DOES RATING HELP MICROFINANCE INSTITUTIONS RAISE FUNDS? A CROSS-COUNTRY ANALYSIS OF THE ROLE OF RATING AGENCY ASSESSMENTS IN MICROFINANCE INDUSTRY Except when reference is made to the work of others, the work described in this thesis is my own or was done in collaboration with my advisory committee. This thesis does not include proprietary or classified information. ______________________________________ Qianfei Wang Certificate of Approval: _________________________________ _________________________________ James R. Barth Valentina M. Hartarska, Chair Lowder Eminent Scholar Assistant Professor Professor Agricultural Economics and Finance Rural Sociology _________________________________ _________________________________ Patricia A. Duffy Joe F. Pittman Associate Professor Interim Dean Agricultural Economics and Graduate School Rural Sociology DOES RATING HELP MICROFINANCE INSTITUTIONS RAISE FUNDS? A CROSS-COUNTRY ANALYSIS OF THE ROLE OF RATING AGENCY ASSESSMENTS IN MICROFINANCE INDUSTRY Qianfei Wang A Thesis Submitted to the Graduate Faculty of Auburn University in Partial Fulfillment of the Requirements for the Degree of Master of Science Auburn, AL December 17, 2007 iii DOES RATING HELP MICROFINANCE INSTITUTIONS RAISE FUNDS? A CROSS-COUNTRY ANALYSIS OF THE ROLE OF RATING AGENCY ASSESSMENTS IN MICROFINANCE INDUSTRY Qianfei Wang Permission is granted to Auburn University to make copies of this thesis at its discretion, upon request of individuals or institutions at their expense. The author reserves all publication rights. ______________________________ Signature of Author ______________________________ Date of Graduation iv VITA Qianfei Wang, daughter of Wei Wang and Jian Zhang, was born in January, 1979, in Shenyang, LiaoNing Province, China. She graduated from Shenyang the 20 th High School in 1997. She attended Nankai University, Tianjin, China the same year, and graduated with a Bachelor of Art degree in Japanese Language in 2001 and a Master of Science degree in Management in 2004. She entered Graduate School, Auburn University, in January 2006 where she held a graduate research assistantship. She married to Zhigang Li in 2004. v THESIS ABSTRACT DOES RATING HELP MICROFINANCE INSTITUTIONS RAISE FUNDS? A CROSS-COUNTRY ANALYSIS OF THE ROLE OF RATING AGENCY ASSESSMENTS IN MICROFINANCE INDUSTRY Qianfei Wang Master of Science, December 17, 2007 (B.A., Nankai University, China, 2001) 62 Typed Pages Directed by Valentina Hartarska This thesis studies whether rating assessments help MFIs raise more funds, using a sample of 315 MFIs operating in 63 countries worldwide during the period of 1999-2005. The main conclusion is that there is no strong evidence that rating affects MFIs fundraising efforts, after accounting for possible endogeneity of rating by two-stage least square procedure, although rating has statistically significantly positive effects on the change of non-deposit liability and total equity of MFIs by OLS procedure. However, the evidence for endogeneity of rating is not strong, which might support the idea that rating helps impose market discipline. Difference of rating impact on raising liability between vi raters is detected. Rating updates, subsidized rating and rater types do not affect fundraising and regulatory authorities? activities do not affect the values of rating. vii ACKNOWLEDGEMENTS I would first like to thank my family for their love, support, and patience throughout this journey. Thanks to Department of Agricultural Economics and Rural Sociology in Auburn University for giving me the opportunity to pursue a graduate study and to receive a graduate research assistantship. Special thanks will be given to my major professor, Dr. Valentina Hartarska for guiding me through my graduate study and having faith in my abilities to complete it. Thanks will also be extended to Dr. James Barth and Dr. Patricia Duffy for their guidance, encouragement and advices through this project. viii Style Manual of journal used American Economic Review Computer Software used Microsoft Word 2003 and Stata 9.1 ix TABLE OF CONTENTS LIST OF TABLES .............................................................................................................. x CHAPTERI: INTRODUCTION ........................................................................................ 1 CHAPTERII: MOTIVATION AND RELATED LITERATURE ....................................... 4 CHAPTERIII: EMPIRICAL MODEL.............................................................................. 14 CHAPTERIV: DATA DESCRIPTION ............................................................................. 22 CHAPTERV: RESULTS................................................................................................... 30 CHAPTERVI: CONCLUSIONS ...................................................................................... 43 REFERENCE.................................................................................................................... 45 APPENDIX....................................................................................................................... 48 x LIST OF TABLES Table 1: Introduction of rating methodology for rating agencies .............................. 10 Table 2: Definition of variables ................................................................................. 18 Table 3: Sample distribution by year ......................................................................... 22 Table 4: Distribution of sample MFIs and raters by geographic region .................... 23 Table 5: Brief description of MFI rating agencies and sample distribution............... 24 Table 6: Summary statistics ....................................................................................... 26 Table 7-1: Summary statistics for non-rated group and rated group ......................... 27 Table 7-2: Summary statistics for N_GRADE by raters............................................ 27 Table 8: Mean comparison of selected indicators between rated group and MBB ... 28 Table 9: The impact of rating on change in non-deposit liability and equity ............ 31 Table 10: The impact of rating on change in non-deposit liability and equity by raters and regions .................................................................................. 33 Table 11: The impact of rating on change in non-deposit liability and equity by two-stage least square procedure .......................................................... 37 Table 12: Heckman Selection Model to predict numerical grades ............................ 39 Table 13: Summary statistics for PHECKMAN and actual N_GRADE ................... 40 1 CHAPTER I: INTRODUCTION Microfinance is the provision of financial services to low-income clients, traditionally exclude from the mainstream financial system. Microfinance Institutions (MFIs) are the financial institutions serving poor households and small enterprises (Hartarska, 2005). Worldwide, MFIs expand the frontier of finance by providing loans and other financial services to the under-served poor. According to Daley-Harris (2006), as of December 2005, 3,133 Microcredit Institutions reported reaching 113.3 million clients, which is 22.8 percent annual growth rate since end of 2004. Today, MFIs are not a rarity any more, but are important members of the financial systems in developing countries. The growing relevance of MFIs has lead to the development of specialized microfinance rating agencies that perform global risk assessments and credit rating for MFIs. As the microfinance industry matures, microfinance rating services have gained increased attention from investors, practitioners, and donors. However, MFIs still find it challenging to obtain funds from prospective donors and financial markets. Rating is expected to play a key role in helping MFIs to improve performance and to help them access commercial capital. Rating generates independent information and could improve efficient allocation of funds. Rating may be especially important in the absence of developed equity and debt 2 markets for microfinance and because there are few alternative mechanisms that can help donors and investors choose the most appropriate MFIs to support. On the other hand, rating may have a limited role since microfinance rating agencies have little competition and there are neither recognized standards nor a consistent rating system. It is not yet clear whether rating plays a disciplining role and, in particular, whether good rating helps MFIs raise funds. This paper uses a database of 315 MFIs across 63 countries and studies the role of rating agency assessments of MFIs. It adopts an empirical approach used in studies on the impact of market forces on performance of financial intermediaries and accounts for possible endogeneity of rating. Consistent numerical rating grades by raters are created in this thesis, which helps capture more information of rating itself. The main conclusion is that there is no strong evidence that MFIs with better ratings were more likely to raise funds, after accounting for possible endogeneity of rating by two-stage least square procedure, although better rated MFIs are more likely to raise funds according to the OLS procedure. However, the evidence for endogeneity of rating grade is not strong, which might support the notion that ratings possibly help impose marketing discipline. Difference of ratings impact on raising liability between raters is detected. Also, evidence is found that NGOs have lower return of rating than profit financial institutions. Rating updates, subsidized rating and the type of rater do not affect fundraising and regulatory authorities? activities do not affect the values of rating. The rest of the paper is organized as follows. The next section of the thesis briefly surveys the literature in rating as a governance mechanism and in microfinance industry. Chapter III presents the empirical specification and discusses methodological issues. 3 Chapter IV describes the data used. Chapter V presents the empirical results by OLS and two-stage procedures accounting for possible endogeneity of rating. Heckman Selection Model is also presented as an extension. The conclusions of the thesis with remarks on possible future research are summarized in the final chapter VI. Some additional empirical tests and informative lists are presented in appendix section. 