WPS6020 Policy Research Working Paper 6020 Impact Evaluation Series No. 54 Does a Picture Paint a Thousand Words? Evidence from a Microcredit Marketing Experiment Xavier Giné Ghazala Mansuri Mario Picón The World Bank Development Research Group Finance and Private Sector Development Team & Poverty Reduction and Economic Management Network Poverty Reduction and Equity Unit April 2012 Policy Research Working Paper 6020 Abstract Female entrepreneurship is low in many developing had lower measured ability and whose wives were less economies partly because of constraints on women’s time educated, the responses to the female brochure were and mobility, which are often reinforced by social norms. more negative, as did female business owners with low This paper analyzes a marketing experiment designed to autonomy within the household. Women with relatively encourage women to adopt a new microcredit product. high levels of autonomy had a similar negative response A brochure with the same content but two different to the male brochure, while there was no effect on female covers was randomly distributed among male and female business owners with autonomy. Overall, these results borrowing groups. One cover featured five businesses run suggest that women’s response to psychological cues, such by men, while the other showed identical businesses run as positive role models, may be affected by their level of by women. Men and women responded to psychological autonomy at home, and more intensive interventions cues. Among men who were not business owners, may be required for more disadvantaged women. This paper is a product of the Finance and Private Sector Development Team, Development Research Group; and Poverty Reduction and Equity Unit, Poverty Reduction and Economic Management Network. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at xgine@worldbank.org, gmansuri@worldbank.org and mpicon@worldbank.org. The Impact Evaluation Series has been established in recognition of the importance of impact evaluation studies for World Bank operations and for development in general. The series serves as a vehicle for the dissemination of findings of those studies. Papers in this series are part of the Bank’s Policy Research Working Paper Series. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Does a Picture Paint a Thousand Words? ∗ Evidence from a Microcredit Marketing Experiment Xavier Giné, World Bank Ghazala Mansuri, World Bank Mario Picón, World Bank Abstract Female entrepreneurship is low in many developing economies partly because of constraints on women’s time and mobility, which are often reinforced by social norms. This paper analyzes a marketing experiment designed to encourage women to adopt a new microcredit product. A brochure with the same content but two different covers was randomly distributed among male and female borrowing groups. One cover featured five businesses run by men, while the other showed identical businesses run by women. Men and women responded to psychological cues. Among men who were not business owners, had lower measured ability and whose wives were less educated, the responses to the female brochure were more negative, as did female business owners with low autonomy within the household. Women with relatively high levels of autonomy had a similar negative response to the male brochure, while there was no effect on female business owners with autonomy. Overall, these results suggest that women’s response to psychological cues, such as positive role models, may be affected by their level of autonomy at home, and more intensive interventions may be required for more disadvantaged women. Keywords: Microfinance; Credit Constraints; Information; JEL codes: G21, D24, D83, O12. Giné: Development Research Group, The World Bank (e-mail:xgine@worldbank.org). Mansuri: Development Research Group, The World Bank (e-mail: gmansuri@worldbank.org). Picón: Development Research Group, The World Bank (e-mail: mpicon@worldbank.org). We are indebted to three anonymous referees for extremely insightful comments which have substantially improved the paper. We also thank Jonathan Zinman and Tahir Waqar for valuable discussions. We are especially grateful to the following for their help and support in organizing the experiment: Shahnaz Kapadia, at ECI Islamabad, for her help with designing the brochure; Irfan Ahmad at RCONs, Lahore, for managing all field operations and data collection; Dr Rashid Bajwa, Agha Javad, Tahir Waqar and the field staff at NRSP for implementing the intervention; Qazi Azmat Isa, Kevin Crockford and Imtiaz Alvi at the World Bank Office in Islamabad, and Kamran Akbar at the Pakistan Poverty Alleviation Program (PPAF) in Islamabad for their support and encouragement. This project was jointly funded by the World Bank (Development Research Group, South Asia Region, Poverty Reduction and Equity Network, Gender Group), the PPAF and the Kaufmann Foundation. Santhosh Srinivasan provided outstanding research assistance. The views expressed herein are those of the authors and should not be attributed to the World Bank, its executive directors, or the countries they represent. 1. Introduction Women in developing countries face numerous barriers to participation in economic life. In addition to constraints on time and limited access to capital, women’s exposure to markets and business networks is often also limited by low mobility and lack of education. Weak decision-making power within the household often reinforces these disadvantages, further limiting women’s ability to secure time or resources for their own productive activities. One manifestation of this is the comparatively low prevalence of female-owned businesses. Even as entrepreneurs, however, women operate businesses that are far smaller in scale and profitability than male businesses. The recognition that women may therefore be most in need of credit for small businesses has played an important role in the woman-centered microcredit movement of the last two decades (Yunus, 1999). In Pakistan, however, microfinance as well as other support for micro-entrepreneurship has focused primarily on men while women remain at the margins of economic life. This paper reports on an experiment designed to encourage female uptake of a new microcredit product that allowed eligible borrowers the opportunity to borrow up to four times the typical loan size.1 In cooperation with the microfinance institution, we designed two different brochures that provided information about the characteristics of the loan and described the application process. The brochures were identical, except for the cover page: one featured 5 different businesses with men operating them, while the other had the exact same five businesses, but with women entrepreneurs, instead. Groups 1 While supply constraints may also be an important determinant of lack of credit to small firms (Banerjee and Duflo, 2008; de Mel et al. 2010), our sample consists of individuals that are already microfinance clients. 1 of borrowers targeted to be offered the new product were randomized to receive either the brochure with the female or the male pictures. We find that this form of marketing affects both male and female clients but quite differently. In the full sample of clients, the brochure has little impact on loan demand among either men or women. When we focus on business owners, however, we find that exposure to the female brochure substantially decreases demand for the larger loan among women clients but has no impact on male clients. Importantly, this is not an artifact of business scale. While women typically operate much smaller businesses, we show that women who operate businesses which are comparable in scale to male businesses also react negatively to the female brochure. However, once we allow the response to the brochure picture to vary by individual characteristics, we find that this negative effect is concentrated among women business owners with low decision making power. For this group of women, we conjecture that it is the male decision maker’s reaction to the female brochure that matters. We find that men also react negatively to the female brochure, but only among non-entrepreneurs, individuals with poor digit span recall (correlated with education and entrepreneurship) or with wives who are relatively poorly educated. This could be interpreted as evidence of affinity (Evans, 1963; Mobius and Rosenblat, 2006), but it is also consistent with men’s low regard for females as business owners (see Beaman et al. 2009 who study perceptions about female politicians). Interestingly, female business owners with high decision making autonomy shown the male brochure also react negatively by roughly the same magnitude, while there is no effect on female business owners with autonomy shown the female brochure. 