Investment in Microfinance Equity: Risk, Return, and Diversification Benefits

This paper is the first to draw a global picture of worldwide microfinance equity by taking full advantage of daily quoted prices. We revisit previous findings showing that investors should consider microfinance as a self-standing sector. Our results are threefold: 1) Microfinance has become less risky and more closely correlated with the financial sector, 2) Microfinance shares exhibit higher market betas than those of conventional financial institutions, and have equivalent currency exposure, and 3) Introducing a self-standing microfinance sector presents few diversification benefits. This paper confirms that microfinance has changed dramatically during the last decade.


Introduction
The microfinance industry offers attractive opportunities to investors who want to help alleviate poverty in developing countries. Furthermore, microfinance is often perceived as different from conventional finance. It is thought to be riskier because borrowers generally have no collateral and little or no credit history to underpin a precise assessment of their credit risk. Despite this, investing in microfinance is also viewed as a way of diversifying from international markets in general and the conventional financial sector in particular (Gonzalez, 2007;Krauss and Walter, 2009;Di Bella, 2011;Galema et al, 2011). This diversification effect stems from the special nature of microfinance institutions, which are driven by a double bottom-line (social and financial), and from the fact that they are less leveraged than traditional financial institutions. By scrutinizing the most recent trends of this rapidly evolving industry, this paper aims to check whether investors still need to consider microfinance as a self-standing sector. To this end, it examines both the risk factors associated with microfinance and the diversification benefits of microfinance investment.
Microfinance has changed dramatically during the last decade, moving from a universe of donor-financed NGOs towards a widely disparate industry (Mersland, 2009), including a growing number of commercial banks. Simultaneously, socially responsible investments have gained momentum on financial markets. We assess the impact of microfinance equity on globally diversified portfolios by using the classical tools of portfolio analysis. For that purpose, we construct original microfinance country indices, analyze their movements, and assess them in reference to comparable indices for the financial sector and to national indices.
We also take international equity indices into consideration. 1 Our findings show that although the microfinance sector has definitely moved closer to conventional finance, the risk-return trade-off has also changed. 2 Microfinance equity is now less risky than before, but it is also much more closely correlated with the financial sector, and yields fewer diversification benefits. We conclude that microfinance can reasonably be considered as part of the financial sector.
Two types of public investments in microfinance are currently available: Microfinance Investment Vehicles (MIVs) 3 and listed equity of Microfinance Institutions (MFIs). For reasons of data availability, this paper concentrates on listed MFIs. Arguably, microfinance equity is not representative of the whole sector. 4 However, while confined to a relatively small number of assets, microfinance equity has the considerable advantage over MIVs of being publicly priced daily on stock exchanges. This makes it more transparent and allows for deeper financial analysis. Conversely, MIVs invest in several MFIs mainly through loans, but the content of their portfolios is often opaque, 5 making it difficult for outsiders to assess the actual level of risk.
The microfinance sector has experienced successful initial public offerings (IPOs) such as the highly publicized flotation of Banco Compartamos in Mexico in 2007. These IPOs have nevertheless been criticized by influential players, including Nobel Prize winner Muhammad Yunus, who views the Compartamos IPO as mission drift 6 that compromises the sector's reputation (see Ashta and Hudon, 2012, for a detailed discussion). Leaving ethical and Indices, computed from book values (see http://www.wallsstreetadvisorservices.com/). By contrast, our indices are based on market prices only. 2 Our paper examines whether this assertion holds true for investors who are seeking financial returns only. Alternatively, financial and social returns can be combined (see Drut, 2010;Dorfleitner et al., 2012). 3 See e.g. Matthäus-Maier and von Pischke (2006). 4 In particular, many MFIs still rely on subsidies for reaching financial sustainability (Hudon, 2010;Nawaz, 2010;Hudon and Traca, 2011). 5 The authors -and colleagues from other universities -have tried for years to obtain data on MIVs with little success. Regarding transparency, MIVs tend to adopt an attitude comparable to that of hedge funds. 6 Microfinance mission drift stems from the double bottom-line (social and financial) embodied by the MFIs. An MFI is said to be mission-drifted when it sacrifices its social mission (typically, poverty alleviation and/or women's empowerment) for financial purposes (see, e.g., McIntosh and Wydick, 2005;Copestake, 2007;Ghosh and Van Tassel, 2008;Mersland and Strøm, 2010;Armendariz and Szafarz, 2011). mission-based considerations aside, this paper is the first, to our knowledge, to start from observable returns of publicly traded MFIs. From a portfolio perspective, these returns are to be judged not only on a case-by-case basis but also in regard to their correlation with other assets.
