An Evaluation of ACCION Chicago's borrower evaluation process:

Transcription

1 1 An Evaluation of ACCION Chicago's borrower evaluation process: Investigating evidence of asymmetric information in the United States microfinance loan market Carolyn Fallert MMSS Thesis 2012

2 A special thanks to: 2

3 3 I. Introduction Microfinance is a hot topic for economists studying developmental economics today. An ideal solution for providing sustainable economic growth in developing economies, microfinance uses existing social capital to provide specialized financial services to entrepreneurs who otherwise would not have access to credit, to savings programs, or to business training. It uses a renewable source of funding and attempts to uplift microentrepreneurs in impoverished communities. In the past decade, many nonprofit organizations have adopted the microfinance model from developing economies and have implemented it in the United States. In the United States, however, this microfinance model faces a new set of challenges and limitations. Today, United States microfinance organizations like ACCION Chicago strive to measure their impact in American communities and to refine their loan process to best fit the needs of the American microenterprise sector. This study reviews existing research on microfinance in the United States and uses borrower records from ACCION Chicago to evaluate ACCION's loan approval process. Using these borrower records, the researcher investigated evidence of asymmetric information in ACCION's microfinance loan market to gather information about potential financial risks for ACCION Chicago as a US microfinance lender. One of the greatest challenges for United States microfinance organizations today is maintaining financial sustainability. By improving their loan approval process and

4 4 understanding the risks of certain borrowers, microfinance organizations can improve their profitability and reduce financial risks. Evaluating the loan process of US microfinance organizations does not only benefit lenders, however. By accurately anticipating the risks of borrowers, microfinance organizations can actually provide more loans for borrowers at more ideal interest rates. If microfinance organizations can predict the default probabilities of borrowers, they are more likely to receive repayments in full. As a result, microfinance organizations with higher repayment rates have more available capital to generate more loans for borrowers. Borrowers benefit not only from an increased source of capital, but also from appropriate interest rates. If a microfinance organization over estimates a borrower s probability of default it will likely assign that borrower an interest rate that is too high. For a borrower who is risk averse and who receives an unusually high interest rate, this high interest rate may discourage them from borrowing altogether. To properly cater to the needs of borrowers, microfinance organizations need to make an effort to accurately anticipate the risk of borrowers and to assign interest rates according to this evaluation. This study is designed to evaluate the lending process of ACCION Chicago. Information from this evaluation is meant to improve ACCION Chicago s sustainability while providing more opportunities and more ideal interest rates for subprime borrowers. II. Literature Review Background on microfinance in the United States

5 5 Microfinance is a relatively recent approach to facilitating economic growth in the United States. By definition, microfinance is the process of providing financial services to low-income individuals who traditionally use these services to expand a microenterprise. Over time this definition has expanded to include not only business entrepreneurs, but consumers as well. In the field of developmental economics, academics have conducted and continue to conduct extensive research that evaluates the process of these financial services and the impact they have on borrowers, on borrowers' communities, on social dynamics, and on economic growth. Because microfinance is a relatively recent approach to stimulating financial development, there is still a great demand for policy research that evaluates its effectiveness. During the last several decades, the scope of microfinance has broadened far beyond the Grameen Bank in Bangladesh and has found its way to the United States. As a potential solution for stimulating not only developing economies in Asian and Latin American countries but the United States economy as well, these specialized financial services for low-income individuals seem even more attractive after the recent financial crisis in the United States. If microfinance can facilitate economic growth by supporting American microenterprises, it may be another means of providing jobs and encouraging economic activity. Mark Schreiner and Jonathan Morduch (2001) recently investigated the outlook and the associated challenges of microfinance in the United States. They discussed

6 6 microfinance in light of "The Third Way" policy that Bill Clinton introduced in According to their perspective, this policy for writing financial policies and programs encourages individuals to be financially responsible and maintains deregulation while addressing social inequalities. Schreiner and Morduch described microfinance as a "compromise between conservative and liberal policies" that "aims to encompass the best of both, coupling market-based economic approaches with a commitment to social justice" (p. 5). Theoretically, microfinance seems like the all-around perfect policy solution for stimulating economic growth. In reality, however, there is not nearly enough empirical analysis or other research to provide evidence that microfinance in the United States will have a significant impact on the development of small businesses and their surrounding communities. In fact, the effectiveness of microfinance in developing countries is still debated. Even as a well-established economic policy for countries with developing economies, microfinance is still a widely debated solution for economic growth. Academic researchers and policy makers continue to argue over the effectiveness and the ultimate impact of microfinance services on borrowers and their local communities. Ram D. Rajbanshi, Meng Huang, and Bruce Wydick claim, there has been a substantial disagreement regarding the impact of microfinance (2012, p. 1). They are currently working on a paper that investigates the discrepancy between data from practitioners in

7 7 the field of microfinance and that from academic researchers by measuring the causal effect of microfinance on villages in the eastern region of Nepal. According to their results, 68.3% of the significant apparent microfinance impact observed by practitioners is an illusion driven by correlated unobservable factors (2012, p. 1). This study illustrates the persistent debate about the impact of microfinance on developing economies, and the importance of designing rigorous experimental studies to capture the causal impact of microfinance on borrowers and on surrounding communities. Like microfinance in developing countries, microfinance in the United States will require robust empirical evidence to prove itself as an effective approach to economic development. Microfinance has a lot of potential to encourage financial growth among lowincome entrepreneurs in America. In order to encourage policy makers and social policy programs to consider microfinance as a viable option in the United States, however, microfinance organizations servicing American entrepreneurs will need to couple with the academic world to provide concrete evidence of their economic and social impact. This research should include an evaluation of the loan approval process of US microfinance organizations and of the potential financial risks of this lending model. The economic model for microfinance organizations in the United States is significantly different from that for microfinance organizations in developing countries. Therefore, it

8 8 will be imperative for policy makers and American microfinance organizations to conduct this research and to evaluate microfinance in the context of the United States. Distinct characteristics of Microfinance in the United States There are several fundamental differences between microfinance in the United States and microfinance in developing economies that present unique challenges and restrictions for US microfinance organizations. Because these distinct characteristics of US microfinance organizations reflect the culture, the legal environment, and the economic forces of American society, they are inevitable aspects and challenges of microfinance in the United States. According to existing research, US microfinance organizations are distinguished by four fundamental characteristics: (1) their limited use of the group lending microfinance model, (2) their reliability on outside sources of funding, (3) the strict legal regulations of their economy, and (4) the small population of microentrepreneurs that utilizes their financial services. US microfinance organizations typically do not implement the group lending model of microfinance organizations in developing economies because is it not compatible with American culture. According to Schreiner and Morduch (2001), the group lending model is not as effective in America due to fundamental cultural differences between borrowers in developing economies and those in America. American, subprime lenders do not experience the same social pressure as those in developing countries and "are less concerned about allegiance to local leaders and about

9 9 the maintenance of an unblemished reputation" (p. 15). Because ACCION Chicago provides loans to borrowers throughout the Chicago area, it is very unlikely that these borrowers know each other, let alone understand each other's riskiness. Schreiner and Morduch explain that many US microfinance borrowers lack the social capital that is necessary to select and to trust other borrowers in a group lending model. Because group lending is often used to mitigate adverse selection and moral hazard, microfinance lenders in the United States that do not use this model could face more financial risk when providing microfinance loans (Amendáriz de Aghion & Morduch, 2005, p. 88). Although some US microfinance organizations have adapted the group lending model in the United States, this model tends to be successful among a group of acquaintances and among borrowers in immigrant communities only. Grameen America, which is a close rival to ACCION USA, has adapted their original group lending model in Bangladesh to their United States microfinance organization. According to the Grameen America website, Grameen America selects borrowers who are business owners below the US poverty line and then requires these borrowers to create a lending group by identifying four other borrowers. In a 2010 Times article about microfinance in America, Barbara Kivait explains that the social support of these Grameen America lending groups is more valuable to borrowers than the actual loans. Although this group lending model did not necessary eliminate all financial risk for

10 10 Grameen America, it provided valuable social capital for its borrowers. According to Jeffrey Ashe, the group lending model of Working Capital s microfinance services was very successful among borrowers in immigrant communities (2000). Once "the United States' largest peer-group lending program," Working Capital also adapted the group lending model to their microfinance services to facilitate business growth through the creation of professional networks and the exchange of business advice. These groups not only encouraged economic growth, but also strengthened the community and the social support for local microenterprises. The social connectivity of immigrant communities in the United States creates a more ideal setting for the group-lending model because the social pressure and personal obligation in the group already exists. It is important to note, however, that Working Capital eventually collapsed due to financial losses (Kiviat, 2009); therefore, this group lending model was unable to reduce all financial risks for the organization. Although the group lending is applicable for some US microfinance organizations, it is not an ideal model for organizations like ACCION Chicago, which lends to borrowers from a variety of social networks and neighborhoods. Therefore, most US microfinance organizations must find other ways to mitigate for adverse selection and moral hazard in their lending processes. The second distinct characteristic of US microfinance organizations is their dependence on generous donations and outside sources of funding. Self-sufficiency is a challenge to all existing microfinance organizations in the United States because these

11 11 organizations cannot demand interest rates as high as those of microfinance organizations in developing economies. Lisa J. Servon, who is currently an associate professor and the director of the Community Development Research Center of the Milano New School for Management and Urban Policy in New York, has written several articles about the development of microenterprise in the United States. While analyzing the current challenges that face microfinance in the United States, Servon lists sustainability as one of the greatest threats to microfinance organizations in America. She recognizes efforts to provide funding for these organizations, but claims that "few programs have devised ways to cover their costs of doing business in a significant way" (2006). What makes the sustainability of these organizations so challenging is the high risk and high cost characteristics of their work (Bhatt and Tang, 2001). A 2002 study by Nitin Bhatt, Gary Painter, and Shui-Yan Tang investigated the financial sustainability of several microcredit programs in California using empirical analysis. To collect data for this research, Bhatt et al. surveyed 87 California organizations listed in "the 1996 Directory of U.S. Microenterprise Programs,... The California Business Resource Guide published by the Economic Development Corporation of Los Angeles County, the membership directory of the National Association of Community Development Loan Funds, the San Francisco-based Development Fund, and various World Wide sites" (p. 216). Their surveys inquired about these organizations' funding, financial services, and

