"MICROFINANCE MISSION DRIFT - A STUDY OF MICROFINANCE INSTITUTIONS IN ASIA AND LATIN AMERICA" by Keely Dempsey B.B.A., Thompson Rivers University, 2008 THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS IN DEVELOPMENT ECONOMICS UNIVERSITY OF NORTHERN BRITISH COLUMBIA December 2011 ©Keely Dempsey, 2011 UMI Number: MR87559 All rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. ttswWioft FtoMstfiriii UMI MR87559 Published by ProQuest LLC 2012. Copyright in the Dissertation held by the Author. Microform Edition © ProQuest LLC. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code. ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 Abstract Existing literature suggests that while pursuing self-sufficiency microfinance institutions (MFIs) become vulnerable to mission drift as a result of increasingly targeting more well-off borrowers and foregoing their poorest clients. This thesis investigates whether or not the pursuance of operational selfsufficiency tends to drive MFIs away from the poorest borrowers. This study employs panel least squares, fixed effects and random effects and uses one sample of 223 MFIs in Latin America and the Caribbean and one sample of 196 MFIs in Asia from 2005-2009. The results show that while pursing operational self- sufficiency all MFIs in Latin America and the Caribbean are susceptible to mission drift. The results are less pronounced for Asia. Table of Contents Abstract ii Table of Contents iii List of Tables v Glossary vi Acknowledgement viii Introduction 1 Chapter II Economic and Socio-economic Analysis of Asia and Latin America and the Caribbean Macro-economic Conditions in Asia and Latin America and the Caribbean Macro-economic Conditions in Asia Marco-economic Conditions in Latin America and the Caribbean Socio-economic Conditions in Asia and Latin America and the Caribbean Socio-economic Conditions in Asia Socio-economic Conditions in Latin America and the Caribbean Financial Sector in Asia and Latin America and the Caribbean Financial Sector in Asia Financial Sector in Latin America and the Caribbean Microfinance in Asia and Latin America and the Caribbean Microfinance in Asia Microfinance in Latin America and the Caribbean 7 7 9 13 16 16 17 19 19 20 23 23 25 Chapter III Review of Literature Economies of Scale Operational Self-Sustainability Mission Drift Asia vs. Latin America and the Caribbean 30 30 31 34 35 Chapter IV Database and Methodology Database Regression Approach Hypothesis Development 36 36 39 42 iii Chapter V Empirical Results Descriptive Statistics Economies of Scale Operational Self-Sustainability Mission Drift Conclusion 43 43 49 53 58 63 Chapter VI Conclusion ^ Bibliography Appendix 1 ^ % Households with Access to a Bank Account iv ^ List of Tables Table 2.1: Macro and Livelihood Indicators for Asia and Latin America and the Caribbean 8 Table 2.1.1: Macroeconomic Indicators for Selected Asian Economies 12 Table 2.1.2: Macroeconomic Indicators for Western Hemisphere Economies 15 Table 2.4.1: Microfinance Growth Indicators for Asia 28 Table 2.4.2: Microfinance Growth Indicators for Latin America and the Caribbean 29 Table 4.1.1: Distribution of Microfinance Institutions in Latin America and the Caribbean 38 Table 4.1.2: Distribution of Microfinance Institutions in Asia 38 Table 5.1.1: Descriptive Statistics of MFIs in Asia and Latin America and the Caribbean: 2005-2009 46 Table 5.1.2: Correlation Coefficients among the Explanatory Variables 48 Table 5.2(a): Analysis of Economies of Scale for MFIs in Asia 52 Table 5.2(b): Analysis of Economies of Scale for MFIs in Latin America and the Caribbean 52 Table 5.3(a): Analysis of OSS for MFIs in Asia 55 Table 5.3(b): Analysis of OSS for MFIs in Latin America and the Caribbean 57 Table 5.4(a): Analysis of Mission Drift for MFIs in Asia 60 Table 5.4(b): Analysis of Mission Drift for MFIs in Latin America and the Caribbean 62 v Glossary Term Definition Assets Total of all gross loans. Commercialization In a microfinance context, commercialization refers to the move by MFIs to provide services on a financially self-sufficient basis and under prevailing commercial principle and regulations. Cross-Subsidization Lending Consists of specifically targeting unbanked wealthier clientele. Average Loan Balance per Borrower Loan Portfolio, Gross / Number of Active Borrowers Borrowers per Loan Officer Number of Active Borrowers / Number of Loan Officers Cost per Borrower Operating Expense/ Number of Active Borrowers , average Financial Self-Sufficiency A percentage which indicates whether or not enough revenue has been earned to cover both direct costs including financing costs, provisions for loan losses, and operating expenses - and indirect costs, including the adjusted cost of capital. Group Lending Lending mechanism which allows a group of individuals - often called a solidarity group - to provide collateral or loan guarantee through a group repayment pledge. The incentive to repay the loan is based on peer pressure - if one group member defaults, the other group members make up the payment amount. Individual Lending Single-client lending where repayment relies solely on the individual. vi Term Definition Loan Portfolio, gross All outstanding principals due for all outstanding client loans. This includes current, delinquent, and renegotiated loans, but not loans that have been written off. It does not include interest receivable. Micro-credit Another name for a micro-loan. A part of the field of microfinance, microcredit is the provision of credit services to low-income entrepreneurs. Microfinance Institution An institution that provides financial services to the world's poor. Number of Active Borrowers The number of individuals or entities who currently have an outstanding loan balance with the MFI or are primarily responsible for repaying any portion of the Loan Portfolio, Gross. Individuals who have multiple loans with an MFI should be counted as a single borrower. Operational SelfSufficiency A percentage which indicates whether or not enough revenue has been earned to cover the MFIs total costs - operational expenses, loan loss provisions and financial costs. Portfolio at Risk > [XX] days The value of all loans outstanding that have one or more instalments of principal past due more than [XX] days. Progressive Lending Pertains to the notion that existing clientele are able to obtain higher loans after achieving a flawless repayment schedule. Return on Assets (%) (Net Operating Income, less Taxes)/ Assets, average Return on Equity (%) (Net Operating Income, less Taxes)/ Equity, average Unbanked A term used to describe the world's working poor who are not able to participate in the formal banking sectors. Yield on Gross Portfolio (nominal) (%) Interest and Fees on Loan Portfolio/ Loan Portfolio, gross, average vii Acknowledgement I would like to sincerely thank my parents, Ed and Jeanne Dempsey, and my sister, Lindsay Dempsey, for their continued support and encouragement throughout this endeavour. I would also like to thank my husband, John Maxwell, for always believing in me. I would also like to acknowledge my supervisor, Ajit Dayanandan, for his commitment to seeing me succeed. I thank you all. viii Introduction The concept of micro-credit, pioneered by Nobel Peace Prize winner Muhammad Yunus of Bangladesh in 1976, relates to the provision of financial services to the poor.1 The microfinance movement has encompassed all parts of the globe and has now reached approximately one hundred fifty million households worldwide. 2 In spite of the industry's double digit growth (approximately ten per cent), microfinance is estimated to satisfy only ten per cent of the entire global market demand (Dieckmann, 2007). Microfinance is now considered as a means of creating employment, enabling households to provide for education/health expenditures, reducing poverty, aiding households to smoothen consumption in the wake of unexpected economic shocks, promoting the bargaining power of women in the household, gender equality, creating social capital and promoting a bottom up growth process (Morduch, 2002; Khandekar, 2005). For microfinance, however, the crucial aspect of the revolution is that it has the ability to provide large-scale outreach profitably (Robinson, 2002). The microfinance industry is a highly unique industry in the sense that it is theoretically subject to satisfying a double bottom line.3 In addition to maintaining a social or developmental mission recent literature has suggested that in order for microfinance institutions (MFIs) to remain operational a movement away from the 'Microfinance embraces not only small loans but includes savings, insurance, fund transfers and other related services. 2http://microfinance.cgap.org/2010/05/17/microfinance-in-2010 ( Retrieved August 31, 2011) 3Ibid 1 reliance on subsidies and donor support and towards that of financial self- sustainability is required (Christen et al, 2004; ADA, 2009). Self-sustainability entails that microfinance institutions are to generate sufficient revenues through the provision of client services to cover the full cost of providing services. For those MFIs that have chosen to embrace the notion of self-sustainability, pursing profits has become an attractive vehicle for doing so (Cull et al, 2007; Berger et al, 2006). In recent years there has been considerable debate of whether or not the laudable goals of the double bottom line have been compromised as microfinance institutions allegedly shift away from serving the poorest borrowers in pursuit of commercial viability (Drake and Rhyne, 2002; Olivers-Polanco, 2005; Copestake, 2007; Cull et al, 2007; Dichter andHarper, 2007; Hishigsuren, 2007;Mersaland and Strom, 2010). The emphasis on the pursuit of profits by microfinance institutions has provoked a heated debate between Mohammed Yunus, founder of Grameen Bank, and microfinance institutions that have formulated using various adaptations of the Grameen Bank model. The provocation for this debate began in 1992 when PRODEM, a Bolivian non-governmental MFI, commercialized and transformed into a share-holder owned BancoSol. The initial public offering of Banco Compartamos in Mexico intensified the debate. Yunus advocated that microfinance institutions be social businesses driven explicitly by social missions and identified Banco Compartamos as a brutal money lender (Cull et al, 2009; Malkin, 2008).4 Yunus warns that by seeking commercial orientation the 4http://www.compartamos.com/wps/portal/Inicio.Retrieved November 14, 2011. 2 microfinance industry is at risk of subverting from the original mission of poverty alleviation (Carrick and Santos, 2007). In recent years the increasing emphasis on financial sustainability rather than on social mission has led to allegations of mission drift (Sinha and Brar, 2005; Armendariz and Szafarz, 2009). However, existing studies have shown mixed results, with most actually showing little or no evidence of mission drift at all (Gonzalez-Vega et al, 1997; Rhyne 1998; Christen, 2001; Christen and Drake, 2002; Mersland and Stram, 2010). Contradictory, indirect evidence provided by the global database Microfinance Information eXchange and from rating agencies for microfinance institutions show that the size of the average loan provided to borrowers has increased in almost all countries; while simultaneously, the numbers of borrowers introduced into the microfinance system has declined.5,6 Given the divergence in theoretical and empirical evidence it is timely to initiate an innovative study focussed on the mission drift debate from a regional perspective. The main research question that this thesis aims to address is whether or not the pursuance of self-sustainability tends to drive microfinance institutions away from the poorest borrowers; and if so, is there evidence to support regional differences between Asia and Latin America and the Caribbean. The subject matter of mission drift from a regional perspective is approached by comparing the experience of MFIs where microfinance activity is the highest in the world; Asia and Latin America and the Caribbean (Armendariz 5 www.Ratingfund.org is an example of a rating agency. 