The Influence of International Board Members and International CEO’s on Performance of MFIs

(1)

The Influence of International Board

Members and International CEO’s on

Performance of MFIs

Abstract:

This paper contributes to the scarce amount of literature on international influences on microfinance by examining the influence of international board members and international CEO’s (Chief Executive Officers) on MFIs (Microfinance Institutions). Agency theory is used to indicate negative effects and resource-dependency theory is used to indicate positive effects of these international influences on MFI performance. The data used contains information from 379 MFIs from 74 countries over the period 2000 to 2006. Results reveal that international board members and international CEO’s positively, yet insignificantly, influence MFIs performance. This paper provides a basis for future research because results suggest complementary effects between international board members and international CEO’s and performance and because a causal link between these international effects and performance remains unclear.

Key words: Microfinance, International, Governance

Master Thesis

(2)

1. Introduction

Microfinance institutions (MFIs) provide financial services to the very poor that do not have access to regular financing methods. Over the past decades, especially after Muhammed Yunus received a Nobel Prize for microfinance, the microfinance sector has developed and grown rapidly.

As the microfinance sector is developing, international influences are increasing simultaneously. International influences refer to the influence of international actors through capital provision and accompanying responsibilities and through knowledge and best-practice transfers to MFIs (Mersland et al (2011)). The reason for international interest is two-fold. First, multilateral organizations such as the World Bank support microfinance as a poverty alleviation method. Furthermore, investments in microfinance to reduce poverty falls under ‘responsible investments’ which is high on the agenda today and enables firms and wealthy individuals to gain reputation. Second, the microfinance sector is growing rapidly thus it is also a profitable investment opportunity. Although the MFIs are often located in less-developed countries, funding comes primarily from developed countries. With the increase in international interests and funding MFIs are increasingly subject to international influences that are expected to influence the microfinance sector. However, to this day, the scope and scale of international influences on MFI performance is unknown.

At the same time, the growth and development of the microfinance industry has made good governance in MFIs imminent (Rock et al (1998), Labie (2001)). Good governance is important, firstly, to control the high growth rates and changes within the microfinance sector, and secondly, because MFIs face a unique double-bottom line in their mission: outreach to the poor and financial sustainability of the MFI. Good governance is needed to ensure that funds received will be used for the initial performance goal of reaching out to the poor.

(4)

This paper addresses international influences more closely by investigating the influence of international board members and international CEO’s on performance of MFIs.

To do so, this paper makes use of the agency theory and the resource-dependency theory. From the perspective of the agency theory, agency costs are expected to increase with the appointment of international board members and an international CEO, thus imposing a negative effect on MFI performance. Resource-dependency states that the MFI will gain access to capital and knowledge from international board members and CEO’s, thus causing a positive effect on MFI performance.

The significance of this paper is that it contributes to the scarce microfinance literature, both in terms of governance and international influences, on the performance of MFIs. The data used incorporates 379 MFIs from 74 countries over the period of 2000 to 2006. The data consists of information obtained from risk assessments and reports from five rating agencies. This study responds to the request by Mersland and Strøm (2009) that more research should be done on international influences on MFIs performance.

The remaining part of this paper is divided into five parts. Section 2 consists of a background of microfinance in which the need for governance and the literature on MFI performance is discussed. Section 3 explains the two main theories underlying international influences. Section 4 uses literature and theory to deduce the hypothesis. Section 5 reviews the data, the model and methodology. Section 6 provides an overview of the results. Finally section 7 provides a discussion and a conclusion.

2. Background

2.1 Microfinance and Governance

The term microfinance refers to financial services, such as credit and saving opportunities, provided to the poor. MFIs ensure the provision of these financial services. Initially MFIs were non-profit organizations that received their funding primarily through donations from charities or through government subsidies.

(5)

funds remains limited, resulting in less outstanding loans. Most importantly, as Elisabeth Rhyne (1998) predicted; “subsidizing MFIs cannot occur indefinitely and this type of funding will not be sufficient in the long-run.”

MFIs are therefore experiencing a shift in the source of funding away from grants and towards finances from the mainstream capital market, relying on amongst others investors, large mutual funds and private investors. This shift allows more funds to become available to the MFIs but it also implies that investors require positive returns on their investments. In accordance with these developments MFIs must therefore take responsibilities to ensure positive returns on the funds received, which they do by monitoring and ensuring both transparency and efficiency of operations. As a result, the new demands accompanying these investments have made MFIs self-sustainable, no longer relying on grants to cover costs.

The developments have resulted in a double-bottom line in the mission of MFIs; reaching out to the poor (outreach) and becoming and remaining financially self-sustainable (sustainability). Due to this double-bottom line, MFIs can be distinguished from other firms because they resemble both banks and non-profit organizations. Due to the double-bottom line in the mission, there is an increase in the risk of diverging objectives of different stakeholders; some focusing on outreach while others focus on sustainability (Brau and Woller (2004)). If the focus is primarily on sustainability, or providing positive returns for investors, then the main goal of microfinance - outreach to the poor - may be neglected. A heated discussion about this trade-off is on-going and referred to as mission-drifting (Mersland and Strøm (2010), Morduch (2000)).

As the microfinance sector is grows the need for good governance is becoming more imminent (Rock et al (1998) and Labie (2001)). Corporate governance ensures the protection of interests of owners by minimizing agency costs arising due to separation of ownership and control (John and Senbet (1998)). For MFIs, governance is of specific importance because of high organizational complexity. High complexity is primarily due to the presence of different types of MFIs, both non- and for-profit, and the double bottom line in the mission (Hartarska (2005)). Governance is for a large part done by the board of directors and CEO’s therefore background information of governance will be provided in this paper.

(6)

influences on MFIs. Literature on what impacts MFI performance, and more specifically, international influences on MFI performance, is scarce and is further investigated the next section.

2.2 MFI Performance Literature

Literature on the determinants of performance of MFIs is limited. Cull et al (2006) find that individual-lending increases profitability whereas Mersland and Strøm (2009) conclude the opposite; that individual lending does not effect sustainability and that group-lending increases outreach. Armendariz de Aghion and Morduch (2005) reveal that group-lending increases repayments rates and thus performance. Hartarska and Nedolnyak (2007) and Hartarska and Mersland (2009) do not discover a relation between regulation and performance, neither through sustainability or outreach. However regulation can indirectly affects savings and clients reached. Hartarska (2005) finds that auditing and rating has minor impact. Several studies (Mersland and Strøm (2008), Cull et al (2006) and Hartarska (2005)) have focused on ownership type of MFIs comparing non-profit to for-profit firms and find that ownership does not influence sustainability or outreach. Hermes et al (2009) conclude that commercial MFIs are expected to promote efficiency and that higher efficiency is negatively related to the outreach. Hartarska and Mersland (2009) further find that competition, in combination with bad internal governance mechanisms, decreases MFI performance.

