Corporate Governance and Performance in Microfinance Institutions

(1)

Corporate Governance and Performance

in Microfinance Institutions

MFI versus MFI Plus

MSc. Thesis IE&B

C.W. de Wilde (s1469061) E-mail: char_de_wilde@hotmail.com

Supervisor: prof. dr. B.W. Lensink University of Groningen Faculty of Economics and Business

(2)

ABSTRACT

Using a global dataset on 290 MFIs from 61 countries rated by third-party rating agencies, this thesis focuses on the relationship between corporate governance and both outreach and financial performance in microfinance institutions. First, it examines what the effect of governance is on MFI performance in general. Second, it specifically examines whether the impact of the governance mechanisms on performance differs between microfinance institutions (MFI) and microfinance plus institutions (MFI Plus). A MFI specialises in financial services only, whereas a MFI Plus complements the provision of financial services with additional non-financial services. The findings indicate that the effects of governance on financial performance do not differ to a great extent between MFI and MFI Plus. For both, the beneficial effect of a female CEO and internal board auditor is confirmed. Regarding outreach however, measured by average loan size and number of clients, the governance mechanisms are of great importance for MFI Plus organizations specifically. An interesting finding is that the effect of international directors depends on the particular type of MFI.

(3)

TABLE OF CONTENTS

1. INTRODUCTION ... 5

2. LITERATURE BACKGROUND ... 8

2.1MICROFINANCE AND MICROFINANCE PLUS ... 8

2.1.1 Defining the concepts ... 8

2.1.2 Microfinance versus Microfinance Plus ... 9

2.2CORPORATE GOVERNANCE IN A MICROFINANCE SETTING ... 11

2.2.1 The dual mission of microfinance institutions ... 11

2.2.2 The ownership type of microfinance institutions ... 12

2.2.3 The limited role of the external governance mechanisms ... 13

2.2.4 Risk assessment in the changing microfinance industry ... 13

2.3PREVIOUS RESEARCH ... 14

3. THEORETICAL FRAMEWORK ... 16

3.1PERFORMANCE MEASURES ... 16

3.2GOVERNANCE MECHANISMS ... 17

3.2.1 Internal governance mechanisms ... 17

3.2.2 External governance mechanisms ... 21

3.3CONTROL VARIABLES ... 22

4. DATA ISSUES AND METHODOLOGY... 24

4.1THE DATASET ... 24 4.2THE METHODOLOGY ... 25 4.3THE ECONOMETRIC MODEL ... 26 4.4DIAGNOSTICS CHECKS ... 28 4.4.1 Multicollinearity... 28 4.4.2 Heteroskedasticity ... 29 4.4.3 Normal Distribution ... 29 4.4.4 Autocorrelation ... 30 4.4.5 Endogeneity ... 30 5. DESCRIPTIVE EVIDENCE ... 31 6. ECONOMETRIC EVIDENCE ... 35 6.1FINANCIAL PERFORMANCE... 35

6.1.1 The general model ... 35

6.1.2 The extended model ... 37

6.1.3 The control variables ... 40

6.2OUTREACH ... 41

6.2.1 The general model ... 41

(4)

(5)

1. INTRODUCTION

Microfinance refers to the provision of financial services to lower-income people, especially the poor and very poor, who traditionally lack access to banking and related services. This lack of access can create persistent poverty traps and income inequality (World Bank, 2008). Microfinance therefore has been greeted with great enthusiasm, and is by now a well recognised and powerful instrument in the fight against poverty.

As the microfinance industry starts to mature, the question arises if merely providing financial services will be sufficient to alleviate poverty. It is argued that aside from financial services, clients need additional non-financial services to be able to make productive use of the loan. Therefore, a new trend in the microfinance industry is a microfinance institution (MFI) that alongside financial services provides social services such as literacy training, health services, and business training (MFI Plus). This new type of MFI offers „microfinance plus‟: it combines microfinance with complementary non-financial services that evolve around the client and aim to develop their practical and relevant skills and knowledge. Although these „plus‟ activities might be beneficial for clients, such services are costly to provide. As a MFI is characterized by the dual mission of providing financial services to the poor (outreach) and financial sustainability of the institution (financial sustainability), providing these plus services challenges the MFI Plus and its governance structure in achieving both goals.

Microfinance is high on the public agenda nowadays and it attracts an increasing interest from academics and practitioners interested in development issues. With the maturation of the microfinance industry, new risks and challenges arise.

C-GAP (2006) considers the lack of strong performing MFIs to be a major constraint on the further development of the microfinance industry. A recent report of CSFI (2008) identifies corporate governance as the principal risk facing the microfinance industry.

(6)

Only a limited number of academically based studies on corporate governance issues in microfinance institutions are available. Except for Hartarska (2005) and Mersland and Strøm (2007, 2009), the influence of corporate governance on MFI performance has not been empirically studied before. Both Hartarska (2005) and Mersland and Strøm (2007) call for future studies that take alternative governance mechanisms into account and consider the new characteristics of the microfinance industry. Further research is deemed relevant as improved understanding of the effects of corporate governance on MFI performance is needed to guide policy making on a global level.

Due to the novelty of the phenomenon, research literature on MFI Plus organizations is still scant. The differences between MFI and MFI Plus organizations therefore remain fairly unexplored. To our knowledge, this is the first study to examine the governance structure of both the MFI and MFI Plus organization specifically. As the MFI and MFI Plus are two different types of organizations, they might require a different governance structure to be successful. Further knowledge is needed to find the optimal governance structure for both microfinance institutions regarding outreach and financial sustainability. Hence, this study should improve the understanding of the effects of corporate governance on MFI performance in this changed microfinance industry.