12 12 borrower populations. Out of these 87 organizations, they chose four organizations with which to conduct an in-depth study using data records and personal interviews with borrowers and loan officers. According to their empirical research, all of the 87 programs depended on external funding, and "about 27 percent of their funding [was] obtained from private donors" (p ). The rest of their funding was provided by government subsidies. This dependence on external funding is prevalent in California, but it is not unique to these microcredit organizations. ACCION Chicago seems to experience a similar dependency on outside funding. On the ACCION Chicago website ( ACCION recognizes a long list of non-government donors who provide generous support for their credit program. Some of these donors made donations of over $100,000 to support the organization in For microfinance to be an effective and efficient program for long-term economic growth in the United States, financial sustainability is a necessity. In order to move closer to self-sufficiency, US microfinance organizations need to focus on minimizing risks and reducing costs while strategically maintaining external funding. Continued policy research will help develop an effective approach to improving this self-sufficiency. To fully understand the distinction between microfinance in developing economies and microfinance in the United States, it is also important to consider the political and economic environment of US microfinance organizations. In the context of the US governing territory, both microfinance organizations and microenterprises face

13 13 much stricter regulations and legal requirements than those in developing economies. According to Schreiner and Morduch's (2001) research, these legal regulations increase the cost of entry for microenterprises by requiring licenses for certain industries and reduce the profitability of microfinance loans by capping interest rates. If low enough, these caps on interest rates can also exclude certain borrowers from the microfinance loan market. Although these legal regulations provide many benefits for consumers and for certain borrowers, they also inhibit the immediate success of microfinance in the United States (p ). In the context of the US economy, microfinance organizations also serve a significantly smaller population of microentrepreneurs. Schreiner and Morduch explain that Americans are often discouraged from becoming microentrepreneurs in the United States because they face difficult competition from large firms and have large opportunity costs to running their businesses (pp ). In the United States, large firms like Wal-Mart and McDonalds dominate most industries for consumer goods. Rather than compete with these industrial giants, many potential microentrepreneurs turn away from microfinance sector or enter service industries. For a low-income American however, it is much easier and less risky to find a low-wage job with one of these large companies. Compared to owning a microenterprise, working a low-wage job typically requires less work, promises a consistent paycheck, and provides sizable benefits. Therefore the financial security of a low-wage job is an attractive alternative.

14 14 The individual lending model of most US microfinance organizations, the dependence of US microfinance organizations on outside sources of funding, the strict legal regulation of American businesses, and the small microenterprise sector in the United States present unique challenges to US microfinance organizations and clearly differentiate them from microfinance organizations in developing economies. As a reflection of American society, these unique characteristics of US microfinance organizations will be inevitable challenges to the success of microfinance in the United States. In order to accurately formulate public policy for microfinance in the United States, it is necessary to understand the implications of these distinct characteristics by conducting empirical research and policy evaluations in the context of the United States specifically. Empirical Research on Microfinance in the United States Most existing microfinance research focuses on models that are very specific to developing economies. Because there are fundamental differences between microfinance in the US and microfinance in developing economies, policy makers and microfinance organizations in the United States cannot turn to this existing research as evidence of the successful economic and social impact of microfinance in America. Therefore it is essential to evaluate the process of providing microfinance services and their impact in the context of the United States economy.

15 15 There is actually very limited academic research on microfinance in the United States specifically. According to Lisa J. Servon, one of the greatest challenges to microfinance in the United States is a lack of sufficient data and empirical analysis (2006, p. 356). Of the existing studies on microfinance in the United States, very few use an empirical approach to analysis. The studies that do use empirical analysis rarely evaluate the loan approval process or the financial risks of these US microfinance organizations. In addition to the empirical research of Bhatt et al. in California that was described in the previous section, there are a few other studies on US microfinance that should be noted. The Self-Employment Learning Project (SELP) is an extensive empirical evaluation of microfinance financial that provided a valuable foundation for research on microfinance in the United States in A five-year evaluation of 405 microfinance clients and their respective microfinance programs, SELP was one of the most extensive statistical studies of microfinance in the United States over the longest period of time. Peggy Clark and Amy Kays conducted this study in conjunction with The Aspen Institute, which designed their research surveys. According to Clark and Kay's report of the study, SELP used The Aspen Institute surveys to track the "employment status and sources of earnings of the respondent," the "business profitability and net worth," the "program participation and loan history," and the "household composition, income and net worth" (p. 80). These surveys measured the income and economic status

16 16 of the microentrepreneurs over time while monitoring the operations of each microfinance program. The results of this study indicated that microfinance clients experienced an average increase in household income of $8,484 during the five-year study. For more than 50% of the subjects in the study, this increase in household income was enough to raise them above the national poverty line. Although those microfinance clients did not always use their microenterprises as their primary source of income, these small businesses contributed an average of 37% to household income of borrowers (p. vii - viii). Although SELP is respected as one of the most extensive empirical studies of microfinance in the United States, it has an important limitation: it does not use a control group to capture the specific impact of the microfinance programs on microentrepreneurs. In the empirical analysis of the study, Clark and Kay could not immediately conclude that the increase in microentrepreneurs household income was due to his or her participation in a microfinance program. Therefore, their research did not capture the causal impact of microfinance on the borrowers of their study. In order to capture this causal impact, Clark and Kays would need to compare the increase in household income of microfinance clients to other microentrepreneurs who did not participate in a microfinance program. This comparison would have captured the true impact of existing microfinance programs. Although this study was a significant contribution to the field of US microfinance, its empirical research methods were somewhat limited.

17 17 The 1998 ACCION impact evaluation by Cristina Himes and Lisa J. Servon was also an important contribution to the field of research on microfinance in the United States. At the time of publication, Himes was the director of operations for ACCION USA and Servon was an associate professor for the Department of Urban Planning and Policy Development at Rutgers University. Their study used data from ACCION loan applications and from interviews with ACCION clients to compile both quantitative and qualitative data for the study. Because most ACCION clients in the study received between two and four microfinance loans over several years, Himes and Servon were able to compile a longitudinal database from loan application data (p. 5). After successfully combining quantitative and qualitative data, Himes and Servon evaluated the impact of ACCION's loans to find that ACCION clients experienced significant financial growth and an increase in household income during their ACCION loan participation. As their analysis illustrated, clients who received an average of three loans during a 17 months period experienced "an average 47% increase in monthly profits," "an average 42% increase in equity," and "an average 38% increase in monthly take home income" (p. 11). These numbers illustrated a positive impact on the financial success and the financial security of microentrepreneurs during their participation with ACCION s microfinance loan program. Himes and Servon recognized that there was a limitation to their empirical analysis on microfinance in the United States, however. According to their perspective, "the field desperately needs some research that uses a control group.

18 18 Anecdotal evidence about the positive impact of participation in a microenterprise program on that group that does not choose to pursue self-employment abounds, but there is no research that systematically and extensively studies that group" (p. 8). Although their empirical analysis does not isolate the causal impact of ACCION s microfinance loans, the results of Himes and Servon's analysis are a great foundation for future academic studies. Josephine M. LaPlante of the Edmund S. Muskie Institute of Public Affairs conducted a 1996 study that also evaluated in the impact of programs that promoted the development of microenterprises. Its analysis focused on the economic impacts of the Small Business Assistance Programs of Coastal Enterprises, Inc. in Wiscasset, Maine while measuring the returns to investment for these programs. Coastal Enterprises, Inc. aims to facilitate sustainable economic growth in Maine by providing loans to small businesses. In her empirical analysis, LaPlante investigated the benefits of these loans to the local, state, and federal governments while estimating their impact of these loans on firms in the economy. Unfortunately, this study also fails to isolate the causal impact of these loans on microenterprises specifically. Although several studies have attempted to measure the impact of microfinance in the United States, the existing empirical studies have several limitations. The most distinct limitation of these studies was the lack of a control group. Of the existing research, there were no empirical evaluations that isolated the causal impact of

19 19 microfinance services on microentrepreneurs by using a treatment and control group set up. It is also important to note that none of these studies attempt to evaluate the process of microfinance lending in the United States. In order to accurately formulate policies for microfinance in the United States, it is necessary to understand not only the positive impact of microfinance on microentrepreneurs but also the risks involved in microfinance lending. In order to address the financial sustainability of microfinance in the United States, researchers need to investigate evidence of asymmetric information in the lending process. By investigating adverse selection and moral hazard in the loan approval process of microfinance organizations in the United States, researchers can evaluate the efficiency and sustainability of US microfinance. ACCION Chicago Although there are many microfinance organizations in the United States, this study focuses on the lending services of ACCION Chicago specifically. As a well established American microfinance organization, ACCION Chicago has the necessary infrastructure and the support from ACCION USA to provide microfinance loans to hundreds of small businesses. By catering its microlending to small businesses and by developing a unique approach to evaluating the financial risks of borrowers, ACCION Chicago has been at the forefront of microfinance in the United States. ACCION Chicago's mission is to encourage economic and community development in the Chicago area by providing loans to small business owners who do

20 20 not have access to traditional financial services through banks. By providing these loans, ACCION Chicago aims to facilitate the growth of small businesses while encouraging them to be self-sufficient and to be financially responsible. ACCION intends to help business owners achieve higher levels of sustainable income while stimulating the local economy. Originally founded as an associate of ACCION International in 1994, ACCION later became part of a network of ACCION organizations in the United States. Known as the US ACCION Network, this group of microfinance organizations includes independent offices in New Mexico, California, Texas, New York, and Chicago. Although each of these organizations works independently, the US ACCION Network is committed to providing microfinance services, especially credit, to microenterprises in the United States. The US ACCION Network, and consequentially ACCION Chicago, provides financial services that focus primarily on providing loans to borrowers who are unable to qualify for a traditional bank loan. According to the ACCION Chicago Credit Policy Manual, ACCION Chicago offers loans to small businesses that "range in size from $500 to $25,000" and range in term "from 3 months to 60 months" (p. 13). As of January 2010, however, ACCION Chicago has increased this loan amount to $50,000 for well established businesses and to $20,000 for start-up businesses. All of these loans are used for specific business projects. Although a few of these loans are "credit builder loans" that help a borrower build credit with ACCION Chicago, these credit builder loans also help with specific business