6The average loan balance per borrower is the most commonly used measure of mission drift; however, is by no means the only indicator and is subject to drawbacks. Other indicators include: the percent of women borrowers, household index surveys, monthly household income per capita, etc. 3 and Morduch, 2010). According to the Microfmance Information eXchange, the top ten MFIs in the world in terms of the number of clients served are all located in one of these two regions. However, despite this similarity, in terms of economic and socio-economic structure Asia and Latin America and the Caribbean are extremely diverse. In regards to estimates of poverty, nearly thirtyone per cent of the world's poor live in South Asia while only eight per cent live in Latin America (Armendariz and Szafarz, 2009). Additively, the gross domestic product per capita of Latin America is nearly six times that of South Asia, which in turn provides a greater scope for cross-subsidization lending in Latin America (Armendariz and Szafarz, 2009). The two regions have also developed extremely diverse models of microfmance. The Asian model of microfmance is driven by a strong sense of developmental focus with emphasis on social impacts and follows closely to that portrayed by the Grameen Bank (Montgomery and Weiss, 2005). The emphasis is on social and economic objectives (Armendariz and Szafarz, 2009). The Asian model of microfmance is focussed mainly on povertyalleviation represented by the fact that the region leads the world in terms of both breadth and depth of outreach (Montgomery and Weiss, 2005). The Latin American model of microfmance is predominately a for-profit model focussed exclusively on pursing and obtaining financial objectives (Berger et al, 2006). MFIs in Latin America prefer to lend their funds to the urban economically-active poor, individuals with established businesses that require capital to grow(Montgomery and Weiss, 2005). This fact raises the issue of the forgotten rural poor and mission drift. For the most successful MFIs in Latin America 4 profitability is on par with the region's major international banks (Berger et al, 2006). Some of the prominent MFIs in Latin America, like BancoSol and BancoComparatamos in Mexico, have issued shares and accessed funds from the capital market. Accessing private capital, especially foreign private capital, has resulted in many MFIs in Latin America changing from their philanthropic roots to profit-seeking commercial paradigms (Dees, 1996). One instance of similarity between the two regions in regards to the microfinance industry indicates that many MFIs in both Asia and Latin America and the Caribbean have moved away from following a group-lending model to that of individual lending. This raises the concern of whether these MFIs are seeking well-off (or mature) borrowers rather than seeking out themore risky ultra-poor borrowers. A second illustration of similarity has to do with the occurrence of microfinance bundling; although this is prevalent in both regions there is a higher frequency in Latin America and the Caribbean than in Asia. Microfinance bundling refers to the concept of bundling micro loans with other microfinance products, i.e. micro insurance. The bundling of microfinance products may have negative implications as it is associated with coercion and could lead to collusion among microfinance service providers. This thesisis structured as follows: Chapter II will review the economic and socio-economic backgrounds for sample regions Asia and Latin America and the Caribbean. Chapter II is divided into for sub-sections in order to independently review the macro-economic, socio-economic, and financial states as well as to review the microfinance industry and the role that it plays in providing credit to 5 the poor. This chapter utilizes the IMF World Economic Outlook Report 2010 and the Human Development Index 2010 as the basis for the macro-economic and socio-economic analyses. Chapter III will develop the hypotheses for this analysis and review the relevant literature pertaining to the sustainability of microfinance institutions, economies of scale, and mission drift. Chapter IV will outline the data and methodology used in this analysis. Chapter Vwill present the empirical results and Chapter VI will conclude. 6 Chapter II Economic and Socio-economic Analyses of Asia and Latin America and the Caribbean This chapter briefly reviews the macro-economic and livelihood conditions for sample regions Asia and Latin America and the Caribbean. This chapter is organized into four sub-sections. Section 2.1 reviews the current macroeconomic state of the two regions based from the recovery phase of the global economic crisis. Section 2.2 reviews the current socio-economic conditions utilizing data from the Human Development Index 2010. Section 2.3 provides a brief summary of the financial conditions in both regions. Section 2.4 reviews the microfinance industry and the role that it plays in providing credit to the poor. 2.1 Macro-economic Conditions in Asia and Latin America and the Caribbean Recovery from the financial crisis fared better than what was originally predicted, with the economies considered as either developing or emerging recovering faster than the advanced economies (IMF, 2010). Among the emerging and developing economies, emerging Asia has been leading the recovery with growth also solidifying in key Latin American economies (IMF, 2010).Table 2.1 provides macroeconomic and livelihood indicators for Asia and Latin America and the Caribbean. 7 Table 2.1: Macro and Livelihood Indicators for Asia and Latin America and the Caribbean Countries Asia Korea Hong Kong Singapore China India Indonesia Thailand Philippines Malaysia Vietnam Population (Millions) 2009 49 7 5 1,331 1,155 230 68 92 27 87 GDP 2009 832,512 215,355 182,232 4,984,731 1,310,171 540,277 263,856 160,476 191,601 91,854 Density People per Sq km 502 6,696 6,943 142 383 125 132 303 82 278 International Poverty Line Survey Population Below $1.25 per day(%) Year - 28.41 2005a - - - - 49.41 24.4' <2 22 <2 24.2 2004-053 2007a 2004a 2006a 2004a 2006a - - - - South Asia 1,568 1,634,623 1,944 6,345,309 EAP Latin America and the Caribbean 324 122 Mexico Brazil Argentina Columbia Venezuela Peru Chile Ecuador Bolivia Uruguay Paraguay 107 4 40 8 28 29 17 4 10 3 6 874,902 1,571,979 308,741 230,844 326,498 126,734 163,670 57,249 17,340 36,093 15,015 55 23 15 41 32 23 23 49 9 19 16 <2 1.6 1.0 6.1 18.4 8.2 <0.5 3.2 9.7 <2 9.3 2004a 2007° 2006bc 2006c 2006° 2007° LAC 572 3,976,530 28 - - 2006c 2007° 2007° 2007° 2007c Note: a. Expenditure base. b. Covers urban area only. c. Income base. f. Weighted average of rural and urban estimates Source: World Development Indicators 8 Dom.Credit Provided by the Banking Sector (% of GDP) 112 125 94 145 73 37 146 46 116 95 73 38 46 118 27 43 20 19 116 20 55 34 21 72 2.1.1 Macro-economic Conditions in Asia Despite that the effect of the economic downturn on Asian economies was sharper than what was initially projected, the region has rebounded quickly and has taken the lead in the global recovery (IMF, 2010). Asia can be segregated categorically into five economies: advanced, newly industrialized, developing, ASEAN-5, and other developing Asia. Emerging Asia is a combination of the newly industrialized and developing economies. According to the IMF's World Economic Outlook there are four factors that are responsible for Asia's rapid and robust recovery. Foreconomies such as China and India the normalization of trade and the swift turnaround in inventory cycles has increased the demand for retail and industrial production. Secondly, the introduction of a proactive policy implemented a priori to the crisis in an effort to off-set the downward pull of exports facilitated the strong recovery. This policy allowed many Asian economies to create strong public and private components as well as a third factor - resilient domestic demand. The newly industrialized economies have capitalized on the rebounding inventory cycle and strong domestic and regional demand to facilitate a rapid recovery. The ASEAN-5 economies have also taken advantage of the increased regional activity, specifically those countries exporting electronics and commodities. Fourthly, the resumption of capital inflows also reinforced domestic demand and created access to external funding. The advanced region of Japan was able to partially capitalize on the success of the export industry; however, unanticipated currency appreciation of the yen, excess 9 capacity, and a weak labour market is projected to negatively affect output growth. In 2009 annual growth in real GDP in almost all of the Asian economies was higher than expected (IMF, 2010). The advanced economies of Asia underperformed relative to the other economies with annual growth in output reporting a 3.0 percent decline from the previous year.7 The substantial fall in growth was highly influenced by the dramatic deceleration of growth in Japan. On the contrary, in 2009 developing Asian economies outperformed other Asian economies with growth reporting a 6.9 percent increase from the previous year. The significant growth in real GDP can be attributed to the exceptionally high levels of growth from India and China. The robust activity from within these two regions is projected to facilitate growth in the rest of Asia. Projected forward into 2010 and 2011, Asia's GDP is expected to grow by 7.9 and 6.7 percent, respectively. Once again, the economies considered as the advanced economies are expected to underperform relative to the other Asian economies. Developing Asia is projected to continue leading the recovery with annual growth rates over the two years ranging between 8.4 and 10.5 percent (Table 2.1.1). In 2009 Asia's inflation reached 2.0 percent and in 2010 is projected to see a year-over-year annual increase of 4.3 percent. In 2011 inflation is projected to be slightly lower at 3.3percent. In 2010 and 2011 Japan is projected to be the only economyin the region that will experience deflation with the change in consumer prices reporting a deceleration of 1.0 and 0.3 percent, respectively. 7A11 data within this section is taken from - IMF, (2010). World Economic Outlook: Recovery. Risk and Rebalancing. 10 Affected by Japan's deflation, in 2010 and 2011, the economies of advanced Asia are expected to have the lowest levels of inflation relative to all other Asian economies. The newly industrialized economies are projected to have the second lowest level of inflation with changes in consumer prices reporting an increase of 2.6 and 2.7 percent, respectively. The ASEAN-5, China, and India are expected to fall in between with inflation ranging between 4.6 and 6.0 percent. The economies considered as other developing Asia are projected to have the highest change in consumer prices with inflation reaching 9.1 percent in 2010 up to as high as 9.6 percent in 2011. Within Asia unemployment rates vary depending on the specific economy; however, within most regions have remained fairly low. Due to the unavailability of data for all economies an aggregate unemployment rate for all of Asia is not available; however, is available for certain autonomous regions. In 2009, the advanced economy in Asia reported an unemployment rate of 4.9 percent. The newly industrialized economies reported a slightly lower unemployment rate of 4.3 percent. Countries within the ASEAN-5 recorded varying unemployment rates ranging as low as 1.4 percent in Thailand up to as high as 8.0 percent in Indonesia.In 2010 and 2011 the unemployment rate for the advanced economies is projected to decline to 4.7 and 4.6 percent, respectively. For the newly industrialized economies the unemployment rate is also expected to decline in 2010 and 2011 to 3.8 and 3.7 percent, respectively. For the countries within the ASEAN-5 the varying levels of unemployment are expected to continue. 11 Table 2.1.1: Macroeconomic Indicators for Selected Asian Economies Real GDP Projections 2009 2010 2011 Consumer Prices' Projections 2009 2010 2011 Unemployment Projections 2009 2011 2010 Advanced -3.0 4.6 2.8 -0.1 0.7 1.2 4.9 4.7 4.6 Advanced - Japan -5.2 2.8 1.5 -1.4 -1.0 -0.3 5.1 5.1 5.0 N1E -0.9 7.8 4.5 1.3 2.6 2.7 4.3 3.8 3.7 Korea Taiwan Hong Kong Singapore 0.2 -1.9 -2.8 -1.3 6.1 9.3 6.0 15.0 4.5 4.4 4.7 4.5 2.8 -0.9 0.5 0.6 3.1 1.5 2.7 2.8 3.4 1.5 3.0 2.4 3.7 5.8 5.1 3.0 3.3 5.3 4.4 2.1 3.3 4.9 4.1 2.2 Developing Asia 6.9 9.4 8.4 3.1 6.1 4.2 - - - China India 9.1 5.7 10.5 9.7 9.6 8.4 -0.7 10.9 3.5 13.2 2.7 6.7 4.3 4.1 4.0 - - - ASEAN-5 1.7 6.6 5.4 2.9 4.4 4.4 - - - Indonesia Thailand Philippines Malaysia Vietnam 4.5 -2.2 1.1 -1.7 5.3 6.0 7.5 7.0 6.7 6.5 6.2 4.0 4.5 5.3 6.8 4.8 -0.8 3.2 0.6 6.7 5.1 3.0 4.5 2.2 8.4 5.5 2.8 4.0 2.1 8.0 8.0 1.4 7.5 3.7 6.0 7.5 1.4 7.2 3.5 5.0 7.0 1.4 7.2 3.2 5.0 4.4 5.3 4.6 11.2 9.1 9.6 - - - 5.8 3.6 9.2 7.9 7.9 6.7 2.8 2.0 5.6 4.3 4.0 3.3 - - - - - - Other Developing Emerging Asia Asia Movements in consumer prices are shown as annual averages. Source: IMF. (2010). World Economic Outlook: Recovery. Risk and Rebalancing. Other Developing Asia: Afghanistan, Bangladesh, Bhutan, Brunei Darussalam, Cambodia, Fiji, Kiribati, Lao, Maldives, Myanmar, Nepal, Pakistan, Papua New Guinea, Samoa, Solomon Islands, Timor, Sri Lanka, Tonga, Vanuatu. 12 2.1.2 Macro-economic Conditions in Latin America and the Caribbean Latin America and the Caribbean (LAC) has also been recovering from the downturn quicker and more robust than originally anticipated (IMF, 2010). The region is categorized into three economies: South America, Central America, and the Caribbean (IMF, 2010). According to the IMF World Economic Outlook, Latin America and the Caribbean's robust recovery can be attributed to various factors: solid macroeconomic policy, policy support, external financing conditions, robust commodity export revenues, and sustained domestic demand. All of these components played a significant role in driving the recovery of the LA-4 (Brazil, Chile, Columbia, and Peru). Mexico is also staging a steady recovery; however, the economy's tight financial links to the United States pose it at risk for uncertainty. The potential uncertainty of the U.S. economy also affects Central America and the Caribbean as they have become partially relianton the income from American tourists as well as U.S. remittance flows. As with Asia, the recovery rates within LAC are diverse and dependent upon the strength of macroeconomic policy, resilience of domestic demand, and the degree of exposure to global trade and financial conditions. For those commodity exporting economies output will continue to grow as a result of trading links with China and other intraregional linkages. In 2009 growth in real GDP for Latin America and the Caribbean declined 1.7 percent from the previous year.8 Central America and South America underperformed compared to the Caribbean with declines in output of 0.5 and 0.2 8A11 data within this section is taken from - IMF, (2010). World Economic Outlook: Recovery. Risk and Rebalancing. 