(7)

optimal. Creditors on the board, however, increase efficiency and clients on the board increase outreach to more clients however these clients tend to be less poor.

To my knowledge, Mersland et al (2011) is the only research conducted on international influences on MFI performance. The indicators used for international influence are: whether the MFI initiator was international, international directorship and international debt and whether the MFI is part of an international network. Results show that an international initiator increases outreach, especially through the outreach to women. International directorship is shown to have a negative effect on sustainability whereas it has a positive effect on outreach. Finally international debt and MFIs that are part of an international network enhance outreach whereas no affect is found on sustainability. Mersland and Strøm (2009) only investigate international directors and find a negative relation to sustainability and a positive relation to outreach.

Based on the above literature the importance of this paper is underlined because it adds to the scarce literature on factors influencing performance, especially in the light of governance variables. Other than Mersland et al (2011) no research has been done on international effects on performance of MFIs therefore this paper seeks to address this issue more closely by investigating the effects of international board members and international CEO’s of performance.

3. Theories

Due to the novelty of microfinance and lack of literature on this specific topic, general corporate finance and governance literature must be used to investigate microfinance. Furthermore, of the theories used for analysis, agency theory has its roots in the economic theory of the firm and is often referred to in corporate governance. The resource-dependency theory comes primarily from business frameworks. This section provides a background to the two theories which will thereafter be combined and applied in the hypotheses section to determine effects of international board members and international CEO’s on MFI performance.

3.1 Agency Theory

(8)

asymmetry. Information asymmetry occurs if, for example, the agent has more specific information than the principle which he can withhold or change to influence the decision-making process without the principle knowing. To minimize these interest divergences, incentives must be created or agents must be monitored to ensure the pursuit of right objectives. This process incurs costs which are referred to as agency costs (Jensen and Meckling (1976)).

Agency costs are often associated with the separation of ownership and control, however they can occur at all levels of the firm and between different actors; each pursuing their own interest at the cost of the firms’ profit maximization. Harris and Raviv (2008) explain how insiders (e.g. CEO’s, executives) at the board level and outsiders on a board of directors will never provide full information because both parties have private incentives in the decision-making process. Agency costs can also arise between the top management and employees, where top managers must rely on information from experts and employees who have knowledge about specific firm operations. At the bottom level there is a discrepancy between the firm and the clients; on the one hand because the clients have less information about the products offered by the firm and on the other hand because the CEO/employees have less information about the clients or the market. Aligning interests and providing complete information calls for monitoring and good governance within a firm to reduce agency costs and maximize firm goals.

Internationalization of firms creates an extra dimension in which information asymmetry arises (Sanders and Carpenter (2002)). Internationalization increases the amount and complexity of information faced by decision-makers. Foremost, culture and institutional differences increase for cross-border firms. Furthermore, local information and specialized knowledge of the local market create informational asymmetries between the firm and its actors and the local environment. Lastly, Berger et al (2000) state that cross-border firms often have extra layers of management inducing more barriers, preventing the smooth flow of information.

(9)

3.2 Resource Dependency Theory

The resource dependence theory as proposed by Aldrich and Pfeffer (1976) is based on the presumption that an organization cannot maintain itself without external resources. A key idea is that interdependencies with other organizations and the environment will influence the strategic choices of management. Furthermore, the resource-based theory states that specific resources within or acquired by a firm can create a competitive edge. Hall (1993) identifies intangible resources such as organization networks, know-how and culture of organizations on which a firm can build their advantage on. Resources can also mean natural resources or access to capital. The idea behind this theory is that knowledge, expertise, information as well as access to capital can be attained through external parties which a firm in itself does not have. A key example is the extent to which (external) board members can add value to a firm through their expertise, access to (knowledge) networks, capital, information and/or experience.

There is some empirical literature on international affects and the resource-dependency theory to underline the relevance of this theory when analysing MFIs. Barney et al (1991) use the resource dependency theory to show that internationalization leads to the introduction of superior knowledge into the firm. Superior knowledge refers to intangible assets such as management skills, routines and expertise. Berger et al (2000) refer to this as the global advantage hypothesis and states that efficiency in firms can increase with foreign ownership through the introduction of expertise and skills from abroad. General effects of foreign ownership are investigated by Wagner (2005) who finds that internationalization of German firms increased performance. From corporate finance literature Lensink and Naaborg (2008) and Gajnoni et al (2003) find results that support positive effects of foreign ownership on performance. Claessens et al (2001) explain that internationally owned firms in developed countries benefit more from external superior knowledge than developed countries that already have access to this knowledge. Although this paper does not focus on ownership, this literature provides evidence of international influences on performance.

(10)

4. Hypothesis

4.1 The Effect of International Board Members

The need for governance in MFIs has been established in section 2.1. The main way to accomplish governance is through the board of directors who have the primary responsibility to provide good governance by limiting agency costs between different agents of the organization (John and Senbet (1998)). Before turning to the specific role played by international board members in MFIs, this section first explains the composition and the role of the board of directors, in general and with respect to MFIs. This is important because it will determine how much influence a board has in an MFI and it will indicate the important role that an international board member can have on the board can have.

In general, the board of directors consists of owners and representatives of owners. Owners can range from managers and CEO’s of the MFI, to shareholders and shareholder representatives and capital providers, including wealthy individuals that own private equity. Managers or CEO’s have an interest in the firm. Shareholders and shareholder representatives are part owners. The main function of a board of directors is to monitor processes and set out strategic plans for the MFIs, usually this is done together with the firms’ management (Solomon (2010)). Specific duties are also evaluating the CEO and representing owners or stakeholder interests. The board of directors is expected to act independently of the firms operations and focus primarily on the long-run sustainability of a firm.

(11)

board it is important to realize that the type of board will affect the extent to which board members can influence performance.