21 21 developments. The majority of ACCION microfinance loans are single-time loans; however, ACCION will occasionally provide refinanced loans. According to Jill Aldridge, who is the Vice President of Lending and Marketing of ACCION Chicago, these refinanced loans extend an existing ACCION loan for borrowers who demonstrate a need for further financing and who are committed to repaying their loans. To be eligible for a microfinance loan, ACCION Chicago borrowers must have avoided bankruptcy during the past twelve months, they must be self-employed, they cannot qualify for a loan through a bank, they must show a strong commitment to their business' development, and their business must not have an "adverse effect on the development of the community." ACCION confirms these and other eligibility requirements through an extensive loan application process (p. 8-9). As a microfinance organization that focuses on lending, ACCION Chicago has a very unique approach to approving loans for its borrowers. Because ACCION lends to local business owners who do not have a very strong credit history, credit score is never the single, deciding factor for loan eligibility. ACCION gauges applicants commitment to their business through interviews and outside recommendations. It evaluates a potential borrower s financial security by looking at his or her expected cash flow and by investigating the applicant s repayment history with other forms of credit. According to Jill Aldridge and to the ACCION Credit Policy Manual, the loan application process includes an online loan application, an interview with a loan officer, a visit to the

22 22 business site by an ACCION representative, a cash flow analysis, and the evaluation of required documentation (p. 22, 24). The ACCION Approval Committee looks at the loan package, which includes all information from the application process, to make an educated decision about the borrowers' eligibility. ACCION Chicago s loan approval process seems like an ideal approach to overcoming problems of asymmetric information in the loan application process. By hopefully distinguishing between risky and non-risky microfinance borrowers, it is designed to limit ACCION Chicago's risk as a lender. If borrowers successfully hide information about their financial riskiness during the application process, adverse selection is likely to create inefficiency in ACCION s market for microfinance loans and to threaten the sustainability of the lending organization. Likewise, if borrowers successfully hide information about their commitment to their business or about the effort they are willing to put into a project, moral hazard will likely increase financial risks for ACCION. Because ACCION tries to measure information that borrowers may deliberately hide, the task of measuring borrowers riskiness and borrowers commitment to financial success can be incredibly difficult. By including interviews, recommendations, and a loan package analysis, ACCION Chicago has developed a risk evaluation of borrowers that is more qualitative than a typical bank process and that hopefully overcomes the inefficiency of asymmetric information. Because it considered a variety of documentation and personal interviews, this loan approval process also

23 23 accommodates borrowers in the subprime loan market who do not have ideal credit scores. Besides using an extensive loan application process to screen its borrowers, ACCION Chicago also requires collateral for certain loans to minimize their risk as a microfinance lender. When determining the necessity of collateral for a loan, ACCION takes into consideration the size of the loan as well as the borrowers access to collateral. As stated in the Credit Policy Manual, when collateral is necessary for a loan contract, its value should be "125% of the monetary value of the loan" (p. 17). Demanding collateral that has a significant value ensures that repaying the microfinance loan is more profitable than losing the promised collateral. Approved forms of collateral include mortgages, automobiles, mobiles homes, and even business electronics such as computers and fax machines. Including a collateral requirement in loan contracts enables ACCION Chicago to incentivize borrowers to repay their loans while reducing losses in the case of default. By using a qualitative approach to approving loans and by accepting a variety of collateral, ACCION Chicago is able to cater to the needs of its borrowers while mitigating its own financial risk. For ACCION Chicago, this is necessary for achieving its mission and for maintaining long term sustainability. As an example of microfinance lending in the United States, ACCION Chicago is an ideal organization to investigate because it has developed extensive methods for mitigating risk in the microfinance lending market and because it is the largest in the

24 24 Chicago area. As a member of the US ACCION Network, it also represents one of the largest microfinance efforts in the United States. According to Ben Bernanke s speech at the 2007 ACCION Texas Summit on Microfinance in the United States, "ACCION has been at the forefront of the development of microfinance in the United States." Because the Chicago branch of ACCION USA is making an earnest effort to accurately measure its impact on its borrowers and on its surrounding community, it is an excellent US microfinance organization to study. Measuring Adverse Selection and Moral Hazard This study is designed to evaluate ACCION Chicago s process of approving microfinance loans by investigating evidence of asymmetric information between lender and borrower in the context of the subprime lending market. In order to accurately measure this asymmetric information, it was necessary to consider the analytical approach of similar studies in the field of economic research. These existing studies address the challenges of distinguishing between adverse selection and moral hazard among loan applicants and of evaluating lending processes in the subprime lending market. Asymmetric information in the lending process results in two potential outcomes: adverse selection and moral hazard. Adverse selection typically occurs when there is asymmetrical information between the borrower and the lender during the loan approval process. For example, the borrower could have more information about his

25 25 probability of defaulting than the lender does. Unlike adverse selection, moral hazard occurs after the loan has been approved, and the borrower finds an incentive to change his or her probability of defaulting during the repayment process. Anticipating the riskiness of potential borrowers is crucial for any bank or lending organization because this risk evaluation determines whether or not the organization receives its money back. Therefore this asymmetric information, whether it is adverse selection or moral hazard, can threaten the financial security and the sustainability of a microfinance organization. The profit of lending organizations depends directly on the repayment of their loans. Although microfinance organizations in the United States are typically non-profit organizations and do not seek to make significant profits from their loans, their longterm sustainability depends on this loan repayment. According to Joseph E. Stiglitz and Andrew Weiss (1981), financial losses from adverse selection are "a consequence of different borrowers having different probabilities of repaying their loan. The expected return to the bank obviously depends on the probability of repayment, so the bank would like to be able to indentify borrowers who are more likely to repay" (pp. 393). Because this risk evaluation is so important to the sustainability of a non-profit microfinance organization, it is essential to investigate evidence of asymmetric information in the loan approval process. According to existing research, there are several challenges to identifying asymmetric information between ACCION Chicago and their borrowers. One of these

26 26 challenges is quantifying the unknown information about ACCION borrowers during the loan approval process. Empirical research on asymmetric information typically measures this hidden information by using a unique approach to data analysis or by finding another, outside source of borrower information that was not originally available to the lender. There are several different experimental and analytical approaches to the measurement of asymmetric information in existing academic research. Stephen Sharpe argued that lenders can acquire this hidden information by observing borrowers during the loan repayment process. He developed a dynamic theory of asymmetric information between banks and firms in his 1990 article, Asymmetric Information, Bank Lending and Implicit Contracts: A Stylized Model of Customer Relationships. According to Sharpe's theory, lenders gather important information about firms during the lending process that leads to almost perfect information between banks and firms during future periods of lending (p. 1070). During this process, firms establish a "reputation" with banks that depends on the outcome of their loan repayment. Sharpe analyzed this situation by setting up a two-period game and by using subgame perfect strategies from game theory (p. 1076). In the first period of the game, banks in a competitive loan market approve loan contracts for several firms. By observing the repayment behavior of these firms, each bank is able to collect information about their current customers. In the second period, banks are no longer

27 27 perfectly competitive because they can offer loan contracts to previous customers based on the outcomes of the first period. Firms in the second period can choose between the offers of their current bank and the offers of other, competing banks. For customers who successfully repaid their loans in the first period, banks can offer more attractive interests rates and loan contracts, which tends to increase retention among reliable firms. Sharpe's model of dynamic relationships between banks and firms is slightly different from the relationship between ACCION and its microfinance borrowers. Unlike the banks in the perfectly competitive market of Sharpe's model, ACCION is one of the few organizations that provide microfinance loans to borrowers in the subprime lending market of Chicago. Despite this difference, Sharpe's model provides a realistic model of the dynamic relationship between borrowers and lenders that is comparable to ACCION's observation of borrower behavior during the microfinance loan process. This model is particularly applicable to ACCION's refinancing process. Each year ACCION offers several refinanced loans, which are loans for borrowers who are in good standing with the organization and who demonstrate a need for an extended ACCION loan. ACCION captures hidden information about borrowers during the initial loan process that enables them to make a more educated decision about approving refinanced loans. Of course, there are some limitations to using this model for investigating evidence of adverse selection because the group of borrowers who apply for refinanced loans at

28 28 ACCION may be a self-selected sample. The application of Sharpe's model in this study would require further investigation. Another challenge to measuring asymmetric information is distinguishing between adverse selection and moral hazard in the loan selection process and the repayment process respectively. In their 2009 article, Observing unobservables: Identifying information asymmetries with a consumer credit field experiment, Dean Karlan and Jonathan Zinman outline an approach to differentiate between these two results of asymmetric information. As defined earlier, adverse selection is the result of asymmetric information during the loan approval process. When adverse selection occurs, the borrower successfully hides information about their probability of defaulting on the loan. Karlan and Zinman refer to this as "hidden information" or as one of the "exante selection effects" of the loan approval process. Moral hazard, on the other hand, is the "hidden action" or the "ex-post incentive effects" (pp. 1994). In the case of moral hazard, a borrower's probability of defaulting may increase because the borrower understands that the lender, or ACCION Chicago in this situation, will bear the burden of a failed business project. Because ACCION will face the financial consequences of a defaulted loan, there might be a greater incentive for borrowers to pursue more risky projects or to spend more recklessly once they have the loan in hand. This is especially true if borrowers do not have collateral on theirs loans or if they face very few consequences to defaulting. This is also more likely to occur if a borrower receives a

29 29 loan with a high interest rate that makes default more profitable than repaying. It is difficult to distinguish between adverse selection and moral hazard because they are often indicated by the same positive correlation between interest rate and probability of defaulting. Karlan and Zinman designed an experimental study that distinguished between these two instances of asymmetric information and provided an excellent example for future natural experiments. To design the study, Karlan and Zinman worked with a South African lender that serves a consumer loan market to set up three randomizations: (1) randomized interest rates offered in mailed promotions that were sent to 57,533 former clients (2) randomized changes to these initial interest rate, either increases or decreases (3) randomized offers to provide better pricing on future loans if the borrower successfully kept up with repayments. Using this multi-dimensional randomization, Karlan and Zinman designed an experiment that measured evidence of adverse selection before the loan was approved and evidence of moral hazard while the loan was being repaid. This experimental design is ideal for measuring asymmetric information because it allows the researcher to separately test for evidence of adverse selection and for that of moral hazard. Although this experiment design provides important insight into the process of distinguishing between adverse selection and moral hazard, there were not enough resources to directly apply this research method to this ACCION Chicago study.