Previous work has already investigated the financial properties of microfinance investment. However, because of data availability issues, authors are bound to use figures extracted from annual accounting statements provided by the Microfinance Information Exchange (MixMarket) rather than high frequency market data. This significantly limits the relevance of their results for mainstream investors. From this perspective, Krauss and Walter (2009) present evidence that, over the period 1998-2006, including microfinance in global portfolios reduced overall portfolio volatility, but that the same result did not hold for domestic investors. Using MixMarket data for the period 1997period -2007period , Galema et al. (2011 apply the spanning test methodology proposed by De Roon et al. (2001), and confirm that investment in microfinance is beneficial in terms of portfolio diversification.
In this paper, we construct microfinance country equity indices and an international Global Microfinance Index (GMI). We analyze the changes in these indices, assessing them in reference to comparable indices for the financial sector and also to national indices. Our findings show that microfinance has been closely correlated with the financial sector since 2003. In terms of risk exposure, estimates of the Capital Asset Pricing Model (CAPM) demonstrate that microfinance shares exhibit higher market beta than those of conventional financial institutions and have equivalent currency exposure. We then turn to spanning tests and show that introducing a self-standing microfinance sector presents little interest regarding banking sector diversification in emerging countries.
The remainder of the paper is structured as follows. Section 2 describes the database and methodology. In section 3, we study the joint movements of the finance and microfinance indices. In section 4, we analyze microfinance risk exposures by estimating both a local CAPM including the foreign-exchange factor and an international four-factor model. Section 5 assesses the diversification benefits of microfinance stocks. Section 6 concludes.

Data and Methodology
We consider the complete universe of listed MFIs. 7 This exhaustive approach is a real advance over the existing literature because the securities in question are the only assets that allow investors attracted by the microfinance sector to gain direct access to the capital of MFIs. The alternative, MIVs, concentrates primarily on loans to MFIs , and the choice of institutions benefiting from them is often unclear.
Three MFIs are quoted in South Africa: African Bank (the oldest quoted MFI, since January 1990), Blue Financial Services (BFS), and Capitec; one in Kenya: Equity; two in Indonesia: Danamon (also one of pioneering quoted MFI, since April 1990) and Bank Rakyat Indonesia (BRI); one in Bangladesh: Brac; and two in Mexico: Compartamos and Financiera Independencia (FI).