For all boards, a specific task is to minimize agency costs and increase objective monitoring. One prominent way to achieve this is by creating higher board independence through outside directors (John and Senbet (1998)). “Independence is referred to as a person having no connection to the firm, thus no family members, CEO’s or any of the firms’ clients as well as owners that own 5% or more of the shares in the company.” (Monks (2008), page 265) Independent directors on a board are assumed to be relatively free from conflicts of interests and therefore more able to monitor diversions of management from firm objectives. Before turning to the role of independent board members on the board of MFIs, the specific characteristics of MFI boards are discussed.

For MFIs, the board of directors is becoming more important as the need for governance is increasing (Hartarska and Mersland (2009), Rock et al (1998) and Mersland and Strøm (2009)). Rock et al (1998) provide an account of responsibilities to which a board of directors of a MFI specifically must cohere to. First, they must maintain a distance from daily operations, second, they must use experience from past directors and provide continuity for the board itself and third they should make decisions as a group. Furthermore, boards are required to take into account country-specific legal requirements in their strategies for the MFIs. Hartarska((2005), page 1630) states: “The board of directors is an internal governance mechanism that helps resolve the agency problems by Principles and Agents” by carrying legal responsibilities and performing effective monitoring. In taking these responsibilities the board distances itself from operations and remains objective in monitoring and guiding the organization.

As with a general board, the board of a MFI represents owners and stakeholders. A key difference between MFI boards and other boards, however, is that the owners and stakeholders (e.g. donors, clients, employees) and diverge from shareholders (e.g. large mutual funds, private investors) (Hartarska (2005)). The different types of stakeholders and shareholders, in combination with the dual-mission of MFIs, make the representative task of the board of directors more complex. Depending on the type of stakeholder or shareholder represented by the board, the board of directors may focus on sustainability (or even positive returns on investments) or on outreach (Labie (2001)).

(12)

lead to information asymmetries because the board of directors relies on information from external actors or internal CEO’s which may not be provided honestly or completely if there are diverging interests. Although the distinction in objectives is not so clear-cut and cannot easily be observed it can influence priorities of a MFIs board and how board members can influence MFI performance.

It is for this reason that independent directors, as a solution to conflicts of interest and information, asymmetries are expected to play an important role in MFIs. Empirical evidence on MFI board independence provides support for this. Hartarska (2005) finds that more independent boards are more effective and Hartarska and Mersland (2009) find that performance increases with a larger proportion of outside directors and independent directors.

Board independency as a solution to effective governance by the board is important in this paper because international board members can be seen as relatively independent board members. Mersland et al ((2011), page 165) expect that “in relation to monitoring the microbank - an international director can take on a special independent role as he/she is less part of vested domestic interests.” Oxelheim and Randoy (2003) also use international board members as an indication of board independence. International board members creates a more objective board and results in less agency costs between the stakeholders and shareholders thus positively influencing MFI performance.

A negative effect of higher independence can occur if information asymmetries between the top management (e.g. the CEO) and the board of directors can grow. This is enlarged when there are more international board members because the distance between them and firm will be even larger due to cultural and organizational differences. These informational distances can give rise to agency costs depressing MFI performance.

(13)

Concluding, international board members play an essential role in the governing of MFIs. International board members are expected to have a more independent role in the board which can diminish agency costs arising due to information asymmetries and conflicts of interests. These effects are expected to be larger than information asymmetries created due to (cultural) distances between them and top management. Furthermore, knowledge and access to funding are expected to increase with international board members, positively affecting MFI performance. In the different effects explained above, the positive effects are expected to outweigh the negative effects. Therefore the first hypothesis is:

Hypothesis I: The effect of international board members on the board is expected to increase MFI performance.

4.2 The Effect of an International CEO

The chief executive officer (CEO) has the highest authority within a firm and has a central position as he is the communicator between the board of directors and the internal firm, the operations and the other key executives. There are four standard responsibilities that a CEO has. First, the CEO communicates and exchanges information between the different internal and external parties. Second, the CEO is the key player in the decision-making process, in which interest of all involved parties (board of directors, shareholders and the internal members) are taken into account. Third, the CEO has a leadership position in managing the firms operations. Finally, the CEO, together with the board of directors, is to design a long-term vision and strategy for the firm. Of this, Lafley (2009) states that the most important role assigned to the CEO, which no one else in the firm can do, is that of communicating between the outside world and the internal firm. Before turning to the resource-dependency theory and agency theory to determine CEO effects of MFI performance, the managerial discretion theory will be used to show why, in general, CEO’s have the power to influence firm performance.

(14)

the task environment, the organization environment and CEO specific characteristics (Finkelstein and Hambrick (1990)).

The task environment refers to environmental munificence, dynamism and complexity (Dess and Beard (1984)) and includes product differentiability, market growth, demand instability and regulations (Finkelstein and Hambrick (1990)). If the task environment becomes more complex with more information, the CEO has more options in the decision-making process providing him with more managerial discretion. Managerial discretion specifically rises if the industry is growing, if regulation is low and uncertainty in demand is high (Haleblian and Finkelstein (1993)). Organization environment refers primarily to the extent to which a firm has a strong legitimacy in their routines and mission. If a firm is well established then the flexibility of the CEO is reduced because he must confer to the standard approaches already in place. Organization environment also relates to the amount of resources available to the firm. If resources availability is low or resources used are scarce then the CEO has fewer options so discretion decreases. CEO specific characteristics refers to demographic and psychological characters ranging from political acumen, cognitive complexity, tolerance for ambiguity, locus of control to aspiration levels and power base.

(15)

also the local government. Although funds are not scarce for MFIs the different acquiring methods and the many actors involved increase complexity. The last level through which a CEO obtains discretion is by specific CEO characteristics. These characteristics are both formal and psychological. I will not provide further analysis of this because it is beyond the scope of this paper.

So far I conclude that a CEO of a MFI has high managerial discretion in the decision-making process with which he impacts performance. First, the CEO has managerial discretion because MFIs and their environments are complex and fast-growing creating more opportunities for a CEO’s. Second, because the CEO has a key role in managing resources and making decision in the best interest of the firm. Empirical evidence for powerful CEO’s with high managerial discretion is found to positively influence performance variability (Galema et al (2009) and Adams et al (2005)).

