female leadership, performance, and governance in microfinance institutions

Journal of Banking & Finance 42 (2014) 60–75

Contents lists available at ScienceDirect

Journal of Banking & Finance

journal homepage: www.elsevier .com/locate / jbf

Female leadership, performance, and governance in microfinanceinstitutions

http://dx.doi.org/10.1016/j.jbankfin.2014.01.0140378-4266/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author. Tel.: +47 92434333.E-mail address: [email protected] (R. Mersland).

Reidar Øystein Strøm a, Bert D’Espallier b, Roy Mersland c,⇑a Faculty of Social Sciences, Oslo and Akershus University College, Norwayb Faculty of Economics and Management, Hogeschool-Universiteit Brussel, Belgiumc School of Business and Law, University of Agder, Norway

a r t i c l e i n f o

Article history:Received 3 May 2012Accepted 17 January 2014Available online 29 January 2014

JEL classification:G32G34

Keywords:Female leadershipFinancial performanceMicrofinanceMFIsCross-country panel data

a b s t r a c t

This paper investigates the relations between female leadership, firm performance, and corporate gover-nance in a global panel of 329 Microfinance Institutions (MFIs) in 73 countries covering the years 1998–2008. The microfinance industry is particularly suited for studying the impact of female leadership ongovernance and performance because of its mission orientation, its entrepreneurial nature, diverse insti-tutional conditions, and high percentage of female leaders. We find female leadership to be significantlyassociated with larger boards, younger firms, a non-commercial legal status, and more female clientele.Furthermore, we find that a female chief executive officer and a female chairman of the board are posi-tively related to MFI performance, but this result is not driven by improved governance.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction Unlike studies in high-income countries (Smith et al., 2006;

The microfinance institution’s (MFI) purpose or mission is toprovide access to financial services to poor families and small busi-nesses situated mostly in developing and newly industrializedcountries. Microfinance is to a large extent a women’s business. Fe-male borrowers are the MFIs’ largest market, and lending to wo-men is considered one of the main reasons for microfinance’ssuccess (Armendáriz and Morduch, 2010). But microfinance isnot only a business for women it is to a large extent also a businessby women. Interestingly, beside Nobel laureate Muhammad Yunus,several women are industry icons: for example, Pilar Ramirez ofBanco FIE in Bolivia and Ingrid Munro in Jamii Bora in Kenya. Thefemale proportion of top executives and directors in MFIs is high.In our sample the CEO is female in 27% of MFIs, the chair is femalein 23%, and 29% of all board seats are held by women. These pro-portions are much higher than corresponding figures in traditionalfirms. For instance, for their very large sample of U.S. companies,Adams and Ferreira (2009) report that only 8.8% of directors are fe-male. In this paper, we investigate whether female leadership im-proves governance and financial performance in MFIs.

Adams and Ferreira, 2009) that often consider only the role ofdirectors, we address these questions for the chief executive officer(CEO), the chair, as well as board directors. Moreover, our study isnovel because it surveys entrepreneurial firms (MFIs) in emergingmarkets. Our data tells us that the MFIs’ median time in operationis eight years. In eight years, the weight of tradition has not settledin a firm, so that a masculine culture has not yet become ingrained,and the male network has not had time to become established.Thus, the ‘‘glass ceiling’’ (Kanter, 1977) between men at the topand women in jobs below has not had time to set. This createsopportunities for able women to rise in the MFI’s leadership hier-archy. Microfinance is also typically a mission-driven organization(Randøy et al., 2014). Thus, we are able to tell if the leadership’sgender matters for governance and performance in circumstancesthat are different from those usually studied.

The female orientation is often a stated goal in many MFIs. TheMFIs in our sample have indicated whether they prefer to lend towomen. 44% state that they have such a female bias. This femaleattention is evident in supporting international organizations formicrofinance as well. The objective of the Microcredit SummitCampaign, which plays a central role in the promotion of microfi-nance, is ‘‘to ensure that 175 million of the world’s poorest fami-lies, especially women, receive credit for self-employment and

http://crossmark.crossref.org/dialog/?doi=10.1016/j.jbankfin.2014.01.014&domain=pdf

http://dx.doi.org/10.1016/j.jbankfin.2014.01.014

mailto:[email protected]

http://dx.doi.org/10.1016/j.jbankfin.2014.01.014

http://www.sciencedirect.com/science/journal/03784266

http://www.elsevier.com/locate/jbf

1 Becker gives various examples of matching: ‘‘. . .the optimal sorting of more ableworkers and more able firms, more ‘‘modern’’ farms and more able farmers, or moreinformed customers and more honest shopkeepers’’.

R.Ø. Strøm et al. / Journal of Banking & Finance 42 (2014) 60–75 61

other financial and business services’’ [our emphasis](www.microcreditsummit.org).

When many MFIs are gender biased, it becomes interesting tostudy how well female leaders perform. Does a woman bring bettergovernance and better financial performance to the MFIs they run?Mersland and Strøm (2009) argue that because a female CEO is bet-ter able to tap into the local, often female information network, fe-male leaders may design product and procedures that better meetthe female users’ needs. We follow this line of argument here, andextend the analysis to the chair and the board of directors. Ourmain hypothesis is that female leadership has beneficial conse-quences for the MFI’s governance as well as its financial perfor-mance. In gender biased MFIs, female leaders could be bettermatched to the challenges and opportunities that female custom-ers face.

But female leadership is possibly endogenous, that is, specificMFIs may attract female leaders. If these are also good performers,we cannot attribute good financial performance to female leader-ship. At most we can state there is a correlation. We attempt to dis-entangle the reverse causality problem by following the Heckman(1978) dummy endogenous variable method. We also test for sam-ple selectivity bias by the inverse Mill’s ratio test.

The research on female leadership is scant in microfinance.Armendáriz and Morduch (2010) argue that female targeting andfinancial sustainability are perfectly compatible, since female tar-geting within microfinance has often been attributed to increasedefficiency due to higher repayment rates among female borrowers.D’Espallier et al. (2011) confirm that the targeting of women leadsto higher repayment rates in MFIs. Both deal with the customer as-pect. However, our study investigates whether female leadershiphas an impact upon the MFI’s governance and its financial perfor-mance in an industry that to a large extent caters to femalecustomers.

In the general governance literature, only Adams and Ferreira(2009) address both the governance and performance issues re-lated to female management, but limit the study to directors. Theyfind that female directors are ‘‘tougher’’ monitors than men, butalso that a positive effect of female directors on performance isonly detectable for firms with weak governance structures wheninstrumental variables (IV) methodology is employed. Further-more, Smith et al. (2006), and Francoeur et al. (2008) investigatethe relationships between a female CEO and female directors onthe one hand and financial performance on the other. These studiesdo not look into corporate governance issues. All studies cited hereuse data from diverse industries in Western countries. In contrast,we investigate the effects of three leadership types (CEO, Chair andDirectors) upon both corporate governance and financial perfor-mance in an homogeneous industry in many developing countries.

The sample consists of 329 MFIs in 73 countries from 1998 to2008. The data are from rating agencies and cover up to six yearsof data per individual MFI. The sample is drawn from the sameindustry, where MFIs largely follow the same mode of operation,focusing on loans to poor people and small enterprises, grantingsmall loans with a short maturity, and demanding frequent repay-ments (Helms, 2006). Borrowers often have little or no collateral orcredit history. Frequent repayments enable the MFI to quickly as-sess the borrower’s repayment ability. Nevertheless, heterogeneitymay arise due to different firm and country characteristics, in par-ticular attitudes to women. We control for heterogeneity by firmand country background variables. Firm controls include the MFI’sbusiness practice, differences in institutional background, andcommonly used controls such as MFI size, age, and risk. The coun-try controls encompass a set of country variables, and also includeworld regional dummies. It turns out that despite cultural diver-sity, the fraction of female leadership positions is remarkably sim-ilar across countries. By controlling for country differences, our

findings are relevant for corporate governance in other thanemerging markets (Aguilera and Jackson, 2010).

We find that female leadership is negatively related to suchgovernance measures as the number of board meetings, internalaudits, and the separation of the CEO’s and chair’s roles, but posi-tively related to MFI financial performance. This is contrary towhat Adams and Ferreira (2009) find for female directors. Thus,the quality of an MFI’s CEO and chair seems to be more importantfor the MFI’s success than general corporate governance. Countryspecific variables complement these findings, as they are corre-lated with corporate governance but not financial performance.The results are robust to variations in estimating methodology,variable definition, and regression specification.

This article proceeds as follows. Section 2 develops hypothesesfrom former literature on female top executives’ and board mem-bers’ influence on firm performance and corporate governance.Section 3 gives a brief introduction to the microfinance industryand its special focus on women together with data descriptives.Section 4 lays out the estimating methodologies and also definesvariables. Section 5 covers the conditions under which female lead-ership tends to arise, and Section 6 examines the relations betweenfemale leadership and corporate governance. Section 7 deals withthe relations between female leadership and the MFI’s financialperformance. In Section 8 we perform a number of robustnesschecks, and Section 9 presents our conclusions.

2. Gender, governance and performance

Mersland and Strøm (2009) find that a female CEO induces ahigher financial performance in the MFI. They assume this is dueto the female CEO’s better understanding of the market in whichthe MFI operates. This is a matching, or sorting, argument implyingthat an MFI that is matched with a leadership that has the sametraits will perform better. In this case, the ‘‘same traits’’ refers togender, so that for instance an MFI favoring female clients ismatched with female leadership. The underlying theory for thisis the Becker (1973) model for the marriage market.1 Thus, thehypothesis is that female managers and directors will improve theMFI’s governance and financial performance due to the better matchbetween the MFI’s leadership team and its market conditions. Inmicrofinance, Ghatak (2000) shows how the Becker model may beapplied to the matching of good borrowers in a group lendingscheme. In Thomas and Ramaswamy (1996) the matching of leaderswith specific traits and the firm’s strategy increases firmperformance.