2 These findings contribute to the relatively limited evidence on the impact of marketing on behavior change and in this sense, Bertrand et al. (2010) is the paper closest to ours. They analyze a direct mail field experiment in urban South Africa that randomized advertising content, loan price, and loan offer deadlines simultaneously. Their subjects, like ours, were existing microfinance clients. Bertrand et al. find that advertising content matters, especially among men, but it is mostly treatments that appeal to intuition (such as a picture) as opposed to reason (like a comparison of interest rates across lenders) that influence behavior. The reason why reflexive cues are less relevant is perhaps because individuals that received the loan offers had rationally decided not to borrow, and so were already familiar with the terms of the loan made salient by reflexive treatments (Kahneman, 2003). In Bertrand et al (2010) men respond positively to the image of women credit officers in a context where this creates no threat to relative male authority within the household or within the community. In our experiment, however, the picture of a female entrepreneur on a brochure advertising a larger loan challenges local norms of relative power. In this sense, the paper is relevant to the growing literature on social norms which uses traditional marketing techniques to alter attitudes and behavior by changing individual perceptions (see, for example, La Ferrara et al., 2008; Jensen and Oster, 2007 and Paluck, 2009). The remainder of the paper is structured as follows. Section 2 describes the context in Pakistan and the marketing experiment. Section 3 discusses the data, Section 4 describes the empirical strategy and Section 5 reports the results of the experiment. Section 6 discusses the policy implications of the results and concludes. 3 2. Setting and Intervention Pakistan has a population of over 162 million, with over 60% living in rural areas. Although the agricultural sector continues to be important for overall growth, 45 percent of the rural poor rely on non-farm activities as important sources of income.2 Pakistan’s financial system has grown significantly in the past few years due in part to the successful implementation of various financial sector reforms, including the granting of banking licenses to a number of new private banks in the early 1990s, the modernization of the governance and regulatory framework of the banking sector in the late 1990s, and the privatization of major public sector banks since the early to mid- 2000s. Despite these recent achievements, access to financial services is still quite limited, especially in rural areas. According to the Access to Finance dataset (Nenova et al. 2009), only 14 percent of households interviewed reported using a financial product or service (including savings, credit, insurance, payments, and remittance services) from a formal financial institution.3 When informal financial access is taken into account, however, this figure rises to just over 50 percent. Overall, rural firms and households account for about 7 percent of total credit disbursement (about Rs 130.7 billion) and the bulk of this is for agricultural finance (Rs 108.7 billion), including both farm and nonfarm credit (see Akhtar, 2008). While microcredit volumes are skewed towards rural areas, microcredit currently accounts for only 17 percent of total rural credit and serves some 1.7 million clients. Comparative 2 Source: World Development Indicators database, April 2009 and World Development Report 2008. 3 In comparison, 32 percent of the population has access to the formal financial system in Bangladesh, and this figure amounts to 48 percent in India and 59 percent in Sri Lanka (World Bank, 2008). 4 rates of microfinance penetration in the South Asian region are 35 percent in Bangladesh, 25 percent in India, and 29 percent in Sri Lanka. Among microfinance providers, Khushali Bank is active in 86 districts; National Rural Support Program (NRSP) comes second with a presence in 51 districts while Kashf Foundation has some presence in 24 districts.4 These three microfinance entities account for approximately 70 percent of the sector’s active clients (MicroWatch, 2008). Unlike most other countries, the microfinance sector has focused primarily on men rather than women on the grounds that there is less demand for credit from women given their low mobility levels and cultural norms around women as economic actors. Consistent with this, Pakistan continues to under-perform on a range of social indicators relative to other countries at similar levels of per capita income and rural development. According to the 1998 Human Development Report, for example, Pakistan ranked 138 out of 174 on the Human Development Index, 131 out of 163 on the Gender Development Index (GDI), and 100 out of 102 on the Gender Empowerment Measure (GEM). 2.1 National Rural Support Program Established in 1991, NRSP is the largest of the Rural Support Programs in the country in terms of outreach, staff and development activities. It is modeled after the Aga Khan Rural Support Program, established in the early 1980s as a not-for-profit rural development organization. During the early 1990s, NRSP remained small, but the establishment of the Pakistan Poverty Alleviation Fund in 2000, a second-tier funding 4 Both NRSP and Kashf obtain a large fraction of their loan funds from the Pakistan Poverty Alleviation Fund (PPAF) which supported this work. 5 and capacity-building apex, provided critical funding that fueled NRSP and other partner NGOs’ growth. As part of its growth strategy, NRSP applied for a microfinance bank license in 2008 and became a microfinance bank in early 2010, falling under the supervision of the State Bank of Pakistan. Microfinance banks now account for almost half of the outreach of the microfinance sector. NRSP makes loans largely to members of a community organization. Its staff supports the creation of community organizations (CO) by a process of social mobilization which includes the creation of a community co-financed and co-managed infrastructure project and skill and group management training. Members of a CO typically live close to each other and meet regularly. Most also contribute towards individual and group savings. To date, NRSP has organized more than a million poor households into a network of more than 100,000 COs across the country. Roughly one- half to two-thirds of CO members are also active borrowers and group meetings serve as the venue for the receipt and repayment of loans for most members. NRSP has three main credit products: a single installment loan for agricultural inputs (fertilizer, seeds, etc.) with maturity of 6 to 12 months; enterprise loans and loans for livestock that have 12 monthly installments each. The maximum amount that can be borrowed depends on the number of loans successfully repaid (loan cycle). A new borrower starts with a small loan limit of Rs 10,000 (USD 117)5 which can increase in intervals of up to Rs 5,000 per loan cycle. As a point of comparison, a cow costs around 5 Currency converter accessed online on July 2nd, 2010. 6 Rs 60,000. All loans have joint liability at the CO level although new loans are issued even if some CO members are overdue.6 Besides credit, NRSP offers training in various vocational skills and provides up to 80 percent financing for infrastructure projects in the village. 2.2 The Experiment and Marketing Intervention The study was conducted in five branches in the districts of Bahawalpur, Hyderabad, and Attock, spanning different agro-climatic regions of Pakistan.7 Figure 1 shows the location of the study districts. NRSP staff conducted a complete listing of the occupation of CO members in the study branches to identify those who were engaged in a non-farm activity. After the listing, a baseline survey was conducted in November 2006 in a sample of 747 COs, selected so that their membership was between 5 and 26 members. The original sampling framework included all CO members that according to the listing exercise had a non-farm business and five other members selected at random from each CO. In practice, enumerators ended up interviewing everyone that attended a special CO meeting that was called to conduct the baseline survey. The resulting sample consisted of a total of 4,162 members interviewed, and 2,284 members (54.9%) that were in good standing. The timeline of the experiment is presented in Figure 2. Using data from the listing exercise, COs were randomly allocated into two groups, one of which was assigned to receive business training. Training sessions were 6 Borrowers are required to find two guarantors, who can be members of the same CO. NRSP staff uses guarantors as a means of exerting peer pressure, rather than enforcing repayment from them. 7 These branches are as follows: Matiari and Tando Muhammad Khan in Hyderabad, Attock in Attock and Bahawalpur (rural and urban) in Bahawalpur. 7 held, from February to May 2007. Each session lasted for 6 to 8 days (see Giné and Mansuri (2011a) for more details about the business training intervention). After completing the business training sessions, members from all study COs were invited to an orientation meeting that introduced the possibility of borrowing a larger loan amount. Most orientation sessions took place in regularly scheduled CO meetings and lasted for about an hour and a half. Attendance at these sessions was high, with more than 90 percent of members attending. Message consistency during the orientation was maintained by providing training to all NRSP credit officers and other staff who were in charge of delivering the orientations.8 During the orientation meeting, members who were in good standing, i.e., those who had successfully repaid at least one loan on time, received one of two versions of a marketing brochure. Orientations occurred successfully in 596 COs. In the remaining 151 COs orientation meetings could not be held because the CO had either disbanded or was newly formed so that none of its members was eligible for the lottery.9 The brochure was identical in all respects except one. In one version, the entrepreneurs manning the business were male while in the other they were female. To ensure that only the gender of the businessperson differed between both versions of the brochure, the exact same business was photographed twice, the first time with a man as owner and the second with a woman. Figure 3 shows the front and back of the brochure along with the picture of the businesses first with men (Male brochure) and then with women (Female Brochure). The businesses in the brochure were chosen to be representative of the type of businesses typically run by NRSP microentrepreneurs. The brochure thus contained two 8 There were 12 teams of two NRSP staff each in Attock, 29 in Bahawalpur and 7 in Hyderabad. 