Final report. Mavis Amponsah Innovations for Poverty Action House number C149/14 2nd, Dzorwulu Crescent, Dzorwulu, Accra

Transcription

1 Final report Universite AN Laval RCT ON AN INNOVATIVE LOAN PRODUCT FOR FEMALE ENTREPRENEURS IN GHANA Mavis Amponsah Innovations for Poverty Action House number C149/14 2nd, Dzorwulu Crescent, Dzorwulu, Accra 0

2 Summary Few developing country firms manage to grow and scale-up, especially those run by female entrepreneurs. Lack of financial capital has often been cited as a reason, but studies have shown that, even when given access to appropriate capital, entrepreneurs often fail to apply it to productive firm investments. This research examines whether marketing an unlocked cash loan product or a locked asset loan product (capital investment is dedicated ex ante to a specific firm asset) leads to better productivity and growth outcomes for female micro enterprises. We launched a randomized-controlled trial with 2,000 micro entrepreneurs in Accra, Ghana to evaluate the impact of this intervention on credit access and product adoption, as well as differences for female versus male entrepreneurs. 1

3 JEL: C93, G21, O16 Please consult the following link: Keywords: Saving Investment, Entrepreneurs, Diversion, Loan, Finance, Microfinance, Developing Country Please consult the following link : and look at the second level of classifications by clicking on the JEL code links; there, you will find examples of keywords associated to specific domains Author Mavis Amponsah R.A, Innovations for Poverty Action Accra, Ghana mamponsah@poverty-action.org Acknowledgements This research work was carried out with financial and scientific support from the Partnership for Economic Policy (PEP) ( with funding from the Department for International Development (DFID) of the United Kingdom (or UK Aid), and the Government of Canada through the International Development Research Center (IDRC). 2

4 FINAL REPORT Contents Executive summary Introduction Context of the study Research Questions and Objectives Literature review Design and Methodology Study area Data collection and randomization Intervention Features of the unlocked loan product (Green group) Features of the locked loan product (Brown group) Marketing the loan products Controlling contamination and spillover effects Loan requirements Requirements for Susu Loan: Requirements for Micro Enterprise Loan: Requirement for SME loans Analysis and Results TABLE 1: COMPARISON OF EXPERIMENTAL GROUPS RANDOMLY ASSIGNED OFFER OF A LOCKED/ASSET LOAN OR AN UNLOCKED/CASH LOAN TABLE 2: COMPARISON OF EXPERIMENTAL GROUPS BY GENDER TABLE 3: RESULTS OF PRODUCT OFFERS ON ADOPTION CONCLUSION REFERENCES

5 Executive summary Of the two billion people living on less than $2 per day, roughly half run a business (Collins et al. 2009). As such, scholars agree that substantial economic growth in developing countries can occur if significant numbers of micro-entrepreneurs scale their businesses and transition into small or medium enterprises (SMEs). Yet, the reality is that few developing country firms manage to grow and scale-up (Ardagna and Lusardi 2008; Schoar 2010). We ask if this missing middle problem is more pronounced for female entrepreneurs and, if so, how can it be overcome? Given female entrepreneurs may experience limited credit access due to discrimination or reduced risk appetites due to household constraints, financial markets may not offer products that increase access to capital or ensure it is invested optimally in productive firm assets. Thus, we design a novel assetlocked loan product that aims to address financial capital constraints to female micro enterprise growth from both the supply-side and the demand-side. Specifically, this research examines whether marketing a traditional unlocked loan product (no restrictions on how capital from loan proceeds is invested) or our new locked loan product (capital investment is dedicated ex ante to a specific firm asset) leads to better productivity and growth outcomes for female-run micro enterprises. We launched a randomized-controlled trial with roughly 2,000 small business owners (mainly microentrepreneurs) who save with Sinapi Aba Savings and Loan (SASL), to evaluate the impact of this intervention on credit access and investment in firm assets, as well as on the female entrepreneur s savings behavior, loan repayment behavior, business performance and household welfare. We will measure these outcomes by conducting a sampling survey, baseline survey, and an endline survey with the entire sample of 2,000 micro-entrepreneurs. Additional data will be collected through follow-up interviews conducted during the intervention phase (i.e. loan application, approval and disbursement) as well as bank administrative data that provides detailed financial information (e.g. weekly savings, deposit and repayment activities). 4