The importance of this thesis is two-fold.

First, it will examine what the effect of the governance mechanisms is on MFI performance in general.

We follow the theoretical framework of Mersland and Strøm (2007), who differentiate between internal and external governance mechanisms. We use recently released data from third-party rating agencies that provide a unique global dataset on 290 MFIs from 61 countries. With the use of this larger and more comprehensive global dataset, we try to respond to Morduch‟s (1999) and Hartarska‟s (2005) request for the use of better data in analysis of microfinance questions.

(7)

(8)

2. LITERATURE BACKGROUND

In this section, we will provide background information on the main concepts of (1) microfinance and microfinance plus and (2) corporate governance in a microfinance setting.

2.1 Microfinance and Microfinance Plus 2.1.1 Defining the concepts

A good starting point for defining the concept of microfinance comes from the Grameen Foundation: "Microcredit refers specifically to loans and the credit needs of

clients, while microfinance covers a broader range of financial services that create a wider range of opportunities for success. Examples of these additional financial services include savings, insurance, housing loans and remittance transfer”.

Hence, microfinance is the provision of credit and other financial services to the poor.

As the microfinance industry starts to mature, the question arises if merely providing financial services will be sufficient to alleviate poverty. It is argued that aside from financial services, clients need additional non-financial services to be able to make productive use of the loan. Therefore, a new trend in the microfinance industry is a microfinance institution that alongside credit and other financial services provides social services such as literacy training, health services and business training. This new approach to microfinance carries names such as „holistic‟, „integrated‟ or, as it is called in this study, „microfinance plus‟: it combines microfinance with complementary non-financial services that evolve around the client‟s needs and aim at developing their practical and relevant skills and knowledge.

That said, the Grameen Foundation defines microfinance plus as follows: "The local

MFI might also offer microfinance plus activities such as entrepreneurial and life skills training, and advice on topics such as health and nutrition, sanitation, improving living conditions and the importance of educating children".

(9)

2.1.2 Microfinance versus Microfinance Plus

In this study, we would like to make the difference in the approach the MFI chooses to take in delivering its services to the poor. The distinction will be made between (a) the „specialist‟ approach taken by the microfinance institution (MFI) and (b) the „plus‟ approach taken by the microfinance plus institution (MFI Plus).

(a) The specialist approach (MFI)

The specialist approach states that a MFI must concentrate on the provision of financial services only. The call for specialization is based on a win-win proposition: specialized MFIs will obtain better financial results and subsequently will improve their capability of accessing funding and servicing more customers. Therefore, the specialized MFIs will enjoy economies of scale. Important policy makers argue that the only way for microfinance institutions to become sustainable and reach scale is to concentrate on financial services only (Otero and Rhyne, 1994). Although training and technical assistance might be beneficial for micro-enterprises, such services are costly to provide and may undermine the financial sustainability of microfinance institutions. More importantly, when providing additional services management becomes more complex. Otero and Rhyne (1994) indicate that the level of professionalism and specialization necessary to become financially sustainable is unlikely to be compatible with the provision of plus services alongside microfinance. Hence, additional services and programs should not be treated as a “necessary” condition for successful microfinance delivery of the institution. Support for this argument is provided by accounts of several successful MFIs that follow the „specialist approach‟ (Biggs, Snodgrass & Srivasta, 1991).

(b) The plus approach (MFI Plus)

The plus approach states that a MFI should complement the provision of financial services with additional non-financial services. Poverty is claimed to be multidimensional, with the poor needing access to a coordinated combination of microfinance and other development services to overcome poverty (Khandker, 2005). According to advocates of the plus approach, microfinance plus providers will reach out to the poorest customers and alleviate the most poverty.

(10)

A recent study by Karlan and Valdivia (2009) confirms that plus activities in the form of business training improve the repayment and retention rates and thereby improve the institution‟s overall financial results. Moreover, a plus activity can improve the customers‟ human capital to the extent necessary to service larger loans which leads to economies of scale as well.

MFI Plus organizations have taken on a wide variety of these plus services. In general, plus services belong to one of two groups. The first group consists of services designed to strengthen the customer‟s entrepreneurial activity, for example, business or vocational training (business development services). The second group consists of services related to the general well-being of the customer and its family, such as, health awareness, literacy training and family planning (social services). Many different combinations of plus services can be provided, but common to all is that they evolve around the client‟s specific needs. These plus services may augment the complexity of the institution, yet an increasing amount of successful MFI Plus examples like the BRAC and the Grameen bank can be named.

(11)

2.2 Corporate governance in a microfinance setting

In microfinance, governance refers to the mechanisms through which donors, equity investors and other providers of funds assure themselves that their funds will be used according to the intended purposes. Such control mechanisms are necessary because the goals of managers may differ from the goals of the providers of funds. The underlying problem is known as the agency problem, and the associated costs known as the agency costs1. The goal of every governance mechanism is to minimize agency costs by aligning the objectives of managers with the objectives of the providers of funds (Hartarska, 2005). Once a microfinance institution reaches a larger number of clients, manages an increasing volume of financial resources, borrows substantial amounts from financial markets and starts to earn profit, governance becomes an important aspect.

What makes governance of MFIs different from and more challenging than that of other types of institutions are four unique issues:

1) the dual mission of microfinance; 2) the ownership type of MFIs;

3) the limited role of the external governance mechanisms; 4) the risk assessment in MFIs (Rock et al., 1998).

Before discussing specific governance mechanisms in the theoretical framework, these four special governance issues will be taken into consideration.