The matching hypothesis of female leadership and the MFI con-tains two sub-hypotheses. The first is that female leadership ismore likely to be found in MFIs with a bias towards female custom-ers. The second is that an MFI’s governance and financial perfor-mance improves with female managers and directors. But wehereby encounter two potential endogeneity problems. The firstis the reverse causality case (Hermalin and Weisbach, 1998) whenthe MFI performing financially well attracts a female CEO. We con-trol for reverse causation of female leadership in financial perfor-mance regressions by the Heckman (1978) model for anendogenous dummy variable. The second endogeneity problem issample selectivity, that is, the selection of a female CEO, chair ordirector might be related to the emphasized focus on female cus-tomers i.e. MFIs hire a female leader because most clients are wo-men, not because of their qualifications. We handle this secondendogeneity problem by the inverse Mill’s ratio (IMR) test. In the

http://www.microcreditsummit.org

2 The dataset consists of rating reports from five rating agencies: MicroRate,icrofinanza, Planet Rating, Crisil, and M-Cril. It is an updated version of the dataset

sed by Mersland and Strøm (2009) and variables on gender leadership have beencluded. At the time of collecting the data the rating agencies were supported by theting Fund of The Consultative Group to Assist the Poor (CGAP) and the Interamer-an Development Bank (IDB) and the reports were made publically available at theorld Wide Web. The donor support has now shrunk and the public availability

olicy has changed. Some examples of reports are, however, still available atww.ratinginitiative.org and www.ratingfund2.org or at the websites of the rating

gencies.

62 R.Ø. Strøm et al. / Journal of Banking & Finance 42 (2014) 60–75

gender literature, the Kanter (1977) ‘‘window-dressing’’ or ‘‘token-ism’’ is an argument for the endogeneity of female leadership. Sheproposes that a company’s election of female directors is done inorder to demonstrate commitment to gender-neutral policies.

The dominant approach in gender studies seems to be that bothgovernance and financial performance improves with more wo-men in management and board. Teigen (2000) puts the positivegender effect down to what may be called the underused resourceeffect and the underused diversity resource effect. The total resourcepool of equally able men and women is underused when candi-dates for management and director positions are pulled from thesub-set of men only. The underused diversity resource means thatmanagers and directors are not optimally matched to the condi-tions of their firm. This last effect could be especially importantin microfinance, where the majority of clients are female. Femaleleaders should therefore increase governance quality and financialperformance in MFIs. Beneficial leadership is often seen as anintrinsic quality among women. Thus, Shrader et al. (1997) sum-marize this in an early investigation: ‘‘There is evidence that wo-men are more oriented toward supporting and maintainingrelationships than men. . . . Therefore, as more and more womenassume managerial positions, organizational learning, climate,and performance should improve.’’ Bertrand and Schoar (2003)conjecture that female managers and directors represent a newmanagement style. If this is true, it should only increase the effectson governance and performance. In particular, in the Adams andFerreira (2007) model directors perform two functions, monitoringand advice. Thus, female directors could be better advisors to theCEO than male. This could be especially important in the youngand entrepreneurial MFIs, where growth rates are unusually high.

It appears that most empirical investigations start from the re-source effects, but that the advantages of matching go unnoticed.Furthermore, only Adams and Ferreira (2009) address governanceand financial performance, although predictions are for both is-sues. They find that women on the board generate better gover-nance, but better firm performance follows only in firms withweak overall governance. Moreover, besides the above-mentionedShrader et al. (1997) study, we are aware of only Smith et al. (2006)and Francoeur et al. (2008) who study different management anddirector positions. These find that female managers (e.g. the femaleCEO) improves firm performance, but that female directors areonly weakly or negatively linked to financial performance. Besidesthis, some studies investigate the role of gender in the top manage-ment team. Welbourne et al. (2007) find that short-term and long-term financial performance (Tobin’s Q) improves when women arepart of the top management teams in firms undertaking an initialpublic offering (IPO).

By far the most numerous studies concern the relationship be-tween female directors and financial performance. This has beenaddressed in cross-sectional and panel data studies, in event stud-ies using time series, and in a natural experiment setting. First,cross-sectional and panel data studies show conflicting results.Some studies find a positive relation (Carter et al., 2003; Campbelland Minguez-Vera, 2008), while others detect a negative relation(Shrader et al., 1997; Smith et al., 2006; Rose, 2007; Adams andFerreira, 2009; Bøhren and Strøm, 2010; Carter et al., 2010; Galle-go-Álvarez et al., 2010). Second, the event studies measure stockprice reaction to the announcement of a female director. Here,the results lean towards a positive relationship. Farrell and Hersch(2005) report no wealth effect, while Campbell and Minguez-Vera(2010) and Kang et al. (2010) find positive reactions. Third, Ahernand Dittmar (2012) utilize the natural experiment setting that thequota regulation in Norway from 2003 to 2008 affords, and find anegative relationship. Thus, different methodological approachesyield conflicting results, both within methodological approachand between approaches.

Can conflicting results be due to the measures used? For in-stance, Konrad et al. (2008) forward the ‘‘critical mass’’ hypothesisthat at least three women must be on a board for their femaleadvantages to be realized. The female director effect is possiblymost evident in board committees (Carter et al., 2010). We are ableto test the first possibility, but lack information about the second.Furthermore, it is difficult to accept psychological explanations forsuperior female leadership. In fact, the Alvesson and Billing (2009)comprehensive survey of the psychological literature on manage-ment style reveals no large differences along gender lines.

The conflicting results may also stem from different institu-tional and country heterogeneity (Aguilera and Jackson, 2010). LaPorta et al. (1998) underline differences in law traditions in com-mon law and Roman (Civil) law countries. Terjesen and Singh(2008) find cross-country differences in the fraction of femaledirectors using a 43 country sample. Furthermore, a robust resultin the general board literature is that the more complex the firmis in terms of the size and span of its operations, the more outsidedirectors are recruited (Baker and Gompers, 2003; Boone et al.,2007; Linck et al., 2008). Speckbacher (2008) argues that sincenonprofit firms often have complex objectives, they will typicallyhave larger boards. Thus, we include a number of institutionaland country controls for the regressions.

3. Women in microfinance: data and variable definitions

The microfinance mission is to provide low-income families andsmall businesses access to financial services. MFIs have a doubleobjective, to serve the poor and to do so in a financially sustainableway (Morduch, 1999). Starting as experimental developmentschemes in Asia and Latin America in the 1970s, microfinancehas become a major industry today. More than 3000 MFIs reporttheir numbers to the Microcredit Summit (www.microcreditsum-mit.org), and they provide more than 150 million people with cred-it. More than 100 international funds invest in microfinanceoffering equity, loans, bonds, and collateralized debt obligations(www.mixmarket.org). The industry is young and entrepreneurial,in fact, the median age is 8 years in our sample, 25% are underbanking authorities regulation, and the incorporation ranges fromshareholder ownership (25%) to cooperatives (16%) and non-gov-ernmental organizations (52%).

Our data set is based on rating assessment reports gathered byspecialized rating agencies and encompasses 329 MFIs operating in73 different countries worldwide in the years 1998–2008.2 At eachrating, the raters collect data for the rating year and years immedi-ately preceding. In this way, up to six years of data for an MFI areavailable for the period 1998–2008. The amount of detail varies inthe reports, resulting in different numbers of observations. No data-set is perfectly representative of the microfinance field. In particular,our dataset contains relatively few megasized MFIs, and does notcover the virtually endless numbers of small savings and credit coop-eratives. The former are rated by such agencies as Moody’s and Stan-dard & Poor’s, while the latter are not rated at all. Ratings data are,however, considered among the most representative available forthe microfinance industry (Mersland and Strøm, 2009).

Information in the rating reports are collected during on-site

MuinraicWpwa



http://www.mixmarket.org

http://www.ratinginitiative.org

http://www.ratingfund2.org


visits to the MFI by specialized evaluators working in the ratingagencies. The information is thus not self-reported by the MFIbut is collected by the evaluator and further screened by the ratingcommittee at the rating agency’s main office. The information inrating reports is therefore regarded to be of high quality, and ratingof MFIs is one of the main transparency initiatives in the microfi-nance industry (Beisland et al., 2014). The MFI rating assessmentsare much wider than traditional credit ratings, as they aim to mea-sure the MFIs’ ability to reach their multiple sets of objectives(Beisland and Mersland, 2012). The purpose of rating reports isto present independent information that stakeholders, such aslenders, donors, owners or managers, can use to make informeddecisions. Even if a rating agency argues that its methodology isdifferent from that of other agencies, the core information usedin this study consists of standard indicators that are calculatedsimilarly across the industry and by all rating agencies.

Table 1 provides a detailed description of the main variablesused in our analyses, as well as a number of summary statistics.The table contains definitions of female leadership, financial per-formance, and governance, the set of MFI and country control vari-ables, and social mission and institutional variables that enter asinstruments.

3.1. Female leadership

The table shows the measures for the three female leadership3

categories, CEO, chair, and director. The female director is defined infour different ways in order to accommodate for potential variationsin results when definitions change (Adams and Ferreira, 2009; Carteret al., 2010), and for the Konrad et al. (2008) critical mass of at leastthree female directors measure. Consequently, we use the indicatorvariable as the main female director definition, and try the otherthree in robustness checks. In some cases it has been impossible toascertain the fraction of female directors. When information on thenumber of female directors is missing, we often know the genderof the chair, in which case we are able to construct a binary variableshowing whether women are on the board or not. Both the femaledirector fraction and the binary for female director are used in ourregressions.

3.2. Corporate governance

We choose the number of board meetings, CEO/Chair duality,internal auditors, and board size as governance mechanisms.4

The number of board meetings measures the intensity of the board’swork, and thus, the more meetings, the higher is the monitoringfunction. The variable may be a relevant measure of monitoringintensity in the fast-growing microfinance industry. The number ofboard meetings is close to eight on average with a median of four.This is fewer than Monks and Minow (2008) report, but then MFIsdo often not have sub-committees. Entrepreneurial firms often com-bine the CEO and chair functions. However, governance recommen-dations (Cadbury, 2002; Organization for Economic Co-operationand Development, 2004) warn against such power concentration.CEO/Chair duality is therefore an indication of less monitoring. Aninternal audit linked to the board can give the board independentinformation on goal fulfillment. An internal auditor means moremonitoring. The overview of the literature on audit fees by Hay

3 For most MFIs, the female leadership variables have been collected only once,namely in the year in which the assessment by the rating agency was done. When thisis the case, we assume female leadership variables to be constant over the years inwhich other time-varying information is available.

