Open Access

THE DETERMINANTS OF MICROFINANCE INSTITUTIONS EFFICIENCY: THE ROLE OF WOMEN BORROWERS

HAZWAN HAINI

School of Business and Economics, Universiti Brunei Darussalam, Bandar Seri Begawan, Brunei

E-mail Address: hazwan.haini@ubd.edu.bn

Corresponding author.

Search for more papers by this author

and

ZOË ANASTASIOU

School of Business and Economics, Loughborough University, Leicestershire LE11 3TU, UK

Search for more papers by this author

https://doi.org/10.1142/S2737566821500067Cited by:0 (Source: Crossref)

Abstract

Microfinance institutions (MFIs) are instrumental in enabling the economic empowerment of women. We examine the efficiency and performance of 84 Indian MFIs from 2016 to 2018 using a two-stage double bootstrap approach. Our results show that MFIs with increased outreach and actively target female borrowers achieve higher efficiency. Furthermore, we find larger MFIs and higher leverage intensity to be positively associated with efficiency. Government policies should be encouraged to support current MFIs to grow larger, actively target female borrowers and increase outreach to the poor to support India’s financial inclusion agenda and facilitate the economic empowerment of women whist revitalizing less efficient MFIs.

Publisher's Note:

At the author's request, on page 113, to remove the following "We acknowledge the direction of the Editor-in-Chief, Professor Ahmed Khalid and the Editorial Desk Manager, Frankie Ong". Acknowledge should read as "We are grateful for…" This information has been updated on 27th July 2022.

Keywords:

1. Introduction

The lack of access to financial services remains a major challenge for India. India has the second-largest unbanked population after China (Haini, 2020a), with over 190 million unbanked out of the 1.7 billion globally, with women overrepresented among the world’s unbanked, making up to 60% of the unbanked in India (Demirgüç-Kunt et al., 2020). In developing countries, the poor, elderly and women lack access to formal financial services that are affordable, fair and safe. It is long argued that commercial banks have not fulfilled the credit needs of the relatively poor people as they do not have collateral and are considered high risk. To overcome these challenges, financial inclusion is an important agenda for policymakers and governments. Financial inclusion can be defined as a process that serves to remove the barriers to specific societal groups and individuals. It aims to alleviate poverty, minimize social exclusion and enhance economic growth through various solutions (Chibba, 2009). One solution to promote financial inclusion is through the development of Microfinance Institutions (MFIs).

MFIs offer financial services to low-income populations that are generally excluded from the formal financial system. The advocates of microfinance argue that access to finance can help reduce poverty, by providing the poor opportunities to invest in income-generating activities and diversify their sources of income (Littlefield et al., 2003). From a wider perspective, there is consensus that financial development promotes economic growth through the early seminal works of Goldsmith (1969) and King and Levine (1993). Although economic growth benefits the majority, the rapid financial development of many developing countries is at the expense of the relatively poor (Haini, 2020b). As a result, reaching the poorest is an important objective in many developing countries and microfinance is one way to promote outreach (Lønborg and Rasmussen, 2014). Thus, the performance of MFIs at its objective of promoting financial inclusion is of interest to governments, charitable institutions and socially responsible investors.

This study examines the efficiency of Indian MFIs from 2016 to 2018 and the determinants of its efficiency. India has the most mature MFI industry compared to other developing countries as its industry has grown rapidly since 2005 (Ghosh, 2013). However, the rapid growth of MFIs has come at a cost as increased competition and profit-oriented firms have affected the social mission of MFIs. As a result, the MFIs (Development and Regulation) Bill, first drafted in 2012, was proposed by the Finance Ministry to override all other laws with the objective of consumer protection. Despite the Indian MFI industry experiencing a crisis, MFIs can still be a means of expanding credit access to those excluded from the formal financial system. The examination of Indian MFIs is interesting, as India has the most mature MFI industry and experienced the positive and negative aspects of microfinance.

Previous studies examining the efficiency of Indian MFIs employed standard data envelopment analysis (DEA) scores and censored (Tobit) regression for the second stage (Babu and Kulshreshtha, 2014; Kar and Deb, 2017). There may be issues with interpreting the results of standard DEA scores which are known to overestimate scores if not calculated with a bootstrap (Simar and Wilson, 1998). In addition, the use of censored (Tobit) regression in the second stage has been criticized by Simar and Wilson (2007), who suggested that Tobit estimates are serially correlated if not treated with a bootstrap and truncated regression. We avoid these issues by employing a two-stage double bootstrap approach, providing better results for inference. Moreover, previous studies that examine the determinants of efficiency employed a cross-country dataset. It was suggested that the determinants of efficiency may be country-specific, as different regulations may affect the operation of MFIs (Gutiérrez-Nieto et al., 2007). Cross-country studies may not fully acknowledge the significance of country characteristics such as macroeconomic environments and differences in regulatory framework and level of competition. We contribute to the MFI efficiency literature by the examination of an individual country that has had a unique experience with microfinance.

We find that the average bias-corrected efficiency score for Indian MFIs is 48.7%, suggesting there is room for improvement. Our findings are comparable to earlier studies that also examined MFI efficiency with women borrowers. For example, Gutiérrez-Nieto et al. (2009) report an efficiency score of 35.53% and 28.16 for MFIs across many countries. Meanwhile, other studies that examine MFI efficiency for general borrowing also show comparable findings such as in Haq et al. (2010), which reports an average efficiency score of 38.30%. Consequently, our study employs recent data and shows that MFIs in India are relatively more efficient at reaching women borrowers, yet our results imply that there can be further cost savings.

Subsequently, we identify the determinants that may improve efficiency. Our estimates suggest that the size of firms affects the efficiency of MFIs, which is consistent with previous MFI examinations (Wijesiri et al., 2015). In addition, we support the literature that suggests outreach and efficiency may be compatible (Morduch, 2007; Bos and Millone, 2015) as we find that MFIs that actively target female borrowers and have higher outreach are positive and significant determinants of MFI efficiency. We find that the age, legal status and profitability of MFIs are insignificant to efficiency scores. A large segment of MFI borrowers originates from self-help groups, which began to develop in the 1970s. Furthermore, previous studies suggest that MFIs that are non-governmental organizations (NGOs) tend to be more efficient at outreach compared to others. In our case, we find that legal ownership makes no difference to the efficiency of MFIs. This may be a country-specific effect where the regulation of MFIs in India may affect the corporate governance of MFIs. In addition, profitability is insignificant to efficiency scores this may be because MFIs exist to promote financial inclusion, and their financial focus is to be self-sustainable as opposed to achieving large profits.

The rest of the paper is organized as follows. Section 2 provides a brief literature review on the importance of financial inclusion and microfinance. This section focuses on microfinance in India and the benefits of financial inclusion on gender empowerment. Section 3 provides a review of DEA studies of MFIs. Section 4 describes our data sample, methodology and model specification. Section 5 presents our results and the discussion follows in Sec. 6. We conclude with policy implications and avenues for future research.

2. Financial Inclusion and Microfinancing Institutions

MFIs are generally overseas non-profit organizations that make loans to villagers, micro-entrepreneurs and poor families. Their main aim is to improve access to finance and help fund self-employment projects for income generation. Alongside the provision of microfinance in the form of loans, they offer savings products, fund transfers and insurance facilities (Haq et al., 2010). The original mission of MFIs could be regarded as achieving financial inclusion through outreach to the poor. As the industry has developed, some MFIs have shifted this mission toward becoming financially sustainable whereby they no longer have to rely on the support from donors to operate efficiently. This has given rise to an interesting debate of whether MFIs are able to achieve both financial inclusion and financial sustainability. Alongside this, the financial sustainability of MFIs has provoked an interest in their efficiency and whether the money lent to MFIs is being used effectively.

Lack of access to finance is often the cause of persistent income inequality, as well as slower economic growth (Beck et al., 2009). Academics suggest that access to affordable finance may enable the poor to undertake economic activities and to take advantage of growth opportunities necessary for financial development (Swamy, 2014). Hence, expanding access to finance and promoting financial inclusion remains an important objective for many governments, policymakers and organizations. MFIs play a significant role in bridging the gap between the formal financial institutions and the rural poor thus promoting access to finance.

The worldwide growth of MFIs has had a positive impact on reducing poverty as the core poor are generally risk averse to borrow for investment in the future. Many MFI programs are conducted through peer group loan methodology, in which members accept joint liability for the individual loans made. The group loan methodology is a risk-sharing innovation that achieves peer monitoring (Stiglitz, 1990) since the members of the group loan now have a direct interest in ensuring that no individual member defaults. This effectively transfers all costs and risks of the MFIs to the borrowers. Hence, MFIs allow formal institutions to operate in an industry that is otherwise dominated by informal arrangements. Furthermore, it is long argued that commercial banks, as a formal institution, have not provided the credit needs to the relatively poor who do not have the necessary conditions to offer loan guarantees but may have promising investment ideas (Gutiérrez-Nieto et al., 2007). Thus, MFIs are special financial institutions with both a social and for-profit nature.

