All in the family? The impact of founder directors and family governance on microfinance institutions' social performance

Research Question/Issue: This study investigates the impact of family governance, including founder directors and their ties to family members on the board, on the social performance of microfinance institutions (MFIs), a special kind of social enterprise with dual objectives. Research Findings/Insights: Using a dataset of 735 MFIs operating in Bangladesh from 2007 to 2017, we find that founder directors and board members with family ties to the founder have an adverse impact on MFIs' social performance. These findings hold when we perform several robustness tests and endogeneity tests. Theoretical/Academic Implications: We contribute to the corporate governance literature on MFIs and social enterprises in two ways. First, our findings suggest that, when MFIs are confronted with dual performance objectives, founder directors may “ trade off ” social outcomes in favor of economic outcomes and therefore adversely affect MFIs' social performance. Second, our findings extend the literature by showing that the presence of board members with family ties to founder directors also adversely affects MFIs' social performance. Practitioner/Policy Implications: This study suggests that MFIs' board composition influences their governance and ability to oversee their social and financial performance effectively. If MFIs' social performance is a major concern of national policy makers, then regulation should be put in place to limit board recruitment with family ties.


| INTRODUCTION
Motivated by the fact that the microfinance sector serves the banking needs of the poorest citizens of the world (Hermes et al., 2011;Lopatta & Tchikov, 2016), recent research has highlighted the need to focus on the social performance of microfinance institutions (MFIs) (Alon et al., 2020).The extant literature has suggested that good governance is vital for the performance of social enterprises such as MFIs (Centre for the Study of Financial Innovation [CSFI], 2014, 2016), but their governance is complex due to their simultaneous pursuit of both financial and social goals (Armendáriz & Morduch, 2010;Mersland & Strøm, 2009).This dual purpose may produce trade-offs in the extent to which both goals can be achieved equally well (Jensen, 2002).
Therefore, this study scrutinizes the effect of corporate governance factors (family governance in particular) on the performance in the microfinance sector.
Our starting premise is that MFI board members-pursuing both financial and social performance objectives-can be exposed to conflicting interests when facing multitask goals (Hahn et al., 2015;Holmstrom & Milgrom, 1991;Lynn, 2021).Although past research has addressed the benefits of independent directors for good MFI governance (Bassem, 2009;Hartarska, 2005;Hartarska & Mersland, 2012;Kyereboah-Coleman & Osei, 2008), the impact of founder directors and their ties to family board members has received far less attention in the MFI literature (Mori et al., 2015).Accordingly, this study takes the opportunity to address a research gap.
We posit that founders on MFIs' boards of directors face a tradeoff dilemma between financial and social interests that leads them, in effect, to make decisions that have an adverse impact on their social performance.An adverse relationship is posited because founder directors on the board are more likely to have invested significant time, effort, and finance and, to protect their capital interests, are likely to place greater emphasis on the financial outcomes than on the social outcomes, advancing the mission of MFIs.We also posit that this adverse relationship with social performance may be prevalent when the founder has ties to family members on the board.When more than one member of the founder's family is present on the board, these family members may try to dominate deliberations overall.Attempts to dominate board activity and decision making stem from the fact that founding members bring funds from their family resources to the MFI.We argue that, due to their embedded economic interest, board members who have a family connection to the founder exacerbate the trade-off problem, thereby adversely affecting the MFI's social performance.
To test our hypotheses, we use the setting of Bangladesh and employ an unbalanced panel dataset of 735 non-governmental organization (NGO) MFIs from 2007 to 2017.We chose Bangladesh because the modern version of the microfinance movement started here, the sector has grown fast, and the market is mature and now one of the largest in the world (Mia et al., 2017).The findings of our study suggest that founder directors on the board do adversely affect social performance.For instance, for a one-standard-deviation increase in the proportion of founder directors to total directors, the breadth of outreach decreases by 4.40% (i.e., a decrease of 3522 borrowers from the loan portfolio) and the depth of outreach decreases by 3.89% (i.e., an increase in the average loan balance of BDT373.68 or US $4.99) relative to the mean value. 1 An increase in the average loan balance suggests that the most economically underprivileged citizens receive less focus, thereby undermining the social mission of MFIs.
Our study also demonstrates a negative effect of family members' dominance of the board.For instance, for a one-standard-deviation increase in the proportion of founder directors with ties with family board members to total directors, the breadth of outreach decreases by 3.52% (i.e., 4254 fewer borrowers) and an inverse (adverse) relationship with the depth of outreach is found, whereby the average loan balance increases by BDT467.19 (or US$6.23) relative to the mean value.The results remain robust after controlling for endogeneity and using a lagged regression model, alternative measures of key variables, and seemingly unrelated regression (SUR) analysis.
The findings contribute to the literature in several ways.First, our findings provide insights into how MFIs, a particular kind of social enterprise, manage a dual objective.Our findings suggest that, when confronted with a dual objective, founder directors may "trade off" social outcomes in favor of economic outcomes and therefore adversely affect MFIs' social performance.Second, our study extends the literature by exploring founders' ties to family members on the board.Like founder directors, we find that their ties with family members on the board have an adverse impact on MFIs' social performance.Lastly, our study has implications for policy makers.Specifically, policy makers overseeing the microfinance sector may need to limit the presence of MFI founders and ties to family members on the board.

| INSTITUTIONAL SETTING: MICROFINANCE IN BANGLADESH
The modern version of the microfinance program was established in 1976 in Bangladesh by Muhammad Yunus.Microfinance in Bangladesh is pioneering because, after the success of the Grameen Bank's microcredit model, many new players entered the market and the microfinance movement spread worldwide (Yunus, 2013).Modifications to the core principles of the Grameen Bank took place to meet the market demand and provide better services to the world's poorest people (Cull et al., 2009;Mia et al., 2017).The microfinance market in Bangladesh is now the world's second largest (after India) in terms of clients served, with at least 32 million borrowers (Microcredit Regulatory Authority [MRA], 2017).
With respect to the microfinance sector, MFIs in Bangladesh have specific requirements regarding formation, regulatory requirements, and corporate governance practices.For example, the MRA established the "MRA Rules 2010" for the licensing and operation of MFIs operating in Bangladesh.Regarding corporate governance practices, the boards of MFIs need to comprise a minimum of 5 and a maximum of 10 members and should include female members.The rules also state that an individual should not be a board member for more than three consecutive terms and that the total number of new members should be lower than the total number of board members for any term.