30 30 Because this randomization method requires extensive method preparation, data collection, assistance from the organization, and follow-up data on loan repayments, it was not the most realistic approach to measuring adverse selection among ACCION Chicago borrowers. However, if the appropriate resources were available, this would be a recommended research method. Distinguishing between adverse selection and moral hazard in the ACCION data in this study required a slightly different approach. While designing a model to identify adverse selection and moral hazard among ACCION Chicago borrowers, it was important to remember the difference between ACCION's microfinance borrowers and typical bank customers. The microfinance loans from ACCION Chicago are fundamentally different from the consumer loans in the Karlan and Zinman study because ACCION Chicago borrowers are not eligible for bank loans and are part of the subprime lending market. Lending to ACCION borrowers is naturally riskier than lending to a consumer who is eligible for a typical bank loan. In their 2007 article, Liquidity constraints and Imperfect information in Subprime Lending, William Adams, Liran Einav, and Jonathan Levin investigate data from subprime lenders specifically while successfully distinguishing between adverse selection and moral hazard. Their study used data from an auto sales company that serves clients from the subprime lending market. This is an important study to consider because it focuses on a subprime lending market that is similar to that of ACCION Chicago. Because the auto sales company works primarily with used car dealerships, customers

31 31 tend to be low-income workers who have low credit scores, ranging between 350 and 800. Less than half of the credit scores for these borrowers were above 500 (p. 5). Because this auto sales company worked with customers who have such a poor credit history, it experienced a high rate of default and it demanded high interest rates. According to Adams et al., there was strong evidence of both adverse selection and moral hazard in the subprime lending market for auto loans. By regressing the probability of loan repayment on the loan amount for each level of risk in their data, Adams et al., found a significant negative correlation between loan size and repayment probability. In the study, they defined evidence of both adverse selection and moral hazard by this negative correlation; therefore the effects of asymmetric information were, in fact, prevalent in the subprime lending market. This study successfully distinguished between adverse selection and moral hazard by using two distinct variations in the loan sizes: (1) the borrower's chosen down payment, which could be at or above the minimum requirement (2) other "exogenous variations" in loan size that depended on the cost of the car (p. 7). The first variation in loan size, which is correlated with the chosen down payment, indicates unobservable borrower characteristics that may affect the probability of default (p. 29). Therefore, Adams et al. used this variation to successfully measure evidence of adverse selection (p. 30). To measure moral hazard, the researchers looked at the correlation between loan default and the exogenous variation in the loan size that did not reflect the down payment.

32 32 Unfortunately, this empirical method was not perfectly applicable to this ACCION Chicago study because microfinance loans involve a different approval and payment process than consumer loans. The most fundamental difference between the two is the lack of a down payment for microfinance loans. The two loan processes could be compared if the cost of the business project for a microfinance loan were compared the cost of a car in the Adams et al. research. In this perspective, any savings that the ACCION borrower uses to cover the initial cost of a business project would be similar to a down payment for a car. Although ACCION Chicago requires that borrowers have a certain amount of savings before they receive an approved loan, ACCION does not record the amount of money that a borrower saves to cover the initial cost of a business project before using a microfinance loan. If there were a record of this savings, this variation in savings might be used to distinguish adverse section from moral hazard among ACCION borrowers in the same way that the down payment was used to indicate adverse selection among car consumers. In this model, a negative correlation between the amount of borrower savings for the initial cost of the business project and the probability of default would be evidence of adverse selection. The exogenous variation in the cost of business projects could also be compared to the exogenous variation in the cost of a car. Ideally, the cost of a business project should be completely independent of a borrower s probability of defaulting. This cost depends on the needs of the business, the industry of the business, and the growth of the

33 33 business. However, there might be a variation in the requested loan size among ACCION loan applicants that is correlated with the hidden information or risk of the borrower. If a business owner anticipates defaulting on a loan, he or she may request a loan that is larger than the anticipated cost of the project. If borrowers have a high probability of defaulting on the loan, it would make sense for them to request a loan that is higher than necessary for their business project. Hopefully ACCION Chicago has a way of measuring the actual, anticipated cost of a business project; however this can be difficult to measure. Although Adams et al.'s analytical approach has potential in a microfinance setting, there are some limitations to the ACCION data and some fundamental differences between the two loan markets that make the application of this research method difficult. Because both studies address subprime markets, however, the results of the Adams et al. study provide useful background for the ACCION Chicago study. It is important to note that, though there was significant evidence of adverse selection and moral hazard in the Adams et al. study, they also argued that "risk-based pricing" can counter the effects of asymmetric information in these subprime markets. (p. 3) If riskbased pricing can, in fact, mitigate the effects of asymmetric information for ACCION Chicago as well, ACCION's loan pricing model is a great example to study because the loan contracts are written to fit the expected risk of borrowers.

34 34 To measure evidence of asymmetric information in ACCION Chicago s microlending process, this study addresses the challenge of distinguishing between adverse selection and moral hazard while evaluating the ACCION lending process in the context of the subprime borrower market. In order to address these challenges, however, this study selected a specific dataset from ACCION Chicago s records and from the Reference USA database. III. Data To empirically measure asymmetric information in ACCION Chicago's selection process, this study used background information about ACCION's borrowers, records of ACCION borrowers' loan repayments, and an outside source of information about these borrowers' riskiness. One of the greatest challenges of measuring adverse selection and moral hazard is finding information about borrowers that is unknown to the lender. In an attempt to compare ACCION's borrower data to ACCION's unknown information about potential borrowers, the researcher gathered data from the Reference USA database. Using data from ACCION Chicago and Reference USA, the researcher created a comprehensive data set that included ACCION's knowledge of borrowers, Reference USA's knowledge of borrowers, and records of borrowers loan repayments. ACCION data ACCION's borrower background information included records of borrowers' level of income, education, ethnicity, birth country, savings, credit score, business size,

35 35 number of employees, and details of their approved microfinance loan. ACCION collected this data for borrowers from the year 2008 to 2011; however, because ACCION Chicago recently reevaluated its approach to data collection and recordkeeping, the data from 2009 to 2011 provided the most detailed borrower records. The ACCION loan repayment data was saved as weekly records of defaulted loans and loans that were currently being repaid. Because the repaid loans were not included in these weekly repayment records, it was necessary to infer default rates. To determine whether or not the loans were actually repaid, the researcher used existing repayment records to track loans during the repayment process. It is important to note that the default rate in this study measured the number of borrowers who officially defaulted on loans. Often microlenders refer to default in terms of net losses rather than the number of borrowers who officially default on loans. Net losses are a more accurate measurement of the financial risk for a microfinance organization, but this measurement does not look at the individual cases of defaulted loans. Because the researcher did not have specific data on the value of net losses and wanted to investigate borrowers probability of default, this study s default rate measurement was the next best estimation. The data on default rates was paired with the background data on borrowers to provide a comprehensive data record of ACCION's borrower background information and of the loan repayments of each borrower. Limitations to ACCION data

36 36 Although ACCION's database included extensive information about each borrower and his or her loan repayments, it was important to note three limitations of this dataset. First, the set of variables in the ACCION data was someone inconsistent from year to year from 2008 to From 2008 to 2011, the data collection became more detailed, providing the most background information about borrowers in 2010 and Therefore, while running analysis on this data, it was important to consider the empirical results according to the year of the observations. The second limitation of this data was its purely quantitative focus. These borrower records did not attempt to capture the qualitative information about borrowers that ACCION used during its loan approval process. Because ACCION uses a unique and partially qualitative process for approving its loans, these data records may not represent all of the information that ACCION used to make its final loan approval decision. For example, part of the loan approval process included an interview with the potential borrower to gauge the borrower's commitment to his or her business and to repaying his or her loan. The information that ACCION gathered about borrowers during these interviews was not completely captured in the borrower database. Because the ACCION interviewer's judgment about a borrower's commitment was a more qualitative measurement, it may have been difficult to capture this information empirically. It is important to understand that the existing dataset represents ACCION's measurable and recorded information about potential borrowers during the loan approval process. Hopefully, as ACCION'S approach to data collection

37 37 continues to develop over time, they will be able to capture this qualitative information numerically to strengthen the empirical analysis of future, similar studies. The third limitation of the ACCION dataset was the uncertain measurement of default in the later years of data. Because it usually takes between one and two years to repay a loan, many of the borrowers in 2011 ACCION background data did not have the opportunity to fully repay their loan in full. Therefore, the respective repayment data for 2011 borrowers was incomplete and insufficient for data analysis. Although it may not perfectly represent ACCION'S knowledge of potential borrowers, this comprehensive dataset of borrower background information and loan repayments was the most feasible and reliable measurement of ACCION'S known borrower information. In order to address some of these limitations in the dataset, the researcher decided to use the ACCION data from 2010 because it provided the most information about borrowers and provided the most accurate measurement of default rates. Reference USA data In this study, the Reference USA database was the source of unknown information about ACCION Chicago borrowers. This database included data records of over 14 million businesses in the United States. In order to refine this database to best fit the needs of this study, the researcher focused on businesses that had an annual sales volume of less than one million dollars, that were located in the Chicago area, and that