2.2.1 The dual mission of microfinance institutions

Governance is about achieving corporate goals. For most MFIs, dual goals exist. The fundamental goal is to contribute to development. This involves reaching more clients and poorer population strata, the main outreach “frontiers” of microfinance (Helms, 2006). The second goal is to do this in a way that achieves financial sustainability and independence of donors. Hence, MFIs are characterized by the dual mission of providing financial services to the poor (outreach) and financial self-sufficiency of the institution (financial sustainability) (Armendariz de Aghion and Morduch, 2005). Institutions might differ in the extent to which they want to maintain the dual focus. Their mission is mirrored in the corporate governance structure, as they align their internal governance mechanisms accordingly. Through its composition and the

1

(12)

priorities it sets, the board of directors directly shapes the extent to which MFIs seek to maintain the dual focus of sustainability and client coverage.

The difficult task of balancing a dual mission becomes even more complex when the microfinance activity is not the sole or main activity of the institution. Regarding MFI Plus, it is argued that although the provision of plus activities might be beneficial for clients, such services are costly to provide and may undermine the financial sustainability of the MFI. The consequence could be that outreach will be undermined as well, as the MFI can decide to serve the more profitable clients in the market, the upper poor, to compensate for financial losses. This is called „mission drift‟. Therefore, providing plus services challenges the MFI to keep pursuing both goals and forces them to step up their governance composition and structure.

2.2.2 The ownership type of microfinance institutions

The assumption made in corporate governance literature is that the governance framework is implicitly defined by the corporate structure it governs. Of particular importance is whether the microfinance institution itself operates within a shareholder framework (SHF) or within a non-profit framework (NPO). For both the SHF (for-profit) corporate structure and NPO (non-(for-profit) corporate structure, factors exist that may strengthen or weaken its governance structure.

NPOs are often considered to be weaker structures since they lack owners with a financial stake in the operations (Jansson and Westley, 2004). Their board members tend to include donors and clients, who are likely to have limited financial knowledge and experience. Both facts lead to a lower financial performance than that of SHFs. On the other hand, the strength of a NPO is that they are more capable of tapping into local information networks.

The strength of the SHF is that it has a focus on profit maximization which enhances financial performance. However, its weakness is that they lack a representation of microfinance stakeholders at the board table.

(13)

2.2.3 The limited role of the external governance mechanisms

The manager of a corporation is disciplined through the labour market for managers and through the market for takeovers. As the market for MFI managers is thin and most MFIs do not have true owners, the role of these external market forces in microfinance is limited. As a result, the board of directors is of particular importance in disciplining the manager. Furthermore, other external market forces have started to play an important role in promoting manager accountability, including competition and regulation (Hartarska, 2005).

The upcoming presence of for-profit firms and the increased competition for donations and customers in the microfinance industry both motivate the MFI to find state-of-the-art governance structures. The competition for donations in particular has contributed to the emergence of „rating agencies‟. Hence, the increased competitive pressure leads to the fact that MFIs and their managers are gradually becoming more transparent. Competition therefore serves as an alternative external governance mechanism these days. Another external governance mechanism playing a part is regulation. Some MFIs that provide deposit services are increasingly subject to regulations or are under the supervision of a government agency. Although the positive consequence is that regulation increases transparency, it also has the negative consequence of promoting more conservative behaviour of the manager. This could shift the manager‟s attention from optimizing both outreach and financial performance.

2.2.4 Risk assessment in the changing microfinance industry

Providing financial services in general involves a set of risks that the board of directors must be able to assess. The fact that many MFIs operate in developing countries adds an additional layer of risk. The nature of microfinance has always made risk assessment important, but there are several reasons for the fact that boards are required to possess an even greater ability to assess risk than previously.

(14)

2.3 Previous Research

Good corporate governance has been identified as a key bottleneck in strengthening MFIs‟ financial performance and increasing its outreach (Rock et al., 1998; Labie, 2001; United Nations, 2006). Although good governance has been recognized as critical for the success of MFIs, only a limited number of academically based studies on governance issues are available.

There are several reasons for the lack of studies on the effects of governance on MFI performance. First, performance data are considered proprietary and are therefore hard to obtain. Secondly, the organizations that provide microfinance services (NGOs, banks, cooperatives, and non-bank financial institutions) are very diverse, which makes it difficult to choose an appropriate conceptual framework. Thirdly, there are few market mechanisms that promote transparency, which make governance practices hard to observe. Finally, the unique characteristics of MFIs also complicate the analysis of governance issues.

Hartarska (2005) investigates the influence of corporate governance on financial performance and outreach, using unrated Eastern European data. Governance mechanisms include board characteristics, CEO compensation and ownership type. She includes several institutional and firm control variables and finds that a more independent board obtains better financial performance. Moreover, the results indicate that board diversity improves both outreach and financial performance, whereas larger and less independent boards weaken financial performance. The difference in financial performance and outreach between various ownership types is negligible.

(15)

(16)

3. THEORETICAL FRAMEWORK

Corporate governance, understood as the system or the set of mechanisms by which organizations are directed and controlled (OECD, 2004), influences organizations‟ performance. In this section, we will spell out hypotheses concerning the relationship between governance and MFI performance. The predictions are based upon the corporate governance literature of banks as well as of non-profit organizations, applied to the microfinance field.

3.1 Performance measures

Governance is about achieving corporate goals. For most MFIs, dual goals exist. One goal is the social mission of reaching poor clients (outreach) and the second goal is financial sustainability. Performance will be measured along both the outreach and financial sustainability dimension.

