0.202 0.010 0.278 0.257 -0.240 0.325 0.163 0.296 -0.082 0.080 0.427 0.099 0.294 0.287 0.442 0.227 0.134 0.390 0.367 0.450 0.592 0.620 0.327 0.353 0.266 0.358 0.430 0.094 0.651 0.216 0.338 0.209 0.112 0.270 0.290
SC
Average MPCE
74.0 32.6 20.3 31.3 24.3 60.7 36.6 49.1 29.6 64.3 61.9 40.3 47.0 62.1 27.4 42.4 36.5 66.5 55.4 53.9 19.8 17.2 22.6 36.5 33.6 34.6 48.2 44.4 31.3 46.6 36.4 43.7 36.1 52.2 40.7 34.9
EP
Andaman & Nicobar (UT) Andhra Pradesh Arunachal Pradesh Assam Bihar Chandigarh (UT) Chhattisgarh Daman and Diu (UT) Delhi Dadra Nagar Haveli (UT) Goa Gujarat Haryana Himachal Pradesh Jharkhand Jammu & Kashmir Karnataka Kerala Lakshadweep (UT) Maharashtra Manipur Meghalaya Mizoram Madhya Pradesh Nagaland Orissa Pondicherry (UT) Punjab Rajasthan Sikkim Tamil Nadu Tripura Uttar Pradesh Uttaranchal West Bengal All-India
Urban
Concentration Index
Rural Utilization*
State/UT
‘(UT)’ indicates union territory. * Utilization refers to percentage of households using formal financial services. Source: Estimation based on All-India Debt and Investment Survey,2002-03
35
ACCEPTED MANUSCRIPT
Table 3: Variables Associated with Household Use of Formal Financial Services Household Characteristics
Availability of Banking Services Infrastructure Urbanization
Income Employment Status Education Household Size
0.212 0.205 0.199 0.202 0.182 0.210 0.488 0.302 0.408 0.119 0.137 0.136 0.200 0.110 4.885 0.305 0.208 0.081 0.407 0.747 7.789 0.226 0.161 4.767 25.795
AC C
EP
Consumption1 Consumption2 Consumption3 Consumption4 Consumption5 Salaried Self Employed Laborer Illiterate Literate Primary Middle Secondary and above Female Household Size General SC ST OBC Rural Bank Penetration Outreach Index Outreach Index PCA Teledensity Urbanization Number of Observations
Standard Deviation 0.409 0.404 0.399 0.401 0.386 0.408 0.500 0.459 0.491 0.324 0.344 0.342 0.400 0.313 2.524 0.460 0.406 0.273 0.491 0.435 3.981 0.144 0.138 3.040 11.069
Minimum
Maximum
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 3.57 0.02 0.01 1.10 9.79 129026
1 1 1 1 1 1 1 1 1 1 1 1 1 1 35 1 1 1 1 1 24.64 0.64 0.61 15.25 49.77
M AN U
Mean
TE D
Variable
SC
Table 4: Summary Statistics of Explanatory Variables
36
Social Group Gender Place of Residence
RI PT
State-level Variables
ACCEPTED MANUSCRIPT
Table 5: Determinants of Use of Formal Financial Services by Households Marginal Effect 0.068 0.125 0.196 0.320 -0.036 -0.114 0.072 0.092 0.116 0.194 -0.018 0.029 -0.034 -0.075 -0.027 0.065
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.023 0.000 0.000 0.000 0.000 0.000 0.000
RI PT
Coefficient 0.343 0.628 0.983 1.606 -0.179 -0.571 0.363 0.460 0.585 0.977 -0.092 0.145 -0.169 -0.376 -0.136 0.328 -2.222
SC
Explanatory Variables Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Rural Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
M AN U
129026 3913.01 0.000 -106100000 0.118
Note: State dummies are included in the regression analysis.
Table 6: Sector Wise Determinants of Use of Formal Financial Services by Households Coefficient
Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
0.332 0.602 0.953 1.542 -0.024 -0.433 0.381 0.474 0.560 0.847 -0.122 0.127 -0.226 -0.445 -0.158 -1.864
Rural Marginal Effect 0.065 0.119 0.188 0.304 -0.005 -0.085 0.075 0.093 0.110 0.167 -0.024 0.025 -0.045 -0.088 -0.031
AC C
EP
TE D
Explanatory Variables
84411 2323.04 0.000 -78640822 0.096
p-value
Coefficient
0.000 0.000 0.000 0.000 0.613 0.000 0.000 0.000 0.000 0.000 0.014 0.000 0.000 0.000 0.000 0.000
0.451 0.847 1.225 1.861 -0.363 -0.769 0.365 0.500 0.750 1.271 0.054 0.206 0.027 -0.146 -0.075 -2.865
Urban Marginal Effect 0.091 0.171 0.247 0.375 -0.073 -0.155 0.074 0.101 0.151 0.256 0.011 0.042 0.005 -0.029 -0.015 44615 1902.36 0.000 -27123534 0.149
Note: State dummies are included in the regression analysis.