4 The corporate governance variables have typically been collected only once,namely in the year in which the assessment by the rating agency was done. When thisis the case, we assume governance variables to be constant over the years in whichother time-varying information is available.

et al. (2006) shows that fees are related to the firm’s size, complexity,governance, and independence. The CEO/Chair duality and internalauditor fractions are both low on average. The average board sizeof seven members (median six) seems to be on par with interna-tional experience. Board size has turned out to be hard to categorize.Early studies in Yermack (1996) and Eisenberg et al. (1998) find anegative relation between board size and firm performance. Adamsand Ferreira (2009) and Mersland and Strøm (2009) confirm the neg-ative sign. A possible explanation is that firms lose business oppor-tunities due to longer decision time in larger boards. However,studies of endogeneity of governance mechanisms (Baker and Gom-pers, 2003; Boone et al., 2007; Linck et al., 2008) find that board sizetends to vary with the firm’s size and complexity. Therefore, the signis hard to predict. Furthermore, we note the large dispersion in thedata on the number of board meetings and board size. In a youngindustry experiments with the best governance setup are to be ex-pected. Adams and Ferreira (2009) use board attendance and CEOturnover as measures of governance quality. We believe our proxiesmore directly measure relevant governance issues in microfinance.Thus, we expect that board meetings, internal audit, and board sizeincreases with female leadership, and that the CEO/Chair duality isless prevalent with female leadership.

3.3. Financial performance

We use return on assets (ROA) and return on equity (ROE) asfinancial performance measures together with operational andfinancial self-sufficiency (OSS and FSS, respectively). Market per-formance measures are impossible since no MFI in our sample islisted. ROA, ROE, and FSS are all taken directly from the raters’ re-ports. OSS is defined as portfolio revenues divided by operationalexpenses. This measure is free from bias resulting from differentcapital structure, access to subsidized funding and possible differ-ences in default policies in the MFI. FSS is an adjusted measure ofOSS taking into account financial costs, default costs, subsidies andother MFI specific adjustments. OSS and FSS are commonly usedmetrics in MFI evaluations (Armendáriz and Morduch, 2010, p.243). Table 1 shows that, on the whole, microfinance is not a lucra-tive business. On average, both ROA and ROE are low and FSS is lessthan 1.0. We expect financial performance to increase with femaleleadership.

3.4. Social mission

The social mission variables encompass the MFI’s gender bias, itsrurality bias, and the average loan. Table 1 reveals that the averageloan is USD 734 and the median is USD 351. These are often usedmeasures of poverty outreach in the microfinance literature (Schre-iner, 2002). In 44% of cases, the rating agencies attest that MFIs havea female gender bias in their lending practices. Unfortunately, manyMFIs do not report their percentage of female customers. Thosewho do, however, show a percentage in the 70–75% range, whichis close to that reported in Cull et al. (2009). Thus, the female frac-tion is high in MFIs on both the customer and the leadership sides.Especially when the MFI has a gender bias, it is important for theMFI to recruit a leadership team that understands its chosen mar-ket. But given that women are often at the lowest income level indeveloping countries, the advantage of having women in leadershippositions should also carry over to the rural/urban measure andaverage loan.

3.5. Institutions

The institutional variables are the MFI’s regulatory status, itsfounding status (internationally initiated or not), the competitionin its product market, and its ownership type. MFIs are diversely

Table 1Summary statistics for the variables used in this study. The number of observations (N) is stated in terms of firms for categorical variables and firm-years for continuous variables.

N Mean Median St. dev. Min. Max.

Female leadershipFemale CEO Binary: 1 if female CEO 329 0.27 0.00 0.44 0.00 1.00Female chair Binary: 1 if female chair 238 0.23 0.00 0.42 0.00 1.00Female director Binary: 1 if one or more female directors 167 0.75 1.00 0.44 0.00 1.00# Female directors Number of female directors 148 1.69 1.00 1.76 0.00 9.00Female dir. fraction Female directors as fraction of all directors 148 0.29 0.20 0.28 0.00 1.00Dumfemdir P 3 Binary: 1 if # female directors is 3 or more 148 0.23 0.00 0.42 0.00 1.00

Corporate governanceMeetings Number of annual board meetings 197 7.64 4.00 7.51 1.00 52.00CEO duality Binary: 1 if CEO and chair are same person 304 0.15 0.00 0.36 0.00 1.00Internal audit Binary: 1 if MFI has an internal auditor 280 0.39 0.00 0.49 0.00 1.00Board size Number of directors 303 7.04 6.00 3.48 1.00 23.00

Financial performanceROA Return on assets 1075 0.010 0.022 0.11 �0.89 0.34ROE Return on equity 987 0.053 0.070 0.27 �0.91 0.56OSS Operational self-sufficiency 1061 1.56 1.49 0.68 0.004 3.00FSS Financial self-sufficiency 667 0.96 0.96 0.37 0.06 3.00

General characteristicsTA Total assets (USD 1000) 1125 5365 2227 9856 19.288 144,000TLP Total loan portfolio (USD 1000) 1140 3786 1665 5801 3.425 59,700PaR30 Fraction of loan portfolio 30 days overdue 1043 0.066 0.034 0.094 0.000 0.973Average loan TLP divided by credit clients 1037 734 351 1330 1.00 24.589Age Number of years in operation 1081 9.36 8.00 7.23 0.00 79.00Rural Binary: 1 if emphasized area is rural 319 0.26 0.00 0.43 0.00 1.00Urban Binary: 1 if emphasized area is urban 319 0.30 0.00 0.46 0.00 1.00Gender bias Binary: 1 if priority is on female clients 324 0.44 0.00 0.49 0.00 1.00Regulated Binary: 1 if regulated by banking authority 322 0.25 0.00 0.43 0.00 1.00Internationally initiated Binary: 1 if internationally initiated 327 0.36 0.00 0.48 0.00 1.00NGO Binary: 1 if type is NGO 329 0.52 1.00 0.50 0.00 1.00COOP Binary: 1 if type is cooperative 329 0.16 0.00 0.36 0.00 1.00Competition index Index from no comp. (1) to high comp. (7) 303 4.26 4.00 1.54 1.00 7.00


incorporated, covering ordinary shareholder-owned firms, mutu-ally held institutions (COOP), non-governmental organizations5

(NGOs), and state banks. We use two indicator variables, one indicat-ing the MFI is an NGO, the other indicating it is a COOP. The owner-ship variable is potentially important, since women may more easilyenter leadership positions in the often more mission-driven NGOsand COOPs. The competition variable is the rater’s assessment ofthe MFIs’ competitive challenge in its area, and converted to a com-mon scale of 1–7. Finally, 36% of the MFIs in our sample are interna-tionally initiated (prodded to start by a Western organization). Thisrich institutional background provides an opportunity for studyingfemale leadership under a diversity of external conditions. Merslandand Strøm (2009) find that these institutional variables are not re-lated to the MFI’s financial performance, but did not investigate gov-ernance issues.

3.6. MFI and country controls

Aguilera and Jackson (2010) point out that country specific tra-ditions and institutions can be important in corporate governancestudies. We include a number of MFI level and country level vari-ables for the governance and performance regressions in order toaccount for MFI heterogeneity. The firm level controls are the MFI’sage, the portfolio at risk (more than 30 days outstanding), and theMFI’s size. The specification of size is the natural logarithm of totalassets, which reduces outlier bias. We expect that the larger theMFI is, the more complex it becomes, and the more it will adoptformal governance mechanisms, that is, monitoring becomes moreimportant and advising less. The MFI’s age is an important control

5 MFIs incorporated as Non-Governmental Organizations (NGOs) are not-for-profitfirms where no particular group or person can legally claim ownership of it or receiveresidual earnings from it (Mersland, 2009).

variable, since an MFI is likely to learn what governance mecha-nisms work and how to achieve profitability.6 Last, the risk is spec-ified as the default rate, that is, the fraction of the portfolio 30 daysoverdue. Cultural predispositions are likely to be found betweencountries. The country control variable includes the Human Devel-opment Index (HDI) from UN’s Development Programme. Further-more, GDP per capita (adjusted for purchasing power parity), GDPgrowth, and the Heritage Foundation’s index of economic freedomenter regressions as controls along with the HDI index. Finally, wecheck for country differences in gender inequalities by includingthe UN index GII (gender inequality index).

The 73 countries rank low on development indices, and arehighly dispersed. Appendix A shows the countries in our sample,the frequency of MFIs in each country and also in main world re-gions. The MFIs in our sample range from countries ranked 35(Argentina) to 135 (Niger) in 1998 and from 39 (Chile) to 134(Democratic Republic of Congo) in 2008 out of 135 countries inthe HDI. Thus, the MFIs in our sample are situated in developingcountries. At the same time, the heterogeneity within the samplewith respect to the home country’s development level underlinesthe need for country controls.

The emphasis on choosing a female leadership is not limited toparticular segments of countries in our sample. This is evidentfrom the scores on HDI and the gender inequality index (GII), seeTable 2. The percentage of female leadership categories in differentworld regions is set out in the lower part of the table.

The table shows that only for the female CEO the score differ-ence is significant in panel A, and even this at the rather low 10%significance level. Thus, we cannot find clear country differenceswith respect to female leadership. Likewise, the female CEO differ-

6 The oldest MFI in our sample is DCC bank in India which started to offermicrofinance services in the early 1920s.


ences across world regions are significant, but not the female chairand the female director. The lowest incidence of female leadershipin any category is in the Eastern Europe and Central Asia (EECA) re-gion. Appendix A shows that the EECA countries constitute 19.5% ofthe sample. Evidently, Table 2 shows that controlling for country orregional differences is important. But the table also shows that fe-male leadership is not confined to a particular region, but shows afairly similar pattern in all regions. The female leadership similar-ity and country controls should ensure that our results are not dri-ven by specific conditions in a few countries.

We run multivariate regressions to analyze relationships. Thereis always a danger that multicollinearity occurs in such a setting.Furthermore, we run corporate governance mechanisms and finan-cial performance analyses separately. A first way to check for mul-ticollinearity and independence is to run a correlation analysis forthe main explanatory variables.