6 The sampling and baseline surveys have now been conducted. The baseline report submitted to PEP in April showed that there are no systematic differences between the treatment and control groups across roughly 30 pre-treatment covariates, including: economic outcomes (e.g. employees, sales, cost and profits), business characteristics (e.g. years in operation, capital at start up, access to credit, registration status, business practices), entrepreneur demographics (e.g. age, gender, number of children, dependents, race/ethnicity), social outcomes (e.g. household satisfaction, household confidence, household education, household food, household health, household big purchases, household improvements) and surveyor evaluations (e.g. owner s aspiration to grow, owners English level, owner s literacy level, owner s numeracy level). We therefore conclude that randomization holds, and the control group appears to represent a valid counterfactual for the treatment group. Further baseline analyses showed that there is a statistically significant difference between female- and male-owned enterprises in our sample across the following 14 covariates: economic outcomes variables (e.g. employees and profits), business characteristics variables (e.g. capital at start up, registration status, some business practices), entrepreneur demographics variables (e.g. gender, number of children, race/ethnicity), social outcomes variables (e.g. household satisfaction, household food, household big purchases) and surveyor evaluations (e.g. owners English level, owner s literacy level, owner s numeracy level). This final report outlines the findings from our study s intervention: i.e., adoption of the two loan products. Specifically, we use intervention data collected to date to analyze demand for credit and product take up. Our results show that there is a significant effect of gender on loan adoption: female business owners are 20.6% less likely to sign up for a loan product (either the locked/asset or unlocked/cash loan product) compared to their male counterparts. The main purpose of this project, however, was to understand if we could develop a credit product that reduces the diversion of loan funds away from one s business. We therefore designed the locked/asset product to address the pressures a business owner might face to divert loan funds and spend money on non-business purposes. We analyzed heterogeneous treatment effects using three different business owner characteristics that tend to drive increased adoption of the locked/asset loan product (compared to the unlocked/cash loan product): (a) more self-awareness of temptation failures leads to a 26.2% increase in adoption; (b) higher consideration of future consequences results in a 29.6% bump in take up; and (c) greater lack of diversion strategies leads to a 30.5% increase in product sign ups. 5

7 The results of this project have important implications for policy makers wishing to stimulate greater access to credit for small businesses in developing economies and hoping to increase the likelihood that loan funds are indeed invested in business purposes. 1. Introduction A microenterprise is considered a small business employing nine people or fewer, and having capital assets worth less than $ (Micro Focus, International Wikipedia). Most micro-entrepreneurs are primarily interested in earning a living to support themselves and their families. Scholars have, however, agreed that substantial economic growth in developing countries can occur if large numbers of micro-entrepreneurs scale their businesses and transition into small or medium enterprises (SMEs). But how can these businesses grow when financial institutions have been cautious about lending to micro entrepreneurs due to the high default rates and risks associated with such ventures (Mensah, 2004)? Women in most cases face this challenge. Thus, there is a need to examine whether women entrepreneurs who take loans from Microfinance Institutions (MFIs) actually invest the entire loan proceeds into their businesses and as a result realize the optimal benefits for which the credit facilities are acquired. This study therefore sought, using micro-entrepreneurs views and responses, to measure the impact of an innovative loan product on the decisions and business performance of micro-entrepreneurs. The findings can motivate business owners to access savings services, overcome temptation and social demands, and build savings habits. Our new asset-locked loan product is explicitly designed to increase the likelihood that loan proceeds are used toward investments in productive assets, and it has the potential to increase profitability and growth. These outcomes might, in turn, stimulate new policies aimed at promoting even greater financial access and investment for micro-entrepreneurs in developing countries. Given the sheer number of micro enterprises across the globe, our intervention has the potential to improve firm productivity and, in turn, the lives of millions of poor business owners. 6

8 1.1 Context of the study Few developing country firms manage to grow and scale-up, especially those run by female entrepreneurs. Many researchers believe this is due to the inability of female micro-entrepreneurs to obtain the capital necessary to make lumpy investments in productive assets. Perhaps most surprisingly, even when given access to appropriate capital they often fail to apply it to productive firm investments. Some factors reside at the individual level, such as fear of debt, lack of self-control, financial illiteracy, or insufficient risk-taking when making investment decisions. Other factors arise at the environmental level. For instance, the income shocks and high uncertainty faced by female microentrepreneurs in developing countries might lead them to utilize capital for needs outside their business, e.g. household items, education fees, or health expenditures. From the point of view of financial institutions marketing loan products to these micro-entrepreneurs, females who invest in productive business assets might have lower odds of default. These customers may also achieve potentially greater profitability than those who apply the capital toward other uses. The prospects of such outcomes might, in turn, encourage greater financial access for female microentrepreneurs in developing countries many of whom are starved for capital. However, overcoming the challenge of ensuring micro-entrepreneurs (female or male) utilize capital for productive business investments has proved elusive so far. Financial institutions typically give the female micro-entrepreneur an unlocked (cash) loan that is not committed (at least not contractually) to the purchase of a specific productive asset, thereby allowing reallocation of the capital to non-business uses. Standard economic theory suggests that an unlocked loan product should be superior because it allows the micro-entrepreneur to determine the optimal use of the money. Such unconstrained use might produce superior outcomes compared to a financial product that forces a female microentrepreneur to use the money for a specified asset. This view, however, assumes that microentrepreneurs are fully rational actors who can (and will) make utility maximizing choices. In reality, given the individual and environmental factors noted above, many female micro-entrepreneurs are unable to commit to using loan proceeds for optimal business investments. Instead they divert some (if not all) of the capital to other purposes. As a result, constraining the investment choices of female 7