37
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.442 0.000 0.664 0.298 0.124 0.000
ACCEPTED MANUSCRIPT
Table 7: Availability of Banking Services and Use of Formal Financial Services
0.318 0.564 0.876 1.469 -0.170 -0.626 0.353 0.431 0.565 0.987 -0.163 0.144 -0.141 -0.332 -0.107 0.295
Bank Penetration
0.088
Consumption2* Bank Penetration Consumption3* Bank Penetration Consumption4* Bank Penetration Consumption5* Bank Penetration 0.052 -0.016 -2.490
0.017
0.010 -0.003
129026 4256.32 0.000 -105200000 0.125
TE D
Teledensity Urbanization Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
p-value
Coefficient
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.313 0.760 1.529 2.655 -0.160 -0.640 0.350 0.426 0.567 0.981 -0.166 0.145 -0.144 -0.345 -0.108 0.308
AC C
EP
Note: Regression analysis includes all Indian states.
38
0.000
0.000 0.000 0.000
Model II Marginal Effect 0.061 0.149 0.300 0.521 -0.031 -0.126 0.069 0.084 0.111 0.193 -0.033 0.029 -0.028 -0.068 -0.021 0.060
RI PT
Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Rural
Model I Marginal Effect 0.063 0.111 0.173 0.290 -0.034 -0.123 0.070 0.085 0.111 0.194 -0.032 0.028 -0.028 -0.065 -0.021 0.058
SC
Coefficient
M AN U
Explanatory Variables
p-value 0.019 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.168
0.033
0.000
-0.004 -0.039 -0.106 -0.174
-0.001 -0.008 -0.021 -0.034
0.857 0.057 0.000 0.000
0.057 -0.016 -2.995
0.011 -0.003
0.000 0.000 0.000
129026 4421.12 0.000 -104900000 0.127
ACCEPTED MANUSCRIPT
Table 8: Availability of Banking Services and Use of Formal Financial Services: Sensitivity Analysis I
0.331 0.589 0.914 1.523 -0.172 -0.628 0.357 0.432 0.562 0.983 -0.176 0.148 -0.145 -0.254 -0.105 0.303
Outreach Index (Consumption2* Outreach Index) (Consumption3* Outreach Index) (Consumption4* Outreach Index) (Consumption5* Outreach Index) Teledensity Urbanization Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
3.698
0.728
0.000
0.001 -0.005
0.544 0.000 0.000
TE D
0.007 -0.023 -2.377
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.000
Coefficient
129026 4151.55 0.000 -105100000 0.126
AC C
EP
Note: Regression analysis includes all Indian states.
39
Model II Marginal Effect 0.050 0.155 0.303 0.522 -0.031 -0.127 0.069 0.083 0.111 0.192 -0.036 0.030 -0.030 -0.050 -0.021 0.062
RI PT
Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Rural
p-value
0.254 0.792 1.552 2.674 -0.158 -0.652 0.354 0.423 0.570 0.985 -0.185 0.151 -0.152 -0.256 -0.105 0.317
SC
Model I Marginal Effect 0.065 0.116 0.180 0.300 -0.034 -0.124 0.070 0.085 0.111 0.193 -0.035 0.029 -0.028 -0.050 -0.021 0.060
Coefficient
M AN U
Explanatory Variables
6.226 0.136 -1.393 -3.352 -5.276 0.011 -0.020 -2.950
1.216 0.027 -0.272 -0.655 -1.030 0.002 -0.004 129026 4553.99 0.000 -104500000 0.131
p-value 0.009 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.788 0.004 0.000 0.000 0.314 0.000 0.000
ACCEPTED MANUSCRIPT
Table 9: Availability of Banking Services and Use of Formal Financial Services: Sensitivity Analysis II
0.329 0.586 0.909 1.512 -0.172 -0.619 0.356 0.434 0.562 0.980 -0.163 0.145 -0.149 -0.289 -0.111 0.303
Outreach Index PCA (Consumption2* Outreach Index PCA) (Consumption3* Outreach Index PCA) (Consumption4* Outreach Index PCA) (Consumption5* Outreach Index PCA) Teledensity Urbanization Constant No. Observations Wald chi2(18) Prob> chi2 Log pseudo likelihood Pseudo R2
2.580
Coefficient
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
0.276 0.735 1.396 2.389 -0.158 -0.646 0.353 0.424 0.568 0.982 -0.177 0.149 -0.155 -0.281 -0.111 0.317
0.509
0.000
0.009 -0.004
0.000 0.000 0.000
129026 4147.88 0.000 -105300000 0.124
TE D
0.046 -0.021 -2.140
p-value
AC C
EP
Note: Regression analysis includes all Indian states.
