Self help groups, microfinance, financial inclusion and social exclusion: Insight from Assam

Abstract

In order to link the unorganised with the formal financial sector in India, the National Bank for Agriculture and Rural Development (NABARD) introduced the Self-Help Group Bank Linkage Programme (SHG-BLP) as a trial initiative in 1992 and mainstreamed it in 1996. Microfinance services are available in Assam through SHG-BLPs registered with the Deendayal Antyodaya Yojana-National Rural Livelihoods Mission (DAY-NRLM). In central Assam, there are the most active SHG-BLP units. In light of this, the present study aims to explore contribution of SHG-BLP in financial along with the social inclusion of the marginalised rural people of central Assam. The impact of SHG-BLP is facilitated by the application of the propensity score matching method. The empirical results show that the SHG-BLP considerably reduces social exclusion among participants relative to their non-participant counterparts and ensures financial inclusion too. Based on this empirical finding, the study recommends that the coverage of SHG registration under DAY-NRLM be expanded. Simultaneously, efforts must be made to expand the alternative model of SHGs, the MFI-Bank Linkage Model, for expanding SHG coverage.

Highlights

• •The present study aims to explore the contribution of Self Help Groups- Bank Linkage Programme in financial along with the social inclusion of the marginalised rural people of central Assam.
• •Two self-developed indices are constructed to measure the level of financial inclusion and social exclusion of both the treated and control groups.
• •Indices are developed by utilizing Multiple Correspondence Analysis and the impact analysis of the Self Help Groups- Bank Linkage Programme is facilitated by using Propensity Score Matching method.
• •The empirical result ensures the positive role of the NABARD architected project- Self Help Groups- Bank Linkage Programme on financial as well as social inclusion for the participants compared to their non-participant's counterpart.
• •The study corroborates the successfulness of the NABARD's project- Self Help Groups- Bank Linkage Programme for the upliftment of the vulnerable and weaker section of the society, particularly in a backwards region.

1. Introduction

All the United Nations Member States adopted the 2030 Agenda for Sustainable Development in 2015 with the aim of prosperity for people and the planet, now and in the future [1]. The principles of the Sustainable Development Goals (SDG) are embedded in the 17 manifested goals. The SDGs aim to make the globe hunger- and poverty-free. Obviously, it is difficult to achieve such a goal for the poorest and most vulnerable sections of society without active institutional intervention. Thus, the agenda embraces broad targets aimed at promoting social justice and fostering inclusive and participatory decision-making for all. In a nutshell, the accomplishment of the SDGs implies that people everywhere, irrespective of their social status, will be socially included. Although there is no universal concurrence regarding the definition or benchmark of social exclusion [1], defines " … social exclusion as a state in which individuals are unable to participate fully in economic, social, political and cultural life, as well as the process leading to and sustaining such a state".

Thus, to examine how much SHG-BLP is successful in fulfilling its founding objectives, SHG-BLPs impact on simultaneous achievements in financial and social inclusion should be explored. Only this way, we can establish the prosperousness of the SHGs.

1.1. Earlier literature and existing research gap

Comprehensive studies concentrate on the efficacy of microfinance on employment and income generation, poverty reduction, asset creation, the social-economic empowerment of rural poor women, and the advancement of financial inclusion. However, the earlier studies lacked consensus in their findings.

[[2], [3], [4], [5], [6]], pinpoint microfinance as a successful instrument for reducing poverty and inequality among the participants. Earlier studies documented that Self-Help Group (SHG) led microfinance programme becomes a successful model for income and employment generation [[7], [8], [9]]. However, existing literature also shows the partial success of the SHG-led microfinance model [10,11]. The SHG bank linkage model helps in improving the formal institutional credit flow among landless and marginal farmers, discourages non-institutional borrowing through thrift creation, and results in a higher level of financial inclusion than that of non-members [[12], [13], [14]]. In fact, SHG-led microfinance also helps in reducing multidimensional poverty [15].

The theoretical foundation of SHG lies in the “Difference Principle” of “Rawls Theory of Justice, (1971)", which demands boosting the advancement of society's “least-advantaged” group. In light of that, the role of SHGs in promoting social inclusion must be examined to validate SHG operations. Unfortunately, no such attempt is ever made at the national or international levels. Thus, there exists a research gap. However, it is crucial to explore the role of SHGs to corroborate social inclusion. The same ideology is also manifested in the 17 embedded goals of the SDG. This idea is also reflected in the objectives of initiating SHG operations in India, namely, the upliftment of the poor, marginalised and vulnerable ruralities. Assam is a North-eastern state of India and SHGs were introduced from its beginning. The popular and only operational model of SHGs in Assam is SHG-BLP, in which all SHGs are currently registered under DAY-NRLM. The emphasis is thrust on SHG-BLP performance for accomplishing the goals and targets of the SDGs in India and Assam. Hence, it will be pertinent to explore the impact of SHGs simultaneously on financial inclusion and social exclusion, particularly for a backwards region of India, like Assam. Unfortunately, very few researchers in India are exploring this field, considering Assam. Studies by Refs. [16,17] are worth mentioning in this case.

Moreover, the theoretical foundation of the microfinance model is the “Rawls Theory of Justice, (1971)", in general and the “Difference Principle” of the same in particular. Accordingly, the essentiality of the SHG-BLP can only be examined through exploring its impact on the social inclusion of the participants. Unfortunately, the potency of SHGs for both financial and social inclusion has not yet been investigated in Assam or in any other state or country on the globe. Thus, there exists a research gap, and to shed light on this research gap, the present paper attempts to test the following hypotheses:

H0

SHG-BLP results in more financial inclusion for the participants than their non-participants’ counterparts.