4 CHAPTER II: MOTIVATION AND RELATED LITERATURE Cross-country empirical studies on the role of credit ratings on MFIs are rare. Hartarska&Nadolnyak (2007b) study whether microfinance rating agencies were able to impose market discipline on MFIs during the period 1998-2002. The results indicate that subsidizing rating did not help MFIs raise more funds and not all rating agencies had equal impact on MFI abilities to raise extra funds ? while some helped MFIs raise funds, others did not. However, the Consultative Group to Assist the Poor (CGAP), the Interamerican Development Bank and the European Union have established and still support a Rating Fund which is a special fund that subsidizes rating of MFIs. Since 2002 microfinance raters adopted a numerical rating scales but the impact of letter grade rating has not yet been examined. This thesis focuses on whether microfinance rating agencies were able to impose market discipline on MFIs, examining the effectiveness of rating assessments post 2001 after letter grade scales were introduced. It creates a consistent rating scale with numerical values from the diverse rating scales of all rating agencies and uses this index to study whether the rating an MFI obtained helped fundraising. This section reviews the literature on rating as an external governance mechanism and provides an overview of microfinance rating in the context of general rating practices. 5 The empirical approach explores the impact of rating on MFIs? ability to raise additional funds and is based on a similar approach employed in the study of bank performance. 1. RATING AS AN EXTERNAL GOVERNANCE MECHANISM A credit rating assesses the credit worthiness of an individual, corporation, or even a country. Credit ratings are calculated from financial history and current assets and liabilities. Typically, a credit rating tells a lender or investor the probability of the subject being able to pay back their debt. In recent years, credit ratings have also been used to adjust insurance premiums, determine employment eligibility, and establish the amount of a utility or leasing deposit. There are three types of credit ratings: personal credit ratings, corporate credit ratings and sovereign credit ratings. Corporate credit rating had its origins in the first rail-bond ratings of John Moody, who started Moody?s Investors Service in 1900. Today, a credit rating is both necessary for access to capital and a key determinant of the price of funding for many companies in developed capital markets. In most countries credit ratings have weight for regulators as well as for the capital markets, and banking supervisors often require them to determine such things as deposit insurance and minimum capital requirements (Farrington, 2005). Loffler (2004) explores the usefulness of credit rating agencies? rating systems and their rating migration policies as tools for formulating governance rules. Such rules, which consist of buy and sell restrictions, are predicated on rating stability and are commonly used in investment management. Loffler suggests there may be many circumstances where credit ratings-based governance rules may be more effective. More 6 generally, he finds that many statistical measures that are currently used to assess rating quality may be insufficient to judge the economic value of rating information in a specific context. Arnoud (2006) has shown that credit ratings could play a key role as ?focal points? once institutional rigidities are considered. This paper shows that credit rating can coordinate investors? beliefs and together with the implicit contract and monitoring relationship between credit rating agencies and the firm and ratings have a real impact. Some studies focus on the difference between unsolicited ratings and solicited ratings. Roy (2006) finds evidence that unsolicited ratings tend to be lower than solicited ones, after accounting for differences in observed bank characteristics by using a sample of Asian banks rated by Fitch Ratings. This downward bias does not seem to be explained by the fact that better-quality banks self-select into the solicited group, rather, unsolicited ratings appear to be lower because they are based on public information. Poon (2001) finds evidence of a significant difference in distributions between solicited and unsolicited ratings and that the ratings of the unsolicited group tend to be lower for the overall sample and the matching sub-sample, by employing S&P?s sample rating from pooled time-series cross-sectional data of 265 firms from 15 countries. The paper also indicates that the relatively lower unsolicited ratings may not be caused by downward bias in assigning ratings because this bias can be due to differences in rating standards and the scales used by rating agencies, or systematic differences in their rating procedures or by self-selection issues or other latent factors. Ferri (2001) examines the behavior of issuer ratings in developing countries, and find that bank and corporate ratings appear to be strongly related in an asymmetric way with changes in sovereign ratings. Bongini (2002) study the power of credit ratings to 7 predict bank insolvency in developing countries. Morgan (2002) examines the banks opacity from the lack of consensus among main rating agencies. Morgan (2000) investigates the disciplinary role of markets using bond spreads, ratings, and bank portfolio data on over 4,100 new bonds issued between 1993 and 1998, and finds that the bond spread/rating relationship is the same for the bank issues as for non-bank issues, especially among the investment grade issues. This suggests the bond market prices incorporate publicly available measures of bank risk efficiently. Bond rating agencies focus on predicting risk and are able to predict future problem loans and bank performance (Berger et al., 2000). Equity studies also find that investors promptly incorporate relevant rating information into bank stock prices (De Young et al., 2001). Morgan (2002) finds evidence that external rating exercises market control for both banks and non-financial institutions by providing independent information. But because financial institutions are generally regulated and supervised by the authorities, external rating usually serves to help regulators strengthen the banks. 2. RATING IN MICROFINANCE In microfinance, governance refers to the mechanisms through which donors, equity investors, and other providers of funds ensure themselves that their funds will be used according to the intended purposes (Hartarska, 2005). Such control mechanisms are necessary and important because MFI managers may have different objectives from the providers of funds. Hartarska (forthcoming) finds that regulatory involvement and external audit do not impact performance but that rating may hold the potential to play a disciplining role in microfinance. Credit rating agencies may serve as an external 8 governance mechanism because they operate as independent market entities and exercise market control by providing signals about the quality of financial firms and their debt (Barth et al., 2006). Hypothesis 1: rated MFIs get more funds than MFIs whose performance is not rated. Hypothesis 2: regulation, audit and rating play a role to impose market discipline as external governance mechanisms in microfinance. Many MFIs operates as NGOs. Manne (1999) proposed that governance mechanisms are much weaker in the nonprofit sector due to the absence of shareholders and a market for corporate control. Also, however, it indicates that ratings serve as control mechanisms of NGOs. Hypothesis 3: the impact of rating in NGOs differs from that in for-profit institutions. There are two types of rating services that perform evaluations for MFIs: credit risk ratings and global risk assessments. Credit risk ratings are provided by both microfinance and traditional (mainstream) rating agencies. Global risk assessments, or performance evaluations, are provided mainly by agencies which are active exclusively in the microfinance field as specialized agencies. Global risk assessments provide a more comprehensive picture of an MFI?s performance level. These performance evaluations, in comparison to more traditional credit risk ratings, place more weight on operational elements such as appropriateness of lending methodologies and governance issues. Ratings produced by specialized credit agencies allow an MFI to be compared with other MFIs. 9 The idea of creating a specialized rating agency for MFIs emerged in 1996, as a result of the pilot study for MicroRate. Theoretically, the goal of specialized rating was to create an enabling market mechanism that would reduce information asymmetries, help bring perceived risks in line with actual risks, and increase capital flows to the emerging microfinance sector (Farrington, 2005). It has been shown that the total number of ratings and assessments has grown by 84% in the past four years, from 152 in 2001 to 281 in 2005 (The Rating Fund Market Survey 2005). Currently, 16 rating agencies are active in this market, suggesting that the industry has reached a certain level of maturity and that there is a need to identify if and how rating can serve as an effective mechanism of market discipline. Market forces through the market for managers and through the market for takeovers have a limited role in microfinance because the market for MFI managers is thin and most MFIs do not have true owners. This forces donors and creditors are increasingly relying on information from audited financial statements and rating agencies (Hartarska, 2005). The main objective of such external governance mechanisms is to reduce information asymmetries between the different stakeholders and the firm (Healy&Palepu, 2001). Empirical work has touched the factors explaining the rating of MFIs. Nieto&Cinca (2005) find a positive and significant relationship between rating and profitability, rating and size, rating and productivity, a negative and significant relationship between rating and risk, while no significant relationship between rating and social performance. These findings are consistent with theory, enhancing the value of ratings for investors. The evidence of the impact of external governance mechanisms on MFI financial 10 performance is scarce (Hartarska and Holtmann, 2006). One of the seminal studies using empirical data of MFIs rating is Hartarska (2005). She finds that external governance mechanisms such as auditing, rating, and regulation, have a limited impact on outreach and sustainability of microfinance institutions in Eastern Europe and the Newly Independent States. The evidence that rating improves performance is scarce. In this study, rating is significant at the 10% level only in the outreach regression but not in the regressions on sustainability variables. It is mentioned that although these results can be idiosyncratic to the sample period when rating in the region was not widely used and usually was donor mandated, they are consistent with some developments in microfinance, with raters struggling to survive perhaps because they failed to become effective external governance mechanism. Furthermore, the author notes that more research and better data is needed to ensure that strong organizations direct scarce resources to the entrepreneurial poor. Rating methodologies differ significantly across raters. Table 1 provides a brief description of the rating methodologies of 13 rating agencies. It reveals that different rating agencies use different methods to assess MFI financial performance. Table 1: Introduction of rating methodology for rating agencies Agency Analysis Apoyo&Asociados It issues a report containing information about: equity performance, credit risk, funds diversification, market situation, operational and technological risks, management and ownership, and future trends. Accion Interational It has adapted the CAMEL rating methodology to perform global risk assessments of MFIs. The CAMEL methodology assesses 21 indicators under 5 areas: Capital adequacy, Asset quality, Management, Earnings and Liquidity management. CRISIL It has developed the MICROS methodology, with six indicators: Management 25%, Institutional Arrangement 15%, Capital Adequacy & Asset Quality 20%, Resources 10%, Operational Effectiveness 15%, and Scalability & Sustainability 15%. 