38 38 were established between the years 1970 and The Reference USA database was a trusted source of business data because it was a comprehensive census of businesses and was used by several government agencies in Illinois to conduct existing policy research. To compile this comprehensive database, Reference USA keeps records of every phone number in every existing phonebook in the country. According to Eric Ebers, who is a Reference USA product trainer and representative for Illinois libraries, Reference USA makes approximately 26 million phone calls to listed businesses every year to update their existing database. For those businesses that refuse to provide specific information about their business or that do not respond to phone calls, Reference USA uses an economic model to predict information about the business, such as its size and its number of employees. Ebers explained that this model used a matrix to estimate business data according to the statistics of other similar businesses. When checked with existing data, the results of this model were fairly accurate. The Reference USA database provided comprehensive background information about the businesses in its database that included, but was not limited to, the company's description, establishment date, NAIC industry code, number of employees, categorized credit score, accounting expenses, advertising expenses, insurance expenses, labor expenses, and sales. In the analytical model of this study, Reference USA's information about company expenses was as the "unknown information" about ACCION borrowers. The composition of a business' expenses measured the business' riskiness and the

39 39 business commitment to financial success. This study assumed that a business owner who spent significantly less on insurance expenses than other business owners in his or her industry was probably riskier than most other business owners. It also assumes that a business owner who spent significantly more on advertising expenses was probably more committed to the success of his or her business. According to Jill Aldridge, who is the Vice President of Lending and Marketing at ACCION Chicago, it is necessary to compare these business expenses within industries because some industries are legally required to spend more on certain expenses than on others. For example, insurance expenditure is often correlated with the risks and the legal requirements of the business s industry. In order to compare ACCION's evaluation of borrowers to this "unknown information" in the Reference USA database, the researcher merged these two databases and analyzed the business expenses according to industry. Limitations of Reference USA data Recall that the Reference USA dataset of this study included businesses in the Chicago area that earned less than $1 million in annual revenue and were established between the years 1970 and Although these parameters matched the parameters of the ACCION database, they actually eliminated an important sample of businesses in the Reference USA dataset. During the data analysis process, the researcher discovered that these parameters excluded businesses that did not have a recorded date of establishment. Therefore, there could have been a Reference USA business in the

40 40 Chicago area that had an annual sales less than $1 million and was established after 1970, but was not listed in the study s Reference USA dataset simply because Reference USA did not record its date of establishment. Unfortunately, when the researcher attempted to collect these data observations with missing dates of establishment, the resulting dataset was too large for the study s available resources. Because this data was unattainable, the researcher was unable to measure any statistical difference between the existing Reference USA dataset and the missing Reference USA observations. It is possible that the study s Reference USA dataset is a slightly biased representation of the total population of Chicago business owners who earn less than $1 million in revenue and who established their companies after Assuming that Reference USA was unable to gather detailed information about businesses who were recently established or who were underdeveloped, the missing data observations in the Reference USA dataset are probably new business with limited public information. Because these missing observations are likely smaller and new businesses, this dataset may under-represent the newest and smallest businesses in Chicago. Unfortunately, this underrepresentation may slightly bias the comparison between existing ACCION businesses and other, small businesses in Chicago. Although the study s Reference USA dataset was a slightly biased representation of small businesses in Chicago, it provided the best data comparison given the study s available resources.

41 41 In the quantitative analysis of this study, the researcher focused on the Reference USA data that also described ACCION business from the 2010 database. Unfortunately, the Reference USA database did not include all of these 2010 ACCION businesses because many of ACCION business are quite small and are still developing their reputation in the market. Reference USA updates its database annually, but if a company is just beginning, it may take a few years for Reference USA to locate and to collect data on the company. Of the 198 businesses that were listed in the 2010 ACCION Chicago borrower database, only 62 businesses merged with the Reference USA database. Although a 33% merge rate is far from perfect, the 62 businesses that did merge provided a sufficient sample size for OLS regressions. When the researcher compared the merged ACCION businesses to ACCION businesses that did not merge with the Reference USA database, there was an insignificant difference between the credit scores, the income levels, the ethnic compositions, the business sector compositions, the gender breakdowns, and the geographical distributions of the two groups. There was, however, a significant difference between the default rates of the two groups. Businesses from the 2010 ACCION data set that defaulted on their loans were 62% less likely to merge with the Reference USA data set; therefore the merged sample may under represent the sample of defaulted borrowers. This bias in the merged sample is somewhat expected because borrowers who default on a microfinance loans are likely to be financially insecure and rather small in the overall market. Reference USA may

42 42 have trouble collecting data on companies who are struggling financially or who do not survive in the market for a very long time. In this study s data analysis, this merged sample could have created a bias in the final empirical results. Because it was less likely for defaulted loans to appear in the merged sample, it was difficult to accurately test the probability of default. Because the merged sample included so few of the defaulted loans, regressions would have underestimated any positive correlations between explanatory variables and the probability of default. Therefore, it was necessary to interpret the final empirical results according to this bias. Before applying the results of this study to ACCION s policies or to other microfinance organizations in the United States, the researcher would need to reevaluate the hypotheses with more representative data from other years and from other organizations. IV. Analysis Background data about ACCION borrowers To fully understand the implications of this study s data analysis, it is necessary to consider the context of the data. ACCION clients are unique borrowers because they represent a subprime market of borrowers that otherwise would not have access to formal financial services.

43 43 According to ACCION s records from 2008 to 2011, the average credit score of ACCION Chicago borrowers is 589. Of these borrowers, 33.3% have a checking account without a savings account, and 53.0% were classified has having a low income (an income less than $37,400, or less than 50% of the HUD Area Media Income for Chicago MSA). These borrowers typically run microenterprises in the food and beverage, retail, and professional service sectors. For the past four years, the ratio of male to female borrowers has been relatively even, and the majority of borrowers are of African American (42% of borrowers), Caucasian (35% of borrowers), or Latino/a (15% of borrowers) ethnicity. During the past four years, ACCION has provided loans to almost 500 borrowers. On average, the total amount financed for these loans is $7, during a two year period. The average interest rate of these loans is 14.7% and the average default rate is 16.8%. Because the data analysis of this study focused primarily on the data for 2010 ACCION borrowers, the researcher also investigated the statistics of this data specifically. In 2010, ACCION Chicago provided loans for 165 borrowers. These loans had an average interest rate of 14.9%, had an average total amount financed of $7, 071, and lasted a little less than two years. The average default rate in 2010 was 8.9%. Despite the slightly lower default rate in 2010, these data statistic are relatively similar to those that describe the data for 2008 and The 2010 default rate might have been slightly lower because some of the 2010 ACCION loans were refinanced to extend their due date

44 44 and some of them lasted longer than two years. Therefore, a few of the 2010 loans may be still been in the process of repayment when the final default rates were calculated. When these default rates were calculated, loans that were in the loan repayment process would not be officially defaulted, even if they were late on loan repayments. Overall, however, the 2010 ACCION data closely resembles data from the earlier ACCION datasets. (For more detailed background statistics for ACCION Chicago borrowers and ACCION Chicago loans, see Appendix A and Appendix B.) ACCION and Reference USA data comparison The researcher quantitatively compared the ACCION and Reference USA datasets to understand the difference between ACCION Chicago businesses and other small businesses in Chicago. Although the Reference USA dataset provided a slightly biased representation of young businesses earning less than $1 million in the Chicago area, this comparison was the best available base comparison for the study. According to this quantitative comparison, businesses in the Reference USA database were slightly older than ACCION businesses on average. The average age of businesses in the Reference USA database was seventeen years, but the average age of ACCION business from 2008 to 2011 was only 6 years. This discrepancy in business age seemed to reinforce the assumption that Reference USA was unable to collection information about some of the newer Chicago businesses during its data collection process. In addition to being slightly older, the Reference USA businesses also have a higher

45 45 average credit rating. The average credit score for ACCION businesses from 2008 to 2011 was 590, and the average credit rating for the Reference USA business was between a B+ and a B rating. According to BCSalliance.com, an online blog that provides financial information for consumers, this B+/B credit rating typically corresponds to credit scores ranging from 620 to 670 (Boyd). Assuming this conversion, businesses in the Reference USA dataset have an average credit score that is higher than that of ACCION businesses. Because ACCION businesses are subprime borrowers by nature, however, this disparity in credit scores is not surprising. Despite the difference between average business age and average credit scores of the two datasets, the datasets had relatively similar business sector representation. According to ACCION s business sector categorization, the most common industries among ACCION borrowers are the food and beverage industry (14.63% of total borrowers), the professional services industry (14.23%), and the retail sales industry (10.22%). In this categorization, professional services included financial advisers, attorneys, architects, tax advisers, accountants, and other similar professions. Retail sales included business that sold clothing, jewelry, beauty products, gifts, and other similar retail goods. Because these industries have relatively low barriers to entry, it is not surprising that a significant percentage of ACCION borrowers fall into these three categories. Similar to the population of ACCION borrowers, the sample of businesses in the Reference USA dataset included a significant percentage of businesses in the retail

46 46 industry (24% of the total dataset) and in the professional services industry (14%). The food and beverage industry was not as strongly represented in the Reference USA dataset as it was in the ACCION data, however. Overall, there was not a large discrepancy between the industries represented in the ACCION data and those represented in the Reference USA dataset. Like the representation of business sectors, the geographical representation in the two datasets was also similar. 71.7% of the ACCION businesses and 60.30% of the Reference USA sample were located in Cook County. Of the ACCION business and of the Reference USA businesses within Cook County, many were located in the loop or in the neighborhoods surrounding the loop. According to these statistics, the two datasets had very similar geographical representation. (See Appendix C for more specific statistics on geographical representation) If the information had been available, it would have been helpful to make a more detailed comparison between the average annual revenues of the two datasets. Because Reference USA recorded the annual revenues of businesses in ranges, its data provided little detail about the variation in annual revenues among its sample of businesses. The Reference USA businesses were categorized as either less than $500,000 in annual revenue or between $500,000 and $1 million in annual revenue. Therefore, the comparison of the average annual revenues of each dataset was inconclusive.