H0

SHG-BLP results in more social inclusion for the participants than their non-participants’ counterparts.

Both hypotheses have their origins in Refs. [18,19] the theoretical foundation of the microfinance programme.

Moreover, the studies related to SHGs are area-specific because the implications of SHGs performances depend on social, cultural, economic, and political (SCEP) factors. The similar social, economic, cultural, and political backgrounds are a priori conditions for footprint analysis. Thus, there is a research gap for Assam in this respect, and it will be appropriate to explore and study the aftermath of SHGs in Assam, considering a specific region where a priori conditions are fulfilled. With this backdrop, the current paper concentrates on exploring the efficacy of the SHG-led microfinance programme on improving financial inclusion and abbreviating the social exclusion of the participating vulnerable rural poor households. While unfolding, this objective signifies that the study will involve itself in the impact evaluation of the SHG-BLP by corroborating more financial and social inclusion simultaneously for the participants than their non-participant counterparts.

The paper is organised into six parts, as follows: The first section is the introduction, and section 2 discusses the justification of the study. Section 3 sketches the sample design and enlists data sources; the geographical scenario of the study area and the methodologies adopted for the study are discussed in Section 4. The empirical results are presented in Section 4, and the discussion of the empirical findings is presented in the subsequent Section 5. The research findings in terms of conclusion as well as suggestions for induced policy measures are provided in Section 6.

2. The rationale of the study

In Assam, SHG formation started in the early 1990s. In fact, the state became the epicentre of the SHG-Bank Linkage Programme (SHG-BLP/SBLP) [36] since its inception. Assam is divided into five administrative divisions, viz., Barak Valley (3 districts), Central Assam (6 districts), Lower Assam (11 districts), North Assam (4 districts), and Upper Assam (9 districts). In Central Assam, there is a large chain of active SHGs compared to other parts of the state. These SHGs are registered under earlier SGSY and presently DAY-NRLM (https://daynrlmbl.aajeevika.gov.in/). It is noteworthy that Central Assam comprises three plain and three hill districts; Dima Hasao, East Karbi Anglong, and West Karbi Anglong are the hill districts, and Nagaon, Morigaon, and Hojai are the plain districts. However, the proposed division is planning to separate the hill region from the plain of Central Assam, as Assam Hills.

The socio-economic conditions, population structures, and literacy rates of all three plain districts are similar. Among all the three plain districts, Nagaon recorded the top slot, followed by Morigaon and Hojai district in providing group loans [20]. In fact, Nagaon has the highest share of operative SHGs in Assam [21]. The total numbers of operative SHGs in Nagaon, Morigaon, and Hojai are 15868, 7699, and 6104, respectively, aggregating to 29,671, which accounts for almost 74% of total operative SHGs in Central Assam and 13% of total operative SHGs in Assam [21]. The population structure of these three districts is diversified. These districts are the residence of the vulnerable rural population, viz., the Dimasa tribes and Hindu and Muslim lower class and lower caste people. This background motivates us to consider these three districts of Central Assam for our study.

The study is novel in assorted ways. The uniqueness of the study lies in the choice of the study area. This is perhaps the first of its kind in the concerned area. Although some studies address Nagaon for SHGs impact assessment, no earlier research explored all three plain districts for SHGs impact analysis. Secondly, the objectives of the current study are to make it nonpareil; earlier studies only concentrated on analysing the impact of SHGs on financial inclusion for the vulnerable rural poor. In this sense, the present study is complementary to the ongoing research on the impact of SHGs on financial inclusion for the vulnerable rural poor. But the present study contributes by taking a step beyond the SHGs impact analysis on financial inclusion to examine their potential for social inclusion. The achievement of seventeen SDGs implies social inclusion, which is the contemporary goal of the central government, as reflected in “Pradhan Mantri Mann Ki Baat” (What is on the Prime Minister's mind?) as “Sabka Vikas Sabka Saath” (Development for all). Accordingly, this paper is an attempt to scrutinise the role of SHGs in Assam in accomplishing the central goal of “Sabka Vikas Sabka Saath".

The uniqueness of the study is assorted by the choice of the study region as well as the extension of the impact evaluation to social participation. The study includes three districts of Central Assam, viz., Nagaon, Morigaon, and Hojai. The study area is one of its strengths. The simultaneous impact evaluation of SHG-BLP participation on financial as well as social inclusion also makes the study unique. The application of appropriate econometric and statistical tools is another strength of the study.

3. Methods

The data sources and the methodology to explore the said objectives are presented in this section.

3.1. Conceptual framework

Globally, microfinance is a successful model for uplifting the marginalised and vulnerable. Generally, participants are women, and commonly, participation ensures employability, income generation, financial inclusion, and in some favourable cases, women empowerment. The SHG bank linkage programme (SHG-BLP/SBLP) and the Microfinance institution (MFI) Bank Linkage model are two widely accepted microfinance models. However, the theoretical underpinning of both models is the “Difference Principle” of Rawls' Theory of Justice (1971). In accordance with the “Difference Principle” [18,19], endorsed boosting the advancement of society's “least-advantaged” group. In addition to ensuring “fair equality of opportunity” [18,19], he would also implement other feasible measures, such as a minimum wage or guaranteed minimum income. The “Difference Principle”, according to Refs. [18,19], gave his theory of social justice a liberal bent. Social justice can only be achieved through social inclusion. Thus, the ultimate goal of any microfinance model is to achieve social justice for all, which means social inclusion for all. The impact analysis of any microfinance programme is incomplete without exploring its role in achieving social inclusion for participants. This conceptual framework is the foundation of this study.