11 Agency Analysis ClassRating The assessment of bonds, debt, shares and financial strength (global risk assessment) of financial institutions takes 5 steps: information analysis, solvency analysis, liquidity analysis, issuer's contract analysis and final classification. Equilibrium It performs a quantitative analysis, focused on asset quality, capital adequacy, profitability, liquidity, balance sheet mix, funding strengths and weaknesses, cash flows, and so on. On the other hand, qualitatively, it assesses the management quality, business diversification and financial flexibility. Feller Rate The rating is based both in solvency classification and product?s own characteristics. For debt titles assessments, Feller examines guarantees, which can lead to different repayment capacities. FitchRating The rating is a comprehensive qualitative and quantitative assessment of strengths and weaknesses of the institution. Quantitative aspects e.g. balance sheet integrity, or profitability and risk management are counterbalanced by qualitative considerations about strategy, management quality, environment issues and future perspectives. JCRVIS It uses a methodology called MIRACLES, the acronym for Management, Information Systems, Reputation, Asset quality, Capital, Liquidity, Earnings and Supervisory systems (internal and external). MCRIL It uses a rating tool with three categories of indicators: governance and strategy, management systems, and financial performance. MicroRate For this agency, there is no unique criterion applying equally to all MFIs. It tries to identify this hierarchy correctly for each analysis. But the criteria ranked most frequently are: portfolio quality, operational effectiveness, management and governance. Microfinanza It performs a quantitative and qualitative assessment of strengths and weaknesses of the MFI, to grade the risk on two categories: fiduciary risk (related to governance and management) and credit risk (obligations e-payment ability). PlanetRating The rating is a comprehensive qualitative and quantitative assessment of strengths and weaknesses of the institution. Quantitative aspects e.g. balance sheet integrity, or profitability and risk management are counterbalanced by qualitative considerations about strategy, management quality, environment issues and future perspectives. Pacific Credit Rating The rating exercise studies quantitative and qualitative information. Qualitative aspects are considered very important and are based on fundamental principles. Complete analytic revisions are undertaken to assess the financial health of the institution. Then, future financial results are estimated, which will allow future rating revisions. Source: Nieto&Cinca (2007) ?Factors explaining the rating of Microfinance Institutions? Nonprofit & Voluntary Sector Quarterly. Among these rating agencies, Equilibrium, FitchRating, Microfinanza, PlanetRating and Pacific Credit Rating perform qualitative as well as quantitative analysis on MFIs. Qualitative assessments focus on MFIs strategies including weakness and strength 12 identification, management quality, environment issues and future perspective, which have special meaning for microfinance as a rapidly growing industry. Apoyo&Asociados, JCRVIS, MCRIL, MicroRate, Microfinanza and PlanetRating consider governance as a part of the evaluation in addition to assessing financial performance. Hypothesis 4: the impact of rating differs among different raters. MFIs are close to banks because they both collect deposits and provide loans. On the other hand, they differ from banks because a considerable part of the assets of MFIs comes from the donors, although some investors infuse funds through public capital market. Compared to regular banks, MFIs have lower level of capitalization, they rarely have publicly traded debt, and the asset base for MFIs comes from grants. Although more MFIs are transforming into commercial banks, overall, NGOs and non-bank financial institutions constitute the overwhelming majority of MFIs. Rating of MFIs differs from rating of regular banks. First, microfinance rating should, at least in part, consider the social performance of MFIs because MFI have a mission to provide financial services to the poor and must cover their costs to maintain sustainability. Three rating agencies MicroRate, PlanetRating and MCRIL claim to study MFIs social performance (Nieto&Cinca, 2007). For example, MicroRate proposes that striving to balance a clear and rational relationship among the social, financial and operational considerations of sound microfinance practice deserves a good grade in rating. Moreover, as MFIs rating agencies mature along with development of microfinance industry, the assessment of social performance and the development of specific methodologies to measure MFIs social impact have become increasingly important. Secondly, mainstream raters put more emphasis on credit risk and solvency, 13 benchmarking against the banking sector. Alternatively, specialized raters focus more on portfolio structure and quality and operational risk and efficiency, benchmarking against other MFIs. Hypothesis 5: the impact of MFIs specialized rating agencies differ from that of mainstream rating agencies. Finally, since many MFIs are regulated, regulatory involvement may affect the ability of rating agencies to discipline MFIs because implicit guarantees that an MFI can be recapitalized after bad performance might be provided by regulators, therefore, the value of the information offered by a rating agency may be diminished (Hartarska&Nadolnyak, 2007a). Moreover, donors can play the same role by deciding to recapitalize or provide grant(s) to a failing organization if it fulfils a mission important to them. Hypothesis 6: the role of rating is diminished for regulated MFIs. 14 CHAPTER III: EMPIRICAL MODEL Cross-country empirical studies on the impact of credit ratings on MFIs performance are rare. Hartarska (2005) finds rating as an external governance mechanism play a limited role on MFIs profitability and sustainability in Eastern Europe, perhaps because of unique transition environment. Hartarska&Nadolnyak (2007b) find that subsidized rating does not help MFIs raise additional funds but that the ability to raise funds is affected by who provides rating services. Therefore, it is important to examine how the rating scales used by different raters impact raising funds by MFIs. To explore the impact of rating on MFIs ability to raise additional funds, this paper adopts an empirical approach similar to the approach employed to study bank performance (Samolyk, 1994; Barth et al., 2003 and Hartarska&Nadolnyak, 2007b). This thesis uses the model developed by Hartarska&Nadolnyak (2007b). They argue that rating will play a disciplining role if it helps the MIFs raise additional funds. The null hypothesis formed by the authors is that rated MFIs were not able to raise additional funds either equity or borrowed funds. The alternative hypothesis is that after all information on the performance of the MFI is controlled for, rating still produces new information and, therefore, affects MFIs ability to raise additional equity or debt. In particular, the empirical model is: it ChF =constant+? ? t R +? ? 1?it B +? ? t M + it ? (1) 15 where the dependent it ChF is the log difference of the change in funds. Two dependent variables are used as it ChF . The first one is the log difference in borrowed funds other than deposits (LiabCh) and it is used to study the impact of rating on the ability of MFIs to attract loans. The second is the log difference in equity (EqCh), which captures the ability of MFIs to raise equity. The independent variables in 1?it B are MFI specific variables available from the financial statements and help isolate the direct impact of rating, captured in the t R variable so that t R is assumed to capture the impact of new information available only through rating but not available from other sources. The variables in 1?it B vector includes measures of financial performance, return on total assets (ROA) and operational self-sufficiency (OSS); a measure of outreach, number of active borrowers (NAB); the capital ratio as a measure of the level of leverage (CAPITAL); loans to total assets ration (LOAN) as a measure of the focus on lending; portfolio at risk (PA R ) as a measure of risk exposure; savings to total assets (SAVINGS); MFI size (MFISIZE) in terms of logarithm of total assets and MFI?s age (AGE). The increment of equity and non-deposit liability in the current period is influenced by the indicators of financial performance and structure one year back, so the values of t-1 period are used. Such specification avoids possible endogeneity 1 . Since MFIs can be non-profit NGOs, as well as banks and non-bank financial institutions, a dummy for non-profit type is included while the banks and nonblank financial institutions are treated as the reference group. t R is a vector of variables that represent rating. Several specifications are included in 1 OLS is unbiased and consistent in applying for recursive models due to the absence of endogeneity. 