Having established that the CEO has high managerial discretion in MFI we turn to the expected effects themselves. The role of a CEO within a MFI is similar to the role explained above. An

international CEO is a CEO in a firm located in a country which is not their own. Specifically,

the international CEO is expected to come from a more developed country than the country in which the MFI is located. Sanders and Carpenter (2002) state that internationalization increases complexity, thus discretion is expected to increase simultaneously. The agency theory and resource-dependency theory will be used to analyze whether an international CEO with his discretionary power positively or negatively affects the performance of an MFI. The extra dimension analyzed is that the CEO is international.

(16)

mostly female (Armendariz de Aghion and Morduch (2005)). Lack of client knowledge is also provided as an explanation for the negative effects of international directors on performance. I expect that international CEO’s face larger (cultural) distances between them and the clients providing international CEO’s with an informational disadvantage which has negative effects on performance.

At a second level, informational asymmetry can occur between an international CEO (the principle) and his employees (the agents). This is most evident in how a CEO intends to run an MFI, by for example imposing European or American systems such as centralization in decision-making or not can influence MFI performance. Armendariz de Aghion and Morduch (2005) explain that, with hindsight, top executives of MFIs were too centralized in the decision-making process. Furthermore, Mersland and Strøm (2009) state that international directors bring a culture of high costs to MFIs. If an international CEO introduces foreign systems into a firm without sufficiently adapting to local employees efficiency can also decrease.

The other side of this coin is however is that international CEO’s are expected to bring resources and expertise into the MFI. The CEO can introduce an efficient system and ensure adaption to the local firm and market then this can decrease costs and increase performance. Based on the resource-dependency theory the CEO may also have access to international networks and funds, either directly or through the board of directors, which can further increase performance.

Concluding, the managerial discretion theory is used to show that a CEO has the power to impact performance. In any firm the CEO is expected to positively influence performance or else he will be removed from the CEO position. International CEO’s in MFIs add two extra dimensions to the CEO position. Firstly, negative agency costs diminish the positive effects on performance whereas, secondly, resources and expertise can increase positive effects on performance. The overall positive effects are expected to outweigh the negative agency costs. Therefore the second hypothesis is:

(17)

5. The Data, model and Methodology

5.1 Data

The dataset consists of information obtained from risk assessment reports from five rating agencies. These rating agencies are Microrate, Microfinanza, Planet Rating, Crisil and M-Cril. The reports are approved by the Rating Fund of the Consultative Group to Assist the poor and are publicly available. (C-gap: www.ratingfund2.org1) The rating agencies use common indicators when rating MFIs performance, compiled by the Inter-American Development Bank and Microrate. Using data from third parties is expected to be more reliable than using self-reported data.

The dataset contains information from 379 MFIs from 74 countries. Data over the period 2000 to 2006 is used. The dataset contains both financial variables and governance variables of MFIs for up to four consecutive years. When necessary data has been annualized and converted to US dollars using official exchange rates at that time.

The dataset is subject to many missing values and the effect on this should be noted as it will influence the results considerably. The large amount of missing values is both because information is not available and because rating reports are only constructed every four years. For governance variables this means an observation is only available once in every four years. The research presented here entails a cross-section analysis to overcome this issue. An average is calculated over the period of 2000 to 2006. The average consists of the average of the observations divided by the number of observations. In the case of one observation available over the entire period then this value is used. If there are 3 observations over the period, an average of these 3 observations is used. The average is calculated both for financial as well as governance variables. Mersland and Strøm (2009) use a random effects method to transform the data to overcome the missing value problem and treat governance variables as constant using one observation available. Assuming governance variables to be constant is not done in this paper because for some governance variables different values are given and arbitrarily choosing one of these to be constant ignores information available.

1_{C-gap: The consultative Group to Assist the Poor. It includes development agencies to provide financial}

(18)

A selection bias is present because the sample only includes officially rated MFIs. Using rated MFIs ensures that these firms prioritize improvement of receiving access to funding by providing transparency of their firms through the rating reports. Especially commercial for-profit firms must provide this information for their investors. This selection bias may be an advantage to this research as I expect international influences to increase with more professional MFIs that engage in ratings. However it may decrease the comparative effect of international influences on MFI performance because the MFIs included in this sample do not include small or local run MFIs, less influenced by international factors.

5.2 The model

Two hypotheses about the effects of the international influences on MFIs performance have been formulated. The model used to test the hypothesis is shown below in equation (1).

PERFORMANCE = α + β1MEMB +β2CEO + β3CONTROL +

ε

(1)

PERFORMANCE is measured in terms of outreach and financial sustainability. This is to verify

whether there are discrepancies in performance results as there is a double-bottom line in the mission. The dependent variable used to measure financial performance is operational-self-sustainability (OSS) and for outreach performance it is the number of clients (lnCLIENTS).

MEMB is the number of international board members on the board of a MFI. Based on

Hypothesis 1, this effect is expected to be positive so β1 is expected to be positive. CEO indicates

whether the MFI has an international CEO or not. Based on hypothesis 2, if there is an international CEO then he will have a positive effect on performance so the expected sign of β2 is

positive. As verified earlier, the term international is used to refer to influences from abroad thus influences from a country outside the country in which the MFI is located. CONTROL is a vector of control variables each exerting a different effect on performance therefore the expected sign of

β3 is ambiguous. The control variables are portfolio at risk, firm size, lending methodology,

(19)

5.3 Variable Description

Table 1

Variables Description

Dependent

OSS Operational-self sufficiency

lnClients The natural log of total clients

Independent

IntMemb Number of international board members on the board

IntCEO Dummy stating whether CEO is international (DM=1) or not (DM=0)

Control

Par30 Portfolio at Risk: fraction of the portfolio that is delinquent for at least 30days

FirmSize Firm size: measured by the natural log of total assets

LoanMeth Dummy for lending methodology

Comp A scale ranging from 1-7 indicating market competition faced by the MFI

MFIage The number of years the MFI exists

5.3.1. Dependent Variables

Operational self-sufficiency (OSS) measures how well the MFI is financially sustainable and entails that operating revenues should be larger than the costs incurred. Operational-self-sufficiency is a common measured use in the microfinance literature (Mersland and Strøm (2009)) and Hartarska (2005)) and is measured by the following equation:

OSS = operating revenue / (financial expense + loan loss provision + operating expense)2

Previous research used several measures for financial performance variables. Therefore I tested the correlations between OSS, return on assets and return on equity (see appendix A for results). The three measures are highly correlated therefore I choose to use only one measure as it reflects the effects on the other performance indicators.