9 micro-entrepreneurs by locking the financial product to investments in specific productive assets may lead to better outcomes. 1.2 Research Questions and Objectives In conducting this project, we aim to address the following research questions: Which marketing strategy (promoting locked versus unlocked products) is more effective in encouraging female entrepreneurs to adopt formal financial services in developing countries? In general, do locked or unlocked loan products lead female entrepreneurs to carry out more consistent savings behaviour and to increase their ability to raise substantial lump sums? Will a locked loan product result in better productivity and growth outcomes for micro enterprises owned by females? Which types of female micro-entrepreneurs benefit more from adoption of a locked loan product, relative to an unlocked loan product i.e. do heterogeneous treatment effects exist? What are the social and household-level outcomes of increasing access to lock versus unlock loan products for female run micro enterprises? To address these research questions, we launched a randomized-controlled trial with roughly 2,000 small business owners (mainly micro-entrepreneurs) who are current clients of our partner bank: Sinapi Aba Savings and Loan (SASL). Our study design includes extensive measurement (pre and post intervention) of key economic and social outcomes, as well as business and entrepreneur characteristics. We will conduct a sampling survey, baseline survey, and an endline survey with the entire sample of 2,000 micro-entrepreneurs. Additional data will be collected through follow-up interviews conducted during the intervention phase (i.e. loan application, approval and disbursement) as well as bank administrative data that provides detailed financial information (e.g. weekly savings, deposit and repayment activities). As at the time of this report, the sampling and baseline surveys were completed and the intervention (and related data collection) are in progress. 2. Literature review A small number of developing country firms succeed to grow and scale-up, especially those run by female entrepreneurs. A field experiment conducted by de Mel, McKenzie and Woodruff (2008) which used randomised grants to generate shocks to capital stock to a set of Sri Lankan 8

10 microenterprises found that the average real return to capital in these enterprise is 4.6% - 5.3% per year, substantially higher than market interest rates. They then examine the heterogeneity of treatment effects. Returns are found to vary with entrepreneurial ability and with household wealth, but not vary with measures of risk aversion or uncertainty. Treatment impacts are also significantly larger for enterprises owned by males; indeed, they find no positive return in enterprises owned by females. Many researchers believe this is due to the inability of female micro-entrepreneurs to obtain the capital necessary to make lumpy investments in productive assets (Agier and Szafarz 2012; Banerjee and Duflo 2008; Beck and Demirguc-Kunt 2006, 2008; Klapper and Parker 2010). Perhaps most surprisingly, even when given access to appropriate capital they often fail to apply it to productive firm investments (Banerjee et al. 2010; Karlan and Zinman 2010). Prior research suggests a variety of reasons for this failure. A study conducted by Fafchamps et al. (2011) found that standard models of investment predict that credit-constrained firms should grow rapidly when given additional capital, and that how this capital is provided should not affect decisions to invest in the business or consume the capital. They randomly gave cash and in-kind grants to male- and female-owned microenterprises in urban Ghana. Their results cast disbelief on the ability of capital alone to inspire the growth of female microenterprises. First, while the average treatment effects of the in-kind grants are large and positive for both males and females, the gain in profits is almost zero for women with initial profits below the median, suggesting that capital alone is not enough to grow subsistence enterprises owned by women. Second, for women they strongly reject equality of the cash and in-kind grants; only in-kind grants lead to growth in business profits. The results for men also suggest a lower impact of cash, but differences between cash and in-kind grants are less robust. The difference in the effects of cash and in-kind grants is associated more with a lack of self-control than with external pressure. As a result, it appears that the manner in which funding is provided affects microenterprise growth. Some factors reside at the individual level, such as fear of debt, financial illiteracy, or insufficient risktaking when making investment decisions (Banerjee and Mullanaithan 2010; Bertrand, Mullainathan and Shafir 2004; Karlan, Morduch and Mullainathan 2010; Field, Pande and Papp 2009; Minnit and Naude 2010). Other factors arise at the environmental level. For instance, the income shocks and high 9