3.2. Geographical scenario of the study area

Assam is one of the seven states in the northeastern region. The state is broadly divided into two river valleys, viz., the Brahmaputra and Barak River valleys. Assam shares international borders with Bhutan and Bangladesh. The state is bordered by Nagaland and Manipur in the east; Arunachal Pradesh and Bhutan in the north; Meghalaya, Tripura, Mizoram, and Bangladesh in the south; and in the west, there is West Bengal. The state is connected with mainland India through the 22-km Siliguri corridor, which is known as the chicken neck. Fig. 1 displays an image of the Assam map.

Fig. 1.

Map of Assam.

3.3. Data: study area and sampling

Presently, the state of Assam is divided into 33 administrative districts under five regional divisions. We consider “Central Assam” as our study area. “Central Assam” consists of six districts. Among them, we have selected Nagaon, Morigaon, and Hojai as the maximum numbers of revived and operational SHGs are located in these three districts, and there is a proposal to separate the rest of the three districts from Central Assam to form a new division, “Assam Hills”. Moreover, these three districts are similar in terms of occupational and population structure, literacy rate, standard of living, etc. [35], (see table-A.3 in appendices [35]). Thus, we restrict our attention to these three districts only purposefully. The total numbers of operative SHGs in Nagaon, Morigaon, and Hojai are 15868, 7699, and 6104, respectively, aggregating, to 29,671, which accounts for almost 74% of the total number of operative SHGs in Central Assam [21].

The present study utilises secondary data sources, such as the NABARD Report, (2016–17) [20] and the DRDA unpublished report of respective district block offices, (2019–20), to draw the profile of the study area. However, the present study mainly relies on the novel data set that has been especially collected for exploring the mentioned objectives. We adopted multistage, random sampling with stratified and purposive sampling techniques. The data is collected in different stages (multistage). In the first stage, out of six districts in central Assam, considering socioeconomic and demographic conditions, three districts are selected purposively. From the selected districts, only intensive blocks are selected for sampling (see Table-A.3 in the appendices) in the second stage. Out of six, two intensive blocks are selected purposively from Nagaon district, with the characteristics of one being nearest and another being farthest from district headquarters. On the contrary, all the intensive blocks from Marigaon and Hojai (see Table-A.3 in appendices) districts are selected for the study. In the third stage, 4% of the total SHGs from the selected intensive blocks are selected purposively with the characteristics that 50% of the SHGs are nearest and the rest are farthest from the block headquarters. By utilizing [22] formula, we determine the number of participants and non-participants for each selected SHGs to be sampled in the fourth stage. In the fifth stage, we use the random number table to select the participants and non-participants randomly for the interview. The selection of the participants and non-participants is stratified into different intensive blocks. In the sixth or final stage, the relevant information is gathered from both participants and non-participants. The total numbers of SHGs, participants, and non-participants are presented in Table 1.

Table-1.

Profile of the sample SHGs.

District	Blocks	Respondents	SHGs	Members sampled
Nagaon	Barhampur	Participants	29	58
		Non-Participants		87
	Dolongghat	Participants	30	60
		Non-Participants		90
Morigaon	Laharighat	Participants	48	96
		Non-Participants		144
	Mayong	Participants	8	16
		Non-Participants		24
Hojai	Binakandi	Participants	40	80
		Non-Participants		120
Total			155	775

Source: Authors' own specification based on unpublished secondary data collected from District DRDA Office, up to January 2019

The field survey was conducted from April 2019 to January 2020. To explore the objectives, we need two groups, viz., the treated and the control groups. As per the AIMS Guidelines, the control group is constituted to reflect the socio-economically comparable group to remove biases while estimating the programme impact [23,24]. The treatment group comprises those SHG participants who received the benefits of the SHG loan at least two years before the survey. On the contrary, the non-participants are those who have not availed any benefit from the SHG programme. They create the control group. To perform the matching without replacement, we need more non-participants than participants to appropriately explore the said objectives. Because of that, our sample involves a total of 310 participants and 465 non-participants. Thus, a total of 775 respondents comprise the sample frame for the study. The number of non-participants exceeds that of participants by 155, and this large number ensures the appropriateness of the date to explore the said objectives. The relevant information from the selected treated and control group members is collected through the interview based on the pre-tested questionnaire to facilitate the investigation of the objectives. The characteristics of the two group members and their households are presented in Table 2.

Table 2.

Summary Results of t-test Conducted in Quantitative Variables.

Variables	Respondents	N	Mean	S.D	S.E	Mean Difference	Std. Error Mean Difference	t-statistics
Age	Participants	310	31.474	6.503	0.369	1.343	0.461	2.913*
	Non-Participants	465	30.131	5.950	0.276
Education	Participants	310	5.577	3.362	0.191	0.483	0.255	1.896***
	Non-Participants	465	5.095	3.634	0.169
Agricultural Land Holding (In Bigha)	Participants	310	1.881	1.554	0.088	0.239	0.112	2.145*
	Non-Participants	465	1.641	1.473	0.068
Distance from Bank (in Km)	Participants	310	4.861	2.099	0.119	0.373	0.145	2.565*
	Non-Participants	465	4.488	1.798	0.083
Consumption expenditure	Participants	310	1906.710	484.884	27.540	496.839	30.389	16.349*
	Non-Participants	465	1409.871	277.053	12.848

Source: Authors' Own Calculation Based on Primary Data

*Significant at 10% level, **Significant at 5% level and *** Significant at 10% level.