Kennedy (2008) puts the danger level for multicollinearity to acorrelation between two variables at about 0.80. Only the correla-tion between the MFI’s assets and its loan portfolio reported inTable 3 reaches this level. This simply means that these variablesare substitute definitions of the MFI’s size. The next highest levelsare found between the financial performance variables, the highestbeing 0.75 between ROA and ROE. This is as expected. Since theremaining correlations are low, the explanatory variables used inregressions are independent on a satisfactory level, that is, theymay be used independently of each other. The table also showsthat the correlations between corporate governance mechanismsand financial performance variables are weak, the highest being0.22 between OSS and the number of board meetings. This result

Table 2Female leadership and mean scores on the human development index (HDI) and thegender inequality index (GII) in Panel A and the percentage of female leadershipcategories in world regions in Panel B. Higher HDI value means higher development, ahigher GII value means more inequality. HDI is an average over relevant years for theMFI, the GII is indexed in 2008. Source: UN Development Programme and own data.

Female

CEO Chair Director

Table AGII:

No 0.653 0.639 0.602Yes 0.639 0.642 0.629t-Test mean difference 1.276 �0.147 �1.352MFIs 295 214 133

HDI:No 0.604 0.626 0.660Yes 0.633 0.625 0.651t-Test mean difference �1.820* 0.048 0.423MFIs 325 235 147

Table BRegion:

Latin America 31.3 26.7 73.5Africa 24.1 19.7 76.5

MENA 36.4 29.6 75.0EECA 14.1 11.6 60.7

Asia 34.8 28.1 82.4Total 27.1 22.7 73.0MFIs 329 238 148Chisqrd(4) 9.60** 5.27 3.15

The HDI combines indicators of life expectancy, educational attainment (mean ofyears of schooling for adults aged 25 years and expected years of schooling forchildren of school going age) and income (the logarithm of GNI per capita (PPPUSD)) into a composite human development index. The GII includes three maincomponents, reproductive health (maternal mortality and adolescent fertility),empowerment (parliamentary attainment and educational attainment) and labormarket (labor force participation rate). EECA is Eastern Europe and Central Asia;MENA is Middle East and North Africa.* Significance at 10% level.** Significance at 5% level.

is in fact in line with Mersland and Strøm (2009) who find hardlyany significant relationships between governance variables and theMFI’s financial performance. This indicates that governance andfinancial performance variables may be run independently, whichis the procedure we follow here. Nevertheless, we run robustnesschecks on the potential relationships between governance andfinancial performance. These robustness checks parallel the testingprocedure in Adams and Ferreira (2009).

4. Methodology

We use a straightforward probit method to predict the femaleleadership variables. Thus, in the case of the female CEO we runthe probit regression:

PrðFemaleCEOÞ ¼ f ðSocial mission; Institutions; Controls; ErrorsÞð1Þ

and then similar regressions for the chair and female directors. Thisrelation has an interest in itself, but is also fundamental in financialperformance regressions where female leadership may be endoge-nously determined.

When either a corporate governance (CG) variable or a financialperformance (FP) variable is the dependent, the basic estimatingrelation is as follows for the case of return on assets (ROA):

ROA ¼ f ðFemale leadership; MFI controls; Countrycontrols; ErrorsÞð2Þ

The corporate governance variables are meetings, internal audi-tor, CEO duality, and board size. Financial performance is one of thefinancial performance measures (ROA, ROE, OSS or FSS). The CEOand the chair variables are both dichotomous. We also turn the fe-male director variable into an indicator variable in the main regres-sions. For the governance regressions we use the probit methodwhen the governance mechanism is dichotomous (internal auditor,CEO duality), and tobit-censored regressions for the discrete gover-nance variables (board meeting and board size). For instance, theboard size has a truncated distribution, since it cannot be smallerthan 1. The financial performance regressions are performed withinstruments in two different setups that we explain below. Allregressions are performed with a constant, but this constant isnot reported.

In the financial performance estimations we build upon theHeckman (1978) endogenous dummy variable model and followthe two-step procedure laid out by Wooldridge (2010, pp. 937–945). In the first step, an instrument for each female leadershipindicator variable is generated from (1). This is the probability thata given MFI has a female CEO, say. The extracted probability is thenused in the second step random effects model as an instrument forfemale leadership, say the female CEO.

The two-step method yields the added advantage that femaleleadership is regressed on variables that are likely to proxy forthe match with female leadership. This concerns social missionvariables, such as the MFI’s gender bias, but also institutional vari-ables, such as the MFI’s ownership type. The generated instrument,that is, the likelihood that the CEO is female, is likely to be highlycorrelated with the female CEO, but not with any measure of finan-cial performance. Wooldridge (2010) furthermore shows that thegenerating regression (1) does not need to be correctly specifiedin order to generate a useful instrument.

The two-step method has the further advantage that an inverseMill’s ratio test for endogeneity due to sample selection is easilydevised. Sample selection may arise in our context if it is the casethat some able women do not seek leadership positions in theMFIs, despite being as qualified as the observed female CEO anddirectors.

Table 3Correlation matrix. Pairwise correlations between all continuous variables.

1 2 3 4 5 6 7 8 9 10 11

1 # Fem. dir.2 Meetings �0.063 Board size 0.33 �0.114 Roa �0.11 �0.06 0.025 Roe �0.03 �0.06 0.08 0.756 Oss �0.08 0.22 0.01 0.51 0.517 Fss �0.13 0.02 0.07 0.65 0.50 0.438 TA �0.01 �0.01 0.04 0.07 0.13 0.18 0.179 TLP �0.01 �0.01 0.01 0.08 0.14 0.20 0.18 0.98

10 par30 �0.04 0.18 �0.17 �0.09 �0.08 �0.01 �0.22 �0.10 �0.1211 Average loan �0.12 0.19 �0.09 �0.01 0.10 0.38 0.02 0.01 0.03 0.0112 Age 0.01 0.30 �0.05 �0.11 �0.19 �0.04 �0.01 0.23 0.19 0.26 �0.08

Variables are defined in Table 1.

7 The results are unaffected when the HDI control is replaced by the Heritage indexor by a more elaborate set of country controls including per capita GDP and GDPgrowth.


Wooldridge (2010, pp. 809–813) shows how a general test forsample selection can be performed. In our case, when the femaleleadership category under consideration is endogenous, the firststep is to estimate (1) as before. From this estimation one savesthe predicted values for the likelihood that the CEO is female. Thisis the inverse Mill’s ratio. The second step is to insert this predictedvariable into the financial performance regressions (2). If the pre-dicted variable is significant, measured by an ordinary t-test, thereis a case for sample selection.

In both the governance and financial performance regressions anumber of control variables are included so as to remove MFI spe-cific heterogeneity as much as possible. First of all, we estimatewith MFI clustered standard errors to correct for heteroskedasticityand autocorrelation. This alone in fact takes away most of the MFIlevel heterogeneity (Petersen, 2009). Furthermore, the MFI leveland country level variables control for country heterogeneityamong MFIs. Last, we include indicator variables for the mainworld regions, defined in Table 2, and for time. Time indicator vari-ables control for market-wide impacts in financial performanceregressions. With panel data, instruments, and a wide set of controlvariables the estimations should at least vouch for reliable correla-tions, but perhaps causality only in the financial performanceregressions, where we have taken account of the endogeneity of fe-male leadership.

We report on robustness checks in Section 8 varying estimationmethod, variable specification, especially the female director, andthe use of lagged variables. Estimations in Adams and Ferreira(2009) and in Carter et al. (2010) show that results are not robustto estimation method. For instance, the Adams and Ferreira resultsvary with OLS, fixed effects at industry and firm level methods, andfinally, with a two-stage least squares IV method. We specify a dif-ferent IV model for the financial performance estimations, whereinstruments are the significant variables in the regressions of fe-male leadership on background variables in relationship (1). Wecheck for the validity of the instruments with the Sargan test ofoveridentifying restrictions.

Adams and Ferreira (2009) and Carter et al. (2010) also showthat the financial performance results may be sensitive to the def-inition of female director. We perform regressions with differentdefinitions. In particular, we examine the Konrad et al. (2008) crit-ical mass theory with an indicator variable being one if the numberof female directors is equal to or larger than three. Other femaledirector measures put to the test are the percentage of femaledirectors and the absolute number of female directors.

Governance and financial performance may be closely related,although the correlation matrix in Table 3 and evidence in Mers-land and Strøm (2009) cast doubt if this is the case in microfinance.In the Hermalin and Weisbach (1998) model board independenceis endogenously determined by past performance. Accordingly,

Adams and Ferreira (2009) always use governance and financialperformance variables together in regressions. We run two robust-ness checks for governance importance. The first is for governancevariables where former financial performance is an explanatoryvariable. In the second robustness check we include governancemechanisms among the explanatory variables in financial perfor-mance regressions.

Adams et al. (2010) recommend the fixed effects method for pa-nel data, which will remove time-invariant heterogeneity in thedata completely, but then recommend the random effects modelif the board attribute enters nonlinearly, and if different firms facedifferently shaped tradeoffs between governance mechanisms. Ourrandom effects methodology with MFI level standard errors speci-fications are linear, but also MFI specific. Furthermore, a number ofour explanatory variables are themselves time-invariant, andwould be wiped out with fixed effects. This concerns the femaleleadership variables. Accordingly, the random effects method isour preferred method.

The microfinance promise (Morduch, 1999) says that the MFIfollows its social mission to offer financial services to the poorwhile being financially sustainable. This could pose a problem forestimations, if there is a large, negative correlation between finan-cial performance and outreach to poor customers. In our sample,the correlation between ROA and average loan is only 0.111, indi-cating goal independence. This lack of correlation allows separateregressions for governance mechanisms and financial perfor-mance. Nevertheless, we include social mission variables to controlfor any influence.

5. The match of female leadership and the MFI

The first step in the IV procedure outlined above is to generatethe probability that the leader in the MFI is female. We test thehypothesis that female leadership is more likely the greater weightthe MFI places on its social mission. This concerns first and fore-most the MFI’s gender bias, but also the MFI’s choice of urbanand rural market and its average loan size.

We use straightforward probit regressions for the CEO and chairregressions and OLS regression for the female director fractionusing data from the rating year. We have related the three femaleleadership roles of CEO, chair, and director to the background char-acteristics of MFIs and a country control,7 first by using the wholesample, and then using the sample for the rating year only. The re-sults appear in Table 4.

Table 4Female leadership and MFI characteristics.