| Theoretical background and literature
In general, the perspectives that address trade-off problems consider circumstances that involve choosing between two options that effectively balance the costs and benefits.For example, as far back as the 1950s, some scholars raised concerns that firms attempting to engage in social responsibilities were diverting their attention away from profit-maximizing activities or the overarching purpose of the firm (Levitt, 1958).In effect, there is a view that time, attention, and resources committed to social responsibilities can result in suboptimal efforts to maximize financial profits (or vice versa), resulting in a trade-off (Berens et al., 2007;Krishnamurti et al., 2021;Van der Byl & Slawinski, 2015;Z. Wang et al., 2018).Specifically, relative to financial objectives, social objectives are believed to be voluntary, yet engagement in voluntary objectives is not cost free (Barnett, 2007;Brammer & Pavelin, 2008;Friedman, 1970;McWilliams & Siegel, 2001).Friedman (1970) posited that firms' only responsibility is to maximize profits and that social responsibility is, in effect, a corporate tax.Others, such as Galbreath (2017), have suggested that engaging in social responsibility places an increased burden on alreadyscarce resources.Hence, managers are believed to face trade-offs in terms of the extent to which they attempt to maximize social outcomes while also optimizing financial returns (Hahn et al., 2014;Jensen, 2002;Krishnamurti et al., 2021;Lynn, 2021).While it has been recognized that MFIs are social enterprises and operate differently from commercial firms (Hermes et al., 2011), we submit nonetheless that they are likely to face potential trade-offs in fulfilling their dual mission because of bounded rationality and satisficing behavior (Cyert & March, 1963).
Research into the trade-off dilemma has presented some interesting findings.For example, Galbreath (2017) found that the presence of insiders on boards of directors is negatively associated with social performance, implying that they are likely to favor financial results.Such a perspective was confirmed by Graham et al. (2005), who reported that over 80% of corporate executives (insiders) whom they surveyed said that they would decrease voluntary projects, such as those related to social responsibility, to meet quarterly or annual financial earnings targets.Using a sample of 63 MFIs, Mori et al. (2015) established that founder directors have an insignificant effect on social performance in the short term.In a related study, Randøy et al. (2015) concluded that MFI entrepreneurial CEOs have a positive association with both financial and social performance.However, executives such as CEOs and members of boards of directors can have very different roles and impacts, as we explain below.
In the first instance, CEOs are full-time employees who have overall responsibility for outcomes and are dependent upon a firm for their income.If they are unable to deliver on outcomes, the board has the authority to fire them (Hillman & Dalziel, 2003).Boards of directors, however, are an elite decision-making group that is not largely involved in the day-to-day activities of firms.They are generally not full-time employees of the focal firm; rather, they are concerned with the overall strategy, they make major policy decisions, and they oversee and monitor the management, including the performance of the CEO (Hillman & Dalziel, 2003). 2 With boards occupying such an elite position, power structures can emerge in which board members assume dominant roles and exert considerable influence to protect their own agendas and interests (Finkelstein, 1992;Finkelstein & Mooney, 2003;Golden & Zajac, 2001).MFI founder directors and their family members on the board may seek to influence strategies and decisions that favor their interests, such as financial interests, as already noted (Randøy & Goel, 2003). 3

| Hypotheses
The extant literature has suggested that founding members on the board can have a positive influence on the performance of firms (Chen et al., 2012;He, 2008).Founders are the initial architects of a firm's structure and of its direction (Gimeno et al., 1997;Nelson, 2003).They have a strong psychological attachment to the firm (Gimeno et al., 1997), and they imprint their beliefs on the business (e.g., Baron et al., 1999;He, 2008;Nelson, 2003).Therefore, their involvement in the board is believed to create positive benefits.
Others have argued, however, that founding members may have an adverse effect because they can interfere with decision making to benefit themselves or their friends or their extended stay on the board can provoke them to engage in opportunistic behavior (Anderson et al., 2003; Anderson & Reeb, 2003; Randøy &   Goel, 2003). 4  In the case of MFIs, founders on the board are likely to have high levels of power and political persuasion (Finkelstein, 1992;Finkelstein & Mooney, 2003;Golden & Zajac, 2001).As MFIs generally have limited access to the capital market, they cannot easily secure funds from public sources.Hence, at the inception of an MFI, a key source of capital is founder-provided funds.In such circumstances, MFI founding directors can take substantial equity stakes and act as block holders.As a result, founding directors tend to play an active role in board functioning and decision making (e.g., Chen et al., 2012).Hence, founder directors' influence is believed to be stronger than that of the other members of the board.However, MFIs must account for a double bottom-line objective, and we believe that this can cause the attention and focus of founder directors to become diffused.Specifically, within the microfinance context, we expect founders' presence on the board to have an adverse influence on social performance based on the following arguments.
First, founders are considered to be the most influential board members of MFIs as they define the goals of the firm and design the structure of the organization (Gimeno et al., 1997).Hence, they possess a strong psychological attachment to economic outcomes because they are intrinsically motivated to see their financial investment generate a return (Demsetz & Lehn, 1985;Li & Srinivasan, 2011).Further, MFI founder directors have a strong economic interest in the firm as they provide funds during inception, possibly the only source of capital for the MFI at that stage.These founder directors act as block holders since most MFIs do not generally issue shares publicly for the establishment of the firm.Hence, MFI founder directors have a high stake in the firm's economic performance as they are the residual claimants for it (Chen et al., 2012;Fama & Jensen, 1983).This can affect the extent to which social objectives are met at their highest levels.In other words, when financial outcomes absorb founder directors' attention, board deliberations and decisions can ultimately be skewed and trade-offs that favor economic over social outcomes can be made.
Second, as additional funding is needed, founder directors create a secure network through their personal relationships and interactions with potential fund providers (Ben-Ner & Van Hoomissen, 1994).
However, with additional funding comes the pressure to repay this capital in a timely manner to maintain continuity with the providers.
To ensure the timely repayment of these funds, we argue that founding board members are likely to place greater emphasis on the financial performance of MFIs, which may have an adverse effect on their social performance.This is because founder directors will favor financial performance by reducing operating costs (Hermes et al., 2011).
The average loan size per borrower is small for MFIs, which results in high operating costs per dollar of credit and difficult or unpredictable repayment collection.Therefore, if founder board members place a greater priority on financial outcomes, they may direct the MFI to provide a larger amount of credit per client.To achieve this, an MFI tends to attract clients with more collateral or those who are considered "wealthier" by relative standards and can more readily repay their loans on time.Attracting these clients has the advantage of lowering the cost of credit per dollar (Hermes et al., 2011).In this way, providing a larger amount of credit reduces the operating costs and increases the economic performance of MFIs.
Nonetheless, we argue that providing a larger average loan size has an adverse impact on a firm's social outreach.This is because the poorest and lowest income earners have a limited ability to use larger loan amounts and find it challenging to manage the repayment of a larger loan amount (Collins et al., 2009).Thus, providing larger loan amounts to economically disadvantaged and vulnerable borrowers may result in borrower over-indebtedness and loan defaults (Guha & Chowdhury, 2013).As MFIs provide credit without formal collateral and the methods of enforcing contracts are informal (Collins et al., 2009;Morduch, 1999), MFIs have difficulty with enforcement against defaulted clients to ensure the repayment of credit in a timely manner.This scenario complicates lending practices for MFIs because, if relatively "wealthier" or more financially stable borrowers are served to reduce loan costs, the social objective can be undermined.
We expect founder directors to emphasize economic outcomes through the provision of larger amounts of credit, which in turn reduces the possibility to reach poorer borrowers (Cull et al., 2007).
Such practices result in a trade-off and are likely to undermine social goals.Therefore, the presence of founding directors on the board is predicted to have an adverse impact on MFIs' social performance.