16 this thesis. First a dummy for being rated (RATING) is used, and alternatively, separate dummies for each specific rater. Dummies for individual raters are included in some selected models because empirical studies have shown that credit agencies differ in their evaluation of financial intermediaries (Morgan, 2002 and Hartarska&Nadolnyak, 2007b). Secondly, a numerical index that assigns values to each letter grade (N_GRADE) is used. It has a scale from 1 to 10. Higher value presents better rating. Numerical values were assigned based on the scale of letter grades issued by individual raters. N_GRADE is created to be consistent across raters since individual raters? grading scales are very different. Thirdly, a dummy variable for a rating update (UPGRADE) is used. This variable is important because it has been found that 49% of MFIs worldwide have rating updates (The Rating Fund market Survey 2005, 2006). Fourthly, a dummy for rater type (MFISPECIAL) is used, which equals 1 if the rating agency is a specialized MFI rating agency and zero otherwise. This variable is applied to detect whether there is a different effect between specialized MFI rating agencies and regular credit rating agencies. Finally a dummy for whether the rating is subsidized by the Rating Funds (SUBSIDIZED) is used to examine the effectiveness of subsidized rating because previous study shows that subsidized rating has no impact on raising funds. In their study, Hartarska&Nadolnyak (2007b) use dummy variables for being rated or not and for being rated by a particular rater because letter grades were introduced only in 2001. This thesis focuses on the role of letter grades issued by the raters. Detailed description and clarifications are presented in the next section. t M are macroeconomic country-specific variables, including inflation (INFLATION), logarithm of GDP in current US dollars, (ECONSIZE), logarithm of GDP per capital in 17 US dollars, (GDPC), property rights (PR), business freedom (BF), financial freedom (FF), the external ratings and creditor monitoring index (ERC), index of the size of informal market in a country (INFORMAL) and the number of MFI competitors in the country (COMPET). Business freedom (BF) is the ability to create, operate, and close an enterprise quickly and easily. Burdensome, redundant regulatory rules are the most harmful barriers to business freedom. Microfinance serving clients include the members who are willing to start a small business by getting loans from MFIs. Therefore, it is controlling for the demand of this group of clients to MFIs. Property rights (PR) is an assessment of the ability of individuals to accumulate private property, secured by clear laws that are fully enforced by the state. It is controlling for the institutional structure that is expected to control the supply of funds to MFIs, from private investors or donors. Financial freedom (FF) is a measure of security of banking system as well as its independence from government control. It may affect funding choices of MFIs that operate as regular commercial banks. The index of the size of the informal market (INFORMAL) measures the level of market economy. Higher values represent larger informal markets. This variable controls demand for loans from MFIs because many MFIs serve non-registered business operating in informal market. The sources of these indices are described in Table 2. They come from the Heritage Foundation, located in the website www.heritage.org and the first three are the components of general economic indicator, economic freedom. The number of competitors in microfinance market (COMPET) comes from www.microcreditsummit.org. This variable captures market impact by the competitive pressure generated by other 18 MFIs operating in the same the market. The external ratings and creditor monitoring index (ERC) is an index created by Barth et al. to measure the role of external credit monitoring of the banking system in a country. It is on a scale from zero to three and is constructed by adding one for an affirmative answer to the following question: a.) Is subordinated debt allowable or required as a part of capital? b.) Do regulations require credit ratings for commercial banks, and c.) Are the top 10 banks in the country rated by an international credit rating agency? (Barth et al., 2006) Table 2 presents definitions of the variables used in the analysis. Table 2: Definition of variables 2 Variable Definition LiabCh Difference of logarithm of non-deposit liability; capture the ability of MFIs to attract loans. EqCh Difference of logarithm of total equity; capture the ability of MFIs to raise equity. N_GRADE Numerical value of letter grades consistent across different rating agencies 3 . RATING 1 if the MFI is rated by a rating agency in the current year, zero otherwise; usually based on previous years financial statements. MFISPECIAL 1 if the MFI is rated by a MFI specified rating agency in the current year, zero otherwise. UPGRADE 1 if the MFI has at least one rating update, zero otherwise. SUBSIDIZED 1 if the rating is subsidized, zero otherwise. AUDIT 1 if the financial statement of the MFI is audited, zero otherwise. REGULATED 1 if the MFI is regulated by a government regulatory agency, zero otherwise. OSS Operational self-sufficiency = Operating revenue / (Financial expense + Loan Loss Provision + Operating Expense). Measures how well the MFI can cover its costs through operating revenues. NAB Logarithm of the number of current borrowers, which is the number of individuals that currently have an outstanding loan balance with the MFI or are responsible for repaying any portion of the gross loan portfolio. ROA Return on total assets. CAPITAL Ratio of total equity to total assets. SAVINGS Ratio of saving to total assets. LOAN Ratio of loans outstanding to total assets; measures risk exposure and how much MFI focus on lending. 2 Region dummies and rater dummies are not listed in this table. A dummy variable that indicates whether the country?s legal origin is English Common Law or not has been tried but was not significant. 3 The scale is from 1 to 10. Higher value represents better rating level. The numerical values are assigned based on the levels of grades issued by raters. For example, if the grade scale is ABCDE, then A=10, B=8, C=6, D=4 and E=2. 19 Variable Definition PAR Portfolio at risk >30 days. NGO 1 if the MFI is organized as a NGO, zero otherwise. AGE Age of the MFI, equals to number of years since inception. MFISIZE Logarithm of the total assets of the MFI. INFLATION Inflation rate, change in consumer prices, source: www.mixmarket.org ECONSIZE Logarithm of GDP in current US dollars. PR Property Rights Index, higher values mean lower protections of private rights; scales from 0 to 100; source: www.heritage.org BF Business Freedom Index, higher values mean more freedom; scales from 0 to 100; source: www.heritage.org FF Financial Freedom Index, higher values mean more freedom; scales from 0 to 100; source: www.heritage.org EconFreedom Economic Freedom Index, the average of ten components with the same weight, higher values mean more freedom; scales from 0 to 100; source: www.heritage.org GDPC GDP per capital current prices in US dollars; source: World Economic Outlook, April, 2007. www.imf.org ERC The External Ratings and Creditor Monitoring Index, source: World Bank Survey of Bank Regulation and supervision, versions 1999-2000 and 2003. INFORMAL Index of the size of the informal market; Scales from1 to 5; 1 equals market economy, 5 represents the informal market size is larger than that of the formal market; source: www.heritage.org COMPET The number of MFI competitors in the country; Source: www.microcreditsummit.org SAMPLE REPRESENTATIVNESS It is important to note that if sample selection is entirely random in the sense that we begin with a random sample and randomly drop observations, OLS estimator will be still consistent. However, there are may be some systematic reasons why MFIs not rated have chosen not to get rated and estimation of the impact of rating is based only on observations with ratings. MFIs choosing to be rated know they have achieved good financial performance and probably expect to get good ratings. Therefore the rating information is not collected randomly and the impact of rating will be overestimated if we only use the rated sample instead of the full sample. Consequently, the parameter estimates obtained are biased and inconsistent. Self-selection bias may need to be corrected for before the rating information currently available is used. 20 Sample selection bias can arise in practice for two reasons. First, self-selection is result of individuals? choices or data units? characteristics (Heckman, 1979). In microfinance, whether MFI chooses to be rated in the current year is determined by its management, the board or, in some cases, by potential donors. It is possible that the financial performance the MFI obtained in previous year affects the decision of getting the external rating in current year. For example, poorly performing MFIs may be less likely to obtain credit ratings from rating agencies because they do not have confidence in receiving good ratings or they do not want to pay for a possible speculative grade rating. Nonrandom sample selection can also arise in panel data. In the simplest case, we have several years of data, but, due to attrition, some MFIs leave the sample. This is particularly a problem in policy analysis, where attrition may be related to the effectiveness of a program. To correct for possible sample selection, a Heckman's sample selection model can be estimated. Assume 1. 1 Y = ?'X + 1 U 2. 2 Y = ?'Z + 2 U where X is a k-vector of regressors, Z is an m-vector of regressorss, possibly including 1's for the intercepts, and the error terms 1 U and 2 U are jointly normally distributed, independently of X and Z, with zero expectations. The first model is the model we are interested in. However, the latent variable 1 Y is only observed if 2 Y > 0. Thus, the actual dependent variable is: Y = 1 Y if 2 Y > 0, Y is a missing value if 2 Y ? 0. 21 Heckman model is a two-stage estimation procedure, beginning with a probit model (selection equation) that produces the inverse mills ratio that is then included in the equation of interest. In addition to the two equations, Heckman estimates rho, the correlation of the residuals in the two equations and sigma, the standard error of the residuals of the outcome equation. Lambda is the product of rho and sigma. Similar to regular two-stage least square procedure, at least one identified instrumental variable is necessary in the probit model of selection equation. Ideal instruments should strongly determine whether MFIs select to be rated, simultaneously, not correlated with what grade MFIs will receive. The instruments selection will be interpreted with details in the next section. The major hypotheses tested with empirical models are as follows: 1) rating helps MFIs raise funds; 2) the impact of rating differs between different raters; 3) better ratings attract more funds; 4) the role of rating differs between for-profit and non-profit institutions; 5) upgrade does not improve fundraising; 6) subsidized rating does not help; 7) MFI specialized rating agencies have more impact than mainstream rating agencies; 8) the impact of rating differs among different regions; 9) audit, regulation and rating play a role to impose marketing discipline as external governance mechanisms; 10) the role of rating is diminished for regulated MFIs. Some accessorial hypotheses will be tested as well: 1) less capitalized MFIs attract more funds; 2) smaller MFIs are more likely to get additional funds; 3) better operational self-sustainability helps MFIs raise debts but not equity. 