There are several reasons why OSS is used instead of other financial indicators. Firstly, because financial subsidies prevail in the microfinance industry, OSS is an accurate measure because it can be adjusted for the financial subsidies (Grau (2004)). Secondly, OSS is used rather than other financial indicators because standard accounting measures are often subject to country-specific regulations and accounting practices which make it hard to compare profitability if a global

2

(20)

dataset is used (Hartarska and Nadolnyak (2007)). From table 2 we can see that the mean of OSS exceeds 1 which indicates that the average MFIs in this sample are self-sustainable.

CLIENTS represents total clients and is a measure used to evaluate the MFIs performance in the

light of outreach to the poor. It includes all clients, both saving and credit clients. This measure of performance is included to investigate the extent to which microfinance reaches clients, following Schreiner (2002) in his depth measure. Important to keep in mind is that this measure in itself neglects the size of the loan or the type of clients reached however those aspects are beyond the scope of this paper. In table 2, lnCLIENTS is the natural logarithm of total clients. The natural log is taken to minimize effects due to the large spread in the number of clients between the different MFIs. For all regressions lnCLIENTS is used as the dependent variable for outreach.

5.3.2 Independent Variables and Control Variables

The independent variables of interest consist of international board members and international CEO’s present in MFIs. The first hypothesis tests the effect of international board members on MFI performance. IntMEMB is the number of international board members on the board of a MFI. The variable ranges from 0 to 6. International board members as a share of the board, as Mersland and Strøm (2009) do, is not used because I assume that international influence can already be exerted by one international board member. The second hypothesis predicts a relationship between an international CEO and the performance of an MFI. The variable IntCEO consists of a dummy variable which is 1 if the CEO is international and 0 if the CEO is national.

Five control variables are included to capture firm- or country-specific effects. These are based on the findings in section 2.2 of what general variables influence performance or on control variables that other research has used. Par30 is the fraction of the portfolio that is more than 30 days late in repayments. This is included as a risk indicator and an indirect performance measure.

FirmSize is included to control for effects that are specific to a firm being large; for example a

(21)

measured with a scale ranging from 1 to 7. Low competition is close to 1 and 7 reflects high competition. Competition is expected to increases environmental complexity possibly providing CEO’s with more power to influence performance. Economic theory expects competition to lead to higher efficiency, affecting performance. For MFIs, however, Hartarska and Mersland (2009) find a negative effect on performance. LoanMeth is a dummy variable referring to the type of lending methodology a MFI engages in. Three types are identified; village banking (DM=1), solidarity groups (DM=2) and Individual loans (DM=3). The main distinction I will make in the analysis of this paper is between group- and individual-based lending because the effect of this distinction remains ambiguous in their effect on performance. Group lending includes both village banks and solidarity groups. MFIage is included as a proxy for the experience MFIs have gained over the years. MFI age can further reflect the organization environment which is more structured for older MFIs providing CEO’s in young MFIs with more managerial discretion, thus more effect on performance.

An important note on the control variables is that many proved to be insignificant indicators of performance when running regressions in this paper. More control variables were tested, such as regulation and inflation to capture more country-specific effects and CEO-chairman duality and board size to capture MFI specific effects; however coefficients and significance did not change (Results not shown). See section 6.4 for further analysis of possible control variables.

5.4 Methodology

5.4.1. Explorative Analysis

First an explorative analysis is done to provide a general picture of international and non-international oriented firms. An non-international oriented MFI entails those where either or both international board members or international CEO’s are present. This is interesting to determine differences that may not be captured in regression analysis. It is important to keep in mind that these results do not report about the strength or importance of relations found with respect to performance. These results merely show whether there are differences between international oriented MFIs and non-internationally oriented MFIs.

5.4.2. Cross-section OLS

(22)

lost when doing regressions because it diminishes the amount of observations that do not match with the observations available of other variables or years. The downside of this approach is that the time variable is dropped therefore changes in variables cannot be taken into account. This approach is contrary to Mersland and Strøm (2009) and Hartarska (2005) who use a random effects panel data estimation to determine the effects of governance and financial variables on financial performance and outreach. These studies overcome the missing data problem by opting to use a constant for the time-invariant governance variables for which many observations are missing. However this approach does not overcome the loss observations due to mismatches between observations and variables.

The regression analysis will be approached as follows. First bivariate regressions will be run between the two independent variables and the dependent variables. These regressions will not include any control variables. Thereafter extended regressions will be run including control variables. Each regression is subject to a few diagnostic tests to validate statistical significant.

5.4.3. Regression Diagnostics

In order to use an OLS regression there are three assumptions that must be fulfilled. The most important assumption for this paper is multicollinearity. Correlations between different variables can reduce significance of variables therefore they must be treated with caution if included in the regression. Correlations are tested for using pearsons pairwise correlations. If significant correlations are above 0.8 they will be excluded from the regression. This threshold is also used by Mersland and Strøm (2009) and Mersland (2011). The results are shown and discussed in the descriptive statistics section. To double-check for multicollinearity a post-estimation VIF (variance-inflation-factor) test is performed. This test determines whether coefficients of other variables change due to the inclusion of another variable. If problems arise then specific correlated variables will be removed from the regression. A second assumption underlying an OLS is equal variances of the error terms, referred to as heteroskedasticity. The post-estimation test, Breusch-Pagan test for heteroskedasticity was performed on all regressions. The null-hypothesis assumes equal variance thus if the results are significant, the null-hypothesis is rejected and I assume heteroskedasticity. If heteroskedasticity is present the regression will be made robust. This should control for heteroskedasticity and allows the regression to be used for further analysis3. The results will state whether regressions

3

(23)

are made robust however results for each test conducted will not be shown. A third assumption for a standard OLS is a normal distribution of the error terms. The Shapiro-Wilk W test is used to test the normality of the error terms. For all tests the p-value was significant indicating non-normal data. The analysis will continue using this data as I assume that most data will be non-normality distributed however it is important to keep in mind that the data may be bias.

6. Empirical Results

This section begins with a short explanation of the descriptive statistics of the variables used in this paper. It continues with a brief overview of results from the explorative analysis of the dependent and independent variables of interest. Thereafter the results from the econometric analysis are presented. I will conclude this section by providing some drawbacks of the analysis.

6.1 Descriptive Analysis

Table 2 Descriptive Statistics

Variable Obs. Mean Std. Dev. Min. Max.