9 First time borrowers were not eligible to participate in the lottery. NRSP felt it did not have sufficient credit history for this group to allow them access to the much larger loans available to lottery winners. 8 agribusinesses, two retail businesses and one tailoring business. According to our baseline data, 49.41 percent of male businesses were agribusiness, 26.87 percent were in retail and 8.53 percent were involved in handicrafts and tailoring, thus accounting for almost 85 percent of all male businesses. Among female businesses, 19.88 percent were agribusinesses, 17.60 percent were retail businesses and 56.90 were in handicrafts and tailoring, accounting for almost 95 percent of all female businesses. All members of a CO were given one of the two brochures, which were randomly allocated across COs. The goal of the brochure was to explain how to apply for a larger loan via a lottery. Appendix A provides the translated text of the brochure. According to this, all eligible members could make a loan request of up to Rs. 100,000. The request was subject to all the usual technical and social reviews conducted by NRSP credit officers, who could also determine the loan amount they were willing to approve for each borrower. Approved loans which were larger than the usual limit of Rs. 30,000 were to be forwarded to headquarters, where the results of the lottery were maintained.10 Lottery winners could borrow the approved amount, while those who lost the lottery could borrow up to their regular loan size. Although the brochure encouraged members to borrow for productive purposes, in practice there were no restrictions on the use of the loan. In addition, qualifying members who already had an outstanding loan with NRSP were allowed to apply for the larger loan, subject to the condition that part of the new loan would be used to pay off the outstanding debt. Eligible CO members had seven months, from November 2007 to June 2008, to apply for the larger loan. Of the 2,284 eligible CO members, 713 (31.2 percent) applied. 10 The lottery was designed so that the chance of winning was 50 percent. See Giné and Mansuri 2011 for more details. 9 NRSP approved 532 loans (74.6 percent). Most applicants had their loan amounts reduced. Credit officers reported that this was due to concerns that borrowers would not be able to make the required monthly installments. Of the customers approved, 254 were assigned to win the lottery (47.7 percent) and 211 ended up borrowing (83 percent). Among the 278 loan applicants that lost the lottery, only 161 borrowed (58 percent). Among the reasons cited for changing their mind are time elapsed from request to approval (average time was 2 months), and for losers the fact that the new loan size was not too different than the loan they currently had. Figure 4 shows the distribution of loans by disbursement date and gender. The vertical axis measures number of loans applied for in each disbursement month. It is clear that women did the bulk of the borrowing in the first three to four months after the lottery started. Males, on the other hand, waited for the months of May and June to ask for loans, coinciding with the agricultural season. A follow-up survey was conducted in December 2008. This was six months after the loan lottery concluded and about 13 months after the loan orientation meetings. In the follow up, some 45 percent of eligible CO members recalled attending the loan lottery orientation meeting. Among those who recalled attendance, about 70 percent recalled receiving a brochure and of these about half were able to correctly recall the picture they were shown. Of the 211 lottery winners who took the larger loan, 125 reported loan use in the follow up. Table 1 reports the share of loans used for different purposes by gender, along with the p-value of a test of differences by gender. On average, about 48 percent of the loan was used for working capital, with men significantly more likely to use loans for this 10 purpose (50 percent as compared to 36 percent for women). In contrast only 6 percent of loan proceeds were used for purchasing business equipment, but here women were 3 times more likely to report this use (12 percent compared to about 4 percent for men). For other uses, men and women look roughly similar, though women use loans more frequently for consumer durables (8 percent of loan proceeds on average versus about 3 percent for men), while men use loan proceeds more frequently for housing improvements (8 percent versus about 3 percent for women). 3. Data Baseline data collected in November 2006, prior to the business training and loan lottery orientations, included questions on the CO member, the member’s household and the business. Besides the usual set of variables, such as age, education, marital status etc, individual characteristics include measures of entrepreneurship, digit span recall, risk preferences and decision making autonomy across a range of household and business outcomes. Household characteristics include information on the income generating activities of all household members, total household assets including livestock and past and current borrowing and saving of household members. Business characteristics include age, location and type of business activity, as well as the scale of the business as measured by its assets, hired workers and monthly sales. Summary statistics from the baseline survey are presented in Table 2, and variable definitions are provided in Appendix B. CO members are about 38 years old, have 4.1 years of education, own some 3.5 acres of land and have average household expenditures of about Rs 5,200 per month (roughly 61 USD). This places them significantly above the 11 bottom half of the population of the villages in which they reside (see Mansuri, 2011). Women are only slightly less likely than men to report owning a business (61 percent of sample men and 58 percent of sample women owned a business at baseline).11 However, there are significant differences between male and female CO members in almost all other dimensions, and these differences are sustained when we focus only on those who owned a business at baseline. Women tend to have less education, perform significantly below men on digit span recall and are less risk tolerant (on a 0 to 10 scale). Women members are also more likely to come from households that have less land wealth, as compared to the households of male CO members. This indicates some selection of women CO members by wealth and is consistent with more stringent female seclusion practices among landed rural households (see Jacoby and Mansuri, 2011). Importantly, women members also report far less decision making autonomy than men do on a range of household, individual and business outcomes. Among business owners, women report having complete autonomy over roughly 1.5 decisions out of a total of 8, whereas men claim to have complete autonomy over more than 3 decisions.12 Women are also more likely than men to belong to a mixed gender CO in our sample (about 63 percent of members in mixed COs are women). While this is a small sample overall, given that only 7 percent of COs have both men and women, there are some interesting differences between women in mixed and single gender COs. In particular, the former seem to have less ‘voice’ in the COs they belong to. While the odds of holding office in the single gender COs are about the same for men and women, at 20 and 18 percent respectively, 48 percent of men and only 7 percent of women in mixed 11 This is not surprising given that the men and women in our sample are all microfinance clients. 