13 percent, respectively. Within the LA-4, Chile and Brazil underperformed compared to Columbia and Peru. Growth rate projections for 2010 and 2011 are expected to be significantly higher. In 2010, South America is projected to experience a 6.3 percent increase in output. The substantial increase is due to high projected outputs for Brazil, Argentina, Peru, Uruguay, and Paraguay. Overall, in 2010 and 2011 Latin America and the Caribbean is expected to experience an increase in output of 5.7 and 4.0 percent, respectively (Table 2.1.2). In 2009 inflation for the region was reported at 6.0 percent. This value was highly influenced by Venezuela's incredible 27.1 percent inflation rate. Central America and the Caribbean reported substantially lower rates of inflation than the other economies with 3.8 and 3.5 percent, respectively. With the exception of the Caribbean, projections for 2010 and 2011 remain fairly consistent for all economies. Unemployment within the region was fairly high in 2009 compared to Asia. Unfortunately, the aggregate unemployment rate for all economies within Latin America and the Caribbean is unavailable due the inconsistency of data; however, unemployment rates for many regions individually are provided. For the LAC economies for which data is available, the majority of unemployment rates in 2009 ranged between 7.3 percent and 9.6 percent. With the exception of Columbia and Paraguay, unemployment rates within South America were fairly similar for all regions. Aggregate figures for the Caribbean are not available; however, it is expected that due to a high reliance on the agricultural sector unemployment rates will remain an on-going challenge. 14 Table 2.1.2: Macroeconomic Indicators for Western Hemisphere Economies Mexico Real GDP Consumer Prices1 Projections Projections 2009 2010 2011 20139 201C 2011 -6.5 5.0 3.9 5.3 4.2 2009 3.2 South America -0.2 6.3 Brazil Argentina Columbia Venezuela Peru Chile Ecuador Bolivia Uruguay Paraguay -0.2 0.9 0.8 -3.3 0.9 -1.5 0.4 3.4 2.9 -3.8 Central America The Caribbean LAC -0.5 0.4 -1.7 4.1 6.4 6.8 6.9 7.5 7.5 4.7 -1.3 8.3 5.0 2.9 4.0 8.5 9.0 4.1 4.0 4.6 0.5 6.0 6.0 2.3 4.5 5.0 5.0 4.9 6.3 4.2 27.1 2.9 1.7 5.2 3.3 7.1 2.6 5.0 10.6 2.4 29.2 1.7 1.7 4.0 1.7 6.5 4.6 4.6 10.6 2.6 32.2 2.5 3.0 3.5 4.1 6.4 5.2 3.1 2.4 5.7 3.7 4.3 4.0 3.8 3.5 6.0 3.9 7.2 6.1 4.1 5.5 5.8 Unemployment Projections 2010 2011 5.5 5.0 4.5 - - - 8.1 8.4 12.0 7.9 8.6 9.6 8.5 7.2 8.0 12.0 8.6 8.0 9.0 8.6 7.5 8.6 11.5 8.1 7.5 8.7 8.5 - - - 7.3 5.6 7.0 5.3 6.9 5.2 - - - - - - - - - Movements in consumer prices are shown as annual averages. Source: IMF. (2010). World Economic Outlook: Recovery. Risk and Rebalancing. Central America: Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua, Panama. The Caribbean: Antigua, Barbuda, The Bahamas, Dominican Republic, Belize, Dominica, Grenada, Guyana, Haiti, Jamaica, St. Kitts, Nevis, St. Lucia, St. Vincent, Grenadine, Suriname, Trinidad, Tobago. 15 2.2 Socio-economic Conditions in Asia and Latin America and the Caribbean Poverty is measured using the Multidimensional Poverty Index (MPI), developed in 2010 by Oxford Poverty & Human Development Initiative and the United Nations Development Programme.9 The index complements the monetary based measure of less than US$1,25/day by considering multiple deprivations and their overlap.10 The MPI follows the same dimensions as the Human Development Index. Health, education, and income and are measured using 10 indicators: assets, floor, electricity, toilet, water, cooking fuel, children enrolled, number of years of schooling, child mortality, and nutrition." The MPI is calculated by multiplying the incidence of poverty by the average intensity of poverty.12 2.2.1 Socio-economic Conditions in Asia In Asia, an average of 33.8 percent of the population is considered to be MPI poor (they are deprived in at least 30.0 percent of the weighted indicators, by definition). Those 33.8 percent considered as MPI poor suffer from deprivation in 50.2 percent of the indicators used for this measure. That is to say that those individuals classified as MPI poor suffer from deprivation in an average of five out of the ten indicators listed above. Other developing Asia is the poorest 'Multidimensional Poverty Index (MPI): An index of acute multidimensional poverty. MPI has 3 dimensions: health, education, and standard of living which are measured using 10 indicators. The MPI is calculated by multiplying the incidence of poverty by the average intensity of MPI poverty across the poor. Incidence of poverty: the proportion of people who are poor according to the MPI (those who are deprived in at least 30% of the weight indicators). Intensity of deprivation: percentage of weighted indicators in which an average poor household is deprived. 1 ""Multidimensional Poverty Index" Oxford Poverty and Human Development Initiative. Retrieved 03/01/2011. " Ibid 12 Ibid 16 economy in the region with an average 48.1 percent of the population considered MPI poor. Those 48.1 percent suffer from deprivation in 52.0 percent of indicators. At 12.3 percent, the ASEAN-5 has the lowest average headcount of those considered MPI poor; however, those 12.3 percent suffer from deprivation in 47.6 percent of indicators. Developing Asia ranks in between with 34.0 percent of the population considered to be MPI poor. Those 34.0 percent suffer from deprivation in 49.2 percent of indicators.13 Given that the economy of other developing Asia has the highest headcount of MPI poor it is appropriate that they are deprived in the largest percentage of indicators. This is due to a correlation between headcount and average intensity in which countries with higher MPI headcounts tend to have higher average intensity (Alkire and Santos, 2010). 2.2.2 Socio-economic Conditions in Latin America and the Caribbean In regards to MPI poverty and intensity of deprivation, Latin America and the Caribbean isconsidered far better-off than Asia. Since other developing Asia has a higher MPI headcount than any of the LAC economies it is expected that none of the LAC economies will have a higher intensity of deprivation. InLatin America and the Caribbean, on average, 16.6 percent of the population is MPI poor (they are deprived in at least 30.0 percent of the weighted indicators, by definition). Those who are MPI poor suffer from deprivation in 44.8 percent of indicators. The North American economy has the lowest MPI headcount at 4.0 percent;however, those 4.0 percent suffer from deprivation in 38.9 percent of the indicators. Central America has the largest MPI headcount at 33.1 percent of the 13 Advanced and newly industrialised Asia did not report MPI values. 17 population. Those considered MPI poor in this economy suffer from deprivation in 50.0 percent of the indicators. 18 2.3 Financial Sector in Asia and Latin America and the Caribbean: 2.3.1 Financial Sector in Asia The financial system underdeveloped compared in developing Asia continues to remain with that of the industrial economies (Asian Development Bank, 2010). Since the Asian crisis in 1997 developing Asia has been making significant strides to establish a sounder and more efficient financial system through the efforts of extensive post-crisis reform and restructuring (Asian Development Bank, 2010). Asia's commercial banks have responded remarkably well to the post-crisis reform. They have not only improved and expanded their existing financial products and services but have undertaken new business ventures, including: investment banking, consumer lending, and real estate. However, recent research from the Asian Development Bank (ADB) signifies that the financial sector is becoming less dependent on banks as the system develops and diversifies. The Asian Development Bank analyzed the regions financial depth (the size of the financial system relative to GDP) as a proxy for financial system development (Asian Development Bank, 2010). The results indicated that Asia's aggregate financial depth had increased since the 1990's, signifying that the financial system has been developing. Concurrently, the growing importance of equity markets further signifies that the region appears to be heading towards a predominately market-based system (Asian Development Bank, 2010). However, despite the increased diversification in system structure access to the financial services within the region lags far behind that of high income OECD countries (Asian Development Bank, 2010). According to a recent study, approximately 19 59.0 and 58.0 percent of the total population for East Asia and South Asia, respectively, are financially un-served (Chala et al, 2009). See Appendix 1 for the percentage of households with access to a bank account for selected Asian countries. 2.3.2 Financial Sector in Latin America and the Caribbean The financial sector in Latin America and the Caribbean has undergone tremendous trials and tribulations since the early 1990's. From one debt crisis to the next the regions limited access to bank credit and lingering financial system uncertainties have impacted the regions constrained economic growth (Belaisch et al, 2005). Financial liberalization and the promise of reforms spurred credit growth in the early 1990's; however, due to an onslaught of activities banking crises erupted throughout the region and banking quickly deteriorated (Stallings and Studart, 2002; Belaisch et al, 2005). In an attempt to rectify the severely damaged financial system banks were restructured and/or recapitalized; however, this was successful in only some Latin American countries but not all. For those countries where it was successful the restructuring strengthened the financial system and decreased the chance a recurring crisis (Belaisch et al, 2005). Unfortunately, simultaneously, the restructuring allowed for an overhaul of the banks regulatory systems (Belaisch et al, 2005). This coupled with the relaxation and/or elimination of foreign bank regulatory limitations led to foreign banks gaining an increased market share within the system (Belaisch et al, 2005). By the year 2000 in Chile, Argentina, Mexico, Peru, Paraguay, and Venezuela foreign banks owned more than one half of the banking system (Belaisch et al, 2005). As 20 of recently, a second wave of financial crises hit Latin America. The countries hit included Ecuador (1999), Argentina (2001), Uruguay (2002), the Dominican Republic (2001) and Bolivia in (2003). Today, Latin America and the Caribbean is primarily a bank-based financial system with minute and illiquid security markets (Stallings and Studart, 2002). In times of uncertainties banks have retained a comparative advantage in obtaining information that is crucial to financial intermediation (Belaisch et al, 2005). World Bank literature analyzed the market classification by evaluating the regions structure-size (stock market capitalization to GDP/bank credit to GDP) and structure-activity (stock market value traded to GDP/bank credit to GDP) to verify the predominance of a bank-based financial system.14 Given that both values for Latin America and the Caribbean declined confirms that the region continues to operate mainly as a bank-based system (Beck and Demirgiic-Kunt, 2009). In most Latin American countries the private sector's use of bond and equity markets to raise financial resources remains limited relative to its resource to banks (Belaisch et al, 2005). In regards to financial depth (the size of the financial system relative to GDP), Latin America is much smaller than Asia and consequently considered to be less developed (Cuadroe/ al, 2002). Deposit to GDP ratios are less than 50.0 percent compared with the typical ratio of 90.0 percent in the emerging markets of East Asia (Belaisch et al, 2005). Given that financial depth is correlated with financial efficiency it is highly likely that Latin America's financial system is also less efficient (Bossone and Lee, 2002). Latin America has a larger proportion of unbanked individuals than Asia; 65.0 percent 14 A higher value for both indicators specifies a more likely market-based financial system 21 of the total population in Latin America is un-served bythe formal financial system (Chala et al, 2009). See Appendix 1 for the percentage of households with access to a bank account for selected Latin American and the Caribbean countries. 22 2.4 Microfinance in Asia and Latin America and the Caribbean The microfinance industries in Asia and Latin America and the Caribbean were created out of extremely diverse ideological, political, and economic conditions. As a result, the microfinance industries that currently exist within the two regions today portray very distinct differences. The microfinance industry in Asia was born in Bangladesh in the early 1970's when Professor Muhammad Yunus undertook a research project that involved providing small amounts of credit to the poor. This endeavour was purely socially driven and has since embedded the roots for which the microfinance industry in Asia stems from today. On the contrary, microfinance in Latin America and the Caribbean developed under significantly dissimilar conditions. In Bolivia, when a collapsing populist regime threatened to unleash widespread unemployment Banco Sol, a microfinance institution, stepped-in to provide credit to the cash constrained informal sector. Not so long after, the region was introduced to the potential of commercial profitability. As a result, the regions lending methodologies are less concerned with the rural poor and targeting poverty and are more so focussed on the economically active poor and supporting microenterprise endeavours. 