More formally:
Hypothesis 1. Founding members on the board reduce MFIs' social performance.
Our second hypothesis is based on a classic debate within the corporate governance literature revolving around board of director composition (Fama & Jensen, 1983;Finegold et al., 2007;Galbreath, 2017;Jensen & Meckling, 1976).The most common recommendation is for board independence or for boards to have a majority of independent directors who have no current or previous ties to a firm and no shareholding positions (Finegold et al., 2007).
Independence is prized because managers are believed to be opportunistic and need to be monitored to ensure that resources are not misappropriated (Fama & Jensen, 1983;Jensen & Meckling, 1976).
Nevertheless, boards continue to consist of a variety of directors with various degrees of independence from managers (Galbreath, 2017).
Our interest is in members of the boards of MFIs who are related to the founder (founder family ties) (Mori et al., 2015).Specifically, Mori et al. (2015) posited that founding MFI board members have a positive impact on social performance.However, they found an insignificant effect.We note that Mori et al. (2015) did not account for members of the board who might be part of the founder's family, even though previous research has suggested that examining such board members is of value (Gomez-Mejia et al., 2001;Thomsen & Pedersen, 2000).We build on their study to offer an alternative perspective that suggests an adverse relationship between founder family ties on the board and MFIs' social performance.
First, members of the boards of MFIs who are part of the founder's family (i.e., family ties) have characteristics of family ownership and family control as well as having both financial and physical investments in the firm (Randøy & Goel, 2003).Given the small capital base of MFIs and minimal shareholder involvement, the family members of founders not only provide funds but also are known to have a high propensity for risk taking (Randøy & Goel, 2003).Investing funds in the MFI is an exercise in which returns are put at risk; therefore, the expectation is that the founder's family members on the board will emphasize the firm's economic outcomes and financial sustainability (Randøy & Goel, 2003).While we certainly do not discount the interest of founder family members in the social mission, we argue that, owing to the investment of family funds, they are highly motivated to monitor the management to ensure that their economic interests are not misappropriated but rather maximized (Anderson et al., 2003;Anderson & Reeb, 2003).This effort is likely to divert attention from broader social outreach, creating a trade-off that results in suboptimal social outcomes for the MFI.
Second, founding directors work to reduce the costs per dollar of credit provided while providing larger loan amounts to more financially stable borrowers, ensuring that their personally invested capital is not at risk.When the founding director has a family relationship with other member(s) of the board, those members may come together not only to invest more funds but also to secure positive financial outcomes.Hence, their disproportionate focus on financial outcomes is likely to jeopardize MFIs' social outreach.Further, founding directors use their personal and family networks as well as their family assets and reputations as collateral to attract external funds (Barontini & Caprio, 2006;Patel & Cooper, 2014).Here, they often rely on their family connections.Founding directors and their family members on the board therefore place a higher priority on financial performance to ensure that the MFI can repay the external funds.
However, such a focus on financial sustainability may, in turn, result in MFIs reducing their operating costs by serving more stable clients with larger loan sizes (Hermes et al., 2011).As this occurs, there is a loss of the more economically disadvantaged borrowers at the bottom of the pyramid, lowering the number of total borrowers from the loan portfolio and hence undermining the social mission.Given these perspectives, we argue that members of the board who have family ties to the founder have an adverse effect on both the breadth and depth of social outreach.Therefore, we develop the following hypothesis: Hypothesis 2. Members of the board who are tied to the founder director through a family connection reduce MFIs' social performance.

| Sample and data sources
The data for this study come from MFIs operating in Bangladesh.
Drawn from the Microcredit Regulatory Authority (MRA), our study uses data from 735 MFIs with 5531 firm-year observations. 5Performance data are collected from yearly statistical publications by the Management Information Systems (MIS) department of the MRA.
Data on the boards of the MFIs are hand collected from the MIS department and the licensing department as MFIs have to supply relevant information about directors to the MRA to obtain a license to operate. 6The sample covers a period of 11 years (from 2007 to 2017). 7The dataset in the panel is unbalanced as some of the MFIs are newly established and have only been licensed in recent years and some have ceased to operate at different times in the study period.

Number of active borrowers (NAB Log)
The breadth of outreach is measured using the natural logarithm of the total number of active borrowers (NAB Log) (Hartarska, 2005;Hossain et al., 2020).Borrowers include those individuals who currently have an outstanding loan balance with the MFI or are responsible for paying any portion of the loan or interest included in the gross loan portfolio (Hartarska, 2005;Mersland & Strøm, 2009).The larger the NAB Log, the greater the social orientation of the MFI.We expect that an MFI will be more socially oriented, reaching more clients, when the breadth of outreach (NAB Log) is larger.

Average loan balance per borrower (ALB Log)
The depth of outreach is measured as the natural logarithm of the average loan balance per borrower (ALB Log) (Mersland & Strøm, 2009).The smaller the loan provided, the deeper the outreach to the poorest and most socio-economically disadvantaged people (Collins et al., 2009).This idea is based on the argument that poor people have limited capacity to use larger loans and virtually no means to provide collateral (Collins et al., 2009).Their ability to repay a larger loan is limited due to their small and irregular incomes.Relative to the poorest of borrowers, "wealthier" borrowers are generally provided with larger loans, an action that appears counter to the ultimate aim of MFIs' social mission (Hossain et al., 2020).
However, the measures discussed above are not free from criticism as they may not fully capture MFIs' social performance.For example, the number of active borrowers does not indicate how active the clients are (Copestake, 2007).Further, the average loan balance per borrower does not provide information about the income and wealth level of microfinance borrowers (Quayes, 2012).
Moreover, this measure may be distorted by changes in the exchange rate, inflation, or a shift of MFIs toward richer clients (Copestake, 2007).In addition, the average loan amount may be distorted as devoted borrowers become wealthier over time and because of the accumulation of a few very big loans with many little ones (Bibi et al., 2018).Finally, the number of borrowers and the outstanding loans per borrower consider only the lending side of MFI activities and may reflect their aggressiveness (e.g., providing multiple loans to the same borrowers and overlapping borrowers), among other factors.
Importantly, in many countries, MFIs provide micro savings facilities, suggesting that MFIs' activities are not only tied to credit.The microfinance literature has argued that poor people want to save from their tiny income, through either commitment savings (Ashraf et al., 2010) or voluntary savings (Morduch, 1999), as a precaution against future uncertainties and unwanted economic shocks (Collins et al., 2009) or to use accumulated savings as collateral to obtain credit (Armendáriz & Morduch, 2010).Following the literature, we argue that poor people at the bottom of the pyramid have the smallest of incomes, making them unattractive to the banking system.
Hence, they are unlikely to be able to open savings accounts with commercial banks and remain excluded from formal banking channels.

Number of saving clients (NSAV Log)
We use the log of number of saving clients (NSAV Log) as an additional measure of the breadth of outreach.We argue that, with their savings products, MFIs can reach more clients, particularly those excluded from the formal banking channels.Therefore, a higher value of NSAV Log suggests a greater breadth of outreach of MFIs.

Average savings balance per saver (ASB Log)
We use the log of the average savings balance per saver (ASB Log) as an additional measure of the depth of outreach.We argue that MFIs provide micro savings facilities to the extremely poor by enabling their clients to hold very small amounts in saving accounts.Therefore, the smaller the ASB Log, the deeper the outreach of MFIs.