22 CHAPTER IV: DATA DESCRIPTION Data used in this study come from four sources. Individual MFI data and MFI rating data come from the database collected by MIXMARKET information platform (www.mixmarket.org). MIXMARKET is a global, web-based, microfinance information platform. It provides information to sector actors and the public at large on microfinance institutions (MFIs) worldwide, public and private funds that invest in microfinance and raters/external evaluators. At the time of data collection, it had posted MFI profiles of 954 MFIs across 89 countries, for period 1998-2006. Country-specific data mainly come from MIXMARKET, while some general economy and bank regulation indices come from Heritage Foundation, World Bank Country Variables, World Bank Banking Survey and MicrocreditSummitCampaign, with details in Table 2. After merging the three major databases, the resulting database includes 315 MFIs across 63 countries from 1999-2005, and forms 875 annual individual MFI observations. Table 3: Sample distribution by year 1999 2000 2001 2002 2003 2004 2005 Total Observations 34 42 88 119 161 170 261 875 Full Sample % of sample 3.9 4.8 10.1 13.6 18.4 19.4 29.8 100 Observations 0 0 8 10 28 33 50 129 Rated Sample % of sample 0 0 6.2 7.8 21.7 25.6 38.8 100 Table 3 presents the sample distribution by year. The MFIs in the sample are 23 increasingly distributed across time. 3.9 percent of the annual observations are from 1999, 4.8 percent from 2000, 10.1 percent from 2001, 13.6 percent from 2002, 18.4 percent from 2003, 19.4 percent from 2004 and 29.8 percent from 2005. Over 85 percent of the ratings were completed between 2003 and 2005, while around 65 percent of the annual observations are from this period. In this study, observations of rating without grades are excluded because the analysis focuses on the impact of letter grades. There are no rating observations before 2001 because letter grades rating started in 2001. The concentration of ratings in the last three years of the study period is consistent with the industry developments as rating has become more popular since 2003. Table 4: Distribution of sample MFIs and raters by geographic region Table 4 presents the distribution of sample MFIs and raters by geographic region. The highest numbers of MFIs in the sample are in Africa, Asia and Latin America and The Caribbean (LAC) three regions. They account for over 80 percent of the whole sample. Rated MFIs as a percentage of a region?s MFIs represented in the sample are Region Number of countries OBS % of sample Serving raters Rated OBS Rated % within region Africa 21 230 26.3 MicroRate 26 11.3 Asia 10 175 20 MCRIL CRISIL 20 11.4 Eastern Europe and Central Asia 11 87 9.9 Microfinanza MicroRate Planet Rating MCRIL 14 16.1 Latin America and The Caribbean 17 333 38.1 Microfinanza MicroRate Class&Asociados S.A. FitchRating Equilibrium 59 17.7 Middle East and North Africa 4 50 5.7 Planet Rating MicroRate Microfinanza 10 20 Total 63 875 100 129 24 relatively evenly distributed. The percentage rated MFIs ranges from 11.3 percent for African MFIs to 20 percent for MFIs from the Middle East and North Africa (MENA) region. Table 5 is a brief description of MFI rating agencies included in this study and reveals the rated sub-sample distribution by rating agencies. MicroRate and PlanetRating rated over 55 percent of the sample, followed by Microfinanza and MCRIL with 17.8% and 13.2% respectively. The other agencies, CRISIL, Class&Asociados S.A., Equilibrium, and FitchRating accounts for around 13% totally. MicroRate, PlanetRating and Microfinanza are founded in 1997, 1999, and 2000 respectively but they have been very active in the market for rating service for microfinance. Development of microfinance industry has increased the demand of rating service, consequently boosts the growth and maturity of rating service industry. Table 5: Brief description of MFI rating agencies and sample distribution Agency Type Found Year Serving Region Rated OBS % Rating Grades Scale CRISIL Mainstream 1987 Southeast Asia 2 1.6 mfR1; mfR2; mfR3; mfR4; mfR5; mfR6; mfR7; mfR8 ClassRating Mainstream 1995 LAC 3 2.3 ABCDE Equilibrium Mainstream 1910 LAC 7 5.4 A; B+; B; C+; C; D+; D; E+; E FitchRating Mainstream 1913 LAC 6 4.7 AAA;AA;A BBB;BB;B CCC;CC;C; DDD;DD;D MCRIL MFI specialized 1983 Asia; CEE/NIS 17 13.2 ?+++; ?++; ?+; ?; ??; ?+; ?; ??; ?+; ? MicroRate MFI specialized 1997 LAC; MENA; CEE/NIS 37 28.7 ?++; ?+; ?; ??; ?+; ?; ??; ?+; ?; ?? Microfinanza MFI specialized 2000 MENA; CEE/NIS; LAC 23 17.8 AAA;AA;A BBB;BB;B CCC;CC;C; DDD;DD;D 25 Agency Type Found Year Serving Region Rated OBS % Rating Grades Scale PlanetRating MFI specialized 1999 LAC; MENA CEE/NIS; 34 26.4 A+;A;A-;B+;B B-;C+;C;C-;D;E G5*; G5; G4*; G4; G3*; G3; G2*; G2 G1*; G1 Total 129 100.0% Table 5 also shows that microfinance raters do not have a consistent letter grade system and that grade scales vary by rater. For example, CRISIL has 8 levels, in terms of mfR1-mfR8 (mfR1 represents best rating), while Microfinanza uses 12 levels from AAA to D (AAA represents best rating). In this study a single consistent numerical scale is created using the individual grade scales of each rater. The scale used in this analysis varies from 1 to 10. Higher value represents better rating. For example, if the grade scale is ABCDE by Equilibrium, then A=10, B=8, C=6, D=4 and E=2. During the study period, of the 315 MFIs, 97 were rated at least once and 27 were rated at least twice. In total, the database contains 129 ratings. Principle components analysis is one method for economists to create economic indices, such as economic freedom index. An alternative way is converting qualitative indicators to quantitative ones, which helps detect the impact of qualitative levels. In this thesis, equal margin is applied to each pair of border categories within individual raters, and is used here. The shortcoming of this approach may be that marginal effect might be different due to different point locations in the array but it is assumed that this is not the case here because each rating grade should have equal impact. 26 Table 6: Summary Statistics Variable Obs Mean Std. Dev. Min Max LiabCh 832 0.412 1.637 -10.646 11.263 EqCh 875 0.222 1.498 -7.994 8.094 N_GRADE 129 7.116 1.809 1.9 10 REGULATED 875 0.578 0.494 0 1 L_AUDIT 875 0.392 0.488 0 1 L_OSS* 875 1.165 0.411 0.107 3.769 L_NAB 875 9.200 1.759 2.639 15.124 L_ROA 875 0.078 0.113 0 1.359 L_CAPITAL 875 0.424 0.280 0.0003 1 L_LOAN 875 0.721 0.176 0.055 1.089 L_SAVINGS 875 0.167 0.265 0 1.120 L_MFISIZE 875 15.309 1.754 9.454 22.151 AGE 875 12.395 10.407 2 111 NGO 875 0.454 0.498 0 1 INFLATION 875 0.063 0.069 0.0003 0.961 PR 875 37.817 10.999 10 90 BF 875 33.726 11.129 10 70 FF 875 51.349 16.994 10 90 EconFreedom 875 56.890 6.315 37 79 GDPC 875 6.861 0.925 4.755 8.980 ERC 379 1.525 0.679 1 3 INFORMAL 828 4.089 0.712 2 5 COMPET 842 73.524 143.904 1 666 * L stands for one period lag. Table 6 presents summary statistics of the variables used in the empirical analysis for the full sample. Table 7-1 presents summary statistics of the key variables used in the empirical analysis for years when the MFI were not rated versus years when the MFIs were rated. Rated MFIs differ from non-rated MFIs in terms of several indicators. In the year preceding rating, rated MFIs have higher and statistically significant OSS (1.252 versus 1.149). OSS measures how well the MFI can cover its costs through operating revenues. Similarly, there were differences in terms of the outreach measure of rated and non-rated MFIs. Outreach is measured as the logarithm of the number of active borrowers 27 (NAB), which is the number of individuals that currently have an outstanding loan balance with the MFI. In the year preceding rating, rated MFIs had better outreach indicators than MFIs that did not get rated (9.719 versus 9.110) 4 . Other indicators statistically different for the two groups include loan to total assets ratio (LOAN) (0.768 versus 0.712), and risk profile (PA R ) (0.039 versus 0.055). Change in non-deposit liability (LiabCh) and in equity (EqCh) do not show difference between non-rated MFIs and the full sample MFIs, however, two-sample t test only considers sample means with degree of freedom respectively, but no other compound variables controlling MFI-specific and macroeconomic-specific characteristics are not considered. Table 7-2 summarizes the statistics for N_GRADE by raters. In the sample we studies, FitchRating has the highest average of N_GRADE 9.87 while MicroRate has the lowest one 6.19. However, since the sample includes limited observations by rater, these results should not be interpreted as representatives. Table 7-1: Summary statistics for non-rated group and rated group Variable Non-rated Rated LiabCh 0.395 0.509 (0.060) (0.164) EqCh 0.205 0.320 (0.054) (0.138) L_OSS 1.149 1.252*** (0.015) (0.032) L_NAB 9.110 9.719*** (0.066) (0.116) L_ROA 0.080 0.062** (0.004) (0.006) L_CAPITAL 0.430 0.390 (0.010) (0.024) L_SAVINGS 0.170 0.147 (0.010) (0.009) L_LOAN 0.712 0.768*** (0.006) (0.015) 4 Another industry standard outreach indicator is ?depth of outreach?, which is calculated as the ratio of average outstanding loan size divided by GDP per capita. However, depth of outreach has much less observations in our database, so if it was used as an independent variable, the sample becomes very small. 