The table divulges that the treated and control group members are significantly different in terms of age, education, agricultural landholding (in bighas), distance from a bank (in km), and consumption expenditures (implied by the significant t value). The table also ensures the quality of the data. The impact of microfinance will be overestimated if we simply run naïve regression, as the participants are significantly different from the non-participants. Thus, without correction of the bias in terms of observable characteristics, the impact assessment will be overestimated.

3.4. Methodology

In this section, we will discuss the methodologies to be used to scrutinise the said objectives.

3.4.1. Financial inclusion and social inclusion index: Multiple Correspondence Analysis (MCA)

The concept of financial inclusion is multidimensional, and there is no consensus among the researchers regarding the measures of financial inclusion. In a nutshell, it may be interpreted as poor and low-income households' access to and usage of basic financial services, viz., savings, credit, and insurance available from formal institutions. The whole system must be convenient, flexible, and most importantly, reliable in the sense that savings are paid and insurance claims are received with certainty [13,25,26]. Hence, financial inclusion indicates households' sustainable access to formal financial services, viz., money transfer, savings, credit, insurance, etc. These are also integral parts of microfinance services. In the absence of a well-accepted financial inclusion index (FII), we have developed an FII considering indicators like households’ access to credit, transaction services, savings, and insurance (for details, see table-A.1 in appendices) following [12,13,15,24].

Social exclusion is a multidimensional phenomenon. The concept is wider than materialistic deprivation. Poverty is one of the dimensions of social inclusion. [1,27], defined social inclusion as “a state in which individuals are unable to participate fully in economic, social, political, and cultural life, as well as the process leading to and sustaining such a state”. United Nations, (2016), identified three areas of exclusion, viz., “Economic exclusion, exclusion from public services, and exclusion from civic and political participation”. Following [1,27], we have also constructed a self-developed social exclusion index. The modalities utilised for this purpose are presented in Table-A.2 in appendices. Following [28], the higher the value of the index implies lower social exclusion and/or more socially included (for details see table-A.2 in appendices). The separate index for each dimension is calculated first and then a weighted average of the three dimensions is calculated to obtain the SEI. The weights are equal for all the dimensions [1,27]. For indexing, the two well-accepted methods are Preference Indexing and Indexing by Principal Component Analysis (PCA). In the first method, the weights of the index, are selected according to the preference of the researcher and therefore largely involve the discretion of the researcher. This method of indexing is highly criticised by researchers. On the contrary, PCA for indexing is accepted by almost all researchers. However, PCA is not applicable if the modalities are categorical. The way out is the MCA, as the MCA allows one to analyse the pattern of relationships among several categorical variables [29]. The modalities utilised for indexing (both FII and SEI) are categorical, and therefore we used MCA for indexing. The equation for indexing is presented as equation-1.

When all the indicators are dichotomous the index for ith participant is obtained as follows:

$$FI{I}_i=\frac{1}{K}\left(W_1{I}_{i1}+W_2{I}_{i2}+.......+W_p{I}_{ip}\right)$$ (1)

where.

$W_p$ = the weight (score of the first standardised axis, (score or $\sqrt{{\lambda }_1}$) of category p.

$I_p$ = binary indicator (0 or 1), which takes on the value 1 when the household has the modality, and 0 otherwise.

The index so constructed lies between 0 and 1. In the case of FII “0” means complete financial exclusion and “1” means total financial inclusion. On the contrary, in the case of Social exclusion, index value “0” means socially excluded and “1” means complete social inclusion.

3.4.2. Propensity score matching method and impact assessment

The propensity score matching (PSM) method is the well-accepted method for comparing the outcomes of the treated and control groups. The method allows the researchers to draw causal inferences after correcting the observational biases between the two groups [30,31]. The propensity score [P(X)] is the conditional probability calculated based on the observable characteristics of an individual or household who is treated or selected (Rosenbaum and Rubin, 1983). Being probability the balancing score value lies between 0 and 1. The probability of being treated can be obtained by using any binary choice model (Probit or Logit) [32], and in our case, we use the Probit model. The Probit model specification is presented by equation-2.

$$P\left(Y\right)=P\left(D=1|Y\right)=\phi \left({\alpha }_1{Y}_1+.....+{\alpha }_k{Y}_k\right)=\phi \left(\beta Y\right)$$ (2)

Where, $0<\phi \left(\beta Y\right)<1$ for all values of X and $\phi$ is the cumulative distribution function of the standard normal distribution. The equation-2 is non-linear and therefore parameters $\beta$ are to be estimated by the maximum likelihood estimator.

The impact of the microfinance on the outcome indicators can be assessed by comparing treated and control with valid matches. Since propensity score is a balancing score it provides valid matches between two groups.

Thus if $Z^1$ is the outcome indicator for the SHG programme participants (D = 1), and $Z^0$ is the outcome indicator for the non-participants (D = 0), then the mean impact is given equation-3.

$${}^{ATT}{\Delta }_{PSM}=E_{P\left(X\right)}\left(Z^1|Y,D=1\right)-E_{P\left(X\right)}\left(Z^0|Y,D=0\right)$$ (3)

where ATT represents the Average Treatment effect on the Treated group. Two matching algorithms, viz., the Nearest-Neighbour Matching (NNM) and Kernel Matching (KM) are employed in this study to circumvent any shortcoming and to check the robustness of the estimated impact on the outcomes.