As a measure for financial performance we use the return on assets (ROA)2. This is the ratio of financial income to average total assets, and states the return on average total assets at the end of a given period. The outreach measures are the MFI‟s average outstanding loan (average loan) and the number of credit clients served (credit

clients). The average outstanding loan is a measure of the so called "depth" of

outreach, that is, how poor the clients are that are being served. It is the ratio between average loan size and GDP per capita. The number of credit clients is a measure of the "breadth" of outreach, that is, how many clients are being served. It is measured as the total number of clients that are active within the MFI at the end of the year.

These performance measures should cover a number of interesting features of the microfinance reality.

Table 1 gives an overview of the dependent variable definitions.

Table 1 Definitions of dependent variables

Variable Explanation

Financial performance

ROA Return on assets

Outreach

Average loan Clients average outstanding loan Credit clients The number of credit clients

2

(17)

3.2 Governance mechanisms

We follow the theoretical framework of Mersland and Strøm (2007), who differentiate between the internal and external governance mechanisms. The internal mechanisms are board/CEO characteristics and ownership type. The external mechanisms are product market competition and regulation. Both types of governance mechanisms are used in the analysis to identify their relationship with outreach and financial performance.

Table 2 summarizes the independent variables, their definitions and hypothesized signs to the dependent variables in table 1.

Table 2 Definitions of independent variables and their hypothesized sign to financial (FinP) and outreach performance

Hypothesis

Variable Explanation FinP Outreach

Board size The number of directors -International directors The number of international directors + -CEO/chairman duality CEO and chairman are the same person -Female CEO A dummy being 1 if the CEO is female + + Internal board auditor A dummy being 1 if the internal auditor

reports directly to the board + -/+

SHF Shareholder firm + -/+

NPO Non-profit organization -/+ +

Competition A subjective scale from 1 to 7 with

higher values indicating stronger competition - + Regulation A dummy being 1 if the MFI is regulated

by banking authorities -/+ -/+

Loan methodology A dummy being 1 if loans are mainly made to individuals

Rural/urban market A dummy being 1 if main market is urban MFI experience Years of experience as a MFI

Labour productivity The total number of loan clients divided by the total number of employees

Firm size The natural logarithm of assets

Human Development Index A composite index covering life expectancy, education, and income (GDP per capita)

3.2.1 Internal governance mechanisms

(18)

The board and CEO characteristics include board size, international directors, CEO/chairman duality, female CEO and internal board auditor. In one way or another, these variables turn out to be significantly related to MFI financial performance or outreach and most are suggested in former studies. Ownership type includes being a shareholder firm (SHF) or non-profit organization (NPO).

The ownership-board relationship concerns (a) how well the board is aligned to owner interests (b) how well the board is informed and (c) how decisive the board is (BØhren

and StrØm, 2005). The higher the score on these three dimensions, the better the

financial performance. However, it might have a different influence on outreach.

(a) Board alignment to owner interests

In MFIs, the board is supposed to be better aligned to owner interests if the CEO and chairman are different persons, and if the number of international directors increases. In short, the role of the board of directors is to provide guidance to management in the overall strategic direction of the organization, leaving the CEO to actually manage the organization. The chairman serves as an intermediary between the CEO and the board of directors, and he or she should maintain a degree of detachment that allows him and the board to make independent, responsible decisions regarding the organization and particularly regarding management performance.

There are many compelling reasons for splitting the positions of CEO and chairman of the board3. First, it avoids concentration of power in one person. Secondly, separation ensures the voicing of two opinions and underscores the fact that the CEO reports to the board, which increases board independence. Independent boards are considered to be better able to monitor the CEO on behalf of the owners, and moreover ensure a better alignment of interests. The opposite situation is called CEO/chairman duality, which may be a sign of CEO entrenchment (Hermalin and Weisbach, 1991, 1998). On the one hand, duality might have a negative impact on board independence as the CEO may pursue policies that give him private benefits. On the other hand, the positive effect of duality is that it may enhance decision-making effectiveness.

3

(19)

This ambiguity may explain why Brickley et al. (1997) do not find that firms with a separate CEO and chairman outperform those with a CEO/chairman duality.

Regarding international directors, Oxelheim and RandØy (2003) find that firm

performance improves with the presence of international directors as they are considered to be an indication of independence. On the other hand, international directors may be more focused on financial performance. They might wish to increase the average loan size and thereby shift the focus to the more profitable client, which consequently has a negative effect on outreach.

(b) Informing the board

We expect that the better the CEO and the board are informed, the better financial performance and outreach will be. Regarding this aspect, we will include gender and the internal board auditor.

Women‟s leadership has been central to microfinance from the inception of the industry. One of the innovations in microfinance has been the targeting of predominantly female customers (Armendariz de Aghion and Morduch, 2005). A female CEO is presumed to be better at obtaining information from female customers about which products women want and against what terms. This improved knowledge is expected to positively influence both financial performance and outreach. Moreover, female CEOs make powerful role models. The fact that a female CEO inspires other women to work harder concerns both their female clients as well as the other female board members. Finally, Allen and Gale (2000) mention that there is evidence that women directors spend more time on monitoring activities which increases performance.

(20)

assessments on the appropriateness of the organization‟s internal governance structure and the MFI operations. This should improve financial performance and outreach.

Other variables that could lead to a better informed CEO and board are the CEO‟s experience and educational background, as well as the representation of various stakeholders. A more experienced and educated CEO is likely to bring better and more relevant information to the board‟s attention. However, there is a lack of good directors with the right education and expertise in the microfinance industry. Regarding stakeholders, it is not unusual that several major stakeholders are represented in the board in order to increase its ability to tap into local information networks. The consequence might be that MFI performance will be affected by the relative power of these various stakeholders. However, as stakeholder representatives are almost absent in this dataset, we unfortunately cannot include them in our regressions in a meaningful way. Mersland and Strøm (2007) noticed this surprising lack as well, and they recommend future studies to explore the existence and the role of stakeholders in microfinance‟ governance.