2.4.1 Microfinance in Asia Today Asia's microfinance institutions continue to remain socially engaged and lead the world in terms of both breadth (number of borrowers) and depth (relative poverty of borrowers) of outreach. In 2009 the microfinance industries in South and East Asia reached 50.0 million and 13.9 million 23 borrowers, respectively.15 The size of the loan balance provided to borrowers in Asia is lower than the rest of the world, likely indicating a more dedicated focus on low-income clients (Microfinance Information eXchange, 2010). In 2009, the median size of the average loan balance provided to borrowers from MFIs in South and East Asia equated to SUS141 and SUS331, respectively. Outreach to women is higher in Asia than the rest of the world with the average institution at 94.4 percent. The vast majority of credit products are used to support microenterprise activities, with small shares going towards consumer lending, education, and mortgage or housing loans (Microfinance Information eXchange, 2010). In 2009 the region accrued $US4 billion in deposits from 38.0 million depositors. These deposits were used as a significant source of financing for Asian MFIs. Typically MFIs have three types of financing available: deposits, borrowings, and equity (Microfinance Information eXchange, 2010). Historically, equity has been the main source of funding for the region, providing over 60.0 percent in 2003. Over the past few years growth in equity has been surpassed by both growth in deposits and growth in borrowings. By 2008 equity provided only 30.0 percent of the funding to the region with deposits and borrowings comprising two-thirds of total financing (Microfinance Information eXchange, 2010). The cost of providing loans to borrowers in Asia is significantly less than in Latin America and the Caribbean. In 2009 the median cost per borrower for MFIs in South and East Asia equated to $US18 and $US65, respectively. The 15 All data in this section is taken from the Microfinance Information eXchange. Retrieved September 1, 2011.From http://www.mixmarket.org/. 24 densely populated low income areas in Asia allow for loan officers to provide and monitor loans at relatively lower costs. In 2009 the portfolio at risk over 30 and 90 days remained relatively stable for MFIs in South Asia; however, for MFIs in East Asia and the Pacific the median value for both indicators increased slightly.In 2009 MFIs in South Asia experienced an increase in return on equity from the previous year; however, for MFIs in East Asia and the Pacific return on equity declined. 2.4.2 Microfinance in Latin America and the Caribbean In 2009 the cumulative credit portfolio for microfinance institutions in Latin American and the Caribbean reached approximately SUS20 billion.16 The cumulative number of active borrowers reached 14.3 million, a growth rate of 9.0 percent over the previous year. Although the MFIs in Latin America and the Caribbean reached substantially fewer borrowers than MFIs in Asia; the median value for the size of the loan balance per borrower was significantly higher. In 2009 the median value for the size of the loan balance per borrower equated to $US917. The regions contribution of women borrowers is significantly fewer than Asia, with the median institutions at 62.0 percent. The number of depositors reached a cumulative total of $US17 million, a growth rate of 27.0 percent. In 2009 deposits were the main sources of funding comprising 52.3 percent of the total funding to the region (Microfinance Information eXchange, 2010). In Latin America and the Caribbean the two major credit portfolio types are microenterprise and consumption, with microenterprise being the more I6A11 data in this section is taken from the Microfinance Information eXchange. Retrieve September 1, 2011. From http://www.mixmarket.org/. 25 favourable portfolio of the two (Microfinance Information eXchange, 2010). In 2009 the number of microenterprise loans increased to over 8,000 while the loan portfolio microenterprise loans increased to just under SUSIO.O million. In 2009 consumption credit saw a solid increase in growth in both the number of loans disbursed and the loan portfolio. Consumption credits grew at 21.2 percent, to just over $US4 million, while the number of loans rose by 9.3 percent, to just over 5,000. Housing credit also experienced a substantial increase in growth; however, the size of the portfolio and number of active loans remains far below microenterprise and consumption at this time (Microfinance Information eXchange, 2010). Despite the increase in the number of loans and size of the credit portfolios the quality of the loans has been decreasing. In 2009 the portfolio at risk over 30 days increased to a median value of 6.0 percent while portfolio at risk over 90 days increased to a median value of 4.0 percent. In 2009 return on equity declined significantly, down from a median value of 9.0 percent in 2008 to that of 6.0 percent in 2009. Prior to the global economic crisis the downturn in profitability was attributed to an increase in industry competition; however,postcrisis the hurdles in microfinance have been accredited to the slowdown in economic activity (Microfinance Information eXchange, 2010). As mentioned above, the cost of providing loans in Latin America and the Caribbean is substantially higher than in Asia. In 2009 the median value for the cost per borrower equated to SUS175. The higher cost per borrower can be 26 attributed to the lower percentage and sparse location of poor individuals residing in Latin America and the Caribbean. For Latin America and the Caribbean MFIs social management is becoming an increasingly recognized area of assessment and monitoring; however, only 27.0 percent of MFIs currently have a standing social performance committee that regularly reviews social performance concerns (Microfinance Information eXchange, 2010). Despite the challenges, the region has made considerable progress towards integrating the social component into their operations. 27 Table 2.4.1: Microfinance Growth Indicators for Asia South Asia 2003 2004 Number of Borrowers Number of Depositors Gross Loan Portfolio ($) Assets ($) Portfolio at Risk > 30 Days Portfolio at Risk > 90 Days ROE Loan Balance/Borrower ($) Cost/Borrower ($) 2005 2006 2007 2008 2009 13,512,280 13,403,380 17,882,185 15,941,327 24,383,439 19,136,200 29,956,551 26,434,335 36,388,113 30,165,506 42,405,238 32,018,765 50,022,228 33,109,727 782,096 817,537 1,322,993 1,944,613 3,738,409 3,957,776 4,391,870 1,178,303 1,083,073 1,793,442 3,429,815 5,195,352 5,801,421 6,314,966 2.7% 1.9% 1.2% 1.2% 1.9% 1.8% 1.9% 1.6% 1.2% 0.7% 1.1% 1.4% 1.1% 1.3% 4.9% 10.8% 12.8% 14.0% 10.5% 8.6% 10.5% 72 76 92 107 137 123 141 11 12 11 12 16 18 18 4,493,322 1,304,673 5,408,477 1,542,817 9,468,511 2,097,382 10,725,120 6,593,152 8,783,244 3,255,613 15,448,626 4,614,980 13,897,499 4,899,512 965,457 1,124,755 1,325,122 1,520,924 1,767,031 1,953,239 3,789,136 1,464,558 1,598,702 1,840,395 2,181,396 2,535,085 2,924,964 5,256,007 5.1% 3.8% 4.3% 5.0% 3.5% 1.7% 4.1% 1.9% 2.2% 2.7% 3.2% 2.4% 1.1% 2.7% 10.1% 15.6% 11.8% 14.2% 13.8% 14.0% 11.2% 200 156 204 245 288 319 331 43 36 42 51 58 66 65 East Asia and the Paci 1c Number of Borrowers Number of Depositors Gross Loan Portfolio ($) Assets ($) Portfolio at Risk >30 Days Portfolio at Risk > 90 Days ROE Loan Balance/Borrower ($) Cost/Borrower ($) Source: MIX Market Database Note: Number of Borrowers, Number of Depositors is a sum value; Gross Loan Portfolio, Assets, Loan Balance/Borrower, Cost/Borrower ROE, Portfolio at Risk at both 30 and 90 days is a median value 28 Table 2.4.2: Microfinance Growth Indicators for Latin America and the Caribbean 2003 2004 2005 2006 Number of Borrowers Number of Depositors Gross Loan Portfolio ($) Assets ($) Portfolio at Risk >30 Days Portfolio at Risk > 90 Days ROE Loan Balance/Borrower ($) Cost/Borrower ($) 2007 2008 2009 3,464,294 943,675 4,742,094 3,210,057 7,805,509 6,350,983 9,440,623 7,640,896 12,053,183 9,710,388 13,064,519 13,564,741 14,278,727 17,168,351 2,866,309 3,290,834 3,325,504 4,076,073 4,730,981 4,741,481 6,165,183 3,786,335 4,075,007 4,254,230 4,957,230 6,188,646 6,391,176 7,810,931 4.7% 5.8% 4.5% 4.1% 3.8% 4.9% 5.7% 3.2% 3.4% 2.6% 2.6% 2.1% 3.0% 3.7% 12.6% 12.6% 10.7% 11.5% 10.3% 8.8% 6.3% 514 609 641 659 746 828 917 124 120 144 146 146 177 176 Source: MIX Market Database Note: Number of Borrowers, Number of Depositors is a sum value; Gross Loan Portfolio, Assets, Loan Balance/Borrower, Cost/Borrower ROE, Portfolio at Risk at both 30 and 90 days is a median value 29 Chapter III Review of Literature This section will provide a review of the existing literature and provides the various hypotheses to be tested in the empirical investigation. This thesis utilizes three empirical estimations: (1) do MFIs attain economies of scale as they increase their outreach;(2) what are the determinants of operational selfsufficiency of MFIs - interest rates, repeated loans to seasoned borrowers or cost; and (3) doespursing self-sustainability tend to drive microfinance institutions away from the poorest borrowers; and if so, is there evidence to support regional differences between Asia and Latin America and the Caribbean. Empirical estimations (1) and (2) provide support and additional information in order to more accurately answer the main research question, (3). 3.1 Economies of Scale Existing literature has indicated that attaining economies of scale through increasing the average loan size provided to borrowers leads to a MFIs ability to attain cost-efficiencies and thus increases the likelihood of attaining selfsufficiency. The work of Brau and Woller (2004) concludes that MFIs are able to attain cost-efficiencies by capturing economies of scale by extending larger loans to mature clients. Navajase? al (2000) suggest that in an attempt to attain commercial funding MFIs will seek to capture economies of scale by extending larger loans to wealthier target markets. The work of Crombrugghe, Tenikue and Sureda (2008) find from their study an opposite result. They conclude that the elasticity of the operating cost per borrower to the size of the average loan 30 provided to borrowers is positive as a result of higher selection and monitoring costs associated with lending larger loans. This leads to the following hypothesis: Hi: MFIs that increase the size of the average loan provided to borrowers will experience cost increases as a result of higher selection and monitoring costs associated with lending larger loans. The final hypothesis in the analysis of economies of scale indicates that there exists a relationship between increasing the scale of an institution and lower average costs.17 Crombrugghe, Tenikue and Sureda (2008) utilize the number of borrowers to represent the scale (outreach) of MFIs. The results conclude that the operating cost per borrower is sensitive to the number of borrowers. Therefore, increasing the number of borrowers would lower the average operating costs and would raise total operating costs less than proportionately with the number of borrowers. This leads to the following hypothesis: H2: MFIs that increase the scale of their institutions in the current state of affairs will decrease the average cost per borrower; therefore, capturing economies of scale. 3.2 Operational Self-Sustainability Existing literature has indicated the presence of a positive relationship between higher interest rates and increased self-sufficiency. Brau and Woller (2004) indicate that as a result of large overhead costs associated with providing microloans, compared with that of formal financial institutions, MFIs are required to charge relatively higher interest rates in order to achieve self-sufficiency. Augsburg and Fouillet (2010) suggest that in order for MFIs to achieve independence from subsidies and donor support cost-covering interest rates are a necessity. Christen et al (1995) studied the performance of 11 leading l7This thesis uses total assets as a proxy to represent the outreach of a microfinance institution. 31 microfinance institutions and found that the most financially viable programs differed from their peers in their willingness to set interest rates that would allow for the full recovery of costs. Cull, Demirgiic-Kunt and Murduch (2007) test this hypothesis by analyzing if higher interest rates lead to agency problems and subsequently to less profitability. Their results conclude that individual lending, as opposed to group lending, tend to be more profitable when average interest rates are higher. However, when MFIs charge interest rates that are too high it could lead to lower profitability. Cull et al (2007) suggest further that when MFIs charge interest rates that are above a certain threshold rate only the low-quality high risk clients will borrower. The work of Crombrugghe, Tenikue and Sureda (2008) find similar results. Their results suggest that raising interest rates can be done without harming profitability; however, charging interest rates that are too high could lead to concerns of moral hazard and charging interest rates that are too low could lead to excess demand. This leads to the following hypothesis: H3: MFIs charging higher interest rates will have a higher level of operational self-sufficiency; however, there exists a threshold rate that once surpassed increasing rates any farther could influence self-sufficiency negatively. Next in the analysis of sustainability, empirical research has indicated that there exists a relationship between increasing the number of borrowers for a given number of field officers and attaining a higher level of self-sufficiency. The work of Crombrugghe, Tenikue and Sureda (2008) test this hypothesis in their study and conclude that increasing the ratio of the number of borrowers per field officer does contribute to increased profitability. This leads to the following hypothesis: H4: MFIs with a higher ratio of the number of borrowers per loan officer tend to have higher self-sufficiency. 