OSS 258 1.738 2.120 .134 14.382 CLIENTS 339 17806 43688 183 513000 lnCLIENTS 339 8.768 1.390 5.209 13.148 IntMEMB 236 .561 1.186 0 6 IntCEO1 262 .061 .235 0 1 Par302 372 .069 .088 0 .614 FirmSize 400 14.743 1.338 10.724 19.031 LoanMeth 379 2.308 .787 1 3 Comp 278 4.328 1.588 1 7 MFIage 400 16.545 7.641 5 88 1

Because an average is taken for over time for all variables, dummy variables can also take a value between 0 and 1.

(24)

Firstly, it can be noted that the number of observations for the independent variables are lower than the number of observations of the dependent and control variables. This can be attributed to that fact that governance variables and fixed firm characteristics and come from the rating reports only done every 4years and are therefore subject to many missing values.

The mean of OSS exceeds one therefore I conclude that the average MFI in the sample is financial sustainable. The range of OSS is from 0.13 to 14.38 which is quite large. This is similar for the number of clients reached. The large differences signify large differences between MFIs.

CLIENTS varies from 183 to 51300 and is heavily skewed. The majority of MFIs are at the lower

end however there is a large tail reaching to the right due to some MFIs with a large outreach. By taking the natural log these skewed results are minimized. (See graphs in Appendix A). Therefore

lnCLIENTS is used in the regression. These ranges are high because the sample includes MFIs

from all over the world and because the diversity in MFI sizes is high.

IntCEO is a dummy therefore the mean is the percentage of CEO’s that are international; 6.1% in

this case of IntCEO. IntMEMB ranges from 0 to 6 therefore the mean indicates the average number of international board members on the board. The average is 0.56 however this is heavily skewed because of the many boards that have no international board members. Further analysis shows that 27.5% of the boards have at least one international member. For boards with at least one international member the average international members on the board is 2. The percentage of international board member and international CEO’s present in MFIs seem low. However I expect it to be representative of the number of international CEO’s in MFIs worldwide. It may even be that due to the bias created because the sample only includes rated firms, the international influences may even be exaggerated.

Par30 has a low mean indicating repayment rates are relatively high. FirmSize is the natural log

(25)

Comparative Analysis

Due to the large sample size of which only a small portion includes international observations the data in table 2 does not reflect information on the international-oriented firms. Therefore the comparative analysis is done to explore the data with respect to international influences. Table 3 supports the analysis. The table exhibits three subgroups and their means for the different variables. The four subgroups consist of data specific to: international board members presence (IntMEMB >0), international CEO presence (IntCEO>0) and their non-international counterparts (IntMEMB =0 and IntCEO = 0 respectively). The prior two groups will be referred to as international oriented MFIs whereas the latter two groups will be referred to as non-international oriented MFIs.

Table 3 Mean Comparison

Mean

Variable IntMEMB>0 IntMEMB=0 IntCEO>0 IntCEO=0

OSS 2.301 1.402 2.128 1.684 CLIENTS 10967 19121 15749 19894 lnCLIENTS 8.686 8.699 9.131 8.785 IntMEMB 2.038 0 2.730 .417 IntCEO .190 .009 .941 0 Par30 .036 .083 .016 .065 Assets 4885669 6946351 5507361 6917405 FirmSize 15.023 14.936 15.137 14.931 LoanMeth 2.354 2.360 2.176 2.307 Comp 4.398 4.419 4.093 4.377 MFIage 14.107 18.499 14.764 17.654

(26)

shows that of the 17 MFIs with an international CEO, only 2 have no international board members. This is not the case for international board members, who are present even without the presence of international CEO’s.

Of the control variables, Par30, Comp and MFIage are lower for international influences compared to non-international influences. Of these, competition and the age of the MFI are exogenous and results will not be influenced by the international orientation of MFIs. Par30 is lower, signifying better repayments for international oriented MFIs. International orientated MFIs seem to be slightly larger, indicated by firm size values. For all control variables, T-Tests do not show significant differences in means. Important to note when simply comparing means are the differences in sample size between international oriented MFIs and non-international oriented MFIs. The latter group is much larger, stretching data over a larger range of observations and reducing large differences in means.

Correlations analysis

The next step in the descriptive analysis is to investigate the correlations between variables as this can significantly influence relations between variables, affecting performance. Table 4 shows pairwise Pearson correlations of which there are several significant correlations, indicated by the star. Only the important or interesting ones will be discussed.

Table 4 Pearson Correlations

Pearson Correlation Dependent and Independent variables

OSS lnCLIENTS IntMEMB IntCEO Par30 LoanMeth FirmSize Comp MFIage

OSS 1.0000 lnCLIENTS -.0565 1.0000 IntMEMB 0.1258 0.0317 1.0000 IntCEO 0.0305 0.0688 0.4915* 1.0000 Par30 -.0781 -0.1080 -0.2390* -0.1526* 1.0000 LoanMeth 0.0330 -0.2644* -0.0555 -0.0458 0.1100* 1.0000 FirmSize -.0524 0.6473* 0.0784 0.0393 -0.0483 0.2240* 1.0000 Comp -.1308 0.0333 -0.0080 -0.0565 0.0618 0.0473 0.1220* 1.0000 MFIage -.0815 0.2854* -0.2209* -0.0953 0.1816* 0.0238 0.3666* 0.0904 1.0000 * indicates significance at 5%.

(27)

therefore regressions are run with each of them included separately and one regression including both to determine possible individual and joint effects.

An interesting observation is that IntMEMB and IntCEO are both negatively and significantly correlated to Par30. Par30 means repayments are made more than 30days in arrear; so if Par30 increases, the risk of repayment default increases, so performance is expected to decrease. This is reflected by the negative correlations between Par30 and OSS and lnCLIENTS, although these results are not significant. Based on these correlations, a preliminary conclusion can be made: as international orientation increases, Par30 decreases, thus performance increases. Par30 is also positively and significantly related to MFIage and LoanMeth. A possible explanation for this could be that information asymmetries are reduced and/or knowledge of the local market has increased over the years of experience and with respect to efficient lending methodologies. This decreases associated uncertainty and leads to higher repayments.