(c) Board size/decisiveness

(21)

Ownership type may play a role in MFI performance. Each ownership type has its structural weakness. NPOs are often considered to be weaker structures since they lack owners with a financial stake in the operations, which leads to a lower financial performance than that of SHFs. Accordingly, some practitioners argue for the transformation of NPOs into SHFs. On the other hand, NPOs are believed to be more effective in reaching poor customers. This is confirmed by Gutierrez-Nieto et al. (2007), using data from 20 Latin American MFIs. The structural weakness of SHFs is that they focus on profit maximization but lack a representation of microfinance stakeholders at the board table. Hence, SHFs are expected to have a better financial performance but they will reach less poor clients than NPOs. Thus in theory, NPOs are expected to excel in outreach, whereas SHFs are expected to excel in financial performance.

Nevertheless, Mersland and Strøm (2007) find that SHFs and NPOs perform equally well. They state that the dichotomy along ownership type as described above may not be the best description. The incentive problems between owners and managers may be more pronounced in NPOs, but NPOs have the compensatory benefit of reducing customer adverse selection and moral hazard (Hansmann, 1996) since they are better able to tap into local information networks. Evidence in Cull et al. (2007) confirms this. On the other hand, many SHFs are not managed along the shareholder value model, as they may be strongly committed to reaching the poor. Recent comparisons of performance in different ownership types as well as historical evidence suggest that performance does not vary systematically between ownership types. Therefore, we expect to confirm that SHFs and NPOs perform equally well.

3.2.2 External governance mechanisms

(22)

Based on the information regarding competition provided in the rating reports, the raters have created a subjective measure of competition on a 1 to 7 point scale. Since the raters have multi-country experience and rated dozens of MFIs, they should be able to provide well-judged information. Furthermore, since many MFIs only have local or regional coverage, proxies for national level of competition will be less reliable than this proxy used. Having stated that, our proxy may not be reliable in individual cases, but for the time being we consider this measure to be the best one available as it should serve as a rough guide to the relative competitive pressure in microfinance markets.

Most MFIs are not regulated. A regulated MFI is more likely to earn customer trust, which should lead to better financial performance. On the other hand, regulation is associated with costs like security requirements, investments in information technology, and the stifling of MFI innovations. Higher agency costs may pull financial performance into the opposite direction. Hence, the final outcome of the effect of regulation on financial performance is uncertain. The findings on outreach effects are contradictory. When regulated, the MFI gains access to low-cost depositor funding through the right to mobilize savings. This gives the MFI the opportunity to increase the number of clients, but also the opportunity to increase the average loan size of existing customers. Therefore, the effects upon depth and breadth of outreach are uncertain as well.

3.3 Control variables

Finally, we include several control variables that are specific for the MFIs. The inclusion of these variables will also help to inform the ongoing debate in microfinance literature on matters such as the advantages of group lending. We will comment on the results for these aspects when they yield interesting insights, although the focus is on issues concerning the internal and external governance mechanisms.

(23)

local knowledge and trustworthiness. A dummy variable will indicate whether the main loan methodology of the MFI is to individual borrowers or group lending4.

Secondly, MFIs often target their lending at the rural population to a greater extent than ordinary banks. If MFIs have to serve the poor in remote rural areas, it may be difficult for them to achieve financial sustainability. Having a mainly rural or mainly urban market therefore may influence financial performance.

Furthermore, the average labour productivity, the MFI experience and the firm size are included as MFI-specific controls. And finally, the Human Development Index (Human Development Report, 2006) will control for country-specific effects. The index is a composite of a country‟s average results in three areas, being (i) life expectancy, (ii) education and (iii) income (GDP per capita).

4_{Group lending encompasses village banks and solidarity groups. A village bank normally consists of}

(24)

4. DATA ISSUES AND METHODOLOGY

4.1 The Dataset

In this paper we utilise an extended version of a self-constructed global data set on MFIs, provided by Mersland and Strøm. The data are collected from risk assessment reports of third-party specialized rating agencies. A major advantage of the assessment reports is that they are worked out by a third party and cover a wide range of organizational features alongside financial data and social and financial indicators. We use observations of 290 rated MFIs from 61 countries. At each rating four years of data are obtained, at best. The ratings are performed for the period 2001 to 2007, which means that we have data running from 1998 to 2007. Most of the data stem from the period 2001 to 2005.

The dataset contains information obtained from risk assessment reports of five microlender rating agencies: MicroRate, Microfinanza, Planet Rating, Crisil, and M-Cril, and their reports can be found at www.ratingfund.org. All five are approved as official rating agencies by the Ratingfund of the C-GAP. It is assumed that each rating agency has made the necessary corrections to its financial reports to enable a reasonable comparison of MFIs. This assumption is in line with the benchmarking objective outlined by The Rating Fund. Hence, the rating methodologies reveal no major difference in MFI assessment relevant to the variables used in this study.

The use of rating data may introduce sample selection bias, since only rated MFIs enter. Moreover, this dataset contains relatively fewer of the mega sized MFIs and does not cover the virtually endless numbers of small savings and credit cooperatives. The former are rated by such agencies as Moody‟s and Standard and Poor‟s, while the latter are not rated at all. The 290 MFIs in the dataset represent commercial and professionally oriented institutions that have decided to be rated to improve access to funding, benchmark themselves against others, and increase transparency. Some MFIs in the dataset chose to be specialists while others decided to deliver plus services alongside microfinance, which allows for realistic comparisons.