40
Model II Marginal Effect 0.054 0.144 0.273 0.467 -0.031 -0.126 0.069 0.083 0.111 0.192 -0.035 0.029 -0.030 -0.055 -0.022 0.062
RI PT
Consumption2 Consumption3 Consumption4 Consumption5 Self Employed Laborer Literate Primary Middle Secondary and above Female Household Size SC ST OBC Rural
Model I Marginal Effect 0.065 0.116 0.179 0.298 -0.034 -0.122 0.070 0.086 0.111 0.193 -0.032 0.029 -0.029 -0.057 -0.022 0.060
SC
Coefficient
M AN U
Explanatory Variables
6.091 0.015 -1.756 -4.019 -6.051 0.037 -0.019 -2.609
1.191 0.003 -0.343 -0.786 -1.184 0.007 -0.004 129026 4538.18 0.000 -104600000 0.130
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.979 0.002 0.000 0.000 0.001 0.000 0.000
ACCEPTED MANUSCRIPT
Table 10: Contribution of Various Factor Groups to Pseudo-R2 Based on Shapley Decomposition Outreach Index
Household Factors
84.86
83.54
Availability of Banking Sector Regional/Supply-side Factors
4.41
4.40
10.73
12.06
Outreach Index (PCA)
RI PT
Bank Penetration
84.41 3.70
11.89
SC
Factors
Note: The regression models considered for the decomposition are Model I reported in Table 7, Table 8 and Table 9.
M AN U
Table 11: Income and Financial Inclusion No. of Respondents
No Account
Not Saved
Not Enough Money
Too far Away
Expensive
0- 20 %
398
52.26
78.14
58.80
23.15
27.22
20-40 %
594
51.35
70.20
54.26
20.82
47.55
40-60 %
669
50.52
58.59
48.73
24.08
34.96
60-80 %
609
80-100 %
730
TE D
Income Group
42.36
50.90
41.97
17.88
31.61
29.32
45.07
29.65
14.16
20.22
AC C
EP
Source: Global Financial Inclusion (Global Findex) Database 2014 Notes: ‘No Account’ and ‘Not Saved’ gives us the percentage of people in that income group having no account and no saving, respectively, in last 12 months. The remaining three column gives the percentage of people that gave mentioned reason for not having account.
41
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
Figure 1: Concentration Curve of Financial Inclusion and Lorenz Curve of Consumption Expenditure (All-India)
Source: Estimation based on All-India Debt and Investment Survey, 2002-03
AC C
EP
TE D
Figure 2: Concentration Curves of Financial Inclusion and Lorenz Curves of Consumption Expenditure for Rural and Urban Sectors in India
Source: Estimation based on All-India Debt and Investment Survey,2002-03
42
ACCEPTED MANUSCRIPT
November 10, 2018 Title:
Usage of Formal Financial Services in India: Demand Barriers or Supply
RI PT
Constraints? Abstract:
Estimating the role of both demand-side and supply-side factors in financial inclusion and its distribution is important for policy making. However, existing literature has primarily focused on
SC
supply-side factors. In this context, this paper estimates relative importance of removing demand-side barriers and eliminating supply constraints to enhance financial inclusion in India. It also measures the extent of concentration of usage of formal financial services among richer
M AN U
households. Results suggest that, while availability of banking services has a significant positive effect on usage of formal financial services, its contribution in inducing households to use formal financial services is very small compared to the contribution of factors, such as education, income, employment status, gender and social norms, that influence the demand for formal financial services. It highlights the importance of placing greater emphasis on addressing demand-side barriers, rather than on improving physical availability of banking services, to
TE D
promote financial inclusion in India.
Key words: Financial inclusion; income-related inequality; demand-side barriers; supply
EP
constraints; India
JEL classification: O16; O17; D39; G21; O53
AC C
Highlights:
1. Demand-side barriers, along with availability factors, affect financial inclusion 2. Removal of demand barriers is much more important than relaxing supply constraints 3. Bank outreach contributes below 5% in explained variation in financial inclusion 4. Gender and caste based discrimination in financial inclusion is less in urban area 5. Self-employed households in urban area rely more on informal financial services Authors Ordering: Abhishek Kumar, Rama Pal, Rupayan Pal.