32 Subsequently, empirical research has also indicated that there exists a relationship between higher average loan balances per borrower and increased self-sustainability. The work of Crombrugghe, Tenikue and Sureda (2008) test this hypothesis and conclude that increasing the size of the average loan offered to borrowers is a benefit; however, to attain maximum sustainability MFIs should lend neither very small nor very large amounts. The work of Brau and Woller (2004) conclude that those MFIs proving self-sufficient do so by extending larger loans to marginally poor clients. Navajase? al (2000) suggested that most MFIs demonstrating self-sufficiency have tended to loan to borrowers who were either slightly above or slightly below the poverty line. This leads to the following hypothesis: Hs: MFIs with larger average loan sizes provided to borrowers will attain higher levels of self-sufficiency. The last hypothesis in the analysis of sustainability indicates that there exists a relationship between lower average costs and increased self-sustainability. The work of Paxton et al (2002) observes not only the effect of costs on sustainability but of costs on outreach as well. Their results suggest that due to high transaction costs associated with providing smaller loans there exists a trade­ off between serving the poorest segments and attaining financial viability. The work of Ylinen (2010) indicates that in order to increase sustainability MFIs need to work towards reducing operational expenses. As mentioned above, the work of Brau and Woller (2004) conclude that MFIs that are able to attain costefficiencies through economies of scale will prove self-sustainable. This leads to the following hypothesis: 33 H6: MFIs with lower costs associated with providing loans to borrowers will attain higher levels of self-sufficiency. 3.3 Mission Drift Existing literature has indicated the existence of a relationship between the size of the average loan balance provided to borrowers and higher levels of profitability (Freixas and Rochet, 2008). The work of Mersland and Strom (2010) test the hypothesis and confirm that average loan size increases with profit per client; therefore, MFIs are able to earn larger absolute profits with large loan sizes. The work of Ylinen (2010) further confirms this hypothesis by concluding that MFIs with higher levels of profitability tend to have higher average loan balances, indicating that MFIs face a trade-off and are therefore susceptible to mission drift. This leads to the following hypothesis: H7: MFIs with higher average loan balances tend to provide more loans to well-off borrowers as compared with ultra-poor borrowers and hence are susceptible to mission drift. Subsequently in the analysis of mission drift, Cull et al (2007) find a positive coefficient for institution size and average loan size, indicating that large individual based lenders perform relatively poor in terms of outreach. Ylinen (2010) suggests that older and more mature MFIs trying to increase access to commercial funding may become susceptible to disbursing larger average loans than younger MFIs. Similarly, Ylinen (2010) suggests that given the size of an institution is representative of its age, larger MFIs are more susceptible to mission drift. This leads to the following hypothesis: H8. Mission drift is suspect to occur as institutions mature; therefore, larger and older institutions are suspect to demonstrate higher average loan sizes. 34 Lastly in the analysis of mission drift, existing literature has suggested that average loan size increases with average cost. Mersland and Strom (2010) confirm this hypothesis and suggest that the positive relationship indicates that inefficient MFIs find it necessary to shift their loan portfolios toward larger average loans; therefore, inefficient MFIs are more susceptible to mission drift. This leads to the following hypothesis: H9: 1If the average loan balance per borrower is influenced by the average cost per borrower more so than operational self-sufficiency, cost efficient MFIs should not fall suspect to mission drift. 3.4 Mission Drift in Asia vs. Latin America and the Caribbean Given that South Asia continues to have one of the largest markets of poor individuals in the world, this region will continue to have the ability to attract massive numbers of poor unbanked individuals into the microfmance industry every year. As a result,for MFIs in Asia mission drift is hard to prove industry-wide. On the contrary, given the lower percentage and sparse location of poor individuals in Latin America and the Caribbean mission drift is more likely occur. Due to an underdeveloped financial system MFIs in Latin America have become attractive sources of funding for unbanked wealthier clients in the regions. As a result, MFIs in this region are susceptible to crosssubsidization lending and show vulnerability of mission drift (Armendariz and Szafarz, 2009; Kar, 2010). This leads to the following hypothesis: Hio: The results for mission drift in Latin America and the Caribbean will be more consistent than what is found in Asia. 35 Chapter IV Database and Methodology 4.1 Database The data for the empirical investigation is based on the Microfinance Information eXchange (i.e, MIX Market) a not-for-profit private organization. MIX Market is one of the largest sources of financial and social performance data on MFIs globally. This empirical research analyzes financial outreach data, collected from performance and the MIX Market database, for 419 microfinance institutions between the periods of 2005-2009. The institutions were selected based on the completeness of reporting for the required time-period of study and are therefore not representative of all microfinance institutions operating within the two regions. These institutions do, however, collectively account for 57 percent of the MFIs in the two regions combined. The data are used to form two independent panels, 1) Latin America and the Caribbean, and 2) Asia.18Due to the variability in the completeness of data the sample size between the two regions differ. The sample for Latin America and Caribbean consists of 223MFI's over 5 time periods for a total of 1,115 observations. Table 4.1.1 shows the variance in the number of microfinance institutions within the sample for Latin America and the Caribbean. The cumulative percentage for the regions of Ecuador and Peru equal 35 percent of the MFIs in the sample. It is necessary to keep in mind this variance when analysing the results. 18 Asia consists of South Asia and East Asia and the Pacific 36 The panel for Asia consists of 196 MFIs over 5 time periods for a total of 980 observations. Table 4.1.2 shows the variance in the number of microfinance institutions within the sample for Asia. Regions where microfinance is more developed; including Bangladesh, the Philippines, and India represent a substantial proportion of the sample. These three areas combined account for 56 percent of the MFIs within the sample. It is important to make note that these institutions will be over represented in the analysis compared to the others. 37 Table 4.1.1: Distribution of Microfinance Institutions in Latin America and the Caribbean Region Number of MFIs Argentina Bolivia Brazil Chile Colombia Costa Rica Dominican Republic Ecuador El Salvador Guatemala Haiti Honduras Mexico Nicaragua Panama Paraguay Peru Venezuela 5 17 6 3 13 8 3 34 11 12 6 12 21 19 3 5 44 1 Source: MIX Market Database Table 4.1.2: Distribution of Microfinance Institutions in Asia Region Number of MFIs 13 23 13 4 1 40 13 15 11 1 47 1 7 1 6 Afghanistan Bangladesh Cambodia China East Timor India Indonesia Nepal Pakistan Papua New Guinea Philippines Samoa Sri Lanka Thailand Vietnam Source: MIX Market Database 38 4.2 Regression Approach This study adopts a panel estimation framework. The equations mentioned in chapter three are estimated by pooled ordinary least squares (OLS), fixed effects, and random effects. The initial estimation procedure employed is pooled OLS. This form of estimation is the simplest estimator for panel data and is in most cases unlikely to be sufficient. Despite the expected inadequacy estimating pooled OLS provides a baseline for comparisons with more complex models (Cottrell, 2005). Pooled OLS assumes a constant intercept and slope regardless of the cross-section type (Park, 2009). The pooled OLS specification is written in the following form: Y« = Xifi+uit Where _yi(is the observation of the dependent variable for MFI i in period t, Xitis a 1 x k vector of the independent variables observed for MFI i in period t, is a A: x 1 vector of parameters, and w„is an error term specific to MFI i in period t (Cottrell, 2005). The error term in pooled OLS must be orthogonal with the independent variables and be independent from one another. As a result of the same cross-section being observed over time this typically does not hold in panel data causing error terms of specific units to be correlated with each other (Cottrell, 2005). The next two models employed are the random and fixed effects models. The random effects model can be written as uit =o, + elh yielding Yn = Xitfi + Vi + €n 39 That is, Vi, is treated as a random drawing from a given probability distribution (Cottrell, 2005). It therefore imposes a dummy variable inclusive of an intercept that represents a common mean value for all groups, plus a group specific error-term that will capture group deviations from that common mean. The random effects model assumes that they entity error term is not correlated with the independent variables. The standard entity-fixed effects model is a special case of the random effects model and is used to control for unobservable heterogeneity when this heterogeneity is constant over time. The model can be written as uit = at + eih yielding Yu = Xj$ + at + e„ That is, Uit,is decomposed into a unit-specific and time-invariant component, a,-, and an observation-specific error, 6jt The a, are then treated as fixed parameters (in effects, unit-specific ^-intercepts), which are to be estimated. This is done by subtracting the group mean from each of the variables and estimating the model without a constant (Cottrell, 2005). The fixed effects model assumes that the entity error terms are correlated with the independent variables. This thesis goes further to estimate the time-fixed effects model if this model yields a higher level of significance of the coefficients. This model differs from the standard entity-fixed effects model by assuming that the average value of 7/,changes over time but not cross-sectionally. In this model a, is a time-varying intercept that captures all of the variables that affect 7„and that vary over time but are constant cross-sectionally (Brooks, 2008). The two models, entity-fixed 40 effects and time-fixed effects can be combined and estimated controlling for unobservable heterogeneity when this heterogeneity is constant over time as well as controlling for unobservable heterogeneity when this heterogeneity is constant cross-sectionally. In order to test the model of best-fit a series of tests are conducted. Firstly, to validate the preference for using the fixed effects model over pooled OLS the redundant fixed effects test is employed to test the null hypothesis that the cross-sectional units have a common intercept. A significant pvalue for both regions (Latin America and the Caribbean and Asia) and all estimations (sustainability, mission drift, and economies of scale) counted against the null hypothesis and therefore it was rejected. This confirms that there is a validation to the preference of using the fixed effect approach over pooled OLS. Once it is determined that panel methodology is the structure of best-fit, the Hausman test is employed in order to validate the preference of using the random effect approach against the fixed effect approach. The Hausman test probes whether the entity error terms are correlated with the independent variables. The null hypothesis is that they are not. A significant p-value of less than 0.05counted against the null hypothesis that the random effect approach is consistent and efficient, in favour of the fixed effect approach (Cottrell, 2005). Choosing between the entity effects, time effects, or a combination of the two is also based on the redundant fixed effects test and Hausman test. If there was a situation in which more than one model could be the model of best fit the level of significance, the R2, and adjusted R2 are employed to make the final decision. 41 After the appropriate tests were conducted it was deemed the model of best-fit to be the entity-fixed effects model. Given that panel data typically suffers from heteroskedasticity, meaning the error terms' variance is not constant over observations, the White period method for calculating robust standard errors is used. This method assumes that the errors of a cross-section are heteroskedastic and serially correlated. 