(25)

finance (Solnik and McLeavey, 2004), conversion into USD implies that local inflation has been taken into account. When necessary, all entries in the dataset have been annualised and dollarized using official exchange rates.

No dataset is perfectly representative of the microfinance field. Despite the limitations, we consider our third-party collected data to be more reliable than self-reported data sources like MIX market or questionnaires. Overall, the data seem sufficiently representative for both MFIs and MFIs Plus. Specifically, a large firm bias is avoided. 4.2 The Methodology

The dataset has a panel data structure, such that we have repeated observations on the dependent performance variables for up to four consecutive years, while the independent governance variables are reported only once. As the governance variables represent fixed firm characteristics, we assume them to be constant during the whole period. We will estimate the relationship by using the method of multiple linear regression.

Multiple linear regression enables us to determine the simultaneous effect of several independent variables on the dependent variables by using the least squares principle. By using regressions, we can develop an equation that shows how the independent governance variables are related to the dependent performance variable. In the econometric model, explanatory variables to test the hypotheses are included, as well as a set of variables to control for MFI-specific and country-specific characteristics. The control variables are included as these variables otherwise might influence our results when being omitted. In this study, we use a special case of multiple linear regression, by including interaction variables with a dummy variable Di.

(26)

The functional form is:

Yi =

α

0 +

α

1Xi+ β0 Di+ β1(Xi * Di) +

ε

i

The marginal effect of x on Ey is: δ Ey

α

1 if Di = 0

=

δ x1

α

1 + β1 if Di = 1

The interaction of all variables (here constant and one independent variable x) with the dummy variable allows estimating separate linear relationships simultaneously for the two groups defined by Di. This differs from running two separate regression

models, as the error terms now are assumed to have identical variance across groups. Moreover, simultaneous estimation increases the sample size in the model, which subsequently enhances the possibility of finding significant results.

4.3 The Econometric Model

In this study two models will be used. (1) The general model will estimate the effects of governance on performance for MFIs in general. (2) The extended model includes the interaction variables and thereby will estimate the effects of governance on performance for MFI Plus organizations specifically.

The general model is specified as follows:

i

(2)

The first part of the equation is in accordance with the above explanation of the variables. In the extended model the dummy variable Di is introduced. Diisadummy

variable indicating 1 if it concerns a MFI Plus organization and 0 if it concerns a MFI

specialist organization. With the additional explanatory variables Di, Di *BS,

Di*INTDIR, Di*CEOCHAIR, Di*FEMCEO, Di*INTAUD, Di*SHF, Di*NPO,

Di*COMPET and Di*REGUL, the marginal effect of governance on performance for

(28)

In the extended model the

α

‟s are the marginal effects of the governance variables in general and the

α

+

β

‟s are the marginal effects when being MFI Plus specifically. For

3) CEOCHAIRi +

7) NPOi +

i (3)

Ordinary least squares (OLS) regression is used for fitting both the statistical models in this paper. The underlying assumptions will be tested in the following subsection. 4.4 Diagnostics checks

In order to test the validity of our model, i.e. the underlying assumptions of the OLS regression methodology, we perform several diagnostics checks here.

4.4.1 Multicollinearity

The standard assumption of multicollinearity is to exclude cases of perfect correlation between independent variables. In the presence of multicollinearity, the estimated coefficients might not be statistically significant as a result, even when the relationship might be quite strong. We check for multicollinearity using a pairwise correlation matrix. In this matrix estimates of the linear relationships between the independent variables are given. The results are presented in table 8 (general model) and table 9 (extended model), which can be found in the Appendix. The scores range between 0 and │1│, the threshold value for multicollinearity applied here is │0,8│. Overall, the independent variables show no signs of multicollinearity. However, we do notice that in the extended model the governance variables tend to correlate with their interaction variable, i.e. BS and Di*BS. As these correlations stay below the

threshold value of multicollinearity, this is not a problematic finding. Among the interaction variables themselves there however are multiple pairs that report a correlation larger than │0,8│, i.e. Di*BS and Di*NPO. When we introduced the

(29)

aware of the fact that this could bring multicollinearity into the model. To check if the inclusion of the interaction variables would be problematic, we calculated the variance inflation factors (VIFs) for the independent variables after each regression. The VIF value reflects the degree to which other coefficients‟ variances (and standard errors) are increased due to the inclusion of that particular variable. As a rule of thumb, a variable whose VIF value is greater than 10 may merit further investigation. As the VIF results repeatedly were below the threshold value of 10, we conclude that this collinearity turns out not to be problematic.

4.4.2 Heteroskedasticity

Heteroskedasticity occurs when the variance of the error terms are not the same. This would be a violation of one of the assumptions made for least squares and thus we should check here for heteroskedasticity. In order to get a precise answer to the heteroskedasticity question White tests are used. The hypotheses posed of this are

2 2

0 :i 

H H₁:_i2 2_j

When the p-value found for the null hypothesis is within the 5% rejection region i.e. < 0,05, H0 is rejected and we have to conclude that the data are heteroskedastic.

The p-value found for the models are 0.0406 and 0.0471 respectively. As this is inside the rejection area of a 5% significance test, we have to conclude that the data are heteroskedastic. The remedy for heteroskedasticity is simple. White developed an estimator for standard errors that is robust to the presence of heteroskedasticity. It should be noted that this solution applies only to large samples, which is indeed the case in this study. By using the Eicker-Huber-White robust estimator, we computed the robust standard errors in order to get consistent (i.e. asymptotically unbiased) results. Hence, we took the possibility of heteroskedasticity into consideration by presenting standard errors based on a robust estimation, which instantly corrects for heteroskedasticity.