Dep. var All observations Rating year only

Fem. CEO Fem. chair Fem. dir. fraction Fem. CEO Fem. chair Fem. dir. fraction

Social missionGender bias 0.445*** 0.912*** 0.122*** 0.389** 0.901*** 0.118*

Urban 0.137* 0.202* 0.002 0.059 0.292* 0.017ln(Average loan) �0.072 0.083 �0.010 �0.095 0.061 0.002

Institutional variablesRegulated 0.125 0.281 �0.103*** 0.095 0.371 �0.074Competition �0.056 �0.065* �0.008 �0.067 �0.092 �0.021Int. initiated �0.395*** �0.127 0.022 �0.358** �0.064 0.021NGO 0.378** 0.320* 0.081** 0.128* 0.566* 0.051COOP 0.494*** 0.235** 0.325*** 0.436** 0.492* 0.232**

Firm/country controlsAge �0.009* �0.027*** �0.001 �0.004 �0.041** 0.001PaR30 �0.869 0.229 �0.006 �0.190 0.331 0.015lnTA �0.011 �0.118** �0.032** �0.049 �0.011 �0.039HDI 0.734* �0.964 �0.223* 1.113* �0.562 �0.427

N 731 536 332 254 178 106Wald v2/F-stat. 57.30*** 87.83*** 9.78*** 18.86* 32.20*** 2.42***

Pseudo-R2 0.07 0.12 0.25 0.07 0.14 0.17

Method Probit Probit OLS Probit Probit OLS

We analyze the characteristics that are associated with female leadership in terms of a female CEO, a female chair, a female director, and the fraction of female directors(continuous) by means of pooled probit regressions and pooled OLS. Significance levels are based upon standard errors corrected for autocorrelation and heteroskedasticity.* Statistical significance at the 10% level.** Statistical significance at the 5% level.*** Statistical significance at the 1% level.


Table 4 shows that female leadership is more likely in MFIs thathave a gender bias toward female clients, in MFIs concentrating onthe urban market, in NGOs and cooperatives more than in share-holder-owned MFIs (the omitted category), and in younger MFIs.However, a female CEO is less likely when the MFI has an interna-tional founder, while a female chair is less likely when competitionis high.

It is remarkable that gender bias and ownership type are signif-icant for all three categories of female leadership for the full sam-ple, and that the signs are the same. Gender bias is significant in allleadership categories in the rating year regressions as well. Owner-ship type is significant as well in the rating year regressions, exceptfor NGOs. Thus, the MFI’s choices of social mission and its owner-ship structure are important for understanding the extent of fe-male leadership. The instrument we generate can therefore beused in governance and performance regressions.

The matching regressions from the relationship (1) have paral-lels to the papers on the endogeneity of boards, e.g. Boone et al.(2007), and may contribute to the understanding of the emergenceof female leadership in MFIs. Some of the variables are truly exog-enous, this concerns first and foremost the internationally initiatedand the firm age variables, but also ownership type, regulation andthe social mission orientation are variables that hardly change.

6. Female leadership and corporate governance

Is internal governance better in MFIs with women as the CEO,chair, or director? We perform regressions with the governancevariables as dependent and the female leadership variables asindependent and also include control variables and regional dum-mies.8 Each female leadership variable is introduced into the regres-sion, one at the time. Table 5 gives an overview.

The overall statistics are satisfactory, with high F and Wald chi-square statistics throughout. Thus, we cannot uphold the hypothe-

8 In this regression no time-dummies are taken up since the governance variablesin our dataset do generally not vary over time.

sis that all explanatory variables are insignificant in theregressions.

Contrary to our expectations concerning the advantageousmatch between female leadership and MFI governance, Table 5suggests that female leadership is generally associated with weakercorporate governance. A female CEO is negatively related to boardmeetings but positively to the board size, the female chair anddirector are negatively related to internal audits, but positively toCEO duality. The female director is positively related to board size.The findings are at odds with those of Adams and Ferreira (2009),who find that female directors bring about better governance prac-tices, although they use different measures, in particular boardattendance. We cannot confirm the evidence from a review ofCanadian governance (Brown et al., 2002) that ‘‘boards with morewomen surpass all-male boards in their attention to audit and riskoversight and control’’. Rather, the negative association with bettergovernance by all female leadership categories, not only the CEO,suggests that mechanisms for CEO monitoring have low value infemale-led MFIs.

A number of the control variables have interesting implications,and are generally in line with former research linking governancevariables to firms’ market conditions (Baker and Gompers, 2003;Boone et al., 2007; Linck et al., 2008). MFI size in terms of lnTAis highly significant in all regressions, inducing more meetings,more often an internal auditor and CEO duality, and a larger board.The MFI’s age is likewise highly significant, but the MFI is likely tosplit the CEO and chair role as the MFI gets older. Thus, corporategovernance becomes more important as the MFI is larger and older.This seems reasonable, since formalization often tends to overtakethe entrepreneurial spirit as the firm matures. The internal auditorresults confirm Hay et al. (2006), that is, the same variables areimportant for internal audits. We also find that the COOP owner-ship type comes in combination with more board meetings, less of-ten internal auditor, more often CEO duality, and a larger board.The same is true for the NGO ownership form, except for the num-ber of meetings. Thus, in more complex ownership structures suchas the COOP, governance is discharged by a large board meetingfrequently, and not to supporting mechanisms such as the internal

Table 5Female leadership and corporate governance.

Meetings Internal audit

Female leadershipFemale CEO �1.401*** 0.058Female chair �0.760 �0.282**

Fraction fem. dir. �3.034 �1.336***

Female director �1.016 �0.429***

MFI controlslnTA 0.858*** 0.764** 0.976** 1.017*** 0.328*** 0.265*** 0.408*** 0.381***

Age 0.191*** 0.221*** 0.167** 0.158** 0.013** 0.023*** 0.019** 0.022**

PaR30 0.900 0.370 0.032 0.063 �0.335 �0.773 �0.115 �0.068Competition �0.024 �0.297 �0.383 �0.326 0.056* 0.002 �0.049 0.023NGO �0.371 �0.686 �0.558 �0.341 �0.649*** �0.768*** �0.515** �0.658***

COOP 5.827*** 5.121*** 5.997*** 6.086*** �0.596*** �0.621*** �0.124* �0.702***

Country controlsGDPpercapita 0.002 �0.001 �0.001 �0.001 �0.001 �0.001 �0.001 �0.001Heritage �0.079* �0.056 �0.056 �0.041 �0.001 0.001 �0.021 �0.009GDPgrowth 24.166*** 20.009*** 19.841** 19.086*** 1.421 3.470** 1.823 1.408

Regional dummies Incl. Incl. Incl. Incl. Incl. Incl. Incl. Incl.

N 591 458 336 370 863 623 405 467F stat./Wald v2 12.27*** 12.40*** 11.96*** 10.03*** 181.28*** 131.79*** 129.40*** 150.62***

R2 0.06 0.07 0.07 0.07 0.19 0.17 0.29 0.24Method Tobit Tobit Tobit Tobit Probit Probit Probit Probit

CEO Duality Board size

Female leadershipFemale CEO 0.082 0.761**

Female chair 0.208 �0.309Fraction fem. dir. 0.746** �0.293Female director 0.275* 1.052***

MFI controlslnTA 0.209*** 0.253*** 0.263*** 0.216*** 0.484*** 0.524*** 0.423*** 0.328***

Age �0.026*** �0.034*** �0.074*** �0.039*** �0.009 0.001 �0.042*** �0.051***

PaR30 �0.131 0.100 0.988 0.359 �0.721 0.406 �0.343 �0.552Competition �0.043 �0.031 0.162 0.035 �0.571*** �0.601*** �0.239*** �0.273***

NGO 0.261** 0.208* 0.029 0.131 1.941*** 2.221*** 1.132*** 1.095***

COOP 0.523*** 0.554*** 0.946*** 0.654*** 2.157*** 1.772*** 1.552*** 1.426***

Country controlsGDPpercapita 0.001 0.001 0.001 0.001 �0.001 �0.001 �0.001 �0.001Heritage 0.037*** 0.041*** 0.026** 0.034*** 0.042*** 0.043** 0.019 0.041**

GDPgrowth �2.079 �2.901 �0.453 �2.062 �2.398 0.598 �1.745 �1.985


N 910 685 392 453 883 646 447 500F stat./Wald v2 71.06*** 71.77*** 58.29*** 46.66*** 16.24*** 12.48*** 6.47*** 10.16***

R2 0.10 0.11 0.14 0.09 0.05 0.05 0.05 0.05Method Probit Probit Probit Probit Tobit Tobit Tobit Tobit

We regress corporate governance on female leadership and controls. Governance is measured through the number of board meetings, a binary for internal audits, and a binaryfor CEO duality and the number of board members. For the binary governance variables probit regressions are estimated, for the discrete governance variables, tobit-censoredregressions are estimated. Significance levels are based on heteroskedastic and autocorrelation-corrected standard errors.* Statistical significance at the 10% level.** Statistical significance at the 5% level.*** Statistical significance at the 1% level.

9 The IMR-test was performed for all IV-regressions, we did not find any evidence


auditor. It appears that in member owned firms the members con-trol the firm by sitting on the boards in frequent meetings. Thus,the CEO’s power position is strengthened by the lack of internalauditor and the CEO’s presence at the board, but curtailed by alarge board in frequent meetings. Our ownership type variablesconfirm the main result in Desender et al. (2009), who comparewidely and closely held firms. These findings indicate that our con-trol variables are highly relevant in this study.

Table 5 shows that the country variable GDP growth is signifi-cant and positive in the regressions for board meetings and inter-nal auditor, and that the Heritage Foundation Index is positivelyrelated to CEO duality and board size. More meetings and betterinternal auditing are plausible in a high-growth environment.Thus, country differences are evident for corporate governance,supporting Terjesen and Singh (2008), and complement the femaleleadership findings.

7. Female leadership and performance

How is female leadership linked to MFI performance? Is thematch of the female leadership with social mission oriented MFIsadvantageous for financial performance? Table 6 shows the resultswhen we use the Heckman (1978) model for an endogenous indi-cator variable, the female leadership categories being the indicatorvariable, and the probabilities of being for instance a female CEOderived from regressions in Table 4.