Return on equity (ROE)
Because we explore the potential trade-offs between economic and social outcomes and the extent to which attention may be divided with respect to the dual purpose of MFIs, we include financial performance as a reference point against our focus on social performance.
Because founder directors and their family members on the board have ownership and/or equity positions in an MFI, we use return on equity (ROE) as the measure of financial performance (Strøm et al., 2014).ROE is calculated as the net operating income after tax scaled by the average total equity.As an alternative measure of financial performance, following prior research (Cull et al., 2007;Mersland & Strøm, 2009), we use the yield on the gross loan portfolio (Yield).The portfolio yield is calculated as the net interest income scaled by the average loan outstanding.

| Explanatory variables
To test the first hypothesis, we use the presence of founder directors on the board as the key explanatory variable (Anderson et al., 2003;Nelson, 2003).We define founding directors as those who have resided on the board of the MFI since inception (T.Wang & Song, 2016).In our main analysis, following prior studies (e.g., Kroll et al., 2007;T. Wang & Song, 2016), we use two variants of this variable: the natural logarithm of one plus the total number of founders on the board (Founder Log) and the proportion of founder directors on the board, which is calculated as the number of founder directors scaled by the total number of directors on the board (Founder Prop).In the robustness analysis, we use the total number of founding directors on the board (Founder Total) as the key explanatory variable.
To test the second hypothesis, our study considers founder members with family ties as the key explanatory variable (Anderson et al., 2003;Anderson & Reeb, 2003;Patel & Cooper, 2014).Following Bennedsen et al. (2008), if a founder member on the board has a family relationship (e.g., sibling, child, spouse, or parent) with another board member, they are considered to be founders with a family tie.
To identify family relationships, we check the curriculum vitae of each of the directors.As the MRA has made it mandatory for MFIs to disclose the kinship (any of the following: parent, siblings' child, or spouse) of the directors with their name and the level of kinship, we utilize this opportunity to trace founder directors' family relationships with other director(s) on the board accurately.Following the extant literature (e.g., Kroll et al., 2007;T. Wang & Song, 2016), our study uses two alternative measures of this variable as the explanatory variable: the natural logarithm of one plus the total number of founders with family ties on the board (Family Log) and the proportion of founder directors with family ties on the board, which is calculated as the number of founder directors with family ties scaled by the total number of directors on the board (Family Prop).In the robustness analysis, we also use the total number of founding directors with family ties on the board (Family Total) as the key explanatory variable.

| Control variables
Several variables are included to control for potential confounding effects.Based on the previous literature (e.g., Bennedsen et al., 2008;Dalton et al., 1998), we include the log of the total board size (Board Size) as it may influence the strategic decision making and the effectiveness of governance mechanisms.Based on the extant literature (Adams et al., 2010;He, 2008;Xu et al., 2015), we also include the log of the number of independent directors on the board (Independence).
Given that MFIs in Bangladesh are required to have women on their board, we control for the log of the number of female directors on the board (Female Directors).Because a powerful CEO can exert more influence on the decision making of the board of directors, we include CEO Power as another governance variable, which is a dummy variable that takes the value of 1 if the CEO also acts as the chair of the board and 0 otherwise (Galema et al., 2012).
We also control for firm size since the scale of operation and the performance of MFIs differ depending on their size.Following prior microfinance studies (e.g., Cuéllar-Fernández et al., 2016;Malikov & Hartarska, 2018), we use the natural logarithm of the gross loan portfolio as a proxy for firm size.Following Hossain et al. (2020), we also include the Operating Expense ratio, measured as the operating expenses over the total assets.We include the debt-to-equity ratio (Debt-Equity) to control for the effect of MFIs' risk profile and capital structure on their performance (Chen et al., 2012).Finally, following the MFI literature (e.g., Ahlin et al., 2011), our study includes loan loss provision expenses (Provision Expense) to control for its effect on MFI performance.We measure Provision Expense as the fraction of the loan portfolio retained for bad loans over a year (Ahlin et al., 2011).All the variable descriptions and measurements are presented in Table 1.

| Estimation method
Following the prior literature (e.g., Randøy & Goel, 2003;Xu et al., 2015), we use the following pooled ordinary least squares (OLS) regression models to test our hypotheses 8 : where the subscript i denotes the individual MFI (i = 1, 2, …, 735), t refers to the time in years (t = 2007, 2008, …, 2017), and β i , γ i , and δ i are the vectors of the coefficients for the main explanatory variables, governance variables, and firm characteristic variables, respectively.Moreover, Y d refers to year effects and ε i,t is the idiosyncratic error term.The dependent variables (P it ) are the social performance of the firm (NAB Log, ALB Log, NSAV Log, and ASB Log).We also include financial performance (ROE) as a reference point, as noted above.

| Descriptive statistics, univariate tests, and correlation analysis
Panel A of

NAB Log
Breadth of outreach, measured as the number of active borrowers (natural logarithm) who have an outstanding loan balance at the MFI or have yet to pay the unpaid portion of the credit to the MFI.The larger the NAB Log, the greater the breadth of outreach.

NSAV Log
Breadth of outreach, measured as the log of the number of savings members of the MFIs.The larger the number, the greater the breadth of outreach.

ALB Log
Depth of outreach, measured as the average loan balance per borrower (natural logarithm) in Bangladeshi currency (BDT).The smaller the average loan size per borrower, the deeper the outreach to poor borrowers.

ASB Log
Depth of outreach, measured as the natural logarithm of average savings balance (in BDT) per member.The smaller the number, the deeper the outreach to poor savers.

ROE
Return on equity = net income after tax/total equity.ROE serves as the financial performance measure and a reference point given our focus on social performance.

Explanatory variables
Founder Log Natural log of (1 + total number of founder directors).

Founder Prop
Proportion of founder directors on the board (i.e., total founder directors/total number of directors).

Family Log
Founder directors with a family tie = log (1 + total founders with a family tie).

Family Prop
Proportion of founder directors with a family tie (i.e., total founders with a family tie/total number of directors on board).

Board Size
Total number of board of directors (logarithmic).

Independence
Log of the number of independent directors on the board.

Female Directors
Log of the number of female directors on the board.

CEO Power
CEO Power is a dummy variable that equals 1 if the CEO and chair are the same person and 0 otherwise.

Firm Size
Size of the firm, measured as the natural logarithm of the total loan portfolio in Bangladeshi taka (BDT).

Debt-Equity
Total debt (short-term and long-term debts) to total equity.Equity is the residual interest of the assets after payment of all obligations.

Operating Expense
Operating expenses to total assets.

Provision Expense
Loan loss provision expenses.The proportion of the loan amount set aside as the allowance for the loan losses to cover the costs of loan that the MFI is not expected to recover from the default borrowers.
Variables used in the sensitivity analysis

Founder Total
Total number of founder directors on the board.

Family Total
Total number of founder directors who have a family tie with other directors on the board.

Yield
Portfolio yield.Calculated as the financial revenue from the loan portfolio scaled by the average gross loan portfolio.
perspective.In addition, the mean (median) value of the ratio of founding members with a family tie to total members on the board (Family Prop) is 10.5% (0.00%).Overall, these univariate test statistics support our prediction of an adverse impact of founder directors and their family governance on social performance.Overall, these findings suggest that the higher the level of founding directors on the board, the more the MFIs provide a larger average loan to their borrowers, which in effect undermines (or exerts a negative impact on) their outreach to the poorest people.This is because a larger loan size means that "wealthier" clients are likely to receive funds at the expense of the poorest of borrowers (Collins et al., 2009).