4.4.3 Normal Distribution

(30)

hypothesis of normality. As a result, the error terms do not appear to follow a normal distribution. Hence, our estimations will be slightly biased due to this nonnormality. 4.4.4 Autocorrelation

We speak of autocorrelation when the error terms are serially correlated. Our dataset has a panel data structure. Unfortunately, the Durbin-Watson test of autocorrelation cannot be performed on panel data. We do however have to keep in mind the effects when having a panel. As we notice that the corporate governance variables of our interest barely change over time, we decide not to use a fixed effect estimator5. Therefore, to keep in mind the effects of a panel and moreover to be sure of the fact that no autocorrelation will exist in our regressions, we will cluster each regression at the MFI level as a precaution. Thereby we control for the possibility of error terms correlating at the MFI level.

4.4.5 Endogeneity

It is possible that some of the regressions presented in the following section are subject to endogeneity problems. In terms of our model, endogeneity could arise due to the existence of unobserved variables that could influence both the corporate governance variables as well as the performance variables. As our specified model however is based upon sound theoretical assumptions, we have no compelling reasons to suspect such possible endogeneity. A way to test for endogeneity is by using the Hausman test. This requires the use of a good instrumental variable. A good instrumental variable would be a variable which is highly correlated with governance but not with the error terms of the regressions. However, no instrument known to be useful is available. Therefore, we do acknowledge the possibility of endogeneity in our model, but as there is no cure at hand we do not take further actions concerning the endogeneity assumption.

5_{The use of a random effects estimator would be possible. However, as we do not apply a panel data}

(31)

5. DESCRIPTIVE EVIDENCE

Before proceeding, we need to define the variables used in the analysis. This also gives rise to the opportunity to supply some stylized facts about the MFIs in the data sample. We start by giving some descriptives on the dependent variables. Table 3 summarizes the dependent variables.

Table 3

Better financial performance is associated with a higher ROA. Better outreach is associated with a lower average loan size and a higher number of credit clients. Higher values for average loan point to lending to wealthier people and hence a deterioration in depth of outreach. Higher values for credit clients imply reaching a greater number of clients and hence an increase in breadth of outreach. A brief comment on the average loan size is in order. The lowest loan amount is USD 1.00, while the average loan amount is USD 782.69. The maximum amount of USD 24.589 is an extreme case, but is kept within the dataset since robustness checks indicate that filtering it out does not significantly influence overall findings. The same argument goes for the extreme case regarding the MFI serving 394.374 credit clients.

In this dataset, 82.46% of the organizations are a MFI and 17.54% are a MFI Plus organization. This indicates that nearly one out of every five MFIs in this dataset refuses to follow policy recommendations to specialise in financial services only, and thereby now provide non-financial services alongside microfinance.

Table 3b provides an explorative analysis of the relationship between plus providers and specialists on the one hand, and between financial performance and outreach on the other.

Table 3b

(32)

In terms of breadth of outreach, the mean value of credit clients is slightly higher for plus providers. As the t-test however shows that the results do not differ significantly, we conclude that specialists and plus providers have an equal focus on breadth of outreach.

Table 4 shows the main values on board and management characteristics, regarding MFI and MFI Plus respectively. We use the opportunity to split the data for MFI specialists and MFI plus providers throughout the rest of this section, as this might provide useful knowledge for further explanations along this study.

Table 4

The table shows that the number of observations for most independent variables is far less than for the dependent variables in table 3. The reason for this is that the governance variables represent fixed firm characteristics that are assumed constant over time. They are reported only once in this dataset, but later on in the analysis we will fill up the missing cells as to increase the number of observations used in the regression analyses. The mean of many dummy variables can be interpreted as the percentage of organizations in the category. For simplicity, in the rightmost „% no' column we report the percentage of the variable that scores zero.

(33)

As the female variables turn out to be highly correlated, we prefer to group gender under the female CEO variable. The reason for preferring the female CEO is that she plays a key role in the long-term success of the institution and in the realization of effective governance. Moreover, the female CEO variable has the most observations. And last, obviously, the CEO is not a novice in business. In MFIs and MFI Plus only 7.6% and 10% of the CEOs have no former business experience, where only 15.3% and 29.2% have no business education. However, as these results come from very few observations, we are forced to drop these variables in further analyses. By performing a t-test, only board size and international board directors prove to significantly differ between MFI and MFI Plus on a 5% significance-level. The results of the t-test can be found in table 4b in the Appendix.

Table 5 shows the main values on ownership type, the external governance mechanisms and the control variables.

Table 5

The MFIs in this dataset comprise a number of ownership types, that is, a bank, a non- bank financial institution, a non-governmental organization (NGO), a cooperative/ credit union, a state bank and otherwise. For MFI, at 53,4% the NGO is by far the largest group, followed by the shareholder owned firms being the bank and the non-bank financial institution at 34,7%, and the cooperative union at 11.9%. MFI Plus is even more strongly dominated by NGOs, as they comprise 86% of total MFI Plus organizations, compared to only 4% being a shareholder owned firm and 10% being a cooperative union. In this study, the partition will be made between being a shareholder firm (SHF) and a non-profit organization (NPO). The SHF dummy is defined to be the bank and non-bank financial institution, the NPO dummy comprises the NGOs. As the remaining ownership types do not have a clear ownership structure, and as the great majority of firms is either SHF or NPO by these definitions, this partition is a natural choice. In the regressions we will moreover add an ownership type dummy, indicating a SHF if 1 and NPO if 0.

(34)

already somewhat regulated due to the fact that their accounts are transparent. However, most MFIs are not officially regulated, as the table shows that only 34% of MFIs are regulated by banking authorities, compared to merely 12% of MFI Plus.