Given the small number of listed MFIs, it is important to understand how they compare to the universe of MFIs in the developing world. MFIs are hybrid institutions driven by both financial and social motives. Acknowledging this double-bottom-line specificity, Table I successively reports not only financial characteristics (return on equity, return on assets, operational self-sufficiency (OSS), 8 and debt-to-equity ratio), but also social ones (average loan size, interest rates charged on loans, and share of female borrowers). The non-listed 7 We follow De Mariz et al. (2011) and consider listed institutions with daily liquidity of at least USD 0.2 million. This ensures that prices are not distorted by supply/demand imbalances. Actually, we only exclude SKS, the Indian quoted MFI, because its IPO is too recent (August 2010). 8 OSS is equal to the total financial revenue divided by (financial expenses plus loan loss provision plus operating expenses). An MFI with an OSS value above 100% is said "operationally sustainable," meaning that its revenue from lending is sufficient to cover its costs, including the cost of capital, accounting for bad loans, and paying operating expenses (see Ahlin and Lin, 2006 Regarding financial characteristics, Table I shows that listed MFIs are among the most profitable MFIs, although most differences are not statistically different because the microfinance industry is highly heterogeneous, which implies significant dispersion of financial indicators. The relatively higher profitability of listed MFIs is especially apparent in OSS values. Two factors concur to explain this outcome. First, IPOs impose changes in firms' capital structure. As pointed out by Cull et al. (2011), complying with regulations affects the characteristics of microfinance institutions. Second, many non-listed MFIs still rely on subsidies, which may hinder financial performance. Bogan (2012) points out that long-term use of grants perpetuates inefficiencies by insulating MFIs from competitive pressure. In contrast, listed firms are exposed to competition, as they need to attract market funding. 10 < Insert Table I here > The social performances reported in Table I are summarized by three indicators widely used in microfinance: average loan size, measured by the average loan balance per borrower scaled by GNI per capita; the interest rate charged on loans, proxied by the nominal yield on gross portfolio; and the share of female borrowers. 11 9 The MixMarket is the leading business information provider dedicated to microfinance sector. The organization provides detailed financial and social performance information from MFIs. MixMarket data is self-reported. In addition, a diamond-based ranking system gauges the MFIs' level of transparency. Our sample is restricted to MFIs having at least three diamonds (1,355 observations out of a total of 1,407 MFIs). 10 In microfinance this basic capitalistic principle is somewhat mitigated by the fact that the shareholders of listed MFIs include charitable institutions and socially responsible investors who are ready to tolerate some financial sacrifices in exchange for fulfilling a social mission. 11 Despite being widely used in practice, these indicators are not free from criticisms. For instance, average loan size is a poor indicator for at least two reasons. First, it is abusively penalizing for cross-subsidization and progressive lending (Armendariz and Szafarz, 2011). Second, its use can create a perverse incentive to offer disadvantageous credit conditions to some client segments, such as women and discriminated-against minorities (Agier and Szafarz, 2013).
Listed MFIs exhibit average loan sizes in a range comparable to those of their non-listed counterparts. In contrast, the interest rates charged by listed MFIs (above 50% a year on average) are higher, although not significantly so, than those charged by non-listed MFIs (between 24% and 38% a year on average). The high level of interest rates charged to poor borrowers is a controversial issue in microfinance, especially from an ethical standpoint (Hudon, 2007). Practitioners argue that high interest rates are inevitable without subsidization, mainly because the operational costs associated with managing small loans are high (Fernando, 2006). Strikingly, listed MFIs serve 78% of female borrowers on average, which places them slightly above their non-listed counterparts (from 50% to 76%). Again, the differences are hardly significant. Overall, the social performances of listed MFIs do not depart from those of the rest of the industry. IPOs per se do not seem to entail mission drift. This is a promising sign for the future of microfinance, as commercialization currently represents a strong trend in the industry (Bogan, 2012) and new IPOs are to be expected in the near future.
Listed MFIs represent less than 0.1% of world market capitalization. Relative to their domestic market, however, most microfinance stocks are far from negligible. To get a clear picture, Table II reports data on market capitalizations for all finance and microfinance stocks in our dataset, as well as the shares of their countries in the world's emerging stock markets.

All market capitalizations are computed from Morgan Stanley Capital International (MSCI)
indices. In the countries where microfinance has listed stocks, we give the capitalization of these stocks with respect to both their domestic finance-plus-microfinance (F+MF) sector (Table II,  and accounts for a notable 14% of the domestic stock market. To obtain a scaling perspective, Table II also provides the market shares of the countries at stake in F+MF sectors in the worldwide emerging-finance sector (column 7), and in the emerging-market universe (column 8). Together, the five countries under scrutiny account for 9.61% of the emerging financial sector and 12.4% of the emerging-market universe.