Panel B of
Interestingly, in Columns ( 9) and ( 10 T A B L E 3 Correlation matrix: The Pearson pairwise correlation. (2) Columns ( 5)-( 8) present the results regarding the depth of outreach, as captured by ALB Log and ASB Log.In Columns ( 5) and ( 7), Family Log has a significant positive relationship with ALB Log (β 1 = 0.093, p < .01)and ASB Log (β 1 = 0.084, p < .05).We obtain similar results when using Family Prop as a proxy for family ties in Columns ( 6) and ( 8).Economically, the coefficients in Columns ( 5) and ( 7) suggest that, for a one-standard-deviation increase in Family Log, ALB Log increases by 5.07% (i.e., an increase in the average loan balance of BDT475.46 or US$6.34 relative to the mean) and ASB Log increases by 4.58% (i.e., an increase in the average saving balance per member of BDT129.11 or US$1.72 relative to the mean).These results indicate that, with an increase in founding directors' family ties on the board, the loan size to borrowers and the saving balance per member increase, suggesting a shift in priority toward larger loans and savings.Overall, these findings support our Hypothesis 2 that family ties have an adverse impact on the social mission of MFIs.
Finally, in Columns ( 9) and ( 10), we find that founder directors' family ties have an insignificant association with MFIs' financial performance (ROE).One possible explanation for this insignificant finding is that board members with family ties within the MFI channel their economic interests or focus through the founder director.In effect, persons with family ties to the MFI recognize that founder directors have stronger power or political influence over MFIs than other board members and hence exploit family ties to expropriate personal and family interests, compromising the MFIs' financial performance (Finkelstein, 1992). 10,11

| Two-stage least squares (2SLS) regressions
Although the baseline results suggest an adverse impact of founder directors and their family ties on the social performance of MFIs, these results may be biased if endogeneity is present in the estimations (Wooldridge, 2016).For example, our estimations may omit some variables that are correlated with both social performance and founder directors and their family ties.Moreover, there is a concern that the social performance of MFIs may affect the presence of founder directors and their family ties, resulting in a reverse causality problem.Although exogenous instruments are widely used in instrumental variable regression to account for endogeneity, the difficulty of finding suitable instruments has been well documented (Jiang, 2017).As an alternative, we rely on the Lewbel (2012) heterogeneity-based instrument, which is increasingly used in economics and finance studies (e.g., Broadstock et al., 2018;Caliendo et al., 2017;Gong et al., 2016;Hasan et al., 2022).This technique does not require an external instrument; instead, it exploits heterogeneity in the error term of the first-stage regression to generate instruments from within the existing model.We generate the instruments from the auxiliary equations' residuals and multiply the residuals by each of the generated variables, including the constant in mean-centered form.
Table 5 reports the results.Panel A shows that the effect of founder directors on the depth, outreach, and financial performance remains robust.For example, the coefficient of founder directors is negative and significant for NAB Log (Founder Log = À0.159,Founder Prop = À0.310;both significant at p < .01).Moreover, the coefficient of founder directors is positive and significant (Founder Log = 0.142, Founder Prop = 0.279; both significant at p < .01)for ALB Log.Finally, the coefficient for ROE is positive and significant for Founder Prop (0.07, p < .05).
In Panel B, we find robust evidence of a negative association between founder director family ties on the board and MFIs' social ).We find that our estimation is not subject to underidentification and over-identification bias.The Cragg-Donald Wald F statistic is also larger than the Stock and Yogo (2005) critical values (at 10% maximal IV size), suggesting that our estimated results are not susceptible to weak instruments.Overall, we provide evidence that the main regression results are not driven by endogeneity problems.

| Two-step system generalized method of moments (system GMM)
We further check the potential bias in our baseline results arising from unobserved heterogeneity and reverse causality through system GMM estimation, suggested by Arellano and Bover (1995) and Blundell and Bond (1998).In our estimation, we use a two-step procedure (Arellano & Bond, 1991) and obtain robust standard errors using the Windmeijer (2005) finite sample correction.This procedure removes reverse causality by internally transforming the data through a statistical process whereby the previous value of the variable is subtracted from its present value.Hansen test for orthogonality conditions also meets the requirements, indicating that the instruments used in both the panels are exogenous.

Panel A of
Moreover, the estimated coefficient of the lagged values of social performance variables (NAB Log tÀ1 and ALB Log tÀ1 ) and financial performance (ROE tÀ1 ) are statistically significant in all the columns for both the founder directors (Panel A) and their family ties (Panel B), suggesting that our models do not reject the dynamic nature of the model.
In Panel A of Table 6, we continue to observe a negative and significant impact of founder directors (i.e., Founder Log and Founder Prop) on social performance (i.e., NAB Log, NSAV Log, ALB Log, and ASB Log) but a positive and significant impact of founder directors on economic performance (ROE), corroborating the findings from the main analysis.In Panel B (Table 6), we find that the relationship between family ties and MFIs' social and economic performance remains robust ( p < .05)when using Family Prop as a proxy for family ties.However, this relationship is weaker when Family Log is used as a proxy for family ties.Overall, the system GMM results suggest that the predicted impact of founder directors and their board member family ties on social performance is unlikely to be confounded by unobserved heterogeneity and reverse causality.

| Omitted variable bias test using the impact threshold for a confounding variable (ITCV)
To address the endogeneity concern resulting from omitted variable bias, we estimate the ITCV (Larcker & Rusticus, 2010).This statistical technique assesses the effect of omitted variables on parameter estimates and thereby helps to determine whether confounding variables exist in the data.In the context of our study, ITCV estimates how correlated an omitted variable must be with microfinance performance and founder directors and their board member family ties to invalidate our main result.
Table S3 (Panel A) demonstrates that the threshold value for founder directors (i.e., Founder Log) is much larger than the raw and partial impacts of the control variables, except for firm size, 12 which suggests that our main results are not materially affected by the omission of a confounding variable.Similarly, in Panel B of Table S3, we observe that the size of threshold value for family ties (Family Log) is much larger than the impacts of the controls, implying that our main results are unlikely to be driven by an endogeneity problem arising from omission variable bias.

| Entropy balancing estimates
As a further attempt to address the endogeneity problem, we use the entropy balancing approach in accordance with recent research (e.g., Gounopoulos et al., 2021;Hasan & Uddin, 2022).We divide firm-year data into treatment (high founder directors and high family ties) and control (low founder directors and high family ties) groups based on their median value.We then apply the entropy balancing procedure to match the treatment and control groups.This method corrects for both random and systematic differences between the treatment and control groups to lessen the possibility that design choices might influence our findings (Hainmueller, 2012).
We show the covariate balance and the regression results using the entropy-balanced sample in Table S4.We find that entropy balancing ensures that the treated and control samples have similar mean, variance, and skewness of the variables (see Panels A1 and B1).In the regression results, we continue to see that the findings from our main analysis remain robust in the entropy-balanced sample (see Panels A2 and B2).

| Lagged independent and control variables
One may argue that the use of lagged independent and control variables in the regression analysis is important to alleviate the concern that founder directors and their family ties in the previous year (t À 1) may affect the performance of MFIs in the current year (t).This regression specification also alleviates the concern about reverse causality since the current performance of MFIs is unlikely to drive the previous level of founder directors and their family ties on the board (Hasan et al., 2022;Lin et al., 2018).
We re-estimate the regressions with 1-year-lagged explanatory variables (including controls) to examine the robustness of our results.
Table 7 shows that the coefficients for both founder directors and their family ties remain robust for both the social and financial performance of MFIs, corroborating the findings from the main analysis.This analysis thus provides evidence of the robustness of our findings.