11 uncertainty faced by female micro-entrepreneurs in developing countries might lead them to utilize capital for needs outside their business, e.g. household items, education fees, or health expenditures (Banerjee and Duflo 2007, 2011; Collins et al. 2009; de Mel, McKenzie and Woodruff 2009, 2010). From the point of view of a financial institution marketing loan products to these micro-entrepreneurs, females who invest in productive business assets might have lower odds of default. These customers may also achieve potentially greater profitability than those who apply the capital toward other uses. The prospects of such outcomes might, in turn, encourage greater financial access for female microentrepreneurs in developing countries many of whom are starved for capital. However, overcoming the challenge of ensuring micro-entrepreneurs (female or male) utilize capital for productive business investments has proved elusive so far. Financial institutions typically give the female micro-entrepreneur a cash or unlocked loan which is not committed (at least not contractually) to the purchase of a specific productive asset, thereby allowing reallocation of the capital to nonbusiness uses. Standard economic theory suggests that an untied loan product should be superior because it allows the micro-entrepreneur to determine the optimal use of the money. Such unconstrained use might produce superior outcomes compared to a financial product that forces a female micro-entrepreneur to use the money for a specified asset (see de Mel, McKenzie and Woodruff 2008). This view, however, assumes that micro-entrepreneurs are fully rational actors who can (and will) make utility maximizing choices. In reality, given the individual and environmental factors noted above, many female micro-entrepreneurs are unable to commit to using a savings account or loan for optimal business investments. Instead they divert the capital to other purposes (Karlan and Zinman 2010). As a result, constraining the investment choices of female microentrepreneurs by locking the financial product to investments in specific productive assets may lead to better outcomes (see Fafchamps et al. 2011). 10

12 3. Design and Methodology 3.1 Study area The map above shows the study site drawn with gps data collected from the field In selecting financial institution with which to conduct the study, the criteria included a partner that had a minimum of three branches, had the ability to service 1,000 small business loans, and had the willingness to dedicate at least 1,000,000 GHC in loan capital exclusively to our project. Sinapi Aba Savings and Loans (SASL) was one of the only institutes that met our criteria. With over 45 branches across the country, SASL is a large financial institution in Ghana. This project is conducted in three branches in greater Accra (Achimota, Mataheko and Tema) and one in the central region (Kasoa). 11

13 3.2 Data collection and randomization We first of all collected administrative data from the four branches of SASL. The administrative data contained client information from January 2013 to May The administrative data went through a number of screening processes to ensure clients in the sample were micro enterprise owners. After the first phase of the recruitment process, the sampling survey was administered to about 4,000 clients. The purpose of this survey was to collect background information of businesses and their business practices, which also served as a screening tool in selecting our final sample. The sampling and business practice survey took around 60 minutes to complete. The baseline survey was administered to roughly 2,000 business owners. After the baseline survey was conducted, the recruited clients with all responses completed (and who owned a micro or small business) were randomly assigned into 2 groups: the green (i.e. unlocked/cash loan product) and the brown group (i.e. locked/asset loan product). Stratification was done at the branch level using the data from the sampling survey. T-tests and Hotelling tests were used on 20+ variables to ensure balance was achieved in the random assignment step. The variables used in checking the balance covered the following areas: economic outcomes, business characteristics, entrepreneur characteristics, social outcomes and surveyor evaluations. 4. Intervention The intervention is designed to promote a loan product that locks in an entrepreneur's capital investment ex ante to a productive firm asset. The loan product locks in funds from a customer s loan to the purchase of a capital asset that can enhance firm productivity, as opposed to the large majority of other products available in this market that do not lock in the funds to a specific investment. To assess the effectiveness of our proposed solution, however, we introduced both an unlocked version and a locked version of this new loan product (see Appendix for an example of the product brochures). Thus, the intervention consists of one control group (offered the unlocked loan product) and one treatment group (offered the locked loan product). 12

14 4.1 Features of the unlocked loan product (Green group) Essentially, there will be no restrictions on how the loan proceeds are used by the customer. However, before applying for a loan the customer must first meet these criteria: 1. Make regular micro deposits for at least two months, and save 40% of the desired loan amount (to be held as a down-payment); 2. The customer does not have to make any commitments ex ante on how she will spend loan proceeds; 3. Participants who are assigned to the control group will be offered the unlocked loan product; 4. This product will be marketed using a one-page (front and back) brochure and standard sales script (see Appendix). 4.2 Features of the locked loan product (Brown group) Participants in the treatment group receive an otherwise identical intervention except the promotional brochure offers a loan product that locks the funds into one of four productive assets: equipment, vehicle, construction, and inventory (Appendix). This product is marketed using a similar one-page brochure and the same promotional strategy, including comparable field officer scripts and visits. The terms of the locked loan product are identical to those of the unlocked loan product offered to the control group participants. The only difference in the treatment group is that the locked loan product is restricted in how the loan funds can be used. Customers offered this financial product choose ex ante which productive firm asset they would like to obtain a loan for. On the day the loan is disbursed, an Administrative Assistant accompanies the client to purchase the asset, thereby ensuring compliance to the treatment (receipts and photos of the purchase are also obtained to verify the loan funds are indeed invested fully into the business). 4.3 Marketing the loan products Administrative Assistants hired by Innovations for Poverty Actions go out and market the loan products to clients. Brochures and standard sales scripts are used to market the loan products. The brochure describes the terms and benefits of the loan product, and provides the location of the nearest participating branch office and background information of the partner, Sinapi Aba Savings and Loans. 13