12 Giné and Mansuri (2011a) also find that women members are often not the effective managers of businesses they claim to own. 12 COs report holding any office. Women in mixed COs are also significantly more likely to observe purdah, even though mixed COs are more than twice as likely to be composed of members of the same zaat (caste).13 They are also younger and have more young children under age 9. They are also less risk tolerant, significantly less optimistic and have less trust in the formal institutions. However, they do about as well as other women on digit span recall and the index of entrepreneurial literacy. They are also about as well educated as women in female only COs and come from households that are, on average, wealthier than those of other women in the sample (Appendix Table A1). Both men and women in mixed COs are also far less likely to be eligible for the larger loan, and this is even more the case for men, perhaps because they have been CO members for a much shorter time, on average. Entrepreneurs are also different from other CO members. They are more likely to be older, more educated and to report a family history of business ownership. In addition, business owners have better digit span recall, have a better outlook on life and not surprisingly, also score higher on a business knowledge test (see Giné and Mansuri, 2010 for details). They are also more likely to be risk tolerant. Among women, those that have a business are less likely to report mobility constraints, but are somewhat more likely to observe purdah, perhaps because they are relatively wealthier. Table 2 also suggests that the scale at which male and female businesses operate is quite different (see also de Mel et al. 2009). Men’s businesses yield monthly sales that are more than 4 times as large as monthly sales for women. Women also tend to operate more businesses from home. While there is considerable overlap in the type of businesses 13 Seclusion practices are much less stringent within zaat/biradari (caste) groups. See Jacoby and Mansuri (2011). 13 owned by men and women, men are far more likely to be in the agribusiness sector, with much greater contact with wider markets, while women are concentrated in small scale manufacturing, handicrafts and tailoring in particular. Table 3 shows the main sources of credit for sample households at baseline. Only 5 percent of members have any loans from formal financial sources, mainly commercial banks. In contrast, 34 percent report borrowing from informal lenders, mainly shopkeepers, and 21 percent report borrowing from relatives.14 Not surprisingly, most borrow from an MFI, including NRSP. Average loan sizes vary greatly by source. While the average loan from a commercial bank is around 100,250 Rs (1,172 USD), the average MFI loan is only 12,000 Rs (USD 140) and loans from informal lenders are typically in about the same range. There is little variation in the relative share of lenders by gender, however, though female CO members tend to borrow less overall. Table 3 also reports reasons for not borrowing for those who reported not using a credit source. Interestingly, most respondents without loans report an unwillingness to borrow – either because they have no need for loans or because they dislike borrowing – as the main reason for not taking a loan and this varies little across lenders. Lack of collateral and cumbersome loan application procedures come in next, and are particularly important when dealing with a formal lender. Table 4 presents the means of baseline variables for the sample. Columns 2, 3 and 4 report the means for CO members who were exposed to the male and female brochure, respectively and the p-value of the test that the difference in means is significant. We also report the means and p-value of the same test for the sample of males (columns 5 to 8) 14 Informal lenders also include traders and wholesalers and, to a smaller extent, professional moneylenders and landlords. 14 and females (columns 9-12) separately. Overall, we find balance between the two groups. The difference in means for members receiving the male and female brochure is significant at conventional levels for only two of the 16 baseline variables we consider. Study participants who received the male brochure borrow more from informal sources and are less likely to be members of a mixed group (both significant at the 10% level). For the sample of males, the difference in means in the two groups is again significant for only two of the 16 variables considered. Male participants who received the male brochure are somewhat less likely to belong to a mixed CO (significant at the 10% level)15 and are somewhat less likely to borrow from friends and relatives (significant at the 5% level). For the sample of females the difference is only significant for own education. Women who received the male brochure appear to have higher formal education by about one-third of a year. Table 3 also reports the p-value of an F-test that all baseline characteristics are jointly insignificant. We cannot reject this hypothesis in any of the three samples (lowest p-value is 0.58). 4. Empirical Strategy Because the type of brochure is assigned randomly at the CO level, its impact can be estimated via the following regression equation: Yij = α + βFBj + γXij + εij (1) where Yij is the outcome of interest for individual i in CO j (e.g., an indicator variable that takes the value 1 if the respondent had applied for a larger loan, and 0 otherwise), FBj is an indicator variable that takes the value 1 if the respondent was shown the female 15 Since brochure-type was assigned at the CO level, the lower representation of men in mixed COs likely explains this. 15 brochure, Xij is a vector of individual characteristics collected at baseline and εij is a mean-zero error term. Standard errors are clustered at the CO level throughout since the CO level treatment assignment creates spatial correlation among members of the same CO (Moulton, 1986). The vector Xij includes the following baseline characteristics reported in Table 2: eligibility, being offered business training, a dummy if decision- making power is above the median in the same gender sample, own education, digit span recall, spouse education, landholdings, membership in a mixed group, age, number of children and risk tolerance. It also includes field unit (branch) dummies, our stratification variable. This set of variables includes characteristics for which there is imbalance as well as variables that are likely to affect the decision to borrow. 16 The coefficient β captures the impact of being shown a brochure with pictures of female entrepreneurs on the cover and is the coefficient of interest. We also examine interactions between the type of brochure shown and baseline characteristics which could proxy for attitudes towards women Yij = α + Ï?(FBj * Zij)+βFBj + γXij + εij (2) Zij is a subset of individual characteristics included in the vector Xij that could represent an individual’s attitudes towards women. The coefficient Ï? on the interaction term FBj * Zij reveals the extent to which the impact of the female picture (FB) on loan uptake varies according to the members’ attitudes towards women. 16 The regression results to come are not substantially affected by the exclusion of this set of control variables. 16 5. Empirical Results This section presents evidence on the impact of the female brochure on two main outcomes: the decision to borrow and the loan amount requested. Table 5 presents regression results from the estimation of equation (1) for the decision to borrow. Columns (1)-(3) present results for the full sample, combined and disaggregated by gender. On average, only 14.6 percent of members, roughly 31 percent of the eligible, applied for a larger loan. This is somewhat higher for men, at 20 percent, and correspondingly lower for women, among whom only 8 percent of the eligible applied. Given that the sample consists of seasoned microfinance clients, this number appears low. When asked anecdotally, many borrowers report either that the monthly installment amount for the larger loan was too high or that the maturity period was too short. Both of these indicate that the clients of NRSP are not, by and large, constrained by the current loan size, but that some could benefit from larger loans. Among men, older members and business owners were more likely to apply, while among women, being offered business training increases loan demand. Membership in a mixed CO dampens loan demand for both men and women even after controlling for eligibility, perhaps because group dynamics provide disincentives to borrow, since only half of the members of mixed COs are eligible to borrow (see Appendix Table A1). Exposure to the female brochure does not seem to have any effect, however. In columns (4)-(6) we focus on all CO members who had a business at baseline. We find, perhaps surprisingly, that the female brochure impacts negatively the uptake by women entrepreneurs. The effect is not small. From a base of 8.3 percent, the point estimate indicates a reduction of 39 percent in the probability of applying for the larger 17 loan. In contrast, businessmen appear indifferent to the picture in the brochure and, as before, being offered business training increases loan demand while membership in a mixed CO depresses it, but there are no differences by gender in these or any other characteristics. Taken face value, the result among women appears counterintuitive: if anything, one might expect that a woman who owns a business would be attracted to a loan product that makes salient her identity as a business woman. But since women operate businesses that are much smaller than those of men, it is possible that the negative impact is an artifact of the business scale. Specifically, women may find the scale of the business pictured in the brochure too large and they may be discouraged from applying.17 To explore this hypothesis further, we restrict attention to female business owners who operate at a scale comparable with men.18 Specifically, we construct a sample of matched male and female businesses by sector.19 Given differences in business scale by gender, this matched sample necessarily consists of businesses in the upper (lower) tail of the distribution of female (male) businesses. In particular, female businesses in the matched sample have on average Rs 4,016 higher sales compared to the sample of all female businesses, while male businesses in the matched sample have Rs 8,129 lower sales, on average. Sample size is also considerably reduced. Only 363 female businesses can be matched with a corresponding male business. Columns (7)–(9) report 17 We have some anecdotal evidence that businesswomen felt that some of the businesses featured in the brochure were run on a larger scale than the typical female business and that this fact discouraged them from borrowing. As mentioned, many women operated each of the three business types featured in the brochure, which together account for 95 percent of all businesses run by women. In addition, the brochure clearly stated that the loan could be used for any purpose and for any business, but nonetheless, the picture may have triggered a more deliberative response (Kahneman, 2003). 