< Insert Table II  Compartamos to 79.6% for BFS) and high levels of extreme risks. The returns also display something rarely found in finance: they are nearly always asymmetrical to the right, with skewness as high as 6.5 for African Bank. At the same time, African Bank has an exceptionally fat-tailed distribution, with kurtosis of 73.5, and a maximum of 266% for one month and 233% for one day, both occurring in February 1995. All returns have been positively tested for stationarity. Table III here > 12 The history of the microfinance industry in South Africa is singular (see Porteous and Hazelhurst, 2004;Napier, 2006). After having experienced full deregulation in the post-apartheid period (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999), which enhanced commercial microcredit activities, the sector started to be supervised by the Microfinance Regulatory Council (MFRC) which is "entrusted with the responsibility of regulating the activities of the micro lending sector and to protect consumers against deceptive and unfair lending practices in terms of the Usury Act Exemption Notice (…) of June 1999." (http://www.dti.gov.za/thedti/mfrc.htm). 13 Rhyne (2009) mentions that Equity boasts over a million small savers and was recognized as the best bank in Kenya by Euromoney in 2007.

< Insert
To estimate the local CAPM and the international Fama-French-Carhart model in Section 4, we use country stock indices from MSCI and the four standard global factors (market, size, value and momentum) in USD retrieved from Kenneth French's website. 14 More precisely, the country indices are quoted in domestic currencies and encompass 21 stocks in Mexico, 45 in South Africa, 22 in Indonesia, and 7 in Kenya. 15 The industry composition of each country index reflects the specific features of the local market. Data for interest rates (three-month interbank rates) in each country and exchange rates are from Datastream.
To compare monthly changes in microfinance and finance stocks, we construct original capitalization-weighted indices for both the microfinance and the financial sectors. We proceed in the following way. First, for each country we create a local microfinance return index starting when at least one MFI is quoted in that country. These indices take into account frequently. At the end of the sample period (December 2010), the GMI is composed of nine stocks from five countries with the following geographic weights: 63% for Indonesia, 18% for South Africa, 13% for Mexico, 4% for Kenya, and 1% for Bangladesh. 14 Available at: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html The global factors include 23 countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Hong Kong, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Singapore, Spain, Switzerland, Sweden, United Kingdom, and United States. 15 Bangladesh is excluded from the universe for CAPM estimations due to unavailability of interest rates. Figure 1 charts the movements in the local microfinance indices (in local currency), each normalized to 100 at its starting date. Importantly, the South African index experienced such exceptional growth over the period that a different scale is needed to represent its change (axis on the right side of Figure 1). Figure 1 also charts the change in the GMI (in USD). This graph should be interpreted with caution at the beginning of the sample period because of the lack of geographical diversification.
< Insert Table IV here > with the equivalent MSCI emerging indices, for which volatility ranges from 20.6% for South Africa to 35.9% for Indonesia). Extreme risks are also much higher: skewness is highly positive (from 0.39 for Bangladesh to 6.59 for South Africa), as it is for individual MFIs (except for the Mexican index, where Compartamos is the only MFI to exhibit left-skewed asymmetrical returns); and kurtosis is very high (up to 74.3 for South Africa) for national microfinance indices, except for Bangladesh and Mexico (which has kurtosis of nearly 3). The GMI index has an average annualized monthly return of 20.3% and volatility of 56%, and more moderate extreme risks (skewness of 0.5 and kurtosis of 5.91).
For the finance sector, we construct local and global indices by mimicking the construction of the microfinance indices. More precisely, in each of the five countries under consideration, the local finance index is built from the financial stocks belonging to the corresponding MSCI universe, but excluding microfinance. Consequently, at the end of the sample period the local financial indices are composed of stocks from the following numbers of banking institutions: 5 in South Africa, 3 in Kenya, 3 in Indonesia, 11 in Bangladesh, and 2 in Mexico. Again, each index is weighted by market capitalization. The number of financial institutions in Bangladesh is striking. This is because the financial sector is particularly well developed compared with other sectors in that country (Demirgüc-Kunt and Levine, 1999).