The institutionalist and welfarist perspectives of MFIs expose the possibility of a trade-off between the financial inclusion of the poor and the financial sustainability of MFIs. The welfarist paradigm postulates that MFIs should focus on poverty alleviation and depth of outreach (Brau and Woller, 2004). MFIs vary in terms of their financial sustainability and many rely on donor support to cover their expenses, which enables MFIs to focus on their original mission of reaching the very poor.

The institutionalist paradigm suggests that MFIs should generate enough revenue to meet their operating and financing costs (Brau and Woller, 2004). As the MFI industry has matured and developed over the last decade, the importance of financial viability and sustainability has been highlighted (Hermes et al., 2011). Most MFIs are not financially sustainable and rely on donor support for their operations, as reaching the poor and providing them with credit may be very costly. Microfinance is a costly business, with higher transaction and information costs, due to the lack of background information on borrowers. As a result, there has been a shift in focus to increase efficiency and financial sustainability to mitigate these high costs especially for profit-oriented firms where investors expect returns. Various research studies suggest that MFIs experience mission drift as they increasingly cater toward more profitable customers as opposed to serving the very poor (Mersland and Strøm, 2010). The institutionalist view suggests that although the MFI industry has developed rapidly, it is still young and may have room for cost reduction in the future (Brand, 2000). During the 1990s, the focus of many NGOs gradually shifted from providing microfinance for development and poverty reduction to a business model that emphasized full-cost recovery (Ghosh, 2013). Pursuing such financial objectives can create incentives that encourage mission drift and, in turn, lead to unsustainable patterns of lending that focus on profitability at the expense of the poor (Bateman, 2010).

From a policy perspective, it is very important to examine whether there is a trade-off between outreach and financial sustainability (Hermes and Lensink, 2011). It is suggested, however, that financial sustainability and outreach may be compatible if MFIs focus on their original mission without over-reliance on subsidies (Morduch, 2007). Poorly designed subsidies can create wrong incentives that limit the scale of microfinance outreach. Furthermore, Bos and Millone (2015) suggest that mission drift only occurs if MFIs seek higher financial returns, but this effect is neutralized if MFIs are more cost-efficient. As the industry stresses the importance of financial sustainability, it raises the question of whether investors are willing to accept a decrease in returns to achieve higher outreach.

2.1. Microfinancing institutions and female borrowers in India

The MFI industry in India has seen unprecedented growth. As of 2018, there are 96 MFIs in India compared to 26 in Bangladesh, the second largest in terms of number of MFIs (Microfinance Information Exchange, 2018). The beginning of the microfinance sector in India dates back to the 1970s with the emergence of informal self-help groups to provide access to much-needed savings and credit services (EY paper). Following the liberalization of the Indian economy, MFIs in India began operating in the late 1990s based on Sangam social banking models (Sriram, 2010).

Originally, women would form groups to tackle social issues together as a form of empowerment. These social groups would then form Sangam social banking models which usually consist of five or more members that would be collectively responsible for the repayment of their loans. Social banking models are a type of self-help group program that encourages the development of self-reliance of poor people. This was prominent in India as a banking model that enabled the economic empowerment of women. The development of MFIs in India began as part of a developmental and poverty reduction project with the support of World Bank credits and NGOs. The focus of female borrowers has been a major feature of microfinance provision in India (Sharma et al., 2007), making it important to explore the role of MFIs through the dimension of gender.

Critics of microfinance doubt whether it has any positive impacts on female borrowers, especially in India, as some women borrow credit to pay other debts (Wichterich, 2012). Yet there is a vast body of research that has provided evidence on the positive impacts of microfinancing for female empowerment, especially in developing countries. Early research by Pitt and Khandker (1998) suggested that women have higher pay-back ratios and use the income for health and education (Haini, 2020c). This is supported by various research studies that suggest that women are more likely to use resources in ways that improve family well-being (Rawlings and Rubio, 2005; Handa and Davis, 2006). Pitt et al. (2006) compared the effectiveness of microfinance between male and female borrowers; they find that the number of female borrowers has a significant effect on most measures of empowerment such as control of resources, finance and freedom of movement. On the other hand, the participation of male borrowers has a negative effect on these measures of empowerment for women. Furthermore, it is suggested that female members of MFIs are more empowered compared to non-members, as they have more control over savings, income generation, decision-making and freedom of mobility (Kato and Kratzer, 2013). Accordingly, it is suggested that female empowerment is strongly correlated with microfinance outreach (Laha and Kuri, 2014). In addition, some MFIs actively target female borrowers in India and have extended their social mission by partnering with health insurance companies to provide insurance, and consequently health empowerment, to the poor (Rai and Ravi, 2011). It is fair to conclude, therefore, that MFIs play a significant role in the economic empowerment of women in developing countries such as India.

The industry in India grew rapidly between 2005 and 2010 (Ghosh, 2013). There was a strong demand for loans from borrowers neglected by the banking system and investors were eager to invest funds in the high growth industry. As a result, gross loans of MFIs grew six-fold during this time, suggesting excessive expansion and leading to the Indian microfinance crisis, also known as the Andhra Pradesh crisis. Many private profit-making MFIs leveraged the existing self-help group programs and pressured the repayment of MFI loans. The main problem was that these profit-making MFIs were built on existing self-help group programs that were dependent on NGOs. Ghosh (2013) suggested that competition amongst MFIs forced them to concentrate on certain geographical areas that were more profitable. This not only affected their original mission of outreach, but also created patterns of multiple lending and borrowing to the same consumers which created saturation and excessive competition (Ghosh, 2013). Furthermore, Priyadarshee and Ghalib (2011) suggested that some MFIs even collaborated with consumer goods companies to worsen the indebtedness of poor households. The aggressive lending of MFIs draws parallels between the Indian microfinance crisis and the global financial crisis of 2008, where banks would lend without due diligence. At the peak of its crisis, many borrowers were heavily indebted and some even took their own lives (Ghosh, 2013). The Ministry of Finance in India and Reserve Bank of India responded to the crisis by drafting the MFIs (Development and Regulation) Bill in 2012, to set rules and regulations that focused on consumer protection, such as limiting over-lending, multiple borrowing and coercive means of gaining repayment. This conducive policy and regulatory environment for MFIs were established in order to expand the financial inclusion agenda in India.

3. DEA Studies on MFIS’ Efficiency

Examining the efficiency of financial institutions has been well-studied as it provides insight into managerial practices. MFIs are a special type of financial institution that provides microfinance for the poor. There are two main approaches to examine the efficiency of financial institutions; the parametric approach, stochastic frontier analysis (SFA), and the nonparametric approach, DEA. We use the nonparametric DEA approach as it requires no functional form and fewer assumptions of the underlying technology frontier. Although there are MFI examinations that employ SFA (Paxton, 2007; Hermes et al., 2011; Servin et al., 2012), it may not be suitable in the case of India. Regulatory changes restrict Indian MFIs to collect deposits and this may lead to misspecification of the functional form when employing SFA. We still, however, review MFI examinations that employ SFA and DEA to provide a better insight into the efficiency of MFIs.

The philosophical perspective of what a financial institution does and the definition of its efficiency has allowed different models to emerge through the selection of inputs and outputs. Two basic models are prevalent in the literature: intermediation and production. Under the intermediation model, financial institutions collect deposits and make loans to make a profit. In studies examining financial institutions’ efficiency, deposits and loans are used as the traditional inputs and outputs, respectively. Under the production model, a financial institution uses physical resources, such as labor and capital in order to process transactions, take deposits, lend funds and so on. Labor and assets are treated as inputs whereas deposits and loans are treated as outputs (Vassiloglou and Giokas, 1990; Soteriou and Zenios, 1999). The selection of inputs and outputs is crucial for modeling financial institutions’ efficiency as changes in input and output specification can lead to different efficiency scores (Berger and Humphrey, 1997).

There are several cross-country examinations of MFIs employing DEA. Gutiérrez-Nieto et al. (2007) examine the efficiency of 30 Latin American MFIs and find that efficiency is influenced by the location of MFIs (country effect) and their institutional types, using two inputs (number of credit officers and operating expenses) and three outputs (interest and fee income, gross loan portfolio and number of outstanding loans). Bassem (2008) investigates the efficiency of 35 MFIs in the Mediterranean zone during the period 2004–2005 and concludes that the size of institutions negatively affects their efficiency. In addition, Haq et al. (2010) examine the efficiency of 39 MFIs across Africa, Asia and Latin America and find that institutional type NGO-MFIs are more efficient under the production approach. Segun and Anjugam (2013) examine the efficiency of 75 MFIs in 25 Sub-Saharan African countries and find that MFIs are inefficient in meeting the goals of either providing microfinance-related services to their clients or intermediating funds between borrowers and depositors. There may be weaknesses in examining cross-country studies. Cross-country studies may not fully acknowledge the significance of country characteristics such as macroeconomic environments and differences in regulatory framework and level of competition (Haini, 2019 and 2020f).