15 These administrative assistants follow a standard sales script when promoting and explaining the product to ensure consistent marketing quality. 4.4 Controlling contamination and spillover effects In order to ensure that the loan products are not contaminated and as a result cause spillover effects, systems were put in place to guard against these problems happening. These systems were part of the 3 intervention stages namely: loan offering stage, loan signup stage and loan disbursement stage. Loan offering stage This is the stage where the Administrative Assistants go out and market the loan products to clients. During this loan offering stage, we ensured that the marketing brochures were easily identified to avoid contamination. We did this by giving different colors to the brochures of the two loan products. Clients in the control group received a green brochure, whereas clients in the treatment group received a brown brochure (see Appendix). In addition, to further avoid any mix ups in the field, a name label was attached to every brochure (done at the IPA office in advance), which had the full name of the client and participant ID number printed in either green or brown according to their random assignment. The businesses in our sample are also geographically quite dispersed from one another plus we are only using a portion of the partner s overall client base. Taken together, this makes it less likely that study participants will communicate with one another and as a result we don t foresee contamination happening. Loan sign-up stage This is the stage where clients visit the branch to sign-up for the loan. The clients who visit the branch are taken through a loan orientation interview. Once they are loan ready, the loan application process is initiated by Sinapi Aba loan officers. The loan application and review process is identical for both groups. Final decisions on loan disbursements are made by SASL managers at Headquarters in a different city (Kumasi), so these managers are not aware of the difference between clients in the treatment and control groups. There is therefore no bias or further contamination. 14

16 Loan disbursement stage This is the stage where the client has met all the requirements of the loan and approval has been granted by management of the bank for the loan to be given to the client. Those in the control group are given cash, whereas those in the treatment group are accompanied by an Administrative Assistant to purchase the asset pre-specified by the client. Throughout the 3 stages, great care was taken by the intervention team to ensure contamination and spillovers did not occur. 4.5 Loan requirements Requirements for Susu Loan: This loan is designed to support business owners whose cash flow is very low, especially market women and other traders. To qualify for this type of loan, you have to make a minimum of 2 months (40 days) consistent savings contributions into a Susu account without withdrawal. There s no minimum or maximum contribution amount, the client must contribute an amount that he or she can consistently pay. The amount saved will determine the amount of loan this client qualifies for, as the client s savings constitute the down payment for the loan, and a minimum of 40% of the loan amount is required. For example, a client who contributes GHS 5.00 daily and has saved GHS 200 after 40 days can apply for a GHs500 loan, or a client who contributes GHS daily and has saved GHS 400 after 40 days can apply for GHS 1000 loan. Assessment is done on the client, both at the business and residence. Clients who fall into this category can take loan amounts ranging from GHS 500 to GHS No collateral is needed for this type of loan and the interest rate is 4.2% per month Requirements for Micro Enterprise Loan: This loan is given out to entrepreneurs who hold flexy accounts. For a client to qualify, he or she needs collateral and a guarantor. The collateral can be a landed property, for example. The guarantor is also required to provide 6 months bank statements or pay slips. A processing fee of 3.5% of the loan amount is required as cash collateral. No specific minimum balance is required except that the account should be running consistently for 3 months. The minimum and maximum amount a client can take ranges from GHS 5000 to GHS 19,999, and this is given at an interest rate of 54% p.a. 15

17 4.5.3 Requirement for SME loans This type of loan is for companies registered under liability and clients have been saving with Sinapi Aba for at least 3 months. This type of loan needs a guarantor and a collateral. The requirements for this loan is the same as requirements for Micro Enterprise loan. 5. Analysis and Results Our unit of analysis is the small business owner for which we have roughly 2,000 observations during baseline. Firms in the baseline data have been randomly assign to receive either a locked/asset loan offer (treatment group) or an unlocked/cash loan offer (control group). To ensure confidence in the randomization process, we first checked if the two groups are comparable with respect to individual and firm characteristics that could otherwise account for systematic differences in outcomes (Gelman and Hill 2006, p ). We did this by comparing the locked treatment group to the unlocked group (Table 1), and the male group to the female group (Table 2), on 29 pre-treatment covariates: economic outcomes (e.g. employees, sales, cost and profits), business characteristics (e.g. years in operation, capital at start up, access to credit, registration status, business practices), entrepreneur demographics (e.g. age, gender, number of children, dependents, race/ethnicity), social outcomes (e.g. household satisfaction, household confidence, household education, household food, household health, household big purchases, household improvements) and surveyor evaluations (e.g. owner s aspiration to grow, owners English level, owner s literacy level, owner s numeracy level). One can infer from Table 1 that p>0.10 for the majority of the variables which means there are no systematic differences (except those that otherwise could occur by chance) between the means of the treatment and control groups within these pre-treatment covariates. Based on these results, we can conclude that randomization holds, and the control group appears to represent a valid counterfactual for the treatment group (Angrist and Pischke 2009). On the other hand, Table 2 that focuses on Gender Comparison shows that p<0.10 for 13 variables which are; economic outcomes variables (e.g. employees), business characteristics variables (e.g. 16