The Wald chi-squared test shows that we cannot leave allexplanatory variables out of the specification. The inverse Mill’s ra-tio is not significant in any of the regressions, thus, our estimationsdo not suffer from sample selection bias.9

of sample selection bias.

Table 6Female leadership and performance: a two-step experiment setting.

ROE ROA

(1) (2) (3) (4) (5) (6)

Panel AFemale leadership

Female CEO 0.189** 0.035Female CHAIR 0.178* 0.104***

Female DIRECTOR 0.152 0.092*

MFI controlslnTA 0.070*** 0.072*** 0.059*** 0.027*** 0.024*** 0.019**

Age �0.005** �0.002 �0.007* �0.001* 0.004 �0.001par30 �0.218** �0.026 �0.062 �0.216*** �0.043 �0.004

Country controlsGDP per capita 0.000 0.000 0.000 0.000 0.000 0.000Heritage �0.002 �0.002 �0.003 �0.001 �0.001 �0.001GDP growth 0.226 0.082 0.237 0.298** 0.298* 0.402*

Regional dummies Yes Yes Yes Yes Yes YesTime dummies Yes Yes Yes Yes Yes Yes

Model statisticsN 608 456 348 649 485 366R2 0.13 0.12 0.09 0.17 0.14 0.13Wald v2 920.72*** 27.77*** 21.59*** 121.87*** 32.10*** 24.07***

IMR-test 0.137 �0.247 �0.133 0.082 �0.168 �0.137

OSS FSS

(1) (2) (3) (4) (5) (6)

Panel BFemale leadership

Female CEO 0.196* 0.302**

Female CHAIR 0.013 0.396**

Female DIRECTOR 0.428* 0.549*

MFI controlslnTA 0.216*** 0.213*** 0.212*** 0.101*** 0.112*** 0.121***

Age �0.004 0.003 0.003 �0.001 �0.001 �0.001par30 �0.384 �0.501 �0.351 �0.648*** �0.741*** �0.729***

Country controlsGDP per capita 0.000 0.000 0.000 0.000 0.000 0.000Heritage �0.017 �0.023 �0.018* �0.005 �0.007 �0.006GDP growth 0.35 0.045 0.154 0.088 �0.13 �0.042

Regional dummies Yes Yes Yes Yes Yes YesTime dummies Yes Yes Yes Yes Yes Yes


IMR-test 0.101 �0.479 �0.169 0.213 �0.201 �0.249

Instruments: fitted probabilities from a probit explaining binary variables for female CEO, female chair, and female director.Financial performance in terms of ROE, ROA (panel A), OSS and FSS (panel B) regressed on female leadership, MFI and country controls using IV. As instruments, fittedprobabilities from a probit analysis explaining the dummy female CEO, dummy female CHAIR and dummy female DIRECTOR have been used, in the dummy endogenousvariable model of Heckman (1978). Significance levels based on heteroskedastic and autocorrelation-corrected standard errors clustered at the MFI level.* Statistical significance at 10% level.** Statistical significance at 5% level.*** Statistical significance at 1% level.


The female CEO is positively and significantly related to finan-cial performance for three out of four measures, confirming ourexpectations and Mersland and Strøm (2009). The female chairand female director are also significant with a positive sign in threeout of four regressions. Thus, results are remarkably similar acrossmanager and director roles. Shrader et al. (1997), Smith et al.(2006), and Francoeur et al. (2008) find that performance is en-hanced with a female CEO, but reduced with female directors.Adams and Ferreira (2009) also find that female directors overallhave a negative impact upon performance. However, our findingsfor female leadership in general support arguments for high abilityamong female MFI leaders due to a superior match of leadershipand tasks.

The firm controls show that the firm effects are reasonable,being positive for MFI size, negative for age, and negative for de-fault risk. The only occasionally significant country effects are nei-ther surprising in view of Allen and Gale (2000) finding that firmperformance is quite similar across financial systems of the world.Thus, from Tables 5 and 6 we find that country idiosyncraciespartly associate with governing mechanisms, but few such associ-ations are detectable for financial performance.

8. Robustness

How reliable are the above results? In this section, we reportvarious robustness checks for the estimation method and variable

Table 7Female leadership and performance.

ROE ROA


Female CEO 0.202** 0.136Female chair 0.167* 0.024Fraction fem. dir. 0.245 0.016

MFI controlslnTA 0.067*** 0.067*** 0.063*** 0.029*** 0.025*** 0.021**

Age �0.005** �0.002 �0.006** �0.001 �0.001 �0.002PaR30 �0.266 �0.024 0.036 �0.243** �0.076* �0.095

Country controlsGDPpercapita �0.001 0.001 �0.001 �0.001 �0.001 �0.001Heritage �0.002 �0.001 �0.002 �0.001 �0.001 �0.001GDPgrowth 0.254 0.027 0.189 0.381* 0.254* 0.396**

Regional dummies Included Included Included Included Included IncludedTime dummies Included Included Included Included Included Included


Sargan stat. 1.670 4.195 2.920 6.884 6.786 3.076Sargan p-value 0.64 0.24 0.40 0.07 0.08 0.38

OSS FSS


Female CEO 0.175* 0.423**

Female chair 0.034 0.487**

Fraction fem. dir. 0.116 0.694**

MFI controlslnTA 0.212*** 0.204*** 0.191*** 0.117*** 0.131*** 0.121***

Age �0.005 �0.004 �0.002 �0.003 �0.003 �0.004PaR30 0.191 0.345 0.341 �0.662** �0.724*** �0.549*

Country controlsGDPpercapita 0.001 0.001 0.001 0.001 0.001 0.001Heritage �0.014*** �0.018** �0.011 �0.001 �0.001 �0.005GDPgrowth 0.965 0.502 0.150 0.048 0.016 0.478

Regional dummies Included Included Included Included Included IncludedTime dummies Included Included Included Included Included Included


Sargan stat. 1.979 6.684 5.126 4.659 2.083 1.878Sargan p-value 0.19 0.15 0.16 0.20 0.56 0.60

Instruments usedOwnership type, internationally initiated, dumrural, gender biasAn instrumental variables approach to determine whether female leadership stimulates financial performance. We regress financial performance on female leadership using2SLS. For instruments, the variables that have a clear relation with female leadership according to Table 4 have been used. The validity of the instruments is tested using theSargan test of overidentifying restrictions. Significance levels are based on heteroskedastic and autocorrelation-corrected standard errors clustered at the MFI level.* Statistical significance at 10% level.** Statistical significance at 5% level.*** Statistical significance at 1% level.


specification. This is motivated by the diversity of results in the lit-erature depending on the method used or the variablespecification.

In the first robustness test we use a different instrumental vari-ables method for financial performance, where instruments are ta-ken from the significant variables in the matching regressions inTable 4. At the same time, we use the fraction of female directorsas our definition. Table 7 reports the results.

The signs are everywhere the same as in Table 6, but the signif-icant results are fewer in this case. The MFI control variables be-

10 In Table 7, we have 2 out of 12 regressions were at the 10% significance level inthe ROA regressions. Table 8 shows significance levels at 1% and 2% in two OSSregressions. Thus, the null hypothesis of no overidentifying restrictions cannot beupheld in these cases.

have much as in Table 6 as well. The Sargan test tells us thatinstruments are relevant for these estimations.10 Thus, overall,the results in Table 6 are robust to the estimation method used.

Ambivalent results for the female director in former researchmotivate alternative tests employing different specifications of thefemale director. In particular, we examine the Konrad et al. (2008)critical mass hypothesis that there needs to be at least three femaledirectors to realize their positive impact in full. We implement thiswith an indicator variable being one if the number of female directorsis equal to or larger than three. Other female director measures put tothe test are a binary variable for a gender mixed board and the abso-lute number of female directors. Table 8 reports the results.

It turns out that the new specifications of the female directorgive significant results only for the FSS measure. Thus, we cannotconfirm the critical mass theory or other specifications of the

Table 8Instrumental variables performance regressions with different definitions for female director.

ROE ROA

Panel AFraction fem. dir. 0.245 0.016# Female directors 0.025 0.008dumfemdir P 3 0.071 0.013lnTA 0.063*** 0.058*** 0.056*** 0.021** 0.021*** 0.021***

Age �0.006** �0.006* �0.007** �0.002 �0.002* �0.002*

PaR30 0.036 �0.044 �0.039 �0.095 �0.093 �0.095Sargan stat. 2.920 3.293 4.352 3.076 2.976 2.305Sargan p-value 0.40 0.35 0.23 0.38 0.39 0.51

N 377 377 377 406 406 406R2 0.10 0.10 0.10 0.17 0.17 0.18Wald v2 116.79*** 40.28*** 253.88*** 270.10*** 400.83*** 536.08***

OSS FSS

Panel BFraction fem. dir. 0.116 0.694**

# Female directors �0.005 0.085dumfemdir P 3 �0.019 0.391*

lnTA 0.191*** 0.191*** 0.190*** 0.121*** 0.107*** 0.102***

Age �0.002 �0.003 �0.001 �0.004 �0.003 �0.005PaR30 0.341 �0.034 �0.035 �0.549* �0.521 �0.547Sargan stat. 5.126 11.170 10.028 1.878 1.592 1.562Sargan p-value 0.16 0.01 0.02 0.60 0.66 0.67

N 408 408 408 360 360 360R2 0.22 0.15 0.11 0.18 0.11 0.11Wald v2 122.12*** 199.42*** 52.05*** 27.11*** 37.68*** 380.32***

Country control variables: GDPpercapita, Heritage, GDPgrowthInstruments usedOwnership type, internationally initiated, dumrural, gender biasFraction fem.dir. is the percentage female directors. # female directors is a continuous variable denoting the number of female directors. Dumfemdir P 3 is a dummy that is 1if the MFI has three or more female directors. For instruments, the variables that have a clear relation with female leadership according to Table 4 have been used. The validityof the instruments is tested using the Sargan test of overidentifying restrictions. Significance levels are based on heteroskedastic and autocorrelation-corrected standarderrors clustered at the MFI level. Year dummies and country controls are included.* Statistical significance at 10% level.** Statistical significance at 5% level.*** Statistical significance at 1% level.

Table 9Corporate governance and financial performance.