| Alternative measures of the dependent variables
We further show the robustness of our main results using an alternative measure of the explanatory variables and dependent variables.
Specifically, we use the total number of founding directors on the board (Founder Total) and the total number of founding directors with family ties on the board (Family Total) as alternative explanatory variables.In addition, we use the yield on the gross loan portfolio (Yield) as an alternative measure of financial performance (Mersland & Strøm, 2009).In Table 8, we find that the coefficients of Founder Total and Family Total exhibit a sign and significance that are consistent with our main analysis.In addition, we obtain qualitatively similar results when Yield is used as an alternative financial performance measure.

| Additional macroeconomic controls
As stated earlier, compared with the other internal governance variables, the number of founder directors is relatively time invariant.Hence, there may be a concern about whether our dependent variable (i.e., both the breadth and depth of outreach) is affected by timesensitive variables, capturing macroeconomic developments.Following prior studies (e.g., Cull et al., 2014;Hossain et al., 2020;Malikov & Hartarska, 2018) S5.One potential caveat is that macroeconomic controls and year effects may be highly correlated.
To alleviate this potential concern, we separately re-estimate the regression without year effects and using orthogonal regressions.
In both cases, we continue to obtain robust evidence (results untabulated).

| SUR analysis
Because we rely on multiple dependent variables, there is the possibility that the error terms between the equations may be correlated, resulting in biased results.To account for this possibility, we use SUR analysis, which allows for the incorporation of relevant information from each equation into its final estimate for each regression model (Griffiths et al., 1993;McElroy, 1977).Table S6 demonstrates that the results are in line with our main findings.Hence, SUR analysis further demonstrates the robustness of the main findings.

| Additional analysis: Moderating influence of board independence
Agency theorists and corporate governance scholars have long recognized the importance of board independence for proper board T A B L E 9 Further analysis for three different quantiles of independent directors (pooled OLS).functioning and improving firm performance (e.g., Dalton et al., 1998;Finegold et al., 2007;Jensen & Meckling, 1976).To account for this possibility, we conduct an additional analysis that explores the moderating role of independent directors.
We argue that founder directors and their family ties on the board are likely to possess greater insider information and firmspecific expertise about the firm than outside independent directors (Galbreath, 2017;Liu et al., 2015).Moreover, independent directors are likely to have more generic or other specialized knowledge that is not directly tied to the focal MFI.Outside (independent) directors have to incur significant time and costs to acquire firm-specific information and expertise (Maug, 1997).Therefore, a large proportion of independent directors on the board results in a high information asymmetry problem that could increase suboptimal monitoring and advising (Duchin et al., 2010;Liu et al., 2015).Further, the possibility of "free riders" may increase with the presence of a larger proportion of independent directors if these directors believe that their peers may have better firm-specific information and expertise.In such a situation, the benefits associated with a larger proportion of independent directors may decrease (Liu et al., 2015) and the possibility of the controlling block holders (such as founders and their family members on the board) enjoying substantial (private) benefits at the expense of others may increase (Dyck & Zingales, 2004;Nenova, 2003).
We test our conjecture by partitioning the data into three subsamples of MFIs based on the proportion of independent directors.
We repeat the pooled OLS regressions to determine whether there is a significant difference in the relationship between founder directors (as well as family ties) and MFIs' performance for different quantiles of independent directors.In Panel A of Table 9, we find that the relationship between Founder Prop and NAB Log is insignificant and that the coefficient is lower for the first quantile of board independence (Q1) (Column 1).However, the negative coefficient of Founder Prop becomes larger and significant (β = À0.183,p < .10)for the second quantile of board independence (Q2) (Column 2).Interestingly, the negative coefficient for Founder Prop becomes even larger and more significant (β = À0.227,p < .05)for the third quantile of board independence (Q3) (Column 3).A chi-square test of the difference in Founder Prop between the first and third quantiles of board independence is statistically significant (χ 2 = 3.78, p < .05).
We repeat the analysis for ALB Log and report the results in Columns ( 4)-( 6).We find that the coefficient for Founder Prop is statistically significant (p < .05),positive, and larger for the third quantile of board independence than for the first and second quantiles.A chisquare test of the mean difference of the coefficient of Founder Prop between the first and third quantiles is also statistically significant at the 5% level.Finally, in Columns ( 7)-( 9), we find that the coefficient of Founder Prop with respect to the financial performance (ROE) of MFIs is also positive, larger, and significant for the first quantile of board independence.
In Panel B of Table 9, we examine the moderating role of board independence on the relationship between family ties and MFI performance.In Columns ( 1)-( 3), we find that the impact of Family Prop on NAB Log is negative, significant, and more pronounced (β = À0.644,p < .01)for the third quantile of independent directors.Similarly, the impact of Family Prop on ALB Log is positive, significant, and more pronounced (β = 0.609, p < .01)for the third quantile of independent directors (Columns 4-6).Finally, the relationship between Family Prop and ROE (financial performance) is positive, larger, and statistically significant for the first quantile of board independence (β = 0.131, p < .01)(Columns 7-9).A chi-square test confirms that the difference in the coefficient of Family Prop between the first and third quantiles of board independence is statistically significant at the 1% level.
Overall, the results in Table 9 suggest that the adverse effect of founder directors and their family ties on the breadth and depth of outreach is relatively more pronounced in the presence of a larger proportion of independent directors, whereas the positive impact of founder directors and their family ties on financial performance is more pronounced when independent directors' presence is lower.
Such results suggest a potential paradox in that having more independent directors appears to result in less monitoring of social performance because of a potential free-rider problem, while having fewer independent directors may allow founder directors more latitude to ensure that the financial performance is maximized.
We also find a similar and consistent result when we replace Founder Prop with Founder Log and Family Prop with Family Log (untabulated).Moreover, we obtain consistent evidence when NSAV Log, ASB Log, and Yield are used in the analysis (see Table S7).