47 47 Unfortunately, there was also limited information about the gender and the ethnic identification of businesses owners in the Reference USA sample. If Reference USA had collected this information with their data collection, it would have been interesting to compare the gender representation and the ethnic representation of the ACCION data to the Reference USA sample. Overall, the ACCION dataset and the Reference USA sample represented similar geographical locations, business sectors, and ranges of annual business revenues. However, the representation of younger businesses and of business with lower credit scores in the ACCION dataset clearly distinguishes it from the Reference USA sample. (See Appendix C for ACCION and Reference USA statistics comparison) Analytical reasoning To investigate evidence of asymmetric information among ACCION borrowers in 2010, this study analyzed the relationship between the probability of default and the interest rates among borrowers. In its analysis, this study used a spectrum of success and failure on loan repayments. The loan repayment process could result in one of four outcomes: (1) a defaulted loan, (2) a repaid loan with late repayments, (3) a refinanced loan, or (4) a repaid loan with little to no late repayments. To successfully capture the relationship between interest rate and default probability, the researcher needed to designate a measurement of loan default that provided sufficient variation in the data. In the dataset of merged 2010 ACCION and Reference USA observations, there were

48 48 only three borrowers who officially defaulted on their loans. Because there were so few borrowers who officially defaulted, the default variable provided little information about the relationship between interest rate and the probability of default. Fortunately, the ACCION repayment data included a weekly measurement of the days past due for each loan repayment. This days past due variable measured the number of days that a borrower was late on his or her repayment in the given week. Using the maximum number of days past due for each loan, the researcher compiled this information to measure borrower success on a continuous spectrum of delayed repayments. As expected, the loans with higher maximum days past due were more likely to become defaulted. This maximum days past due variable captured three of the four loan repayment outcomes along a continuous spectrum: (1) a defaulted loan, (2) a repaid loan with late repayments, and (3) a repaid loan little to no late repayments. In addition to using the maximum days past due measurement, the researcher also tested a default categorization variable that divided borrowers into three categories: (1) loans that were not refinanced and had a maximum days past due more than 14 days, (2) loans that were refinanced, and (3) loans that were not refinanced, but had a maximum days past due less than 14 days. These three categories included the possibility of refinancing in the overall default spectrum. The researcher defined this default categorization as an alternate approximation of default because it incorporated refinancing as an indication of repayment success. According to ACCION Chicago s credit policy, refinancing only occurs if the borrower is reliable with loan repayments

49 49 and demonstrates a significant need for further financing to accomplish his or her business project. This study assumes that borrowers who needed and qualified for refinancing were less likely to default than the borrowers who needed but did not qualify for refinancing. It also assumes that borrowers who had no trouble repaying their microfinance loan and who did not need refinancing from ACCION were probably more financially secure than borrowers who could qualify for a refinanced loan, but needed more money from ACCION to successfully complete their business project. Therefore borrowers who did not receive a refinanced loan were divided into two other categories: the borrowers who did not qualified for a refinanced loan and the borrowers who could have qualified for a refinanced loan but did not need it. These categories corresponded with categories (1) and (3) respectively. Because there was no existing data that divided these two remaining categories, the researcher arbitrated a division between the two. Non-refinance borrowers who had a maximum days past due larger than or equal to fourteen days were listed in the unqualified for refinancing category. Nonrefinance borrowers with a maximum days past due less than fourteen days and no record of default were listed in the qualified, but no need for refinancing category. This study assumed that refinanced borrowers were less likely to default than the unqualified for refinancing category, but more likely to default than the qualified, but no need for refinancing category. Obviously this default categorization did not provide the perfect measurement of default rates; however, it incorporated all the available data

50 50 about refinancing to help predict the probability of borrower default and of delayed repayments. The researcher used these two measures of default to test baseline regressions of default. These baseline regressions investigated the relationship between interest rate (i ) and rates of defaulted or delayed repayments (D) while controlling for loan length (length), credit score of business owner (credit), business sector (sect), ethnicity of business owner (ethn), gender of business owner (sex), and geographical location of business (geo) (See model 1). The first of these two regressions investigated the correlation between interest rate and the maximum number of days past due (maxdayspd) for loan repayments (model 1a), and the second regression captured the correlation between interest rate and the default categorization variable (default_cat) (model 1b). Note that the default category regression required an ordered probit model because the dependent variable (default_cat) was a discrete value. This probit model measured the likelihood of a loan s placement along the default categorization given the values of the explanatory variables. (1) D i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + u (1a) maxdayspd i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + u (1b) default_cat i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + u These two models tested for a positive and significant correlation between interest rate and the probability of delayed (or defaulted) repayments. The researcher expected to see

51 51 a positive correlation between the loan interest rates and the probability of delayed or defaulted repayments for three primary reasons. First, ACCION may have been able to anticipate the riskiness of its borrowers, and therefore successfully adjusted its interest rates to account for this riskiness. If ACCION expected a borrower to more risky or to be more likely to default, raising the loan s interest rate would financially compensate ACCION s risky investment in the borrower s business. According to ACCION Chicago s Tier Lending Price Model in the ACCION Chicago Credit Policy Manual, ACCION tends to assign higher interest rates to borrowers with lower credit scores. This interest rate adjustment is also designed to encourage the borrower to repay the loan immediately in order to avoid compounding interest. If ACCION accurately predicted the riskiness of its borrowers, this positive correlation between the interest rate and the default rate would reflect the success of ACCION s borrower evaluation process and interest rate selection process. Besides ACCION s successful evaluation process, this positive correlation could also indicate the two different results of asymmetric information: adverse selection and moral hazard. If adverse selection contributed to this positive correlation, ACCION may have been unable to distinguish between risky and non-risky borrowers. For example, risky borrowers might have been able to convince ACCION that they were reliable and were financially stable, when, in fact, they had no intention or no ability to repay their loan in full. One would hope that ACCION s good intentions as an organization would

52 52 be met with trustworthy clients. However, ACCION cannot make this assumption about all clients, especially in a pool of subprime borrowers. A borrower who is financially insecure may be easily overwhelmed by a high interest rate and may resort to defaulting rather than repaying the loan. In this case, the adverse selection in ACCION s loan approval process would contribute to the positive correlations between interest rate and the probability of default. Besides evidence of adverse selection, this positive correlation between interest rate and the probability of default could also be the result of moral hazard among ACCION borrowers. In the presence of moral hazard, ACCION borrowers may put less effort into their business project after receiving their ACCION loan because they know that ACCION will soften the financial blow of a failed project. When interest rates are particularly high on loans and the cost of defaulting is less than the cost of repaying the loan, the incentive to put less effort into a business project is logically higher. Therefore, when borrowers view interest rates as particularly high and have little incentive to make their business project succeed, the positive correlation between interest rate and default rate could be evidence of moral hazard. The challenge of this study was to determine which of these three reasons could explain the positive correlation between interest rate and default rate among ACCION borrowers. To address this question, the empirical analysis of this study focused on three main variables: the ratio of the loan s approved amount to the loan s requested amount

53 53 (referred to as the amount ratio), an outside measurement of insurance expenses, and an outside measurement of advertising expenses. These variables were selected to measure the impact of ACCION s evaluation process, of adverse selection, and of moral hazard respectively. For testing the impact of ACCION s borrower evaluation process, the researcher used the amount ratio, which attempted to convert ACCION s qualitative evaluation of borrowers into a quantitative measurement. This variable was the ratio of the approved loan amount to the requested loan amount. According to Brett Simmons, most borrower applicants know that they will not get the exact amount of money that they request. Because applicants expected to receive less money than they request, they often request more than necessary. This amount ratio captured the discrepancy between the requested loan amount and the approved loan amount during the loan application process. Based on the assumption that more trustworthy borrowers received a loan amount that was closer to their amount requested, this amount ratio was a measure of ACCION s trust in borrowers. This study therefore assumes that this amount ratio measures ACCION s qualitative evaluation of the potential borrower. Because the amount approved should be closer to the amount requested when ACCION trusts the borrower, this ratio is positively correlated with ACCION s trust in its borrowers. The amount ratio measured this trust while removing any variation in the amount approved that was due to project costs or due to the business sector.

54 54 Starting with the base regression, the researcher added this amount ratio as an additional explanatory variable to control for ACCION s qualitative evaluation of borrowers. To measure the impact of controlling for the amount ratio, the researcher checked for an omitted variable bias in the original regression by testing for a significant difference between the interest rate coefficients of the two regressions. In this case, the base regression was the estimated model with an omitted variable (see model 1) and the regression that controlled for the amount ratio was the complete regression. (1) D i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + u (2) D i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + β 8 amountratio i + u According to the omitted variable bias, the bias on the interest rate coefficient in the base regression should be equal to the coefficient on the amount ratio in the second regression times the correlation between the amount ratio and the interest rate. Assuming that ACCION had an effective process for evaluating borrowers, there would be a negative correlation between the amount ratio and the default rate. Likewise, there should be a negative correlation between the amount ratio and the interest rate because ACCION is likely to give lower interest rates to borrowers who seem trustworthy and low-risk. Therefore, an omitted variable bias on the interest rate coefficient in the base regression would be positive. If the amount ratio were a relevant variable for the regression, there would be a significant decrease in the interest rate coefficient after the researcher controlled for the amount ratio in the regression. This significant decrease

55 55 would suggest that ACCION s qualitative evaluation of borrowers contributes to the positive correlation between interest rate and the probability of default. Testing for a significant difference between the original interest rate coefficient and the new interest rate coefficient indicated whether or not ACCION s qualitative evaluation significantly contributed to this positive correlation. To test for evidence of adverse selection, this study used Reference USA s measurement of insurances expenses as an approximate measurement of unknown borrower riskiness. According to ACCION Chicago s Credit Policy Manual, ACCION Chicago did not require these borrowers to provide information about their business expenses on insurance. Therefore, ACCION was unaware of this measurement of borrower riskiness during the loan selection process. This study assumed that insurance expenses were an approximate measurement of borrower riskiness because business owners who are risk-adverse are likely to spend more on insurance to protect their business. Because different industries require different levels of insurance expenditure, however, this study looks at insurance expenditure within industries that have similar insurance requirements. For example, a business in the trucking industry requires significantly more insurance expenditure than a business in cosmetic sales. Therefore, comparing the insurance expenditure of a trucking business and a cosmetics retailer could lead to biased estimates. By comparing insurances expenses within similar

56 56 industries, this study attempted to accurately capture the variation in insurance expenditure that depends on borrower riskiness specifically. Using the omitted variable bias measurement, the researcher also investigated the impact of this insurance expenditure variable on the relationship between the default rate and the interest rate of borrowers in In order to capture the impact of insurance expenses on this correlation, the researcher compared the original base regression (1) to a similar model that included the insurance expenses as an explanatory variable (3). By controlling for the business insurance expenses, the second regression controlled for the unknown information about borrowers riskiness. (1) D i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + u (3) D i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + β 8 insurance i + u After adding the insurance expenses variable to the original model, the researcher compared the coefficients on the interest rate variable in the each model (here, β 1 for regression (1) and β 1 for regression (3)) and tested for a significant difference between the two coefficients. According to the omitted variable bias, there should be a positive bias on the interest rate coefficient in the base regression if this insurance variable were relevant. Based on the assumption that borrowers with higher insurance expenditure would be less risky and would be less likely to default, the correlation between insurance expenditure and default rate would be negative. Likewise, the correlation between insurance expenditure and interest rate would also be negative.