Meetings Internal audit CEO duality Board size

Lagged financial performanceROA t � 1 �0.213 �0.549 �0.645 �0.023 �0.163 �0.250 �1.391 �1.912

Female leadershipFemale CEO �1.608*** 0.065 0.121 0.694**

Female chair �0.775 �0.296** 0.279* �0.382

MFI controlslnTA 1.076*** 0.956*** 0.344*** 0.270*** 0.201*** 0.242*** 0.529*** 0.621***

Age 0.219*** 0.225*** 0.014* 0.023* �0.032*** �0.039*** �0.011 �0.001PaR30 0.614 0.751 �0.956 �0.324 �0.258 �0.531 �0.989 �0.352Competition �0.007 �0.285 0.071 0.008 �0.055 �0.042 �0.586*** �0.613***

NGO �0.171 �0.601 �0.681*** �0.823*** 0.254* 0.191* 1.880*** 2.259***

COOP 6.144*** 5.101*** �0.623*** �0.667*** 0.495** 0.507*** 2.002*** 1.604***

Country controlsGDPpercapita �0.002 �0.001 �0.001 �0.001 0.001 0.001 �0.002 �0.001Heritage �0.116** �0.088 �0.002 �0.004 0.042*** 0.049*** 0.055*** 0.053**

GDPgrowth 25.865*** 22.055*** 2.483 3.761** �0.601 �1.241 �3.502 �0.102


N 422 325 615 444 650 587 626 458F stat./Wald v2 8.78*** 9.05*** 136.43*** 103.29*** 51.81*** 50.15*** 12.19*** 9.60***

R2 0.07 0.06 0.19 0.18 0.10 0.10 0.05 0.05Method Tobit Tobit Probit Probit Probit Probit Tobit Tobit

We regress corporate governance on lagged financial performance and controls. Governance is measured through the number of board meetings, a binary for internal audits,and a binary for CEO duality and the number of board members. For the binary governance variables probit regressions are estimated, for the discrete governance variables,tobit-censored regressions are estimated. Significance levels are based on heteroskedastic and autocorrelation-corrected standard errors.* Statistical significance at the 10% level.** Statistical significance at the 5% level.*** Statistical significance at the 1% level.



female director. The overall findings on the female director vari-able are inconsistent, partly driven by different specifications ofthe variable. Thus, we confirm findings in Carter et al. (2010) thatthe definition of female director matters. This implies that differentdefinitions of female director may cause the different result wefind in the literature.

Next, we turn to the possible link between governance andfinancial performance, as the Hermalin and Weisbach (1998) mod-el suggests. We add the lagged ROA to the set of right hand sidevariables and rerun Table 5 in the female CEO and chair regres-sions. Results are set out in Table 9.

We find no significant relationships between the lagged ROAand governance. Given the very low correlations in Table 3 between

Table 10Financial performance when board size and the number of board meetings are included.

ROE

(1) (2) (3


Female CEO 0.198**

Female CHAIR 0.223*

Female DIRECTOR 0

MFI controlslnTA 0.071*** 0.078*** 0Age �0.008* �0.007* �0par30 �0.267** �0.339* �0

Board variablesBoard size 0.003 0.007# Board meetings 0.004 0.003

Country controlsGDP per capita 0.001 0.001Heritage �0.002 �0.003 �0GDP growth 0.187 0.028

Regional dummies Yes Yes YeTime dummies Yes Yes Ye

Model statisticsN 374 303 24R2 0.16 0.12 0.Wald v2 72.71*** 53.47*** 51

OSS

(1) (2) (3


Female CEO 0.674*

Female CHAIR 0.083Female DIRECTOR 0

MFI controlslnTA 0.226*** 0.237*** 0Age �0.010* �0.003 �0par30 �0.090 0.059 �

Board variablesBoard size �0.001 0.019 �0# Board meetings 0.019 0.024 0

Country controlsGDP per capita 0.001 0.001Heritage �0.018* �0.024* �0GDP growth 0.223 0.405 0.

Regional dummies Yes Yes YeTime dummies Yes Yes Ye

Model statisticsN 403 328 26R2 0.23 0.20 0.Wald v2 116.69*** 97.40*** 82

Instruments: fitted probabilities from a probit explaining binary variables for female CEThe Heckman (1978) dummy endogenous model is used in estimations. Significance lev* Statistical significance at the 10% level.** Statistical significance at the 5% level.*** Statistical significance at the 1% level.

governance and financial performance variables this comes as nosurprise. Also, the results in Table 9 confirm findings in Merslandand Strøm (2009), where internal governance mechanisms arenot associated with financial performance. These findings supportthe regression specifications we have used here, treating gover-nance and financial performance in separate estimations.

Finally, we turn the question around following Adams andFerreira (2009) in including corporate governance variables intothe financial performance regressions. We include board meetingsand board size among the explanatory variables in the financialperformance regressions, using the Heckman (1978) dummyendogenous variable method from Table 6. The regressions areset out in Table 10.

ROA

) (4) (5) (6)

0.116*

0.088*

.154 0.096*

.087*** 0.032*** 0.027*** 0.029***

.010* �0.002* �0.001 �0.002

.585* �0.075 0.051 0.122

0.004 �0.002 �0.001 �0.0040.002 0.001 0.001 �0.001

0.001 �0.001 0.001 0.001.003 0.001 �0.001 �0.001

0.041 0.165 0.152 0.227

s Yes Yes Yess Yes Yes Yes

6 401 325 26011 0.11 0.11 0.17.26*** 73.61*** 55.54*** 39.36***

FSS

) (4) (5) (6)

0.578*

0.383**

.412* 0.091

.223*** 0.114*** 0.121*** 0.088***

.007 0.001 �0.001 �0.0020.285 �0.625** �0.857** �0.782**

.006 �0.004 0.008 0.006

.028* 0.006 0.001 �0.001

0.001 0.001 0.001 0.001.015* 0.004 �0.004 �0.004

536 0.806 0.721 0.841

s Yes Yes Yess Yes Yes Yes

3 317 286 23620 0.16 0.16 0.15.39*** 56.87*** 61.83*** 43.51***

O, female chair, and female director.els are based on heteroskedastic and autocorrelation-corrected standard errors.


We find only a weak relationship in the OSS regressions whenthe female director is taken to be the female leadership variable.This means that our results from Table 9 are upheld, and that wedo not risk much in not including governance mechanisms in ourmain financial performance regressions in Table 6. An interestingresult in Table 10 is that the female CEO relates significantly toROA when the board size is in the regression. This is in fact thesame result that Mersland and Strøm (2009) obtain, supportingagain the enhanced financial performance in female-led MFIs.

In unreported regressions we use winsorized data, but find nomaterial difference in results from those presented here. Further-more, we also tried regressions without the GDP Growth andregressions with interactions between governance variables andownership type. None of these robustness regressions gave resultsthat differ from those reported.

The overall conclusion of the robustness checks is that our re-sults are confirmed when we vary estimation method, variable def-inition for female director, and regression specifications. Thus, ourconclusions from the governance and financial performance sec-tions are upheld.

9. Conclusions

Microfinance institutions (MFI) are remarkable in that they hiremore female CEOs and elect more female chairs and female directorsthan firms in advanced countries. An MFI’s mission is to supply loansto low-income families and small businesses, especially women, inthe developing world, and it aims to do so in a financially sustainablemanner (Morduch, 1999). This paper investigates the conditions underwhich female leadership tends to emerge, and the relationships be-tween female leadership and corporate governance, as found in Adamsand Ferreira (2009), and its association with the MFIs’ financial perfor-mance. Unlike other studies, that often include only directors, ours spec-ify female leadership as a female CEO, chair, or director. The data arehand-collected from third-party raters’ reports on 329 MFIs in 73 coun-tries, and each MFI rating report provides information for up to six years.

Three main conclusions emerge from the investigation. The first con-cerns the conditions under which female leadership is generally found.Female leadership increases with the MFI’s mission to supply credit par-

No. Country Total No. Country

1 Albania 1 31 Moldova2 Argentina 1 32 Morocco3 Armenia 3 33 Nicaragua4 Benin 6 34 Pakistan5 Bolivia 14 35 Paraguay6 Bosnia Hercegovina 10 36 Peru7 Brazil 13 37 Philippines8 Bulgaria 2 38 Romania9 Burkina Faso 3 39 Russian Feder

10 Cambodia 12 40 Senegal11 Chile 2 41 South Africa12 Colombia 6 42 Sri Lanka13 Dominican Republic 3 43 Tanzania14 Ecuador 16 44 Togo15 Egypt 4 45 Trinidad and16 El Salvador 4 46 Tunisia17 Ethiopia 7 47 Uganda18 Georgia 5 48 Montenegro19 Guatemela 5 49 Cameroun20 Haiti 2 50 Guinee21 Honduras 8 51 East Timor

ticularly to women, with being a cooperative or a NGO. However, havingan international founder tends to reduce female leadership. This sup-ports the matching argument from the Becker (1973) model in the sensethat female CEOs are indeed matched with gender biased MFIs.

The second main conclusion deals with the relationships be-tween female leadership and corporate governance. Unlike Adamsand Ferreira (2009), we find female leadership to be associatedwith weaker corporate governance, since board meetings are few-er, internal audits less common, and CEO duality more commonwhen women hold leadership positions.

Finally, the third main conclusion confirms the findings ofShrader et al. (1997), Smith et al. (2006), and Francoeur et al.(2008) that female CEOs are positively related to firms’ financialperformance. This also applies to female directors, adding evidenceto the just mentioned studies and also to Adams and Ferreira(2009). We also find that the female chair performance effect issimilar to that of the female CEO; financial performance is betterin MFIs with a female chair. The conclusion is that in an industrylargely catering for female customers, having female leadership islikely to improve the MFI’s financial performance. Taken togetherwith the negative relationship between female leadership and gov-ernance mechanisms, this means that female-led MFIs performbetter with less oversight, less monitoring. The upshot is that thequality of leadership is decisive in microfinance institutions.

We believe the contrasting governance and financial perfor-mance results are due to the fact that MFIs are young, entrepre-neurial firms. The optimal governance form has perhaps not beensettled. Furthermore, the abilities of the individual CEO are pivotalin this rapidly expanding segment. Future research would benefitfrom exploring the extent and the implications of the female lead-ership attribute, like education and experience, in more detail.Future research could also fruitfully explore whether female lead-ership is better at meeting the MFI’s outreach goals than male.