28 Variable Non-rated Rated L_PAR 0.055 0.039*** (0.003) (0.004) L_MFISIZE 15.206 15.907 (0.066) (0.117) AGE 12.539 11.566 (0.403) (0.503) Standard errors are in the parenthesis * difference in means between rated and non-rated statistically significant at the 10% level ** difference in means between rated and non-rated statistically significant at the 5% level *** difference in means between rated and non-rated statistically significant at the 1% level Table 7-2: Summary statistics for N_GRADE by raters MCRIL MicroRate Microfinza PlanetRating N_GRADE (Mean) 7.71 6.19 8.04 6.36 OBS 17 37 23 34 ClassRating Equilibrium CRICIL FitchRating N_GRADE (Mean) 6.67 8 8.75 9.87 OBS 3 7 2 6 Table 8: Mean comparison of selected indicators between rated group and MBB 5 2004 2005 Variables Sample MBB Sample MBB CAPITAL 33.9 34.0 33.5 29.5 LOAN 75.8 78.4 80.8 80.1 NAB 44,250 14,426 55,480 16,755 ROA 6.4 2.1 6.0 2.1 PAR 3.5 1.9 3.7 1.8 Total Assets 25,900,000 7,432,540 36,300,000 12,230,758 Table 8 presents comparison of the means of selected indicators between the rated group and benchmarks posted by Micrbanking Bulletin, issue No. 14, Spring, 2007. The benchmarks provided by the bulletin are the most widely used benchmarks in the industry. CAPITAL and LOAN have close means to the benchmarks, without significant difference 5 The MicroBanking Bulletin (MBB) is the premier benchmarking source for the microfinance industry, reaching back as far as 1997. It is a primary output of the Microfinance Information Exchange, Inc. (MIX). The MicroBanking Bulletin's industry commentary, analysis and benchmarks are widely used by investors, donors and other service providers to facilitate greater standardization and a better understanding of developments in the microfinance sector. 29 between the sample and the proposed population. However, NAB, ROA, PAR and MFISIZE (in terms of total assets) have higher averages than the benchmarks. Therefore, presumably, larger MFIs with better social performance and more sustainable operations will be more likely to choose to be rated. Comparison between the risk level of the rated MFIs and that of the Microbanking Bulletin benchmarks shows that the risk measure PAR is higher for rated than for the benchmarks. This may be related to the fact that larger number of borrowers (higher NAB) unavoidably leads to higher risk exposure. 30 CHAPTER V: RESULTS Table 9 presents the results of the impact of rating, on change in non-deposit liability and equity, for the full sample, without considering possible sample selection issues. The results show that rating sends signals that help MFIs obtain loans but not equity. Specifically, according to results of Model 5, Table 9, MFIs will increase their non-deposit liability by 58.8 % if they were rated. Individual RATER dummies are jointly significant in LiabCh model, which is consistent with previous research that raters have different impacts on raising funds. ClassRating, FitchRating and CRISIL have statistically significant positive impact 6 while Equilibrium, MCRIL Microfinanza Microrate and PlanetRating do not. However, individual raters do not affect change of equity. This might be because donors, who are the main providers of equity, do not pay as much attention to rating as creditors because donors may care more about the mission of the MFI, while creditors want to ensure themselves that their loans will be repaid. Auditing turns out to be effective as an external governance mechanism because MFIs that have their financial statements audited are more likely to raise funds. Regulated MFIs are not more likely to get additional equity and liability. External governance mechanisms play an important role in MFIs fundraising activities because RATING, AUDIT and REGULATED are jointly significant in change in equity and in 6 In model 5, the coefficients of ClassRating and FitchRating are 2.765 and 0.710, significant at 0.01 and 0.10, respectively. The constant term is 3.456, significant at 0.01, which indicates that CRISIL has significant impact as the reference group. 31 change in liability. The results are consistent with Barth (2006), which strongly supports the view that more stringent external audits and greater external rating and credit monitoring, individually and collectively, enhance bank profitability in countries around the world. Equity to total assets ratio (L_CAPITAL) affects ability to raise funds, which means that capital leverage help MFIs get well capitalization in microfinance industry. Table 9: The impact of rating on change in non-deposit liability and equity (1) (2) (3) (4) (5) (6) LiabCh EqCh LiabCh EqCh LiabCh EqCh RATING 0.544* 0.507 0.404 0.232 0.588** 0.283 (1.77) (1.64) (0.89) (0.91) (2.01) (1.63) RATER DUMMIES YES NO YES YES SUBSIDIZED -0.183 -0.153 0.038 -0.161 -0.101 -0.126 (0.82) (0.61) (0.12) (0.50) (0.46) (0.54) REGULATED 0.168 0.190* 0.236 0.288* 0.141 0.155 (1.49) (1.75) (1.23) (1.67) (1.33) (1.56) L_AUDIT 0.250** 0.196** 0.454** 0.195 0.250** 0.206** (2.11) (1.97) (2.29) (1.20) (2.20) (2.13) UPGRADE -0.687* -0.077 -0.916* 0.046 -0.681* -0.065 (1.84) (0.39) (1.90) (0.18) (1.82) (0.33) L_OSS 0.505*** 0.320** 0.547 0.004 0.488*** 0.320** (2.75) (2.27) (1.11) (0.01) (2.70) (2.36) L_ROA 0.631 0.152 1.970 0.536 0.715 0.186 (1.05) (0.38) (1.20) (0.83) (1.15) (0.47) L_NAB -0.049 0.081 -0.013 0.143 -0.062 0.079 (0.88) (1.55) (0.12) (1.55) (1.12) (1.53) L_CAPITAL 1.203*** -1.405*** 0.662 -1.440*** 1.084*** -1.570*** (4.30) (6.24) (1.40) (3.59) (3.98) (7.05) L_LOAN 0.534 0.280 0.279 0.244 0.746** 0.206 (1.43) (0.92) (0.57) (0.48) (2.13) (0.72) L_SAVINGS 0.872*** -0.429 0.467 -0.810* 0.892*** -0.546** (2.62) (1.55) (1.05) (1.96) (2.78) (2.08) L_MFISIZE -0.386*** -0.518*** -0.341*** -0.499*** -0.367*** -0.518*** (5.77) (8.22) (3.05) (4.78) (5.73) (8.31) L_PAR 0.240 -0.029 -0.963 0.129 0.049 -0.174 (0.30) (0.04) (1.24) (0.12) (0.06) (0.27) NGO -0.187 -0.154 -0.181 -0.430** -0.159 -0.204* (1.31) (1.29) (0.76) (2.28) (1.16) (1.82) AGE 0.014* 0.021*** -0.004 0.011 0.013 0.021*** (1.72) (3.25) (0.44) (1.11) (1.57) (3.32) ECONSIZE 0.036 -0.044 0.043 -0.186** 0.075 -0.072* (0.57) (0.97) (0.46) (2.50) (1.59) (1.89) PR -0.008 0.000 (1.41) (0.04) 32 (1) (2) (3) (4) (5) (6) LiabCh EqCh LiabCh EqCh LiabCh EqCh BF 0.004 -0.002 (0.71) (0.33) FF 0.006 0.007** (1.54) (2.07) INFLATION -1.245 -1.658** -1.394 -1.722 -1.068 -1.752** (1.61) (2.16) (1.19) (1.45) (1.33) (2.47) GDPC -0.048 0.103 -0.112 0.305 -0.108 0.162* (0.42) (1.09) (0.62) (1.60) (1.13) (1.93) INFORMAL -0.013 -0.011 (0.17) (0.16) COMPET 0.001 0.000 (1.30) (0.11) EconFreedom 0.014 -0.005 0.011 0.007 (0.76) (0.30) (1.16) (0.84) ERC -0.089 -0.125 (0.72) (0.90) Constant 4.418*** 6.971*** 3.732 8.476*** 3.456*** 7.112*** (3.90) (7.71) (1.56) (4.32) (3.19) (8.12) Observations 785 824 355 379 832 875 R-squared 0.23 0.24 0.25 0.23 0.22 0.24 F-Statistics 7.10 4.50 11.00 2.95 10.11 6.92 Robust t statistics in the parentheses. significant at 10%; ** significant at 5%; *** significant at 1% No positive impact of rating update is found in all the specification. This result should be interpreted with caution because in the sample only 27 MFIs out of 94 in total have rating updates while The Rating Funds Survey 2005 shows that overall 49% of MFIs have. Equity to total assets ratio has negative effects on the change in equity, and positive on change in liability which suggests that MFIs may target an optimal capital structure. The smaller an MFI?s size is, the more funds it gets, which is consistent with the notion that MFIs strive to obtain economies of scale. In Models (3)-(6) EconFreedom is used as the general economic freedom index, instead of BF, FF and PR, because these three components are not jointly statistically significant. The ERC index, created based on banking regulations and supervisions, is not statistically significant in model (3) and model (4) and is not used further in the analysis because it is not available for all countries in the sample and including it in the sample of 33 rated MFIs significantly reduces the sample. It is also possible that microfinance industry might differ significantly from the regular bank industry. Additionally, another policy study also shows that any approach to regulation and supervision of MFIs needs to recognize their heterogeneity and accommodate the flexibility and scope for development that MFIs need (Hardy, 2003). Some studies argue that rating has more effects during a longer period of time (Jorion et al., 2005). In this thesis, rating with one period lag, together with all other variables derived from RATING with one lag period was estimated with several model specifications. However, there was no evidence to support the notion that rating has an impact for a period longer than one year. Table 10: The impact of rating on change in non-deposit liability and equity by raters and regions (1) (2) (3) (4) LiabCh EqCh LiabCh EqCh N_GRADE -2.170 0.032 0.185** 0.246** (1.49) (0.05) (2.54) (2.23) SUBSIDIZED -0.124 -0.150 -0.179 -0.043 (0.10) (0.13) (0.63) (0.19) N_GRADE*SUBSIDIZED -0.082 -0.001 (0.49) (0.01) REGULATED -0.621 1.698 0.317 0.788*** (0.35) (1.12) (0.99) (3.01) N_GRADE*REGULATED 0.113 -0.150 (0.47) (0.70) UPGRADE 1.460 -1.393 -0.675 0.061 (0.60) (0.52) (1.21) (0.15) N_GRADE*UPGRADE -0.337 0.193 (1.00) (0.54) MFISPECIAL -25.294 -6.860 -0.238 0.069 (1.63) (0.87) (0.38) (0.18) L_AUDIT 0.708* -0.056 0.735* 0.128 (1.69) (0.19) (1.84) (0.54) L_OSS 1.456** -0.011 1.060** 0.077 (2.14) (0.02) (2.07) (0.18) L_ROA 10.120** 4.028 4.984 2.028 34 (1) (2) (3) (4) LiabCh EqCh LiabCh EqCh (2.00) (1.39) (1.37) (0.87) L_NAB 0.005 -0.032 -0.089 0.037 (0.03) (0.18) (0.55) (0.26) L_CAPITAL -0.846 -3.254*** -0.381 -2.930*** (1.01) (3.32) (0.64) (4.79) L_LOAN -1.141 0.953 -1.095 0.249 (1.01) (0.86) (1.61) (0.32) L_SAVINGS 0.115 -1.419 0.820 -1.115 (0.08) (1.15) (0.87) (1.47) L_MFISIZE -0.823*** -0.663*** -0.746*** -0.647*** (3.28) (3.41) (4.36) (3.92) L_PAR 2.240 2.945 -0.363 3.176 (0.61) (0.77) (0.15) (1.12) NGO 1.114 2.951** -0.055 0.699 (0.69) (2.32) (0.15) (0.67) N_GRADE*NGO -0.274 -0.454** -0.091 (1.11) (2.60) (0.64) AGE 0.015 0.052* 0.036 0.051** (0.48) (1.77) (1.61) (2.28) ECONSIZE 0.373 0.065 0.099 -0.169 (1.13) (0.24) (1.05) (1.63) PR 0.004 -0.002 (0.15) (0.08) BF 0.026 -0.019 (1.02) (0.74) FF -0.007 0.007 (0.30) (0.41) INFLATION -7.398 2.037 -3.717 0.986 (1.06) (0.38) (0.94) (0.25) GDPC -0.309 0.335 -0.035 0.363 (0.44) (0.52) (0.13) (1.60) REGION DUMMIES NO NO RATER DUMMIES NO NO RATERS*N_GRADE NO NO INFORMAL 0.694* 0.200 (1.94) (0.75) COMPET -0.005 -0.002 (1.21) (0.53) EconFreedom 0.029 -0.025 (0.99) (1.02) Constant 23.976* 6.707 7.660** 9.793*** (1.78) (0.92) (2.18) (3.01) Observations 119 123 125 129 R-squared 0.58 0.55 0.43 0.44 F-Statistics - - 2.41 2.82 Robust t statistics in the parentheses. significant at 10%; ** significant at 5%; *** significant at 1% 35 Table 10 presents the results of the impact of rating, measured with a numerical grade, on change in non-deposit liability and equity, for the sample or rated MFIs, with rater dummies, region dummies and interactions. Model (3) and (4) where for some variables and interactions left out are the preferred models after applying some tests for joint significance. N_GRADE has positive impacts in both model (3) and (4). This suggests that rating plays a disciplining role because better ratings help MFIs raise funds. Subsidized rating does not help MFIs raise additional funds considering both main effect and interaction effect, which is consistent with the result of Hartarska and Nadolnyak (2007b). Within the sample of rated MFIs REGULATED also has a positive impact on raising equity but