The highest significant correlation is between FirmSize and lnCLIENTS; this is as expected because as the firm size increases, capacity increases and more clients can be reached. FirmSize is also correlated significantly with MFIage which is expected because MFIs grow over the years. As explained in section 5.4.2. a maximum of 0.8 is a threshold for which variable can be correlated without resulting in multicollinearity in regressions. Therefore no variables will be excluded from the regressions. To further test this claim, each regression is accompanied by a post-estimation test checking for multicollinearity.

Next we turn to the regression analysis to determine marginal effects of international board members and CEO’s on the performance measures whilst controlling for MFI- and country specific effects to give a complete picture of the results.

6.2 Regression Results

(28)

Table 5 Regression Results

International Members International CEO’s _{International Members and CEO’s}

OSS lnClients OSS lnClients _OSS _lnClients

Columns 1 2 3 4 5 6 7 8 9 10 11 12 IntMEMB 0.216* 0.130 0.041 0.007 0.328 0.216 0.036 -0.018 (0.091) (0.373) (0.639) (0.902) (-0.039) (0.229) (0.737) (0.820) IntCEO 0.249 0.328 0.415 0.146 -0.918 -0.663 0.262 0.051 (0.679) (0.647) (0.283) (0.540) (0.167) (0.462) (0.610) (0.878) Par30 -1.328 -1.775** -1.398 -1.226 -0.405 -1.946** (0.304) (0.021) (0.319) (0.222) (0.772) (0.048) LoanMeth 0.181 -0.777*** 0.262 -0.784*** 0.269 -0.798*** (0.280) (0.000) (0.123) (0.000) (0.161) (0.000) FirmSize 0.130 0.883*** 0.039 0.885*** 0.022 0.873*** (0.906) (0.000) (0.661) (0.000) (0.850) (0.000) Comp -0.155 -0.007 -0.194 -0.023 -0.173 0.001 (0.171) (0.861) (0.105) (0.555) (0.169) (0.986) MFIage -0.013 0.023*** -0.013 0.017 -0.021 0.000*** (0.220) (0.000) (0.171) (0.141) (0.072) (-0.014) Constant 1.544*** 1.851 8.676*** -3.077*** 1.699*** 1.551 8.783*** -2.797*** 1.625 1.729 8.643 -2.966*** (0.000) (0.268) (0.000) (0.000) (0.000) (0.259) (0.000) (0.000) (0.000) (0.321) (0.000) (0.000) N 182 154 222 188 187 158 245 211 160 136 198 170 R2 0.016 0.05 0.001 0.693 0.001 0.061 0.005 0.709 0.028 0.068 0.004 0.705

Values in parenthesis are the p-values.

***, **,* indicate significance levels at 1%, 5% and 10% significant level, respectively.

6.2.1. Sustainability

The first column shows bivariate linear regression for IntMEMB on OSS. Results show that

IntMEMB is significantly related to OSS at 10% level. The sign of IntMEMB is positive as

expected in hypothesis 1. This effect however diminishes and becomes insignificant when control variables are included, as shown in column 2. Column 5 and 6 show the results of IntCEO on

OSS. In the simple regression IntCEO is not significant. Including control variables does not

(29)

IntCEO. This means that if there are international board members present, there is high

probability that there are also international CEO’s however regressing only one means the effect is captured by the one included. When both are included the effect will be captured by the variable that has most influence, reducing or changing the sign of the other variable.

For both regressions on OSS the control variables are insignificant. Even after including more diverse control variables or dropping others (not shown) the coefficients changed only marginally and they remained insignificant. Par30, Comp and MFIage have negative coefficients whereas

LoanMeth and FirmSize have positive coefficients. These results are as expected for Par30;

because high repayments rate should lead to higher performance. Although higher competition is expected to increase efficiency, results found here underline the negative results on performance found by Hartarska and Strøm (2009). The positive coefficient for LoanMeth shows support for the results found by Mersland and Strøm (2009) that individual loans have a positive effect on

OSS. The positive relation between FirmSize and performance are confirmed by these results. The

negative coefficient of MFIage, in line with results found by Mersland et al (2011) are surprising as I expect performance to increase with MFIage as experience and (local) knowledge increases. A possible explanation provided by Hermes et al (2011) for the negative coefficient is that new MFIs incorporate mistakes from the past making them more efficient than MFIs in under-going a transition. The results found for control variables are similar for the joint-regression. Again, the coefficients are slightly larger; however this is attributed to the fact that there are less observations in the regression. The constant is only significant in the regression without control variables. The effect of Comp and MFIage are reduced to being close to zero indicating that they have no influence on outreach.

R2 is very low in all regressions for on OSS. This remains low even when including more or less control variables. The low R2 is attributed to the lack of explanatory power of the control variables; however a low R2 is not uncommon for a cross-section OLS analysis.

6.2.2. Outreach Performance

Columns 3 and 4 show the effects of IntMEMB on lnCLIENTS. Columns 7 and 8 present the results of IntCEO on lnCLIENTS. For both the simple regression and the regression including control variables IntMEMB and IntCEO have no significant effect on outreach. The coefficients are however both positive. Indicating some support for hypotheses 2, although it is not

(30)

remain positive. IntMEMB seems to have a negative effect on outreach once control variables were included to the regression.

For the regression of IntMEMB on lnCLIENTS the control variables par30, Loanmeth, FirmSize and MFIage are significant. Par30 has the expected negative effect, FirmSize is positive, and

MFIage is positive. LoanMeth is negative; providing support to Armendariz de Aghion and

Morduch (2005) who state that group-lending increases outreach. Comp is insignificant and negative. For the regression IntCEO on lnCLIENTS, LoanMeth is again negatively significant and

FirmSize is positively significant. The other control variables are insignificant, Par30 is negative, Comp is negative, and MFIage is positive. For the joint regression results for the control variables

are the same as for the distinctive regressions.

R2 for the outreach regressions are very high, 0.69 and 0.70 for IntMEMB and IntCEO regressions, respectively. This can be attributed to the high effect of FirmSize as control variable. Excluding FirmSize reduces R2 considerably.

6.3 Drawbacks

In general statistical analysis, more observations provide better statistical results. Drawbacks to the investigation primarily concern the amount of missing values and low observations for governance variables leading to lower observations. This paper overcomes a large amount of missing values by taking averages and creating a cross-section. This allows for a comparison without too many mismatches due to lacking data. A drawback of this is that changes in variables over time cannot be taken into account. Observations are further diminished because governance variables are only rated once every 4 years.