We identify several individual country examinations of MFIs employing DEA. Nghiem et al. (2006) examine the technical efficiency of 49 microfinance schemes in Vietnam and find that the average technical efficiency of all microfinance schemes is 80% and age and location of the schemes have an influence on efficiency. In addition, Piot-Lepetit and Nzongang (2014) investigate the possible trade-off between outreach and sustainability within 52 village banks in Cameroon and find that majority of the institutions do not show a trade-off. To the best of our knowledge, there have been few studies examining the efficiency of Indian MFIs (Babu and Kulshreshtha, 2014; Kar and Deb, 2017). Babu and Kulshreshtha (2014) examine the Malmquist TFP index for a panel of 34 Indian MFIs during 2005–2011 and find that there is a decline in total factor productivity over the period. Moreover, Kar and Deb (2017), using data of 31 Indian MFIs for the period 2009–2015, find that the technical efficiency of MFIs in India is estimated to be 79% and find that sustainability has a positive impact on efficiency.

These individual examination studies, however, may still have limitations. Many of the studies above, including the cross-country examinations, employed the conventional DEA estimator, which are biased by construction and are sensitive to sampling variations (Simar and Wilson, 1998). Furthermore, the studies employ the censored (Tobit) regression to identify the determinants of efficiency estimates. The findings could be biased as the scores may be correlated. We aim to overcome these weaknesses by examining India as an individual country, as India is one of the largest and regulated MFI industry, it may not be prudent to examine MFI efficiency across countries and makes more sense to examine the efficiency of MFIs within the same country (Balkenhol, 2007). Furthermore, we employ a two-stage double bootstrap approach to avoid biased estimations, which is detailed in Sec. 4.

4. Data and Methodology

The following section outlines the methodology that will be used in the first and second stages and explains the use of VRS as the technology frontier. The choice of input and output variables used in the first stage are then described and justified as well as the environmental variables employed in the second-stage truncated regression.

Based on the work of Farrell (1957), DEA was developed by Charnes et al. (1978) and is a linear-programing method for assessing the relative performance and efficiency of peer decision-making units (Thanassoulis, 1999). DEA is a nonparametric deterministic model that builds a production frontier based on the multiple input–output observations from a sample. DEA is appropriate for examining MFI efficiency as its ability to handle quantitative and qualitative data (Golany and Storbeck, 1999) and it has been employed widely in studies examining bank and MFI efficiency.

4.1. Nonparametric test of returns to scale

We begin our methodology by identifying the optimal scale of global technology. The assumption behind the optimal scale of global technology is important as the results of technical efficiency can vary. Seiford and Zhu (1999) suggested that the sensitivity of returns to scale can be related to the efficient frontier and changes in the position of the efficient firms along the frontier. We formally test the optimal scale of technology as suggested by Simar and Wilson (2002); assuming constant returns to scale when it is not optimal will lead to inconsistent results as it overestimates efficiency scores. They suggest the following tests:

Test #1: $H_{0}$ : Technology is globally CRS $H_{1}$ : Technology is VRS.

Test #2: $H_{0}$ : Technology is globally NiRS $H_{1}$ : Technology is VRS.

If the null hypothesis in Test #1 is rejected, where technology is globally CRS, then we can proceed to Test #2. Likewise, if the null hypothesis is rejected, we can formally assume that the scale exhibits VRS. Test #1 and Test #2 test the following measures:

Ŝo2n=∑Kk=1ˆθok(yk,xk,y,x|CRS)∑Kk=1ˆθok(yk,xk,y,x|VRS),<math display="block" altimg="eq-00005.gif"><msubsup><mrow><mi>Ŝ</mi></mrow><mrow><mn>2</mn><mi>n</mi></mrow><mrow><mi>o</mi></mrow></msubsup><mo>=</mo><mfrac><mrow><msubsup><mrow><mo>∑</mo></mrow><mrow><mi>k</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>K</mi></mrow></msubsup><msubsup><mrow><mover accent="true"><mrow><mi>θ</mi></mrow><mo>̂</mo></mover></mrow><mrow><mi>k</mi></mrow><mrow><mi>o</mi></mrow></msubsup><mo stretchy="false">(</mo><msub><mrow><mi>y</mi></mrow><mrow><mi>k</mi></mrow></msub><mo>,</mo><msub><mrow><mi>x</mi></mrow><mrow><mi>k</mi></mrow></msub><mo>,</mo><mi>y</mi><mo>,</mo><mi>x</mi><mi>|</mi><mstyle><mtext mathvariant="normal">CRS</mtext></mstyle><mo stretchy="false">)</mo></mrow><mrow><msubsup><mrow><mo>∑</mo></mrow><mrow><mi>k</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>K</mi></mrow></msubsup><msubsup><mrow><mover accent="true"><mrow><mi>θ</mi></mrow><mo>̂</mo></mover></mrow><mrow><mi>k</mi></mrow><mrow><mi>o</mi></mrow></msubsup><mo stretchy="false">(</mo><msub><mrow><mi>y</mi></mrow><mrow><mi>k</mi></mrow></msub><mo>,</mo><msub><mrow><mi>x</mi></mrow><mrow><mi>k</mi></mrow></msub><mo>,</mo><mi>y</mi><mo>,</mo><mi>x</mi><mi>|</mi><mstyle><mtext mathvariant="normal">VRS</mtext></mstyle><mo stretchy="false">)</mo></mrow></mfrac><mo>,</mo></math>(1)

Ŝo2n=∑Kk=1ˆθok(yk,xk,y,x|NIRS)∑Kk=1ˆθok(yk,xk,y,x|VRS).<math display="block" altimg="eq-00006.gif"><msubsup><mrow><mi>Ŝ</mi></mrow><mrow><mn>2</mn><mi>n</mi></mrow><mrow><mi>o</mi></mrow></msubsup><mo>=</mo><mfrac><mrow><msubsup><mrow><mo>∑</mo></mrow><mrow><mi>k</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>K</mi></mrow></msubsup><msubsup><mrow><mover accent="true"><mrow><mi>θ</mi></mrow><mo>̂</mo></mover></mrow><mrow><mi>k</mi></mrow><mrow><mi>o</mi></mrow></msubsup><mo stretchy="false">(</mo><msub><mrow><mi>y</mi></mrow><mrow><mi>k</mi></mrow></msub><mo>,</mo><msub><mrow><mi>x</mi></mrow><mrow><mi>k</mi></mrow></msub><mo>,</mo><mi>y</mi><mo>,</mo><mi>x</mi><mi>|</mi><mstyle><mtext mathvariant="normal">NIRS</mtext></mstyle><mo stretchy="false">)</mo></mrow><mrow><msubsup><mrow><mo>∑</mo></mrow><mrow><mi>k</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>K</mi></mrow></msubsup><msubsup><mrow><mover accent="true"><mrow><mi>θ</mi></mrow><mo>̂</mo></mover></mrow><mrow><mi>k</mi></mrow><mrow><mi>o</mi></mrow></msubsup><mo stretchy="false">(</mo><msub><mrow><mi>y</mi></mrow><mrow><mi>k</mi></mrow></msub><mo>,</mo><msub><mrow><mi>x</mi></mrow><mrow><mi>k</mi></mrow></msub><mo>,</mo><mi>y</mi><mo>,</mo><mi>x</mi><mi>|</mi><mstyle><mtext mathvariant="normal">VRS</mtext></mstyle><mo stretchy="false">)</mo></mrow></mfrac><mo>.</mo></math>(2)

In both cases, if the null is rejected then the average distance between VRS and CRS or NIRS is on average large and

$Ŝ_{2 n}^{o}$ is significantly larger than 1. The existence of scale economies means that the VRS model needs to be employed to accommodate scale effect in analysis (Dyson et al., 2001).