Appendix A

Countries in the sample and the frequency in each country, andfrequency of firms in different world regions

Total No. Country Total

2 61 Chad 14 62 Rwanda 49 63 Zambia 11 64 China 11 65 Serbia 1

25 66 Ghana 37 67 Malawi 11 68 Gambia 1

ation 12 69 Kosovo 48 70 Rep of CongoBrazz 11 71 Burundi 11 72 Niger 11 73 DRC – Kinshasa 13 Grand total 329

Tobago 115 Regional distribution2 Code Region Total3 1 Latin America 991 2 Africa 871 3 MENA 33

(continued on next page)

Appendix A (continued)

No. Country Total No. Country Total No. Country Total

(continued on next page)22 India 30 52 Bangladesh 2 4 EECA 6423 Indonesia 2 53 Nepal 5 5 Asia 4624 Jordan 3 54 Vietnam 1 Grand total 32925 Kazakhstan 3 55 Azerbaijan 526 Kenya 6 56 Mongolia 227 Kyrgyzstan 3 57 Nigeria 328 Madagascar 1 58 Mozambique 129 Mali 3 59 Tajikistan 430 Mexico 16 60 Croatia 1


References

Adams, R.B., Ferreira, D., 2007. A theory of friendly boards. Journal of Finance 62 (1),217–250.

Adams, R.B., Ferreira, D., 2009. Women in the boardroom and their impact ongovernance and performance. Journal of Financial Economics 94 (2), 291–309.

Adams, R.B., Hermalin, B.E., Weisbach, M.S., 2010. The role of boards of directors incorporate governance: a conceptual framework & survey. Journal of EconomicLiterature 48 (1), 58–107.

Aguilera, R.V., Jackson, G., 2010. Comparative and international governance. TheAcademy of Managerial Annals 4 (1), 485–556.

Ahern, K.R., Dittmar, A.K., 2012. The changing of the boards: the impact on firmvaluation of mandated female board representation. Quarterly Journal ofEconomics 127 (1), 137–197.

Allen, F., Gale, D., 2000. Comparing Financial Systems. The MIT Press, Cambridge,Mass.

Alvesson, M., Billing, Y.D., 2009. Understanding Gender and Organizations, seconded. Sage Publications, London.

Armendáriz, B., Morduch, J., 2010. The Economics of Microfinance, second ed. MITPress, Cambridge.

Baker, M., Gompers, P.A., 2003. The determinants of board structure at the initialpublic offering. Journal of Law and Economics 46, 569–598.

Becker, G.S., 1973. A theory of marriage: Part I. Journal of Political Economy 81 (4),813–846.

Beisland, L.A., Mersland, R., 2012. Do microfinance rating assessments make sense?– an analysis of the drivers of MFI ratings. Nonprofit & Voluntary SectorQuarterly 41 (2), 213–231.

Beisland, L.A., Mersland, R., Randøy, T., 2014. Transparency and disclosure in theglobal microfinance industry: implications for practice and policy makers. In:Forssbaeck, Oxelheim (Eds.), The Oxford Handbook of Political, Institutional andCorporate Transparency. Oxford University Press, New York, USA.

Bertrand, M., Schoar, A., 2003. Managing with style: the effect of managers on firmpolicies. Quarterly Journal of Economics 118 (4), 1169–1208.

Bøhren, Ø., Strøm, R.Ø., 2010. Governance and politics: regulating independenceand diversity in the board room. Journal of Business Finance & Accounting 37(9–10), 1281–1308.

Boone, A., Field, L., Karpoff, J., Raheja, C., 2007. The determinants of board size andcomposition: an empirical analysis. Journal of Financial Economics 85 (1), 66–101.

Brown, D.A.H., Brown, D.L., Anastasopoulos, V., 2002. Women on Boards: Not Justthe Right Thing. But the ‘‘Bright’’ Thing. The Conference Board of Canada, ReportMay 2002.

Cadbury, A., 2002. Corporate Governance and Chairmanship. A Personal View.Oxford University Press, Oxford.

Campbell, K., Minguez-Vera, A.M., 2008. Gender diversity in the boardroom andfirm financial performance. Journal of Business Ethics 83 (3), 435–451.

Campbell, K., Minguez-Vera, A.M., 2010. Female board appointments and firmvaluation: short and long-term effects. Journal of Management and Governance14 (1), 37–59.

Carter, D.A., Simkins, B.J., Simpson, W.G., 2003. Corporate governance, boarddiversity and firm value. Financial Review 38, 33–53.

Carter, D.A., D’Souza, F., Simkins, B.J., Simpson, W.G., 2010. The gender and ethnicdiversity of US boards and board committees and firm financial performance.Corporate Governance: An International Review 18 (5), 396–414.

Cull, R., Demigüc-Kunt, A., Morduch, J., 2009. Microfinance meets the market.Journal of Economic Perspectives 23 (1), 167–192.

Desender, K.A., Garcia-Cestona, M.A., Crespi, R., Aguilera, R.V., 2009. BoardCharacteristics and Audit Fees: Why Ownership Structure Matters? Universityof Illinois at Urbana-Champaign, College of Business Working Papers.

D’Espallier, B., Guérin, I., Mersland, R., 2011. Women and repayment inmicrofinance: a global analysis. World Development 39 (5), 758–772.

Eisenberg, T., Sundgren, S., Wells, M.T., 1998. Larger board size and decreasing firmvalue in small firms. Journal of Financial Economics 48, 35–54.

Farrell, K.A., Hersch, P.L., 2005. Additions to corporate boards: the effect of gender.Journal of Corporate Finance 11 (1–2), 85–106.

Francoeur, C., Labelle, R., Sinclair-Desgagné, B., 2008. Gender diversity incorporate governance and top management. Journal of Business Ethics 81,83–95.

Gallego-Álvarez, I., García-Sánchez, I.M., Rodríguez-Dominguez, L., 2010. Theinfluence of gender diversity on corporate performance. Spanish AccountingReview 13 (1), 53–88.

Ghatak, M., 2000. Screening by the company you keep: joint liability lending andthe peer selection effect. Economic Journal 110, 601–631.

Hay, D., Knechel, W.R., Wong, N., 2006. Audit fees: a meta-analysis of the effect ofsupply and demand attributes. Contemporary Accounting Research 23 (1), 141–191.

Heckman, J.J., 1978. Dummy endogenous variables in a simultaneous equationsystem. Econometrica 46 (6), 931–959.

Helms, B., 2006. Access for All: Building Inclusive Financial Systems. CGAP,Washington.

Hermalin, B.E., Weisbach, M.S., 1998. Endogenously chosen boards of directorsand their monitoring of the CEO. American Economic Review 88 (1),96–118.

Kang, E., Ding, D.K., Charoenwong, C., 2010. Investor reaction to women directors.Journal of Business Research 63 (8), 888–894.

Kanter, R.M., 1977. Men and Women of the Corporation. Basic Books,New York.

Kennedy, P., 2008. A Guide to Econometrics, sixth ed. Blackwell Publishing, Malden,Mass.

Konrad, A., Kramer, V., Erkut, S., 2008. Critical mass: the impact of three or morewomen on corporate boards. Organizational Dynamics April (3), 145–164.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, T.W., 1998. Law and finance.Journal of Political Economy 106 (6), 1113–1155.

Linck, J.S., Netter, J.M., Yang, T., 2008. The determinants of board structure. Journalof Financial Economics 87 (2), 308–328.

Mersland, R., 2009. The cost of ownership in micro-finance organizations. WorldDevelopment 37 (2), 469–478.

Mersland, R., Strøm, R.Ø., 2009. Performance and governance in microfinanceinstitutions. Journal of Banking & Finance 33 (4), 662–669.

Monks, R.A.G., Minow, N., 2008. Corporate Governance, fourth ed. John Wiley &Sons, Chichester, UK.

Morduch, J., 1999. The microfinance promise. Journal of Economic Literature 37 (4),1569–1614.

Organisation for Economic Co-operation and Development, 2004. Principles ofCorporate Governance. Paris, France.

Petersen, M.A., 2009. Estimating standard errors in finance panel data sets:comparing approaches. Review of Financial Studies 22 (1), 435–480.

Randøy, T., Strøm, R.Ø., Mersland, R., 2014. The impact of entrepreneur-CEOs inmicrofinance institutions: a global survey. Entrepreneurship Theory & Practice39 (4).

Rose, C., 2007. Does female board representation influence firm performance? TheDanish evidence. Corporate Governance: An International Review 15 (2), 404–413.

Schreiner, M., 2002. Aspects of outreach: a framework for discussion of the socialbenefits of microfinance. Journal of International Development 14, 591–603.

Shrader, C.B., Blackburn, V.B., Iles, P., 1997. Women in management and firmfinancial performance: an exploratory study. Journal of Managerial Issues 9 (3),355–372.

Smith, N., Smith, V., Verner, M., 2006. Do women in top management affect firmperformance? A panel study of 2,300 Danish firms. International Journal ofProductivity and Performance Management 15 (7), 569–593.

Speckbacher, G., 2008. Nonprofit versus corporate governance: an economicapproach. Nonprofit Management and Leadership 18, 295–320.

Teigen, M., 2000. The affirmative action controversy. NORA – The Nordic Journal ofWomen Studies 8 (2), 63–77.

Terjesen, S., Singh, V., 2008. Female presence on corporate boards: a multi-countrystudy of environmental context. Journal of Business Ethics 83, 55–63.

http://refhub.elsevier.com/S0378-4266(14)00028-4/h0005






















































































































Thomas, A.S., Ramaswamy, K., 1996. Matching managers to strategy: further tests ofthe Miles and Snow typology. British Journal of Management 7, 247–261.

Welbourne, T.M., Cycyota, C.S., Ferrante, C.J., 2007. Wall street reaction to women inIPOs: an examination of gender diversity in top management teams. Group &Organization Management 32 (5), 524–547.

Wooldridge, J., 2010. Econometric Analysis of Cross Section and Panel Data. The MITPress, Cambridge, Mass.

Yermack, D., 1996. Higher market valuation of companies with a small board ofdirectors. Journal of Financial Economics 40, 185–212.










female leadership, performance, and governance in microfinance institutions

Documents