18 capital at start up, registration status, some business practices), entrepreneur demographics variables (e.g. gender, number of children, race/ethnicity), social outcomes variables (e.g. household satisfaction, household food, household big purchases) and surveyor evaluations (e.g. owners English level, owner s literacy level, owner s numeracy level). This means there are some differences between the males and the females in our sample when considering these variables. Furthermore, looking at the other variables apart from the ones listed above, we can see that p>0.10 for these remaining variables. Which means that there is no difference between the males and the females in our sample considering these variables. Overall, however, it will be important to control for gender in our subsequent analysis on product adoption. 17

19 6.1 TABLE 1: COMPARISON OF EXPERIMENTAL GROUPS RANDOMLY ASSIGNED OFFER OF A LOCKED/ASSET LOAN OR AN UNLOCKED/CASH LOAN Overall sample Locked loan offer Unlocked loan offer uality of means te N Mean Standard Deviation N Mean Standard Deviation N Mean Standard Deviation t-stat p-value Economic Outcomes Employees (total paid staff, fro recent month) Sales (total in Cedis, for recent month) Cost (total in Cedis, for recent month) Profits summary (total in Cedis, for recent month) Business Characteristics Years in operation (total) Capital at startup (total in Cedis) Accessed credit (formal loan: in prior year) Non registered business (Not formally registered with government or local unit: percentage) Practices: Separates business/ personal finances (percentage) Practices:Keeps any form record of finances (percentage) Practices: Keeps formal records of finances (percentage) Practices: Developed a new product/services (percentage) Practices: Attracted new customers (percentage) Entrepreneur Characteristics Age (total years) Gender (Percentage female) Children (percentage children greater than 2) Dependants (total number) Ethinicity: Ghanaian black or colored (percentage) Social Outcomes Household: Satisfaction [1-5 scale] Household: Confidence [1-5 scale] Household: Education [1-5 scale] Household:Foood [1-5 scale] Household: Health [1-5 scale] Household: Big purchases [1-5 scale] Household:Improvements [1-5 scale] Surveyor Evaluations Enumerator evaluation: Owner's aspiration to grow [1-5 scale] Enumerator evaluation: Owner's English level [1-5 scale] Enumerator evaluation: Owner's literacy level [1-5 scale] Enumerator evaluation: Owner's numeracy level [1-5 scale]

20 6.2 TABLE 2: COMPARISON OF EXPERIMENTAL GROUPS BY GENDER Male Female Equality of means test N Mean Standard Deviation N Mean Standard Deviation t-stat p-value Economic Outcomes Employees (total paid staff, fro recent month) Sales (total in Cedis, for recent month Cost (total in Cedis, for recent month) Profits summary (total in Cedis, for recent month) Business Characteristics Years in operation (total) Capital at startup (total in Cedis) Accessed credit (formal loan: in prior year) Non registered business (Not formally registered with government or local unit: percentage) Practices: Separates business/ personal finances (percentage) Practices:Keeps any form record of finances (percentage) Practices: Keeps formal records of finances (percentage) Practices: Developed a new product/services (percentage) Practices: Attracted new customers (percentage) Entrepreneur Characteristics Age (total years) Children (percentage children greater than 2) Dependants (total number) Ethinicity: Ghanaian black or colored (percentage) Social Outcomes Household: Satisfaction [1-5 scale] Household: Confidence [1-5 scale] Household: Education [1-5 scale] Household:Foood [1-5 scale] Household: Health [1-5 scale] Household: Big purchases [1-5 scale] Household:Improvements [1-5 scale] Surveyor Evaluations Enumerator evaluation: Owner's aspiration to grow [1-5 scale] Enumerator evaluation: Owner's English level [1-5 scale] Enumerator evaluation: Owner's literacy level [1-5 scale] Enumerator evaluation: Owner's numeracy level [1-5 scale]