Apart from the internal and external governance variables specified above, we include several control variables which are in accordance with previous literature. Regarding loan methodology, the results show that MFIs mainly make loans to individuals, whereas MFI Plus rather uses group loans. This result is in line with the expectation that MFIs tend to focus on financial performance and MFI Plus tends to focus on outreach. Regarding the categories mainly rural or urban market served, both the MFI and MFI Plus mainly focus on the urban market. This reflects the MFI's trouble in reaching the rural market. Furthermore we notice that the typical MFI is a rather young organization. Surprisingly, MFI Plus organizations seem to be slightly older on average and greater in size. The firm size variable moreover shows a large dispersion of the MFI's size. A brief comment on the Human Development Index (HDI) is in order as well. The minimum and maximum values show that the organizations come from a wide variety of country background. As Gorton and Winton (2003) argue that institutions, regulations, and laws are important in the study of financial

(35)

6. ECONOMETRIC EVIDENCE

In this section we report the results from econometric tests on the relationships between the financial and outreach performance on the one hand, and the internal and external governance mechanisms on the other. First we will present the results for financial performance, followed by the results for outreach. We will comment on the results for the general model, and subsequently on the results found for the extended model. Last we will comment on the control variables. The regression analyses will be performed by using the statistical programme STATA 10.1.

6.1 Financial performance

Table 6a and 6b show the results from the OLS regressions on financial performance, with ROA as the dependent variable. We have performed four kinds of regressions for ROA. The first ROA column presents the effects of the governance variables on financial performance for the respective models. In the second ROA column we drop the international directors and internal board auditor variable, in order to increase the number of observations. By increasing the number of observations vis-à-vis the decreasing number of parameters, we increase the degrees of freedom. Increasing the degrees of freedom is generally seen as desirable as it strengthens causal inference. Hence, this regression can strengthen the process of reaching conclusions about the effects of governance on performance. In the third ROA column we perform a test regarding ownership type. We drop ownership types other than SHF and NPO, and replace these with the dummy variable ownership type, being 1 if the firm is a SHF and 0 if it is a NPO. In the fourth ROA column we perform a robustness check by only including the variables that proved to be significantly related to financial performance. We will comment on the internal and external governance mechanisms per and across regressions.

6.1.1 The general model

The results of the general model, which estimates the effects of governance on performance for microfinance providers in general, are presented in table 6a.

Table 6a

(36)

This implies that in general board size has no effect on financial performance. This result is in line with Mersland and StrØm (2007), who report an insignificant negative

sign as well. It however contradicts Hartarska (2005), who reports a negative but significant sign, which implies that a large board lowers financial performance. International directors carries a significant negative sign, which indicates that the presence of international board directors tends to worsen financial performance. This finding is contrary to our expectations. An explanation may be the fact that international directors may bring along a culture of higher costs, which negatively influences financial performance. CEO/chairman duality has a negative though insignificant effect on ROA. Therefore, we cannot state whether the MFI is better governed when the CEO and chairman are different persons, nor can we reject or confirm the result from Brickley et al. (1997), who find a significant positive effect for CEO/chairman duality on financial performance. Having a female CEO has a strong and significant positive effect on ROA. This confirms the importance of gender for microfinance institutions, where female customers and leaders are often considered to be of special importance. Having a female CEO guiding and monitoring the institution will positively affect financial performance. The internal board auditor has a significant positive sign as well, which emphasizes the positive influence of directly reporting to the board on financial performance.

Ownership type (SHF or NPO) shows positive results for both SHF and NPO. However, the difference lies within the significance of the effect. Since the SHF coefficient is not significant at the 10% level in any of the regressions, it turns out that being a SHF does not improve financial performance as expected. Hence, it is not proven that a SHF is better at obtaining good financial performance as compared to other ownership types. Due to this finding, we disregard suggestions in literature that NPOs should be turned into a SHF as to strengthen financial performance.

(37)

effect on ROA. The proposition of NPOs performing financially better than SHFs is thereby confirmed. Our finding is contrary to the finding of Mersland and StrØm

(2007), who find that ownership type does not matter with respect to financial performance. However, it is in line with the finding in their World Development paper of 2009, where the results do report that NPOs tend to outperform SHFs financially. A possible explanation for our finding might be that in this dataset NPOs in general perform better than SHFs, plus the fact that NPOs are relatively overrepresented as compared to SHFs.

The external mechanisms of competition and regulation show significant negative results for competition. This implies that increased competition tends to worsen financial performance. Thus, even though the competition measure is subjective, it captures the effect that increased competition leads to a lower ROA. For the regulation impact we find no significant results, which is in line with Hartarska (2005) and Mersland and StrØm (2007). The latter opt a possible explanation for the

regulation coefficient not turning out to be of significant influence on financial performance. Maybe the fact that the MFIs in the sample are rated implies a homogeneity among the firms with respect to regulation. Maybe transparency is sufficient regulation by itself.

To check the robustness of these general model results, we perform a separate regression which only includes the significantly related variables. The results are presented in the fourth ROA column and largely conform to our earlier findings. Both the female CEO and internal board auditor clearly come out to be of strong and positive influence on financial performance. Being a NPO will strengthen financial performance as well, although its positive effect is slightly less strong than in previous regressions. International directors and increased competitive pressure have a negative effect on financial performance, although slightly less negative than reported previously.

6.1.2 The extended model

The results of the extended model, which estimates the effects of governance on performance for microfinance plus providers specifically, are presented in table 6b.

(38)

The results show that most governance effects found for MFI Plus (

α

s+

β

s) are in the same direction as the governance effects found for MFIs in general (

α