4.2. First-stage DEA model and bias-corrected procedure

In the first stage, we calculate the biased-corrected radial efficiency scores employing the input-oriented VRS model with a bootstrap. The VRS model, based on the work of Banker et al. (1984), assumes variation between input and outputs and can be formulated by applying the input or output orientation perspective. The input-oriented approach assumes that efficiency is calculated as a comparative reduction in inputs, holding output levels constant. Assuming input-oriented VRS, we calculate the conventional Debreu–Farrell radial nonparametric efficiency measurement (Debreu, 1951; Farrell, 1957). For each data point $k (k = 1, \dots, K)$ vector $x_{k} = (x_{k 1}, \dots, x_{k N}) \in ℜ^{N}$ denotes $N$ inputs, vector $y_{k} = (y_{k 1}, \dots, y_{k M}) \in ℜ^{M}$ denotes $M$ outputs. We assume that under technology $T$ the data $(x, y)$ are such that outputs are producible by inputs

T = {(x, y) : y are producible by x} . <math display="block" altimg="eq-00015.gif"><mi>T</mi><mo>=</mo><mo stretchy="false">{</mo><mo stretchy="false">(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo stretchy="false">)</mo><mo>:</mo><mi>y</mi><mstyle><mtext mathvariant="normal"> are </mtext></mstyle><mstyle><mtext mathvariant="normal">producible </mtext></mstyle><mstyle><mtext mathvariant="normal">by </mtext></mstyle><mi>x</mi><mo stretchy="false">}</mo><mo>.</mo></math> (3)

The technology is fully characterized by its production possibility set,

P (x) \equiv {y : (x, y) \in T}, <math display="block" altimg="eq-00016.gif"><mi>P</mi><mo stretchy="false">(</mo><mi>x</mi><mo stretchy="false">)</mo><mo>\equiv</mo><mo stretchy="false">{</mo><mi>y</mi><mo>:</mo><mo stretchy="false">(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo stretchy="false">)</mo><mo>\in</mo><mi>T</mi><mo stretchy="false">}</mo><mo>,</mo></math> (4)

or input requirement set,

L (y) \equiv {x : (x, y) \in T} . <math display="block" altimg="eq-00017.gif"><mi>L</mi><mo stretchy="false">(</mo><mi>y</mi><mo stretchy="false">)</mo><mo>\equiv</mo><mo stretchy="false">{</mo><mi>x</mi><mo>:</mo><mo stretchy="false">(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo stretchy="false">)</mo><mo>\in</mo><mi>T</mi><mo stretchy="false">}</mo><mo>.</mo></math> (5)

Conditions (4) and (5) imply that the available outputs and inputs are feasible. The frontier is defined by the lower boundary of the input requirement set and the upper boundary of the production possibility set. We employ the input-oriented VRS model of radial efficiency measurement as we assume managers of MFIs have more control over inputs than output quantities (Wijesiri et al., 2015). For

$K$ data points,

$M$ outputs and

$N$ inputs an estimate of the Debreu–Farrell input-based measure of technical efficiency can be calculated by solving a linear programing problem for each data point

$k (k = 1, \dots, K)$ :

where

$x$ is a

$K \times N$ matrix of data on inputs and

$y$ is a

$K \times M$ matrix of data on outputs. Equation (6) is our DEA specification and will result in scores ranging between 0 and 1. The benchmark MFI will have the highest efficiency score of 1 and the MFIs that are assigned a score of lower than 1 are relatively inefficient. In other words, the values of inputs and outputs are located some distance from the corresponding reference point on the production frontier. Note that we discuss our inputs and outputs in the next section.

Conventional DEA techniques suffer from several constraints that may limit the interpretation of results to decision-makers (Ferrier and Hirschberg, 1997). As the production frontier is measured and determined by the input and output sample, efficiency is only measured as a relative and can be biased based on the sampling variation (Simar and Wilson, 1998). The inherent advantage of employing DEA to assess relative units is also its limitation as it assumes a point estimate without a sense of sampling variation (Haini, 2020d and 2020e).

We employ the bootstrap concept introduced by Simar and Wilson (1998) based on Efron (1979), which eliminates the limitation of conventional DEA. The bootstrap repeatedly simulates the data generating process and applies the original estimator, providing confidence intervals and corrections for the bias and converging toward the true efficient frontier without limiting the advantages of DEA.

4.3. Second-stage truncated regression

In the second stage, we examine the effect of various environmental variables on our biased-corrected efficiency scores obtained from the first stage. The most commonly used regression is the Tobit estimator. The use of Tobit regression in the second-stage analysis, however, has been criticized by Simar and Wilson (2007). They argue that the assumption where the error term is independent of the explanatory variables is invalid when employing the Tobit regression, as they are correlated. Furthermore, they suggest that the DEA efficiency scores may be correlated with each other leading to further biased estimations in the second stage. To address these issues, they propose an alternative double bootstrap that allows for valid inference using the biased efficiency scores. We employ the double bootstrap method (Algorithm 2) (Simar and Wilson, 2007) where the bias-corrected efficiency scores ( ${\hat{θ}}_{k}^{*}$ ), produced in the first stage of the analysis, are regressed on a set of explanatory variables ( $z_{k}$ ) using the following regression specification:

ˆ θ * k = α + z k β + ε i, i = 1, \dots, n, <math display="block" altimg="eq-00029.gif"><msubsup><mrow><mover accent="true"><mrow><mi>θ</mi></mrow><mo>̂</mo></mover></mrow><mrow><mi>k</mi></mrow><mrow><mo>*</mo></mrow></msubsup><mo>=</mo><mi>α</mi><mo>+</mo><msub><mrow><mi>z</mi></mrow><mrow><mi>k</mi></mrow></msub><mi>β</mi><mo>+</mo><msub><mrow><mi>ε</mi></mrow><mrow><mi>i</mi></mrow></msub><mo>,</mo><mspace width="1em"></mspace><mi>i</mi><mo>=</mo><mn>1</mn><mo>,</mo><mo>\dots</mo><mo>,</mo><mi>n</mi><mo>,</mo></math> (7)

ˆ θ * i = α + β 1 Age + β 2 Size + β 3 Legal + β 4 ROA + β 5 LVR + β 6 TWB + β 7 OUT + ε i, <math display="block" altimg="eq-00030.gif"><mrow><msubsup><mrow><mover accent="true"><mrow><mi>θ</mi></mrow><mo>̂</mo></mover></mrow><mrow><mi>i</mi></mrow><mrow><mo>*</mo></mrow></msubsup><mo>=</mo><mi>α</mi><mo>+</mo><msub><mrow><mi>β</mi></mrow><mrow><mn>1</mn></mrow></msub><mstyle><mtext mathvariant="normal">Age</mtext></mstyle><mo>+</mo><msub><mrow><mi>β</mi></mrow><mrow><mn>2</mn></mrow></msub><mstyle><mtext mathvariant="normal">Size</mtext></mstyle><mo>+</mo><msub><mrow><mi>β</mi></mrow><mrow><mn>3</mn></mrow></msub><mstyle><mtext mathvariant="normal">Legal</mtext></mstyle><mo>+</mo><msub><mrow><mi>β</mi></mrow><mrow><mn>4</mn></mrow></msub><mstyle><mtext mathvariant="normal">ROA</mtext></mstyle><mo>+</mo><msub><mrow><mi>β</mi></mrow><mrow><mn>5</mn></mrow></msub><mstyle><mtext mathvariant="normal">LVR</mtext></mstyle><mo>+</mo><msub><mrow><mi>β</mi></mrow><mrow><mn>6</mn></mrow></msub><mstyle><mtext mathvariant="normal">TWB</mtext></mstyle><mo>+</mo><msub><mrow><mi>β</mi></mrow><mrow><mn>7</mn></mrow></msub><mstyle><mtext mathvariant="normal">OUT</mtext></mstyle><mo>+</mo><msub><mrow><mi>ε</mi></mrow><mrow><mi>i</mi></mrow></msub><mo>,</mo></mrow></math>

where

$α$ is a constant term,

$β$ is a vector of parameters and

$ε_{i}$ is the error term.

We examine several environmental variables that may affect the efficiency of Indian MFIs. Age refers to the number of years an MFI has been in operation and can be a proxy of experience. There are conflicting studies with regards to the effect of MFI age and efficiency. On the one hand, Hermes and Lensink (2011) suggest that older MFIs are less efficient, as younger MFIs can specialize and acquire successful business models from existing MFIs. On the other hand, younger MFIs may be less efficient due to higher initial start-up costs (Paxton, 2007). It is suggested that MFI age is positively associated with efficiency as they gain knowledge and information about the industry reducing information asymmetry and maintain repeat clients. We use a categorical variable that takes a value of unity if an MFI has been in operation for more than eight years and is defined as mature, based on the Microfinance Information Exchange (MIX) market definition database (Microfinance Information Exchange Inc., 2018). We examine the effects of MFI age on efficiency as previous studies have inconclusive findings. The capacity of MFIs to compete with others and its market awareness may be proxied by the size of MFI (Staub et al., 2010; Nhung and Okuda, 2015). It is suggested that size accounts for differences in technology, investment opportunities and operations which may increase efficiency. Due to data availability, we categorize MFI size by total assets into three categories; small refers to MFIs with total assets of less than US$5 million, medium refers to MFIs with total assets of US$5–50 million and large refers to MFIs with over US$50 million total assets (Microfinance Information Exchange Inc., 2018).

We also employ a dummy variable to account for the effects of legal status on efficiency. We examine whether NGOs have different effects on their corporate governance and operations that can affect efficiency. Many of the earlier self-help group programs in India transitioned into NGOs, which focused more on outreach. Differences in corporate governance and ownership may affect their efficiency through their objectives. As the Microfinancing Bill in India regulates MFI activity and operations, it may be interesting to examine whether legal status may have any effects on efficiency. We also include several control variables in the study. Return on assets (ROA) is a proxy for profitability and provides insight into the objectives of MFIs. Leverage intensity (LVR) measures the differences in risk taking of MFIs (Hermes et al., 2011). Higher leverage intensity may suggest an increase in risk-taking from managers and may lead managers to manage the MFIs in a more cautious manner, consistent with the agency cost hypothesis (Harris and Raviv, 1991). Outreach refers to the total number of borrowers an MFI has. Small outreach refers to MFIs with less than 10,000 borrowers, medium outreach refers to MFIs with 10,000–30,000 borrowers and large outreach refers to MFIs with over 30,000 borrowers. TWB is a dummy variable that takes the value of one if an MFI actively targets female borrowers.