21 6.3 TABLE 3: RESULTS OF PRODUCT OFFERS ON ADOPTION 19

22 Table 3 outlines our main analysis on product adoption, which is operationalized as the formal take-up of a loan: i.e., completion of an initial application form and attendance at an orientation interview with a loan officer (which requires the client to physically show up to a bank location at the scheduled appointment time). For clients randomly assigned to the treatment group, adoption is coded as 1 if the business owner applies for the locked/asset loan product (as detailed in the brown flyer). This binary outcome variable is coded as 0 if the owner does not adopt the offered product. By contrast, for those in the control group, adoption is coded as 1 if the client applies for the unlocked/cash loan product (as outlined in the green flyer) and 0 otherwise. The overall adoption rate is 20.8% with 428 business owners signing up for the loan product offered to them. As expected, this rate does not differ significantly between the two experimental groups (p=0.59): 21.3% adopted the locked/asset product (n=220 from the treatment group) and 20.4% adopted the unlocked/cash product (n=208 from the control group). To further examine the effects of our intervention, we use probit regressions with the binary product adoption outcome variable outlined above (1 = adopted loan product; 0 = did not adopt). In Column A of Table 3, we see there is no difference in the take up rates between the treatment and control groups. This result is expected since the two interventions were purposefully designed to be as comparable as possible in terms of marketing quality: flyer (e.g., content, design, pictures), sales effort (e.g., content of sales pitch, number of site visits and phone calls), and customer service (e.g., dedicated admin assistant to help schedule loan interviews, explain paperwork, etc). This result holds even after controlling for a range of covariates at the individual level (e.g., gender, married, children, education, race) and the firm-level (full-time employees, part-time employees, business partners, monthly sales, monthly profits). 20

23 Interestingly, the only significant impact on product adoption is the coefficient for gender. We find that female business owners are 20.6% less likely to sign up for a loan product (compared to male owners). This effect appears to be robust across model specifications (refer to Table 3, Columns A-C) with females being reluctant (versus males) to adopt the loan product being offered in either of the experimental groups. Overall, it seems that female business owners tend to be credit averse more so than their male counterparts. While it s important to understand the role of gender differences at play in the micro credit space, the main purpose of this project is to understand whether we can develop a credit product to reduce the diversion of loan funds away from one s business. Thus, we designed the locked/asset product (and randomly offered it to those in the treatment group) in order to address the pressures a business owner might face to divert loan funds and spend money on non-business purposes. To examine the effectiveness of the brown locked/asset product, we analyze heterogeneous treatment effects using three different business owner traits i.e., characteristics of the owner measured pre-intervention that could account for differential adoption rates of the locked/asset loan. First, we examine whether adoption rates differ for business owners who (ex ante) have a high selfawareness that they are susceptible to temptations (i.e. they are more likely to experience self-control failures). To measure this trait or characteristic (at baseline), business owners were asked to rate themselves on a scale of 1 (never) to 5 (always) on four statements that could describe their actions: (i) I spend money on nonessential things; (ii) I waste money by buying things I don t really need; (iii) 21

24 After I receive a business loan, I think that buying something extra for myself is okay; and (iv) When I get additional money in my business, I feel tempted to spend the money for personal use. The scores on all four questions were combined into a composite that measured the average level of temptation self-awareness for each participant. This composite was then converted into a dummy variable that identified participants who were higher (coded as 1 ) versus lower (coded as 0 ) on this particular characteristic. Also, the experimental groups were roughly balanced with 540 participants in the treatment group, and 521 participants in the control group, scoring high on the temptation selfawareness trait. Column B in Table 3 provides the results of the probit regressions that analyze these effects. There is a negative main effect of temptation self-awareness on the adoption of loan products. That is, owners who are more aware that they can be tempted easily (e.g., to spend money on non-business purposes) are 23.7% less likely on average to sign up for the loan product being offered in this project irrespective of which type of loan is being marketed to them. However, there is a positive heterogeneous effect in the case of the treatment group. For business owners who are highly aware of their own temptation failures: being offered the locked/asset loan (i.e. the treatment) leads to a 26.2% increase in product adoption (compared to the unlocked/cash loan). In other words, when a business owner is aware she s more susceptible to self-control failures, then the locked/asset loan is a more attractive product for her. This product allows the owner to impose a commitment on herself and ensure the loan funds are invested in her business (instead of possibly getting spent on personal or nonessential purchases). Second, along similar lines, we also analyze if adoption rates differ for business owners who (pretreatment) are more likely to consider the consequences of their choices (i.e., they think about what 22

25 might happen in the future when making a decision today). To measure this trait or characteristic (at baseline), business owners were asked to rate themselves on a scale of 1 (never) to 5 (always) on three statements that could describe their behaviors: (i) Before taking an action today, I consider the consequences it will have on me in the future; (ii) I am aware that I might waste money in the future, so I try to store some money in a secure place (where I can t easily spend it); and (iii) Before making a choice today, I assess if the outcome will be harmful to me in the future. The scores on all three questions were combined into a composite measure of consequence consideration. This composite was then converted into a dummy variable that identified participants who were higher (coded as 1 ) versus lower (coded as 0 ) on this trait. Importantly, the two groups were roughly balanced with 429 participants in the treatment group, and 392 participants in the control group, scoring high on the consequence consideration characteristic. Column C in Table 3 provides the results of this additional analysis. There is a negative main effect of consequence consideration on the take up of credit. It seems that owners who consider the future consequences of their decisions are 20.1% less likely adopt either the cash loan or the asset loan. They avoid credit all together. But consistent with the findings above on temptation self-awareness (Column B), there exists a positive interaction effect for the treatment group (Column C). Marketing the locked/asset loan (i.e. the treatment) to business owners who consider future consequences results in a 29.6% bump in take up (versus those offered the unlocked/cash loan). Basically, the treatment product reverses the negative adoption trend: the locked/asset loan appears to be a more attractive product to business owners who think about the outcomes that might arise in the future as a result of their present day decisions. This may be because the locked/asset product helps the owner to avoid pressures to divert loan funds away from the business and overcome otherwise detrimental consequences in the future. 23