Bangladesh experienced a major financial sector reform, initiated by the World Bank at the beginning of the 1990s and pursued by the government after 1996, with a view to expanding and diversifying the sector and privatizing national banks (Uddin and Hopper, 2003).
The Global Finance Index Comparator 16 (GFI-C) aggregates the five local indices.
However, to allow rational comparisons with the GMI, the country weights in the GFI-C are constrained to be those of the GMI. Specifically, the weight of each country in both the GMI and the GFI-C is dictated by the size of its microfinance sector (converted to USD).
Interestingly, on the Indonesian stock market the microfinance sector (present since 1990) predates the banking sector (present since 1996). As a result, the GFI-C can be defined only for the period starting in November 1996. Average annualized monthly returns range from 22.5% for South Africa since 1990 and 33.5% for Kenya since 1991. However, financial indices' volatility (from 25.9% for South Africa to 51.6% for Indonesia) is much lower than for microfinance indices. These results also apply to the GFI-C index, whose average annualized monthly return (22.5%) is slightly higher than that of the GMI index (20.3%), and with lower volatility (47.5% versus 56%).

Joint Movements of the Finance and Microfinance Global Indices
To facilitate comparison of our two global indices, the GFI-C for finance and the GMI for microfinance, a common base was set at December 1996. The graph of daily cumulative returns of the GFI-C and GMI ( Figure 3) shows that after a period of great disparity between finance and microfinance, with higher instability for microfinance, a phenomenon of convergence appeared. In fact, the correlation between the GMI and the GFI-C rose from 33% over the first half of the sample period (until December 2003) to 79% during the second half.
Volatilities for the two series also differ by sub-period (initially 53% and then 30% for finance, and 76% and 34% for microfinance). The Engle and Sheppard (2001)  To describe the joint movements of the GFI-C and GMI indices, we adopt DCC-MVGARCH 18 modeling (Engle and Sheppard, 2001;Engle, 2002), which enables us to factor in dynamic conditional correlations. This approach is often used to model correlation dynamics between financial series (Kearney and Poti, 2006;Brière and Signori, 2009 Following the conventional approach (e.g., Engle, 2002;Chiang et al., 2007), the univariate GARCH processes for GFI-C and GMI are modeled using a GARCH (1,1) specification of the form: , and the   In the DCC model, the kk  time-varying covariance matrix of   t u denoted by t Q fulfills: where A and B are non-negative parameters satisfying 1 AB  . The proper correlation matrix t R is given by: We follow Engle's (2002) two-step log-likelihood estimation procedure for the DCC.  A and B). With reference to parameter significance and information criteria, the best model is unambiguously the GARCH (1,1) for both series, which is also the most frequent specification for financial returns.
< Insert Table V here > The coefficients of the lagged variance and innovation terms are highly significant, which is consistent with time-varying volatility and the appropriateness of the GARCH (1,1) specification. Both GARCH (1,1) univariate processes present a high degree of persistence (long memory), signaled by i i    being close to 1, which is even higher for GMI (0.996) than for GFI-C (0.986). Figure  What caused the convergence between quoted microfinance and mainstream finance?
Evidently, it is more difficult to interpret analyses based on our indices than it would be for the individual stocks. In particular, post-IPO stock returns are known to exhibit abnormal features such as serial correlation and high volatility driven by the investors' learning process (Lowry and Schwert, 2002;Lowry et al., 2010). In microfinance, these features are likely amplified by the novelty of the activity, which makes learning more intense. We therefore contend that convergence is attributable, at least partly, to investors' learning mechanism. In addition, going public affects the microfinance business model and makes MFIs more like 20 Patten et al. (2001) also exhibit the good performances of BRI during the East Asian crisis. banks. IPOs probably represent the culmination of the commercialization trend occurring throughout the microfinance sector (Christen and Drake, 2001). This trend is associated in particular with savings mobilization, an activity typical of mainstream banks (Woller, 2002).