4.4. Data sources

We compiled data from MIX market database which is available online. The MIX database collects self-reported balance sheet information and has been used in many empirical studies. We used data on 84 Indian MFIs from 2016 to 2018 and average the data, dropping 20 observations from the initial sample set due to missing data. We use the averages based on the suggestion put forth by Ruggiero (2007), which emphasizes that the use of averages has been shown to effectively smooth out measurement errors and reduce the sensitivity of statistical noise.

4.5. Selection of input and output variables (first-stage)

We construct a single DEA model using an input and output measure to estimate the efficiency of MFIs from both an outreach and sustainability perspective. We select the total number of loan officers and operating expenses as our input variables and our three output variables are interest income, gross loan portfolio and total number of active female borrowers. Our input and output selection considers both production and intermediation models and is based on the performance measurement framework suggested by Yaron (1994).

The operating expenses, interest income and gross loan portfolio aim to capture financial sustainability, consistent with an intermediation model. Generally, the traditional intermediation model includes deposits and loans to make a profit (Berger and Humphrey, 1991). However, MFIs in India are restricted by regulation to collect deposits thus we used a modified intermediation model. Furthermore, we are concerned with financial sustainability as opposed to profitability, due to the nature of MFIs. In this case, MFIs should generate enough income to repay the opportunity cost of all inputs and assets to focus on their mission of outreach. Operating expenses considers all the costs related to the operations of MFIs (Gutiérrez-Nieto et al., 2007). Various studies have examined the efficiency of banks and MFIs using operating expenses as an input (Berger and Humphrey, 1997). The inclusion of operating expenses complements the output selection of interest income, which is considered an indicator of operational sustainability. MFIs need to collect enough income and operate at a low cost. This allows an MFI to operate and survive in the long run. Furthermore, the inclusion of gross loan portfolio is consistent with the intermediation model (Berger and Humphrey, 1997).

The total number of loan officers, as an input, and the total number of active female borrowers, as an output, consider the extent of outreach of MFIs and are consistent with the production model. Loan officers have been employed as an input in various studies (Gutiérrez-Nieto et al., 2007; Segun and Anjugam, 2013). As the microfinance landscape in India developed from clients of self-help group programs and communities in rural areas, it is necessary for MFIs to minimize adverse selection through screening potential borrowers. This process of identifying, screening, negotiating, and monitoring loans and repayment is accomplished through employing loan officers. The total number of loan officers may be more appropriate than total number of employees as loan officers’ activities are closely related to essential MFI operations. The total number of active female borrowers is an indirect proxy for the depth of outreach as it considers the lack of access of borrowing for women in India (Yaron et al., 1998).

5. Empirical Results

We identify our returns to scale prior to calculating DEA efficiency scores. Summary statistics of our input–output variables and environmental variables are given as well. We then present our DEA efficiency scores and bias-corrected DEA scores and report the results of our second-stage truncated regression.

5.1. Nonparametric test of returns to scale

We begin by formally identifying the returns to scale using the nonparametric test of returns to scale and analysis of scale efficiency (Simar and Wilson, 2007), Table 1 presents these results. We find that both tests reject the null hypothesis of constant returns to scale and non-increasing returns to scale and therefore we can assume that the global technology frontier exhibits varying returns to scale.

**Table 1. Nonparametric test of returns to scale.**
	Test #1	Test #2
t-statistics	0.877	0.968
p-value	0.00	0.00

5.2. Summary statistics

Tables 2 and 3 present our summary statistics. The first five variables in Table 2 relate to the first stage of calculating bias-corrected efficiency scores, where loan officers and operating expenses are the input variables, and interest income, gross loan portfolio, and female borrowers are the output variables. There is a considerably large variation across the input and output variables as expected, as the nonparametric test of returns to scale suggests that the technology frontier exhibits varying returns to scale. The industry in India is one of the largest in the world, with large MFIs that have been in operation for many years, such as Janalakshmi Financial Services, a NBFI-MFI which operates in over 19 states in India, alongside smaller MFIs such as Bal Mahila Kalyan, an NGO-MFI which only operates in the Bihar state. The summary statistics of the environmental variables for the second stage, are ROA and LVR in Table 2 and all the variables in Table 3. There is considerable variation across MFIs for the environmental variables in the second stage.

**Table 2. Summary statistics ( $N = 84$ ).**
Variable	Mean	SD	Min	Max
Loan officers	758.76	1939.38	3.00	14,067.00
Operating expenses	7158.27	23,622.33	21.39	195,367.00
Interest income	15,178.63	50,005.35	37.30	404,225.80
Gross loans portfolio	102,882.40	291,535.60	262.07	1,974,730.00
Female borrowers	378,411.50	957,059.80	1519.00	5,888,750.00
ROA	0.118	0.902	−0.310	8.280
LVR	4.282	2.860	0.044	16.861

Note: ROA refers to return on assets and is the profit after tax divided by total assets. LVR refers to leverage intensity and is the ratio of total liabilities divided by total equity.

**Table 3. Summary statistics (categorical variables, $N = 84$ ).**
Variable		Frequency	%	Cum. (%)
Age	New and Young	18	21.43	21.43
	Mature	66	78.57	100
Size	Small	31	36.92	36.90
	Medium	33	39.29	76.19
	Large	20	23.81	100
Legal	Other	53	63.10	63.10
	NGO	31	36.90	100
Outreach	Small	15	17.86	17.86
	Medium	15	17.86	35.72
	Large	54	64.28	100
Target women borrowers	No	17	20.24	20.24
	Yes	67	79.76	100

Note: New and young refers to MFIs that are in operation for 1–4 years, and 5–8 years, respectively. Mature refers to MFIs that are over 8 years. Small refers to MFIs with total assets of less than US$5 million. Medium refers to MFIs with total assets of US$5–50 million. Large refers to MFIs with over US$50 million total assets. Legal refers to the legal status of MFIs, where NGO are NGOs. Other refers to non-bank financial institutions, cooperatives, credit unions and others. Small outreach refers to MFIs with less than 10,000 borrowers. Medium outreach refers to MFIs with 10,000–30,000 borrowers and large outreach refers to MFIs with over 30,000 borrowers.

5.3. First-stage efficiency scores summary statistics

We present the summary statistics of our input-oriented VRS model of radial efficiency scores in Table 4. There are 15 efficient MFIs in the conventional DEA scores with some variation across MFIs. The bootstrap procedure, however, reduces this biasedness and provides efficiency scores closer to the true frontier as there are no efficient MFIs in the bias-corrected scores. The bias-corrected efficiency scores present a lower variation of efficiency scores as expected, with a maximum efficiency score of 85.4% and a minimum efficiency score of 48.7%.

**Table 4. First-stage DEA efficiency scores ( $N = 84$ ).**
Variable	Mean	SD	Min	Max	Efficient DMUs
DEA scores	0.677	0.210	0.196	1.000	15
Bias-corrected DEA scores	0.487	0.164	0.134	0.854	0

5.4. Second-stage truncated regression

We run three sets of regression in the second stage. All three sets contain Age, Size and Legal, however, Set (1) drops Outreach as a control variable, Set (2) drops TWB as a control variable and Set (3) drops Outreach and TWB as control variables.

Table 5 presents our second-stage truncated regression results. Age provides statistically insignificant results in all three sets of regressions. This suggests that age of MFIs has minimal effects on MFI efficiency. The size of MFIs also shows positive and significant results in all three sets of regression suggesting that larger MFIs tend to be more efficient. We find that the legal status of MFIs is statistically insignificant with respect to efficiency. It was suggested that the legal status of MFIs affects corporate governance, which has implications on MFI goals and objectives. NGOs are generally not-for-profit and may focus on their social goals of outreach whilst other MFIs may be concerned with their financial sustainability. Our control variables provide expected findings. We find expected results for TWB and outreach; MFIs that actively target women borrowers are associated with higher efficiency as TWB is statistically significant and positive and the same can be said about MFIs with high outreach. Our profitability proxy, ROA, is statistically insignificant and may not affect efficiency scores while leverage intensity, LVR, is statistically significant and positive for MFI efficiency.