not debt suggesting that creditor prefer regulated MFIs, while donors are indifferent to regulation. Neither UPGRADE nor MFISPECIAL have statistically significant impact. There is no well developed rating system for microfinance entities and even specialized raters have no consistent rating grades and rating standards. This suggests that the type of rating agencies an MFI chooses does not matter. In addition, most MFIs receive upgrades either because the rating is subsidized or they were satisfied with the rating grade they obtained in previous year. Therefore, updates of rating do not differ from first-time rating. Operational self-sustainability affects the ability to raise funds because this is one of the missions MFIs strive to achieve and funding providers might care about this criterion. N_GRADE*NGO has statistically significant negative sign, which means that rating grades have lower return on NGOs than for for-profit financial institutions. This is interpreted to mean that for for-profit financial institutions aim at ?profit? and rating 36 might help them attract funds while NGOs concentrates on meeting a mission. REGION DUMMIES, RATER DUMMIES and RATER*N_GRADE are not jointly significant in models (1) and (2). It may be because N_GRADE accounts for difference between raters and was assigned with consistent numerical grades. The results from OLS regressions may suffer from endogeneity since the MFIs who have chosen to be rated probably knew they had better financial performance and expected better letter grades and thus more funds than the ones who did not chose to be rated by external raters. Also, the cost of rating assessment activities is another factor which prevents MFIs with relative poor performance from getting rating. Therefore, the estimates based on the sample of MFIs with full rating information may be biased. It is necessary to consider sample selection problems. HAUSMAN TEST FOR ENDOGENEITY AND TWO-STAGE PROCEDURE Since N_GRADE may suffer from endogeneity, a Hausman test to check for that is performed. In the second stage, P-values of residuals saved from the first stage of Hausman, which regresses N_GRADE on all the other exogenous variables in the clean models, are 0.01 and 0.03 respectively, which indicates endogeneity. Two-stage least square procedure is applied to correct for this problem. REGULATED might be appropriate for an instrumental variable for LiabCh equation in two-stage procedure, using the preferred models. There are two reasons. First, if MFIs get regulated, they need to meet some financial standards otherwise the regulators will impose sanctions. Therefore, it is supposed to be positive correlated with N_GRADE. Second, REGULATED might not be correlated with additional funds (liability) from creditors 37 because donors do not care much about whether MFIs are regulated due to the distortion of regulation in microfinance. L_OSS might be appropriate for an instrumental variable for EqCh equation in two-stage procedure, using cleaned models. OSS captures how well MFIs make themselves operational self-sustainable. It goes into the evaluation of rating, thus, it is correlated with N_GRADE. On the other hand, investors care more about return on total equity, instead of OSS; therefore, it is not correlated with the extra equity MFIs receive. Table 11: The impact of rating on change in non-deposit liability and equity by two-stage least square procedure (1) (2) LiabCh EqCh N_GRADE 1.189 0.350 (0.82) (0.47) SUBSIDIZED -0.232 -0.070 (0.53) (0.30) UPGRADE -1.206 -0.015 (1.36) (0.03) MFISPECIAL 1.671 0.254 (0.59) (0.18) L_AUDIT 0.854* 0.105 (1.73) (0.43) L_OSS 0.432 (0.33) L_ROA 7.385 2.304 (1.20) (1.09) L_NAB -0.433 -0.015 (0.80) (0.05) L_CAPITAL -1.993 -3.173** (0.77) (2.11) L_LOAN -1.612 0.172 (1.09) (0.16) L_SAVINGS 0.597 -1.049 (0.47) (1.35) L_MFISIZE -0.728*** -0.658*** (2.90) (3.88) L_PAR -1.983 2.609 (0.37) (0.79) NGO 0.006 0.090 (0.01) (0.34) 38 (1) (2) LiabCh EqCh AGE 0.027 0.050** (0.62) (2.11) ECONSIZE 0.195 -0.153 (0.89) (1.18) INFLATION -6.264 0.116 (0.93) (0.03) GDPC -0.407 0.305 (0.67) (0.91) EconFreedom 0.101 -0.016 (0.80) (0.27) REGULATED 0.729** (2.03) Constant 0.923 9.534* (0.07) (1.74) Observations 125 129 R-squared - 0.42 F-Statistics 1.25 2.80 Robust t statistics in the parentheses. significant at 10%; ** significant at 5%; *** significant at 1% Table 11 presents the results of two-stage procedure, using the preferred models from the OLS specifications. The results show that the coefficients on N_GRADE are not statistically significant in either of the models. Although it is not exactly clear what goes into rating, in general, rating agencies exist, and thus, they must provide additional independent information. In microfinance, rating matters more for commercial funds providers; however, commercial funds have not been distinguished from other funds sources, such as donation and grants in this study. There is evidence to show that AUDIT and REGULATED help MFIs raise liability and equity, respectively. The impacts of MFISIZE are consistent with all the other analysis and this result is very robust in different models. However, the evidence for endogeneity of N_GRADE is not strong. The instruments selected may suffer from the problem of weak instruments. Therefore, it is possible that 39 rating helps impose market discipline in microfinance as an effective external governance mechanism. HECKMAN EXTENSION Self-selection bias can also be corrected by Heckman Selection Model as well as two-stage procedure. L_AUDIT might be an appropriate instrument identified in the Heckman Selection Model. First, MFIs are likely to choose to both get financial statements audited and rated. In our sample, the correlation between RATING and AUDIT is 0.16 at 0.0000 significant levels. Second, AUDIT is not correlated with N_GRADE because rating system does not account whether MFIs get their financial statements audited. In our sample, the correlation between N_GRADE and AUDIT is 0.04 with p-value 0.63. AGE is another instrument in Heckman Selection Model for the similar reasons. Table 12: Heckman Selection Model to predict numerical grades 7 Heckman selection model -- two-step estimates Number of obs=875 (regression model with sample selection) Censored obs=746 Uncensored obs=129 Wald chi2(16) = 48.99 Prob > chi2 = 0.0000 Outcome equation Selection equation N_GRADE RATING L_CAPITAL 1.742*** -0.196 (2.59) (0.80) L_NAB 0.180 0.014 (1.17) (0.27) L_PAR 3.951 -1.002 (0.97) (0.94) L_OSS 0.588 0.178 (1.19) (1.13) 7 Heckman estimates rho, the correlation of the residuals in the two equations and sigma. In this regression, rho is 0.134, significant at 10% level, which confirms that self-selection problem exists in this case. 40 Outcome equation Selection equation N_GRADE RATING L_SAVINGS 0.354 -0.387 (0.49) (1.47) L_LOAN 1.167 0.625* (1.06) (1.75) L_ROA -1.794 -0.753 (0.69) (0.96) L_MFISIZE 0.297 0.129** (1.51) (2.19) L_AUDIT 0.364*** (3.30) AGE -0.017** (2.08) Constant -1.623 -3.592*** (0.41) (4.99) Observations 129 875 Robust z statistics in the parentheses. significant at 10%; ** significant at 5%; *** significant at 1%. Table 12 presents the results obtained by Heckman Selection Model. By using Heckman Selection Model, we can get a vector of the predicted values of N_GRADE, which is denoted as PHECKMAN. Table 13 reports the comparison of actually observed N_GRADE and the predicted PHECKMAN. Paired two-sample t-test shows that the predicted result is as robust as the original values, which is the preliminary step and the precondition for using PHECKMAN as independent variable, instead of N_GRADE in the following regression specifications. Table 13: Summary statistics for PHECKMAN and actual N_GRADE 8 Variable Obs Mean Std. Dev. Min Max N_GRADE 129 7.116 1.809 1.9 10 PHECKMAN 129 6.975 0.961 1.93 9.98 However, the relative small number of uncensored group will affect the power and the efficiency of the prediction, in our case, the difference of observation numbers 8 Paired t-test for equal mean shows the predicted result is robust with P-value 0.03. The null hypotheses of equal mean is failed to be rejected at 1% significance level. 41 between full sample and uncensored group might be too large. It is not safe enough to use PHECKMAN instead of actual N_GRADE in the models. Therefore, Heckman extension ends up by this step. TESTS FOR IDENTIFIED MODELS Table 10: Model (3) and (4) are identified as cleaned models. Firstly, heteroskedasticity has been examined in the specifications by using Breusch-Pagan tests. Robust errors are reported in parentheses because heteroskedasticity exits in all specifications. Secondly, multicollinearity was assessed in the specifications using the variance inflation factor (VIF). The means of VIFs confirm that multicollinearity is not a problem in the clean models but it is a problem in the extended models of all the regressions. (See Appendix5.) Lastly, specification error and omitted variables bias was tested for using the Ramsey RESET test in all models. The null hypothesis states that the model does not have specification error and the alternative is that the model does have specification error. The F-statistics for the Ramsey RESET of specification (3) and (4) were calculated to be 2.39 and 3.30; consequently, p-values are 0.001 and 0.02 respectively. The results might be acceptable since the p-values reported from Ramsey RESET are usually low and the significance level can be set lower than the regular t-tests in regression procedures. For the RESET test, the lower the F-statistic is, the more certain it can be concluded that specification error or omitted variables test is not a problem. For these two models used in this thesis, it is not highly certain that these problems cannot be proven to exist in the 42 model. It is possible that there are still some other factors which have effects on the change of non-deposit liability and total equity of MFIs, such as outreach depth indicator, and the factor whether the country has deposit insurance schemes. Since the database available does not provide the information mentioned above, they cannot be included in the analysis. The limitations of this thesis come from the limited data sample, as well as the excessive missing values within unbalanced panel data, thus further improvement is hard to be performed. Better database is necessarily to help future work refine the model. 