4.3 Hypothesis Development The hypotheses stated in chapter III were used to formulate the regression questions estimated in this thesis. Hypotheses one and two are used to form the following equation for economies of scale: Log(Cost per Borroweruj = o;•+ (3iLog(A verage Loan Balance per Borrowerit) + j32Log(Total Assetsit) + elt 0;&<0 Hypotheses three, four, five, and six are used to form the following equation for operational self-sufficiency: OSSu = oti+ (31(Yielda) + fafYield2it) + faLog(Borrowers per Loan Officerit) +j34Log(A verage Loan Balance per Borrowerit) + ftLog(Cost per Borrower,J+ e„ /3/> 0 ; 02< 0 ; fa > 0 ; fa >0 / 0s 0 ; f a > 0 ; f a > 0 42 Chapter V Empirical Results This chapter presents the results of the empirical exercise relating to economies of scale, determinants of operational self-sustainability and mission drift of MFIs in Latin America and the Caribbean and Asia. Along with the regional differences (Asia vs. Latin America), the study also presents empirical results of the above mentioned issues with regard to size - small, medium and large MFIs separately. The empirical results presented in this chapter are that of a fixed effect model; the preferred methodological framework. Similar estimations were conducted based on ordinary least squares and random effect model. The results of the estimations were tested based on the Hausman test. The Hausman test confirmed the fixed effect model to be the most robust model; therefore, for the sake of brevity and space conservation solely the fixed effects model results are presented. This chapter is divided into five sections. Section 5.1 provides descriptive statistics of the data; section 5.2 provides the empirical results of economies of scale , section 5.3 provides the empirical results of operational selfsufficiency; section 5.3; section 5.4 provides the empirical results of mission drift and section 5.5 summarises the conclusions of this chapter. 5.1 Descriptive Statistics During the time period 2005 to 2009, the average assets of MFIs in Latin America and the Caribbean amounted to $US67 million, as compared with $US62 million in Asia (Table 5.1.1). However, the standard deviation of assets for MFIs in Asia, at SUS375 million, is relatively higher than the $US237 million in Latin 43 America and the Caribbean. This disparity indicates a larger variation in the size of MFIs in Asia as compared with Latin America and the Caribbean. The average loan balance per borrower of MFIs in Latin America and the Caribbean is $US1,336, nearly five times that of the $US275 average loan balance per borrower in Asia. The higher average loan balance per borrower is an indicator of the possibility of mission drift for MFIs in Latin America and the Caribbean. The average number of borrowers per loan officer, defined as the number of active borrowers/the number of active loan officers, is considerably higher in Asia than in Latin America and the Caribbean at 360 and 291, respectively. The average cost per borrower, defined as the operating expense/the number of active borrowers, at $US58 for MFIs in Asia, is substantially lower than the $US292 in Latin America and the Caribbean. There is, however, relatively higher variation in the average cost per borrower in Latin America and the Caribbean compared with Asia at $US262 and $US120, respectively. As mentioned in chapter two, the densely populated low income areas in Asia allow for loan officers to provide and monitor loans at relatively lower costs than loan officers in Latin America and the Caribbean. The dense population in Asia also allows for loan officers in this region to service a larger number of borrowers than loan officers in Latin America and the Caribbean. The indicators of operational self-sufficiency (OSS) and yield on portfolio reveal interesting contrasts in both Asia and Latin America and the Caribbean. Operational self-sufficiency is a percentage (%), which indicates whether or not enough revenue has been earned to cover the microfinance institution's (MFI's) 44 total costs. The OSS of MFIs in Latin America and the Caribbean, at 1.16, is relatively higher than the 1.10 in MFIs Asia. This indicates thatMFIs in both Asia and Latin America and the Caribbean are able to more than cover costs and earn sizeable profits through their operating revenues; however, MFIs in Latin America and the Caribbean are slightly ahead of MFIs in Asia. The portfolio yield (income from lending as a proportion of the average outstanding portfolio) is relatively higher in Latin America and the Caribbean as compared with Asia at 0.36 and 0.30, respectively. This implies that average interest rate on micro-credit is relatively higher in Latin America and Caribbean. 45 Table 5.1.1: Descriptive Statistics of MFIs in Sample: Asia and Latin America and the Caribbean: 2005-2009 Asia No Variables Mean Median Std. Dev. Skewness Obs Max Min 1. 2. 3. 4. 5. 6. Assets (millions) Avg. Loan Balance/Borrower Borrowers/Loan Officer Cost/Borrower OSS Yield 62 7 6,450 0.00 375 12.17 969 275 149 6,022 5 408 7.32 953 361 58 1.10 0.30 255 30 1.12 0.27 10,775 1,939 4.40 0.88 9 0.34 0.01 0.00 548 120 0.35 0.14 10.94 10.44 1.03 1.28 777 884 945 816 67 Median 9 Max 3,480 Min 0.07 Std. Dev. 237 Skewness 8.92 Obs 1112 1,336 745 64,087 73 3,180 13.47 1106 291 258 1,651 13 174. 2.43 1020 200 1.16 0.36 152 1.13 0.31 4,836 7.20 1.27 11 0.00 0.04 262 0.35 0.18 11.02 6.76 1.48 1043 1073 1004 Latin America and the Caribbean No Variables Mean 1. 2. 3. 4. 5. 6. Assets (millions) Avg. Loan Balance/Borrower Borrowers/Loan Officer Cost/Borrower OSS Yield Source: Data collected from the MIX Market Database 46 In order to test for the presence of multicoUinearity correlation coefficients amongst the explanatory variables are calculated. Kennedy (2008) indicates that correlations need to be in the range of 0.8-0.9 in order to detect the presence of collinearity amongst any two variables. As is evident fromTable 5.1.2, none of the correlation coefficients are in this range; therefore, multicoUinearity is not a factor in this estimation. 47 Table 5.1.2: Correlation Coefficients among the Explanatory Variables Asia Total Assets Avg. Loan Balance/Borrower Borrowers/Loan Officer Cost/Borrower OSS Yield Total Assets Avg. Loan Balance/Borrower Borrowers/Loan Officer Cost/Borrower OSS Yield 1.000 .113** .222** .134** -.185** -.165** -.113** 1.000 -.186** .660** .126** -.056 .222** -.186** 1.000 -.504** .227** _ 219** -.134** -.185** -.165** .660** .126** -.056 -.504** .227** . 219** 1.000 _190** .438** _ 190** 1.000 .052 .438** .052 1.000 Total Assets Avg. Loan Balance/Borrower Borrowers/Loan Officer Cost/Borrower OSS Yield 1.000 .595** .102** 397** 194** -.321** .595** 1.000 _144** .698** 142** -.680** .102** _144** 1.000 -.499** -.165** -.176** .397** 194** -32i** .698** .142** -.680** - 499** 1.00 -.165** - 18i** -.165** 1.000 0.13 -.181** 0.13 1.000 Latin America and the Caribbean Total Assets Avg. Loan Balance/Borrower Borrowers/Loan Officer Cost/Borrower OSS Yield .200** -.176** ** Two-sided Pearson correlation coefficient is significant at the 0.01 level 48 5.2 Economies of Scale Table 5.2(a) provides the results of the analysis of economies of scale for MFIs in Asia. The positive and significant coefficient on the average loan balance per borrower confirms, for the full sample of MFIs as well as all sub-sample groups, that average costs per borrower tend to increase with the average loan balance per borrower. These results are consistent with the results of Crombrugghe, Tenikue and Sureda (2008).The analysis also investigates the effect of total assets (outreach) on the cost per borrower. The negative and significant coefficient of total assets proves and accepts the hypothesis that costs per borrower tend to decrease with total assets. With the exception of MFIs classified as medium inscale, the results are consistent across all sub-sample groups proving that for these MFIs economies of scale does exist. For the MFIs classified as having a medium scale the positive and significant coefficient on total assets indicates that economies of scale is not present. 49 Table 5.2(a): Analysis of Economies of Scale for MFIs in Asia: 2005 to 2009 - Fixed Effect Model Log (Cost/Borrowerji) = 04 + j3iLog(Average Loan Balance/Borrower;,) + /?2Log(Total Assets^) + fit Variables Size# Exp. Sign Full Large Medium Small 1.69 0.50 -1.57 5.33 (0.66)** (0.51) (1.24)* (1.56)*** 0.65 0.74 0.52 0.43 (0.14)*** (0.07)*** (0.10)*** (0.44)* -0.10 (0.06)* -0.07 (0.03)** 0.15 (0.08)* -0.27 (0.11)** 196 883 0.93 0.91 91 331 0.97 0.9 125 323 0.97 0.94 89 229 0.92 0.87 Constant Log(Average Loan Balance/Borrower) Log (Total Assets) + - Cross-sections Observations R-squared Adjusted R-squared Note 1: Figures in brackets are standard errors Note 2: ***, **, * indicates statistical significance at 1%, 5%, and 10% respectively. # The classification of MFIs according to size is based on gross loan portfolio: small (< $4 million), medium ($4-15 million) and large MFI (>$15 million). Table 5.2(b) provides the results on the analysis of economies of scale for MFIs in Latin America and the Caribbean. In regards to the analysis on the average loan balance per borrower, the results are consistent with those found in Asia. The positive and significant coefficient indicates that for the full sample of MFIs as well as all sub-sample groupsthat cost per borrower tends to increase with the average loan balance per borrower. 50 The analysis on the effect of total assets on the cost per borrower for MFIs in Latin America and the Caribbean is also similar to what was found in Asia. For the full sample of MFIs the negative and significant coefficient on total assets indicates that costs per borrower tend to decrease with total assets, proving that for these MFIs economies of scale does exist. With the exception of MFIs classified as large in scale the results are consistent across all sub-sample groups. For the MFIs classified as having a large scalethe coefficient on total assets is positive, however is insignificant. 51 Table 5.2(b): Analysis of Economies of Scale for MFIs in Latin America and the Caribbean: 2005 to 2009 - Fixed Effect Model Log (Cost/Borrowerit) = QS + /3i Log(Average Loan Balance/Borrowerjt) + ^LogCTotal Assetsjt) + fit Variables Sign Full Large Medium Small 2.38 -0.48 3.57 1.96 (0.82)*** (0.55) (0.94)*** (1.23)* 0.65 0.65 0.62 0.67 (0.05)*** (0.08)*** (0.10)*** (0.12)*** -0.10 (0.05)** 0.06 (0.05) -0.10 (0.06)** -0.09 (0.08)* 223 1043 0.94 0.92 96 400 0.96 0.95 100 285 0.94 0.90 109 358 0.89 0.84 Constant Log(Average Loan Balance/Borrower) Log (Total Assets) + - Cross-sections Observations R-squared Adjusted R-squared Note 1: Figures in brackets are standard errors Note 2: ***, **, * indicates statistical significance at 1%, 5%, and 10% respectively. # The classification of MFIs according to size is based on gross loan portfolio: small (< $4 million), medium ($4-15 million) and large MFI (> $15 million). 52 5.3 Operational Self-Sufficiency Given that for both regions (Asia and Latin America and the Caribbean) there is evidence to support economies of scale by increasing the scale of the institution, we proceed to investigate the determinants of operating selfsufficiency (OSS). As mentioned earlier, OSS is a percentage (%), which indicates whether or not sufficient revenue has been earned to cover the microfinance institution's (MFI's) total costs - operational expenses, loan loss provisions and financial costs.Table 5.3(a) provides the results on the analysis of operational self-sufficiency for MFIs in Asia. The results show that for the full sample of MFIs, yield on portfolio (interest rates) is positively associated with increased operational self-sufficiency. With the exception of MFIs classified as large scale institutions, all sub-sample group estimations produce a positive and statistically significant coefficient on the gross portfolio yield. The coefficient on the squared value of the gross portfolio yield is negative and significant for the full sample of MFIs as well as for MFIs medium in size; indicating that for these MFIs there exists a threshold yield (interest rate) that once reached charging interest rates any higher could influence operational self-sufficiency negatively.These results are consistent with that of Cull, Demirgiic-Kunt and Murduch (2007). MFIs that report charging interest rates beyond a threshold value run the risk of attracting high risk borrowers that do not expect to repay and therefore could potentially attain lower levels of operational self-sufficiency. For large and small scale MFIs the coefficient is positive and insignificant. The empirical results show that the average loan size per borrower is also associated with improved operational self-sufficiency for the full sample of MFIs 53 in Asia as well as all sub-sample groups estimated. A positive and significant coefficient indicates that increasing this ratio will lead to increased operational self-sufficiency. The effect of the average cost per borrower on operational self-sufficiency show that for the full sample in Asia, lower costs are associated with increased operational self-sufficiency. A negative and significant coefficient on the cost per borrower indicates that MFIs that are able to reduce costs will experience an increase in operational self-sufficiency. The results are consistent for all subsample groups estimated; indicating that regardless of size operational selfsufficiency tends to increase as costs per borrower decrease. 