4. Results

This section presents the results of this study.

4.1. Propensity score matching estimator

The Probit model is utilised to determine the probability of being treated or participated in the SHG programme. Table 3 presents the possible variables that are supposed to influence the individual's SHG participation decision. The detailed descriptions of the predictors of the Probit model are also present in Table 3. The propensity scores of the participants and non-participants are determined based on these variables.

Table-3.

Probit regression model for the estimation of the propensity score.

Dependent Variable: SHG participation Dummy (1 = participant; 0 = non-participant)		Iteration 0: log likelihood = −521.07279 Iteration 1: log likelihood = −375.25153 Iteration 2: log likelihood = −374.48786 Iteration 3: log likelihood = −374.48758 Iteration 4: log likelihood = −374.48758
Probit regression Log likelihood = −374.48758		Number of observation = 775 LR chi2 (7) = 293.17 Prob > chi2 = 0.0000 Pseudo R2 = 0.3813
Variables	Definition	Coefficient	S. E	Z-Statistics	P>|z
Age	Measured in complete years	0.027	0.009	3.02***	0.003
Education	Total schooling in completed years	0.011	0.015	0.75	0.452
Agricultural Land Holding (In Bigha)	Size of the operational land holding in bigha	−0.119	0.034	−3.48***	0.00
Cast	Cast of the sampled household Dummy, if SC/ST = 1, 0 otherwise	5.042	0.460	10.95***	0.00
Relegion	Dummy, if Hindu = 1, 0 otherwise	−0.032	0.170	−0.19	0.852
Distance from Bank (in Km)	Distance of the bank measured in km	−0.088	0.028	−3.19***	0.001
Consumption expenditure	Monthly household consumption expenditures measured in Rs.	0.002	0.000	13.78***	0.00
Constant	–	−0.091	0.168	−0.55	0.585

Source: Authors' own calculation based on primary data

***Significant at 1% level. **Significant at 5% level. *Significant at 10% level.

The robustness of the specification is reflected by the high Likelihood Ratio Chi-squares of 293.17 (with a p-value of 0.0000) and Pseudo R-squares 0.3813. The estimated model discloses that the probability of participating in and borrowing from the SHGs is significantly related to age, size of agricultural land holding, cast, distance from the bank, and consumption expenditures. Age, cast, and consumption expenditures all have a positive influence on participation decisions. On the contrary, the predictors-agricultural landholding (in Bigha), and distance from the bank (in kilometres)- are negatively influencing the participation decision.

The ranges of the estimated propensity scores are presented in Table 4.

Table-4.

Estimation of Probit for the whole sample: p-Score.

Variable	Observation	Mean	Std. Dev.	Min	Max
P-Score	774	0.3993	0.2827	0.0119	0.9998

Source: Authors' own calculation based on primary data

The table reveals that the propensity score ranges from a minimum of 0.0119 to a maximum of 0.9998. Consequently, our common support region is [0.0119–0.9998]. The mean propensity score is 0.3993, implying that Probit regression successfully predicts the participation of all observations in about 40% of cases.

4.2. Checking for common support

The impact analysis of microfinance is facilitated by comparing the outcome indicators of the participants with those of the non-participants. A reasonable comparison is only possible when there is enough overlap between the treatment and control groups. By plotting the propensity score histogram, we can check the common support region. The common support region is displayed by Fig. 2.

Fig. 2.

Common support region.

The figure confirms the existence of considerable overlap in the propensity score distribution of both the treated and control groups. This indicates the condition for common support is satisfied. In the present paper, we will use Nearest-neighbour matching and Kernel matching algorithms without replacement, therefore, 310 participants will be matched with 310 non-participants. Thus, 155 observations from our analysis will be discarded.

4.3. Checking for balancing between the two groups

Table 2 ensures that the treated and control groups are significantly different in terms of the observable. But for unbiased comparison between treated and control groups, we need two similar groups. The propensity score theoretically eliminates all the differences between two groups (participate and non-participate) in terms of observable. The test of balancing covariates helps us to test that the mean propensity score of treated and control groups is similar. The Table 5 presents the test result.

Table-5.

Identifying the balance between two groups.

	Mean				t-test
Variables	Treated		Control	% Bias	t	p>|t|
Age	31.47		29.77	3.0	0.26	0.797
Education	5.58		5.97	−7.0	−0.63	0.528
Agricultural Land Holding (In Bigha)	1.88		2.33	−9.4	−0.86	0.389
Cast	0.52		0.63	−5.6	−0.53	0.598
Relegion	0.58		0.52	0.0	0.00	1.000
Distance from Bank (in Km)	4.86		4.35	2.5	0.21	0.833
Consumption expenditure	1906.70		1914.50	−10.6	−0.88	0.378
Pseudo- R2		0.037
LR chi2		3.52
p > chi2		0.966

Source: Authors' own calculation based on primary data

The table offers the useful insight that the matching eliminates almost all the existing biases. This is apparent from the t-test results, which indicate that all the differences in terms of observable covariates between the treated and control groups are eliminated after matching. Moreover, the sufficiently low pseudo-R² after matching implies that the matching procedure can balance the characteristics of the treated and control groups.

As all the pre-conditions are satisfied, we can proceed with analysing the impact of the microfinance based on output indicators in terms of the average treatment effect on the treated (ATT), considering two matching algorithms, viz., the Nearest-Neighbour Matching (NNM) and Kernel Matching (KM).