**Table 5. Second-stage truncated regression ( $N = 84$ ).**
Variable	Set (1)	Set (2)	Set (3)
Mature (Age)	0.031	0.024	0.030
	(0.031)	(0.030)	(0.029)
Medium (Size)	0.093**	0.051*	0.094***
	(0.029)	(0.040)	(0.029)
Large (Size)	0.282***	0.236***	0.282***
	(0.036)	(0.045)	(0.035)
Legal	$-$ 0.023	$-$ 0.022	$-$ 0.018
	(0.029)	(0.029)	(0.028)
Target women borrowers	0.006*
	(0.030)
Medium (Outreach)		0.024
		(0.043)
High (Outreach)		0.071**
		(0.046)
ROA	$-$ 0.031	$-$ 0.033	$-$ 0.031
	(0.023)	(0.022)	(0.022)
LVR	0.007*	0.006*	0.007*
	(0.004)	(0.004)	(0.004)
Bootstrap repetitions	2,000	2,000	2,000
Wald Chi2	105.28	104.64	105.45

Note: Definition of variables are in Tables 4 and 5. Absolute values of $z$ -statistics are reported in parentheses *, **, *** indicating significant levels of 10%, 5%, and 1%, respectively.

6. Discussion

We begin our discussion by reviewing the DEA efficiency scores in Table 4. We find that the average bias-corrected MFI efficiency score is lower than the average conventional score, as expected. We then discuss the results of the second-stage truncated regression in Table 5. We find the following environmental variables to be positive and significant to efficiency; size of MFI, leverage intensity, higher outreach and MFIs that actively target female borrowers. We find the following environmental variables to be insignificant, age of MFI, legal status of MFI and profitability. We provide some explanation and discussion to these findings.

In Table 4, we find that the average bias-corrected score of MFIs is 48.7%, which is moderately low on average. These findings contrast those of Kar and Deb (2017) of 79%. These results, however, may be questionable as they only use a small sample size of 31 Indian MFIs and employed conventional DEA and the biased Tobit regression. This upward biased efficiency score can be seen in our average conventional DEA in Table 4, where we obtain 67.7%. Babu and Kulshreshtha (2014), examined 34 Indian MFIs in 2005–2011, find that total factor productivity of MFIs is declining over the 6 years. Although the findings refer to an older period, it may provide insight into the micro-financial crisis that MFIs experience during 2010, before regulatory changes occurred in 2012. The low average bias-corrected scores obtained may suggest that Indian MFIs are consolidating from the crisis and acclimatizing to the regulatory changes that may affect operations. Thus, the findings suggest that Indian MFIs need to enhance their scale of operations and improve their managerial practices to increase their productivity and promote the financial inclusion agenda.

We examine several environmental variables that may enhance the productivity and efficiency of MFIs with interesting findings. First, we find that age is insignificant to MFI efficiency which differs from other studies. Previous studies have suggested that MFI age is positively associated with their efficiency, as they specialized and capture the market in a sustainable way, as well as reducing information asymmetry (Caudill et al., 2009; Wijesiri et al., 2015). There may be a good explanation, however, as to why age is insignificant for Indian MFIs. Informal self-help groups began to establish in India in the 1970s to tackle social issues and MFIs only began operating in the early 1990s building upon and integrating into these existing networks. Furthermore, building on self-help group programs allows Indian MFIs to reduce their risk and information asymmetry as they effectively transfer risk from the lender to the borrowers through group loan methodology (Stiglitz, 1990). Therefore, older MFIs may not benefit as much from asymmetric information.

Size contributes positively to MFI efficiency in all sets of regression, suggesting that as MFIs grow larger they become more efficient. This is consistent with various studies that provide similar findings (Wijesiri et al., 2017). Larger MFIs may be able to reduce their costs through economies of scale and operate their network of branches in a more cost-effective way. Furthermore, larger MFIs may be able to offer repeat borrowers increasingly larger loans with lower screening costs. It was suggested that information asymmetry between a larger MFI and its clients could be low as larger MFIs have higher social awareness (Wijesiri et al., 2017).

The legal status of MFIs is insignificant to MFI efficiency in India in all sets of regression. Previous results suggest that NGOs tend to be more efficient at outreach while NBFIs are more financially efficient (Wijesiri et al., 2015). Babu and Kulshreshtha (2014) find that NGO-MFIs have higher technical efficiency compared to others using a sample of 34 MFIs in India during 2015–2011. However, in our case, we find that legal ownership is insignificant to the efficiency of MFIs. There are issues with previous studies as they have employed a censored (Tobit) regression, which can lead to upward biased results. Furthermore, it may also suggest that the regulatory changes after 2012, which limits the maximum amount of loan an MFI can lend, may have affected the way NBFI-NGOs operate. This may have restricted profit-seeking NBFI-MFIs to operate in ways that are different from NGO-MFIs, resulting in the legal status of MFIs to be insignificant to efficiency.

In Set (1), we find that MFIs that actively target female borrowers have positive and significant efficiency scores. MFIs that have policies to actively target female borrowers may provide a deeper insight into its corporate governance. It creates incentives to further promote outreach and complements the findings from Morduch (2007) who suggested that outreach and sustainability may be compatible. Similarly, in Set (2), we find MFIs with a higher outreach have positive and significant effects on efficiency scores. This is consistent with the view that increasing the depth of outreach may result in increased efficiency (Bos and Millone, 2015). However, we find that MFIs with high outreach, borrowers of more than 30,000, are significant and positive to efficiency, whilst medium outreach, borrowers between 10,000 and 30,000, are insignificant. This may complement the finding on the size of MFIs as it suggests that MFIs with a higher number of borrowers may benefit from economies of scale.

We also find evidence that ROA is insignificant. This may suggest that profitability or loss is not an important determinant of efficiency. Profitability may not be an important objective for MFIs as they are more concerned with sustainability. At its core, sustainability is not profitability, as it refers to the ability of institutions to cover their operating costs using operating revenue generated from their core activities (Ledgerwood, 1999). Although there are some studies that suggest that a profitable microfinance may be able to reach more people with less dependence on donor funds (Christen, 2001) it may be more prudent for MFIs to obtain a balance between sustainability and outreach. Our results, however, may also suggest that a trade-off does not exist as profitability is insignificant. It may suggest that both profitable and non-profitable MFIs can increase efficiency in outreach and sustainability, consistent with previous findings (Cull et al., 2007; Brau and Woller, 2004). Understanding the trade-offs between profitability and outreach has always been an important debate for policymakers, and we find no evidence that a trade-off exists in the case of Indian MFIs.

Last, we find evidence that leverage intensity is positive and significant toward MFI efficiency. It could be suggested that higher leverage intensity, where debt-to-equity ratio is higher, may increase efficiency through an increase in the number of borrowers. In addition, it is also suggested that a higher debt-to-equity ratio results in managing the MFIs in a more cautious way, according to the agency cost hypothesis (Harris and Raviv, 1991). Providing microfinance to the financially excluded is considered a risky business as they do not offer collateral and with minimal background information. The group loan methodology (Stiglitz, 1990), however, reduces this risk as the group is responsible when default occurs. This may allow MFIs to take a considerably large number of borrowers increasing debt-to-equity ratio, complementing the findings on higher outreach and size of MFIs being positive to efficiency.

The average level of MFI efficiency in India is low at 48.7%, suggesting that there may be room for improvement. We identify and examine potential determinants of efficiency that may have policy implications to allow an MFI to increase its productivity, we outline these in the following section.

7. Concluding Remarks

MFIs play an important role in promoting financial inclusion and the economic empowerment of women. Using DEA and a two-stage double bootstrap approach we examine the determinants of MFI efficiency in India from 2016 to 2018. Previous research has suggested that a trade-off exists between financial sustainability and depth of outreach. More recent research, however, has suggested that sustainability and outreach may be compatible; the selection of inputs and outputs in our DEA model takes this perspective into consideration. The total number of loan officers is employed as an input as they are directly involved with MFI activity. Operating expenses are used as an input to consider the financial sustainability of MFIs. We use the total number of female borrowers as an output to proxy for depth of outreach, while interest income and gross loan portfolio, as outputs, measure financial sustainability.

We employ a VRS-input-oriented radial DEA model, as we assume managers have more control over the inputs than outputs. A VRS model is used as both the nonparametric tests of returns to scale rejected the null-hypothesis of CRS and NiRS. We use a bootstrapped DEA to produce bias-corrected efficiency scores as conventional DEA overestimates efficiency scores. We then regress the bias-corrected scores against a set of environmental variables using a truncated bootstrap regression to identify the determinants of efficiency. We find the following to be positive determinants of MFI efficiency: size, outreach, actively targeting female borrowers and leverage. ROA was found to be a negative determinant of MFI efficiency and the age and legal status of MFIs was found to be insignificant.

Our findings have several policy implications for the Indian government and managers of MFIs. As size of MFIs positively affects efficiency, the government should support existing MFIs to gain more industry knowledge and serve the poor more effectively whilst benefitting from economies of scale through growth. The age of MFI is insignificant to efficiency, as it was suggested that MFIs in India were built upon pre-existing self-help group networks. We find that the legal status of MFIs is insignificant to efficiency implying that the MFIs (Development and Regulation) Bill has been successful. The government and Reserve Bank of India are encouraged to continue protecting consumers against unfavorable MFI activity such as over lending. With regards to the management of MFIs, we suggest actively targeting female borrowers and increasing depth of outreach to the poor as this has a positive effect on efficiency. Whilst we encourage the growth of MFIs, this must be done responsibly to ensure they contribute effectively to the financial inclusion agenda.