54 Table 5.3(a) : Analysis of OSS for MFIs in Asia: 2005 to 2009 - Fixed Effects Model Operational Self-Sufficiencyit= 05 + /3i (Yields) + /32(Yield2it) + /53Log(Borrowers/Loan Officerit) + j84Log(Avg. Loan Balance/Borrowerit) + |S5Log(Cost /Borrowerjt) + 6jt Variables Size# Exp. Sign Full Large Medium Small 0.56 -0.46 0.45 -0.46 (0.27)* (0.44) (0.51) (0.26)* Constant Yield + 1.76 (0.86)* 0.42 (1.81) 2.83 (1.68)* 1.17 (0.60)** Yield2 _ -0.83 (1.03)* 1.47 (2.42) -2.80 (2.14)* 0.01 (0.64) Log(Borrower/Loan Officer) 4" 0.02 0.02 0.05 0.06 (0.02) (0.02) (0.06) (0.05) 0.16 0.22 0.16 0.40 (0.08)** (0.08)*** (0.08)** (0.06)*** -0.23 -0.24 -0.28 -0.29 (0.06)*** (0.12)** (0.07)*** (0.06)*** 192 750 0.77 0.69 96 309 0.68 0.53 112 270 0.84 0.72 68 171 0.60 0.59 Log(Average Loan Ralanre/Rnrrnwerl Log (Cost/Borrower) + Cross-sections Observations R-squared Adjusted R-squared Note 1: Figures in brackets are standard errors Note 2: ***, **, * indicates statistical significance at 1%, 5%, and 10% respectively. # The classification of MFIs according to size is based on gross loan portfolio: small (< $4 million), medium ($4-15 million) and large MFI (> $15 million). 55 Table 5.3(b) provides the results on the analysis of operational selfsufficiency for MFIs in Latin America and the Caribbean. The results show that for the full sample of MFIs in Latin America and the Caribbean, as well as all sub-sample groups estimated, raising portfolio yield (interest rates) is associated with increased operational self-sufficiency. The coefficient on the squared value of the gross portfolio yield is negative and significant for all samplesof MFIs. This result is consistent with the full sample and medium sized MFIs in Asia in that that there exists a threshold interest rate that once reached charging rates above and beyond could influence operational self-sufficiency negatively. Consistent with Asia, for the full sample of MFIs in Latin America and the Caribbean, an increased average loan balance per borrower improves operational self-sufficiency. A positive and significant coefficient indicates that increasing this ratio will lead to increased operational self-sufficiency. All sub-sample groups estimated yield a positive and significant coefficient on the size of the average loan balance per borrower. For the full sample of MFIs in Latin America and the Caribbean cost per borrower proves to have a negative effect on the level of operational selfsufficiency. A negative and significant coefficient indicates that MFIs that are able to decrease this ratio will attain higher levels of operational selfsufficiency.With the exception of MFIs small in scale, all sub-sample groups estimated produce a negative and significant coefficient on the cost per borrower. 56 Table 5.3(b) : Analysis of OSS for MFIs in Latin America and the Caribbean: 2005 to 2009 - Fixed Effects Model Operational Self-Sufficiencyit= 04 + /3i (Yield;,) + /32(Yield2jt) + j33Log(Borrowers/Loan Officer;,) + /34Log(Avg. Loan Balance/Borrower;,) + /35Log(Cost /Borroweri,) + elt Variables Size# Exp. Sign Full Large Medium Small 0.62 0.59 0.10 -1.39 (0.43) (0.38) (0.60) (1.45) 1.94 (0.36)*** 1.63 (0.47)*** 1.57 (0.45)*** 3.16 (1.68)* -0.99 (0.35)*** -0.99 (0.27)*** -0.66 (0.29)** -2.32 (1.81)* 0.01 0.04 0.03 0.08 (0.02) (0.02) (0.05) (0.05) 0.20 0.14 0.33 0.29 (0.07)*** (0.07)** (0.07)*** (0.13)** -0.26 -0.18 -0.36 -0.09 (0.15)* (0.07)** (0.08)*** (0.33) 222 995 0.71 0.62 95 388 0.84 0.79 97 275 0.84 0.75 108 332 0.69 0.53 Constant Yield + Yield2 - Log(Borrower/Loan Officer) Log(Average Loan Balance/Borrower) Log (Cost/Borrower) + + - Cross-sections Observations R-squared Adjusted R-squared Note 1: Figures in brackets are standard errors Note 2: ***, **, * indicates statistical significance at 1%, 5%, and 10% respectively. # The classification of MFIs according to size is based on gross loan portfolio: small (< $4 million), medium ($4-15 million) and large MFI (> $15 million). 57 5.4 Mission Drift Table 5.4(a) provides the results on the analysis of mission drift for MFIs in Asia. The results indicate, for the full sample of MFIs, that the average loan balance per borrower tends to increase with operational self-sufficiency. The results support the hypothesis of the positive relationship between average loan balance and profitability. These results are consistent across MFIsof various sizes - medium and large.The positive and significant coefficient on operational selfsufficiency indicates that for these sample groups the size of the average loan balance per borrower tends to increase with operational self-sufficiency. This result indicates that these MFIs could potentially face a trade-off between serving the poorest borrowers and pursuing operational self-sufficiency.For MFIs small in size the negative coefficient on operational self-sufficiency indicates that for some small scaled MFIsa trade-off may not exist; however, given that this coefficient is insignificant there is insufficient evidence to support that a trade-off does not exists. Total assets is also positively related to the size of the average loan indicating that increases in outreach(total assets) tends to have a positive impact on the size of the average loan balance. The coefficient on total assets is positive and statistically significant for all size groups. This result is consistent with Ylinen (2010) indicating that older and more mature MFIs trying to increase access to commercial funding may become susceptible to disbursing larger average loans. The empirical results of the effect of cost per borrower on the average loan balance per borrower shows that for the full sample of MFIs average loan size 58 increases with the cost per borrower. This result is consistent with that ofCull, Demirgiic-Kunt and Murduch (2007) inthat inefficient MFIs need to shift their loan portfolio towards larger average loans (seasoned borrowers). The results are consistent across MFI size groups; however, the coefficient for MFIs small in size is insignificant. 59 Table 5.4(a): Analysis of Mission Drift for MFIs in Asia: 2005 to 2009 Fixed Effects Model Log(Average Loan Balance per Borrowerjt) = 05+ /3iLog(Operational Self-Sufficiency,,) + /32Log(Assets„) + /33Log(Cost per Borrowerit) + eit Size# Variables Sign Full Large Medium Small 1.80 1.58 -0.18 -0.53 (0.41)*** (0.82)** (1.93) (0.82) Constant Log (OSS) + 0.53 (0.12)*** 0.51 (0.21)** 0.08 (0.04)* -0.33 (0.30) Log (Total Assets) + 0.08 (0.02)*** 0.09 (0.05)* 0.27 (0.06)** 0.38 (0.10)*** Log (Cost/Borrower) + 0.62 (0.02)*** 0.65 (0.06)*** 0.35 (0.12)*** 0.05 (0.25) 195 882 0.59 0.58 97 331 0.66 0.66 124 322 0.97 0.95 89 229 0.95 0.91 Cross-sections Observations R-squared Adjusted R-squared Note 1: Figures in brackets are standard errors Note 2: ***, **, * indicates statistical significance at 1%, 5%, and 10% respectively. # The classification of MFIs according to size is based on gross loan portfolio: small (< $4 million), medium ($4-15 million) and large MFI (> $15 million). 60 Table 5.4(b) provides the results on the analysis of mission drift for MFIs in Latin America and the Caribbean. The positive and significant coefficient for the full sample of MFIs indicates that the size of the average loan balance per borrower increases with operational self-sufficiency. Consistent with Asia, the results confirm the hypothesis of the relationship between average loan balance and average profit by Freixas and Rochet (2008) and are in-line with the work of Cull, Demirguc-Kunt and Murduch (2007).For the full sample of MFIs, as well as MFIs of medium or small scale, the results indicate that there exists a trade-off between serving the poorest borrowers and higher levels of operational selfsufficiency. For MFIs large in scale the coefficient is positive; however, is insignificant. The result for the effect of total assets (outreach) on the average loan balance is positive and similar with that in Asia. A positive and significant coefficient on total assets across all samples estimated indicates that the average loan balance per borrower tends to increase with total assets. The result for the effect of cost per borrower on the average loan balance is also similar to that in Asia. A positive and significant coefficient confirms the hypothesis byCull, Demirgiic-Kunt and Murduch (2007) that inefficient MFIs in need to shift their loan portfolio towards larger average loans. The result is consistent across all size groups. 61 Table 5.4(b): Analysis of Mission Drift for MFIs in Latin America and the Caribbean: 2005 to 2009 - Fixed Effects Model Log(Average Loan Balance per Borrowerjt) = a$+ |8iLog(Operational Self-Sufficiencyit) + /32Log(Assetslt) + j33Log(Cost per Borrowerjt) + €jt Size# Variables Sign Full Large Medium Small 0.24 1.22 1.37 0.60 (0.72) (1.59) (1.28) (1.05) Constant Log (OSS) + 0.20 (0.10)** 0.02 (0.17) 0.27 (0.15)* 0.15 (0.12)*** Log (Total Assets) + 0.23 (0.04)*** 0.14 (0.08)* 0.19 (0.07)** 0.25 (0.07)*** Log (Cost/Borrower) + 0.53 (0.05)*** 0.66 (0.10)*** 0.41 (0.15)*** 0.38 (0.06)*** 223 1043 0.98 0.97 96 400 0.98 0.97 100 285 0.98 0.97 109 385 0.97 0.95 Cross-sections Observations R-squared Adjusted R-squared Note 1: Figures in brackets are standard errors Note 2: **•,**, * indicates statistical significance at 1%, 5%, and 10% respectively. # The classification of MFIs according to size is based on gross loan portfolio: small (< $4 million), medium ($4-15 million) and large MFI (> S15 million). 62 5.5 Conclusion This chapter presented results on some of the basic questions posed in this thesis. Do MFIs exhibit economies of scale in their operations across Asia and Latin America and the Caribbean? What are the main determinants of financial performance (OSS) of MFIs across the Asia and Latin America and the Caribbean? Is there any evidence of mission drift among MFIs across Asia and Latin America and the Caribbean? In regards the first question, viz., economies of scale, MFIs in Asia and Latin America showed evidence of reaping economies of scale across various size groups (small, medium and large). The empirical investigation into the determinants of financial performance (OSS) shows that portfolio yield (interest rates) and average loan balance per borrower have a positive and statistically significant impact on financial performance of MFIs in both regions (Asia and Latin America and the Caribbean). In both regions, increases in portfolio yield (interest rates) over a certain threshold seem to negatively impact financial performance (as revealed by the coefficient on yield2). This result is more robust in MFIs across Latin America and Caribbean. The cost per borrower has a negative statistically significant impact on profitability of MFIs across both regions. In regards to mission drift, the empirical evidence shows that mission drift is evident among MFIs in both Asia and Latin America and Caribbean. The results support the hypothesis that there is a positive relationship between profitability of an MFI and average loan size. These results are similar to the one obtained by Cull et al (2007). 63 Chapter VI Conclusion This thesis utilized three estimations to address the following research questions: (1) do MFIs attain economies of scale as they increase their outreach; (2) what are the determinants of self-sustainability of MFIs interest rates, repeated loans to seasoned borrowers or costs; and (3) does the pursuance of self-sustainability tend to drive microfinance institutions away from the poorest borrowers; and if so, is there evidence to support regional differences between Asia and Latin America and the Caribbean. The study uses a panel data (unbalanced) set of MFIs of 419 MFIs in Asia and Latin America and Caribbean and adopts the robust fixed effects model in the empirical investigation. The first research question is whether MFIs are able to exploit economies of scale over recent years (2005-2009). For both regions the average cost per borrower actually increased with the size of the loan, indicating that MFIs are not exploiting economies of scale by increasing the average loan size. This result is consistent with that of Crombrugghe, Tenikue and Sureda (2008) in which they found that costs per borrower increase with the average loan balance per borrower as a result of higher selection and monitoring required when issuing larger loans.The analysis of the effect of total assets on the average cost per borrower also remained fairly consistent for MFIs in both Asia and Latin America and the Caribbean. The negative and significant coefficient indicates that, overall, 64 economies of scale do occur as MFIs increase is size and outreach. However, for MFIs in Asia classified as having medium outreach and for MFIs in Latin America and the Caribbean classified as having large sizeoutreach the positive coefficient on total assets indicates that for these MFIs the results may not support the existence of economies of scale. The second estimation sought to address the research question whether microfinance institutions are in fact able to attain operational selfsufficiency through the pursuance of higher profits. Using yield as the proxy for profits the estimations for Asia and Latin America and the Caribbean yielded similar results. The positive and significant coefficient for both regions across nearly all sub-sample groups ascertains that charging higher interest rates does in fact lead to greater operational self-sufficiency. However, the results indicate that for both regions there is evidence to support a threshold interest rate that once surpassed could influence operational self-sufficiency negatively. As mentioned above, charging rates above the threshold value attracts low-quality