4.4. Impact of microfinance on outcome indicators and sensitivity analysis

The impact analysis of the SHG-BLP lead microfinance programme is generally conducted based on the average treatment effect on the treated (ATT), the average treatment effect (ATE) and the average treatment effect on the untreated (ATU). But since the results of the ATE and ATU are not testable, these estimates cannot be used to find the actual impact of the microfinance programme. Consequently, we consider ATT for exploring the impact of the SHG-BLP led microfinance programme. The result is presented in Table 6.

Table-6.

ATT estimates of SHG participation impact on Financial Inclusion and Social Exclusion.

Impact Indicators	Matching Method	ATT
		Treated	Controls	Difference	Std. Err.	t-statistics
Financial Inclusion	Nearest Neighbour	0.52	0.41	0.12	0.01	7.99***
	Kernel	0.52	0.39	0.14	0.01	10.25***
Social Exclusion	Nearest Neighbour	0.35	0.28	0.07	0.01	6.29***
	Kernel	0.35	0.27	0.08	0.01	8.31***

Source: Authors' own calculated based on primary data

***Significant at 1% level, * Significant at 10% level.

The table discloses that the SHG-BLP-led microfinance programme has a significant impact on both financial and social inclusion. The estimated ATT of the microcredit on financial inclusion of the participants evidences an increment of FII on average 12% more than their similar non-participant counterpart for Nearest Neighbour matching and the result is significant at the 1% level. Kernel matching reports a similar result. However, in the case of Kernel matching, the financial inclusion of the participants is 14% higher than that of their similar counterparts, and the result is also significant at the 1% level. The same table also discloses that the ATT of social inclusion for the participants is 0.35 and that of the non-participants is 0.28, considering Nearest Neighbour matching. A similar result is also reported by Kernel matching. Both financial Nearest Neighbour and Kernel Matching algorithms confirm that the SII of the participants are 7% and 8% higher than that of their similar counterparts. Both results are significant at the 1% level. The output shows that we get a significant positive treatment effect on the treated of 0.52 and 0.35 for FII and SEI, respectively. That is, the rate of improvement, measured in terms of FII and SEI, of participants is 12.0% and 7.0%, considering Nearest Neighbour and 14.0% and 8.0%, considering Kernel matching, higher than that of matched control group members.

The validity of the above empirical result must be tested, and that can be performed only by corroborating that there is no hidden bias. This is performed by conducting the Mantel-Haenszel (1959) [33] bounds test. Under the assumption of no hidden bias ($\mathrm{\Gamma }=1$), the Mantel-Haenszel (1959) [33] bounds, test statistic gives a similar result, indicating a significant treatment effect (Table 7).

Table-7.

Sensitivity analysis for average treatment effect.

Outcome variables	Mantel-Haenszel (1959) bounds
	Gamma	Q_mh+	Q_mh-	p_mh+	p_mh-
Financial Inclusion	1	1.879	14.684	0.030	0.000
Social Exclusion	1	−0.061	0.524	0.000	0.000

Source: Author's own calculation based on primary data

5. Discussion

The possible reasons for the empirical results are discussed in this section. The empirical results imply that the SHG-bank linkage programme under DAY-NRLM [34] has shifted participant households to a higher level of financial inclusion as compared to non-participant households. This result is congruous with the earlier findings by Refs. [12,13,15,24]. This outcome is quite obvious. While implementing SGSY, NRLM and DAY-NRLM special focus was thrust on vulnerable sections among the rural poor. Accordingly, it was aimed that SC/ST and women would account for at least 80% of the “swa-rozgaris” (self-employed). Consequently, the SHGs members are mainly women and majorly housewives. These members earlier did not possess a bank account or rarely had access to banks. But after becoming SHG members, they possess an individual or joint bank account, develop banking habits, and in some favourable cases, they started using ATMs. For the survival of the groups, they repay the loan on time, and in a few exceptional cases, they take insurance policy. All these steps help their transformation from financial exclusion to inclusion. Thus, patently, the SHGs participation assists in attaining higher financial inclusion. In fact, one of the main components of DAY-NRLM [34] is “universal financial inclusion”. Thus, SHGs are formed with this in-built objective, and undeniably, SHGs are performing to accomplish this goal.

The 2030 Agenda enshrined the principle that people everywhere should acquire the benefits of prosperity and should lead to a decent standard of well-being. The announced seventeen “Sustainable Development Goals” aim to achieve this principle for each one. The effective translation of the goals and targets results in the essential components of “Social Inclusion”. Although there is no universally accepted definition and/or benchmark of social exclusion, generally, a lack of participation in society is considered as social exclusion. In the present study, following [28], dual cutoff method, we consider that the greater the value of the Social Exclusion Index (SEI) the greater the opportunities for the individual to participate in social activities, implying that they are more socially included. The table divulges that the estimated ATT of the SHG-BLP-led microcredit on social exclusion of the participants measured by SEI increases on average by 0.07 (Nearest Neighbour matching) and/or 0.08 (Kernel matching) more than their similar non-participating counterpart. The estimated difference is significant at the 1% level in the case of both matching algorithms. Thus, participation in the SHG-led microcredit programme contributes to the social inclusion of the concerned individual and her family. Unfortunately, the lack of such research in the national and international arenas restricts us from juxtaposing our findings with others. However, our findings are consistent with the earlier studies conducted by Refs. [16,17].