43 CHAPTER VI: CONCLUSIONS This thesis studies whether rating assessments help MFIs raise more funds, using a sample of 315 MFIs operating in 63 countries in the world during the period of 1999-2005. The main conclusion is that there is no strong evidence that rating affects MFIs fundraising efforts, after accounting for possible endogeneity of rating by two-stage least square procedure, although rating has statistically significantly positive effects on the change of non-deposit liability and total equity of MFIs by OLS procedure. However, the evidence for endogeneity of rating is not strong, which might support the idea that rating helps impose market discipline. Difference of rating impact on raising liability between raters is detected. Rating updates, subsidized rating and rater types do not affect fundraising and regulatory authorities? activities do not affect the values of rating. The results on the role of rating should be interpreted with caution due to the specific sample. The results need to be viewed with the limitation of data and the unbalanced nature of the panel data. Also, whether the rating is requested from potential donors or current boards may affect the effectiveness of rating itself. However, date associated with above issues is not available. Moreover, more information can be drawn from full the version rating reports, like sub-rating for separate aspects can be examined 44 in other regression specifications once data is available. The role of rating is very important for policy purposes and, since the results in this thesis are only valid for a specific time period and regions, the role of rating needs to be addressed further with better data. 45 REFERENCES Barth, J., Nolle, D., Phumiwasana T., & Yago G. ?A Cross-country Analysis of the Bank Supervisory Framework and Bank Performance.? Financial Markets, Institutions & Instruments, 2003, 12(2), pp. 67-120. Barth, James R., Caprio, Gerard and Levine Ross. Rethinking Bank Regulation: Till Angels Govern. Cambridge University Press, New York, NY, 2006. Barth, James R., Hartarska, Valentina, Nolle, Doniel E. and Phumiwasana, Triphon. ?A Cross-country Analysis of External Governance and Bank Profitability.? BOOK Boot, Arnoud W. A., Milbourn, Todd T. and Schmeits, Anjolein. ?Credit Ratings as Coordination Mechanisms.? The Reviwes of Financial Studies /2006, v19n1. Cantor, R. ?An Introduction to Recent Research on Credit Rating.? Journal of Banking and Finance, 2004, 28, pp. 2565-2573. Daley-Harris, S. ?State of the Microcredit Summit Campaign Report 2006.? Washington D.C. Microcredit Summmit Campaign, 2006. Farrington, Todd. ?The Evolution of Specilized Rating for MFIs.? MicroEnterprise Development Review, 2005, Vol.8 No. 1. Ferri, Giovanni, Liu, Li-Gang and Majnoni, Giovanni. ?The Role of Rating Agency Assessments in Less Developed Countries: Impact of the Proposed Basel Guidelines.? Journal of Banking & Finance, 2001, 25, pp. 115-148. Hardy, D., Holden, P. and Prokopenko, V. ?Microfinance Institutions and Public Policy.? Journal of Policy Reform, 2003, Vol. 6(3), pp. 147?158. Hartarska, Valentina. ?Do Audit, Rating and Regulation help Impose Market Discipline in Microfinance?? Managerial Finance, forthcoming. 46 Hartarska, Valentina. ?Governance and Performance of Microfinance Institutions in Central and Eastern Europe and the Newly Independent States.? World Development, 2005, Vol. 33, No. 10, pp. 1627-1643. Hartarska, Valentina and Holtmann, M. ?An Overview of Recent Developments in the Microfinance Literature,? Agricultural Finance Review, 2006, 66(2), pp. 147-165. Hartarska, Valentina and Nadolnyak, Denis (2007a). ?Do Regulated Microfinance Institutions Achieve Better Sustainability and Outreach? Cross-country Evidence.? Applied Economics, 2007, v 39, pp. 1207-1222. DOI: 10.1080/00036840500461840. Hartarska, Valentina and Nadolnyak, Denis (2007b). ?Do Regulated Microfinance Institutions Achieve Better Sustainability and Outreach? Cross-country Evidence.? Applied Economics, 2007, v 39, pp. 1-16. DOI: 10.1080/00036840500461840. Heckman, James J. ?Sample Selection Bias as a Specification Error.? Econometrica 1979, 47, pp. 153-161. Jorion, Philippe, Liu, Zhu and Shi, Charles. ?Informational effects of regulation FD: evidence from rating agencies.? Journal of Financial Economics, 2005, 76(2), pp. 309-330. Loffler, G. ?Ratings versus Market-based Measures of Default Risk in Portfolio Governance.? In: Cantor, R. (Ed.), Recent Research on Credit Ratings (special issue). Journal of Banking and Finance 2004, 28 (11). Manne, G.A. ?Agency Costs and the Oversight of Charitable Organizations.? Wisconsin Law Review, 1999, 2, pp. 227-272. Meloni, Caterina and Suaznabar, Claudia. ?Key Data on the MF Rating Industry and the Rating Fund.? II Forum on MFI Ratings-Toward a Sustainable Market, Santa Cruz, Bolivia, 2005. Morgan, Donald P. ?Rating Banks: Risk and Uncertainty in an Opaque Industry. American Economic Review, 2002, 92(4), pp. 874-889. Morgan, Donald P. and Stiroh, Levin J. ?Bond Market Discipline of Banks: Is the Market Tough Enough? ? Federal Reserve Bank of New York Staff Report, 2000, No. 95. 47 Nieto, Begona G. and Cinca, Carlos S. ?Factors Explaining the Rating of Microfinance Institutions.? Nonprofit & Voluntary Sector Quarterly, 2007. Poon, Winnie P.H. ?Are Unsolicited Credit Ratings Biased Downward?? Journal of Banking & Finance, 2003, 27, pp. 593-614. Roy, Patrick V. ?Is There a Difference between Solicited and Unsolicited Bank Rating and If so, Why?? Working Paper research n79, National Bank of Belgium Eurosystem, 2006. Sally, Allen N., Davies, M. and Flannery, Mark J. ?Comparing Market and Supervisory Assessments of Bank Performance: Who knows What When?? Journal of Money, Credit, and Banking, 2000, pt. II, 32, pp. 641-67. Samolyk, K. ?U.S. Banking Sector Trends: Assessing Disparities in Industry Performance.? Economic Review, Federal Reserve Bank of Cleveland, 1994, 30(2), pp. 2-17. Special Report: Who rates the raters? - Credit-rating agencies; Credit-rating agencies, The Economist. London: Mar 26, 2005. 374 (8419). pp. 91 Suaznabar, Claudia. ?Key Data on the MF Rating Industry and the Rating Fund.? II Forum on MFI Ratings-Toward a Sustainable Market, Santa Cruz, Bolivia, October 4-5, 2005. Survey of Microfinance 2005, The Economist, London: Nov 5, 2005, 377(8451) p. 1-16. The Rating Fund. The Microfinance Rating Market Outlook: The Rating Fund Market Survey 2005, 2006. Wooldridge, Jeffery M. Introductory Econometrics: A Modern Approach. The 3 rd edition. Thomson South-Western, Mason, OH, 2006. 48 APPENDIX Appendix1. Methods of creating the numerical rating grades by raters. Agency Grades Scale Numerical Values assignments (1-10) CRISIL mfR1; mfR2; mfR3; mfR4; mfR5; mfR6; mfR7; mfR8 mfR1=10; mfR2=8.8; mfR3=7.6; mfR4=6.4; mfR5=5.2; mfR6=4; mfR7=2.8; mfR8=1.6 ClassRating ABCDE A=10; B=8; C=6; D=4; E=2; Equilibrium A; B+; B; C+; C D+; D; E+; E A=10; B+=9; B=8; C+=7; C=6; D+=5; D=4; E+=3; E=2 FitchRating AAA;AA;A BBB;BB;B CCC;CC;C; DDD;DD;D AAA=10; AA=9.2; A=8.4 BBB=7.6; BB=6.8; B=6.0 CCC=5.2; CC=4.4; C=3.6 DDD=2.8; DD=2.0; D=1.2 MCRIL ?+++; ?++; ?+; ?; ??; ?+; ?; ??; ?+; ? MicroRate ?++; ?+; ?; ??; ?+; ?; ??; ?+; ?; ?? ?+++/?++ =10 ?+=9; ?=8; ??=7 ?+=6; ?=5; ??=4 ?+=3; ?=2; ??=1 Microfinanza AAA;AA;A BBB;BB;B CCC;CC;C; DDD;DD;D AAA=10; AA=9.2; A=8.4 BBB=7.6; BB=6.8; B=6.0 CCC=5.2; CC=4.4; C=3.6 DDD=2.8; DD=2.0; D=1.2 PlanetRating A+;A;A-;B+;B B-;C+;C;C-;D;E G5*; G5; G4*; G4; G3*; G3; G2*; G2 G1*; G1 A+=10; A=9.1; A-=8.2; B+=7.3; B=6.4; B-=5.5; C+=4.6; C=3.7; C-=2.8; D=1.9; E=1 G5*=10; G5=9; G4*=8; G4=7; G3*=6; G3=5; G2*=4; G2=3 G1*=2; G1=1 49 Appendix2. Rated sample distribution by country and year Country 2001 2002 2003 2004 2005 Total Argentina 0 0 0 0 1 1 Armenia 0 1 1 0 0 2 Azerbaijan 0 0 0 0 2 2 Bangladesh 0 0 2 0 0 2 Benin 0 0 3 2 1 6 Bolivia 1 1 1 4 5 12 Brazil 0 0 1 0 0 1 Burkina Faso 0 0 0 0 1 1 Cambodia 1 0 2 1 3 7 Colombia 0 0 1 2 2 5 Dominican Republic 0 0 0 0 1 1 Ecuador 0 0 1 3 2 6 Egypt 0 0 0 2 0 2 El Salvador 0 0 0 0 1 1 Ethiopia 0 0 0 0 1 1 Georgia 0 0 0 0 2 2 Honduras 0 0 0 0 2 2 India 2 2 1 1 1 7 Jordan 1 0 0 0 0 1 Kazakhstan 0 1 1 0 0 2 Kenya 1 1 1 0 1 4 Kyrgyzstan 0 1 1 0 2 4 Madagascar 0 0 0 1 0 1 Mali 0 0 1 0 0 1 Morocco 1 2 2 0 1 6 Nicaragua 0 0 2 0 5 7 Pakistan 0 0 0 1 0 1 Paraguay 0 0 0 0 1 1 Peru 1 0 4 8 9 22 Philippines 0 1 0 1 1 3 Russia 0 0 1 0 1 2 Senegal 0 0 1 1 2 4 Tanzania 0 0 0 1 0 1 Togo 0 0 0 1 1 2 Tunisia 0 0 0 1 0 1 50 Country 2001 2002 2003 2004 2005 Total Uganda 0 0 1 3 1 5 Total 8 10 28 33 50 129 Appendix3. List of MFIs included in the rated group ABA Caritas KAFC ACEP DBACD KLF ACODEP EBS KRep ACSI ECLOF PHL KWFT ADOPEM EDPYME Alternativa Kafo AMC de R.L. EDPYME Confianza MDFKamurj AMRET EDPYME Crear Arequipa MFW AMSSF/MC EDPYME Crear Tacna ODEF AREGAK EDPYME Crear Trujillo Otiv Sambava ASA EDPYME EDYFICAR PADME Al Amana EDPYME PROEMPRESA PAMECAS BANTRA Emprender PAPME BASIX Enda PRASAC BTFF FDL PRESTANIC BURO FIE PRISMA Banco Los Andes ProCredit FIE Gran Poder PRODESA BancoSol FIELCO ProEmpresa Bandhan FINADEV ProMujer CEAPE/MA FINCA TZA ProMujer Nicaragua CEB FINCA Uganda ProMujer Peru CMAC Arequipa FMFB Pakistan RCPB CMAC Maynas FMM Bucaramanga SHARE CMAC Tacna FMM Popay SPANDANA CMAC Trujillo FOCCAS TSKI CMF FORA TSPI CMS FUNBODEM UM PAMECAS COAC Jardin Azuayo FinDev UTrust / UWFT COAC Maquita Cushunchic Fundacion Espoir VF CODESARROLLO Fundacion Leon 2000 VMCA CRECER GK WAGES CREDIT HKL WWB Medell CREDO HdH Zakoura CRYSTAL FUND 51 Appendix4. Country list without rated MFIs Albania Malawi Sierra Leone Cameroon Mexico South Africa Chad Moldova Sri Lanka Chile Mongolia Tajikistan Colombia Mozambique Thailand Costa Rica Nepal Togo Ghana Nigeria Venezuela Guatemala Poland Zambia Haiti Romania Zimbabwe Indonesia Rwanda Appendix5. Results of the VIFs (variance inflation factor) for the test of multicollinearity As a rule of thumb, the variables whose VIF values are greater than 10 may merit further investigation. Some variables in the specifications show rather high VIF but the means VIFs are less than 10 for all the specifications, indicating that multicollinearity is not a problem of the models. VIF (Model 3) Variable VIF 1/VIF L_MFISIZE 3.23 0.309442 L_NAB 2.74 0.364687 GDPC 2.19 0.456419 ECONSIZE 2.14 0.468078 L_CAPITAL 1.95 0.513539 NGO 1.76 0.567981 L_SAVINGS 1.74 0.573416 MFISPECIAL 1.65 0.606373 L_OSS 1.65 0.606751 L_ROA 1.63 0.615173 EconFreedom 1.63 0.615193 L_LOAN 1.57 0.636492 REGULATED 1.47 0.680684 52 Variable VIF 1/VIF N_GRADE 1.46 0.685660 L_PAR 1.41 0.706788 INFLATION 1.27 0.788425 AGE 1.27 0.789055 UPGRADE 1.25 0.798007 SUBSIDIZED 1.20 0.832319 L_AUDIT 1.15 0.873276 Mean VIF 1.72 VIF (Model 4) Variable VIF 1/VIF N_GRADE*NGO 22.24 0.044964 NGO 21.16 0.047259 L_MFISIZE 3.35 0.298507 N_GRADE 3.16 0.316456 L_NAB 2.78 0.359712 GDPC 2.26 0.442478 ECONSIZE 2.16 0.462963 L_CAPITAL 1.96 0.510204 MFISPECIAL 1.94 0.515464 L_SAVINGS 1.78 0.561798 L_OSS 1.65 0.606061 EconFreedom 1.64 0.609756 L_ROA 1.62 0.617284 L_LOAN 1.53 0.653595 REGULATED 1.45 0.689655 L_PAR 1.39 0.719424 INFLATION 1.31 0.763359 AGE 1.28 0.78125 UPGRADE 1.24 0.806452 SUBSIDIZED 1.21 0.826446 L_AUDIT 1.17 0.854701 Mean VIF 3.73