26 Third, we attempt to understand whether adoption rates vary depending on the diversion management strategies that a business owner has already been implementing (prior to our experimental intervention) to protect against external pressures. Recall that the brown product locks the client into fully investing the loan funds into a business asset details which are outlined in the flyer text, visually displayed in the photos, and explained by the admin assistant during a site visit with oversight and compliance of this asset purchase managed by a monitoring officer. Thus, in many ways, the locked/asset product (i.e., treatment) represents a strategy for managing diversion pressure. To measure the extent to which a business owner lacks diversion management strategies, we asked participants to answer yes (scored 1 point) or no (scored 0 points) if they had implemented each of the following activities during the past three months: (i) Lying to my family or friends when they ask me about how profitable my business was last month. I do not tell them the true amount that I make each month; (ii) Locking away my money so that I cannot access it for one month or longer (e.g. using a fixed deposit account with a bank); (iii) Hiding my money in a secure place at home (e.g. using a physical lock box or safe) so that no one can access the money but me; (iv) Taking any extra money and immediately investing it back in my business (e.g. in machines and equipment) so that I do not spend it on stuff I don t need; and (v) Not bringing money home with me after work, so family members cannot see it. The scores on all five activities were summed up to construct a composite that measured the total number of diversion management strategies recently used by each participant (minimum score = 0; maximum score = 5). The median score was 2 strategies. Next, the composite was converted into a lacks diversion strategies dummy variable to identify participants who implemented 0-1 activities as more lacking or higher on this characteristic (coded as 1 ) and those who implemented 2-5 such activities as less lacking or lower (coded as 0 ). The experimental groups were approximately balanced on this variable: 406 participants in the treatment group and 375 participants in the control group were high on the lacks diversion strategies characteristic. Column D in Table 3 provides probit regression results for these heterogeneous treatment effects. As hypothesized, there is a positive impact of the locked/asset product on adoption when business owners are lacking ways to manage diversion pressure. In other words, for business owners who are not already implementing strategies to minimize pressures of spending money on non-business purposes: being offered the locked/asset loan (i.e. the treatment) leads to a 30.5% increase in product adoption (compared to the unlocked/cash loan). Thus, this locked/asset product appears to represent a useful 24

27 tool that business owners can use to protect themselves from outside pressures and reduce the chance they will divert loan funds away from business investments. 7. CONCLUSION We launched a randomized-controlled trial with roughly 2,000 small business owners (mainly microentrepreneurs) who hold a savings or current account with Sinapi Aba Savings and Loan (SASL), to evaluate the impact of a new loan product on access to credit. To date, we have measured outcomes by conducting a sampling survey and baseline survey, as well as through bank administrative data that provides details on product adoption. The participants were randomly assigned to one of two experimental groups and, in turn, offered either a locked/asset loan (treatment group) or an unlocked/cash loan (control group). Clients in the treatment group received an offer to sign up for a product that locks them into fully investing the loan funds into a business asset details which are outlined in the flyer text, visually displayed in the photos, and explained by the admin assistant during a site visit with oversight and compliance of this asset purchase managed by a monitoring officer. By contrast, those in the control group were offered the bank s traditional cash loan that is unlocked in the sense that it does not place any restrictions on how the loan funds are spent by the client. In general, we find a significant impact of gender on loan adoption: female business owners are 20.6% less likely to sign up for a loan product (compared to male owners). It seems they are more credit averse than their male counterparts. However, the main purpose of this project was to understand if we could develop a credit product that reduces the diversion of loan funds away from one s business. We therefore designed the locked/asset product to address the pressures a business owner might face to divert loan funds and spend money on non-business purposes. We analyzed heterogeneous treatment effects using three different business owner characteristics that tend to drive increased adoption of the locked/asset loan product (compared to the unlocked/cash loan product): (a) more self-awareness of temptation failures leads to a 26.2% increase in adoption; (b) higher consideration of future consequences results in a 29.6% bump in take up; and (c) greater lack of diversion strategies leads to a 30.5% increase in product sign ups. One, we find support for the conclusion that the locked/asset loan is a more attractive product for business owners who need to impose a commitment on themselves and ensure the loan funds are 25