18 We are grateful to an anonymous referee for suggesting this analysis. 19 The matched sample is generated as follows. For each female business, we compute the absolute difference in sales with each male business in the same sector and then pick the business with the smallest difference. A male business is matched only once with a female business. In the final sample we keep only those female businesses where the absolute difference in sales is less than Rs 1000. 18 these results. The coefficient for female brochure remains negative but we lose precision. The coefficient falls by about 16 percent while the standard error increases by about 70 percent. In sum, we cannot conclusively rule out a negative effect of exposure to the female brochure even among women business owners who run larger businesses. Note, however, that in this matched group, being offered business training increases the probability of a woman applying for a larger loan by 86 percent from a base of 11 percent and appears to be driving the overall increase in loan applications due to the offer of business training. On the other hand, decision making power does not appear to encourage loan applications from women in any of these samples. Table 6 reports the regression results from specification (2), for males (columns 1-3) and females (columns 4-6) on all three samples. All regressions include the baseline controls used in Table 5. The first striking fact is that a subset of men do respond to the psychological content of the brochure. Specifically, business owners with low scores on the digit span recall question, a proxy for ability, and those whose wives are poorly educated (p-value of FB x Years Education of Spouse is 0.11), respond negatively to the female brochure. In this group, loan demand falls by about 13 percentage points (more than a 50% decline over the base demand). While our experimental design does not allow us to distinguish between lack of affinity towards women and an outright distaste for female-run businesses as possible explanations for the negative response, it does suggest one channel through which exposure to the female brochure among female business owners could depress loan demand. Specifically, women who have low decision making power may turn to their 19 husbands for permission to borrow and may face a negative response from them when shown the female brochure. We do find some evidence in support of this. Among female business owners, only those with low decision making power in their household react negatively to the female brochure. Interestingly, female business owners with high decision making autonomy shown the male brochure also react negatively by roughly the same magnitude, while there is no effect on female business owners with autonomy shown the female brochure (p-value is 0.28).20 This suggests that women with decision-making autonomy react negatively to the brochure of opposite sex by about as much as men without a business, with low digit span and poorly educated wives. Given the disadvantaged position of women in rural Pakistan, we conjecture that men may have less respect for female businesses whereas women may feel more alienated when shown the male brochure. In the matched sample (column 6), the results for female business owners with high decision making autonomy become stronger (the point estimate increases in absolute value). Table 7A reports the impact of the female brochure and baseline characteristics on the loan amount requested (columns 1-4) and approved (columns 5-8) among loan applicants. Table 7B reports the same regressions among loan applicants with a business at baseline. In both tables we find a positive and significant effect of the picture on the loan amount requested among males but not females. The result in both tables is driven 20 More formally we test whether the coefficient on the female brochure plus the coefficient on decision- making power above median plus the coefficient on the interaction between the female brochure and decision-making power above median is different from zero, that is β+γ+Ï?=0 following the notation of Equation (2) . 20 primarily by selection because men without a business and low digit span recall tend to borrow less and decide not to borrow when shown a female brochure. 6. Conclusions We designed a marketing experiment to test whether exposure to positive role models could encourage women’s uptake of a new credit product in a context where women face barriers to participation in economic life. Brochures advertising the new product, a much larger loan, were varied such that the cover page featured either men or women as entrepreneurs running five otherwise identical businesses. The brochures were randomly assigned to existing clients in good standing. The results suggest that both men and women respond to psychological cues embedded in this type of social norms marketing. However, men’s response is conditioned by relative economic status and ability while women’s response is conditioned by relative status within the household. In particular, men who are not themselves business owners have lower measured ability and whose wives are less educated respond more negatively to the female brochure, as do women business owners who have low autonomy within the household. Women with relatively high levels of autonomy shown the male brochure have a similar negative response. In contrast, the reaction of high autonomy female business owners shown the female brochure is no different from that of low autonomy women shown the male brochure. We conjecture that loan demand for low autonomy women is mediated through men who respond positively to the male brochure. This suggests that social norms marketing can often be more salient for the more disadvantaged (Paluck and Ball, 2010) but, as our results suggest, this could generate perverse responses in some contexts. 21 Finally, our results also suggest that exposing women to positive role models or information that challenges prevailing norms may meet different levels of success depending on the level of autonomy enjoyed by women. In particular, women with low levels of autonomy may require more intensive interventions, consistent with other work which has used information campaigns to change stereotypes. Giné and Mansuri (2011b), for example, find that the response to a voter awareness campaign directed at rural women in Pakistan was most effective among women who had voted in the past and hence had overcome some of the barriers to participation in public life. References Akhtar, Shamshad, 2008, “Financial Sector–Ten Year Vision and Strategy.â€? Presentation at the inauguration of the Development Finance Conference on “Expanding Frontiers of Financial Access in Pakistanâ€? organized by State Bank on the eve of 60th Anniversary of the State Bank of Pakistan, Karachi Banerjee, Abhijit and Esther Duflo, 2008, “Do Firms Want to Borrow More? Testing Credit Constraints Using a Directed Lending Programâ€?, MIT Department of Economics Working Paper No. 02-25. Beaman, L. Raghabendra Chattopadhyay, Esther Duflo, Rohini Pande and Petia Topalova, 2009, “Powerful Women: Does Exposure Reduce Bias?