We have identified several possible avenues for future research. Our study is limited to examining the efficiency of MFIs in India from 2016 to 2018. Examining the efficiency of Indian MFIs using a Malmquist index can examine the efficiency over multiple periods, which may give further inference on the findings above and provide better policy implications. Furthermore, we examine India as a country-specific study that calculates the relative efficiency of Indian MFIs. It may be interesting to incorporate MFIs from different countries to draw comparisons between their efficiencies. This can provide insights into the way MFIs operate in different countries. In addition, our study focuses on the supply side of microfinancing, as we examine the determinants of MFIs efficiency. Future studies should focus on examining the demand side of MFIs, in particular on whether MFIs can alleviate poverty that it claims to achieve.

Acknowledgments

We are grateful for the suggestion provided by two anonymous referees. Finally, we would like to extend our thanks to participants at the DEA40 conference at Aston University and Finance and Financial Inclusion Workshop at Loughborough University for their comments and suggestions.

References

Babu, M. and P. Kulshreshtha (2014). Productivity Change and Technical Efficiency in Indian Microfinance Institutions. Studies in Microeconomics, 2(2), 165–200. Crossref, Google Scholar
Balkenhol, B. (2007). Efficiency and Sustainability in Microfinance. Microfinance and Public Policy (pp. 3–23). Palgrave Macmillan UK. Crossref, Google Scholar
Banker, R.D., A. Charnes and W.W. Cooper (1984). Some Models for Estimating Technical and Scale Inefficiencies in Data Envelopment Analysis. Management Science, 30(9), pp. 1078–1092. Crossref, Google Scholar
Bassem, B.S. (2008). Efficiency of Microfinance Institutions in the Mediterranean: An Application of DEA. Transition Studies Review, 15(2), pp. 343–354. Crossref, Google Scholar
Bateman, M. (2010) Why Doesn’t Microfinance Work? The Destructive Rise of Neoliberalism. Zed Books, London. Crossref, Google Scholar
Beck, T., A. Demirgüç-Kunt and P. Honohan (2009). Access to Financial Services: Measurement, Impact and Policies. The World Bank Research Observer, 24, pp. 119–145. Crossref, Google Scholar
Berger, A.N. and D.B. Humphrey (1991). The Dominance of Inefficiencies over Scale and Product Mix Economies in Banking. Journal of Monetary Economics, 28, pp. 117–148. Crossref, Google Scholar
Berger, A.N. and D.B. Humphrey (1997). Efficiency of Financial Institutions: International Survey and Directions for Future Research. European Journal of Operational Research, 98(2), pp. 175–212. Crossref, Google Scholar
Bos, J.W.B. and M. Millone (2015). Practice What You Preach: Microfinance Business Models and Operational Efficiency. World Development, 70, pp. 28–42. Crossref, Google Scholar
Brand, M. (2000). More Bang for the Buck: Improving Efficiency. Micro Banking Bulletin, 4, pp. 13–18. Google Scholar
Brau, J.C. and G.M. Woller (2004). Microfinance: A Comprehensive Review of the Existing Literature. Journal of Entrepreneurial Finance and Business Ventures, 9(1), pp. 1–26. Crossref, Google Scholar
Caudill, S.B., D.M. Gropper and V. Hartarska (2009). Which Microfinance Institutions are Becoming More Cost-Effective with Time: Evidence from a Mixture Model. Journal of Money, Credit and Banking, 41(4), pp. 651–672. Crossref, Google Scholar
Chibba, M. (2009). Financial Inclusion, Poverty Reduction and the Millennium Development Goals. European Journal of Development Research, 21(2), pp. 213–230. Crossref, Google Scholar
Charnes, A., W. W. Cooper and E. Rhodes (1978). Measuring the efficiency of decision making units. European Journal of Operational Research, 2(6), 429–444. Crossref, Google Scholar
Christen, R. (2001). The Transformation of Microfinance in Latin America. Working Paper. Consultative Group to Assist the Poorest (CGAP), Washington, DC. Google Scholar
Cull, R., K. Demirguc and J. Morduch (2007). Financial Performance and Outreach: A Global Analysis of Leading Micro Banks. The Economic Journal, 117, pp. 107–133. Crossref, Google Scholar
Debreu, G. (1951). The Coefficient of Resource Utilization. Econometrica, 19, pp. 273–292. Crossref, Google Scholar
Demirgüç-Kunt, A., L. Klapper, D. Singer, S. Ansar and J. Hess (2020). The Global Findex Database 2017: Measuring Financial Inclusion and Opportunities to Expand Access to and Use of Financial Services. The World Bank Economic Review, 34(S1), pp. S2–S8. Crossref, Google Scholar
Dyson, R.G., R. Allen, A.S. Camanho, V.V. Podinovski, C.S. Sarrico and E.A. Shale (2001). Pitfalls and Protocols in DEA. European Journal of Operational Research, 132(2), pp. 245–259. Crossref, Google Scholar
Efron, B. (1979). Bootstrap Methods: Another Look at the Jackknife. The Annals of Statistics, 7(1), pp. 1–26. Crossref, Google Scholar
Farrell, M.J. (1957). The Measurement of Productive Efficiency. Journal of the Royal Statistical Society. Series A (General), 120(3), pp. 253–290. Crossref, Google Scholar
Ferrier, G.D. and J.G. Hirschberg (1997). Bootstrapping Confidence Intervals for Linear Programming Efficiency Scores: With an Illustration using Italian Banking Data. Journal of Productivity Analysis, 8, pp. 19–33. Crossref, Google Scholar
Fisher, T. and M.S. Sriram (2002) Beyond Micro Credit — Putting Development Back into Micro-Finance. Vistar Publications, New Delhi. Google Scholar
Ghosh, J. (2013). Microfinance and the Challenge of Financial Inclusion for Development. Cambridge Journal of Economics, 37(6), pp. 1203–1219. Crossref, Google Scholar
Golany, B. and J.E. Storbeck (1999). A data Envelopment Analysis of the Operational Efficiency of Bank Branches. Interfaces, 29(3), pp. 14–26. Crossref, Google Scholar
Goldsmith, R.W. (1969) Financial Structure and Development. Yale University Press, New Haven, CT. Google Scholar
Gutiérrez-Nieto, B., C. Serrano-Cinca and C. Mar Molinero (2007). Microfinance Institutions and Efficiency. Omega, 35(2), pp. 131–142. Crossref, Google Scholar
Gutiérrez-Nieto, B., C. Serrano-Cinca and C. Mar Molinero (2009). Social Efficiency in Microfinance Institutions. Journal of the Operational Research Society, 60(1), pp. 104–119. Crossref, Google Scholar
Haini, H. (2019). Internet Penetration, Human Capital and Economic Growth in the ASEAN Economies: Evidence from a Translog Production Function. Applied Economics Letters, 26(21), pp. 1774–1778. Crossref, Google Scholar
Haini, H. (2020a). The Evolution of China’s Modern Economy and its Implications on Future Growth. Post-Communist Economies, 33(7), pp. 795–819. https://doi.org/10.1080/14631377.2020.1793610. Crossref, Google Scholar
Haini, H. (2020b). Examining the Relationship between Finance, Institutions and Economic Growth: Evidence from the ASEAN Economies. Economic Change and Restructuring, 53(4), pp. 519–542. Crossref, Google Scholar
Haini, H. (2020c). Spatial Spillover Effects of Public Health and Education Expenditures on Economic Growth: Evidence from China’s Provinces. Post-Communist Economies, 32(8), pp. 1111–1128. Crossref, Google Scholar
Haini, H. (2020d). Spatial Productivity and Efficiency Spillovers in the Presence of Transient and Persistent Efficiency: Evidence from China’s Provinces. Cogent Economics & Finance, 8(1), p. 1735781. https://doi.org/10.1080/23322039.2020.1735781. Crossref, Google Scholar
Haini, H. (2020e). Examining the Productivity of the ASEAN Economies in the Presence of Transient and Persistent Efficiency. Cogent Economics & Finance, 8(1), p. 1805138. https://doi.org/10.1080/23322039.2020.1805138. Crossref, Google Scholar
Haini, H. (2020f). Tourism, Internet Penetration and Economic Growth. Journal of Policy Research in Tourism, Leisure and Events, pp. 1–7. https://doi.org/10.1080/19407963.2020.1854276. Crossref, Google Scholar
Handa, S. and B. Davis (2006). The Experience of Conditional Cash Transfers in Latin America and the Caribbean. Development Policy Review, 24(5), pp. 513–536. Crossref, Google Scholar
Haq, M., M. Skully and S. Pathan (2010). Efficiency of Microfinance Institutions: A Data Envelopment Analysis. Asia-Pacific Financial Markets, 17(1), pp. 63–97. Crossref, Google Scholar