Furthermore the data consisted of a large sub sample of non-international oriented MFIs compared to a small international-oriented sub sample. This means the likelihood that a statistical Type II error is made will increase. A Type II error is made when a false null hypothesis is not rejected. In this case, because the variables are not significant the results suggest that international effects are zero. However, insignificance may occur because the effects of the large non-international subsample may be larger than effects of the small non-international subsample, decreasing significance levels of possible international oriented coefficients.

(31)

be larger for cross-sectional data because OLS assumes homogeneity in variables although these variables may actually change over time. This can inflate heteroskedasticity and correlations. Furthermore, heteroskedasticity most likely arises due to large differences between MFIs, as seen in the descriptive analysis. Large firms probably also have large client outreach and vice versa causing correlations in the error terms to be correlated.

A last drawback of this analysis is that it cannot test for causality. The different independent and dependent variables may have a causal link; especially concerning international variables. For example; does an international board member positively influence performance or does the international member go to a board of a well-performing MFI?

6.4 Robustness Test in Control variables

A robustness analysis is done using more control variables. The reason for this is two-fold. First, due to results for low R2, further testing was done to determine whether more or different types of control variables could increase the explanatory power of the model. Second, a very basic analysis is done to investigate variables that may indirectly influence performance through international board member and international CEO’s. These variables are chosen based on underlying agency and resource-dependency theories because they may provide access to resources, reduce information asymmetries and increase managerial discretion. Although this analysis is far from complete it may add to or provide some explanation to previous results.

Table 6 provides an overview of the variables tested. An explorative analysis as explained in section 5.4.1, and shown in section 6.1 for the main model, is done to determine differences between international oriented and non-international oriented MFIs. Results of the mean comparison and correlations are shown in the appendix and are used here to explain why variables are of interest in this sub-model.

Table 6 New Control Variables Description

Variable Description

International Initiator DM whether the MFI was initiated by an international organization (DM=1) or not (DM=0)

Network DM whether MFI is part of an international microfinance network (DM=1) or not (DM=0)

CEO-founder DM whether the CEO is also the founder of the MFI (DM=1) or not (DM=0)

CEO-tenure The number of years the CEO is with the MFI

CEO-Chair DM whether the CEO is also the Chairman of the board (DM=1) or not (DM=0)

Board Size The number of board members

(32)

Being part of an international network is expected to increase access to funds and knowledge. International oriented MFIs have a higher mean value for international networks which can increase positive effects from the resource-dependency theory. Whether the MFI was an international initiative is much higher for both international oriented subgroups than non-international oriented sub-groups. I expect that non-international initiatives increase benefits due to resources from abroad. Founder and Tenure refer to whether the CEO is also the founder and the tenure of the CEO, respectively. They are included because the longer the CEO is part of the MFI, the more client and local knowledge the CEO must have decreasing agency costs arising from information asymmetries. There is no obvious difference between the different subgroups with respect to these variables, confirmed by a T-Test. CEO-Chair is included to determine whether the board of the MFI has a one-tier or two-tier system. A two-tier system reducing agency costs thus increasing performance. Board size influences the extent to which international board managers have discretion within a board. Regulation is expected to decrease managerial discretion because the CEO has less power in the decision-making process. Although regulation has proved not to be significant on performance directly (which is why it is not included in the main model) I do include it here in combination with the CEO.

First a regression analysis is done to determine the effects of these variables on international board members and international CEO’s. The results are shown below:

Table 7 Regression IntMEMB and IntCEO

Variables IntMEMB IntCEO

Network 0.490*** 0.053 (0.003) (0.120) Initiator 0.883*** 0.091*** (0.000) (0.007) BoardSize -0.029 -0.005 (0.199) (0.248) Regulation 0.437*** 0.062* (0.010) (0.084) Constant 0.182 0.032 (0.350) (0.427) N 222 233 R 0.247 0.076

From the table it can be seen that the new control variables significantly influence IntMEMB and

(33)

initiator and regulation are positively related to international board member and international CEO presence. Important to keep in mind is that this regression only indicates a relation but does not provide causality relations between the variables. Based on this, strong support is found that the variables influence international board members and international CEO’s. Therefore a regression, as was done for the main analysis, is performed to determine possible effects on performance.

Only the interesting features will be highlighted in this regression analysis. First of all, regressions on OSS and lnCLIENTS using IntMEMB do not show any significant results. Coefficients of IntMEMB are smaller. A plausible explanation for this is that some variables explain performance thus they reduce the explanatory power of IntMEMB. These results point to possible indirect influences by the variables included on IntMEMB and performance. The regressions for IntCEO show negative coefficients for OSS and positive coefficients for

lnCLIENTS, although all insignificant. Thus the new variables have an indirect effect that make

the CEO coefficient negative whereas it was positive in the previous case. The coefficients for

lnCLIENTS remain positive and are larger than the main model. Interesting in the regression

analysis is that the variables that proved to have a significant relation with IntMEMB and IntCEO do not show up as significant for the performance measures. This indicates possible international influences, not directly on performance, but through international board members and international CEO’s.

This robustness analysis is limited. The main model of this paper provides a relation between international board members and international CEO’s on performance. This robustness test makes a first step in determining indirect effects of further variables, through possible relations to international board member and CEO’s, on MFI performance . It is beyond the scope of this paper to provide a full account of possible effects and variables influencing international board members and international CEO’s however it can serve as possible explanations to the results found in the main model and it can provide a basis for future research.

(34)

coefficients. Although results indicate that there are indirect effects, the exact direct and indirect effects remain unknown and should be subject to further research.

7. Discussion and Conclusion

7.1 Discussion

Based on the first hypothesis a positive effect of international board members on MFI performance is expected. Performance is measured in terms of sustainability and outreach to account for the double-bottom line in the mission. The effect of the number of international board members on the board was significant at 10% only for sustainability when regressed on OSS, as shown by the positive coefficient. However, this effect diminished and became insignificant when control variables were included in the regression. The number of international directors was not significant for outreach, measured by lnCLIENTS. Including both international board members and international CEO’s to the regression made the effect of board members on performance slightly negative, although it remained insignificant. This result underlines that international board members influence sustainability more than outreach however because results are insignificant, caution must be taken with this conclusion. Results reveal that no significant empirical evidence is found that international board members positively influence MFI performance.

The Influence of International Board Members and International CEO’s on Performance of MFIs