â€?, Quarterly Journal of Economics 124(4): 1497-1540. Bertrand et al., 2010, “What is advertising content worth? Evidence from a consumer marketing field experimentâ€?, Quarterly Journal of Economics, v. 125, issue 1, pp. 263- 306. De Mel, Suresh, David McKenzie and Christopher Woodruff, 2009, “Are Women More Credit Constrained? Experimental Evidence on Gender and Microentreprise Returnsâ€?, American Economic Journal: Applied Economics, 1(3):1-32. De Mel, Suresh, David McKenzie and Christopher Woodruff , 2010, “Getting Credit to High Return Microenterprises: The Results of an Information Interventionâ€? mimeo, World Bank. Evans, Franklin B., 1963, "Selling as a Dyadic Relationship", American Behavioral 22 Scientist 6: 76-79. Giné, Xavier and Ghazala Mansuri, 2011a, “Money or Ideas? A Field Experiment on Constraints to Entrepreneurship in Rural Pakistanâ€?, mimeo Giné, Xavier and Ghazala Mansuri, 2011b, “Together We Will: Evidence from a Field Experiment on Female Voter Turnout in Pakistanâ€?, mimeo Jacoby, Hanan and Ghazala Mansuri, 2011, “Crossing Boundaries: Gender, Caste and Schooling in Rural Pakistanâ€?, mimeo Jensen, Robert and Emily Oster, 2007, “The Power of TV: Cable Television and Women's Status in Indiaâ€?,NBER Working Paper No. 13305 Kahneman, Daniel, 2003, “Maps of Bounded Rationality: Psychology for Behavioral Economicsâ€?, American Economic Review, 93(5): 1449-1475. La Ferrara, Eliana, Alberto Chong and Suzanne Duryea, 2008, “Soap Operas and Fertility: Evidence from Brazilâ€?, BREAD Working Paper No. 172 Mansuri, Ghazala, 2011, “Bottom up or Top Down? Participation and the Provision of Local Public Goodsâ€?, mimeo. Mobius, Markus M. and Tanya S. Rosenblat, 2006, “Why Beauty Mattersâ€?, American Economic Review 96(1): 222-235. Moulton, Brent, 1986, “Random Group Effects and the Precision of Regression Estimatesâ€?, Journal of Econometrics, 32, 3:p. 385-397. Nenova, T., and C. Nenang, with A. Ahmad, 2009, “Bringing Finance to Pakistan’s Poor: A Study on Access to Finance for the Underserved and Small Enterprisesâ€?, World Bank report. Washington, DC: World Bank. Paluck, Elizabeth L., 2009, “Reducing intergroup prejudice and conflict using the media: A field experiment in Rwanda,â€? Journal of Personality and Social Psychology 96(3): 574-587. Sen, A., 1999, Development and Freedom, New York: Anchor Books. UNDP, 1998, Human Development Report 1998: Consumption for Human Development. New York: Oxford University Press. Yunus, M., 1999, Banker to the Poor, London: Aurum Press Ltd. World Bank, 2008, “Finance for All? Policies and Pitfalls in Expanding Accessâ€?, World Bank: Washington, DC. 23 Appendix A: Brochure script NRSP (in collaboration with the World Bank and the Pakistan Poverty Alleviation Fund) has initiated an Enterprise Development Program. As part of this program, the possibility of obtaining a larger loan is now available. Which CO members qualify for this larger loan? All CO members that have a good borrowing record with NRSP (that is, have successfully repaid at least one loan) will be eligible to put in a request for a larger loan to fund their business activity. Who will obtain the larger loan? Among the applicants who qualify for the larger loan, a lottery will be implemented to determine who gets the larger loan. Winners of the lottery will receive the larger loan approved by NRSP. Losers of the lottery will be offered a normal size loan according to their credit history with NRSP. Each qualifying CO member will have a 50% chance of winning the lottery. To ensure transparency and fairness, the loan lottery will be done in the NRSP head office in Islamabad. Procedure to apply for a larger loan The following steps are involved in accessing larger loans through this program 1. COs will pass a resolution identifying a larger loan need for their eligible and interested members, as such demands come in. 2. Each time there is a demand for a larger loan, a social and technical appraisal will be done to assess the applicant’s credit history and repayment capacity and the loan size that the candidate can safely repay. 3. If a larger than normal loan is approved, the loan application will be forwarded to headquarters where the results of the lottery will be checked and disbursement made accordingly. 4. Borrowers who win the lottery will actually get the larger loan. Borrowers who lose in the lottery will be offered the normal loan amount set by NRSP. Amount of Loan Maximum up to Rs. 100,000/- (One hundred thousand only) according to the need Purpose of Loan Loan will be taken and subsequently utilized for productive purposes only Duration of Loan Up to maximum of One year Loan Repayment Loan repayment will be made according to the prevalent NRSP procedures, whereby the borrower will be given a repayment schedule and he/she will have to repay in 24 installments, in accordance with the schedule, in the nearest NRSP village branch and take a repayment receipt. Appendix B: Variable definitions Data used in this paper come from two surveys: a baseline conducted in November 2006, a follow-up survey in December 2008. We also used administrative data about loan take- up, obtained from NRSP’s internal records. Treatments and Take Up (from administrative records) • Female picture brochure takes the value of 1 if the member was shown a brochure with female business owners on the cover, 0 if the brochure showed the same businesses with men. • Eligibility, takes the value of 1 if member is eligible for the loan lottery, 0 otherwise. • Business Training, a dummy taking the value of 1 if the individual was offered business training, 0 otherwise. • Applied for larger loan, 1 if the member actually requested a loan, 0 otherwise. • Approved, conditional on applying for a loan, takes a value of 1 if the member was approved by NRSP, 0 otherwise. • Borrowed, takes a value of 1 if the individual actually borrowed money from NRSP; while conditional on approval, not all those that applied and were approved took actually a loan. • Amount borrowed, is measured in thousands of Rupees. Baseline characteristics Individual • Female equals 1 for women and 0 for men. • Age is respondent’s age in years. • Years of education is years of completed schooling, and is top-coded at 16. • Married, a dummy taking the value of 1 if member is married, 0 if single, divorced or widowed. • Digital span recall reports the number of digits correctly recalled after being shown an eight digit number for 30 seconds. • Member of a mixed group, dummy takes the value of 1 if the member belongs to a borrowing group with mixed gender, 0 if the group is of the same gender. • Index of female mobility and No purdah index are principal components of several variables with negative values indicating less mobility (or observing more types of purdah). 25 • Business owner equals 1 if the member had a business at baseline, 0 otherwise. • Aversion to risk general is measured on a 0-10 scale where 0 indicates the most risk averse and 10 the most risk-tolerant/lover. • Months as member, number of months as member of NRSP group. • Holds Office in Group, takes value 1 if member has or has had in the past a leadership position in group. • Fraction of Members of same Zaat (caste), is a percentage of members in the group that share the same cast of the member. Household • Education of spouse is years of completed schooling of the respondent’s partner, if any. Top coded at 16. • Total HH income and Expenditures, transformed to logs for analysis. • Number of children under 9 • Land is the total owned land inside and outside the village. • Credit constraints, dummy taking a value of 1 if the member faced any type of credit constraint, formal or informal. • Ever in business, captures business experience within the household. Equals 1 when this is the case, 0 otherwise. • Decision Making, is the number of household decisions out of a total of eight that the member usually takes on his or her own. The decisions are: children’s schooling, consumption expenditures, major investments in business or land, the respondent’s participation in community or political activities, the respondent’s spouse participation in community or political activities, whether or not the respondent should work for an income, whether or not the spouse should work for an income and how much the household saves. In the analysis, a dummy is used that takes value 1 if the variable is above the median for each gender subsample. Business characteristics • Type of business, dummy variables for businesses shown on brochure • Business Literacy, scores of component 1 of a PCA for a set of questions about knowledge about how to run a business, and of competition. • Sales in ‘000 rupees, sales of business in an average month at the time of baseline. Analysis-related variables: • Field Unit, refers to the NRSP branch and is the stratification variable. There are six field units in the three study districts. 26 IBRD 39224 TAJIKISTAN CHINA PAK I S TAN FIGURE 1 JAMMU STUDY DISTRICTS KHYBER and PROVINCE CAPITALS PAKHTUNKHWA NATIONAL CAPITAL PROVINCE BOUNDARIES Approximate INTERNATIONAL BOUNDARIES ISLAMABAD Line of Control Peshawar ATTOCK KASHMIR 0 100 200 300 KILOMETERS 0 50 100 150 200 MILES Lahore PUNJAB AFGHANISTAN Quetta BAHAWALPUR BALOCHISTAN INDIA ISLAMIC REP. OF IRAN SINDH HYDERABAD PAKISTAN Karachi This map was produced by the Map Design Unit of The World Bank. The boundaries, colors, denominations and any other information shown on this map do not imply, on the part of The World Bank Group, Arabian any judgment on the legal status of any territory, or any endorsement or acceptance of Sea such boundaries. APRIL 2012 Figure 2. Timeline Nov 06 Febâ€?May 07 Nov 07 June 08 Dec 08 Baseline  Business Training Brochures   Followâ€?up  Survey and Loan  Survey Lottery Figure 3. Brochures Front Back Male Female Figure 4. Distribution of Loans by Date of Disbursement 100 90 80 70 60 50 40 30 20 10 0 Nov-07 Dec-07 Jan-08 Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Females Males Table 1. Reported Larger Loan Use P-val of t-test All Male Female (2)-(3) (1) (2) (3) (4) Household Durables 0.038 0.027 0.080 0.161 Food consumption 0.041 0.041 0.041 0.999 School Supplies 0.003 0.003 0.005 0.584 Festivals and Ceremonies 0.010 0.007 0.022 0.250 Household Repairs 0.070 0.081 0.027 0.293 Previous Loan Repayment 0.059 0.057 0.068 0.787 Savings 0.041 0.036 0.063 0.438 Inventories/ raw materials for main business 0.475 0.505 0.355 0.110 Equipment for main business 0.055 0.038 0.124 0.064 Inventories and Equipment for other household businesses 0.061 0.053 0.096 0.361 Other uses 0.146 0.153 0.118 0.606 Number of Obs. 125 100 25 Note: Data come from the follow-up survey conducted in December 2008. Columns report the average fraction of