57 57 Because the omitted variable bias is the product of these two correlations, there would be an overall positive bias on the interest rate coefficient. Therefore, if insurance expenditure were a relevant variable for the base regression, there would be significant decrease in the interest rate coefficient from regression (1) to regression (3). If such a significant decrease existed, it would provide evidence of adverse selection in the ACCION loan approval process. On the other hand, if there were a significant increase in the interest rate, or if this change were insignificant, adverse selection would not necessarily contribute to the positive correlation between interest rate and default rate. To test for evidence of moral hazard among ACCION borrowers, this study used Reference USA s measurement of advertising expenses to estimate the hidden effort of borrowers. Moral hazard occurs when borrowers put less effort into the success of their business because they know that ACCION will cover the financial loses of a failed business investment. The magnitude of a borrower s advertising expenses is an approximate measurement of the borrower s commitment to his or her business. This study assumed that a borrower who is highly committed to the success of his or her business is likely to put more effort and more money into advertising. Of course, advertising expenditure is not a perfect measurement of the borrower s effort or the borrower s commitment to his or her business. However, given the available data, advertising expenses is the best approximation of borrower effort that ACCION did not otherwise capture in its borrower evaluation process. Like insurance expenditure,

58 58 advertising expenditure likely varies across industries; therefore, this study compares the advertising expenses of businesses within similar business sectors. To understand the impact of hidden borrower effort on the positive correlation between interest rate and default rate, the researcher controlled for advertising expenses in the base regression to create model (4). (1) D i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + u (4) D i = β 0 + β 1 i i + β 2 length i + β 3 credit i + β 4 ethn i + β 5 sect i + β 6 sex + β 7 geo i + β 8 advertising i + u By adding advertising expenses, the researcher attempted to control for the borrower s hidden effort during the loan repayment process. This study assumed that borrowers who spent more money on advertising were less likely to default because they put more effort into their business. According to omitted variable bias, there would be a positive bias on the interest rate coefficient in the base regression if this advertising variable were relevant to the default regression. Assuming that borrowers who put more effort into their businesses are less likely to default, there would be a negative correlation between advertising expenses and the probability of default. In the case of moral hazard, there would also be a negative correlation between interest rate and advertising expenditure because uncommitted borrowers have less incentive to put effort in their work if they face higher interest rates. Therefore, the resulting omitted variable bias on the interest rate coefficient would be positive. Similar to the previous two comparisons, if

59 59 there were a significant decrease in the interest rate coefficient after the researcher controlled for advertising expenses, moral hazard likely explained part of this positive correlation. If there were no significant difference between the two coefficients, advertising may have been an insufficient measurement of borrower effort or there may have been little moral hazard among ACCION borrowers in This study used the loan amount ratio, insurance expenses, and advertising expenses to test for evidence ACCION s effective borrower evaluation process, of adverse selection, and of moral hazard respectively. By controlling for these variables while testing for a significant change in the correlation between interest rate and the probability of default, the researcher investigated the causation of a positive correlation in the study s base regression. V. Hypothesis Based on existing information about ACCION s loan approval process, ACCION s effective evaluation of borrowers riskiness seems like the most likely explanation for the positive correlation between interest rate and the probability of default. By designing a loan approval process that includes an application, an interview, and a business site visit, ACCION considers not only borrowers credit scores but also a qualitative measure of borrowers reliability and commitment to financial success. It seems likely that ACCION s unique approach to evaluating borrowers enabled them to look beyond a credit score and gauge borrower s probability of default through conversation. Because

60 60 credit scores are an imperfect measurement of borrower riskiness and dedication, ACCION seems to have a more accurate method for evaluating borrowers probability of default. Assuming that ACCION can accurately predict borrowers probability of default, ACCION should also be able to assign appropriate interest rates to its borrowers. Therefore, if ACCION s evaluation of borrower riskiness were effective, a positive correlation between interest rate and probability of default is highly likely. Because ACCION used both a quantitative and a qualitative process to evaluate its borrowers, one would also expect little evidence of adverse selection. ACCION s evaluation of borrowers is likely to capture more information about borrower riskiness and to prevent a large discrepancy between borrower and lender information. Hopefully, if ACCION were able to accurately anticipate the riskiness of borrowers, there would be limited asymmetric information between ACCION and their clients. If this were true, adverse selection probably did not contribute to the positive correlation between interest rate and the probability of default. The impact of moral hazard on this positive correlation is slightly more difficult to predict. Even if ACCION can successfully evaluate the riskiness of its borrowers, it is very difficult for ACCION to control the actions of borrowers after their loans are approved. However, because ACCION considers borrowers commitment to their business and financial success, it seems that ACCION makes every attempt to predict the hidden effort of borrowers before their loans are approved. If ACCION s prediction

61 61 of borrower effort is accurate, then the effect of moral hazard on the positive correlation between interest rate and default rate would be minimal. If ACCION s unique process for evaluating borrowers before loan approval were indeed successful, the effectiveness of this process, rather than adverse selection or moral hazard, is likely the cause for a positive correlation between interest rate and the probability of default or of late repayment. VI. Results Unfortunately, the quantitative results of this study were largely inconclusive. Both base regressions had insignificant estimates for the relationship between interest rate and the probability of default. Because these estimates were insignificant, they provided an insufficient foundation for further quantitative analysis. These indecisive results are likely due to the limited available data for the study. The OLS regression of the maximum days past due model estimated a positive, but insignificant correlation between interest rate and delayed (or defaulted) repayments (See appendix D for complete regression statistics). It was reassuring to see a positive correlation between these two variables; however, these OLS estimates were not significant enough to provide a conclusion about the correlation between interest rate and the maximum days past due variable. The insignificant results of this 2010 regression are probably due to the small sample size of the dataset. If the researcher had

62 62 access to a larger sample size with sufficient variation, this OLS estimate may provide more significant results. The second base regression which used the default categorization as its measure of default also failed to provide any significant results. In fact, the coefficient on the interest rate was quite small and negative (See Appendix D for complete regression statistics). This coefficient could have been small and insignificant for several reasons. First, the default categorization may not have been the most accurate measure of the probability of borrower default. Although borrowers with refinanced loans were committed to their loan repayments, the category of non-refinanced borrowers was very vague. It was difficult to distinguish which of these borrowers were likely to default and which were not. Although the maximum days past due did not account for loan refinancing, it was probably a better prediction of default. Second, the default categorization provided significantly less variation than the maximum days past due variable because it was a discrete variable. Due to less variation in the data, it was more difficult to capture the relationship between interest rate and the probability of default. A larger sample size might have provided more significant results for this regression; however, using the maximum days past due measurement of default would probably provided more promising results. After running the two base regressions, the researcher controlled for each of the three variables (amount ratio, insurance expenditure, and advertising expenditure)

63 63 separately in each base regression. These six resulting regressions were then compared to their respective base regressions. These three variables were not expected to have a significant impact on the base regressions because the base regressions produced insignificant results. Despite the insignificant results in the base regressions, however, the researcher tested these additional regressions to search for a relevant model for the data. (See Appendix E for detailed regression statistics) Unfortunately, the results of these regressions were also insignificant. Controlling for the amount ratio caused an insignificant change in the interest rate coefficient for both the maximum days past due estimation and for the default categorization estimation. Surprisingly, controlling for the amount ratio in both base regressions created an increase rather than a decrease in the interest rate coefficient. However, because these increases were insignificant, the researcher could not make any certain conclusions from the data. Controlling for insurance expenses and controlling for advertising expenses in both base regressions also produced insignificant results. Of the eight regressions, the only significant coefficient was the loan length coefficient. This loan length coefficient was significant for six of the eight estimated models; however it was always much smaller than interest rate coefficient and did not provide sufficient information about the relationship between loan length and the probability of default.