| DISCUSSION
We investigate whether the presence of founder directors and their ties to family members on the board has any influence on MFIs' social performance.The findings contribute to the literature in the following few ways.First, the findings of our study provide insights into how firms manage a dual objective in the context of social enterprises, namely, MFIs.Responding to calls for alternative approaches to the study of corporate governance (e.g., Buck & Shahrim, 2005;Jackson & Deeg, 2008), we rely on theoretical perspectives, highlighting that attempts to maximize financial and social performance simultaneously can be fraught with difficulty and that trade-offs are a possibility in terms of which outcome receives the most effort or attention.In the special case of microfinance, in which MFIs have a dual objective and, in theory, would work toward maximizing both economic sustainability and social outreach, we question the extent to which such an objective is advanced or suppressed based on the composition of the board of directors.This is because boards of directors are powerful decision makers and provide critical oversight and guidance for firms, affecting their performance (Hillman & Dalziel, 2003).Founder directors have considerable financial stakes in the MFIs that they have founded (Chen et al., 2012), and we argue that their focus on financial returns is never far from their minds and, in fact, likely to take precedence.If so, then the extent to which social outreach is advanced to the greatest extent possible is likely to be undermined.Further, when additional funds are needed to support the MFI, founder directors use their personal contacts and networks to secure that capital.As such funds need to be repaid, pressure tends to fall on the founders such that they seek to ensure that the MFIs' financial results are as strong as possible, undermining the breadth and depth of social outreach.Our results appear to support these postulates.
In addition to founder directors, we explore founders' ties to family members on the board.We argue that, like founder directors, their family members are a source of capital to fund the MFI.Founder family members also use their personal assets and reputation as collateral to attract funds from external sources.Hence, they have an imperative to ensure that financial risks are minimized such that lending costs are reduced and that sourced funds can be readily repaid.We theorize that, as these factors are accounted for, social outreach, in terms of the breadth and depth of loans to the poorest, will be undermined and will undermine the social mission.Our results offer some evidence for these postulates.
In sum, with respect to the theoretical implications, previous research has focused on founder directors and social performance in MFIs (Mori et al., 2015), highlighting the extent to which these founding directors can be considered valuable resources.However, in fact, MFIs have both an economic objective and a social imperative, that is, With respect to empirical contributions, we advance an understanding of family governance's effect on MFI performance.We extend the findings on founder directors through the use of a different sample of MFIs (Bangladesh vs. East Africa) but also through confirmed empirical findings in the posited direction.Our empirical findings also extend the work of Mori et al. (2015), suggesting that founder ties to family members on the board have an adverse effect on social performance.The empirical findings therefore question the value of governance that is "all in the family" in overseeing the MFI dual objective.
Lastly, our study has implications for practitioners and policy makers.A focus in recent years on the microfinance industry and MFIs has revolved around corporate governance (CSFI, 2014(CSFI, , 2016)).
Part of the corporate governance equation includes board of director composition.Notably, our study finds that founder directors and their ties to family members on the board appear to have an adverse effect on social outreach.For policy makers overseeing the microfinance sector, recommending policies that limit the presence of MFI founders and their ties to family members on the board may be a logical step.
For MFIs, we follow the work of Galbreath (2017), who found that providing incentives (i.e., compensation tied to social outcomes) and training (i.e., training on corporate social responsibility) positively shaped the negative relationship between insiders on the board and social outcomes.MFIs may wish to consider similar approaches when founder directors and their family members reside on the board.
Given that many MFIs struggle to become financially self-sufficient and simultaneously achieve the objective of serving the world's poorest people (Labie & Mersland, 2011), attaining good corporate governance and the right board composition is clearly warranted.

| LIMITATIONS AND FUTURE RESEARCH
Our study is not without limitations.First, our study uses data from a single country and is limited to licensed MFI providers.Thus, it does not include MFIs such as non-licensed MFIs and banks.MFIs operating in other countries may have different organizational structures and legal identities, and founders' interests and personal stakes in the MFIs may also vary (Galema et al., 2012).Furthermore, the regulatory require-

| CONCLUSIONS
MFIs, a form of social enterprise, commonly have a dual objective of economic sustainability and social outreach to the world's poor and those at the bottom of the pyramid (Yunus, 2013).However, we assert that maximizing both dimensions can be challenging and that the preference for different corporate objectives could be moderated by who resides on the board of directors.Specifically, we investigated whether the presence of founder directors and their ties to family members on the board has any influence on MFIs' social performance.
To test the hypotheses empirically, we use a large sample of handcollected microfinance data from 735 MFIs (5531 MFI-year observations for 2007-2017) from Bangladesh, which is considered one of the most mature and largest microfinance markets in the world.The multivariate tests suggest that founder directors are negatively associated with social performance.Further, we found that founder directors with ties to family members on the board have a negative association with social performance.The findings have implications for theory, practice, and policy.
able in our settings for several reasons.First, the FFE regression requires a significant variation in the main explanatory variables of interest to generate an unbiased and consistent estimate.However, in our case, the variables of interest (Founder Prop and Family Prop; alternatively, Founder Log and Family Log) do not vary extensively over time as the founders remain in the MFI since its inception.Therefore, the use of FFE may be inappropriate (Wooldridge, 2016).Second, for a large number of firms over limited periods, an FFE regression provides inconsistent estimates (Baltagi, 2005).
assets to measure firm size, one third of the licensed MFIs in our sample have not reported their total assets to the MRA.Therefore, the use of total assets to measure firm size causes a considerable loss of observations (2256).Nonetheless, in Table S2, we obtain qualitatively similar results when log-transformed total assets are used to measure firm size.
performance.For example, the relationship between family ties and NAB Log remains negative and significant (Family Log = À0.152,p < .01;Family Prop = À0.474,p < .05).Similarly, the positive relationship between family ties and ALB Log remains robust (Family Log = 0.125, p < .05;Family Prop = 0.398, p < .05).Importantly, the 2SLS results show that family ties are positively and significantly related to ROE (Family Log = 0.046, p < .10;Family Prop = 0.212, p < .05 , we augment our baseline regression by adding five time-sensitive macroeconomic variables, specifically GDP Growth (%), log-transformed Per Capita Income in current US dollars, Inflation in the economy captured by the consumer price index, Unemployment Rate based on the ILO estimate, and Stock Market Capitalization as the percentage of the GDP.The results in Table S5 (Panel A) demonstrate that the inclusion of these timesensitive macroeconomic variables does not change our baseline results, suggesting a robust relationship between founder directors and MFIs' performance.We obtain a similar robust negative relationship between founder director family ties and MFIs' social performance in Panel B of Table

a
dual purpose.The theoretical implications of our research extend a singular focus (social performance) to a broader one (economic and social performance) and raise the idea that certain board members who have responsibility for the dual purpose of an MFI can expect to confront difficult trade-offs.In this way, our research extends the existing theoretical perspectives of MFI boards-where founder members and their family members on the board are seen as valuable resources-to account for how such board members confront or tackle a dual objective, one in which trade-offs between the economic and social purposes are likely to arise.
ments, governance mechanisms, socio-economic environment, and cultural values may differ among countries; hence, the findings of our study can only be generalized beyond Bangladesh with caution.Future research should therefore test the results with respect to different types of MFIs and in different legal and cultural environments.Second, the founders not only sit on the board but may also occupy top management positions, such as the CEO or CFO.Hence, founder CEOs (followingRandøy et al., 2015) or other top management roles, such as that of CFO, and their ties to family members on the board warrant future study.Lastly, our study is quantitative and lacks the richness of insights that could be provided through qualitative research.Future research could seek to interview founder directors or family members on the board who have ties to the founder to consider in greater detail their views and perspectives on how economic and social outcomes are perceived and managed and decisions made with respect to any trade-offs.
Description and measurement of variables.
Table 2 presents the summary statistics of the variables.The mean and maximum values of the number of active borrowers (NAB Total) are 34,000 and 740,000, which indicate that, except for a few dominant MFIs, most of the MFIs have a low-level client base.They also indicate that the mean NAB Total is skewed to a few dominant MFIs.In the multivariate analysis, we use the log transformation of NAB Total (i.e., NAB Log), which shows a mean (maximum) value of 8.847 (13.51), thus removing the impact of outliers to a large extent.The natural logarithm of the number of savings members of the MFIs (i.e., NSAV Log) exhibits a mean (median) value of 9.092 (8.733).Moreover, the mean and maximum average loan balance per borrower (ALB Total) are BDT9378 and BDT31,000, respectively, implying that MFIs in Bangladesh target the poor strata of the population who are at the bottom of the economic pyramid.When taking a log transformation, the mean (maximum) ALB Log takes the value of 8.993 (10.333).The average return on equity (ROE) is 0.16, substantially higher thanStrøm et al.'s (2014)average.The mean (median) value of the ratio of founding members to total members on the board (Founder Prop) is 39.4% (33.3%), which is slightly higher than Mori et al.'s (2015) average of 33.5% from an East African microfinance T A B L E 1