loans used for each purpose. Table 2. Summary Statistics P-value of P-value Matched Business P-value All members All members t-test (7)- Business Owners of t-test Owners of t-test N. Obs Mean Std. Dev. 10th Pct. Median 90th Pct. Male Female (8) Male Female (10)-(11) Male Female (13)-(14) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) Take-up Shown Female brochure (Yes=1) 3,451 0.52 0.50 0.00 1.00 1.00 0.49 0.56 0.00 0.47 0.57 0.02 0.45 0.58 0.02 Eligible for loan lotttery (Yes=1) 3,451 0.55 0.50 0.00 1.00 1.00 0.62 0.62 0.00 0.69 0.67 0.00 0.71 0.68 0.01 Offered Business Training (Yes=1) 3,451 0.54 0.50 0.00 1.00 1.00 0.52 0.53 0.00 0.55 0.53 0.00 0.54 0.56 0.04 Applied for loan (1=Yes) 2,149 0.31 0.46 0.00 0.00 1.00 0.42 0.18 0.00 0.45 0.17 0.00 0.44 0.22 0.00 Approved, conditional on applying (1=Yes) 664 0.76 0.43 0.00 1.00 1.00 0.76 0.76 0.01 0.78 0.75 0.05 0.81 0.74 0.35 Borrowed, conditional on being approved (1=Yes) 503 0.69 0.46 0.00 1.00 1.00 0.75 0.52 0.62 0.75 0.55 0.85 0.78 0.63 0.87 Amount borrowed ('000 Rs) 445 33.00 16.94 15.00 30.00 50.00 33.83 30.00 0.06 34.59 29.55 0.02 36.04 32.50 0.43 Baseline Characteristics Individual Characteristics Female (Yes=1) 3,451 0.46 0.50 0.00 0.00 1.00 - - - - - - - - - Age 3,451 37.88 12.17 24.00 36.00 55.00 38.18 37.51 0.01 37.5 36.97 0.29 38.49 39.02 0.98 Years of Education (0-16) 3,451 4.09 4.51 0.00 3.00 10.00 5.31 2.63 0.00 5.53 2.73 0.00 5.29 2.65 0.00 Married (Yes=1) 3,451 0.83 0.37 0.00 1.00 1.00 0.82 0.84 0.12 0.82 0.86 0.13 0.84 0.87 0.09 Digit Span Recall (0-8) 3,451 3.31 2.24 0.00 4.00 6.00 3.84 2.68 0.00 4.03 2.74 0.00 3.87 2.48 0.00 Member of a Mixed Group (Yes=1) 3,451 0.06 0.24 0.00 0.00 0.00 0.04 0.09 0.00 0.05 0.09 0.00 0.03 0.13 0.00 Index of Female Mobility 1,571 0.06 1.35 -1.11 -1.11 2.68 - 0.06 - - 0.12 - - 0.15 - Index of No Purdah 1,571 0.15 1.62 -2.58 0.96 1.72 - 0.15 - - 0.13 - - 0.37 - Business Owner (Yes=1) 3,451 0.60 0.49 0.00 1.00 1.00 0.61 0.58 0.01 - - - - - - Risk Tolerance (0=Risk Averse; 10= Risk Lover) 3,451 3.61 3.03 0.00 4.00 8.00 3.81 3.37 0.00 3.84 3.39 0.00 4.08 3.59 0.00 Months as Member 3,451 26.55 23.95 6.00 19.00 54.00 24.57 24.56 0.00 26.92 22.11 0.00 28.38 25.35 0.01 Household Characteristics Years of Education, Spouse (0-16) 3,451 3.31 4.34 0.00 0.00 10.00 2.12 4.75 0.00 2.24 4.66 0.00 2.06 4.53 0.00 Total Household Income (log) 3,451 11.53 0.81 10.59 11.51 12.54 11.62 11.44 0.00 11.71 11.45 0.00 11.53 11.61 0.99 Expenditures (log) 3,451 8.28 0.63 7.31 8.01 9.10 8.29 8.27 0.00 8.32 8.29 0.00 8.19 8.39 0.40 Number of Children under 9 3,451 2.65 2.17 0.00 2.00 5.00 2.89 2.37 0.00 2.89 2.43 0.00 2.77 2.31 0.01 Land (area) 3,451 4.48 18.69 0.01 0.25 10.38 5.94 2.74 0.00 5.84 2.57 0.01 4.99 3.48 0.01 Credit constraints (Yes=1) 3,451 0.14 0.35 0.00 0.00 1.00 0.12 0.16 0.00 0.12 0.12 0.00 0.12 0.13 0.01 Family ever in business (Yes=1) 3,451 0.61 0.49 0.00 1.00 1.00 0.61 0.61 0.31 0.75 0.7 0.06 0.74 0.69 0.14 Decision Making (0-8) 3,451 2.65 3.11 0.00 1.00 8.00 3.34 1.84 0.00 3.35 1.59 0.00 3.20 1.79 0.00 Sources of Credit % borrowing Formal Sector 2006-08 2,931 0.05 0.22 0.00 0.00 0.00 0.07 0.03 0.01 0.08 0.03 0.02 0.09 0.01 0.01 Amount borrowed 2006 ('000s) 2,931 5.94 50.92 0.00 0.00 0.00 9.67 1.37 0.08 11.57 0.85 0.19 13.35 0.15 0.31 % borrowing Microfinance Institutions / Microfinance Banks 2006-08 2,931 0.82 0.38 0.00 1.00 1.00 0.79 0.86 0.00 0.86 0.93 0.00 0.87 0.91 0.00 Amount borrowed 2006 ('000s) 2,931 31.17 26.35 0.00 27.00 65.00 34.23 27.42 0.00 37.67 28.79 0.00 39.62 31.54 0.13 % borrowing Friends and Family (other than CO members) 2006-08 2,931 0.22 0.41 0.00 0.00 1.00 0.21 0.23 0.11 0.2 0.22 0.30 0.19 0.22 0.47 Amount borrowed 2006 ('000s) 2,931 5.57 18.73 0.00 0.00 17.00 6.19 4.81 0.00 6.19 4.78 0.01 6.04 4.61 0.13 % borrowing Informal Lenders 2006-08 2,931 0.48 0.50 0.00 0.00 1.00 0.50 0.47 0.16 0.47 0.42 0.27 0.49 0.45 0.20 Debt to informal lenders 2006 ('000s) 2,931 15.07 68.88 0.00 0.00 25.00 23.21 5.06 0.00 26.45 4.32 0.00 18.38 4.73 0.02 Businesses Agribusiness, Dairy, Livestock (Yes=1) 1,962 0.40 0.49 0.00 0.00 1.00 0.54 0.22 0.00 0.54 0.22 0.00 0.56 0.56 - Retail and Food Services (shopkeeping) (Yes=1) 1,962 0.24 0.42 0.00 0.00 1.00 0.28 0.19 0.00 0.28 0.19 0.00 0.29 0.29 - Handicraft, Tailoring, Vocational Trade (Yes=1) 1,962 0.29 0.46 0.00 0.00 1.00 0.09 0.55 0.00 0.08 0.55 0.00 0.05 0.05 - Other (Yes=1) 1,962 0.07 0.44 0.00 0.00 0.00 0.09 0.04 0.00 0.09 0.03 0.00 0.09 0.09 - Sales ('000 Rs) 2,065 14.51 103.18 0.80 4.00 25.00 22.3 4.74 0.01 22.3 4.74 0.00 7.06 7.01 0.48 Note: All variables are from the baseline survey (November 2006), except for Sources of Credit information, which is from the followup and recalls credit for the 2006-2008 period. See Appendix B for definition of variables. Table 3. Percentage Borrowing and Reasons for not Borrowing by Credit Source Credit Source Relatives Commercial Informal and Bank MFI Lenders Friends Percent borrowing from [source] in 2006 5.07 83.43 33.77 21.12 The main [reason] for not borrowing from [source]? Do not like/need to borrow 52.55 81.39 70.81 74.69 Inadequate collateral 18.50 1.98 11.94 6.77 Lender's procedures are too cumbersome 14.49 9.31 6.70 5.82 Lender's loan terms are unfavorable 5.48 0.59 3.35 0.91 Lender is too far away 3.18 0.40 0.78 1.35 Need to pay bribes 2.02 0.79 2.18 0.08 Past default with lender 1.96 0.79 0.89 1.15 Members not willing to lend to me 1.25 4.55 2.51 6.53 Bad credit history 0.56 0.20 0.84 2.69 Notes: Data come from the baseline survey (November 2006) Table 4. Verification of Randomization All members Male Female Means P-val of Means P-val of Means P-val of N. Male Female t-test N. Male Female t-test N. Male Female t-test Obs. Brochure Brochure (2)-(3) Obs. Brochure Brochure (6)-(7) Obs. Brochure Brochure (11)-(12) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Characteristics at Baseline Age 3,451 37.95 37.81 0.74 1,880 37.98 38.40 0.89 1,571 37.91 37.20 0.85 Number of children under 9 3,451 1.73 1.83 0.77 1,880 1.92 1.99 0.53 1,571 1.45 1.40 0.17 Business Owner(Yes=1) 3,451 0.60 0.59 0.57 1,880 0.63 0.59 0.75 1,571 0.57 0.60 0.79 Digit Span Recall 3,451 3.36 3.28 0.85 1,880 3.82 3.88 0.83 1,571 2.72 2.65 0.82 Risk Tolerance (0=Risk Averse; 10= Risk Lover) 3,451 3.71 3.52 0.25 1,880 3.81 3.81 0.53 1,571 3.57 3.22 0.49 Land 3,451 4.38 4.58 0.82 1,880 5.83 6.05 0.77 1,571 2.36 3.05 0.55 Years of Education of Spouse 3,451 3.69 3.71 0.69 1,880 2.72 2.73 0.86 1,571 5.05 4.75 0.55 Member of a Mixed Group (Yes=1) 3,451 0.04 0.08 0.10 1,880 0.02 0.06 0.06 1,571 0.07 0.10 0.89 Own Education 3,451 4.26 3.95 0.14 1,880 5.27 5.36 0.78 1,571 2.84 2.47 0.07 Decision Making 3,451 2.74 2.57 0.35 1,880 3.41 3.26 0.77 1,571 1.82 1.85 0.61 Pct. Borrowing from … at the time of Baseline* Commercial Bank 2,931 2.38 1.7 0.64 1,616 3.69 2.71 0.60 1,315 0.51 0.69 0.74 Microfinance Institution 2,931 71.34 67.62 0.30 1,616 69.44 63.74 0.98 1,315 74.06 71.74 0.46 Friends and Relatives 2,931 7.71 6.58 0.48 1,616 7.85 5.16 0.05 1,315 7.5 8.09 0.14 Informal lenders 2,931 0.7 0.2 0.07 1,616 0.83 0.26 0.19 1,315 0.51 0.13 0.23 Pct. Offered Business Training 3,451 52.47 51.82 0.83 1,880 51.29 52.03 0.93 1,571 54.1 51.6 0.48 Member is eligible for loan lottery (Yes=1) 3,451 63.72 69.00 0.54 1,880 64.63 59.69 0.82 1,571 62.45 62.21 0.66 P-val of F-test that all baseline characteristics are jointly insignificant 0.65 0.58 0.65 Notes: * denotes variable measured at follow-up, conducted in December 2008. Pct. Offered Business Training and Member eligibility come from administrative data from NRSP. Table 5. Uptake of Larger Loan OLS Sample: All members All Business Owners Matched Business Owners All Males Females All Males Females All Males Females (1) (2) (3) (4) (5) (6) (7) (8) (9) Female Brochure -0.026 -0.033 -0.015 -0.029 -0.017 -0.051** -0.029 -0.007 -0.043 (0.021) (0.034) (0.019) (0.027) (0.044) (0.023) (0.035) (0.058) (0.039) Offered Business Training 0.048** 0.046 0.046** 0.058** 0.067 0.037 0.071** 0.056 0.095** (0.021) (0.035) (0.019) (0.027) (0.044) (0.024) (0.035) (0.055) (0.047) Female -0.016 -0.034 -0.029 (0.021) (0.031) (0.046) Age 0.006*** 0.008** 0.000 0.006 0.009 -0.003 0.009 0.011 -0.003 (0.002) (0.004) (0.004) (0.004) (0.006) (0.006) (0.006) (0.008) (0.011) Age^2 -0.000** -0.000** 0.000 0.000 0.000 0.000 -0.000* 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) High Decision-making power (1=yes) 0.001 -0.003 -0.022 0.002 0.009 -0.016 -0.002 -0.037 -0.039 (0.002) (0.020) (0.017) (0.003) (0.025) (0.022) (0.005) (0.051) (0.036) Own Education 0.002 0.005 0.000 0.003 0.004 0.001 0.004 0.004 0.007 (0.002) (0.003) (0.003) (0.003) (0.004) (0.004) (0.005) (0.007) (0.006) Digit Span Recall 0.003 0.002 0.001 -0.001 -0.002 -0.002 0.002 -0.001 -0.001 (0.003) (0.005) (0.004) (0.005) (0.008) (0.005) (0.008) (0.013) (0.011) Risk Tolerance 0.001 -0.001 0.002 0.001 -0.001 0.001 0.004 -0.002 0.006 (0.002) (0.003) (0.003) (0.003) (0.004) (0.003) (0.005) (0.008) (0.006) Yrs. Of Education of Spouse 0.000 -0.005 0.002 -0.001 -0.003 0.000 -0.004 -0.011* -0.001 (0.002) (0.003) (0.002) (0.002) (0.004) (0.003) (0.003) (0.006) (0.004) Number of children under 9 0.003 0.003 0 0.007 0.007 0.006 0.01 0.017 0.001 (0.003) (0.004) (0.004) (0.004) (0.006) (0.007) (0.008) (0.011) (0.011) Business Owner (Yes=1) 0.028** 0.040* 0.01 (0.014) (0.022) (0.018) Land 0.000 0.001 0.000 0.000 0.001 0.000 -0.001 0.000 -0.002 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.001) (0.001) (0.002) Member