In sum, both sides of the market offer reasonable arguments to rationalize the convergence process. Interestingly This can be seen as evidence that convergence is still at work rather than already achieved.
Regardless of the precise date of convergence, the reduction in volatilities at the aggregate level is probably due to the diminishing weight of recent IPOs in our two indices.
In conclusion, the microfinance sector has been trending towards gradual integration into mainstream finance. However, it has retained some specific traits. Microfinance tends to develop in countries where the financial sector is relatively weak (Maksudova, 2010;Vanroose and D'Espallier, 2013), 21 such that the regional distribution of listed MFIs differs from that of the traditional financial sector. From this point, although the two sectors are converging, the potential for microfinance to provide diversification in an equity portfolio can be distinguished from the diversification potential of finance by regional bias. Section 4 considers the nature of risks assumed by domestic and international investors in finance and microfinance equities.

Risk Factors of Microfinance Investment
The profitability of the microfinance sector is hotly debated. 22 For instance, Cull et al. (2009) state that: "The evidence suggests that investors seeking pure profits would have little interest in most of the institutions that are now serving poorer customers" (p. 169). Schmidt (2010) is even more pessimistic about the potential for profitable investment in microfinance: "(...) I fear that the high expectations regarding the return on an investment in MFIs, which I consider to be exaggerated, will have a negative impact on activities in the microfinance sector (...)" (p. 125). 23 In this section, we revisit the issue by investigating the sensitivity of both microfinance and finance stock returns to market and foreign-exchange risks. As shown by Solnik (1974) and Adler and Dumas (1983), foreign-exchange risk plays a key role in international asset pricing. Moreover, Crabb (2004) stresses that foreign-exchange remains an important source of risk for MFIs, which are exposed to the devaluation of their funding sources. Two perspectives are analyzed in turn. First, we consider the situation of a domestic investor in a country where microfinance equity is available. Second, we take an international investor contemplating sector-wise investment in finance and microfinance indices.
To address the domestic investor's situation, we follow Harvey (1995) and estimate a CAPM-type model in local currency including two factors: a standard market factor and a foreign-exchange factor. Owing to the unavailability of interest rate data for Bangladesh, the following model is estimated for four countries (Indonesia, 24 Kenya, Mexico, and South Africa): where i t USD R , is either a stock return or an index (domestic microfinance, domestic finance, GMI, or GFI-C) return. different from zero. With a single exception (BFS in South Africa), the loadings on the domestic market factors are all significant, both for microfinance and for finance stocks, lying between 0.63 (Indonesia) and 1.55 (Equity in Kenya) for microfinance, and between 0.39 (Gfinbur in Mexico) and 1.10 (Commercial Bank in Kenya, and Gfnrote in Mexico) for finance. In general, market betas are higher for microfinance than for finance, signaling a higher systematic risk for MFIs than for traditional banks. However, the difference is hardly significant. The only borderline exception is Kenya, where microfinance exhibits a smaller market beta (0.72) than finance (0.95), the difference being significant at the 10% level. For all countries, R-squared values are relatively low (between 19% for South African microfinance and 55% for Indonesian finance), which is a typical feature of market beta estimation in emerging countries (Harvey, 1995).