This result is also not surprising. Following the guidelines of SGSY, NRLM, and DAY-NRLM, the members of the SHGs in central Assam are mainly women who belong to SC/ST. Before forming the group, they were housewives without any short of economic decision-making power and excluded in terms of all the dimensions of social exclusion. But after participation in the SHGs, they become the earning member of the family and participate in family decision-making related to the education of their children, including girl children. The participation also enables them to overcome information asymmetry and thus enjoy improved cooking fuel through the Pradhan Mantri Ujjwala Yojana”, constructed toilet, and improved drinking water facilities through the Swachh Bharat Abhiyan”. They are now empowered to invite their group members to their family occasions and also empowered to participate in those invitations. They are also empowered to cast their votes according to their choice. With the central and state governments' initiatives, these lower-caste and lower-class women contest for “Panchayat” positions as candidates. All such measures increase the corresponding scores of the participants in all three dimensions and, as a whole, improve their SEI score. Therefore, SHG participation makes the participants more socially included than their non-participating counterparts. Accordingly, the SHG-BLP participants achieve the microfinance program's goal in accordance with “Rawls Theory of Justice”, (1971) [18,19].

6. Conclusions

The objective of this research was to examine the role of SHGs in financial inclusion and depletion in social exclusion. The study area of the research is the three districts of Central Assam, viz., Nagaon, Morigaon, and Hojai. The impact evaluation is administered using the PSM method, and the empirical results manifest that the programme is significantly successful in improving the financial and social inclusion status of the treated group members compared to the control group members. The study is novel for several reasons. This is perhaps the first study that explores the theoretical foundation of the “Difference Principle” of “Rawls Theory of Justice, (1971)". The objectives of the study, the applied methodology make it unique. Moreover, the study area also makes it unique. It is the first study to investigate the impact of SHG-BLP on financial and social inclusion among participants versus non-participants. Notably, the theoretical elegance of the microfinance model is supported by the “Difference Principle” of Rawls Theory of Justice (1971). Accordingly, microfinance is a holistic model to achieve social inclusion for all. In this sense, the present study established the essence of the “difference principle” of " Rawls Theory of Justice (1971)". The SHG-led microfinance programme is a globally accepted model for the financial inclusion of vulnerable. In India, the participants are mostly housewives. The participations enable them to earn money, having a bank account, using ATM cards, having insurance, and consequently, financial inclusion through SHG-BLP is a common phenomenon. The real question here is whether or not participation in SHG-BLP entitles them to social inclusion. If it does, then the ultimate goal of SHG-BLP is considered to be satisfied. This is exactly achieved in the present model for the specified study area. Instead of having religion-region-cultural differences, we observed that participants became more socially included than their non-participants' counterparts. This result is in conjunction with the current objectives of the central government. The central objective is very clear from a single line, “Sabka Vikas Sabka Saath” (Development for all), and it is reflected several times in “Pradhan Mantri Mann Ki Baat” (What is on the Prime Minister's mind). To achieve this goal, the central and state governments are working day and night. The result is the registration of SHGs in DAY-NRLM and many more. As a consequence of these continuous processes, we are getting a successful model of “development and social justice for all” through SHGs in interior India. Thus, it is highly recommended to encourage SHGs bank linkage model. Concurrently, the coverage of the SHGs registration under DAY-NRLM is advocated for expansion. Concomitantly, to expand the functioning of SHGs, the alternative model, viz., the MFI-Bank Linkage Model, is required to be encouraged.

One limitation of the study is that it focuses only on Central Assam. If the study could be extended to Assam as a whole, the impact evaluation may be better understood. However, the a priori condition for impact analysis is that the social-economic-demographic scenario must be similar. Based on the fulfillment of this apiori condition, these three districts of central Assam were chosen. Further extension of the study area is in the future planning of the study. Moreover, the group maturity and length of microfinance participation can have an impact on the economic status of the households. Particularly, the length of the microfinance participation is important for analyzing the prominence of social inclusion. To capture this, we need a dynamic framework. This is not performed in the present study due to the non-availability of such data. This may be viewed as another limitation of the study. The study may be extended in the future by applying dynamic panel analysis based on data collection.

6.1. Endnote

• 1.According to National Rural Livelihood Mission (NRLM) and Deendayal Antyodaya Yojana- National Rural Livelihood Mission (DAY-NRLM) Intensive blocks are those blocks where State Rural Livelihood Missions (SRLMs) directly enter with their own staff at block level, and take support of the internal resources from the resource blocks to form new SHGs or promote the existing SHGs by providing assistances. Community Resource Persons (CRPs) are chosen from the resource blocks to accelerate the implementation of programmes in these blocks. On the contrary, non-intensive blocks are those where NRLM strengthens existing SHGs in these blocks with some capacity building and limited financial assistance. No new mobilization would be done in these blocks.

Submission declaration statement

The authors declare that this article is not under consideration of publication elsewhere and no part of the article is published anywhere in any form. Moreover, the authors also state and that the article will not be submitted for publication elsewhere without the agreement of the Managing Editor.

Information on rights and permissions

Not applicable.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Funding

This research did not receive any specific grant from any funding agencies in the public, commercial, or not-for-profit sectors.

Author contribution statement

Shrabanti Maity: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Data availability statement

Data will be made available on request.

Declaration of competing interest

The authors declared and specified that they have no competing interest and/or conflict of interest.

Appendices.

Table-A.1.