Table 3
Descriptive statistics and univariate test.
presents the results of the Pearson correlation between the variables.The results show that founder directors (Founder Log) have a negative correlation with NAB Log (À0.273, p < .01),ALBLog(À0.081,p< .01),NSAVLog(À0.268, p < .01),andASBLog(À0.056, p < .01)and a positive correlation (0.028, p < .05)withROE.Furthermore, the family ties of founder directors (Family Log) are negatively correlated with NAB Log (À0.298, p < .01),ALBLog(À0.066,p< .01),NSAVLog(À0.288, p < .01),andASBLog(À0.047, p < .01).We note that the negative and significant correlations of ALB Log and ASB Log with both Founder Log and Family Log are not consistent with our expectation.A possible reason for this inconsistent finding is that the Pearson correlation does not take into account the effects of other controls that may affect ALB Log.Therefore, we also perform partial correlation analysis and report the results in TableS1, of the Online Appendix.We find that, consistent with our expectation, both Founder Log and Family Log are negatively and significantly correlated with NAB Log ( p < .01)andNSAV Log ( p < .05),whilethey are significantly positively correlated with ALB Log ( p < .01)andASB Log ( p < .05). 9 5.2 | Multivariate analysis 5.2.1 | Founder directors and MFIs' performance Panel A of Table 4 presents the regression results for the relationship between founder directors and MFIs' performance.Columns (1)-(4) show the results for the breadth of outreach, as captured by NAB Log and NSAV Log.We find that Founder Log has a negative impact on T A B L E 2 NAB Log (β 1 = À0.092,p< .01)andNSAV Log (β 1 = À0.068,p< .05).We observe a similar negative impact when using Founder Prop as a proxy for founder directors in Columns (2) and (4).Economically, Column (1) indicates that, for a one-standard-deviation increase in Founder Log, NAB Log decreases by 4.78%, which equals a decrease of 1625 borrowers relative to its mean.Similarly, in Column (2), for Founder Prop, the reduction in NAB Log is 4.40% (i.e., a decrease of 1497 borrowers) relative to the mean.Overall, these findings support our Hypothesis 1 that founder directors have an adverse influence on the breadth of outreach of MFIs.coefficient in Column (5) indicates that, for a one-standard-deviation increase in Founder Log, ALB Log increases by 4.26% relative to the mean, which accounts for a BDT399.50(US$5.33)increase in credit size per borrower.Similarly, Column (6) shows a 3.89% increase (which represents BDT364.80 or US$4.87) in ALB Log for a one-standard-deviation increase in Founder Prop.
), financial performance (i.e., ROE) shows a positive association with Founder Log (β 1 = 0.019, p < .10)and Founder Prop (β 1 = 0.050, p < .05).Economically, Column (10) suggests that, for a one-standard-deviation increase in Founder Prop, ROE increases by 7.69% relative to the mean.Overall, this find-Panel B: Mean difference test of the variables (assuming unequal variance) 5.2.2 | Founder directors' family ties on the board and MFIs' performancePanel B of Table4presents the results for the relationship between founder directors' family ties on the board and MFIs' performance.Columns (1)-(4) consider how family ties affect MFIs' breadth of T A B L E 2 (Continued) Note: Panel A shows the descriptive statistics of the variables.The financial performance variables (ROE and Yield) are in decimals.The social performance variables include the natural logarithm of the number of active borrowers (NAB Log), the natural logarithm of the average loan balance per borrower (ALB Log) in Bangladeshi currency (BDT), the natural logarithm of the total number of savers (NSAV Log), and the natural logarithm of the average savings balance per member (ASB Log) in BDT.The main explanatory variables (Founder and Family) are presented in total, ratio, and logarithmic measures.Descriptions of all the variables are in Table1.We winsorize all the continuous variables at 1% and 99%.Panel B shows the univariate test of the dependent and control variables based on the median value of Founder Prop and Family Prop.***Statistically significant at the 1% level.**Statistically significant at the 5% level.*Statistically significant at the 10% level.
Note: This table presents the correlation matrix.Descriptions of all the variables are in Table1.Baseline regression results using ordinary least squares (pooled OLS) method.
**Statistically significant at the 1% level.*Statisticallysignificant at the 5% level.T A B L E 4 In Columns (1) and(3), we find that Family Log has a negative and significant relationship with NAB Log (β 1 = À0.100,p < .01)and NSAV Log (β 1 = À0.061, Note: Panel A presents regression results of the relationship between founder directors and MFIs' performance.Our main variable of interest is Founder Log (i.e., 1 + total founder directors) and Founder Prop (i.e., total founder directors/total directors on board).Panel B presents regression results of the relationship between founder directors' family ties and MFIs' performance.Our main variable of interest is Family Log (i.e., 1 + total founder directors with family ties) and Family Prop (i.e., total founder directors with family ties/total directors on board).In both panels, social performance is proxied by NAB Log (natural logarithm of the number of active borrowers), NSAV Log (the natural logarithm of number of savings members of the MFIs), ALB Log (the natural logarithm of average loan balance per borrower in Bangladeshi currency), and ASB Log (the natural logarithm of average savings balance in Bangladeshi currency per member) and financial performance is proxied by ROE (return on equity).Heteroscedasticity and autocorrelation adjusted (firm clustered) robust standard errors are in parentheses.Descriptions of all the variables are in Table1.***Statisticallysignificant at the 1% level.**Statisticallysignificant at the 5% level.*Statisticallysignificant at the 10% level.outreachmeasured using NAB Log and NSAV Log.
Lewbel (2012)ports the endogeneity tests results using two-stage least squares (2SLS) regressions.We followLewbel (2012)in generating heterogeneity-based instruments and use them as instrumental variables.Heteroscedasticity and autocorrelation adjusted (firm clustered) robust standard errors are in parentheses.Descriptions of all the variables are in Table1.Endogeneity test with two-stage system generalized method of moments (system GMM).Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/corg.12528 by National Health And Medical Research Council, Wiley Online Library on [20/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Table 6 presents the GMM results for founder directors, and Panel B reports the results for founder directors' family ties.The Arellano-Bond first-and second-order correlation tests are in line with the expectations (AR1 is significant; AR2 is insignificant).The T A B L E 5 Endogeneity test: Two-stage least squares regressions using heterogeneity-based instrument.***Statisticallysignificant at the 1% level.**Statisticallysignificant at the 5% level.*Statisticallysignificant at the 10% level.T A B L E 6 This table reports the results of the system GMM estimates.Details of the estimation method are in Section 5.3.2.Robust standard errors are in parentheses.Descriptions of all the variables are in Table 1.***Statistically significant at the 1% level.**Statistically significant at the 5% level.*Statistically significant at the 10% level.
T A B L E 7 Robustness tests: Lagged regression models.A: Founder directors and MFI social performance: Use of lagged independent and control variables (t À 1) This table presents the main regression results using lagged independent and control variables.Heteroscedasticity and autocorrelation adjusted (firm clustered) robust standard errors are in parentheses.Descriptions of all the variables are in Table1.***Statistically significant at the 1% level.**Statistically significant at the 5% level.
*Statistically significant at the 10% level.
T A B L E 8 Robustness tests: Alternative measures of key variables.PanelA presents regression results for the relationship between founder directors and MFIs' performance using alternative founder director (Founder Total) and financial performance (Yield).Panel B presents regression results using alternative family ties (Family Total) and financial performance (Yield).Heteroscedasticity and autocorrelation adjusted (firm clustered) robust standard errors are in parentheses.Descriptions of all the variables are in Table1.