Exposure to the currency factor produces the intuitively expected result: Betas are negative in most cases. However, many of them are not significant, at least at the stock level. Interestingly, at the sector level, three out of four currency betas are significant for finance and two out of four for microfinance. Banks and MFIs often fund their portfolios through debt in foreign currency, especially USD (Crabb, 2004). When the dollar appreciates, financing becomes more expensive but the institutions' revenues (from repayment of loans contracted in local currency) remain fixed, thus penalizing them. Only a few MFIs are not exposed to foreign-exchange risk, either because they operate in a fully dollarized economy, or because they solely trade in local currency. Foreign-exchange betas are not significantly different for the finance and microfinance sectors, meaning that the two types of institution share similar exposures. Table VII here >  Table VII provides the international FFC estimation results. The factor loadings (market, size, value and momentum) are estimated based on Eq. (4). As a robustness check, we estimate the FFC regression for the global finance and microfinance indices. The results reveal that, except for Bangladesh, most market betas are significant and higher than the beta for both microfinance and finance. This likely reflects the higher systematic risk of equity from emerging countries compared with a well-diversified world portfolio balanced between emerging-and developed-market stocks. Market betas are higher for microfinance than for finance. In South Africa, the HML factor loading is positive and strongly significant for finance stocks but mostly insignificant for microfinance ones. This is in line with the observation from the report by O'Donohoe et al. (2009, p. 33) indicating that quoted microfinance "trades at a premium on a P/BV basis over traditional banks, though this premium has declined considerably since its peak in November 2007." The size and momentum betas exhibit little regularity. Except for Mexico, the Wald tests detect no difference between estimates for finance and microfinance stocks.

< Insert
We check the stability of our results by relying on the testing methodology proposed by Fabozzi and Francis (1977). Each sample is split into two sub-samples of equal length. In each subsample, both the CAPM and the FFC model are estimated as previously (results not reported), and the factor loadings are compared. Then, we perform two groups of tests: joint Wald tests on the stability of all factor loadings, and then individual Wald tests on the stability of each factor loading. Table VIII summarizes the outcomes of all these tests. To facilitate readability, the table reports for each stock/index: 1) the conclusion of the joint test at the 5% level, and 2) the number of unstable betas detected by individual tests at the 5% level.
Remarkably, the estimated betas take quite similar values over the two sub-periods. Pubali Bank in Bangladesh is the only stock for which instability is consistently detected. On the whole, we detect 17 unstable cases by means of 352 individual tests. This may reasonably be interpreted as evidence of overall coefficient stability given that all tests are performed at the 5% level. Table VIII here > Summing up, both the finance and the microfinance sectors exhibit high market betas, locally and globally. Moreover, domestic investors in both sectors are significantly exposed to foreign-exchange risk. Lastly, that microfinance is converging towards mainstream finance is confirmed by the proximity of their market and foreign-exchange betas in domestic markets.

Diversification Benefits
Spanning tests examine whether including an external asset in a portfolio of benchmark assets shifts the pre-existing efficient frontier to the left. Here, we use the mean-variance spanning test 27 proposed by Huberman and Kandel (1987) to assess the diversification benefits of microfinance stocks for optimal portfolios of financial stocks. This test involves running OLS regressions of the returns of the external asset, E R , on the returns of the K benchmark assets, , 1, , k B R k K  , as follows: Under the null hypothesis of spanning, constant  is equal to zero and the sum of the k  coefficients is equal to one.
We consider two different scenarios. The first, and most general, concerns the situation of a U.S. investor whose pre-existing portfolio consists of financial stocks from the 21 countries represented in the MSCI Global Finance Index (GFI). We successively check whether each country's microfinance stocks and our global microfinance index (GMI) are spanned by this universe. Second, we assess the diversification benefits of microfinance stocks for national portfolios. In each country where microfinance stocks are available, we consider the situation of a domestic investor and contemplate the possibility of adding each local microfinance stock to a pre-existing portfolio composed of local financial stocks only.
< Insert Table IX here >

Conclusion
Despite the impressive development of the microfinance sector, the financial performance of microfinance equity is poorly understood. This is probably due to data availability issues.