FII indicators and their corresponding weights

Indicators	Sources of Finance	Nature of the variable	Descriptions	Weights
Formal Credit	From formal agencies directly and/or through SHG during Survey (during 2016)	Categorical	Yes = 1	0.124
			Otherwise = 0	0.005
	From formal agencies directly and/or through SHG during Survey (during 2017)	Categorical	Yes = 1	0.105
			Otherwise = 0	0.020
	From formal agencies directly and/or through SHG during Survey (during 2018)	Categorical	Yes = 1	0.079
			Otherwise = 0	0.055
	Regular repayment of loan	Categorical	Yes = 1	0.094
			Otherwise = 0	0.031
	Received loan during emergency	Categorical	Yes = 1	0.087
			Otherwise = 0	0.038
Savings	Operating SB Account in Bank/Post office/Co-operative Banks	Categorical	Yes = 1	0.120
			Otherwise = 0	0.024
	Fixed Deposit or Recurring Deposit Account with Institutional Agencies	Categorical	Yes = 1	0.087
			Otherwise = 0	0.015
	Savings in SHG	Categorical	Yes = 1	0.101
			Otherwise = 0	0.040
Insurance	Any source/type of insurance	Categorical	Yes = 1	0.110
			Otherwise = 0	0.011
Transaction Services	Usages of ATM/Debit Card/Cheque	Categorical	Yes = 1	0.085
			Otherwise = 0	0.047
Banking Knowledge	Knowledge of banking procedure after joining SHG	Categorical	Yes = 1	0.024
			Otherwise = 0	0.004
SHG Operation	Facing problem of getting loan through SHG	Categorical	Yes = 0	0.070
			Otherwise = 1	0.118

Source: Authors' own calculation based on primary data

Table-A.2.

SEI indicators and their corresponding weights

Broad areas of Exclusion	Indicators	Modalities	Weights
			Yes	Otherwise
Economic Exclusion	Subjective basic needs	In the past 1 year household has not able to offered three meal a day or pay bill regularly, or keep the home adequately warm (if yes = 0, 1 otherwise)	0.06	0.017
	Employment	Being unemployed or discouraged worker (if Yes = 0, 1 otherwise)	0.010	0.067
	Financial service	Bank accounts owns name (if yes = 1, 0 otherwise)	0.050	0.027
	Material deprivation housing	Can't offered bed for every member (if yes = 0, 1 otherwise)	0.033	0.044
		Constructed toilet (if Yes = 1, 0 otherwise)	0.037	0.040
		Has Cement roof (if yes = 1, 0 otherwise)	0.075	0.002
		Has cement floor (if yes = 1, 0 otherwise)	0.065	0.012
	Material deprivation amenities	Household cannot afford anyone or all: washing machine microwave and freeze (if yes = 0, 1 otherwise)	0.003	0.074
		Has mobile phone (if yes = 1, 0 otherwise)	0.068	0.009
		Has TV set (if yes = 1, 0 otherwise)	0.061	0.015
		Has bicycle (if yes = 1, 0 otherwise)	0.043	0.034
		Has four wheeler motored vehicle (if yes = 1, 0 otherwise)	0.074	0.003
	Material deprivation ICT	Household needs a computer or internet but cannot afford one (if yes = 0, 1 otherwise)	0.001	0.076
Exclusion from public Service	Public Utilities	No public water connection (if yes = 0, 1 otherwise)	0.021	0.091
		No Sewerage system (if yes = 0, 1 otherwise)	0.004	0.107
		No Gas connection (if yes = 0, 1 otherwise)	0.044	0.067
		No Electricity (if yes = 0, 1 otherwise)	0.045	0.066
	Education	Could not afford to buy school materials for every child in past 12 months (if yes = 0, 1 otherwise)	0.016	0.095
		Young children not in school or pre-school (if yes = 0, 1 otherwise)	0.022	0.090
	Health	Could not afford medication or dental checks for every child in the past 12 months (if yes = 0, 1 otherwise)	0.035	0.076
		Doctor consulted in case of medical needs (if yes = 1, 0 otherwise)	0.11	0.001
	Social infrastructure	Lack of opportunities to attend events due to distance (if yes = 0, 1 otherwise)	0.051	0.060
Exclusion from Civic and Political Participation	Social contact	Rare or infrequent social contract with family or relatives (if yes = 0, 1 otherwise)	0.052	0.073
		Rare social contract with friends (if yes = 0, 1 otherwise)	0.057	0.068
		lack of support network that could help in the events of emergency (if yes = 0, 1 otherwise)	0.037	0.088
	Social participation	Since last 1 year household has not been able to invite friends or relatives for meal at least once a month (if yes = 0, 1 otherwise)	0.061	0.064
		Has not been able to afford to buy cinema tickets since 12 month (if yes = 0, 1 otherwise)	0.046	0.079
	Civic participation	Inability to vote due to lack of eligibility or distance to polling station (if yes = 0, 1 otherwise)	0.023	0.102
		No participation/membership in associations, teams or clubs (if yes = 0, 1 otherwise)	0.024	0.101
		No participation in political/Civic activities (if yes = 0, 1 otherwise)	0.024	0.101

Source: Authors' own calculation based on primary data

Table-A.3.

Administrative Divisions of the Selected Districts

District	Geographical area (sq Km)	Population	Literacy (%)	Major Occup-ation	HDI	Revenue Circles	Blocks	Intensive Blocks	Villages	Towns/Cities
Nagaon	2287	1892550	71.00	Agriculture & Allied	0.592	7	13	6	1210	4
Marigaon	1450.20	957423	69.37	Agriculture & Allied	0.576	5	7	2	638	2
Hojai	21.219	36,638	81.08	Agriculture & Allied	0.695	3	5	1	409	3

Source: DRDA and Municipality Board of Nagaon, Morigaon and Hojai, Census of India, Assam Human Development Report.