64 64 Overall, these regressions did not provide significant results and were an insufficient description of the correlation between interest rate and the probability of default in the 2010 ACCION and Reference USA merged data. Consequentially, the researcher was unable to test the causation of a positive relationship between interest rate and the probability of default in the available data. (For the detailed statistics of these regressions and for a graphical representation of these regressions, see Appendix E and Appendix F respectively.) VII. Conclusions The results of this study were inconclusive. They did not provide sufficient evidence of adverse selection or of moral hazard and failed to evaluate ACCION s borrower evaluation process as a predictor of borrower default. The uncertain results of this study are most likely due to insufficient data and to weak data measurements. Although this study was originally intended to evaluate ACCION Chicago s loan approval process, it uncovered a more important policy issue: the necessity of consistent and strategic data collection among microfinance organizations in the United States. Without sufficient data and reliable variables, the researcher could not conclude significant results and could not apply these results to general policies in the field of microeconomics in the United States. Because microfinance in the United States is a relatively recent approach to American economic development, it is not surprising that existing data records provide a limited amount of information. By continuing to

65 65 improve data collection methods, however, ACCION Chicago can provide invaluable information for the field of US microfinance. Repeating this analysis with more detailed data could produce more significant results and may provide sufficient evidence for either promoting ACCION s lending process or for improving its existing lending model. To improve future data collection processes, it is important for microfinance organizations to understand the needs of academic researchers and for academic researchers to be realistic about organizations resources. VIII. Extensions ACCION has made many attempts to improve its data collection during the past several years. For example, ACCION Chicago has considered using The Aspen Institute to help them survey previous clients and to develop more detailed records of current clients. It has also kept more detailed records of client applications and has investigated several methods of empirical analysis for these developing data records. In order to accurately measure asymmetric information in their loan approval process and to accurately measure the impact of their microfinance loans, however, ACCION Chicago will need to make some specific modifications to its record keeping protocol. ACCION should pursue on five main goals to help them improve their data collection process: (1) to create a more systematic approach to recording the defaults and repayments of clients, (2) to quantify the qualitative side of the loan application process, (3) to develop

66 66 a follow-up method to measure future growth and success of current clients, (4) to record information about borrowers who express interest in a loan, but do not actually choose to take a loan from ACCION, and (5) to develop a record of any borrowers who apply for a loan, but whom are turned down. These five modifications could strengthen the content of ACCION's database and improve the significance of future research results. By maintaining a more systematic record of default rates, ACCION will be able to easily measure the repayment records of their borrowers and determine which components of the loan application process best predict a successful loan repayment. The current documentation of loan repayments maintains a weekly record of defaulted loans and of loans that are currently in the repayment process. This record does not include a confirmation of which loans were repaid in full. Therefore, it would be very helpful for ACCION and for future researchers if ACCION maintained a record of repaid and defaulted loans. This would be easy to do if, every time a loan were repaid in full and every time a loan were determined delinquent, ACCION relocated the client in their database and recorded their loan as "repaid" or "defaulted." This could become a regular protocol for ACCION employees who maintain client records. If possible, it would also be helpful to record the borrower s reason for defaulting. For this study, the researcher wrote programs in excel and STATA to compile the existing weekly documentation and to infer default rates. If ACCION developed a more direct way to

67 67 measure default rates, however, these programming steps could be eliminated to help ACCION keep a more accurate record of their default rates. In addition to systematically recording the default rates of borrowers, ACCION could record other measures of loan delinquency in this final report of the borrower. These other measure could be the maximum number of days past due for the client or simply an indication of whether or not the client was ever 30 days late for a loan repayment. By maintaining a record of these other measures of delinquency in addition to a record of loan defaults, ACCION could compile a comprehensive dataset that has sufficient variation for evaluating loan delinquency. In order to gather a more accurate measurement of adverse selection, ACCION should also attempt to quantify the more qualitative aspects of their loan application process. According to the ACCION Chicago Credit Policy Manual, there are several qualitative traits of potential clients that may be used to predict their chances of default. During the loan application process, ACCION reviewers take into consideration the applicant's character, his or her "determination to repay," and his or her "commitment and motivation to enhance [his or her] business" (p. 25). In order to gauge these abstract qualities of applicants, reviewers may consider the applicants' involvement in the surrounding community or religious organization, their education level, their financial literacy, and the sustainability of their business site. Although these observations could have a significant impact on the approval of a loan, these observations are not recorded

68 68 quantitatively in the ACCION database. To improve ACCION's loan approval process, it may be helpful for ACCION to measure these qualitative aspects of the loan application process and to look at the relationship between these observed traits and the clients' default rates. This data would also be useful for understanding any potential difference between ACCION microenterprises and other similar microenterprises in Chicago. In addition to improving their measurement of default rates and their quantitative record of qualitative observations, ACCION should also consider a followup survey for its past clients. The survey could investigate increases in business income, business growth, any resulting employment opportunities, and other potential improvements for the client and the surrounding community. By developing this followup survey, ACCION would be able to measure changes in the microenterprises of its clients. Comparing these changes to changes in other, similar microenterprises that are not ACCION clients would measure the impact of ACCION loans specifically. The follow-up survey process would be essential for any impact measurement. The Aspen Institute would provide an ideal format for this follow-up data collection. Besides following up with existing client records, maintaining records of borrowers who decide not to use ACCION s services would also be helpful for future quantitative analysis. If this is a common occurrence, this data would provide valuable information about borrowers who are discouraged from participating in Chicago s microfinance loan market. If these discouraged borrowers appear less risky that

69 69 ACCION s existing clients, they might have received an interest rate that was too high for their level of risk. If these borrowers choose to use other forms of lending, a record of this would also illustrate whether or not ACCION s loan process meets the current needs of subprime borrowers. This data would be relatively easy to collect if ACCION designed a short, simple survey for each potential borrower who requests information from them, or if ACCION maintained a record of information from applicants who did not end up taking a loan. Data about these discouraged borrowers would be of interest to not only researchers, but also to ACCION. Using this information, ACCION could understand how to modify their lending process to cater to discouraged borrowers. In addition to maintaining records of discouraged borrowers, ACCION should also record information about borrowers who did not qualify for a loan. This data would capture some of the unmet demand for ACCION s microfinance loans in Chicago and would help evaluate ACCION s loan approval process. Knowing why these borrowers did not qualify for a loan would enable ACCION to understand characteristics of borrowers for which they cannot provide loans. Depending on the characteristics of these borrowers and the reasons why they were denied loans, ACCION may want to reevaluate its loan approval process. Measuring this information would be relatively easy if ACCION employees kept the applications of denied borrowers rather than discarding them. This would a simple task that would provide valuable information about the unmet demand of potential microfinance borrowers in the Chicago area.

70 70 Most of these modifications to ACCION s data collection process would involve minimal extra effort, but would significantly improve the quantitative analysis of ACCION s loan approval process and of their positive impact on borrowers in the Chicago area. Improving ACCION s data collection process according to these recommendations would assist not only academic researchers, but ACCION Chicago as well. If ACCION Chicago can maintain records of its default rates and regulate the success of is borrower evaluation process, it can strengthen its financial sustainability while providing better opportunities for microentrepreneurs in the Chicago area. Significant results from ACCION s data records could also provide valuable recommendations for ACCION Chicago and for other microfinance organizations in the United States.

71 71 Appendix A: ACCION background statistics (Jan- Oct) 2011 All 4 years Average gender of borrowers Male 43.9% 58.8% 58.9% 64.0% 57.4% Female 56.1% 41.2% 41.1% 36.0% 42.6% Average credit score Ethnicity of borrowers American Indian or Alaskan Native 0.9% 0.0% 0.5% 0.0% 0.3% Asian 4.4% 9.7% 7.9% 8.0% 7.8% African American 42.1% 40.0% 44.1% 12.0% 42.2% Hispanic or Latino(a) or Spanish Origin 19.3% 12.1% 13.4% 16.0% 14.7% Native Hawaiian or Pacific Islander 0.0% 0.0% 0.5% 0.0% 16.0% Caucasian 33.3% 37.6% 33.7% 34.0% 34.7% Not given 0.0% 0.6% 0.0% 0.0% 0.2% Industrial sector of borrowers Arts/Crafts Production 0.9% 0.0% 2.5% 0.0% 1.0% Auto/Vehicle 0.0% 0.0% 2.5% 4.0% 1.7% Beauty/Salon Services 0.9% 3.0% 6.9% 4.0% 4.1% Cleaning Services 0.0% 0.0% 4.5% 6.0% 2.9% Clothing/Material Production 0.0% 1.2% 2.0% 0.7% 1.1% Construction/Repair 2.6% 1.2% 4.0% 3.3% 3.7% Cosmetics Sales 0.0% 0.0% 0.5% 0.0% 0.2% Day Care 1.8% 0.6% 3.0% 3.3% 2.2% Education 0.0% 0.0% 1.5% 0.0% 0.5% Electronics 0.0% 0.6% 0.5% 0.0% 0.3% Entertainment 0.9% 0.0% 3.0% 3.3% 1.9% Farming 0.0% 0.0% 0.0% 0.7% 0.2% Food/Beverage 1.8% 1.8% 14.9% 14.7% 9.0% Manufacturing 0.0% 0.0% 0.5% 1.3% 0.5% Media 0.0% 0.0% 0.5% 0.7% 0.3% Medical Services 0.0% 0.0% 3.5% 3.3% 1.9% Professional/Off Svc 0.0% 0.6% 17.3% 17.3% 9.8% Retail Stores 2.6% 1.2% 11.9% 12.0% 7.5% Telecommunications 0.0% 0.0% 2.0% 0.0% 0.6% Transportation 1.8% 0.6% 6.9% 6.7% 4.3% Visual Arts 0.0% 0.0% 0.5% 0.0% 0.3% Wholesale 0.9% 0.0% 3.0% 2.7% 1.7% Other 2.6% 2.4% 8.4% 12.7% 6.8% Not given 83.3% 86.7% 0.0% 0.0% 37.72%

72 72

73 73 Appendix B: Background statistics for ACCION loans and loan repayments (Jan Oct) Average Interest rate 13.6% 14.7% 14.9% 9.0% 14.7% Average total amount financed $9, $7, $7, $6, $7, Average length of loan repayment (months) Average default rate 28.1% 15.8% 8.9% 1.3%* 16.8% Total number of borrowers *This number is a biased measurement of default because the majority of 2011 borrowers did not complete their loan repayment process Appendix C: Table and graphical representation of ACCION and Reference USA data comparison 24 months ACCION ( ) Reference USA Total number of businesses Max sales volume $1 million $1 million Average age of business 6 years 17 years Three most common business sectors Sector Food/Beverage Retail sales Percentage of total businesses 14.63% 23.51% Sector Professional services Professional services Percentage of total businesses 14.23% 14.27% Sector Retail sales Manufacturing Percentage of total businesses 10.22% 8.05% Percentage of total businesses in: Cook County 71.70% 60.30% Loop 2.80% 4.00% West loop 1.20% 1.90% Oak Park 0.80% 1.15%

74 74 Near North Side 2.00% 2.00% Rogers Park 1.40% 1.10% Evanston 1.40% 1.96% Near South Side 1.40% 0.51% South Chicago 1.40% 0.80% Average number of employees 3 4 Maximum number of employees Average credit score *

75 75