Although suffering from data limitations, this paper takes full advantage of daily quoted prices of microfinance stocks since their issuance, and draws a global picture of worldwide microfinance equity. Two main messages stand out. First, we have demonstrated that microfinance had largely converged on the mainstream financial sector. This is consistent with the evidence that the MFIs that issue stocks are those that behave most like banks without really intending to serve the poorest of the poor (Cull et al., 2007). Although the situation is far from homogenous across countries, the picture that emerges from factor analysis is consistent with the convergence outcome. Moreover, we have confirmed the intuition that MFIs and banks are both highly exposed to exchange rate risk, probably because they obtain most of their funding from international capital markets.
Second, spanning tests show that the diversification potential of microfinance stocks within the financial sector is minimal. Microfinance equity may be disregarded at virtually no cost by profit-oriented investors seeking new investment opportunities in developing countries. Obviously, this evidence does not hold for investors who are concerned (even slightly) by social outcomes in general 28 and financial access for the poor in developing 28 Microfinance is now seen as part of a broader asset class called "impact investment" designed to yield both a financial return and a broader benefit to society ("Impact Investing: Happy returns, the Emergence of a New Asset Class", The Economist, September 2011). countries in particular. In this respect, listed microfinance can be a halfway solution between a purely profit-oriented investment and NGO-type financing through donations/subsidies or subsidised loans. This paper delivers an encouraging message to those investors, namely that listed microfinance equity generates social performances on a par with those of its unlisted counterpart. Even though interest rates are higher on average, listed MFIs grant loans that are similar in size to those of other MFIs, and they serve the same proportion of women.
Accordingly, IPOs do not seem to be a major reason for mission drift.
A seminal contribution in many respects, this paper also suffers from econometric drawbacks. The main problem likely relates to the underlying probability distributions of the returns. As the descriptive statistics have amply shown, we are dealing with series that depart significantly not only from normality, a common feature of most financial series, but also from the typical heavy-tail distributions that financial econometricians are used to dealing with. Moreover, MFIs have very different size, geographic location, and other characteristics compared with the entire financial sector. Finding a more carefully-matched sample might yield more insights. We therefore view our empirical results as challenging but still preliminary, and we hope they will motivate further studies in this area. Indeed, the quoted microfinance sector is still in its infancy, and a clearer understanding of what drives its performance is likely to emerge with time.
Lastly, emerging financial markets have been documented as a key source of diversification for Western portfolio holders (Bekaert and Harvey, 2003;Quisenberry and Griffith, 2010). However, the optimal composition of portfolios made up of emerging-country stocks remains largely unexplored, especially with respect to their sensitivity to global crises.
In this respect, one promising avenue for research is the way sector-specific and/or countryspecific investments could help robustify global portfolios (Brière and Szafarz, 2008;Brière et al., 2010). For this, adequate sector delineation is a prerequisite. Our paper has also taken steps in that direction.
Hermes, N., Lensink, R., and Meesters, A. (2009)  This table reports the average financial and social characteristics of listed and non-listed MFIs, except for debtto-equity (D/E) ratio where we provide the median. "Average loan size" is the average loan divided by the gross national income per capita. The interest rate charged on loans is proxied by the nominal yield on gross portfolio. Standard deviations are in parentheses.

and market shares of listed finance and microfinance institutions
The table presents market capitalizations in million USD at the end of the sample period (Dec 31, 2010) and market shares of listed finance (F) and microfinance (MF) institutions. "% in domestic F+MF" is the market share of the institution in its domestic F+MF sector (MSCI Finance Country index). " % in domestic market" is the market share of the institution in its domestic market (MSCI Country index). "% F+MF in finance emerging" is the market share of the country's F+MF sector in the World emerging-finance sector (MSCI Emerging Finance index). "% country in emerging" is the market share of the country in the World emerging market (MSCI Emerging index

Figures
"r" indicates that the series is displayed on the right axis. "r" indicates that the series is displayed on the right axis.