Harris, M. and A. Raviv (1991). The Theory of Capital Structure. The Journal of Finance, 46(1), pp. 297–355. Crossref, Google Scholar
Hermes, N. and R. Lensink (2011). Microfinance: Its Impact, Outreach, and Sustainability. World Development, 39(6), pp. 875–881. Crossref, Google Scholar
Hermes, N., R. Lensink and A. Meesters (2011). Outreach and Efficiency of Microfinance Institutions. World Development, 39(6), pp. 938–948. Crossref, Google Scholar
Kar, S. and J. Deb (2017). Efficiency Determinants of Microfinance Institutions in India: Two Stage DEA Analysis. Central European Review of Economics and Management, 1(4), pp. 87–115. Crossref, Google Scholar
Kato, M.P. and J. Kratzer (2013). Empowering Women through Microfinance: Evidence from Tanzania. ACRN Journal of Entrepreneurship Perspectives, 2(1), pp. 31–59. Google Scholar
King, R.G. and R. Levine (1993). Finance, Entrepreneurship and Growth. Journal of Monetary Economics, 32(3), pp. 513–542. Crossref, Google Scholar
Laha, A. and P.K. Kuri (2014). Measuring the Impact of Microfinance on Women Empowerment: A Cross Country Analysis with Special Reference to India. International Journal of Public Administration, 37(7), pp. 397–408. Crossref, Google Scholar
Ledgerwood, J. (1999) Microfinance Handbook: An Institutional and Financial Perspective. The World Bank Publications, USA. Google Scholar
Littlefield, E., J. Morduch and S. Hashemi (2003). Is Microfinance an Effective Strategy to Reach the Millenium Development Goals? CGAP Focus Note No. 24. World Bank, Washington, DC. Google Scholar
Lønborg, J.H. and O.D. Rasmussen (2014). Can Microfinance Reach the Poorest: Evidence from a Community-Managed Microfinance Intervention. World Development, 64, pp. 460–472. Crossref, Google Scholar
Mersland, R. and R.Ø. Strøm (2010). Microfinance Mission Drift? World Development, 38(1), pp. 28–36. Crossref, Google Scholar
Microfinance Information Exchange Inc. (2018). MIX Market. Available at www.mixmarket.org/mixmarket (Accessed: January 2018). Google Scholar
Morduch, J. (2007) Smart Subsidies. In: Balkenhol B. (ed.), Microfinance and Public Policy, International Labour Organization Century Series, pp. 72–85. Palgrave Macmillan, London. Crossref, Google Scholar
Nghiem, H., T. Coelli and D.S.P. Rao (2006). The Efficiency of Microfinance in Vietnam: Evidence from NGO Schemes in the North and the Central Regions. International Journal of Environmental, Cultural, Economic and Social Sustainability, 2(5), pp. 71–78. Google Scholar
Nhung, L.T.P. and H. Okuda (2015). Effects of State Ownership on Companies’ Capital Structure and Profitability: Estimation Analysis before and After the Lehman Shock. Journal of Asian Economics 38, pp. 64–78. Crossref, Google Scholar
Paxton, J. (2007). Technical Efficiency in a Semi-Formal Financial Sector: The Case of Mexico. Oxford Bulletin of Economics and Statistics, 69(1), pp. 57–74. Crossref, Google Scholar
Piot-Lepetit, I. and J. Nzongang (2014). Financial Sustainability and Poverty Outreach within a Network of Village Banks in Cameroon: A Multi-DEA Approach. European Journal of Operational Research, 234(1), pp. 319–330. Crossref, Google Scholar
Pitt, M.M. and S.R. Khandker (1998). The Impact of Group-based Credit Programs on Poor Households in Bangladesh: Does the Gender of Participants Matter?. Journal of Political Economy, 106(5), pp. 958–996. Crossref, Google Scholar
Pitt, M., S.R. Khandker and J. Cartwright (2006). Empowering Women with Microfinance: Evidence from Bangladesh. Economic Development and Cultural Change, 54(4), pp. 791–831. Crossref, Google Scholar
Priyadarshee, A. and A.K. Ghalib (2011). The Andhra Pradesh Microfinance Crisis in India: Manifestation, Causal Analysis, and Regulatory Response. Brooks World Poverty Institute Working Paper No. 157. Google Scholar
Rai, A. and S. Ravi (2011). Do Spouses Make Claims? Empowerment and Microfinance in India. World Development, 39(6), pp. 913–921. Crossref, Google Scholar
Rawlings, L. and G. Rubio (2005). Evaluating the Impact of Conditional Cash Transfer Programs. The World Bank Research Observer, 20(1), pp. 29–55. Crossref, Google Scholar
Ruggiero, J. (2007). A Comparison of DEA and the Stochastic Frontier Model using Panel Data. International Transactions in Operational Research, 14(3), pp. 259–266. Crossref, Google Scholar
Segun, K.R.S. and M. Anjugam (2013). Measuring the Efficiency of Sub-Saharan Africa’s Microfinance Institutions and its Drivers. Annals of Public and Cooperative Economics, 84(4), pp. 399–422. Crossref, Google Scholar
Seiford, L.M. and J. Zhu (1999). An Investigation of Returns to Scale in Data Envelopment Analysis. Omega, 27(1), pp. 1–11. Crossref, Google Scholar
Servin, R., R. Lensink and M. van den Berg (2012). Ownership and Technical Efficiency of Microfinance Institutions: Empirical Evidence from Latin America. Journal of Banking and Finance, 36(7), pp. 2136–2144. Crossref, Google Scholar
Sharma, J., S.K. Parthasarathy and A. Dwivedi (2007) Examining Self-Help Groups: Empowerment, Poverty Alleviation, Education, A Qualitative Study. Nirantar, New Delhi. Google Scholar
Simar, L. and P. Wilson (1998). Sensitivity Analysis of Efficiency Scores: How to Bootstrap in Nonparametric Frontier Models. Management Science, 44(1), pp. 49–61. Crossref, Google Scholar
Simar, L. and P. Wilson (2002). Nonparametric Tests of Return to Scale. European Journal of Operational Research, 139, pp. 115–132. Crossref, Google Scholar
Simar, L. and P.W. Wilson (2007). Estimation and Inference in Two-Stage, Semi-Parametric Models of Production Processes. Journal of Econometrics, 136(1), pp. 31–64. Crossref, Google Scholar
Soteriou, A. and S.A. Zenios (1999). Operations, Quality and Profitability in the Provision of Banking Services. Management Science, 45(9), pp. 1221–1238. Crossref, Google Scholar
Sriram, M.S. (2010). Commercialisation of Microfinance in India: A Discussion of the Emperor’s Apparel. Economic and Political Weekly, 45(24), pp. 65–73. Google Scholar
Staub, R.B., G. da Silva e Souza and B.M. Tabak (2010). Evolution of Bank Efficiency in Brazil: A DEA Approach. European Journal of Operational Research, 202, pp. 204–213. Crossref, Google Scholar
Stiglitz, J. (1990). Peer Monitoring and Credit Markets. World Bank Economic Review, 4(3), pp. 351–66. Crossref, Google Scholar
Swamy, V. (2014). Financial Inclusion, Gender Dimension and Economic Impact on Poor Households. World Development, 56, pp. 1–15. Crossref, Google Scholar
Thanassoulis, E. (1999). Data Envelopment Analysis and its Use in Banking. Interfaces, 29(3), pp. 1–13. Crossref, Google Scholar
Vassiloglou, M. and D. Giokas (1990). A Study of the Relative Efficiency of Bank Branches: an Application of Data Envelopment Analysis. The Journal of the Operational Research Society, 41(7), pp. 591–597. Crossref, Google Scholar
Wichterich, C. (2012). The Other Financial Crisis: Growth and Crash of the Microfinance Sector in India. Development, 55(3), pp. 406–412. Crossref, Google Scholar
Wijesiri, M., L. Viganò and M. Meoli (2015). Efficiency of Microfinance Institutions in Sri Lanka: A Two-Stage Double Bootstrap DEA Approach. Economic Modelling, 47, pp. 74–83. Crossref, Google Scholar
Wijesiri, M., J. Yaron and M. Meoli (2017). Assessing the Financial and Outreach Efficiency of Microfinance Institutions: Do Age and Size Matter? Journal of Multinational Financial Management, 40, pp. 63–76. Crossref, Google Scholar
Yaron, J. (1994). What Makes Rural Finance Institutions Successful? The World Bank Research Observer, 9(1), pp. 49–70. Crossref, Google Scholar
Yaron, J., M. Benjamin and S. Charitonenko (1998). Promoting Efficient Rural Financial Inter-Mediation. The World Bank Research Observer, 13(2), pp. 147–170. Crossref, Google Scholar

Vol. 04, No. 02

Metrics

Downloaded 743 times

History

Received 19 December 2020

Revised 14 December 2021

Accepted 18 December 2021

30 May 2022

Information

This is an open access article published by World Scientific Publishing Company. It is distributed under the terms of the Creative Commons Attribution 4.0 (CC BY) License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Keywords

PDF download