high-risk borrowers. For both regions the results indicate that increasing the size of the average loan provided to borrowers could also increase operational self-sufficiency. Hence, for both regions the concern is introduced that as MFIs pursue selfsufficiency it potentially becomes necessary to move away from serving the poorest borrowers towards serving wealthier more mature borrowers. The results are also consistent in regards to the effect that the average cost per borrower has on operational self-sufficiency. The negative and significant coefficient for both regions across nearly all samples estimated indicates that 65 MFIs that are able to attain cost-efficiencies are more likely to attain operational self-sufficiency. The third estimation sought to directly address if MFIs are abandoning serving the poorest borrowers in pursuit of self-sufficiency (so called mission drift hypothesis). The analysis on the effect of operational self-sufficiency on the average loan balance provided to borrowers from Asia and Latin America and the Caribbean yielded slightly different results. For Latin America and the Caribbean the coefficient on operational selfsufficiency is positive and significant for all size groups indicating that all MFIs in Latin America and the Caribbean potentially run the risk of mission drift. The positive and significant coefficient on operational self-sufficiency indicates that MFIs could potentially face a trade-off between serving the poorest borrowers and pursuing operational self-sufficiency. For MFIs in Asia the results indicate that MFIs classified as small in size may not be at risk of mission drift. The negative coefficient on operational self-sufficiency indicates that for these MFIs there may not be sufficient evidence to support that a trade-off exists. The result of the effect of total assets on the size of the average loan provided to borrowers is similar for both regions. The positive and significant coefficient indicates that as MFIs increase in size and outreach they potentially run the risk of mission drift. The result of the effect of the average cost per borrower on the average loan balance per borrower is also similar for both Asia and Latin America and the Caribbean. The positive and significant coefficient on the average cost per borrower 66 indicates that MFIs with higher costs could potentially fall victim to mission drift. Overall, the results indicate that MFIs do attain higher levels of operational self-sufficiency through the pursuance of increased profitability. The results also indicate that more cost-efficient MFIs will be more likely to attain operational self-sufficiency as opposed to those MFIs that are less cost-efficient. However, contradictory to what existing literature suggests, MFIs in neither Asia nor Latin American and the Caribbean seem to be targeting wealthier clients in an attempt to exploit economies of scale. Rather,the results suggest that by increasing the overall size and outreach of the institution MFIs are able to attain cost-efficiencies. Therefore, rather than focussing on increasing individual loan size focus should be paid to increasing outreach (especially bringing ultra-poor into the ambit of the program). As mentioned above, the results indicate that all MFIs in Latin America and the Caribbean are potentially at risk of mission drift while pursuing higher profitability in an attempt to attain operational selfsufficiency. However, consistent with existing literature, the results indicate that not all MFIs in Asia will be at risk of mission drift while pursuing the same conditions. As mention above, MFIs classified as small in scale may not be at risk. The deviation in results could potentially be due to the variation in existing economic and socio-economic structure between the two regions. As noted, the gross domestic product per capita in Latin America is nearly six times 67 that of South Asia. Also, due to theunderdeveloped financial system MFIs in Latin America have become attractive sources of funding for unbanked wealthier clients in the region. As a result, MFIs in this region are susceptible to cross-subsidization lending leading to vulnerability of mission drift. MFIs in both Asia and Latin America and the Caribbean are susceptible to mission drift as the outreach of the MFI or the cost of providing loans increases. However, for all MFIsin consideration the size of the coefficient on the cost per borrower outweighs that of both operational self-sufficiency and total assets indicating that MFIs that are more cost efficient could potentially avoid mission drift. This result is consistent with the work of Mersland and Strom (2010) which implies that when an MFI is run more efficiently, the MFIs able to reduce average loan size and potentially prevent mission drift. Overall microfinance is following a troubling trend. From a policy perspective, in order to outright prevent mission drift increased regulation is required. However, given that typically the increased costs associated with introducing new regulation gets passed on to the borrowers this may not be the best solution. 68 Bibliography ADA, (2009). Does Foreign Ownership in Microfinance Interfere with Local Development. Discussion Paper No. 1. Alkire, S., Santos, M., (2010). Acute Multidimensional Poverty: A New Index for Developing Countries. Human Development Research Paper 2010/11. Armendariz, B., Szafarz, A., (2009). Microfinance Mission Drift? CEB Working Paper No. 09/015. Asian Development Bank. (2010). Asian Development Outlook 2010 Update.The Future Growth of Asia.Retrieved March 1, 2011. From http://www.adb.org/documents/books/ado/2010/update/ado2010update.pdf Augsburg, B., Fouillet, C., (2010). Profit Empowerment: the Microfinance Institution's Mission Drift.Perspectives of Global Technology and Development. Special Issue, Microfinance and Society, PGDT 9 323-351. Beck, T., Demirgiic-Kunt, A., (2010). Financial Institutions and Markets Across Countries and Over Time. World Bank Economic Review 24(1), 77-92 Berger, M., Goldmark, L., Miller SanaBria, T., (2006)Jn Inside View of Latin American Microfinance. Inter-American Development Bank. Brar, A., Sinha, F., (2005). India, Scaling-up Access to Finance for India's Poor. Report No. 30740-IN. World Bank, Washington DC. Belaisch, A., Collyns, C., De Masi, P., Krieger, R., Meredith, G., Rennhack, R., Singh, A., (2005). Stabilization and Reform in Latin America: A Macroeconomic Perspective on the Experience since the Early 1990's. International Monetary Fund.Occasional Paper 238. Bossone, B., Lee, J., (2002). In Finance, Size Matters. International Monetary Fund, Working Paper 02/113. 69 Carrick, A., Santos, F., (2007). Social vs. commercial Enterprise: The Comparators Debate and the Battle for the Soul of Microfinance. Social Innovation Center. Chala, A., Dalai, A., Goland, T., Gonazalez, M., Morduch, J., Schiff, R., (2009).Half the World is Unbanked. Financial Access Initiatives Framing Note.Retrieved March 1, 2011.From http://financialaccess.org/sites/default/files/110109HalfUnbanked_0. pdf. Christen, R., (2001). Commercialization and Mission Drift.Occasional Paper, CGAP, Washington DC. Christen, R., Drake, D., (2002). Commercialization.The New Reality of Microfinance. In D. Drake and E. Rhyne (Eds.), The Commercialization of Microfinance. Balancing Business and Development, Chapter 1, pp. 2-22. Bloomfield: Kumarian Press. Christen, R., Rosenberg, R., Jayadeva, V., (2004). Financial Institutions with a "Double Bottom Line": Implications for the Future of Microfinance. Occasional Paper, CGAP, Washington DC. Cottrell, A., (2005). GNU Regression, Econometrics, and Time Series Library.Department of Economics.Wake Forest University. Crombrugghe, A., Tenikue, M., Sureda, J., (2008).Performance Analysis for a Sample of Microfinance Institutions in India .Annals of Public and Cooperative Economics 79(2), 269-299. Cuardo, L., Egea, C., Herrero, A., Gallego, S., Santillan, J., (2002).Latin American Financial Development in Perspective.Banco de Espana Servicio de EstudiosDocumento de Trabajo No. 0216. Cull, R., Demigiic-Kunt, A., Morduch, J., (2007). Financial Performance and Outreach: A Global Analysis of Leading Microbank. Economic Journal 117(517), 107-133. Cull, R., Demigiic-Kunt, A., Morduch, J., (2008). Microfinance Meets the Market. Journal of Economic Perspectives 23(1), 167-192. Dichter, T., Harper, M., (2007). What's Wrong with Microfinance. In T. Dichter and M. Harper (Eds.), What's Wrong with Microfinance. Essex, England: Practical Action Publishing. 70 Dieckmann, R., (2007). Microfinance: An Emerging Investment Opportunity - Uniting Social Investment and Financial Returns.International Topics.Deutsche Bank Research.Frankfiirt. December 19. Fernando, N., (2004). Micro Success Story, Transformation on Nongovernmental Organizations into Regulated Financial Institutions. Asian Development Bank, Manila. Freixas, X., Rochet, J., (2008) Microfinance of Banking (2nded.). Cambridge, Mass.: MIT Press. Green, W., (2010).Econometric Analysis, 10th Edition, Prentice Hall. Gonzalez-Vega, C., Screiner, M., Meyer, R., Rodriquez, J., Navajas, S., (1997).Bancosol: The Challenge of Growth for Microfinance Organizations.Retrieved November 18, 2011.From http://ideas.repec.Org/p/ags/ohsesp/28333.html. Hishigsuren, G., (2007). Evaluation Mission Drift in Microfinance: Lessons for Programs with Social Mission. Evaluation Review 31(3), 203260. International Monetary Fund. (2010). World Economic Outlook: Recovery, Risk and Rebalancing. Retrieved March 1/ 2011. From http://www.imf.org/external/pubs/ft/weo/2010/02/ Jain, A., Norton, C., (2009). Microfinance: Where do we Stand Today? Microbanking Bulletin, Issue 18. Microfinance Information eXchange. Kar, A. Sustainability and Mission Drift in Microfinance. PhD diss., Hanken School of Economics, 2010. Khandker, S., (1998).Fighting Poverty with Microcredit: Experience in Bangladesh. Oxford University Press, Inc. New York. Littlefield, E., Morduch, J., Hashemi, S., (2003). Is Microfinance an Effective Strategy to Reach the Millennium Development Goals? Focus Note No. 24., CGAP, Washington DC. Mersland, R., Strom, R., (2009). Microfinance Mission Drift? World Development 38(1), 28-36. 71 Microfinance Information eXchange. (2010). Asia 2009 Microfinance Analysis and Benchmarking Report. Microfinance Information eXchange, Inc. Washington DC. Microfinance Information eXchange. (2010). Latin America and the Caribbean 2009 Microfinance Analysis and Benchmarking Report. Microfinance Information eXchange, Inc. Washington DC. Montgomery, H., Weiss, J. (2005). Great Expectations: Microfinance and Poverty Reduction in Asia and Latin America. Tokyo. Asian Development Bank. Morduch, J., (2002). Analysis of the Effects of Microfinance on Poverty Reductions.NYU Wagner Working Paper No. 1014. Navajas, S., Schreiner, M., Meyer, R., Gonzalez-Vega, C., and RodriguezMeza, J., (2000). Microcredit and the Poorest of the Poor: Theory and Evidence from Bolivia. World Development.28 (2). From http://www.microfinance.com/English/Papers/Bolivia_Poorest.pdf Panjaitan-Drioadisuryo, D., Cloud, K., Cloud, R., (1999). Gender, selfEmployment and Microcredit Programs: An Indonesian Case Study. Quarterly Review of Economics and Finance 39(5), 769-779. Park, H., (2009). Linear Regression Models doe Panel Data Using SAS, Stata, LIMDEP, and SPSS.University Information Technology Services.Center for Statistical and Mathematical Computing.Indiana University. Paxton, J., Graham, D., Thraen, C., (2000). Modeling Group Loan Repayment Behaviour: New Insights from Burkina Faso. Economic Development and Cultural Change 48(3), 639-655. Pit, M., Shahidur, R., Khandker, S,. Omar, H., Millimet, D., (2003). Credit Programs for the Poor and Health Status of Children in Rural Bangladesh .International Economic Review 44(1), 87-118. Rhyne, E., (1998). The Yin and Yang of Microfinance: Reaching the Poor and Sustainability. MicroBanking Bulletin 2, 6-9. Seibel, H., Torres, D., (1999). Are Grameen Replications Sustainable, and Do They Reach the Poor? The Case of CARD Rural Bank in the Philippines, Journal of Microfinance 1(1), 117-130. 72 Stallings, B., Studart, R., (2002). The Financial Sector in Latin America ; The Challenges of Stability, Growth, and Equity. Retrieved March 1, 2011 .From https://project.iss.utokyo.ac.jp/nakagawa/members/.../2(4)Stallings.final.pdf UNICEF, (1997).Give Us Credit. Division of Evaluation, Policy and Planning. University of Oxiord.Oxford Poverty and Human Development Initiative.Retrieved March 1, 2011. From http://www.ophi.org.uk/policy/multidimensional-poverty-index/ Gonzalez-Vega, C., Schreiner, M., Meyer, R., Rodriguez-Meza, J., Navajas, S., (1997).BancoSol: The Challenge of Growth for Microfinance Organizations. 129-170. Hartmut Schneider Ed. Microfinance for the Poor?Paris: OECD. Woller, G., (2002). The Promise and Peril of Microfinance Commercialization.Sma// Enterprise Development 13(4), 12-21. World Bank., (2011). Access to Financial Services and the Financial Inclusion Agenda around the World. Policy Research Working Paper, No. 5537. Wright, G., (2000). Microfinance Systems: Designing Quality Financial Services for the Poor. ZED Books Ltd. London & New York and the University Press Ltd. Dhak 73 Appendix 1 Country % Households with Access to a Bank Account Asia 98 32 11 59 48 59 26 Singapore Bangladesh Pakistan Sri Lanka India Thailand Philippines Latin America and the Caribbean 44.1 39.7 39.2 4.5 35 9.9 3.7 17.8 35.2 Mexico Brazil Columbia Peru Ecuador Bolivia Paraguay Guatemala Panama Source: The World Bank 74