of a Mixed Group (Yes=1) -0.092*** -0.086* -0.105** -0.150*** -0.121* -0.196*** -0.166** 0.021 -0.272*** (0.032) (0.047) (0.048) (0.043) (0.066) (0.070) (0.073) (0.236) (0.100) Eligibility 0.188*** 0.246*** 0.130*** 0.203*** 0.299*** 0.111*** 0.219*** 0.304*** 0.148*** (0.015) (0.026) (0.016) (0.019) (0.033) (0.018) (0.024) (0.040) (0.032) Mean Dependent Variable 0.15 0.20 0.08 0.15 0.20 0.08 0.18 0.25 0.11 N. Observations 3,451 1,880 1,571 2,065 1,149 916 726 363 363 R-squared 0.17 0.18 0.11 0.18 0.17 0.13 0.17 0.17 0.16 Note: The dependent variable takes value 1 if the member applied for a larger loan. All regressions include branch fixed effects and are estimated using OLS methods. Standard errors are clustered at the borrower group level.The following symbols *, * * and ** * denote significance at the 10, 5 , and 1 percent level, respectively. See Appendix B for definition of variables. Table 6: Heterogeneous Effects of Larger Loan Uptake OLS Male Female All Business Matched Business All Business Matched Business All Members All Members Owners Owners Owners Owners (1) (2) (3) (4) (5) (6) Female Brochure (FB) -0.132** -0.134* -0.156 -0.01 -0.071* -0.077 (0.051) (0.071) (0.113) (0.034) (0.038) (0.054) Business Training 0.061 0.097* 0.075 0.027 0.013 0.033 (0.045) (0.054) (0.070) (0.026) (0.032) (0.052) FB x Business Training -0.042 -0.064 0.002 0.03 0.022 0.054 (0.065) (0.080) (0.108) (0.035) (0.044) (0.071) High Decision-making power (1=yes) -0.017 -0.010 -0.039 -0.022 -0.053* -0.101* (0.024) (0.031) (0.068) (0.024) (0.030) (0.054) FBx High Decision making power 0.008 0.008 0.006 0.023 0.077** 0.138** (0.005) (0.007) (0.014) (0.031) (0.039) (0.068) Digit Span Recall -0.005 -0.014 -0.016 0.008 0.008 0.015 (0.006) (0.010) (0.017) (0.005) (0.007) (0.013) FB x Digit Span Recall 0.017** 0.024* 0.023 -0.009 -0.013 -0.019 (0.008) (0.012) (0.021) (0.006) (0.009) (0.016) Yrs. Education of Spouse -0.038 -0.063* -0.111 -0.011 -0.01 0.034 (0.028) (0.038) (0.069) (0.025) (0.033) (0.064) FB x Yrs. Education of Spouse 0.039 0.09 0.123 0.043 0.041 -0.012 (0.040) (0.056) (0.101) (0.032) (0.039) (0.071) Business owner (Yes=1) 0.031 0.045** (0.030) (0.022) FB x Business owner 0.033 -0.057* (0.042) (0.031) Mean Dependent Variable 0.20 0.24 0.25 0.08 0.08 0.11 N. Obs. 1880 1149 363 1,571 916 363 R-squared 0.19 0.19 0.18 0.12 0.13 0.18 Note: The dependent variable takes value 1 if the member applied for a larger loan. All regressions are estimated using OLS methods and control for eligibility, own education, landholdings, membership in a mixed group, age, number of children, risk tolerance, as well as branch fixed effects. Standard errors are clustered at the borrower group level.The following symbols *, * * and ** * denote significance at the 10 , 5, and 1 percent level, respectively. See Appendix B for definition of variables. Table 7A. Loan Size, All Members OLS Log of Amount Requested Log of Amount Approved All Members Males Females All Members Males Females (1) (2) (3) (4) (5) (6) (7) (8) Female Brochure 0.088 0.085* 0.161*** -0.035 0.0036 -0.006 0.008 -0.098 (0.054) (0.049) (0.052) (0.104) (0.029) (0.030) (0.030) (0.069) Offered Business Training -0.041 -0.032 -0.092 0.004 -0.01 0.007 (0.042) (0.046) (0.104) (0.029) (0.031) (0.068) Female 0.078 -0.054 (0.079) (0.055) Age 0.010 0.015* -0.007 0.010** 0.013** 0.002 (0.007) (0.009) (0.018) (0.005) (0.005) (0.017) Age^2 0.000 -0.000* 0.000 -0.000* -0.000** 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) High Decision-making power (Yes=1) 0.015*** 0.087*** 0.024 0.008* 0.021 0.1 (0.005) (0.028) (0.071) (0.005) (0.026) (0.064) Own Education 0.018*** 0.016*** 0.016 0.016*** 0.014*** 0.016^ (0.005) (0.004) (0.010) (0.003) (0.003) (0.010) Digit Span Recall -0.012 -0.018** 0.024 -0.013** -0.012* -0.001 (0.009) (0.008) (0.024) (0.006) (0.006) (0.016) Risk Tolerance 0.003 0.001 0.005 -0.002 -0.004 -0.001 (0.004) (0.004) (0.013) (0.004) (0.005) (0.008) Yrs. Of Education of Spouse 0.002 0.003 0.002 0.004 -0.004 0.018** (0.004) (0.005) (0.010) (0.005) (0.005) (0.008) Number of children under 9 0.004 0.004 0.007 0.004 0.003 -0.017 (0.009) (0.009) (0.029) (0.006) (0.006) (0.022) Business Owner (Yes=1) 0.023 -0.006 0.064 0.050 0.058 0.027 (0.032) (0.036) (0.069) (0.035) (0.042) (0.061) Land 0.006*** 0.005*** 0.007 0.001 0.000 0.005** (0.001) (0.001) (0.005) (0.001) (0.001) (0.002) Member of a Mixed Group (Yes=1) 0.096 0.377** -0.233 0.025 -0.06 0.076 (0.153) (0.170) (0.167) (0.176) (0.176) (0.197) Mean Dependent Variable 10.93 10.93 10.92 10.97 10.67 10.67 10.72 10.52 N. Observations 664 664 491 173 503 503 372 131 R-squared 0.21 0.27 0.32 0.33 0.27 0.34 0.26 0.44 Note: The dependent variable is log of loan amount requested (columns 1-4) and log of loan amount approved (columns 5-8). The sample includes loan applicants. All regressions include branch fixed effects and are estimated using OLS methods. Standard errors are clustered at the borrower group level.The following symbols *, * * and ** * denote significance at the 10 , 5 , and 1 percent level, respectively. See Appendix B for definition of variables. Table 7B. Loan Size, Business Owners OLS Log of Amount Requested Log of Amount Approved Business Owners Males Females Business Owners Males Females (1) (2) (3) (4) (5) (6) (7) (8) Female Brochure 0.093 0.08 0.135** -0.061 0.027 0.011 0.002 -0.068 (0.057) (0.050) (0.052) (0.129) (0.032) (0.033) (0.030) (0.081) Offered Business Training -0.032 -0.058 -0.021 0.054** 0.026 0.032 (0.046) (0.047) (0.145) (0.027) (0.026) (0.098) Female 0.089 -0.094 (0.082) (0.064) Age 0.013 0.017 -0.024 0.007 0.015** -0.006 (0.009) (0.011) (0.021) (0.005) (0.006) (0.023) Age^2 -0.000* -0.000* 0.000 0.000 -0.000** 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) High Decision-making power (Yes=1) 0.025 0.095** -0.022 0.029 -0.007 0.094 (0.038) (0.038) (0.091) (0.030) (0.024) (0.070) Own Education 0.022*** 0.020*** 0.027 0.012*** 0.013*** 0.012 (0.005) (0.005) (0.017) (0.004) (0.004) (0.010) Digit Span Recall -0.014 -0.019* 0.003 -0.009 -0.009 0.001 (0.011) (0.010) (0.034) (0.006) (0.007) (0.020) Risk Tolerance 0.002 0.001 0.014 -0.001 0.000 -0.003 (0.005) (0.005) (0.017) (0.005) (0.006) (0.011) Yrs. Of Education of Spouse 0.001 0.007 -0.007 0.002 -0.005 0.013 (0.005) (0.005) (0.013) (0.006) (0.006) (0.010) Number of children under 9 0.004 0.008 -0.032 0.001 -0.004 -0.011 (0.011) (0.011) (0.037) (0.007) (0.007) (0.019) Land 0.005*** 0.005*** 0.005 0.002 0.000 0.004 (0.001) (0.002) (0.005) (0.001) (0.001) (0.003) Member of a Mixed Group (Yes=1) 0.138 0.314 -0.293 -0.007 -0.256*** 0.030 (0.167) (0.213) (0.293) (0.239) (0.086) (0.331) Mean Dependent Variable 10.92 10.92 10.92 10.92 10.69 10.69 10.69 10.69 N. Observations 459 459 356 103 355 355 278 77 R-squared 0.23 0.30 0.34 0.42 0.28 0.35 0.29 0.58 Note: The dependent variable is log of loan amount requested (columns 1-4) and log of loan amount approved (columns 5-8). The sample includes loan applicants. All regressions include branch fixed effects and are estimated using OLS methods. Standard errors are clustered at the borrower group level.The following symbols *, * * and ** * denote significance at the 10 , 5 , and 1 percent level, respectively. See Appendix B for definition of variables. Table A1. Baseline Member Characteristics of Mixed vs single gender COs Male Female N. Obs. N. Obs Means N. Obs. N. Obs Means Mixed Single Mixed Single P-val of Mixed Single Mixed Single P-val of gender gender Gender gender t-test gender gender Gender gender t-test CO/CG CO/CG CO/CG CO/CG (3)-(4) CO/CG CO/CG CO/CG CO/CG (8)-(9) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Individual Characteristics Age 82 1798 36.26 38.27 0.16 137 1434 34.06 37.85 0.00 Own Education 82 1798 7.22 5.22 0.06 137 1434 2.54 2.65 0.61 Business Owner (Yes=1) 82 1798 0.63 0.61 0.66 137 1434 0.59 0.58 0.84 Digit Span Recall 82 1798 3.87 3.85 0.94 137 1434 2.39 2.71 0.12 Risk Tolerance (0=Risk Averse; 10= Risk Lover) 82 1798 4.09 3.80 0.38 137 1434 2.80 3.43 0.02 Married (Yes=1) 82 1798 0.77 0.82 0.01 137 1434 0.83 0.85 0.48 Index of Optimism 82 1798 -0.23 -0.02 0.21 137 1434 -0.87 -0.29 0.00 Business Literacy 82 1798 0.97 0.66 0.07 137 1434 0.14 0.23 0.67 Index Female Mobility 82 1798 - - - 137 1434 0.08 0.00 0.36 Index No Purdah 82 1798 - - - 137 1434 -0.30 0.19 0.00 Trust in Formal System 82 1798 -0.01 -0.03 0.31 137 1434 -0.28 -0.14 0.00 Months as Member 82 1798 16.48 28.74 0.00 137 1434 20.28 24.98 0.45 Holds office in Group (Yes=1) 82 1798 0.48 0.20 0.00 137 1434 0.06 0.18 0.01 Eligibility 82 1798 0.49 0.63 0.00 137 1434 0.53 0.63 0.00 Household Characteristics Household size 82 1798 7.88 7.93 0.70 137 1434 7.45 7.00 0.00 Years of Education of Spouse 82 1798 3.87 2.67 0.02 137 1434 4.93 4.88 0.91 Number of children under 9 82 1798 1.78 1.97 0.40 137 1434 1.69 1.40 0.05 Land 82 1798 4.38 6.01 0.39 137 1434 4.48 2.58 0.30 Fraction of Members of same Zaat (caste) 82 1798 0.41 0.43 0.12 137 1434 0.45 0.21 0.00 Ever in Business (Yes=1) 82 1798 0.60 0.62 0.11 137 1434 0.54 0.61 0.76 Log of Value of Livestock 82 1798 5.77 9.36 0.00 137 1434 6.99 4.78 0.00 Distance to meeting place 82 1798 6.24 7.50 0.07 137 1434 5.91 9.17 0.00 Credit Constraints (Yes=1) 82 1798 0.09 0.12 0.28 137 1434 0.19 0.16 0.31 Expenditures (in 1,000 Rupees) 82 1798 5.42 4.98 0.91 137 1434 4.86 4.58 0.36 Decision Making (0-8) 82 1798 3.10 3.35 0.71 137 1434 1.95 1.83 0.39